Contents. 1. Opening Lectures. - pressures and responses - past, present and future. 3 Integrated River B~sin Management - planning for t~e future

----~~----- -~--

Contents 1. Opening Lectures 'Jeert Janssen, Gerda Dreise and Leonard Terwey Tte Implementation of the Water Framework Directive in the Netherlands _ljarrinmc Garrigle C;;. aterquality in Ireland 1998-2000 1 15 WaterQuality - pressures and responses - past, present and future 3'"!ice D R Misstear and NickS ::eltic Groundwater: Robins CUITentPressures and Future Responses.?::ml Shand and Bjorn Fr;engstad 3aseline groundwater quality : Contrasting geochemical evolution in selected ::eltic and Viking Aquifers. 3rient Land Bertru G S ~"veil1ance et contrôle des efflorescences à cyanobactéries: seuils de vigilance et d'alerte :' our les eaux récréatives et eaux de, boisson. :; }.1artin, L Aquilina, C Gascuel - Odoux, J Molénat and L Ruiz \ a..c-iationssaisonnieres des concentrations en nitrate dans les bassins versants agricoles..,;,oesser C, Robinson Rand Soulsby C Sources ofironand Manganese in public water supplies: A case study of the ~.1egget Reservoir, Scotland. ~;aud Valerie 3uffering capacity of landscape structures inrural catchments: deflnition and assessment..".!. Ryan and J.P. 0 'Kane ~2e Blue City Project- Water quality measurement on the Lee (POSTER). 16 24 36 41 51 57 58 3 Integrated River B~sin Management - planning for t~e future.'ljlthony J Tollow.-\TI Alternative Approach to the Management of Water Resources. 59 Jôanne Cullen l ne Use of End Member Mixing Analysis in Identifying the Relative. Contributions of Different Runoff Pathways. I1's Potential Value for Future Catehment Management Studies. 64 Sarah M Dunn, Mark Stalham, Neil Chambers and Bab Crabtree Spatial and temporai in the development of policies for abstraction control in Scotland. 74.=. E T Jenkin, S Janman and P Long Envrronmental impacts of past, present and future water management ::netal contaminated river catchment. within a 85 vii

JonesJ AA Natural soil piping, water quality and catchmentt management in the British Uplands. 98 Tony Cawley, D Mc Dermottand E Ki/cullen Predicting the hydrological impact of infill development on a karst wetland system at Oranmore, Co Galway. E. Bocher, M J Penven and K Mathieu (POSTER) Les écoulements de surface sur un versant bocager Modélisation et simulation à l'aide du SIG. 115 121 4 The Scientific support for managemep.t: Hydrological processesand models -limitations and uncertainties TomaszDysarz and Jaroslaw J Napiorkowski 123 Application of sequential optimisation for flood control- Nysa Reservoir System case study. P Hubert, I Tchiguirinskaia, H Bendjoudi, D Schertzer and SLovejoy 131 Modelisation rnultifractale des debits du Blavet a Guerledan. J Martin, L Migliori, J A Smyth and Jp O'Kane 139 Flood Alleviation Planning in a Virtual Water World- De-watering the Lower Feale Catchment A. Creating a Virtual Water World. Monomoy Goswami, Kieran M O'Connor and Asaad Y.Shamseldin 151 Rainfall-RunoffModelling oftwo Irish Catchments. (One Karstic and One Non-Karstic). Dr Rachel Helliwell and SJuggins 165 Predicting biological and chemical recovery in the Galloway region of SW Scotland. Bas Krijgsman and J P Lobo-Ferreira 176 A New Methodology for Delineating Wellhead Protection Areas. Dr P Purcell, Heather Scully and T Gleeson 188 Flow Characteristics of natural sail liners for the containment of landfillleachate and agricultural Slurries. Alaa El-Sadek, Ingeborg Joris and Jan Feyen 197 Water quality rnodelling of an agricultural field: from sail surfac,e ta groundwater. AH Hm'ia, Paul Shand and Chris Soulsby 206 Bedrock groundwater processes during streamflow generation in the riparian zone of an upland catchment in Wales, UK. A Jigorel and J P Morin 214 Evaluation des dépôts sédimentaires dans les retenues: mesures directes et indirectes dans le barrage de Keme Uhel sur le Blavet (Bretagne). G Marjolet, A Artur et M Freslon 223 Périmètres de protection des captages d'eau souterraine dans le massif armoricain. Effets sur la qualité des eaux. viii -

Véronique Maleval 233 Nature, origine et taut de sédimentation dans un lac oligotrophe: le lac de Saint-Paradoux, France. and rate of sedimentation in a oligotrophic lake: The lake of Saint- Pardoux, France. LaurenceA Boorman 241 Feeding the Fish - ACeltic Perspective: Changing potentials for land..:...seaexchanges of organic matter and other materials in selected Scottish Sea Lochs. J Martin, L Migliori, J A Smyth and J P O'Kane 257 Flood Alleviation Planning in a Virtual Water World- De-watering the Lower Feale Catchment. B. Planning in a Virtual Water World. Keshav P Bhattarai and Kieran MO 'Connor 269 The effect of land-use changes and arterial drainage on the runoff component of the hydrological cycle: in the Irish context (POSTER). 5 Risk assessment, perception and management: extreme rainfails, floods, droughts and climate change Pierre Javelle, Eric Sauquet and Jean-Michel Gresillon Describing actual and future flood hydrological regimes. Jerry Grant and David Wilson Communication of Flood Hazard to the Public. Salomon Seyoum Demissie and Conleth Cunnane Representation of elimate change in flood frequency estimation. Keshav P Bhattarai iill investigation of the use of partial L-moments for analysing censored flood samples. D G George, M Rouen, B O'Hea, P McGinnity and N Allott Usingautomatic water quality monitoring systems to monitor the transport of suspended sediment from an upland catchment in the west of Ireland. P Duband Crues Decembre 2000 et Janvier 2001 en Bretagne et references historiques. Keshav P.Bhattarai, Monomoy Goswami and Kieran MO 'Connor An Introduction to the Galway River Flow Forecasting System (GFFS) (poster). 270 281 290 302 312 322 336 IX

THE IMPLEMENTATION OF THE WATER FRAMEWORK DIRECTIVE IN THE NETHERLANDS Geert Janssen B.Sc. Principal hydrogeologist Dept. Hydrology, FUGRO, The Netherlands Treasurer of The Netherlands Hydrological Society NHV Gerda Dreise B.Sc. Head Dept. Integrated Water management & Planning, Water Board Wilck and Wiericke, The Netherlands Leonard Terwey Civ Eng Dip Env Sr. Consulting Engineer Water & Environment, FUGRO, The Netherlands Elected Member of the General Council of the Water Board of Delfland, The Netherlands KEY WORDS European Water Framework Directive (E)WFD, The Netherlands, Water Boards, River Basin Management, Integrated Water Management, Water legislation and Regulations. SUMMARY On the 22nd December 2000 the WFD was published. Since then, the Members of the European Community are facing a deadline of 22nd December 2003, to implement the necessary legislative and administrative provisions or regulations. The purpose of the Directive is to establish a framework for the protection of inland waters, transitional water, coastal waters and groundwater. Water is no longer regarded as a product like any other, but as a heritage which must be protected, defended and treated as such. To prepare the Water Managers in The Netherlands for the consequences of the WFD, a Project Group Implementation of the WFD was established on the 1st of November 1998. This Project Group is a co-operation of the Ministry of Transport, Public Works and Water Management, Ministry of Housing, Spatial Planning and the Environment, Ministry of Agriculture, Nature Management and Fisheries, the Inter Provincial Consultative Body, the Union of Water Boards and the Consultative Body of the Dutch Municipalities. The Project Group divided their study in 9 fields of attention. Main conclusions of the Project Group (so far) are: With the Minister of Transport, Public Works and Water Management as competent authority, relations and responsibilities between the State, Provinces, Water Boards and Municipalities, can be left unchanged; Dutch laws cover most of the WFD. What is needed is a staple through the Dutch legislation ; River Basin Management Plans can be incorporated in the 4th National Policy Document on Water Management. The revision-period of this document can be adapted to 6 years in conformity with the WFD; National emission policy fits within the WFD. A working-out of the emission-approach will be necessary. Dutch assessment-systems for the chemical condition of water generally fit in the WFD. Regarding the ecological status, a new assessment-system will be developed; Currently displayed Monitoring efforts meet the requirements of the WFD, except for monitoring in protected areas. This, the groundwater issue and the reporting obligations will be dealt with in Phase 3; Water Management costs in The Netherlands are (except from abstractions from surface water) already financed by direct or indirect taxes or prices. The economic evaluation in conformity with the WFD is new for The Netherlands but seems feasible. The Project Group will compile a manual with the contributions of the various Working Groups that will be released in the end of 2002, to promote the implementation of the WFD. The WFD is (for The Netherlands) considered to be a good supplement to the existing international treaties. The WFD enhances the international co-operation within river basins to tackle important issues in water management. The WFD integrates earlier enacted Directives and might even set an example for more integration of national laws. Being the sink of a large part of Europe, The Netherlands thus benefits from the WFD. But still, water management is not a separate aim of the European Union but embedded in European Environmental Law. The WFD is clear in its strategies in fighting, preventing and controlling water~ and groundwater-pollution, but there are no such strategies for fighting floods and droughts, which are only mentioned briefly. In The Netherlands, the WFD will be linked with the newly adopted strategy Water Management Policy in the 21st Century (WB21) [11] to clear this omission. WB21 proposes a different approach of water management in The Netherlands: a water-driven, area-specific strategy and integrated, sustainable water management approach and priority to water management in area planning. A strategy in which water is regarded not as enemy, but as ally, with room to move. The implementation of the Water Framework Directive in The Netherlands. Page 1 of 14

SOMMAIRE Le 22 décembre 2000 la WFD a été publiée. Depuis, les membres de la Communauté Européenne doivent respecter la date limite du 22 décembre 2003 pour mettre en application «les dispositions ou les régulations législatives et administratives nécessaires». Le but de la directive est l'établissement d un cadre pour la protection des eaux intérieures, de l'eau transitoire, des eaux côtières et des eaux souterraines. L'eau n'est plus considérée comme un produit comme un autre mais comme un héritage qui doit être protégé, défendu et traité en tant que tel. Pour préparer les Managers de la gestion de l'eau aux Pays-Bas aux conséquences du WFD, un Groupe de Projet «Application de la WFD» a été créé le 1er novembre 1998. Ce groupe de projet est une coopération du ministère du transport des travaux publics et de la gestion de l'eau ; du ministère du logement, de l aménagement du territoire et de l'environnement ; du ministère d'agriculture, de la gestion et de la pêche de nature ; de l'organe consultatif interprovincial; de l'union des agences de l'eau et de l'organe consultatif des municipalités néerlandaises. Le groupe de projet a divisé leur étude en 9 domaines spécifiques. Les conclusions principales du Groupe de Projet (jusqu'à ce jour) sont: Avec le ministre du transport des travaux publics et de la gestion de l'eau en tant qu autorité compétente, les relations et responsabilités entre l'état, les provinces, les agences de l'eau et les municipalités restent inchangées; Les lois néerlandaises couvrent la majeure partie de la WFD. Il est nécessaire «d assembler la législation néerlandaise» ; Des plans de gestion de bassin de fleuve peuvent être incorporés dans le 4ème document national de politique sur la gestion de l'eau. Le temps de révision de ce document est de 6 ans conformément à la WFD; La politique Nationale d'émission couvre la WFD. L approche de l émission doit être élaborée; Les systèmes d évaluation néerlandais sur l'état chimique de l'eau sont généralement adaptés à la WFD. Concernant le statut écologique, un nouveau système d évaluation sera développé ; Les efforts actuels de surveillance répondent aux exigences de la WFD, excepté la surveillance dans des secteurs protégés. L'issue des eaux souterraines et les engagements de rapportage seront traités dans la phase 3 ; Les coûts de la gestion de l'eau aux Pays-Bas (excepté les extractions de l'eau de surface) sont déjà financés par les impôts ou les prix indirects ou directs. L'évaluation économique, conformément à la WFD, est nouvelle pour les Néerlandais mais semble faisable. Le Groupe de Projet compilera un manuel avec la contribution des différents groupes de travail qui sera présenté à la fin de 2002 pour favoriser l'exécution de la WFD. La WFD est considérée (par les Néerlandais) comme un plus aux traités internationaux existants. La WFD intensifie la coopération internationale sur les bassins des fleuves pour aborder les questions importantes dans la gestion de l'eau. La WFD regroupe les directives décrétées antérieurement et pourrait même être un exemple pour plus d'intégration des lois nationales. Étant l'évier d'une grande partie de l'europe, les Néerlandais bénéficient ainsi de la WFD. Notamment, la gestion de l'eau n'est pas un but isolé de l'union Européenne mais elle est incluse dans la loi environnementale européenne. Le combat de la WFD est clair : la prévention et le contrôle de la pollution de l'eau de surface et de l eau souterraine, il n'y a pas de stratégie comparable concernant la lutte contre les inondations et la sécheresse, seulement mentionné brièvement. Aux Pays-Bas, la WFD sera liée à la stratégie adoptée récemment dans «la politique de gestion de l eau dans le 21 ème siècle (WB21) pour corriger cette omission. WB21 propose une approche différente de la gestion de l'eau aux Pays-Bas : une stratégie dominée par l eau et spécifique pour les différents secteurs, une approche de la gestion durable de l'eau et une priorité de la gestion de l eau dans l aménagement du territoire. Il s agit d une stratégie dans laquelle l'eau est considérée non comme un ennemi mais comme un allié, avec un espace pour se déplacer librement. The implementation of the Water Framework Directive in The Netherlands. Page 2 of 14

INTRODUCTION The (European) Water Framework Directive (WFD) is a document that affects all countries in Europe that are united in the European Community. This paper illustrates the way The Netherlands is dealing with the implementation of the WFD. First, the contents and the Aims and Objectives of the WFD will be outlined, as well as the chosen River Basin approach that is used in the WFD and the time-schedules and deadlines that are set. For a better understanding of the implementation of the WFD in The Netherlands the geography and geology of The Netherlands, the climate and hydrology and the man made environment for which the Dutch are famous of are highlighted. Interesting is how dealing with water is institutionalised in The Netherlands. Next will be the implementation of the WFD in The Netherlands. How are the Dutch organised, who is involved in the implementation process, what is their strategy, how are the proceedings of the implementation process, do the Dutch think that they can meet the required time schedule and, last but not least, how do the Dutch inform the public. Purpose Water is not regarded as a product like any other, but as a heritage which must be protected, defended and treated as such. The purpose of the Directive is to establish a framework for the protection of inland waters, transitional water, coastal waters and groundwater which: a) Prevents further deterioration and protects and enhances the status of aquatic ecosystems and other related ecosystems; b) Promotes sustainable water use based on a long term protection of available water resources; c) Enhances protection and improvement of the aquatic environment through the progressive reduction of discharges, emissions and losses of priority substances; d) Ensures the progressive reduction of pollution of groundwater; e) Contributes to mitigating the effects of floods and droughts. Finally the impact of the WFD on the situation in The Netherlands is described. Does the WFD interfere with the Dutch situation or can the Dutch benefit from the WFD. THE WFD: A RESUME OF THE DOCUMENT By the mid 1990s a need to pull together the range of existing European legislation on water in a coordinated manner was identified. The end result of this process is the Water Framework Directive - widely recognised as one of the most ambitious and comprehensive pieces of European environmental legislation to date. The Directive is supposed to be a comprehensive and co-ordinated package to ensure that all European waters are protected according to a common standard. On the 22nd December 2000 the WFD was published. According to the articles 24 and 25 this means that the WFD is in operation since that day. From that day, the Members of the European Community are facing a deadline of 22nd December 2003, to implement the necessary legislative and administrative provisions or regulations. By the end of 2003, the necessary adaptations of the National laws must have passed by national parliament as well. River Basin Approach The Directive has two key components: A system of management of the natural water environment based around natural river basin districts; and Co-ordinated "programmes of measures" to achieve at least a good status for most of the rivers, lakes, estuaries, coastal waters and underground waters. Within each river basin district the ecological condition of rivers and lakes will have to be assessed. For groundwater the key factors are chemical contamination and quantity. Objectives will then be set for each water body and measures to achieve the objectives put in place. Many of these measures are set out in the Directive. They include controls on water pollution, on abstraction and impoundment of water and on other impacts such as engineering works in and around water. The implementation of the Water Framework Directive in The Netherlands. Page 3 of 14

The condition of the water body will then be reassessed to determine whether the specified objective has been met. This process will be repeated on a continuing basis. Time schedules The first stage is transposing the Directive's provisions into the National Laws. Given the scope and fundamental nature of the Directive this should be done by primary national legislation. As with all legislative commitments this will be subject to finding space in the legislative programme. Important deadlines that are set, are: 2003 - Directive's provisions must be transposed into National Law 2009 - Publish River Basin Management Plans & establish programmes of measures 2012 - Programmes of measures fully operational 2015 - Environmental objectives to be met inhabitants. It would probably be more factual to suggest that the Dutch and their ancestors have, for thousands of years, adapted their way of life to suit their environment. In the beginning by taking advantage and relief of the land as they found it. Later by changing the natural landscape, deliberately or not. Geography and Geology The size of the land area of Netherlands is approx. 34.000 km 2. The total territory, including inland lakes, estuaries and territorial sea, measures 41.160 km 2. The Netherlands comprises the deltas of the rivers Rhine (Rijn), Meuse (Maas), Scheldt (Schelde) and Eems (Figure 1). In general The Netherlands slope from south-east to north west. Figure 2 INTRODUCTION TO DUTCH SITUATION The Netherlands is a flat Delta covered with alluvial deposits and, due to its unusual location, with a fascinating geological history. At and below the surface there is an enormous variety of sediments. Some of the sediments are useful while others present a challenge to anyone wanting to build and live on the land or use it in other ways. The deposits that make up The Netherlands, were formed by rivers, the sea, wind, ice-sheets, meltwater, and volcanoes. It is said that The Netherlands is made by its The highest point in the south-east is 322 above m.s.l., not one third of the height of the Irish Carrauntoohill. Barring the dunes, the western and northern parts of the country have elevations varying from slightly above mean sea level, to 6,7 m below sea level at the lowest point near Rotterdam. In the absence of dunes and dikes more than 65% of the country would be flooded at high sea and/or high river levels as illustrated in Figure 2. Figure 1 Tertiary and Mesozoic deposits are situated at great depths and outcrops only occur in the south-east and eastern areas. Marine Tertiary clays form the base of the groundwater aquifer system, and are covered by Pleistocene and Holocene deposits as illustrated in Figure 3. Loamy and clayey material of marine and fluviatile origin, together with local peat soils and fine sands, dominate at the surface in The implementation of the Water Framework Directive in The Netherlands. Page 4 of 14

Figure 3 the south-west, west, north and central river districts. Draw-down of the groundwater table causes shrinkage and oxidation of clayey and peat soils. In the east and south parts of the country loamy cover-sand, medium and coarse sand and gravel are found. Climate and hydrology Due to strong maritime influences the climate of The Netherlands is milder than expected at 52 0 N latitude. The annual average temperature of 9 to 10 0 C is far beyond the 4 0 C that is normal on this latitude. We call the climate of The Netherlands a coastal climate. The average annual rainfall in The Netherlands is 750 mm. Nowhere in the country do values deviate from this by more than 10 to 15%. While the aerial variation in precipitation is small, the seasonal variation is more pronounced. Early spring is driest and summer and late autumn are wettest. The heaviest showers occur inland during summer. Interannual variability in precipitation can be quite large with lowest amounts as low as 400 mm and highest nearly 1200 mm. Mean wind-direction is south-southwest (70% off the time). The mean annual evapotranspiration of The Netherlands varies from 600 mm in coastal areas to 500 mm inland. The seasonal variation of evaporation is very large, due to its dependence on solar radiation and temperature. Evaporation in summer is 9 times higher than in winter, resulting in a water surplus in winter (300 mm from October to March) and a deficit in summer (150 mm from April to September). Surface water covers one fifth of the area of The Netherlands. Almost everywhere in the low polder areas water levels are artificially controlled by a forced discharge, but on higher grounds the drainage of water is mostly by gravity. Smaller ditches in sandy areas might fall dry whereas most ditches in polder areas remain filled all year around. The largest terms in the water balance of the whole of The Netherlands by far are the in- and outflow of the river Rhine (63% of 109.500 10 6 m 3 /year). In 1993 and 1995, the river Meuse inundated large areas and Rhine water reached dangerous levels, causing considerable economic damage. The groundwater hydrology is controlled by the sediments, deposited in a subsiding basin that dips to the north-west (see Figure 4). The Quarternary deposits often have high permeability. Public drinking water is mostly abstracted from these deposits. Upper Tertiary layers are not exploited due to low permeability and often the presence of brackish to saline groundwater as is shown on the hydrogeological cross-section of Figure 3. The groundwater recharge in The Netherlands vary much depends on the local topographical situation. Infiltration of rainfall is the predominant form of recharge in the sandy areas. In those regions The implementation of the Water Framework Directive in The Netherlands. Page 5 of 14

Figure 4 drainage problem. Dikes were built to protect the land against the sea-water. To avoid high water levels in the embanked areas, excess water was released through outlets at low tide. Local embankments were connected by dams closing tidal creeks. Names of Dutch cities e.g. Amsterdam and Rotterdam, remind of these events. Behind the dikes and closure dams the embankment of small areas, polders, was started. The stepwise drainage system, as shown in Figure 5, is very typical in The Netherlands. At first the artificial drainage tools were man- and horse-powered, followed by windmills which made it possible, in various stages, not only to drain polders, but also to reclaim land from shallow and deeper lakes. The waterwheels of the windmills were replaced by Archimedian Screws in the 17th century. Eventually wind-power was replaced by successively steam-, diesel- and electrically driven pumping stations. covered with clay or peat layers practically the full excess precipitation is discharged by surface flow to nearby open water courses. The groundwater level varies with the surface level, and will be lowest near Rotterdam at approx. 7 m below mean sea level. The history of The Netherlands is particularly characterised by floods, reparative work and reclamation. The continuing subsidence of the surface in the polders and the rising of the sea level have resulted in a land with 25% of its area lying below mean sea level and 65% below high sea and river level. Figure 5 The man made environment Water management greatly influences the face of The Netherlands. Thousand years ago large parts of The Netherlands flooded regularly. Due to cultivation, the soil sank, and large area subsided to such a degree that the tide could stream in. Through the centuries, large lakes formed as peat was dug up to serve as fuel. By using embankments and windmills, lakes could be drained and the land reclaimed. However, the permanent need to lower the groundwater table for agricultural production, provoked an irreversible subsidence process. Besides this man-made process, the rising of the sea level also affects the The implementation of the Water Framework Directive in The Netherlands. Page 6 of 14

Organisation of Water management in The Netherlands In the democracy in The Netherlands, the people choose their representatives in three governing bodies. The Central Government, Provincial Government and Regional Authorities, including the Municipalities and Water Boards. The Central Government formulates the main features for the strategic policy with respect to water issues at a national level and is responsible for the operational management of the state managed waters and major flood protection works. The Provincial Government defines the strategic policy for non-state managed waters and for the regional flood protection works, and controls for groundwater extraction and waterways serving navigation. Municipalities are responsible for operational (water)management (sewerage system, drainage in urban areas), and the implementation of (other) policy issues. Water Boards play a most important role in Dutch water management. The Boards are responsible for flood protection, drainage and rural areas, water quantity and quality water management and waste water treatment. area, defined by the boundary of a water-system, and not by historical borders. The administration of a Water Board consists of: General Council, chosen by tax payers every four years; Executive Board, chosen by the General Council; Chairman (Dyke Reeve), appointed by the Crown for a six year period. All taxpayers (following the principle of interestpayment-say/participatory control ) may choose a representative in the administration of the Water Board. In the middle-ages the first Water Boards appeared in The Netherlands. The communities at that time, known as Boroughs, selected Governors to promote the interest of the local population. Water management also fell under their jurisdiction. The very first water management workers were farmers and landowners, who were responsible for local water management, engineering works, such as the dikes, canals, watercourses and roads. There was a continuous need for the building and rebuilding of dikes, dams and river estuaries, drainage locks and canals. A gigantic undertaking in a time when everything had to be done with manor horsepower. Dike building and waterdrainage often needed to be done over the borders of one Borough. Regional co-operation between the Boroughs was the next step. And so the first district Water Boards were formed. The district Water Boards worked in a democratic fashion, involving the local communities not only in the maintenance of public works, but also through participation in the governing body. In 1850 there were 3500 Water Boards. Mergers reduced the number of Water Boards. Following the flood disaster of 1953, the process of centralisation took on a new urgency. By January 2000 there were 57 Water Boards left. It is expected that by 2003, The Netherlands have about 48 Water Boards. Figure 6 In Figure 6 the institutional structure regarding water issues in The Netherlands is illustrated. The Water board A Water Board is an independent lower tier of the government, overseen by the Provincial Authorities. Each Water Board is responsible of the essential aspects of water management in a given Water management in The Netherlands costs 2,8 b per year ( 175/capita) at the end of the 20th century. The greater part of this budget is spent on operating costs. The rest is used for investments, such as sewage treatment plants, pumping stations, and dike strengthening. The Water Boards finance their work entirely from taxes. There are 5 categories of tax payers: owners of real estate, owners of land, industries, leaseholders, and residents. A tax payer may belong to several categories. The Water Board charges for owners of real estate, a sum depending on the value of the property, the owners of land a fixed sum per hectare, and for residents a fixed sum determined by each Water Board and the pollution levy. The implementation of the Water Framework Directive in The Netherlands. Page 7 of 14

IMPLEMENTATION OF THE WFD IN THE NETHERLANDS The Water Management Act (1989) formulates a framework that harmonises the objectives and contents of the water management plans and the coordination of the planning of the participating authorities in water management. By law, the plans are updated every 4 years. Figure 7 Figure 7 shows the relationship of water resource planning with area and environmental policy planning. The system works both top-down and bottom-up. The last few years, quite a lot of new plans were released in The Netherlands e.g. 4th National Policy Document on Water Management, Room for Rivers, A Different Approach to Water, Water Management Policy in the 21st Century, the 5th Memorandum Spatial Planning, Nature for People, People for Nature etc. The mutual attuning of these plans therefore becomes an increasing challenge. General introduction, participants and time- schedule To prepare the water-managers in The Netherlands for the consequences of the implementation of the WFD, a Project Group called Implementation of the WFD was established on the 1st of November 1998. The P roject Group is a co-operation of the Ministries of Transport, Public Works and Water Management (V&W), Housing & Spatial Planning and the Environment (VROM), Agriculture, Nature Management & Fisheries (LNV), Inter Provincial Consultative Body (IPO), Union of Water Boards (UVW) and the Consultative Body of Dutch Municipalities (VNG). Important questions (among others) to be answered by the Project Group were: What are the consequences of the WFD for a Delta like The Netherlands? How does the WFD interfere with the 4th Bill Water management (later followed by Water management 21st Century)? Can the unique system of Water Boards be preserved within the WFD? What are the consequences of the proposed partition in 4 River Basins? Can we meet the WFD requirements regarding water quality? Is there anything mentioned in the WFD concerning flood defence? Does our financial system fit in the system of levies and taxes as described in the WFD? Phase 1 started in November 1998 with a reconnaissance of the possible consequences of the implementation of the WFD was carried out. Phase 1 was concluded with 2 reports with overviews of legislative, administrative, political and technical implications and an overview of various issues that have to be dealt with in near future. Phase 2 started off with a workshop in October 1999. Aim was to deal with the various issues as mentioned above. In the beginning of 2001 this Phase was concluded with a report of the intermediate results of the various Working Groups. Phase 3 started in January 2001 to continue where Phase 2 ended. Under the responsibility of the Project Group a manual will be released at the end of 2002, to promote the implementation of the WFD. This document will disclose the information necessary to meet the requirements of the WFD. Preparing the implementation of the WFD The assignment of the Project Group that started off in 1998, was to present a document that describes the preparation for the implementation of the WFD. Target date for this document was the date of the actual Directive itself, expected by the end of 2000. The Project Group divided their tasks in 9 fields of attention, each dealt with a separate Working Group: Legislation Geographical Division Co-ordination International Implementation Integrating the WFD in existing Dutch Planning-system Emission and Human Influences Aims and Objectives and Monitoring Finances and Economical Analysis National Administrative Organisation Communication These fields are highlighted hereafter regarding the three earlier mentioned Phases up to the end of this year 2002. The implementation of the Water Framework Directive in The Netherlands. Page 8 of 14

Working Group Legislation Article 24 describes that the Member States shall bring into force the laws, regulations and administrative provisions necessary to comply with the Directive at the latest three years after the date of entry into force of the Directive (22nd December 2003). By this time these Laws must have been passed by national Parliament and published in the Dutch Gazette. In The Netherlands the WFD affects at least 10 different Acts, varying from the Water Management Act (1989) to the Environmental Protection Act (1993). These Acts fall under three Ministries (LNV, V&W and VROM). As described in the WFD the Member States shall identify the Competent Authority that will be charged with the supervision over the implementation and the observance of the WFD. These and other issues will be implied in an Introduction Act that describes the amendments necessary. Apart from this Introduction Act a few regulations and by-laws will be introduced. The preparation for this Act, the regulations and the bylaws was finished in the third quarter of 2001. Hereafter, legislation procedures take up to 6 months before introduction of the Bill to the Second (Lower) Chamber. Preferably the text of the Bill has to be ready in the first quarter of 2002, because Parliamentary Discussions in the Second Chamber might take up to a year. Netherlands the River Basins coincide with the River Basin Districts). In The Netherlands four River Basins and can be distinguished: the Rhine, the Meuse, the Scheldt and the Eems. The four River Basins are indicated on Figure 1. The delimitation of the Scheldt seems to be quite obvious where the delimitation of the Eems is more complicated. Deliberation with the German authorities in Phase 3 will be necessary to discuss the transition between transitional water and coastal water, the assignment of coastal water to the River Basin District of the Eems and the border between the districts of the Rhine and the Eems. The border between the Meuse and the Rhine is even more complicated. Both rivers flow into the same estuary where the water mingles. There is no clear hydrological border. For both rivers (and for the Scheldt) international treaties and river committees exist, but those treaties cover only the rivers themselves and not the whole of the river basin and catchment area. The assignment of coastal water to the Meuse or the Rhine is therefore arbitrary. The four River Basin Districts are divided in 17 Figure 8 The intention is that the final text of the Act is passed by the Second and First (Upper) Chamber of the Dutch Parliament in the first quarter of 2003. Then there will be ¾ of a year left to prepare the necessary implementation-measures and regulations before the end of the deadline on 22nd December 2003. To be less dependent of the proceedings in the Second and First Chamber, the Project Group will outline the necessary legal measures to be taken by National, Regional and Local authorities beforehand. Regarding legislative complications of the WFD, the conclusion is that most of the WFD is covered by the Dutch Laws. In one of the documents the Project Group speaks of a staple through the Dutch legislation, meaning an mutual attenuation of the laws. Working Group Geographical Division This Working Group indicates the borders of the sub-basins in The Netherlands and the assignment of these sub-basins to the International River Basin Districts. In Phase 1 it was (not surprisingly) concluded that, being a delta, the whole of The Netherlands would be assigned to a River Basin (or ~District: in The sub-basins as indicated on Figure 8. This subdivision is based on earlier studies. Shallow groundwater that is strongly influenced by open water, can easily be assigned to a certain sub- between basin or basin. Deeper groundwater can cross borders sub-basins, basins and even frontiers which makes assignment to basin (-districts) difficult. Not only technically, but also administrative. The implementation of the Water Framework Directive in The Netherlands. Page 9 of 14

If the subdivision would be made from a groundwater point of view, there would be 15 sub-basins (according to the Netherlands Technical Committee for Soil Protection) but for the WFD the division in 17 parts will be used. Currently the Ministry of Transport, Public Works and Water Management is revising the National Watercourse Maps, taking into consideration the division in river basins and sub-basins. This Working Group almost finished completely their work in Phase 2, apart from the discussions with the Germans and some detailed descriptions of the borders of the sub-basins; those are left for Phase 3. Working Group Co-ordination International Implementation The aims for this Working Group were to coordinate the making of one River Basin Management Plan per international river basin, to assess the position of the international river committees and the discrepancies between the WFD and international water treaties. The international river committees are intergovernmental : all decisions are made by consensus. All rivers, except the Eems, are covered and there are committees for coastal issues as well. The contribution of these committees, the conveyance of responsibilities and the process of decision-making is momentarily subject to discussions. The basin of the River Rhine covers most of The Netherlands. An adequate co-ordination of an international River Basin Management Plan for the Rhine, is therefor of great importance for The Netherlands. During Phase 2 at a conference of the Rhine Ministers in Strasbourg, it was agreed that the Rhine Water Directors of the Rhine Bank Countries will take the lead for drafting the international River Basin Management Plan for the Rhine. These agreements are supported by a strategic document (EU-Water Directors) and an Implementation Working Plan (Rhine Water Directors). During Phase 3 the Working Group will draw up a framework instruction to assure a consistent Dutch contribution during the process of designing the River Basin Management Plan by the international committees. A significant role for the international river committees seems ahead for the rivers Meuse, Scheldt and Eems (not established yet) to come to one international river basis management plan per river as drafted for the river Rhine. Working Group Integrating the WFD in existing Dutch Planning-system This Working Group finished their work in Phase 2 and is dissolved. In Phase 3 two pilots are carried out to test the ideas of this Working Group for workability in every days practice. In their report in Phase 2 the Working Group concluded the following. Generally, with the Minister of Transport, Public Works and Water Management as competent authority, the existing relations and responsibilities between the State, the Provinces, the Water Boards and the Municipalities, can be left unchanged. The ambition is to realise a collective, international River Basin Management Plan. If this fails, for whatever reason, at least plans for the national parts of the river basin districts have to be drawn, four in the case of The Netherlands. Therefore the Working Group chose to incorporate the plans (recognisable) in the National Strategic Water Bill (4th National Policy Document on Water Management) and to change the Water management Act via amendments as described earlier. The Dutch Environmental and Water management Plans have a revision-period of four years. The WFD uses a revision-period of 6 years. Adaptation of the revision-period in conformity with the WFD has consequences for the fashion of data-collection and reporting of progresses. The Dutch Provinces make Provincial Environmental Plans for their Provinces, concerning water management, environmental policy and area planning. The same applies for the Water Boards. In most cases the borders of the (sub-)basins do not correspond with the Provincial or Water Board borders. With the help of checklists, it will be possible to make sure the various plans contain all the necessary items as described in the WFD, which solves this administrative matter in general. The minimum contents of a River Basin Management Plan are described in appendix VII of the WFD. For each of the eleven requirements, the Working Group investigated if the necessary information is already acquired and registered somewhere, or, if not, what actions need to be taken. Currently all the Provinces, in co-operation The implementation of the Water Framework Directive in The Netherlands. Page 10 of 14

with Water Boards, with the help of Consultants, and supported by the Project Group, are drafting the river (sub-)basin management plans. A first draft, containing the next four issues, will be submitted to the EU in 2004: a) Description of river basin(s) on a national level. The discrimination in the four categories (coastal water, transitional water, rivers, lakes) and three types (natural, artificial, heavily modified), and the definition of the reference- circumstances are currently under study. b) Provide insight in the human influences on surface water and groundwater. Groundwater registers exist in the Provincial Environmental Plan, the Groundwater-register and annual reports of Drinking Water Companies. Surface water registers do not exist and have to be made. c) An economical analyses of river basin(s). This analysis is new for The Netherlands. The information required is available and, as described earlier, The Netherlands already have a system in which provision is made for the payment for water services. d) A register of the Protected Areas within the river basin(s). For Drinking Water Area Protection maps do exist. For other areas (Bird and Habitat-areas, shell-fish waters, swimming international waters etc.) national or previous Directives exist. Working Group Emissions and Human Influences In Phase 1 this Working Group concluded that the national emission policy, containing precautionprinciples, a concatenation-approacpriority assessment based on risk-analysis, fits (chain~) and a within the WFD. It was also concluded that a further elaboration of the emission-approach would be necessary. In article 10 of the WFD, a combined approach for point source discharges (like industrial and sewage water treatment plants) and diffuse sources (like agriculture, recreation, transport) is indicated. For point sources the Best Available Technique (BAT) is prescribed, for diffuse sources the Best Ecological Practice (BEP). If these measures do not result in the required water quality as indicated via an imission-test, supplementary measures have to be taken. This immision-test is a test that takes the cumulative effect of emissions into account. In Phase 2, the Netherlands Commission on Integrated Water Management (CIW) presented a report that deals with the priorities of sources that are liable to pollute inland waters, as well as an assessment of point source discharges. The report describes a phased action plan for these sources that will be integrated in the manual for the water managers. The report does not cover emissions in estuaries and incidental emissions. That will be dealt with in Phase 3. The termination of the emission of priority substances is already subject to Dutch Law as in the Pollution of Surface Waters Act (1970), the Environmental Protection Act (1993), the Pesticides Act (1962), and the agreements of the North-Sea Ministers Conference in Esbjerg (1995). However, the practical approach of the Pollution of Surface Waters Act does not quite match with the WFD. Apart from this it seems that the Dutch standards for emission-control are more stringent than those described in the WFD, due to Dutch weather circumstances and special condition of water bottoms. In Phase 3 this Working Group will investigate other significant human influences like the abstraction of groundwater, water level control, morphology transformations and land-use, and activities in coastal waters. Further the Working Group will implement the outcome of the project analysis of pressures and impacts and prepare a new chapter Hazardous Substances and Products for the Environmental Protection Act. Working Group Aims and Objectives and Monitoring All types of water, groundwater, inland water, transitional water and coastal water, have their own quality-references or criteria and assessmentsystems. The Dutch assessment-systems generally fit in the WFD. The standard for artificial water bodies or heavily modified water bodies, as most of all Dutch waters are classified, is a good ecological potential instead of a good ecological status. Following general economical principles it is also allowed to restrict oneself to a good ecological potential if the costs to gain a good ecological status are disproportional. On a national level The Netherlands know the Maximum Admissible Risk (MAR, target for 2006), and Negligible Risk (NR, target for 2010), as aim for the chemical condition of water. Both are comparable with good. Regarding the ecological status, the possibilities to develop a new assessment-system to distinguish very good, good, moderate, insufficient and bad, have been studied in Phase 2. The development of a new system will take years and requires a broad support from the Dutch water managers. In Phase 3 the Working Group will continue their study with e.g. the standards for drinking water, the need of additional quality demands regarding the oxygen percentage of surface water in relation to the suitability for fishes like salmon, a reference for The implementation of the Water Framework Directive in The Netherlands. Page 11 of 14

the quality (chemically and ecologically) of groundwater and subsequently groundwater quality monitoring. Three types of monitoring (surveillance ~, operational ~ and investigative ~) are distinguished in the WFD. Good monitoring measures require adequate frames of references and good communication between neighbours. The elaboration of the monitoring obligations of the WFD was almost finished at the end of Phase 2. Generally the conclusion is that the currently displayed monitoring efforts meet the requirements as stated in the WFD, except for monitoring in protected areas. Further, Phase 3 will concentrate on the groundwater issue and the reporting obligations. For this last issue the Working Group will contact the international EU-monitoring Working Group. Working Group Finances and Economical Analysis The costs of water management in The Netherlands are for the greater part already financed by direct or indirect taxes or prices for water consumption. The water quality management knows a pollution levy for the dumping of effluent in sewerage of rivers. The State and Water Boards levy various taxes to an annual amount of approx. b 1,1 based on the amount of pollution, in conformity with the polluter pays principle as stated in the WFD. Households, industries and farms are charged with a head tax by the Water Boards to finance water quantity management (approx. m 370 annually). The maintenance and management of the Municipal Sewerage is financed via levies paid by households and industry for an annual amount of b 0,64. The Provinces levy m 15 annually for groundwater discharges and adhere to the policy that good quality groundwater has to be artificially recharged to prevent the loss of groundwater to less eminent purposes than e.g. drinking water. Drinking water is not free in The Netherlands. The drinking water companies charge the users of the water per cubic meter (approx. 1,- to 1,5) resulting in an annual amount of b 1,6. Approx. 70% of the drinking water in The Netherlands originates from groundwater. Furthermore there is an Environmental Tax (approx. m 165 annually) on groundwater extractions per cubic meter. This tax is not (yet) used for water management purposes but is used for general purposes (public treasury). The total recovery of the costs of water services in The Netherlands, although collected decentralised through levies and prices, is calculated at b 3,9. In contrast with groundwater discharges, abstractions from surface waters are in most cases free of charges. These surface water abstractions do require at the same time a permission, so it can not be too difficult to introduce a levy per cubic meter discharged water. Apart from this issue there is a question of fairness regarding the contribution that is expected from households, industries and farmers, for water services. In Phase 3 the economic implications of the WFD will be further investigated. National Administrative Organisation This Working Group was established in Phase 3 to integrate WB21. Task of this Working Group is to elaborate an implementation model for the period after Phase 3 and to co-ordinate the junction of the 4 separate national river basin district management plans, if possible merged with the existing and new to erect consultative bodies emanating from the WB21. Further the Working Group will make a register of Protected Areas or Reserves and define the responsibilities for protective measures. Finally this group will provide insight in the costs that are involved with the implementation of the obligations of the WFD and suggest ways to gain the necessary funds. Working Group Communication and/or Manual WFD Because of the comprehensive and sometimes indefinite wording in the articles and appendixes of the WFD, it can not be excluded (or must be expected) that various interpretations will evolve regarding the possible consequences of the WFD for the (inter)national policies for water, environment, nature and spatial planning. Lucid exposition and univocal communication is therefore necessary. For example: it has to be clear that the WFD does not result in changes of the administrative system. The communication also must lead to awareness of the future tasks for the water managers so that they can take into account the necessary budgets and personnel. The implementation of the Water Framework Directive in The Netherlands. Page 12 of 14

The Working Group distinguishes 3 levels of communication: At international level the activities are connected with the Working Group Co-ordination International Implementation. At national level the idea is to organise workshops for national organisations or platforms of Water Managers and ~Boards (UVW), Provinces (IPO) and Municipalities (VNG) but also for the inland navigation, (fishery-) industry, the recreational sector, agricultural organisations, drinking water companies, the mineral resources sector and preservation- and environmental organisations. The approach for the regional communication will very much be the same on but on a regional level. The Working Group will compile a manual that will be released in the end of 2002, to promote the implementation of the WFD. This document will disclose the consequences of the WFD (part I) and the information necessary to meet the requirements of the WFD, and what will be expected from the persons and organisations concerned (part II). The manual will be put together with the contributions of the various Working Groups and contains all relevant information that is compiled during the 3 phases. The manual will not be a static document but will be more dynamic (frequently updated) and made available via the Internet. With regard to water there is quite a lot of information available on the Internet on the webte www.waterland.net that holds numerous sites si (both commercial and not commercial). Since September 2000 the Project Group releases bulletins with the title the State of Affairs. The bulletins are very accessible and quite understandable for the public. THE WFD: OPINIONS The next few opinions are a mixture of the authors and others, who s names will be mentioned in between brackets [..]. The opinion that water is not a commercial product like any other but, rather, a heritage which must be protected, defended and treated as such, is fully supported by the authors of this paper. This message in the WFD casts up a dam against the excessive commercial attitude that we made ourselves familiar with [Havekes, 5]. At the same time it is worthwhile to recognise of the considerations with which the WFD-document begins, rather than the actual text of the Directive only. Water management is not a separate aim of the European Union but embedded in European Environmental Law. However, water management appears to be broader than that, and it can be expected that in the near future, water quantity management will claim a more eminent role. Water management deserves to be an aim by itself within the European Union-treaty. Some do even plead for a special European Commissioner for Water. For now it must be concluded that the WFD is clear in its strategies in fighting, preventing and controlling water~ and groundwater-pollution (articles 16 and 17), but that there are no such strategies for fighting floods and droughts, which are only mentioned briefly in Article 1.e and 11.3.e. Saager [5] stated that the WFD, as a document that is focussed on water quality only, is a set back for The Netherlands with over 10 years of experience in Integral Water management, neither does it add something to the Treaty of Helsinki (1992). The reason for this partiality is both simple and disappointing. For water quantity unanimity of votes is required in compliance with the Treaty of Maastricht, and that was considered unfeasible by the European Council in an attempt to reach an ambitious and broadly founded Directive within a reasonable period of time. In The Netherlands, the WFD will be linked with WB21 to clear this omission. The WFD integrates earlier enacted Directives (as stated in Article 22). Havekes [5], mentioning the WFD as an example, pleads for a Dutch Integrated Water Act to do the same on a national level, to obtain a transparent, effective and coherent legislative framework. As mentioned earlier in this paper, the WFD affects at least 10 different laws in The Netherlands and three Ministries are involved. We could have saved ourselves the staple through the Dutch legislation if we have had one Integrated Water Act. Being the sink of a large part of Europe, The Netherlands benefits from the WFD. It is plausible that the International Committee for the Protection of the Rhine, with its successful Rhine Action Programme (RAP, in the beginning focussing of water quality, later extended with high water, ecological restoration and groundwater), has been a source of inspiration for the realisation of the WFD. The WFD is (for The Netherlands) The implementation of the Water Framework Directive in The Netherlands. Page 13 of 14

considered to be a good supplement to the existing Sciences (IAHS); international treaties. 3. Unie van Waterschappen (2000) CD-Rom of Dutch Water Boards The WFD advances the international co-operation within river basins to tackle important issues in 4. Gans, de. W. (2000 NITG TNO) The Geology of The Netherlands water management. Not as usual by rigid and juridical coercive regulations by a non-co-operative EU Commission, which apparently does not work very well, but by breaking with traditions and encouraging a process of international co-operation. In the next few years during the implementation of 5. Havekes H., Spier J., Saager P., Sprundel van L., Latour P., and Velde van de O. (July 001) Europese Kaderrichtlijn Water : A syllabus of articles concerning the WFD as printed in magazine Het Waterschap ( the Water Board ) the WFD it will become clear if this is really working. [5] 6. Ministry of Transport, Public Works and Water management (2000) The WFD: an intermediate report WB21 proposes a different approach to water 7. Project Group Implementation of the WFD management in The Netherlands to deal efficiently (1st March 2001); Implementation of the with climate-change, sea-level-rising and WFD: Phase 2 continuing soil-subsidence and not only with safety, in-conveniences, material and immaterial damage 8. Project Group Implementation of the WFD (5th March 2001); Implementation of the and cost. What is required is a water-driven strategy WFD: SOW Phase 3 and priority to water management in area planning. 9. Project Group Implementation of the WFD (September 2000, June 2001, December A strategy in which water is regarded not as enemy, but as ally, with room to move. 2001, March 2002); WFD: The State of Affairs 1 to 4 10. Tielrooij F. et al. (31st August 2000) Water management in the 21st Century REFERENCES 11. Ministry of Transport, Public Works and 1. European Parliament / European Council Water management (December 2000); A (23rd October 2000) Directive of the Different Approach to Water, Water European Parliament and of the Council Management Policy in the 21st 2000/60/EC Establishing a Framework for Century (WB21) Community Action in the field of Water Policy 12. Christa Jesse; artist Illustrations/cartoons 2. Huisman P., Cramer W., Ee van G., Hooghart J.C., Salz H. and Zuidema F.C. (from 10) (1998) Water in The Netherlands, Figures 2, 3, 4, 5, 6 and 7 from [2] Netherlands Hydrological Society (NHV) and Figures 1 and 8 from [10] The Netherlands National Committee of the International Association of Hydrological Retain. Storage. Discharge. The implementation of the Water Framework Directive in The Netherlands. Page 14 of 14

Celtic Water in a European Framework: Pointing the Way to Quality Galway 8-10 July 2002 Celtic Groundwater: Current Pressures and Future Responses Bruce D.R. Misstear Department of Civil, Structural & Environmental Engineering, Trinity College, Dublin 2, Ireland, email: bmisster@tcd.ie Nick S. Robins British Geological Survey, Crowmarsh Gifford, Wallingford, Oxfordshire OX10 8BB, UK Abstract The Celtic regions of Ireland and Britain have much in common with regard to groundwater occurrence and groundwater management. The hydrogeology is dominated by fracture flow in bedrock formations, and by an extensive covering of glacial and other Quaternary deposits. Groundwater development is mainly on a small scale for rural water supplies, but there is a very large number of wells and springs. Groundwater plays an important part in the cultural heritage of the Celtic lands, as indicated by the many holy wells. Groundwater also has an important environmental role, for example, as baseflow to rivers and wetlands. There are many pressures on Celtic groundwater, on both its quality and its quantity. Groundwater quality is under threat from numerous point sources of pollution, including septic tank systems, farmyards and abandoned mines, and from diffuse pollution from agricultural fertilisers. In response, comprehensive groundwater protection policies have been introduced in England, Wales, Scotland and in both parts of Ireland. Regarding groundwater quantity, the bedrock aquifers generally have a low storage coefficient and so are susceptible to dewatering effects from over-pumping. Making aquifer withdrawals subject to a licensing process can alleviate some of the pressures on groundwater quantity; however, to date, abstraction licensing procedures are in place only in those Celtic aquifers lying within England and Wales. This situation will change with the implementation of the European Union Water Framework Directive. One of the challenges that hydrogeologists will face in quantifying groundwater resources is the scarcity of reliable hydrogeological data throughout the Celtic regions. Also, much remains to be done in meeting the directive s aim that groundwater and surface water should be managed in an integrated fashion. This will necessitate an improved understanding of aquifer-stream relationships. Gearrchuntas Uisce Talún Ceilteach: Brú agus an Todhchaí. Is cosúil le chéile iad na ceantair Cheilteacha in Éirinn agus sa Bhreatain Mhór maidir le cursaí uisce talún. Lonnaíonn usice talún na dtíortha Ceilteacha seo, den chuid is mó, i scoilteanna sa bhuncharraig agus ins an dríodar a leagadh síos i rith na hoighearaoise. Is beag forbairt ar uisce talún atá déanta go fóill ach ar scála ana bheag faoin tuath: ach tá cuid mhór toibreacha agus fuaráin againn. Bhí uisce talún riamh tábhachtach i ndúchas cultúrtha na dtíortha Ceilteacha, mar is léir, ó na toibreacha beannaithe atá fairsing. Cothaíonn an t-uisce talún na h-aibhneacha agus na riasca comh maith. Is iomaí brú atá ar an uisce talún Ceilteach, ó thaobh truaillithe agus flúirse de. Tá foinsí truaillithe ag bagairt air - dabhaigh múnlaigh, clósanna feirme agus mianaigh tréigthe, leasúchán talmhaíochta. Tá polasí cuimsitheach chun uisce talún a cosaint, dírithe ar na 1

fadhbhanna seo, curtha I bhfeidhm sa Bhreatain Bhéag, in Albain, i Sasana agus sa dá chuid d'éirinn. Maidir leis an méid uisce talún atá ar fáil, de gnáth ní bhíonn taisce mhaith uisce ins na buncharraigreacha agus dá bhrí sin is fuirist an taisce atá ann a laghdú go ró mhór tré iomad phumpála. Más gá ceadúnas do uisce talún a bhaint, d fhéadfaí aon laghdú a smachtú. Faoi láthair áfach níl ceadúnas riachtanach ach amháin i Sasana agus sa Bhreatain Bheag. Ach réiteofar sin nuair a chuirfear european Union Water Framework Directive (Ordú Phlean Uisce an Aontais Eorpaigh) i bhfeidhm. Ceann de na dúshláin is mó atá ag na h-eolaithe a bheidh ag gabháil do chursaí uisce talún ná a laghad eolais atá ar fáil faoi uisce talún ins na réigiúin Cheilteacha. Tá a lán le déanamh comh maith chun aidhm an Directive, go mbeadh uisce talún agus uisce srutha riartha i slí iomhlánach, a chur i bhfeidhm. Ní foláir mar sin, tuiscint níos fearr a bheith againn ar chursaí uisce in a iomláine INTRODUCTION The aims of this paper are to provide an overview of the hydrogeology of the Celtic regions of Ireland and Britain, and especially to highlight the importance of groundwater to the Celtic peoples and the environment. Threats to groundwater quality and quantity are described, together with existing or planned responses to counter these pressures. The paper ends with a discussion of the European Water Framework Directive and its implications for the future management of groundwater resources in the Celtic regions. 2

HYDROGEOLOGY The hydrogeology of the Celtic regions of Ireland and Britain is dominated by fracture flow in ancient bedrock formations. These include Precambrian igneous and metamorphic rocks, Lower Palaeozoic sandstones, quartzites and volcanics, and Carboniferous limestones. Groundwater yields from most of these formations are relatively low, with the major exception of the Carboniferous limestones where these rocks have undergone permeability enhancement through karstification or dolomitisation. The Permo-Triassic sandstones and Cretaceous Chalk that form the main aquifers of central, eastern and south eastern England do not occur extensively in the Celtic lands. Nevertheless, Permo-Triassic sandstones form important local aquifers in the Lagan valley, the Vale of Clwyd and in south-west Scotland including the Dumfries basin. There is a significant Chalk deposit present in Northern Ireland, the Ulster White Limestone, but this tends to be more indurated and less permeable than the Chalk of England. Alluvial and fluvio-glacial deposits are important sources of groundwater in the Celtic lands. Lower permeability glacial tills are also important hydrogeologically, in that they provide protection against pollution to the underlying aquifers. A generalised three-fold classification of aquifers is given in Table 1, including some typical values for aquifer properties (transmissivity and storativity) and well yield. The isotropic nature of fractured bedrock aquifers, even those classified as regionally important, is illustrated by the data on well performance for 225 Carboniferous limestone wells in the west of Ireland (Table 2). Well yields range from 76 l s -1 down to zero, with a relatively low mean value of 2.4 l s -1. Table 1 Aquifer characteristics in the Celtic Regions (adapted from Robins & Misstear 2000) Class Geology Properties Transmissivity (m 2 d -1 ) Storativity Borehole yield (l s -1 ) Regionally important bedrock aquifers Carboniferous Limestone; Permo- Triassic basins; some Devonian sandstones; Some volcanics Anisotropic; fracture flow dominant; regional and local flow paths 100 to 4 000 0.01-0.20 5 to 40 Locally important bedrock aquifers Precambrian and Lower Palaeozoic; Some Upper Palaeozoic; Some volcanics Anisotropic, secondary porosity dominant; local flow paths 20 to 100 <0.05 1 to 5 Superficial aquifers Alluvium; granular glacial deposits; Raised beach; and wind-blown sand Anisotropic, rarely isotropic; primary porosity; resource potential is limited by geometry. 50 to 5 000 Variable 3 to 40 3

Table 2 Summary of performance characteristics of 225 Carboniferous Limestone wells in the west of Ireland (after Drew & Daly 1993) Variable Maximum Minimum Mean Depth (m) 177 3 57 Yield (l s -1 ) 76 0 2.4 Specific capacity (l s -1 m) 7.6 0 0.8 Alluvial aquifers are usually regarded as relatively isotropic, but this assumption is not necessarily valid. Table 3 shows well performance data for alluvial wells in Scotland. Even within a single wellfield along a 3 km reach of the River Spey, there are considerable variations in yield and specific capacity (Table 3a). Data from other alluvial sites across Scotland naturally show even more variation (Table 3b). Table 3 Summary of performance characteristics of alluvial wells in Scotland (test data taken from Jones & Singleton 2000) a) Spey wellfield (36 wells) Variable Maximum Minimum Mean Depth (m) 16.9 8.9 13.2 Yield (l s -1 ) 24.4 6.7 16.7 Specific capacity (l s -1 m) 19.3 1.6 3.4 b) Other alluvial wells (27) Variable Maximum Minimum Mean Depth (m) 34 2.8 15.4 Yield (l s -1 ) 27.3 0 7.6 Specific capacity (l s -1 m) 28.6 0 4.5 IMPORTANCE OF GROUNDWATER The importance of groundwater to the Celtic peoples can be considered under three headings: drinking water, culture and environment. Drinking Water Groundwater provides between 3 and 15% of public water supplies in the Celtic regions (Robins & Misstear 2000). In addition, groundwater is used widely for private water supplies and may account for close to 100% of the water supply in some rural areas. Significantly, there is a very large number of wells e.g. Wright (1999) has estimated that there may be as many as 200 000 wells in Ireland alone. The majority of the public schemes supplied from groundwater are relatively small. However, there are some major groundwater schemes e.g. the Schwyll Carboniferous limestone spring near Brigend in South Wales supports an abstraction licence of nearly 22 Ml d -1 (Hobbs 2000), and the Spey alluvial wellfield in Morayshire has a design yield of 27 Ml d -1 (Jones & Singleton 2000). 4

In addition to public and rural water supplies, groundwater is widely used for agriculture (notably livestock watering) and industry (including distilleries, breweries and creameries). A relatively recent phenomenon has been the development of a large number of bottled water plants in Ireland, Scotland and Wales. Most modern groundwater exploitation is via drilled wells or springs, with hand dug wells remaining important for private supplies. Shallow alluvial groundwater is occasionally abstracted using infiltration galleries e.g. in Athy and Castlecomer in south-east Ireland, at Ely wells near Cardiff and at Ordiequish on Speyside (Misstear et al. 1980, Jones & Singleton 2000). Culture Holy wells are an important feature of local culture throughout the Celtic lands. Logan (1980), quoting an earlier source, suggests that there may be 3000 holy wells in Ireland. Healy (2001) points out that the Irish Townlands Index lists 163 place names beginning with the word tobar (or some version of this word). Although there are few Scottish place names that include the word tobar, the Welsh version ffynon is more common. Many of the wells are still visited regularly, especially on the pattern (or patron) day. Votive offerings such as rags, statues and coins are common. Great healing powers are attributed to many holy wells, especially for curing eye ailments, but also e.g. backache, whooping cough and toothache. Holy wells seem to occur in a huge variety of hydrogeological environments: in superficial and bedrock aquifers, on mountain tops and in lowland bogs; on islands and along shorelines. Many appear not to be wells at all, but are shallow rock depressions fed by rainwater or are ponds or small lakes. Environment It has long been recognised that groundwater contributions to baseflows in rivers are important in lowland areas underlain by major aquifers. Recent research suggests that low permeability bedrock in upland areas may also make a significant contribution to baseflow. This is the case in the upper Severn and Wye catchments in Wales, where the rivers flow over low permeability Ordovician and Silurian argillaceous rocks (Neal et al. 1997; Shand et al. 1999). Shallow groundwater is present owing to the development of intense shallow fracturing and weathering formed during periglacial activity at the end of the last glaciation. Deeper groundwater occurs in fractures developed along bedding planes and structural discontinuities. As well as river baseflow, groundwater discharge can be critical to maintaining sensitive wetland habitats. An important and topical (see below) example in Ireland is Pollardstown fen in County Kildare, a proposed Special Area of Conservation. This is fed by springs and by groundwater seepages rich in calc tuffa, crucial to the habitats present (including the vertigo geyeri snail). PRESSURES The main pressures on groundwater quality and quantity are described in turn below. In view of the modest aquifer yields and limited size of most existing groundwater developments, 5

together with high potential recharge, the pressures on groundwater quantity are arguably less critical than those on groundwater quality. Groundwater Quality Pressures on water quality are summarised in Table 4, and encompass both point and diffuse sources of pollution. The discussion that follows is summarised from Robins & Misstear (2000). Table 4 Main sources of groundwater pollution in the Celtic regions Category Point sources Diffuse/linear sources Source Septic tank systems Farmyards, silage pits, etc Sheep dips Landfills/waste dumps Industry Petrol stations Mining Fertilisers (inorganic) Landspreading of organic wastes Pesticides Urban areas/housing developments Roads Many of the small private well abstractions are affected by local point source contaminants, especially microbiological. The problems are often exacerbated by a combination of factors: poor well construction, especially the absence of a proper sanitary seal to the upper casing; high aquifer vulnerability; the presence of a nearby contamination source such as a septic tank system, itself often badly constructed; and a low well abstraction rate and hence low potential for dilution. Microbiological contamination is usually detected by the presence of faecal bacteria. However, there is growing concern that other microbiological pollutants such as viruses and protozoa (including cryptosporidium) could be present in wells that are susceptible to pollution from livestock activities. Chemical contaminants associated with point sources in rural areas include chloride, sodium, potassium, iron, manganese, ammonia and nitrate. High nitrate concentrations also result from diffuse pollution by inorganic and organic fertilisers. In Scotland, high nitrate levels are found widely in areas of intensive cultivation or grassland in eastern and southern Scotland (Ball & MacDonald 2001). In Ireland, groundwater contamination by nitrate is generally only widespread in the intensive arable areas in the east and south (Daly 1994). A national survey of groundwater quality in Ireland between 1995 and 1997 identified nitrate at above the MAC of 50 mg l -1 NO 3 at only 2.5% of the sites sampled, and concluded that nitrate pollution was generally a localised problem (Environmental Protection Agency 1999). Pollution from synthetic trace organic compounds has been a growing problem over the last two decades. The most prevalent compounds are petroleum hydrocarbons, chlorinated solvents and pesticides. The risk posed to the aquatic environment by the disposal of spent sheep dip is a particular Celtic concern. 6

The abandonment of old mines has led to serious pollution problems in the Celtic regions. Important historical mining activities included: tin in Cornwall; copper in Ireland, Wales and Cornwall; coal in South Wales, Northumbria and Scotland. One of the worst pollution incidents occurred in Cornwall in 1992, when a large discharge of acid mine drainage from the former Wheal Jane tin mine contaminated the River Carnon, resulting in concentrations of cadmium and zinc in the river of up to 600 μg l -1 and 500 mg l -1 respectively (National Rivers Authority 1994). At the Avoca copper mines in County Wicklow, which were finally closed in 1982 after two centuries of working, mine drainage with a ph of about 3.5 is discharging to the surface rendering the Avoca river effectively biologically dead for a reach of several kilometres (Wright et al. 1999). In Wales, over 50 km of rivers are adversely affected by discharges from abandoned coal mines (National Rivers Authority 1994), and there are reported to be more than 80 adits from coal and oil shale mines in the Midland Valley of Scotland that are releasing polluted mine water to surface water sources (Wood et al. 1999). Groundwater Quantity Pressures on groundwater quantity may arise from several sources, including water supply over-abstraction, dewatering for construction sites and climate change. The fractured bedrock aquifers have low storativity and so can be vulnerable to dewatering effects due to overpumping; however, these problems are likely to be localised in view of the low transmissivity and small size of many of the Celtic aquifer units. Potential impacts from such schemes warrant serious study especially where sensitive wetland habitats may be affected. In Ireland, the ongoing Kildare town bypass involves a cutting through the regionally important Curragh gravel aquifer which feeds the important habitat site of Pollardstown fen, as well as springs that supply the Japanese Gardens in Kildare, a major tourist attraction. Detailed field studies and groundwater modelling have led to a novel method of lining the cutting to minimise the long-term impacts of dewatering once construction is completed (Hayes et al. 2001). The implications of climate change on recharge need to be considered. Although there is significant rejected recharge in Celtic situations, and winter rainfall is likely to increase according to many prediction scenarios (e.g. Arnell & Reynard 2000), a decrease in summer rainfall and increase in evapotranspiration (again as predicted in a number of scenarios) would lengthen the period of soil moisture deficit and thus prolong the period when relatively little recharge occurs. This might have a significant impact on groundwater levels, baseflows and spring discharges in late summer, since most of the bedrock aquifers have low storativity. RESPONSES Groundwater Quality In response to the pressures on groundwater quality, comprehensive groundwater protection policies have been introduced in each of the Celtic regions. The policies all recommend the zoning of the land surface according to the risk to groundwater from potentially polluting activities. The land surface zoning encompasses the aquifer vulnerability, the aquifer class and the delineation of protection zones around individual major wells and springs (Table 5). Policy statements are included for different land uses. These policies are intended to provide guidance for planners, developers and water resource legislators, and can be effective tools for protecting groundwater quality from future land use developments. 7

Table 5 Comparison of the main features of land surface zoning for groundwater protection in Ireland and Britain (after Misstear & Daly 2000) Republic of Ireland 1 England and Wales 2 Scotland 3 Northern Ireland 4 Groundwater vulnerability classification Topsoil (high, intermediate and low leaching potential) Topsoil (high, intermediate and low leaching potential) Topsoil (high, intermediate and low leaching potential) Subsoil characteristics (thickness and permeability) Drift (usually shown as undifferentiated stipple on maps) Drift (usually shown as undifferentiated stipple on maps) Drift (usually shown as undifferentiated stipple on maps) Depth of unsaturated zone (sand and gravel aquifers only) Depth of unsaturated zone (not shown on maps) Depth of unsaturated zone (not shown on maps) Depth of unsaturated zone (not shown on maps) Nature of recharge (point or diffuse) Vulnerability maps 1:50 000 scale 1:100 000 scale 1:100 000 scale 1:250 000 scale (regional map) Source protection zones Inner Source Protection Area (SPA) (100 day travel time) Zone I (Inner) (50 day travel time, minimum 50 m radius) Zone I (Inner) (50 day travel time, minimum 50 m radius) Zone I (Inner) (50 day travel time, minimum 50 m radius) Zone II (Outer) (400 day travel time) Zone II (Outer) (400 day travel time) Zone II (Outer) (400 day travel time) Outer SPA (zone of contribution) Zone III (Source Catchment) (zone of contribution) Zone III (Source Catchment) (zone of contribution) Zone III (Source Catchment) (zone of contribution) Aquifer classification Regionally important Locally important Major Minor Highly permeable Moderately permeable Highly permeable Moderately permeable Poor Non-aquifers Weakly permeable Weakly permeable Sources: 1 Department of the Environment and Local Government et al. (1999) 2 Environment Agency (1998) 3 Scottish Environment Protection Agency (1997) 4 Department of the Environment for Northern Ireland (1994 and 1999) The groundwater protection policies do not address the issue of pollution from existing or historical land uses. A risk-based strategy is usually adopted in deciding on the need for groundwater remediation and the best method(s) to achieve this, on a site by site basis. Decision-making can be aided by the introduction of Environmental Quality Objectives (EQOs) and Environmental Quality Standards (EQSs). Although EQOs and EQSs have 8

tended to focus on surface waters (e.g. EPA 1997), work is underway to develop specific guideline values for the protection of groundwater in Ireland. Exceedances of these guideline values would act as a trigger for a risk-based assessment of any further actions required. The GVs are likely to be based on drinking water criteria, but also to take account of the expected background concentrations for different parameters in Irish groundwaters. The setting of GVs can be regarded as a bottom up approach to groundwater protection, in that it is responding to what is actually detected in the groundwater. As such, it can augment the top-down approach to groundwater protection that underlies the groundwater protection policies in Table 5. Standard procedures are needed for delineating protection areas around private groundwater sources, where abstractions are normally small and intermittent, and there is a scarcity of hydrogeological data. There are a number of logical schemes for creating source capture zones, the area of which is given by the infiltration from rainfall recharge and soil interflow as well as overland flow transported into the zone which contribute to the source, i.e. abstraction rate or discharge divided by recharge (Robins 1999). The shape of the zone depends on the prevailing topographical and geological controls but can be deduced qualitatively to create a sensible capture zone. Groundwater Quantity With the exception of the larger Permo-Triassic aquifers, it is uncommon for serious consideration to be given to the water budget of aquifers in the Celtic lands. The reasons are twofold: the perceived higher rainfall of much of these lands and the perception that the aquifers are too small to allow a sensible understanding of renewable resource. Besides usage is generally small and detailed analysis may be unwarranted. There are, however, a number of issues that do need to be addressed to enable proper management of Celtic groundwater in the future; in particular, there is a need for a better understanding of: aquifer properties, especially storativity, of the bedrock aquifers found in the Celtic regions; the interrelationships between groundwater and surface water, and the links between these water sources and ecology; groundwater recharge, including recharge processes and rates through the low permeability tills that are so common in the Celtic regions. THE EU WATER FRAMEWORK DIRECTIVE AND ITS IMPLICATIONS FOR FUTURE GROUNDWATER MANAGEMENT Many of the future requirements are encapsulated in the new EU Water Framework Directive. This will require River Basin District (RBD) Plans to be drawn up and implemented (European Commission 2000). The directive requires an integrated approach to monitoring and managing all waters in a RBD. Groundwater must be characterised in terms of hydrogeology, protection, pressures and dependent ecosystems. Integrated management of groundwater with surface water will not be straightforward in the Celtic regions where aquifer units tend to be small and localised, and consideration of individual catchments within each river basin unit will be required to enable resource management. This will necessitate a far more extensive network of flow gauging, for example, to assess the baseflow contributions from individual aquifer units. Implementation of the Water Framework 9

Directive will also involve the introduction of abstraction licensing in Ireland, Northern Ireland and Scotland. The directive begs a number of questions, not least in defining otherwise accepted terms such as aquifer and groundwater. The term aquifer is interesting in the Celtic sense when is an aquifer so small that it no longer constitutes an aquifer? The directive s answer is that it can be very small indeed and needs only to be capable of yielding 10 m 3 d -1. The significance of the small aquifers lies in the emphasis in the directive on environment and habitat, not just abstraction. The throughput of groundwater (of a specific quality and quantity) to surface water is, therefore, going to be the key to identifying and characterising the smaller Celtic aquifers. Some of the diverse and isolated aquifer units may need to be gathered together, perhaps as groups of similar or like groundwater conditions occurring within a given surface water catchment, which together will provide units of a suitable size for management purposes. Whatever the future may hold, it is clear that the holistic management of surface and groundwater on a catchment wide basis will become the central feature. The next few years will be an exciting period in the development of the hydrogeological understanding of the Celtic lands. References Arnell, N. & Reynard, N. 2000. Climate change and UK hydrology. In: Acreman, M. (ed) The Hydrology of the UK: A Study of Change. Routledge. Ball, D. F. & MacDonald, A. M. 2001. Groundwater nitrate vulnerable zones in Scotland. British Geological Survey Report CR/01/250. Daly, D. 1994. Chemical pollutants in groundwater: a review of the situation in Ireland. In: Chemicals A Cause for Concern? Cork 3-4 November 1994, Sherkin Island Marine Station. Department of the Environment and Local Government, Environmental Protection Agency & Geological Survey of Ireland. 1999. A scheme for the protection of groundwater. Geological Survey of Ireland, Dublin. Department of the Environment for Northern Ireland. 1994. Groundwater vulnerability map for Northern Ireland. Environment Service, Department of the Environment for Northern Ireland. Department of the Environment for Northern Ireland. 1999. Policy and practice for the protection of groundwater in Northern Ireland. Environment Service, Department of the Environment for Northern Ireland Drew, D. & Daly, D. 1993. Groundwater and karstification in mid-galway, south Mayo and north Clare. Geological Survey of Ireland report RS 93/3, Dublin. Environment Agency. 1998. Policy and practice for the protection of groundwater. HMSO, London. Environmental Protection Agency. 1997. Environmental Quality Objectives and Environmental Quality Standards: the aquatic environment. A discussion document. Environmental Protection Agency, Wexford. Environmental Protection Agency. 1999. Water quality in Ireland 1995-1997. Environmental Protection Agency, Wexford. 10

European Commission. 2000. Directive 2000/60/EC of the European Parliament and of the Council of 23 October 2000 establishing a framework for Community action in the filed of water policy. Official Journal of the European Communities, L327. Hayes, T., Sutton, S., Cullen, K. & Faherty, J. 2001. The Curragh aquifer: current conceptual understanding and numerical modelling. In: Proc. 21 st Annual Groundwater Seminar, International Association of Hydrogeologists (Irish Group), Tullamore, 16-17 October 2001. Healy, E. 2001. In search of Ireland s holy wells. Wolfhound. Hobbs, S. 2000. Influent rivers a pollution threat to Schwyll Spring, South Wales? In: Robins, N.S. & Misstear, B.D.R. (eds) Groundwater in the Celtic Regions: Studies in Hard Rock and Quaternary Hydrogeology. Geological Society, London, Special Publications, 182, 113-121. Jones, C. & Singleton, A. 2000. Public water supplies from alluvial and glacial deposits in northern Scotland. In: Robins, N.S. & Misstear, B.D.R. (eds) Groundwater in the Celtic Regions: Studies in Hard Rock and Quaternary Hydrogeology. Geological Society, London, Special Publications, 182, 133-139. Logan, P. 1980. The Holy Wells of Ireland. Colin Smythe, Buckinghamshire. Misstear, B.D.R. & Daly, D. 2000. Groundwater protection in a Celtic region: the Irish example. In: Robins, N.S. & Misstear, B.D.R. (eds) Groundwater in the Celtic Regions: Studies in Hard Rock and Quaternary Hydrogeology. Geological Society, London, Special Publications, 182, 53-65. Misstear, B.D.R., Daly, E.P.D., Daly, D. & Lloyd, J.W. 1980. The groundwater resources of the Castlecomer Plateau. Geological Survey of Ireland Report RS 80/3, Dublin. National Rivers Authority 1994. Abandoned mines and the water environment. National Rivers Authority, Bristol. Neal, C., Robson, A.J., Shand, P., Edmunds, W.M., Dixon, A.J., Buckley, D.K., Hill, S., Harrow, M., Neal, M. & Reynolds, B. 1997. The occurrence of groundwater in the Lower Palaeozoic rocks of upland Central Wales. Hydrology and Earth System Sciences, 1, 3-18. Robins, N.S. 1999. Groundwater occurrence in the Lower Palaeozoic and Precambrian rocks of the UK: implications for source protection. Journal of the Chartered Institution of Water and Environmental Management, 13, 447-453. Robins, N.S. & Misstear, B.D.R. 2000. Groundwater in the Celtic regions. In: Robins, N.S. & Misstear, B.D.R. (eds) Groundwater in the Celtic Regions: Studies in Hard Rock and Quaternary Hydrogeology. Geological Society, London, Special Publication, 182, 5-17. Scottish Environment Protection Agency. 1997. Groundwater protection policy in Scotland. SEPA, Glasgow. Shand, P., Edmunds, W.M., Wagstaff, S., Flavin, R. & Jones, H.K. 1999. Hydrogeochemical Processes Determining Water Quality in Upland Britain. Hydrogeology Report Series of the British Geological Survey, Keyworth, Nottingham, 38 p. Wood, S.C., Younger, P.L. & Robins, N.S. 1999. Long-term changes in the quality of polluted minewater discharges from abandoned underground coal workings in Scotland. Quarterly Journal of Engineering Geology, 32, 69-79. Wright, G. 1999. How many wells are there in Ireland? The GSI Groundwater Newsletter, 35. Wright, G., Misstear, B.D.R., Gallagher, V., O Suilleabhain, D. & O Connor, P. 1999. Avoca mines: uncontrolled acid mine drainage in Ireland. In: Proc. International Congress of the International Mine Water Association Mine, Water and Environment for the 21 st Century: Mine/Quarry Waste Disposal and Closure, Sevilla, Spain, 13-17 September 1999. 11

Baseline groundwater quality: Contrasting geochemical evolution in selected Celtic and Viking aquifers. Paul Shand 1 and Bjørn Frengstad 2 1 British Geological Survey, Crowmarsh Gifford, Wallingford, UK. 2 Norwegian University of Science and Technology/Geological Survey of Norway, Trondheim, Norway. Abstract The geology of the Celtic regions and Norway display many similarities being dominated by hard-rock lithologies and a substantial surface water resource. Minor aquifers are also important in these countries and provide an important groundwater resource, particularly for rural communities. Water quality in many of the minor aquifers of both countries is often poorly documented and therefore the baseline conditions are difficult to assess. The effects of anthropogenic pollution may be important because the dominance of fracture flow may allow rapid transport to the water table with little time for pollutant attenuation or degradation. An attempt has been made to assess the baseline water quality in selected British and Norwegian aquifers. There is considerable overlap in water quality between the different aquifers, but several chemical differences are apparent which can be interpreted in terms of aquifer mineralogy and residence time. In addition, there are some significant differences between the UK and Norwegian aquifers studied. Most of the Norwegian groundwaters have relatively high ph compared with those of the UK (including carbonate aquifers), which is surprising considering that they are mainly composed of hard rocks. In addition, Na-HCO 3 type waters are much more prevalent in Norway compared to a dominance of Ca-HCO 3 type waters in the UK. The reasons for some of these differences may relate to subtle differences in mineralogy and exchangeable cations as a consequence of differences in sea-level following the last glaciation. Differences are also noted in several trace elements e.g. high F, U and Rn concentrations in many of the granite and sedimentary groundwaters in Norway and high barium in the UK sedimentary aquifers. 1. Introduction The Celtic regions contain abundant minor aquifers which provide important local resources, particularly in rural areas. Many of these are classed as hard-rock aquifers with storage largely confined to fracture networks. This is also the case in Norway, where much of the bedrock comprises Pre-Cambrian and Caledonian rocks. Such aquifers are often at risk from pollution because of rapid transport in lithologies dominated by fracture flow. These areas are often overlain by thin acidic soils making the rivers and groundwater prone to the effects of acidic deposition. Many such aquifers are poorly understood in terms of hydrogeology and the quality of the groundwater they contain. The complex and variable mineralogy and structure of many hard-rock aquifers make it difficult to constrain flow-pathways, to determine the natural baseline concentrations and make a reasonable risk assessment. The only waters which are likely to have escaped the effects of anthropogenic pollution are those which have been isolated from the atmosphere for some time and these generally occur as palaoewaters or those which are present under confined conditions. Such resources are limited and abstraction may be considered as groundwater mining and thus unsustainable. However, younger waters may be present for which many individual components are naturally-derived. Conversely, many waters contain concentrations of elements which, although naturally derived, exceed drinking water standards. Such natural contamination of groundwater should not be confused with anthropogenic pollution, and should also be considered as representing baseline. 1

The concept of chemical baseline is useful to discriminate natural inputs from those of anthropogenic origin. In addition, the baseline sets limits on the extent to which remedial measures may be necessary or indeed possible. The variations in bedrock geology, soil type and thickness, residence time of the groundwater, and the presence of drift deposits give rise to a range in the natural baseline concentrations. Therefore, it is necessary to establish criteria for defining precisely the term baseline: a simple average concentration is of little use in establishing a meaningful tool for use by policy makers and end users. However, it is possible to define baseline using simple mathematical tools or the use of simple statistical methods. 2. The Baseline Concept The term Baseline can be defined in different ways and a clear definition is required. The definition used here for groundwater baseline is the range in concentration of an element, species or chemical substance present in solution which is derived from natural geological, biological, or atmospheric sources. Terms such as background or threshold can have a similar meaning, but these terms are not generally used to infer no anthropogenic influence. There are several approaches to determining background water chemistry relative to pollution from a mining and smelter site which are of use in baseline studies: 1) up-gradient and cross-gradient sampling 2) extrapolation using historical water quality data 3) extrapolation from similar geochemical environments 4) geochemical modelling 5) statistical methods All of these methods have application in the determination of baseline geochemistry and can be used in conjunction to deduce the baseline chemistry of groundwaters. It is unlikely that there will be sufficient data available for all of these approaches to be successful, and it is recommended that these be used together to determine the natural baseline. One aspect of baseline geochemistry that has proved difficult to get across to policy makers and stakeholders is the variation that exists in nature; therefore, it is necessary to provide ranges in baseline values and not simply a single concentration. This range can be described in many ways e.g. by mean and standard deviation; the total range in values; or by describing a maximum baseline concentration. The value of traditional statistical parameters such as mean, median or mode are of use when comparing the baseline chemistry of different aquifers, but use of the median value is preferred because it is much more robust and less affected by outlying values. Many authors have used a limit to define an upper baseline e.g. Edmunds et al., (1997) and Langmuir (1997) chose values of the 95 th percentile and 97.7 % respectively. However, there is no a priori reason that anomalous or non-baseline data should fit such an arbitrary choice and the problems of representative sampling are paramount. Most of the outcrop areas of UK aquifers, for example, show the effects of enhanced concentrations of nitrate from agricultural practices. A simple cut-off value or threshold concentration would be meaningless as values below the threshold may not represent baseline. In addition, such techniques do not take into account the fact that anomalous and background concentrations often show significant overlap or have natural multimodal distributions. Traditionally, the use of statistical techniques has assumed that geochemical distributions are either normal, or more generally, lognormal (Ahrens 1954). Such an assumption was heavily criticised because the datasets studied did not generally display normal or lognormal distributions and therefore statistical techniques which assume such distributions are invalid 2

for their treatment. In most natural systems geochemical distributions are generally polymodal and are usually skewed (Reimann & Filzmoser, 2000). Probability plots are useful in highlighting different populations of data. In this context they could be of use in baseline studies to discriminate natural baseline chemistry from that due to pollution. In addition, they display very effectively the range and distribution of data and allow clear comparisons of water quality from different aquifers. However, it needs to be assessed to what degree natural variations produce multi-modal populations. There are several types of geochemical reaction which will alter distributions by removing or limiting concentrations in solution (Box 1). These include redox-processes e.g. in reducing environments nitrate may be completely removed from solution even from polluted waters; adsorption onto solid mineral phases may remove many trace metals or contaminants; and saturation with respect to a mineral phases will limit the solubility of one of more elements e.g. Ca and F through precipitation of fluorite. Box 1 Use of cumulative frequency diagrams to indicate baseline characteristics in groundwaters 90 CUMULATIVE FREQUENCY % 50 MEDIAN Below analytical detection limit Removal of trace concentrations Normal distribution Rapid solution of mineral to solubility limit Upper limit controlled by mineral solubility Saline mixing or pollution Bimodal distribution 10 LOG CONCENTRATION i) The median concentration is used as a single reference point for the element baseline which can be compared regionally or in relation to other elements. ii) Normal to multi-modal distributions are to be expected for many elements reflecting the range in recharge conditions, water-rock interaction and residence times under natural aquifer conditions. iii) Narrow ranges of concentration may indicate rapid attainment of saturation with minerals (e.g. Si with silica, Ca with calcite). iv) A strong negative skew may indicate selective removal of an element by some geochemical process (e.g. NO 3 by in situ denitrification). v) A narrow range in concentration at the upper limit may indicate a mineral solubility control (e.g. F by fluorite) vi) A positive skew most probably indicates a contaminant source for a small number of the groundwaters and this gives one simple way of separating those waters above the baseline. Alternatively the highest concentrations may indicate waters of naturally higher salinity. Other techniques which have been used to illustrate hydrogeochemical datasets include boxplots which display the median and/or the mean, specified percentiles and outliers; histograms (but these can show different distributions depending on buckets used for concentration); Piper diagrams, X-Y plots and maps of data showing the areal variation in concentration. In this report boxplots and probability plots will be used since these show 3

unbiased statistical data (although it must be remembered that analytical and sampling errors still exist). One other problem in characterising the groundwater chemistry of an aquifer or groundwater body is that most analyses relate to samples which are collected as discharge by pumping. Such samples may have been collected from a large screened interval and represent a mixture of chemically different groundwaters. 3. Aims of paper The aims of this paper are to present chemical data from selected aquifers in the UK (Cornish granites, Ordovician-Silurian metasediments of Wales, Devonian Sandstone of Scotland) and Norway (Pre-Cambrian granites, anorthosites and Ordovician-Silurian metasediments); to assess the variations in natural baseline quality, and to compare and contrast groundwater chemistry in these countries. 4. Summary of studied aquifers The aquifers selected from the UK include shallow groundwaters in the Cornish granites, groundwaters from the Ordovician-Silurian metasediments of Wales and groundwaters from the Devonian Sandstones of Fife in Scotland (Figure 1). Figure 1. Map of the UK showing the localities of the selected aquifers. The aquifers selected from Norway include the Precambrian Iddefjorden granite, the Precambrian Egersund anorthosite and Cambro-Silurian metasediments (Figure 2). 4

Figure 2. Map of southern Norway showing selected aquifer study areas. The Lower Palaeozoic metasedimentary rocks of Wales comprise mudstone and siltstones with minor amounts of greywacke and provide local supplies of groundwater, particularly for farms, but the resource is not significantly exploited. The data used here (Figure 1) are for groundwaters sampled in the Teifi Valley of south Wales (Robins et al., 2000) and from the Plynlimon experimental catchment in central Wales (Shand et al., 1996). Groundwater occurs both in deep fractures (generally bedding plane fractures) and in a shallow (< 10 m) weathered and fractured zone. Vertical stratification is apparent in some of these boreholes with acidic oxidising waters in the shallow system and alkaline reducing waters at depth (> 20 m). Groundwater also occurs in fractures and lodes of the Permian Carnmenellis granite in Cornwall. Water yields are generally small and sufficient only for small-scale water supply. Geothermal waters are found in deep fractures within the granite but the data presented here are from the shallow aquifer (Smedley et al., 1989). The Devonian aquifer in Fife represents one of the most important Scottish aquifers and is situated in a graben structure, the Midland Valley. The aquifer supplies 20 Ml d -1 during winter periods, increasing to 40 Ml d -1 in the summer due to abstraction for irrigation. Unlike the aforementioned aquifers, the Fife aquifer has relatively high porosity varying between 10 and 20% due to varying degrees of cementation by calcite and dolomite. The Precambrian Iddefjorden granite was formed during the Sveco-Norwegian orogeny in Middle Proterozoic (900 million years BP) and intrudes older metamorphic rocks. Most of the terrain has emerged from the sea during the last 10,000 years due to about 200 m of net isostatic uplift following the last glaciation. The primary porosity of the granite is negligible and practically all water flow takes place in fractures. Some of the most prominent fracture zones are filled with low-permeability clay minerals and thus yield little water. The boreholes have generally low yield and mainly supply single households and farms. The Egersund anorthosite province is part of the Rogaland igneous complex of Middle Proterozoic age. These also provide groundwater to single farms and households. The boreholes have a median normalised yield of 5.1 L h -1 per drilled metre (Morland 1997), one of the lowest normalised yields among the different lithological groups in Norway. The lower Palaeozoic 5

metasediments of Norway overlap with in terms of age with those in Wales, but show a much larger range in rock types from mudstones, shales to phyllite and mica-schist. The aquifers chosen show a large range in rock types and mineralogy and it is to be expected that large chemical differences will exist between the groundwaters. The selection of study areas was chosen to maximise similar rock types. However, anorthosites are rare in the UK being confined to a small area of outer Hebrides. Sedimentary aquifers are not very common in Norway which is dominated by hard-rock aquifers of Lower Palaeozoic or older age. 5. Hydrochemical characteristics of the groundwaters Details on the hydrochemistry have been presented elsewhere (Shand & Frengstad, 2000) and this paper will deal with comparisons between the study aquifers. This section will deal with the main hydrochemical characteristics of groundwater in the aquifers. The Chalk aquifer of England has been included on plots as a comparison of a well characterised major carbonate aquifer. 5.1 Major elements The relative proportions of major cations are shown on a piper diagram on Figure 3. Although there is a large variation and significant overlap, there are clearly defined groupings. 100 100 Ord/Sil metasediments (UK) Devonian sandstone UK) Permian granite (UK) PC Anorthosite (N) PC granite (N) Ord/Sil metasediments (N) 90 80 70 60 50 90 80 70 60 50 40 40 30 30 20 20 10 10 0 0 100 0 0 100 90 10 10 90 80 20 20 80 70 30 30 70 Mg 2+ 60 40 40 60 SO 4 2-50 50 50 50 40 60 60 40 30 70 70 30 20 80 80 20 10 90 90 10 0 100 90 80 70 60 50 40 30 20 10 100 100 0 0 0 10 20 30 40 50 60 70 80 90 100 Ca 2+ Cl - Figure 3. Piper diagram showing the relative concentrations of major elements. 6

The major element data are summarised using boxplots on Figure 4. One of the most striking features of the dataset is the variable range in ph of the UK aquifers and the consistently high ph of the Norwegian hard-rock aquifers. The Carnmenellis granite groundwaters and to a lesser degree, the Welsh groundwaters, are relatively acidic, typical of hard rocks with little buffering capacity. The Norwegian granites, in contrast have a much higher ph. The carbonate cemented Fife aquifer and the carbonate chalk aquifer have ph ranges close to that imposed by carbonate buffering but with median ph significantly less than the Norwegian groundwaters. The UK hard rock aquifers show many similarities being of low total dissolved solids (TDS), Na as an important cation and low Ca and HCO 3 concentrations. The Fife sandstone aquifer shows many characteristics of carbonate aquifers e.g. high Ca and HCO 3 (c.f. the Chalk) due to the dominant control of its carbonate cements. One distinctive feature of these groundwaters is the very high Mg concentrations as a consequence of dolomite dissolution. Often, these waters are supersaturated with respect to dolomite. 10 1000 9 ph Na 8 7 6 5 4 1000 Concentration (mg l -1 ) 100 10 1 1000 Cl HCO 3 Concentration (mg l -1 ) 100 10 Concentration (mg l -1 ) 100 10 1 1 100 Ca 100 Mg Concentration (mg l -1 ) 10 Concentration (mg l -1 ) 10 1 1 100 SO 4 NO 3 Concentration (mg l -1 ) 100 10 Concentration (mg l -1 ) 10 1 0.1 1 UK LP shale D sandstone P granite K chalk NORWAY PC anorthosite PC granite LP shale 0.01 UK LP shale D sandstone P granite K chalk NORWAY PC anorthosite PC granite LP shale Figure 4. Boxplots comparing ph and major elements in groundwater samples from selected British and Norwegian aquifers. 7

The Norwegian groundwaters also have relatively high HCO 3 concentrations but low Ca and the cations are typically dominated by Na giving rise to waters of Na-HCO 3 type. Although Na-HCO 3 type waters are found in the UK, these are normally present in deeper parts of the aquifers where the groundwaters are old and have undergone significant ion-exchange. Recent studies in the Lower Palaeozoic rocks at Plynlimon have found such waters at shallow depth (c. 20 m) which may be due an important input of sea salts from atmospheric deposition. The distribution of element concentrations is best displayed on cumulative probability plots and two examples are shown on Figures 5 and 6. The plot for Na shows that all of the aquifers display a considerable range in concentrations. The ranges are high in all of the Norwegian aquifers stretching over at least one order of magnitude whilst the three UK study areas show a much smaller range. The Chalk aquifer shows that two distinct populations are present, the higher concentrations coinciding with confined groundwaters which have undergone mixing with an older Formation water. These ranges are typical of most elements. The high concentrations of Mg in the Fife aquifer derived from dissolution of dolomite are clearly displayed on Figure 6 in comparison with the other study areas. In terms of baseline chemistry it is probable that most of the elements considered above represent baseline concentrations. Samples were collected only from boreholes which did not show any signs, or have a history, of point source pollution. However, it is clear from the nitrate concentrations (Figure 4) that most of the aquifers have been affected by diffuse agricultural pollution. Cumulative Frequency 99 98 95 90 80 70 60 50 40 30 20 10 5 LP metaseds UK D sandstone UK P granite UK K chalk UK PC anorthosite N PC granite N LP metaseds N 2 1 0.1 1 10 100 1000 Na (mg l -1 ) Figure 5. Cumulative probability plot showing the range in Na concentrations in the groundwaters. Note that although several groups have similar median concentrations, their ranges can be very different. The use of probability plots to determine a polluted population is difficult because the natural process of denitrification will lead to modification of such distributions as shown in Box 1. There is little historical data available for most of these aquifers, but on-going studies in the UK indicate that baseline concentrations of c. 2-6 mg l -1 are to be expected in oxidising 8

aquifers prior to agricultural intensification. It is difficult to estimate the degree to which other elements have been affected (e.g. Cl, K) by diffuse pollution because the ranges in the natural baseline are large and the additional inputs are within this range of natural concentrations concentrations of Mg in the Fife aquifer and the bimodal distribution in the Chalk aquifer. 5.2 Minor elements Boxplots for the minor elements are shown on Figure 7. Each of the aquifers shows distinctive ranges (often over several orders of magnitude). Some of the more obvious differences include the high Ba concentrations in the UK sedimentary aquifers and the very high Sr in the Chalk. The large ranges in Fe and Mn reflect the redox conditions in the aquifer with low concentrations in the unconfined oxidising parts and high concentrations where the waters are reducing. The neutral to alkaline ph in most groundwaters has kept Al concentrations low but the higher concentration in the acidic groundwaters in the Carnmenellis granite is evident. The Precambrian granites of Norway generally have high F concentrations but this is not seen to the same degree in the Cornish groundwaters. The granite groundwaters of both countries also have high U. There is a contrast in the metasedimentary aquifers with regard to U in that the Norwegian groundwaters have high U concentrations but the Welsh ones low concentrations (Figure 8). Cumulative Frequency 99 98 95 90 80 70 60 50 40 30 20 10 5 LP metaseds UK D sandstone UK P granite UK K chalk UK PC anorthosite N PC granite N LP metaseds N 2 1 0.1 1 10 100 Mg (mg l -1 ) Figure 6. Cumulative probability plot for Mg in the study aquifers. It is clear from this comparison of similar aquifers in the UK and Norway that rock type is not the only control on trace element species in these groundwaters. In order to understand why this is the case, it is necessary to look in detail at the potential reactions which occur in the aquifer. It is useful to study the changes which take place along flowlines, however, in hardrock aquifers where flowlines are complex or poorly understood this makes unravelling the hydrochemical evolution very difficult. Nevertheless, a knowledge of the aquifer mineralogy and the geochemical environment provide insights into the geochemical controls and evolution of groundwaters in these aquifers. 9

1000 10000 Concentration (µg l -1 ) 100 10 1 Ba Concentration (mg l -1 ) 1000 100 10 1 Fe 0.1 10000 Sr 0.1 1000 Mn Concentration (µg l -1 ) 1000 100 Concentration (µg l -1 ) 100 10 1 10 0.1 10000 F 1000 Al Concentration (µg l -1 ) 1000 100 Concentration (µg l -1 ) 100 10 1 10 1000 100 U 100 Li Concentration (µg l -1 ) 10 1 0.1 Concentration (µg l -1 ) 10 1 0.01 0.001 UK LP shale D sandstone P granite K chalk NORWAY PC anorthosite PC granite LP shale UK LP shale D sandstone P granite K chalk NORWAY PC anorthosite PC granite LP shale 0.1 Figure 7. Boxplots comparing minor elements in groundwater samples from selected British and Norwegian aquifers. 6. Controls on the baseline major and minor element chemistry The primary control on geochemical evolution in groundwater systems is the reaction between recharge and the aquifer minerals present along flowpaths. A series of reactions dominate over the time scales involved in groundwater flow including mineral dissolution, ion-exchange, adsorption-desorption and redox reactions. However, it is not always the major mineral phases which dominate the types of reaction involved because different reactions occur at different rates. The dissolution of evaporite and carbonate mineral for example are relatively fast whilst reactions involving silicate minerals are generally very slow. The dominance of Ca, Mg and HCO 3 in the Fife and Chalk aquifers is related to a dominant control by carbonate (calcite and dolomite) dissolution. Although the Fife aquifer is dominated by quartz, the kinetics of dissolution of quartz is extremely slow. The Fife aquifer does contain some silica but this is derived from the dissolution of other silicate minerals such as feldspar. The Welsh metasedimentary rocks are dominated by illite and chlorite which are relatively unreactive. The buffering capacity of the rocks to acidic rainfall is, therefore, relatively low and this is reflected in the relatively low TDS of the groundwaters. The 10

groundwaters are of variable type but many are of Ca-HCO 3 type due to the presence of minor amounts of calcite in fractures or present in drift above the bedrock (Robins et al., 2000). The larger range of concentrations found in the Norwegian metasedimentary aquifer groundwaters is partly due to the larger range in rock types and mineralogy. The anorthosites of Norway are basic rock types with mineralogy dominated by plagioclase (Ca-Na) feldspar and orthopyroxene (Mg-Fe silicate) and one would expect that the groundwaters would be of Caor Mg-HCO 3 type. The granites on the other hand would be expected to have poor buffering capacity and low ph similar to the Cornish granites. The relatively high ph and dominance of these Na-HCO 3 types in Norway require a different interpretation to the Ca-HCO 3 dominated groundwaters of the UK. The most likely interpretation is ion exchange reactions (Frengstad & Banks, 2000) during aquifer freshening (a process which is still seen in deeper UK groundwaters). Ion exchange reactions occur relatively rapidly in comparison with mineral dissolution. During aquifer freshening, Ca in the groundwater is exchanged for Na which is the dominant cation present on clays where saline or seawater has been present. Cumulative Frequency 99 98 95 90 80 70 60 50 40 30 20 10 5 LP metaseds UK D sandstone UK P granite UK K chalk UK PC anorthosite N PC granite N LP metaseds N 2 1 0.0001 0.001 0.01 0.1 1 10 100 1000 10000 U (µg l -1 ) Figure 8. Cumulative probability plot showing the range in U concentrations in the groundwaters. The Norwegian and UK aquifers studied all have relatively low Cl concentrations indicating that extensive flushing has occurred. The differences between the two sets of aquifers, therefore, must relate to the timing or intensity of flushing. Many of the Norwegian coastal aquifers were below sea level until relatively recently. Although these aquifers have subsequently been well flushed by fresh water, the dominance of Na left behind on the exchange sites of aquifer minerals, due to more recent sea-level decline, and has allowed the waters to evolve to Na-HCO 3 compositions (Figure 3) which occurred relatively recently. The differences between trace element concentrations in the granite and metasedimentary aquifers also needs to be studied in more detail. The complexity of the metasediments in Norway and their original sources may be largely responsible for the lack of similarity. In the case of the granites, it should be noted that these are of significantly different age and also tectonic history. The differences may therefore relate to differences in minor mineral phases such as fluorite and uranium-bearing minerals. In addition the glacial history may have had a significant effect in that the upper weathered parts of the Norwegian granites have been scoured during glaciation providing access to fresh mineral surfaces, whereas the Cornish 11

granites are overlain by relatively altered granite where leaching may have depleted the rocks in certain trace elements. 7. Summary and conclusions The groundwaters in both countries show large ranges in baseline concentration within and between different aquifer rock types. Although similarities exist in the hydrochemistry of groundwaters in these two regions, some chemical distinctions are apparent: groundwaters from Norway have relatively high ph compared with those in the UK (including carbonate aquifers). In addition, Na-HCO 3 type waters are much more prevalent in Norway compared to a dominance of Ca-HCO 3 type waters in the UK. This is most likely caused by cationexchange and freshening of the studied Norwegian aquifers as a consequence of more recent isostatic readjustment following the last ice age. Many trace elements show variations in concentration ranging over several orders of magnitude within aquifer units. Much of the variation in trace elements is related to the occurrence of accessory or secondary mineral phases and their associated trace constituents e.g. F, Rn, Be, U, and Ba. The concentrations of such elements are generally related to tectonic setting and provenance e.g. in the case of granites, those from convergent margin boundaries are geochemically distinct from within-plate orogenic granites. Therefore, each aquifer needs to be assessed individually and the application of the baseline concept to problems in groundwater quality such as pollution or remediation needs to take into account the natural variation in both lithology and groundwater chemistry. Indeed, the variation of some parameters in natural waters may be as great or greater than that produced by pollution and water quality standards may be breached by entirely natural processes. References Ahrens, L.H. 1954 The lognormal distribution of the elements (a fundamental law of geochemistry and its subsidiary). Geochimica et Cosmochimica Acta 5/2, 49-74. Edmunds, W.M., Brewerton, L.J., Shand, P. & Smedley, P.L. 1997 The natural (baseline) quality of groundwater in England and Wales. Part 1: A guide to the natural (baseline) quality study. British Geological Survey Technical Report, WD/97/51. Frengstad, B. & Banks, D. 2000 Evolution of high-ph Na-HCO 3 groundwaters in anorthosites: silicate weathering or cation exchange? In: Sililo et al. (Eds): Groundwater: Past achievements and future challenges. Proceedings of XXX IAH Congress, Cape Town. Balkema, Rotterdam, 493-498. Langmuir, D. 1997 Aqueous environmental chemistry. Prentice Hall, New Jersey. 600pp. Morland, G. 1997 Petrology, lithology, Bedrock structures, glaciation and sea level. Important factors for groundwater yield and composition of Norwegian bedrock boreholes? Norges Geologiske Undersøkelse Report 97.122 I, 274 pp. Neal, C., Robson, A.J., Shand, P., Edmunds, W.M., Dixon, A.J., Buckley, D.K., Hill, S., Harrow, M., Neal, M. & Reynolds, B. 1997 The occurrence of groundwater in the Lower Palaeozoic rocks of upland Central Wales. Hydrological and Earth System Sciences, 1, 3-18. Reimann, C. & Filzmoser, P. 2000 Normal and lognormal data distribution in geochemistry: Death of a myth. Consequences for the statistical treatment of geochemical and environmental data. Environmental Geology 39/9, 1001-1014. Robins, N.S., Shand, P. & Merrin, P.D. 2000 Shallow groundwater in Drift and Lower Palaeozoic bedrock: The Afon Teifi Valley in West Wales. Geological Society Special Publication, 182, 123-131. Shand, P., Darbyshire, D.P.F. & Gooddy, D.C. 1999 The application of strontium isotopes to catchment studies: The Plynlimon upland catchment of Central Wales. British Geological Survey, Technical Report WD/99/56. Shand, P. & Frengstad, B. Baseline groundwater quality: A comparison of selected British and Norwegian aquifers. British Geological Survey Internal Report, IR/01/177. Smedley, P.L., Bromley, A.V., Shepherd, D.J., Edmunds, W.M. & Kay, R.L.F. 1989 Geochemistry in relation to Hot Dry Rock development in Cornwall, Volume 4 Fluid circulation in the Carnmenellis granite: hydrogeological, hydrogeochemical and palaeofluid evidence. British Geological Survey Research Report SD/89/2. 12

Résumé Surveillance et contrôle des efflorescences à cyanobactéries : seuils de vigilance et d'alerte pour les eaux récréatives et de boisson. L. Brient et G. Bertru U.M.R. Ecobio Université de Rennes I, France En raison de ses spécificités hydrogéologiques et climatiques, la Bretagne assure ses besoins en eau à partir des ressources superficielles (80 %). La demande sans cesse croissante a nécessité la création de grands réservoirs depuis les années 1970. Ceux-ci sont soumis à des flux importants d'azote et de phosphore générés par l'agriculture intensive et les élevages hors-sol dans les bassins versants. Ces réservoirs sont de nouvelles niches écologiques très favorables au développement du phytoplancton et notamment des cyanobactéries. Leur prolifération entraîne des désagréments aux usagers (loisirs) et oblige les traiteurs d eau à utiliser des méthodes adaptées. La mise en place de schémas de vigilance des plans d eau s est trouvée renforcée en raison de la présence saisonnière de cyanobactéries hépatotoxiques. Les études concernent l'identification et le dénombrement des algues, la détection des toxines, notamment des microcystines. Les méthodes mises en oeuvre doivent nécessairement s'appuyer sur une stratégie d'échantillonnage prenant en compte les répartitions verticales et horizontales des microalgues ainsi que la périodicité des prélèvements. Ces données s avèrent nécessaires pour programmer les actions curatives dans les retenues (épandages de sulfate de cuivre) et adapter les traitements de l'eau brute (élimination des toxines). Cet article montre l'intérêt d'une telle surveillance et propose de l'étendre à tous les plans d'eau utilisés pour l'eau potable et/ou pour les loisirs. Abstract Because of its specific hydrogeological and climatic nature, Brittany uses surface water as its principal water resource (80 %). Increasing water demand has led to the construction of large reservoirs since the 1970 s; these are subjected to a significant influx of phosphorus and nitrogen generated by the intensive agriculture and livestock rearing in the catchments. The reservoirs are new ecological niches very favourable to the development of the phytoplankton and in particular to cyanobacteria. Their proliferation produces pollutants and requires the water supply authorities to use adapted treatment methods. Planned monitoring of the reservoirs was introduced due to the seasonal presence of the hepatotoxic cyanobacteria. These studies concern the identification and enumeration of algae and toxins, in particular the microcystins, and must be based on a strategy of a vertical and horizontal distribution of sampling at regular time periods. These data prove to be necessary to plan remedial actions in the reservoir (surface dosing with copper sulfate) and to adapt treatment processes of raw water (elimination of toxins). This paper shows the need for such monitoring approaches and proposes to extend its use to all reservoirs used for drinking water and/or leisure activities. 1 Introduction En Bretagne, les besoins en eau sont actuellement de plus de 200 millions de m 3 par an. Les eaux de surface représentent environ 80% des approvisionnements contre 37% en moyenne en France. La dégradation des eaux superficielles est continue depuis au moins trois décennies mais aucune politique efficace et volontariste n'a été à ce jour engagée. Depuis 1993, les principales actions reposent sur la maîtrise des rejets d'effluents des élevages et sur d'hypothétiques "bonnes pratiques agricoles". A ce jour aucun résultat probant n'a été enregistré alors que plus de 310 millions d'euros de fonds publics ont été dépensés (Anonyme 2002). Les apports excessifs de nutriments entraînent des proliférations de cyanobactéries potentiellement productrices de toxines (dermatotoxines, neurotoxines, hépatotoxines) dans les eaux douces, la multiplication des sites favorables à Alexandrium dans les estuaires et d'abondantes marées vertes sur le littoral. Les gestionnaires de la ressource en eau ont beaucoup tardé pour prendre en compte ce risque sanitaire. 1

2 Les risques liés aux cyanotoxines Parmi les hépatotoxines, la microcystine-lr est probablement l'une des plus dangereuses car elle s'est révélée être un promoteur de cancers primitifs du foie. Si l'on se réfère à l'étude épidémiologique réalisée dans la province chinoise du Jiangsu, sa toxicité chronique serait redoutable: les populations ayant consommé 0.19pg par jour pendant 4 mois présentent des fréquences élevées de cancer du foie (100.13 contre 4.28 pour 100 000 habitants)(lambert T.W. 1994; Ueno Y et al. 1996). Les microcystines renforceraient l'action d'autres carcinogènes comme l'aflatoxine B 1 et le virus de l'hépatite B. Il a fallu attendre les publications de l'oms (Chorus & Bartram 1999) pour que ce risque sanitaire soit pris en considération. Aussi en l'absence de normes européennes, le décret français du 20/12/2001 relatif aux eaux destinées à la consommation humaine prévoit à l'horizon 2003 la recherche de la microcystine-lr en cas de proliférations algales; il propose une valeur limite de 1 µg.l -1 pour la microcystine-lr, mais par contre, aucune exigence n'est formulée à l'égard des eaux récréatives. Dans ces conditions, faut-il attendre la réalisation d'un bloom à cyanobactéries, comme le stipule le décret du 20/12/2001, pour prendre les dispositions adéquates afin de respecter la valeur limite de 1 µg.l -1. Cette concentration est manifestement trop élevée au regard de l'étude épidémiologique précitée; elle devra être revue à la baisse et fixée en deçà de 0.1 µg.l -1. Le choix actuel du législateur a probablement été fixé pour tenir compte des filières de traitement qui ne sont pas équipées par d'unités de charbon actif ou d'autre dispositif technique permettant d'éliminer ou de réduire les toxines. On sait que les traitements conventionnels comme la coagulation, la floculation, la clarification et la filtration présentent souvent le défaut de solubiliser les toxines. Parmi les techniques reconnues pour leur efficacité, il faut surtout citer les techniques d'ozonation et l'utilisation des charbons actifs en grain ou en poudre. Toutefois, (Hitzfed et al. 1999) indique qu'elles peuvent être prises en défaut lorsque la charge en composés organiques est suffisamment forte pour réduire les performances de l'oxydant et pour diminuer les capacités d'adsorption des charbons. Ce même auteur note que les produits formés, non identifiés, présentent encore une activité inhibitrice des phosphatases, ce qui indiquerait la présence de microcystine. Faute d'avoir privilégié les moyens préventifs nécessaires pour améliorer la qualité de l'eau, les autorités locales mettent en place de technologies coûteuses surtout payées par le consommateur, pour respecter les normes sanitaires. Les techniques gagneraient en efficacité si les utilisateurs disposaient d'informations précises sur les cibles biologiques et leur dynamique. Il devient urgent pour les autorités sanitaires, d'établir un réseau de surveillance permettant de prévoir et d'évaluer les risques sanitaires à la fois pour les eaux d'alimentation et les eaux récréatives. 3 Situation en Bretagne En Bretagne toutes les retenues eutrophisées connaissent de fortes proliférations algales. Les populations évoluent au cours des saisons et dans de nombreux réservoirs les blooms à cyanobactéries apparaissent surtout en été et au début de l'automne. Ainsi de juin à octobre, la quasi-totalité des réservoirs utilisés pour l'eau de consommation, connaît au moins une période de 10-20 jours de proliférations de cyanobactéries. Pour certaines retenues, ces évènements peuvent être répétés 2 à 3 fois en fonction des conditions climatiques et des écostratégies des espèces. Cette situation peut même devenir permanente et à des seuils élevés pour les réservoirs colonisés par des espèces du type Planktothrix agardhii. Certaines collectivités épandent du sulfate de cuivre dans les réservoirs pour inhiber le développement des cyanobactéries (figure 1). L'étude effectuée en septembre 2001 montre que la microcystine-lr a été détectée dans 38 prises d'eau sur 52 étudiées (figure 2). 4 Discussion L'Administration (DDASS) et les collectivités qui ont en charge le contrôle sanitaire des eaux, la qualité des eaux brutes destinées à la production d'eau potable et la qualité des eaux de baignade doivent aujourd'hui mettre en place un réseau de surveillance. Les mesures à prendre pour éviter tout accident pourront être préventives et/ou curatives. Elles pourront être déterminées à partir des résultats des analyses de la microflore algale. 4.1 Propositions pour la mise en place d'un système de surveillance Le système de surveillance doit nécessairement reposer sur l'identification et le dénombrement des espèces potentiellement productrices de toxines (Chorus & Bartram 1999). Aujourd'hui, il peut être complété par un test de type ELISA qui permet d'évaluer globalement les microcystines. Celui proposé par Dietrich (2001) assure une détection de 0.03-0.07 µg.l -1 des différents variants des microcystines. L'identification doit être associée au dénombrements des espèces afin d'évaluer d'une part le niveau de contamination et prévoir leurs potentialités de croissance. Eu égard au nombre d'espèces pouvant être mises en 2

Figure 1 Evolution des populations algales dans la retenue de la Meaugon sur le Gouet en relation avec l'application de sulfate de cuivre. Année 2000. Figure 2 Situation des 52 prises d'eau étudiées et détection de la microcystine-lr (38 points noirs). Septembre 2001. 3

cause (10-14), on peut considérer que l'effort de formation en faveur des personnels techniques concernés n'est pas démesuré. Les techniques de comptage sont également bien documentées (Chorus & Bartram 1999) et peuvent être très facilement mises en œuvre. Cette surveillance doit débuter selon une fréquence hebdomadaire, au plus tard à la mi-mai pour se terminer en septembre ou début octobre en fonction des conditions climatiques. Si les prélèvements doivent être prioritairement effectués à proximité de la prise d'eau, il ne faut pas pour autant négliger les zones en amont de celle-ci qui peuvent constituer des zones favorables au développement des cyanobactéries. La configuration de la cuvette et son orientation par rapport au vent dominant doivent être prises en considération pour l'établissement du protocole d'échantillonnage. A partir d'une mesure simple, à l'aide du disque de Secchi (Z s ), il est possible d'estimer l'épaisseur de la couche euphotique, qui est comprise entre 2 à 3 fois la valeur de Z s puis d'effectuer sur l'ensemble de celle-ci les prélèvements dans la totalité du volume d'eau occupé par le phytoplancton. Il est souvent judicieux de compléter les résultats précédents par une recherche qualitative sur des échantillons prélevés au filet à plancton de maille 20-40 µm. Ces données peuvent s'avérer très utiles pour surveiller certaines zones des réservoirs. En combinant, l'ensemble de ces résultats et la recherche des microcystines, il devient possible de proposer un protocole de suivi aux gestionnaires des plans d'eau. Pour illustrer les différents seuils de vigilance et d'alerte, nous avons fait référence à Planktothrix agardhii et Microcystis aeruginosa qui sont très souvent les espèces dominantes. La première est une forme coloniale et la seconde un filament cylindrique: 100 colonies de M. aeruginosa constituées de 100 cellules sont sensiblement équivalentes à 100 filaments et représentent entre 0,5 et 1.2 µg.l -1 de microcystine ou encore 2.5-3.5 µg.l -1 de chlorophylle a. Figure 3 Protocole de suivi des cyanobactéries et détection de la microcystine-lr dans les milieux aquatiques 4.2 Seuil de vigilance En deçà de 30-50 colonies ou filaments par ml, les actions curatives par application de sulfate de cuivre sont possibles. L'expérience (Brient L. et al. 2001) montre qu'en agissant rapidement sur les stades initiaux de croissance des cyanobactéries, le sulfate de cuivre inhibe leur développement tout en permettant encore celui des autres composantes de la communauté phytoplanctonique (figure 2). Bien évidemment, l'usage du sulfate de cuivre doit être réservé exclusivement à ce niveau de vigilance et en aucun cas il ne peut être utilisé lorsque les proliférations deviennent abondantes car le sulfate de cuivre provoquerait la solubilisation des toxines. Ces traitements devraient faire l'objet d'une plus grande surveillance de la part des autorités sanitaires. En effet, lorsqu'ils sont pratiqués dans les zones très en amont de la prise d'eau, les toxines solubilisées se retrouvent peu de temps après au niveau de la prise d'eau, compte tenu de leur dégradation relativement lente, 15 à 20 jours 4

(Harada et al. 1996). Au cours de cette période de vigilance, la fréquence des prélèvements doit être hebdomadaire. Les activités récréatives peuvent être poursuivies. 4.3 Seuils d'alerte En l'absence de traitements curatifs, les proliférations cyanobactériennes dépendent essentiellement des conditions climatiques.a partir de 200 colonies ou filaments par ml, la situation est critique pour les baignades, notamment pour les enfants qui risquent de " consommer" leur dose journalière admissible par une ingestion de 50 à 100 ml d'eau.lorsque les cyanobactéries sont très abondantes, le traitement par le cuivre devient également problématique du fait de la solubilisation des toxines et surtout de la production de matières organiques qui nuisent à l'efficacité de l'ozonation et des charbons actifs. Dans tous les cas, la filière de traitement doit être optimisée afin de réduire sinon éliminer les microcystines à un seuil inférieur à 0.1 µg.l -1. Lorsque la situation continue à se dégrader et que l'eau contient 2000 colonies ou filaments par ml, il serait souhaitable de rechercher une autre source d'eau brute pendant la durée de la prolifération. Bien évidemment, toutes les activités nautiques devraient être interdites. 5 Conclusion En publiant le décret du 20/12/2001, le législateur reconnaît la dangerosité des cyanotoxines. On peut regretter qu'il n'a retenu que les critères proposés par l'oms, à savoir la seule microcystine-lr qui est l'une des 60-65 molécules constituant l'ensemble des microcystines. Sans doute, faudra-t-il très rapidement abaisser la valeur limite à 0.1 µg.l -1 pour la microcystine-lr mais également pour l'ensemble des microcystines. Actuellement, l'application du décret suppose un effort de formation pour les personnels chargés de la gestion des réservoirs. La mise en place d'un réseau de surveillance est nécessaire et celui-ci devrait être opérationnel avant le printemps 2004. Références bibliographiques Anonyme (2002) La préservation de la ressource en eau face aux pollutions d'origine agricole: le cas de la Bretagne. Rapport février 2002 - Cour des Comptes Brient L., Le Rouzic B., Vézie C. & G. B. (2001) Conditions d'utilisation du sulfate de cuivre pour limiter les proliférations des cyanobactéries et réduire ses effets sur l'environnement/. aghtm, 718-733 Chorus I. & Bartram J. (1999) Toxic Cyanobacteria in Water: A guide to their public health consequences, monitoring and management. Geneva: World Health Organization, 416 Dietrich D.R. (2001) Détecter les cyanotoxines des eaux. Biofutur, 209, 44-47 Harada K.-I., Tsuji K. & Wanatabe M.F. (1996) Stability of microcystins from cyanobacteria. III. Effect of ph and temperature. Phycologia, 35 (6 Supplement), 83-88 Hitzfed B.C., Höger S.J. & Dietrich D.R. (1999) Cyanobacterial Toxins: Removal during Drinking Water Treatment, and Human Risk Assessment. Environmental Health Perspectives, 108, 113-122 Lambert T.W. H.C.F.B., HRUDEY S.E. (1994) Microcystin class of toxins: health effects and safety of drinking water supplies. Environ. Rev., 2, 167-186 Ueno Y, Nagata S., Tsutsumi T., Hassegawa A., Wanatabe M.F., Park H-D, Chen G-C., G. C. & S-Z Y. (1996) Detection of microcystins, a blue -green algal heptatotoxin, in drinking water sampled in Haimen and Fusui, endemic areas of primary liver cancer in China by highly sensitive immunoassay. Carcinogenesis, 17, 1317-1321. Vezie C., Bertru G., Brient L. & Lefeuvre J.C. (1997) Blooms de Cyanobactéries hépatotoxiques dans l'ouest de la France. TSM, 10, 39-46 5

Sources of Iron and Manganese in public water supplies:a case study of the Megget Reservoir, Scotland CORINNA ABESSER, RUTH ROBINSON Department of Geography and Geosciences, University of St. Andrews, St Andrews, KY16 9AL, Scotland, UK, E-mail: ca22@st-andrews.ac.uk CHRIS SOULSBY Department of Geography and Environment, University of Aberdeen Aberdeen AB24 3UF, Scotland, UK Abstract High iron and manganese concentrations are a common feature in many upland water supplies. Peaty soils covering such catchments favour metal mobilisation processes resulting in high iron and manganese loading in the streams draining into lakes and reservoirs. When removed from the water column by sedimentation, the metals accumulate in the bottom sediments from where they can be remobilised when conditions in the water body change. Water quality problems experienced in the Megget reservoir during a period of major draw down have led to extensive investigations at the reservoir body and its surrounding catchment area. This paper presents integrated findings from regular water quality sampling, catchment studies and process-based laboratory experiments. While catchment characteristics control iron and manganese fluxes into the reservoir, especially during high rainfall episodes, in-reservoir processes appear to have significantly contributed significantly to the experienced water quality problems. This has important management implications for the water supply itself and also highlights the importance of catchment-based management strategies for water supplies in Celtic Uplands. Résumé De hautes concentrations en fer et en manganèse caractérisent fréquemment les réserves d eau situées en terrain élevé. Les sols tourbeux recouvrant de tels bassins hydrographiques favorisent les processus de mobilisation des métaux, résultant en de hauts contenus de fers et de manganèses dans les cours d eau qui alimentent les lacs et les réservoirs. Lorsqu ils sont retirés de l eau par sédimentation, les métaux s accumulent dans les sédiments inférieurs, d où ils peuvent être mobilisés à nouveau quand les conditions dans la masse d eau changent. Les problèmes de qualité d eau apparus au réservoir Megget durant une période où le niveau d eau était particulièrement bas ont conduit à de nombreuses investigations au réservoir et dans ses régions avoisinantes. Cet article présente les résultats d échantillonnages réguliers de la qualité de l eau, d études du bassin hydrographique et d études expérimentales conduites en laboratoire. Tandis que les caractéristiques du bassin contrôlent les flux en fer et en manganèse dans le réservoir, surtout durant les périodes de pluies 1

intensives, les processus se déroulant à l intérieur du réservoir semblent avoir contribués aux problèmes de qualité d eau de façon significative. Cela a des implications importantes pour la gestion de l approvisionnement en eau, et souligne aussi l importance de stratégies de gestion à l échelle du bassin pour l approvisionnement en eau dans les hautes terres celtiques. Introduction The Celtic uplands in the north and western parts of the British Isles are extremely important areas for the provision of public water supplies. The wet, cool climate, glaciated valleys and impermeable geology have resulted in preferential reservoir location in these areas. However, the peaty nature of the soils covering such catchments often results in high levels of colour, together with elevated Fe and Mn concentrations, in drainage waters. Undesirable concentrations of these elements have been reported in various upland water supplies throughout the UK (e.g. Heal et al., in press, Gavin et al., 2001) and can create serious problems for water resource managers in that the EC Drinking Water Directive has quite stringent permissible limits on these determinants in public water supplies. Thus treatment costs can be high, particularly if very high levels are unpredictable, as is the case during runoff events after prolonged drought. This paper presents the preliminary results from a study, which was initiated following water quality problems experienced at the Megget Reservoir in the Southern Uplands of Scotland in 1997-1998. Following a period of extensive reservoir draw down, the water quality in the reservoir deteriorated dramatically, in particular iron and manganese levels increased in conjunction with a discoloration of the water. As a consequence, a comprehensive monitoring programme was established which aimed to (a) identify sources of iron and manganese mobilisation in the catchment area and assess how they influence external inputs to the reservoir, and (b) study the reservoir system in order to identify the controls on internal iron and manganese cycling between the water and the sediments. By assessing these factors insight was sought into processes which may have contributed to the problems experienced in 1997-1998. A better understanding of these processes is an important prerequisite for developing a general strategy for managing the Megget reservoir and similar water bodies in upland catchments. Study Area: geological and hydrological characteristics The Megget reservoir is situated in the Southern Uplands between Peebles and Moffat, about 50 km south of Edinburgh. It has a surface area of 2.59 km 2 with a maximum depth of 55 m and drains a catchment area of 40 km 2 with an altitudinal range of 330-830 m and a mean annual precipitation of around 1767 mm (Lothian Water Supply Services, 1989. 2

The bedrock consists predominantly of Silurian greywackes and shales that are enriched in iron and manganese (British Geological Survey, 1993; Peach & Horne, 1899). The bedrock is widely covered by superficial or drift deposits, such as moraines and boulder clay (till) which dominate most of the lower grounds of the catchment. Morainic sands and gravels are mantled by imperfectly drained brown forest soils. Poorly drained peaty gley soils have developed on the fine textured till, mainly found in the area south of the reservoir. The higher grounds are often covered by postglacial peat deposits, which occupy most of the southern part of the catchment. The area north of the reservoir is predominated by peaty podzols, as is illustrated in the simplified soil map in Figure 1. Land use in the catchment area is generally confined to hillfarms and moorland. Sheep farming is the dominant agricultural activity in the area and is concentrated along the hillslopes of the reservoir. Methodology and Analysis Regular water quality sampling has been carried out in the reservoir and on the feeder streams between 1999-2001. Water samples from the reservoir were collected monthly at various locations and depths throughout the reservoir. Stream water samples were collected on a bi-monthly basis and this has been supplemented by storm event sampling on three streams: Linghope Burn, Shielhope Burn and Winterhope Burn. The catchments are characterised by differences in size and topography as well as in soil type distribution as is summarised in Table 1 and Figure 1, respectively. Automatic water samplers were available for two of the catchments and were installed at Shielhope Burn and Winterhope Burn allowing 24-h sampling. At Linghope Burn samples were collected manually, up to 3 times a day, depending on the hydrological conditions. All water samples were returned to the ESW laboratory in Edinburgh for analysis of the following parameters: ph, conductivity, alkalinity, colour, Ca, Mg, Fe, Mn, Al, TOC, SO 4 and SiO 2. Laboratory-based process studies were carried out on undisturbed sediment cores to investigate the relative importance of reservoir internal processes, namely sediment re-suspension and diffusion processes, on the water quality of the Megget Reservoir. Secondary cores were collected for the analysis of metal distribution in the sediment prior to the experiments. slope angle( ) catchment catchment area (km 2 ) catchment length (m) catchment width (m) length /width ratio max mean stdev. Winterhope 10.2 5197 2545 2.04 43.8 15.6 8.7 Linghope 4.4 2827 1362 2.08 40.5 15.9 7.02 Shielhope 4.1 2890 1424 2.03 27.7 9.5 5.08 Table 1 Catchment characteristics (derived from GIS analysis) 3

Results and Discussion Temporal and spatial variability in Reservoir water chemistry Regular sampling at the reservoir and feeder streams has shown distinct seasonal variations in internal and external iron and manganese loading into the reservoir as illustrated in Figure 2. The most significant changes in the reservoir occur during the summer period when mixing of the reservoir is limited due to temperature stratification in the water body. At these times, a distinct concentration gradient could be observed between the wellaerated epilimnion and the isolated hypolimnion, with highest concentrations (Fe up to 207 μg/l, Mn up to 126 μg/l) being found near the sediment-water interface at the bottom of the reservoir. Several studies have suggested that elevated water column concentrations of metals most likely result from sediment-water exchange processes (Warnken et al., 2001, Davison, 1993). In the case of the Megget reservoir, this hypothesis is supported by the increase in the total dissolved (<0.45μm) metal distribution towards the sediment. However, concentrations of iron in the water column observed during the sampling programme were not sufficiently high enough to explain the water quality deterioration experienced in 1997-1998. This suggests that in particular the high iron concentrations (up to 467 μg/l) in the water column during the draw down period are not solely attributed to sediment diffusion and exchange processes and that other sources and release mechanisms are important. Stream Chemistry and catchment response to rain storm events Stream water chemistry shows substantial spatial and temporal variations, which in general appear to reflect differences in soil type distribution, drift and bedrock geology within the catchment area as well as differences in hydrological conditions, respectively. Spatial and temporal pattern of iron and manganese concentrations for Winterhope Burn, Shielhope Burn and Linghope Burn are shown in Figure 3 and clearly illustrate that higher fluxes of iron and manganese occur in the streams draining the southern part of the catchment relative to those draining the north. Intensive sampling on those streams further revealed the differences in the response of the individual catchments to changes in flow conditions, which were evident for a number of solutes. Matrixplots were used to investigate these differences in runoff behaviour and to identify pattern in the runoff chemistry of individual catchments. Inter-correlations between some of the determinants are displayed in Figure 4. The plots clearly indicate two-component mixing of base-rich groundwater with more acidic water derived from soil sources during rainfall events. The linear relationship between Alkalinity and Ca suggests that the baseflow component in all three catchments is derived from a similar source. The influence of weathering processes on stream chemistry appears to be most important in the Linghope catchment as 4

indicated by high alkalinity and the relative enrichment in solutes indicative of bedrock weathering (Ca, SiO 2 ). As the weathering component decreases high concentrations of TOC and metals dominate the runoff chemistry, particularly in Shielhope Burn and Winterhope Burn, indicating that soil water runoff is an important contributor during high flow events in these catchments. The most distinct changes occur in Shielhope Burn, which also exhibits the highest concentrations/widest concentration range in TOC and Fe. As shown in Figure 1 this catchment is predominated by organic rich and poorly drained soils (peaty gleys), which favour reduction processes and mobilisation of iron and manganese, resulting in a soil water component strongly enriched with these elements. The good correlation between metals and TOC suggests that much of the iron and manganese in the stream water is organically complexed. The negative linear relationship between TOC and alkalinity suggests conservative mixing between the components. However, the mixing lines for the different streams have different gradients. Considering the topography of the individual catchments (see table 1) the following interpretations are suggested: (a) the steeper gradient seen for Winterhope catchment indicates a more flashy runoff regime, which results from the higher percentage of steep slopes in the catchment area. (b) Slopes in the Shielhope catchment are generally much shallower and thus, the response during events is slower (less flashy). The slope distribution may also explain the higher contribution of groundwater inputs into the stream indicated by the higher alkalinity values. Effect of sediment re-suspension and diffusion processes on reservoir water quality The diffusion experiment showed that the sediment could release significant amounts of iron (2.0 mg/l at 9 C) and manganese (3.9 mg/l at 9 C) into the overlying core water column even when the water remains oxygenated. Dissolved iron was found to oxidise rapidly, forming a red precipitate on the inside of the core tube. In contrast, manganese remained in the water column for considerable time and removal from the water column was found to be controlled by microbial activity. Re-suspension of the reservoir bottom sediment introduced significant amounts of iron and manganese into the core water column whereby the release of iron (up to 80 mg/l = measured immediately after re-suspension interval) is elevated relative to manganese (up to 20 mg/l). This suggests that re-suspension has more affect on the reintroduction of iron into the water column whereas sediment diffusion processes appear to be more important for manganese release in the oligotrophic Megget system. The presence of fine humic material in the water column is indicated by the strong discoloration of the water, which developed during the initial resuspension interval and remained present in the water column during the entire experimental period. This not only produces serious discolouration of the water but also extends the residence time of iron, complexed with the humic material, in the water column due to the low settling velocity of humic particles. 5

Conclusion It is apparent that the generic features of catchments (geology, soil types and hydrologic regime) are important controls on the iron and manganese fluxes into the Megget reservoir. Detailed sampling on three feeder streams has shown that catchment response to rainfall periods can vary greatly and is also influenced by topographical characteristics. However, further analysis of the collected data is necessary before more detailed conclusions about the functioning of the Megget catchment can be drawn. The results from the laboratory study have indicated that re-suspension, promoted by the draw down of the water level, may have significantly contributed to the water quality problems experienced in 1997/98. These finding have important management implications for the water supply itself and also indicate the importance of catchment-based management strategies for reservoir water quality in the Celtic uplands. The research summarised here has been able to identify the factors that may have been responsible for the problems encountered at the Megget Reservoir during the draw down event in 1997/98. The next stage of the research is the more detailed analysis of the data, using statistical techniques as well as techniques that allow prediction of episodic variability in stream chemistry and contribution from different sources in order to estimate fluxes delivered to the reservoir during different flow regimes. These results together with the results from regular sampling programme and laboratory-based process studies will provide a good basis for outlining a general strategy for managing the Megget reservoir and similar water bodies in upland catchments. References British Geological Survey (1993) Regional geochemistry of Southern Scotland and part of Northern England. British Geological Survey, Keyworth. Davison, W. (1993) Iron and manganese in lakes. Earth Science Reviews, 34, 119-163. Gavin, K.G., Farmer, J.G., Graham, A.C., Kirika, A. & Britton, A. (2001) Manganese-humic interactions in the catchment, water, sediment of Loch Braden, S.W. Scotland. In: Swift, R.S. & Spark, K.M. (eds) Understanding and managing organic matter in soils sediments and waters, IHSS. Heal, K., Kneale, P.E. & McDonald, A.T. (in press) Manganese in runoff from upland catchments: temporal patterns and controls on mobilisation, Hydrological Sciences Journal. Lothian Water Supply Services (1989) Megget Scheme. Lothian Regional Council, Water Supply Services. Edinburgh. Peach, B.N. & Horne, J. (1899) The Silurian rocks of Britain, Vol.1: Scotland. Memoirs of the Geological Survey,U.K. 6

Warnken, K.W., Gill, G.A, Griffin, L.L & Santschi, P.H. (2001) Sediment-water exchange of Mn, Fe, Ni and Zn in Galveston Bay, Texas. Marine Chemistry, 73, 215-231. 7

300 250 0-10 m 11-28 m 29-50 m Iron concentration (µg/l) 200 150 100 50 0 Aug-99 Oct-99 Dec-99 Feb-00 Apr-00 Jun-00 Aug-00 Oct-00 Dec-00 Feb-01 Apr-01 Jun-01 Aug-01 Oct-01 Dec-01 Figure 2. Spatial and temporal distribution of iron in the Megget Reservoir 8

Iron concentration (µg/l) 900 800 700 600 500 400 300 200 100 Linghope Burn Shielhope Burn Winterhope Burn 0 Aug-99 Oct-99 Dec-99 Feb-00 Apr-00 Jun-00 Aug-00 Oct-00 Dec-00 Feb-01 Apr-01 Jun-01 Aug-01 Oct-01 Dec-01 daily rainfall 75 50 25 0-50 -100-150 Rainfall (mm /day) -200 Figure 3. Spatial and temporal distribution of iron in streams 9

TOC Winterhope Burn Linghope Burn Shielhope Burn Fe Al Alkalinity Ca SiO2 Figure 4: Matrixplots for three streams in the Megget catchment 10

Buffering capacity of landscape structures in rural catchments : definition and assessment Valérie Viaud (1,2), Philippe Merot (1) 1 UMR Sol, Agronomie et Spatialisation, INRA-ENSAR 2 INRA, Unité SAD Armorique Rennes 65 rue de Saint Brieuc CS 84215 35042 Rennes cedex France Fax: 02-23-48-54-30; Tel: 02-23-48-52-25; e-mail: viaud@roazhon.inra.fr Abstract Surface water pollution is a prominent environmental issue in rural catchments, for human and animal consumption as well as for stream ecological functions. Intensive farming is considered as the greatest cause of pollution since it has gone both with an increasing use of fertilizers and a gradual disappearance of wetlands, riparian forests and hedgerows, located between cultivated fields and stream water. Thus strategies designed to improve water quality take places at two levels: on the one hand, change in farm practices in cultivated fields like low input farming, organic farming or winter plant cover are promoted to prevent pollutant generation at its source and to reduce pollutant excess; on the other hand, management or re-establishment of buffer landscape structures is proposed in the margin of production areas. It is generally accepted that buffer zones contribute to pollution decrease and their conservation is promoted by many environmental or landscape policies. However buffer zone remains a fuzzy concept: it is widely used although few quantified. In this context we aimed at investigating and discussing accurately buffer zone concept and more especially its characterization by buffering capacity i.e. the assessment of buffer landscape structure functions toward water quality. We have first considered buffering capacity formal definition; then reviewing literature about buffer zones for pollution control, we have emphasized limits of buffering capacity classical definition in the environment field and we have attempted to clarify this concept with an alternative approach. We defined a buffer zone by the target it protects and by a function, i.e. the alteration of a signal between inflow and outflow through the buffer zone. We proposed to classify the buffering capacity according to the type of signal alteration. In a second part of this work, we applied this classification to four actual rural catchments differing by their buffer zone areas. Résumé La pollution des eaux par les nitrates, pesticides, phosphore et sédiments est un problème majeur dans les bassins versants agricoles, pour la consommation humaine et animale comme pour la fonction d habitat des cours d eau. La pollution est principalement attribuée à l intensification de l agriculture, qui s est traduite à la fois par l augmentation de l utilisation d intrants et par la disparition des ripisylves, haies et zones humides entre les parcelles et le cours d eau. Deux types de solutions sont donc mis en œuvre pour améliorer la qualité de l eau : d une part une diminution de l utilisation d intrants sur les parcelles cultivées, d autre part la conservation ou l aménagement de zones tampon en marge des surfaces de production. Il est admis que les zones tampon participent au contrôle de la qualité de l eau et leur aménagement est soutenu par un grand nombre de politiques environnementales ou de gestion des paysages. Néanmoins le concept de zone tampon reste vague : il est largement employé mais l effet des zones tampon est peu quantifié. Dans ce cadre, notre objectif est de discuter et d approfondir les notions de zone tampon et de capacité tampon par rapport à la qualité de l eau. Dans un premier temps, nous considérons la définition classique de la capacité tampon. Puis, à partir d une étude

bibliographique, nous mettons en évidence les limites de cette définition dans le domaine de l environnement. Nous proposons alors de clarifier ce concept par une approche alternative. Nous définissons une zone tampon par la cible qu elle protège et par sa fonction, c est-à-dire par la modification d un signal entre l entrée et la sortie de la zone tampon. Nous proposons une classification des zones tampon en fonction du type de modification du signal. Dans un deuxième temps, nous appliquons cette classification à quatre bassins versants ruraux.

Blue City Project Water Quality Measurement on the Lee Ryan, M. & O Kane, J.P. Department of Civil & Environmental Engineering, National University of Ireland, Cork, Ireland Abstract The overall objective of the Blue City project is to make a complete model of the hydraulic infrastructure of a city situated in its catchment. Cork city and the Lee catchment is our laboratory. The model will include the physics, chemistry and some biology of all the man-made and natural surface and ground-waters on which the city depends for its functioning, including, for example, the supply of water for domestic and industrial use, the collection of waste water for treatment and disposal, the amelioration of flooding and possible rise in sea-level, the generation of hydroelectricity, the enhancement of fisheries, the provision of water based sport and amenities, the development of port and harbour activities, and so on. The model will be made available in various forms, for example on the web, in immersed virtual reality facilities, etc. Such a model demands quality data. The goal is to discover and answer engineering and scientific questions whenever model and data fail to agree. We already have a DEM on a 1 m grid covering 300 square kilometers centred on Cork city and geo-referenced to the national grid with a relative accuracy of 20 cm in all three directions. We have matching multi spectral data with a pixel size of 12 cm. In this paper we describe the current monitoring methods employed by various bodies throughout the Lee catchment and their inability to capture the algae blooms occurring annually in the upper and mid-estuary eutrophication is one of the most important water quality issues. We have installed a state of the art, high frequency, GSM telemetered solution to water quality measurement. This paper describes the equipment chosen for the task, the problems encountered in the set-up and general operation of the station, and the output data. We also present some of the data in graphical format, which clearly shows the need for high frequency water quality measurement. The effect of external influences, such as dam releases and rainfall events, on the system chemistry is also presented and discussed. We examine the future development of water quality measurement, how it can be improved, and our commitment to further technological advances.

An Alternative Approach to the Management of Water Resources by Anthony J Tollow Department of Civil Engineering Technikon Natal Abstract To manage water resources efficiently several different aspects of operation require careful consideration. The initial phase is managing the resource. As demand increases the resource is developed and expanded until fully exploited. When the demand exceeds the available water then either new resources have to be developed or demand restricted. Originally restrictions only occurred during times of water shortages. However, with the increasing cost of developing new resources, management of consumption is, of necessity, being given greater consideration. New approaches are being developed. One such is an approach which combines both resource and consumption management. In this strategy there are two different philosophies either when operating under normal conditions or in times of water scarcity. Nevertheless both may be combined during the operating period. An outline of both the individual components and a suggested approach to their combined use will be developed. Conclusions are drawn. Résumé La gestion efficace des resources en eau amène à prendre en compte plusieurs composantes opérationnelles. La demande allant croissant, la gestion de la ressource passe d abord par une augmentation de l offre disponible, jusqu à atteindre son maximum. Lorsque la demande dépasse cette offre, de nouvelles ressources sont exploitées ou des restrictions sont imposées à la demande. Habituellement, ces restrictions ne concernent que les périodes de déficit en eau. Mais le coût de développement de ces nouvelles ressources allant en augmentant, la gestion de la consommation apparaît de plus en plus nécessaire. De nouvelles approches sont mises en oeuvre à cet effet, dont certaines combinent la gestion de l offre et de la consommation. Les stratégies choisies varient en fonction de la rareté de la ressource dans le temps, tout en étant éventuellement combinées. Ces article présente ces différentes stratégies et propose une approche permettant leur utilisation conjointe, avant de tirer quelques conclusions. INTRODUCTION To manage water resources efficiently a number of very different aspects require careful consideration. The strategy has evolved over many years. Initially the emphasis was placed on resource management with the objective of meeting all demands. Once one resource had been fully developed the next was ready to be brought into use. However, even in the Celtic World with its abundant clean water, consideration needs to be given to other aspects of water management, not necessarily because the resources are becoming exhausted, but on other grounds, such as the costs of transporting water and the costs of the subsequent cleaning up of the water once it has become polluted. Once the initial phase of resource management has been put into practice then as demand exceeds the available water, either new resources have to be developed or somehow the requirement for water may need to be reduced. Originally the only time demand modification was practised was during time of severe water shortage. This had usually been

caused by abnormally long periods of dry weather that had given rise to a severe diminution of rainfall and runoff. The results were low groundwater levels and low reservoir and river levels. However, with the increasing cost of developing new resources, and the greater awareness of the environmental demands for water, the management of consumption is, of necessity, being given greater consideration. New approaches are being developed. One such is an approach which combines both resource and consumption management. In this strategy there are two different philosophies, either when operating under normal conditions, or in times of water scarcity. Nevertheless both may be combined. An outline of both the individual components and a suggested approach to their combined use will be developed. Some of the components have been formulated and tested using generated data sets. In addition a means of comparing the results has been suggested. RESOURCE MANAGEMENT The management of water resources has evolved over many years. Many different approaches have been used. However, most are focussed on the problem of what to do in times of scarcity. Very few are concerned with sophisticated operational policies during periods of normality. However, with increasing energy and other costs, efficient management, even in times of plenty, is becoming essential. Strategies will also need to be developed for periods of high flow, when the water quality deteriorates due to heavy sediment loads in the rivers and possible pollution. Underground resources too, especially those in karst topography may also be affected. In the case of reservoir management during a flood it may only be necessary to divert the flow around the storage, or if that is not possible to consider off stream bankside storage. The latter may also serve to even out the inflow from a river, permitting the operation of pumps at constant speed and constant flow rate. Examples may be found throughout the Celtic World but one that comes to my mind is the bankside storage in connection with the transfer from the Upper Stour to the Pant, (the name for the Upper Blackwater) in Essex. This forms an essential part of the Ely Ouse - Essex transfer. Other forms of management may be less conspicuous but no less important. Control bands The concept of system operation using different mathematical models of varying complexity has been applied in many formats over a number of years. However, since the development of ever more powerful computing facilities the scope of mathematical models has been extended. Nevertheless this may lead to over complex solutions. One development that offers may advantages is the application of control bands. The ideal is a family of control bands, not only to cope with operation under extreme conditions but also to plan operations under more normal variations of climate - the average year - or perhaps the more clearly definable median year. Although they have been developed for single and multiple reservoir operation the concept is also applicable to groundwater and conjunctive use. The concept of the control band may be compared in some ways to the fuzzy logic approach to water pricing (Hall 1991). It may be used where there are either ranges of optimal solutions or there is no one acceptable optimal solution. Both these ranges of scenarios may occur in resource management problems. In itself the control band at first sight allows for sub-optimal operations. However, the flexibility that the band allows may in fact help to achieve a desired optimal solution. The key lies in the band width. This is set up

to suit the resource and the circumstances. A wide band width allows for a large range of solutions. Conversely a small band width may in the extreme result in only one solution. When the actual operation moves out of any particular band then a series of constraints may be activated depending on:- i) whether the point lies above or below the band ii) the characteristics of the various resources forming the system iii) forecasts of demands and future availability iv) the overall trends of the past and current operations v) the desired risk of failure, either to achieve complete optimality or impose restrictions and be unable to meet demand. One approach in designing the relevant sets of control bands is to use simulated sets of data (Tollow 1989a). These may be made up as five year sets of median, high and low flows so that the appropriate control bands may be developed. Then the bands are tested on 100 year sets of data, either generated (Tollow 1989b) or original. Multiple sets of generated dat often help identify weaknesses in the original strategy (Tollow 1991a) which then allows consideration of modifying the control band and operating policy or modifying the infrastructure to make it more efficient. Control bands are usually considered as an operational tool but they may also become a management planning tool too. Control bands have been developed (Tollow 1988) to control the abstraction of water from multiple rivers to fill a large reservoir by taking advantage of off peak electricity and suitable flows. Part of the operating strategy would also indicate when poorer quality water could be allowed to flow to the sea and when it was important to maximize the intake to refill the reservoir after a major drawdown. It is relatively straightforward to allow for other constraints as well as the use of off peak electricity. Control bands are especially useful in controlling multiple -use reservoirs that may form part of a larger number of reservoirs. Each reservoir may have its own sets of control bands that operate in conjunction with those of other reservoirs or even groundwater. With more sophistication and cheaper and more efficient computer control it is possible to build control bands in to operating systems. This is the first step to real time control. It also accepts that there may be more than one possible option. By using appropriate multi -year data sets it is possible for the system to check, either that the best solution is reached or that a range of options is possible. This may be compared to the scenario approach to allow more and varied options. Decisions may be controlled in a number of ways. Linear Programming One of the methods of validating the control bands uses Linear Programming (Tollow 1991b). However, apart from the ability to vary the constraints, a very useful feature, there difficulties were encountered because of the number of possible options. This was due in part to the complexity of even a relatively straightforward system and in part to the need to handle relatively large data sets. To be effective at least ten sets of 100 years of data were found to be needed. The problems encountered included the handling of and interpretation of so many results. The method adopted was to use a global approach and to try and arrange for only the most essential data to be saved to file. The end result was a combination of Linear Programming with sets of constraints, which could be allowed to vary and Simulation.

However, there were limitations caused by non-linearity but although Quadratic Equations were tried there was no apparent advantage. Other non-linear forms did not lend themselves to this particular application (Tollow 1991b). Once validated the control bands may then be applied and actions taken. One often needed action was the ability reduce the total water demand, usually because of a shortage caused by persistent lower than average reservoir levels. REDUCING THE REQUIREMENT FOR WATER The need to reduce water consumption may be as a result of a short term stoppage in supply, or a longer period of non - availability. There are two different approaches to the reduction in the requirement to supply water. The first is short term the second is long term.as an example taken from domestic use, consider the short term measure of putting a brick in the WC cistern to reduce capacity and the long term of designing and fitting a more efficient smaller cistern complete with a dual flush action. It has been interesting to note that as a result of short term action the total consumption has been suppressed for a number of years afterwards (Tollow 1993). Short term reductions Initial short term reductions in the requirement for water may be intended to last from a few days to a few months. The action may be triggered by a temporary disruption in supply or as a result of reservoir levels being below the relevant control band. The reduction may be as simple as a ban on the use of hose pipes. The severity of the required reduction dictates the methods employed, from a temporary increase in charges, when all supplies are metered, to the imposition of standpipes, when there is no alternative available. Water meters make it easier to select the appropriate action as there is an increased number of suitable options available both for short and for long term reductions. In one case some years ago a major flood destroyed part of the water transmission system which resulted in there being no water available for ten days. The only solution was to supply water by tanker. This was in fact more severe than the impositions imposed some three years earlier, when there was a major drought and customers were restricted to 400 litres per household per day (Tollow 1995a). Long term reductions Longer term reductions can take many forms. One is the introduction of rising block tariffs (Tollow 1994). Another is the planning and implementation of policies of encouraging industry to practice waste minimisation (Tollow 2000). Waste minimisation tackles not only the problem of reducing the requirement for water but also of reducing the discharge of polluted effluent. In addition these actions often stimulate byproduct recovery, which may show greater cost savings to industry than the reduction in cost of the water supplied. This may also be the way the Celtic World, with its usually abundant supplies may decide that this would be the best approach to greater overall efficiency. In some circumstances other improvements may be required as there is often good dilution of the effluent. With modern chemicals it is often better to recover them and re-use them rather than discharge them to the river or sea. Waste Minimisation may take many different forms. One uses Linear Programming to demonstrate the advantages of various options. However, there has been some resistance (Tollow 2000). Other means of demonstrating the effectiveness of waste minimisation have

been shown using Pinch analysis (Buckley et al. 2000) or other forms of analysis used for specific industrial applications (Friedrich & Buckley 2000). In addition support systems have been set up. These include waste minimisation clubs (Wynne et al. 2001). These, despite original industry scepticism due to the commercial in confidence problem (Tollow, 2000), appear to be working in certain industries. Considerable savings, due in part to byproducts and also due to reduction in use in chemicals, has taken place (Buckley et al. 2000). Indirectly the desired result of reducing the requirement for potable water has been achieved. Another alternative to individual re-use has also been developed. This is the conversion of treated sewage effluent from an industrial area to potable quality water for reuse in industry. The effluent would otherwise have been discharged to the sea. The reclaimed potable quality water is then sold to nearby paper, petro-chemical and other industries. The result of this action has been to postpone the development of alternative resources which otherwise have been needed. Considerable savings are anticipated. However, the initial effect has been an increase in the cost of potable water to the domestic consumer and to other industries, since less fresh water is needed. This is a problem that only occurs where water is metered and charges levied on actual consumption. In other cases the resulting costs are offset against an income derived from a tax of some description. RELIABILITY, RESILIENCY & VULNERABILITY Alternative measures are required to indicate the effectiveness of the results from the analyses. One approach has been the development of credibility factors and a credibility index (Tollow 1991b).The objective was to quantify in numerical terms the reliability risk, resiliency and vulnerability of the system using a single index the credibility of the actions. This was successfully applied in simulation terms to a relatively straightforward catchment using at least ten sets of 100 years of generated data. However, scope for further development remains. Other means of describing the resilience, reliability and vulnerability have been formulated for the same Catchment (Kjeldson & Rosbjerg 2001). These included samples of some of the system users, including irrigators and environmental demands. In addition both demand management and resource provision were included. A global value for the system as a whole was estimated using a sustainability index and sustainability criteria. The need for further development was highlighted. MANAGING THE DEMAND In practice managing the demand is managing the perceived requirement for water by all users, including the environment. An example, indicating the required constraints was developed. Highlighted was the need for of a different approach both in system planning and operation (Tollow 1995b). Although costs of developing resources is increasing worldwide, in one instance the sustainability criteria indicated that it would be better to develop new resources rather than practice requirement management. However, this phenomenon was explained as being due to the selection of selected users and other shortcomings. An alternative approach was needed to see if this conclusion could be verified (Kjeldson & Rosbjerg 2001). From research on the same catchment, using different criteria, a different research more in line with global trends was suggested as being more likely (Tollow 1995b). CONCLUSIONS On their own neither Resource nor Requirement management seem capable of satisfying

system needs. Both may be actively involved in meeting the overall strategy. One such approach is the combination of control bands and the management of the requirements for water by all consumers. More research and development are essential if this approach is to be implemented. Methods of assessing the results of both the proposed strategies and the results for actual operations need to be developed. These need to address the perceived reactions of both the consumers and the operators. The credibility index was developed as a means of postulating consumer reaction to an enforced reduction in availability. While sustainability may be equated to the definition of the reactions of the operator. To achieve the goal of Water Resource Management a means of defining satisfactory outcomes as well as means of managing the system, both from a resource and from a requirement aspect still needs to be developed further. This is especially important where there is a perception that there is an abundance of the raw material as in the Celtic World, where there may be, as a consequence, a greater reluctance to reduce the requirement for water. ACKNOWLEDGEMENTS To Vicky Crookes Faculty Officer for Science and Engineering Durban Institute of Technology for the translation into French of the Abstract and the Former Computer Centre For Water Research, Pietermaritzburg, for assistance with computer facilities. REFERENCES Buckley, C., Brouckaert, C. and Gianadda, P. 2001. Water Pinch Analysis: Minimisation of water and Wastewater in the Process Industry, European Commission COST 624 Optimal Management of Wastewater Systems, Working Group 5 Treatment Scenarios, Bologna. Friedrich, E., and Buckley, C. 2000. The Use of Life-Cycle assessment in comparing two Water Treatment Methods for the Production of Potable Water, 9 th National meeting SAIChe, Secunda. Hall, M. J. 1991. A study of residential water demands using fuzzy linear regression analysis Proc. Third British Hydrological Symp. Southampton September, 1, 1.29-1.34. Kjeldson, T.R. and Rosbjerg, D. 2001. A framework for assessing the sustainability of a water resources system Regional Management of Water Resources Proc Symp. 6 th IAHS Assembly Maastricht, IAHS Pubn. 268, 107-113. Tollow, A.J. 1989a. "Operation of water supply reservoirs by 'control bands' derived by simulation" Hydrological Sciences Journal, 34( 4), 449-463. Tollow, A.J.1989b. "An alternative Approach to the Generation of Low Flow Data Sequences (with selected examples from Natal)" Proc. SANCIAHS Conf. Pretoria November, 45-46. Tollow, A.J.1991a. "Operating Strategy Comparison using Validated Generated Data (with special reference to the Umgeni)" SANCIAHS 91 Stellenbosch November 8A-4.1 Tollow, A. J. 1991b. "Optimisation of system demand by Linear programming and assessment of subsequent reliability (based on the Umgeni system)" Proc. Third British Hydrological Symp. Southampton, September 2, 2.1-2.8. Tollow, A.J. 1993. "Using constrained optimisation to control growth in demand by means of a combined water and effluent pricing policy (with reference to the Greater Durban region)" Proc. Fourth British Hydrological Society Symp. - Cardiff University, September, 1.31-1.36. Tollow, A. J. 1994. "Conservation of Water Resources - the next step - a totally managed

system" Proc. 50 Years of Water Engineering in South Africa - University of Witwatersrand July, 77-92. Tollow, A. J. 1995a. "A demand management policy dictated by resource availability" IUGCXXI General Assembly, (IAHS International Assembly) Boulder, Colorado, 231 (3), 281-290. Tollow, A. J. 1995b. "Sustainable Water Resources - Matching the demand" Fifth National Hydrological Symp. Edinburgh, (September) 1.15-1.19 Tollow, A.J. 2000. Industrial Water Management - Constrained Optimisation is it the Answer 2 nd Inter Celtic Colloquium, Aberystwyth, July BHS 11(1), 13-20. Wynne, G., Maharaj, D., Buckley, C. 2001. Cleaner Production in the Textile Industry - Lessons from the Danish Experience, Proc. NATCON 2001, Durban.

for Celtic Water Colloquium, Galway, July 2002 Spatial and temporal issues in the development of policies for abstraction control in Scotland Sarah M. Dunn 1, Mark Stalham 2, Neil Chalmers 1 and Bob Crabtree 1 1. The Macaulay Institute, Craigiebuckler, Aberdeen, AB15 8QH, Scotland 2. Cambridge University Farm Potato Agronomy Unit, Cambridge, England e-mail: s.dunn@macaulay.ac.uk Abstract Abstractions of surface and groundwater for irrigation in Scotland are currently subject to control in only two small catchments. Under the terms of the EU Water Framework Directive, it will be necessary to introduce new legislation to control abstractions elsewhere. In this paper a study is carried out on two catchments in the east of Scotland, the Tyne and West Peffer, to examine how different irrigation control strategies would modify the surface water flow regimes in the catchments. The results of the study demonstrated that the West Peffer catchment in particular is significantly affected by abstractions for irrigation of potatoes. Control mechanisms based on allowable abstraction volumes and flow-based abstraction bans could be of considerable help in restoring streamflows to their natural levels, but would modify the hydrological regime in slightly different ways. A spatial analysis of streamflows demonstrated that implementation of controls based on a single monitoring point in a catchment may be ineffective at maintaining acceptable levels of flow in small streams, because of the significance of abstractions relative to the flows. Keywords: abstraction, irrigation, hydrology, streamflow, spatial, temporal Résumé Le prélèvement des eaux souterraines et de surface pour l irrigation en écosse est en ce moment sujet à contrôle dans seulement deux bassins versants. Il sera nécessaire d introduire ailleurs, sous les termes de la loi cadre sur l eau (2000/60/EC), une nouvelle législation pour contrôler le prélèvement. Dans cet article, deux bassins versants (Tyne and West Peffer, écosse de l est) ont été étudiés en vue d examiner comment différentes stratégies de contrôle d irrigation pourraient modifier le débit des cours d eau. Les résultats de l étude démontrèrent que le West Peffer en particulier est significativement affecté par le prélèvement pour l irrigation des pommes de terres. Les mécanismes de contrôle basés sur des volumes de prélèvement autorisés et une interdiction de prélèvement à certains débits pourraient aider considérablement à restaurer les débits à leurs niveaux naturels, mais modifiraient le régime hydraulique dans des directions quelque peu différentes. Une analyse spatiale des débits démontra que la mise en place des contrôles, basés sur un seul point d échantillonage dans le bassin versant, peut être inefficace au maintien de débits acceptables dans les petits ruisseaux, à cause des prélèvements significatifs relatifs aux débits. Mots clés: prélèvement d eau, irrigation, hydrologie, débit, spatiale, temporel Introduction Legislation controlling abstraction of water from surface and groundwater sources in Scotland has historically been relatively relaxed. Although specific statutes govern abstraction of water for specific purposes, such as public water supply, the right to abstract water from surface and groundwater is founded in common law. The complex issues surrounding ownership of water are described by Wright (1995), but essentially they mean that riparian landowners are entitled to make use of water in a watercourse that flows through their land. In order to meet the requirements of the European Water revised 05/04/02

Dunn et al. Spatial and temporal issues in abstraction control Framework Directive (2000/060/EC), new legislation must be introduced to control abstractions of this type. One of the primary forms of abstraction that is currently subject to only limited control in Scotland is irrigation. Irrigation in Scotland is carried out mainly for potato crops, and less commonly for salad crops, grass and soft fruits. It takes place in all parts of the country where potato production occurs, but is most significant in the east, in areas such as Angus, Perthshire, Fife and East Lothian. These areas have a drier climate than other parts, making water resources more scarce, and hence both the need for and impact of irrigation abstractions is exacerbated. A form of legislation to control abstraction of water for irrigation has existed since the introduction of the Spray Irrigation (Scotland) Act 1964, subsequently replaced by the Natural Heritage (Scotland) Act 1991. Although a number of catchments are perceived to be significantly affected by irrigation abstractions, in practice, control orders have only ever been issued to cover two small areas because of the cumbersome nature of the application procedure. Set against a background of increasing demand for irrigation water (Stansfield, 1997 and Weatherhead et al., 2000), it is important that a future strategy is developed to control irrigation abstraction in Scotland. There are many different mechanisms by which abstractions may be controlled. Adeloye (1996) discusses the pros and cons of some approaches that might be introduced in Scotland. Examples that have been used elsewhere include a simple licence system based on a maximum rate of abstraction, a system based on a flow-linked abstraction ban, or a volume based approach giving a monthly or seasonal abstraction allowance. To help in the development of appropriate policy for abstraction control in Scotland, a study has been carried out to examine the effectiveness of different types of control strategies in terms of the economics of potato cropping and stream hydrology. The study was focussed on two catchments in East Lothian, the Tyne and the West Peffer, where potatoes are an important crop. The West Peffer is one of the two catchments in Scotland where a control order for irrigation abstractions currently exists. A spatial modelling approach has been applied to simulate flows in the catchments under natural conditions and different scenarios of irrigation management, based on data pertaining to potato cropping in the area. This paper presents the findings of the hydrological study and highlights some of the spatial and temporal issues that need to be considered in the selection of control mechanisms, if they are to be successful in achieving objectives for environmental improvement. Modelling Approach Hydrological modelling was employed within the irrigation study for two purposes. The first requirement of the analysis was to provide input data to a potato growth model to identify the availability of water for surface abstraction at different times. The output from the potato growth model subsequently provided feedback to the hydrogical analysis by establishing optimal irrigation scheduling under those conditions. Hydrological analysis was also necessary to examine how different scenarios of irrigation control would modify the hydrological regime, and hence how effective they would be in achieving environmental improvements, in terms of stream flows. A modelling approach was used to derive spatial predictions of naturalised flow, which were then superimposed with spatial mappings of irrigation abstraction scenarios. This allowed the effectiveness of the control instruments to be assessed across the whole catchment, rather than just at a single point on the watercourse. Knox et al. (2000) demonstrated the importance of this in their study of financial implications of irrigation restrictions in eastern England. The DIY model (Dunn et al, 1998) was used as the basic tool to derive the naturalised flows. It is illustrated in Figure 1 with a summary of model parameters given in Table 1. The model uses a GIS approach to spatially categorise a catchment into 50x50 m cells, on the basis of a range of physical properties. These properties typically include rainfall, elevation, hill-slope, distance to stream and soil type. For each characteristic category, a hill-slope routing model is used to calculate a time-series signature of the flow, which that category contributes to the stream. The flows from each category are combined and weighted by their relative numbers to determine the catchment stream flow. This means that stream flows can be estimated for any location within the catchment, by determining the contributing areas to the stream at each point. 2

Dunn et al. Spatial and temporal issues in abstraction control One of the aims of the study was to evaluate the manner in which typical irrigation abstractions affect stream flows, not just over the catchment as a whole, but also at any specific location in the stream network. A spatial mapping of potential irrigation abstractions was therefore derived, to link to the spatial predictions of natural stream flows. From the potato crop model different time-series scenarios of irrigation abstraction were derived to correspond to a range of potential policies that might be used to control surface water abstraction on the basis of flows at a particular point in the system. Information on potato production in Scotland was available from the June Census, the British Potato Council (BPC) census and from IACS (Integrated Administration and Control System) returns. The IACS returns are incomplete, because potatoes are not a supported crop under the area payments scheme, but they are of value in providing a precise geographical reference for locations of potato fields. Areas of potato cropping from the IACS data were associated with their geographical reference and then weighted to give the correct total area according to the June Census. It was assumed that abstractions for irrigation were from surface water and that they would take place from the nearest location on the stream network to the geographical reference of the potato field. This made it possible to spatially combine irrigation abstractions with the predictions of natural flow from the DIY model, using GIS analysis. The resulting model maps the net flow across the entire catchment stream network. Case Study Catchments The Tyne and West Peffer case study areas are both situated in East Lothian (Figure 2). The Ordnance Survey 50m digital elevation model (DEM) was used to derive the boundaries of the two catchments and to generate a digital stream network. The catchments were selected on the basis that they had a high proportion of agricultural land used for potatoes, and were perceived to be potentially vulnerable in terms of low flows in the rivers. The West Peffer is already subject to abstraction controls, whereas the Tyne is not. Statistics on the area of agricultural land and potatoes given in Table 2 highlight the intensity of potato production in the West Peffer in particular. However, the distribution of potato production as shown by the IACS locations of potato fields (Figure 2) also show parts of the lower Tyne catchment to be quite intensively used. Topographically, the West Peffer catchment is a low-lying area with minimal relief. The West Peffer Burn flows east to west into the Firth of Forth, with stream flows gauged at Luffness. This point is upstream of the main northern tributary in the catchment, and hence gauges an area of only 26km 2, compared with the 46 km 2 of the whole catchment. The catchment has a mean annual rainfall of around 600mm. When combined with potential evapotranspiration rates of around 500mm per year, it is clear that there is little net runoff in the catchment, even without irrigation abstractions. Although controls have been placed on agricultural abstraction in the catchment for many years now, this has not prevented very low flows from occurring in the stream. Historical data from 1989-1998 show that for 10% of the time, the flow in the West Peffer is equivalent to less than 12mm of runoff per year, a volume that might typically be used to irrigate the whole area in only a few days. The Tyne catchment covers a much larger area (329 km 2 ) with the main river flowing from south-west to east and discharging into the North Sea to the north of Dunbar. Stream flows are monitored near the outlet at East Linton and also upstream at Spillmersford, encompassing an area of around 161km 2 (Figure 2). Lower parts of the catchment are quite similar to the West Peffer, but the source of the Tyne lies in the Lammermuir Hills at an elevation of 527m. As a consequence, both the hydrology and land use of the Tyne are more varied than the West Peffer. Precipitation in the upper part of the catchment averages around 800mm per year, generating much greater specific runoff rates. In this respect, flows from the upper part of the catchment are likely to be beneficial in maintaining acceptable levels of streamflow in the lower catchment, where irrigation abstractions for potato production are more likely to occur. By comparison with the figure of 12mm for the West Peffer, there is an equivalent runoff of 60mm per year that is exceeded for 90% of the time in the Tyne at East Linton. Model Application and Results The DIY model was applied together with the potato growth and irrigation scheduling model to a 10- year period of meteorological data covering 1989-1998. Spatial categorisation for the DIY model was derived from maps of elevation, flow-length to stream, hill-slope and up-slope contributing area. Soils 3

Dunn et al. Spatial and temporal issues in abstraction control were assumed to be homogeneous for the purposes of the hydrological analysis, as the model could be expected to be insensitive to the relatively subtle differences in hydrological behaviour of the soils across the catchments, and the parameter values were to be calibrated. DIY Simulation of Natural Flows Naturalised mean daily flows were simulated from an input of daily precipitation and evapotranspiration, with parameter values calibrated using streamflows at Spillmersford measured during 1989 and 1998. The effects of irrigation abstractions above Spillmersford will be very small relative to the total flow, as the majority of potato cropping occurs lower in the Tyne catchment. No other forms of management of the river system were considered likely to have a significant impact on stream flows. The predicted natural flows for points in the catchment corresponding to each of the three gauging stations were compared with measured managed flows over the full 10-year period (Figure 3). Nash and Sutcliffe efficiencies (Nash and Sutcliffe, 1970) for the full simulations were good, with values of 0.83, 0.85 and 0.88 for Spillmersford, East Linton and Luffness respectively, despite the influence of abstractions on the measured flows. A closer examination of the low flow periods reveals that predicted natural flows at Spillmersford follow the measured flows closely, although there is a tendency for the model to slightly over-predict. The predictions of natural flow at East Linton are also only slightly higher than measured flows during low flow periods, suggesting that irrigation abstractions also have a limited impact on flows in the main stem of the river at this point. However, at Luffness there is a very large discrepancy between the modelled natural flow and measured data. The model predictions suggest the presence of a natural baseflow that does not fall below 0.03 m 3 s -1, whereas in practice during most summers flows have been reduced to, or very close to, zero. The shape of the early hydrograph recession for the modelled flows suggests that the model may be slightly over-predicting low flows, but certainly not to the extent that the measured data would suggest. From this it is clear that summer abstractions have an important influence on flows in the main stem of the West Peffer Burn. Impact of irrigation abstraction scenarios on stream flows The potato growth and quality model developed at Cambridge University Farm (Stalham and Allan, in press and Stalham et al., in press) enables the relationships between irrigation amounts and potato yield and quality to be modelled. This means that irrigation scheduling for optimum quality can be predicted using information about variations in catchment flow and further used to quantify the effect of any water restrictions on grower returns, as a result of reduced quality or yield. A range of irrigation scenarios was studied based on different flow criterion for restricting abstractions. These were then combined with the modelled natural flows to demonstrate the effect of the irrigation abstractions on stream flows. The scenarios included: 1) optimal irrigation of crop for growth and quality 2) a ban on abstraction when flows fall below the level of the 98%ile natural flow 3) a ban on abstraction when flows fall below the level of the 95%ile natural flow 4) a ban on abstraction when flows fall below the level of the 90%ile natural flow 5) a maximum allowed abstraction of 10% of the mean monthly available flow volume 6) a maximum allowed abstraction of 20% of the mean monthly available flow volume There are several issues that need to be examined in relation to the effectiveness of the abstraction controls, including: 1) How the natural flow statistics are modified by abstractions up to this level 2) The impact of abstractions that occur when flows are just above the limiting level 3) The spatial effectiveness of the abstraction controls The potato growth model was used to determine the area of potato cropping that could be optimally irrigated, and the time-scheduling for that irrigation, based on the water available for abstraction according to each scenario. These data were passed back to the hydrological model to calculate time- 4

Dunn et al. Spatial and temporal issues in abstraction control series of managed flows under each scenario. Table 3 summarises the flow duration statistics for the West Peffer at Luffness, calculated over the period 1989-98, for the six scenarios described above. The results show that all of the instruments significantly improve the levels of low flows over the historic measurements and the optimal irrigation scenario. The natural flows that are exceeded for at least 60% of the time are largely unaffected by the abstractions. There is still a significant reduction in the levels of the low flows when compared to the predictions of natural flow, but this is inevitable as long as any abstraction occurs. An abstraction allowance based on 10% of the mean monthly flow volume results in a flow regime that is closer to the natural flow of the river than any of the percentile based instruments, apart from at the very driest times. However, an allowance based on 20% of the mean monthly flow volume is ineffective at controlling abstractions during dry periods. The duration over which stream flows are affected by the irrigation abstractions is quite significant and is summarised in Table 4 for each of the scenarios when applied to the West Peffer. Even with a ban on abstractions imposed at the 90%ile flow level, flows fall below the 98%ile level for 10% of the time, which is the same length of time as the optimal irrigation scenario. However, the amount by which the flows drop below the 98%ile is reduced compared with the optimal irrigation. In addition, potential levels of abstraction are such that, even when there is no ban in place just above the 90%ile level, the demand for irrigation in the West Peffer could drop streamflows to a very low level. Although the instruments would minimise the amount of time that streams are dried out, they would not prevent the situation from arising. A ban would have to be introduced at an unrealistically high level for this to be achieved. The instruments based on allowable monthly abstraction volumes are more successful in terms of the period of time that the instruments fail to achieve particular flows. The allowance based on 10% fails the 98%ile flow for only 5% of the time and the allowance based on 20% fails the same level of flow for 9% of the time. However, from the flow duration statistics in Table 3 it is clear that when the instruments based on allowable monthly abstractions fail, that the severity of the failure will be greater. The instruments that have been studied are all based on monitoring of stream flows at one particular point in a catchment. Even if the control mechanism operates satisfactorily at this point, there is a danger that it may not be effective in controlling flows in other parts of the catchment. To examine this issue, the effect of a 2.5mm / day irrigation abstraction on flows across the Tyne and West Peffer has been calculated, for an example day when flows at Luffness and East Linton are at their 90%ile level. The results of this are illustrated in Figure 4. For the Tyne, flows in most of the catchment are greater than 75% of the natural flow. However, there are several small tributaries where flows are significantly reduced to less than 50% of their natural level. For the West Peffer, flows throughout the catchment are grossly affected by the abstraction, and the main southern tributary does not have the capacity to meet the demand. Discussion and Conclusions The case study analysis of abstraction for potato irrigation in the Tyne and West Peffer catchments confirmed that optimal irrigation of potato crops in the West Peffer catchment has a very significant impact on stream flows during the summer, and can very easily cause the streams to dry out. However, in most parts of the Tyne catchment stream flows are high enough to buffer the impact of irrigation abstractions, despite quite intensive potato cropping in the lower catchment. The severity of the impact in the West Peffer catchment is notable particularly because the magnitude of the individual abstractions is quite small. Yet, the combined effect of the individual abstractions is large, because the stream drains a small area, has a low specific runoff and a high proportion of the land is used for potatoes. Factors of this type need to be taken into consideration in establishing legislation for abstraction control, where it is likely that a lower limit may be set on the magnitude of abstractions that require licensing. The study of the different flow-based control mechanisms demonstrated how the hydrological regime would be modified in slightly different ways, by abstraction bans at specific flow thresholds as compared to monthly allowable abstraction volumes. Selection of one mechanism in preference to another will depend on the practicalities of implementation as well as the specific effect on flow statistics. 5

Dunn et al. Spatial and temporal issues in abstraction control Clearly, the major problem associated with abstractions for irrigation is that the water is required at times when stream flows are at their lowest. In this respect, direct surface abstraction will have the greatest impact on flows. It was assumed in this analysis that all abstractions were from surface water, but in practice a survey of potato growers in the West Peffer showed that most had either a storage reservoir or borehole to supplement the supply from the river. However, no reservoirs or boreholes were reported in the Tyne. The use of storage reservoirs in particular could be seen as a good solution to the irrigation problem, enabling abstractions to be made during the winter and spring when stream flows are higher. From the growers perspective this would also seem to be advantageous, as it would ensure that water was actually available at the times when it is needed, removing some of the uncertainty caused by unknown future weather. Decreases in the demand for irrigation water could also be achieved through more efficient water use, for example by employing trickle irrigation methods in place of spray irrigation, or by the use of short-term weather forecasts in irrigation scheduling, as illustrated by Gowing (2000). The preferential use of groundwater over surface water abstraction may also be less disruptive to stream flows as it smooths the effect of the abstraction over time. However, controls on the use of both impounding reservoirs and abstractions from groundwater will also be required by the EU Water Framework Directive. Temporal aspects are also important when it comes to practical implementation of controls. The regulatory authority will need to give growers warning about probable imposition of an abstraction ban, when flows drop close to the limit. However, the response is likely to be an immediate increase in abstractions, thus exacerbating the situation. The spatial analysis of the irrigation impacts highlighted the problem of using a point measurement of flow as the criterion for triggering an irrigation ban. Due to the heterogeneity of abstraction activities and the different sizes of streams from which they occur, the level at which the ban is triggered may be inadequate to prevent smaller tributaries from being dried out. This effect was observed in the analysis of the Tyne catchment as well as the West Peffer. In practice, to achieve maintenance of flows at, say, the 95%ile across all of the catchment, it is probably necessary to set a level of the 90%ile flow as the criterion at a downstream point. Similar spatial issues were highlighted by Bonvoisin and Moore (1993), in relation to the assessment of discharge consents and abstraction licences. Despite the fact that as a nation Scotland has plentiful water resources, there are areas of the country where streams are presently under threat from over-abstraction. Under the terms of the EU Water Framework Directive it will be necessary to introduce new legislation in Scotland to control abstractions from surface and groundwaters. This study has highlighted some of the management issues in relation to control of irrigation abstractions and demonstrated how spatial and temporal variability need to be taken into consideration, if the controls that are introduced are to be effective in achieving environmental improvement. Acknowledgements This research was funded by the Scottish Executive Environment and Rural Affairs Department. References Adeloye, A.J. and Low, J.M., 1996, Surface-water abstraction controls in Scotland, J. CIWEM, 10, 123-129. Bonvoisin, N.J. and Moore, R.V., 1993, The use of GIS techniques to assess discharge consents and abstraction licences, HydroGIS93: Application of Geographic Information Systems in Hydrology and Water Resources, ed. K.Kovar and H.P. Nachtnebel, IAHS Publ. 211, 345-354. Dunn, S.M., McAlister, E. and Ferrier, R.C., 1998, Development and application of a distributed catchment-scale hydrological model for the River Ythan, NE Scotland, Hydrol. Processes, 12, 401-416. Gowing, J.W. and Ejieji, C.J., 2000, Real-time scheduling of supplemental irrigation for potatoes using a decision model and short-term weather forecasts, Agric. Wat. Man., 47, 137-153. 6

Dunn et al. Spatial and temporal issues in abstraction control Knox, J.W., Morris, J., Weatherhead, E.K. and Turner, A.P., 2000, Mapping the financial benefits of sprinkler irrigation and potential financial impact of restrictions on abstraction: A case-study in Anglian Region, J. Env. Man., 58, 45-59 Nash, J.E. and Sutcliffe, J.V., 1970, River flow forecasting through conceptual models. Part 1 a discussion of principles, J. Hydrol. 10, 282-290. Stalham, M.A. and Allen, E.J., in press, Development of an irrigation scheduling model for the potato crop, J. Ag. Sci. Stalham, M.A., Gaze, S.R. and Allen, E.J., in press, Comparison of irrigation scheduling models for the potato crop. J. Ag. Sci. Stansfield, C.B., 1997, The use of water for agricultural irrigation, J. CIWEM, 11, 381-384. Weatherhead, E.K. and Knox, J.W., 2000, Predicting and mapping the future demand for irrigation water in England and Wales, Agric. Wat. Man., 43, 203-218. Wright, P., 1995, Water resources management in Scotland, J. CIWEM, 9, 153-163. Tables Table 1: DIY model parameters Parameter Variable Function Active zone conductivity KACT Control rate of sub-surface flow Threshold storage THMAX Soil moisture level at which surface runoff response is initiated Total soil porosity PORE Define relationship between soil storage and head Fast response distance FASTD Define density of fast drainage network for surface runoff Groundwater recharge fraction RCHGF Control recharge rate to groundwater Groundwater conductivity KGW Control rate of groundwater flow Slope to stream SLOPE Define hydraulic gradient for hill-slope model Flow path distance to stream DIST Define hill-slope routing distance for each cell Up-slope contributing area UPAR Control soil storage through balance between upslope inflow and down-slope outflow Snow accumulation and melt parameters TS,TM,K,F G,FT Partition precipitation between rainfall and snow and determine snowmelt rates Table 2: Statistics describing agricultural land use in the Tyne and West Peffer catchments Total area (km 2 ) Agricultural area (km 2 ) Area of potato production (ha) Tyne 329 227 631 300 West Peffer 46 37.9 484 380 Potentially irrigated potato area (ha) 7

Dunn et al. Spatial and temporal issues in abstraction control Table 3: Flow duration statistics for flow controlled abstraction scenarios, compared with both predicted natural and historical measured flow statistics. % time flow exceeded Natural flow Predicted flows at Luffness (West Peffer) for different scenarios (m 3 s -1 ) Optimal 98%ile 95%ile 90%ile 10% 20% irrigation ban ban ban month month Measured flow scenario scenario scenario volume volume 60 0.062 0.061 0.060 0.060 0.060 0.061 0.061 0.043 80 0.056 0.039 0.036 0.037 0.036 0.042 0.040 0.020 90 0.036 0.032 0.030 0.032 0.032 0.035 0.033 0.010 95 0.034 0 0.023 0.026 0.028 0.032 0.016 0.004 98 0.032 0 0.008 0.012 0.015 0.012 0 0.002 Table 4: Duration of failure to achieve natural flow thresholds for flow controlled abstraction scenarios Natural flow %ile Optimal irrigation scenario Percentage of time scenario fails flow threshold 95%ile ban 90%ile ban scenario scenario 98%ile ban scenario 10% monthly volume 20% monthly volume 90 17 20 20 20 13 16 95 12 16 15 15 8 12 98 10 14 11 10 5 9 Figures Figure 1: Schematic of the DIY model used for rainfall-runoff modelling Figure 2: Location of the case study catchments showing the stream network, locations of potato cropping and flow gauging stations Figure 3: Predictions of natural flow compared with measured managed flows for low flow periods at East Linton, Spillmersford and Luffness Figure 4: Effect of a 2.5mm/day irrigation abstraction on flows across the Tyne and West Peffer, when natural flow is at its 90%ile level at East Linton and Luffness 8

for Celtic Water Colloquium, Galway, July 2002 Cell model and routing Precipitation Evapotranspiration Snow Net rain Surface Surface runoff Category parameter set Subsurface Ground water Hill-slope routing Cell signature of flow Subsurface storm flow Total flow calculation Cell water balance Groundwater flow Figure 1. Schematic of the DIY model used for rainfall-runoff modeling revised 05/04/02

Dunn et al. Spatial and temporal issues in abstraction control W. Peffer Catchment Luffness East Linton Spillmersford Gauging stations Location of potato cropping Tyne Catchment Figure 2 0 5 10 kilometres N Figure 2: Location of the case study catchments showing the stream network, locations of potato cropping and flow gauging stations 10

for Celtic Water Colloquium, Galway, July 2002 East Linton, 1989-98 Measured flow Modelled natural flow 6 5 4 3 2 1 0 01/01/89 01/07/89 01/01/90 01/07/90 01/01/91 01/07/91 01/01/92 01/07/92 01/01/93 01/07/93 01/01/94 01/07/94 01/01/95 01/07/95 01/01/96 01/07/96 Flow (m 3 s -1 ) 01/01/97 01/07/97 01/01/98 01/07/98 Date Spillmersford, 1989-98 Measured flow Modelled natural flow 3 2.5 Flow (m 3 s -1 ) 2 1.5 1 0.5 0 01/01/89 01/07/89 01/01/90 01/07/90 01/01/91 01/07/91 01/01/92 01/07/92 01/01/93 01/07/93 01/01/94 01/07/94 01/01/95 01/07/95 01/01/96 01/07/96 01/01/97 01/07/97 01/01/98 01/07/98 Date Luffness, 1989-98 Measured flow 0.5 Modelled natural flow Flow (m 3 s -1 ) 0.4 0.3 0.2 0.1 0 01/01/89 01/07/89 01/01/90 01/07/90 01/01/91 01/07/91 01/01/92 01/07/92 01/01/93 01/07/93 01/01/94 01/07/94 01/01/95 01/07/95 01/01/96 01/07/96 01/01/97 01/07/97 01/01/98 01/07/98 Date Figure 3 Figure 3: Predictions of natural flow compared with measured managed flows for low flow periods at East Linton, Spillmersford and Luffness revised 05/04/02

for Celtic Water Colloquium, Galway, July 2002 Managed flow as % of natural flow when 2.5mm/day irrigation water abstracted at 90%ile flow 75-100 50-75 25-50 0-25 W. Peffer Catchment < 0 Tyne Catchment Figure 4. Figure 4. Effect of a 2.5 mm/day irrigation abstraction on flows across the Tyne and West Peffer, when natural flow is at its 90%ile level at East Linton and Luffness revised 05/04/02

Natural soil piping, water quality and catchment management in the British uplands J A A Jones Institute of Geography and Earth Sciences, University of Wales Aberystwyth SY23 3DB, UK Abstract Hydrological studies of natural soil piping in the Cambrian Mountains have shown pipeflow to be an important contributor to streamflow in headwater catchments, on occasions contributing as much as 70% of quickflow in the stream. Reconnaissance surveys of the distribution of piping in Britain have identified over 70 basins that possess pipe networks and analyses of the climatic, physiographic and edaphic properties of these basins suggest that up to 30% of Britain may be subject to pipe development. With minor exceptions, for example in pelosols with high shrinkage potential in East Anglia, these basins lie within the upland zone dominated by histosols and spodosols. In the Maesnant basin, Plynlimon, soil pipes have been found to double the average partial contributing area. Moreover, this additional contributing area can lie well beyond the riparian zone source area, which is commonly seen as the main contributing area. On Maesnant, the longest stretch of pipe network extends 750 m from the stream, the pattern of contributing area is highly irregular, and in many areas the location of the piping bears little relationship to standard topographic indices, like a/s and its variants, which have been used in physically-based runoff simulation models, like TOPMODEL, to define contributing areas. Equally, standard algorithms for calculating flood discharges ignore soil piping, and may significantly underestimate concentration times, or else approximate them for the wrong reasons, e.g. where pipeflow is an effective substitute for overland flow. These pipe networks can play an important role in determining water quality within the streams. Deficiencies in modelling the impact of acid rain on upland streams may be partly due to ignoring the role of soil pipes. Pipes may transmit faecal pathogens, heavy metals and other pollutants from extensive areas of hillslope classically regarded as non-contributing areas. They therefore have important implications for landuse planning. Afforestation on a piped hillslope may cause greater surface water acidification than expected from standard theory. Ploughing and reseeding of upland pasture can produce a negative result if pipe networks are destroyed and waterlogging results. The paper analyses field observations of water quantity and quality in piped catchments and considers the implications for basin management. Keywords: Catchment management, soil piping, hillslope hydrology, upland water quality 1. Introduction The way we use our uplands in Britain is the subject of change and current debate. The impetus behind conifer afforestation has faltered as the need for the product has declined, the economics have changed and the contribution to acidification and aesthetic despoliation of the landscape have been recognised. Sheep-rearing and hill farming in general have suffered under the 2001 foot-and-mouth epidemic, which hit an industry already reeling from unprofitability and the prospect of reduced subsidies as the EU expands in the coming years. Plans are now afoot for widespread reengineering of the upland landscape, nurturing biodiversity and perhaps recreating a more natural land cover, returning to a pre-industrial and maybe even prehistoric landscape. In Wales, Tir Coed has put forward a proposal to develop a national forest based on deciduous trees and shrubs, similar to that already being developed in the English Midlands (Tir Coed, 2001). These changes in landuse are likely to have many hydrological implications. Yet, despite some of the most detailed and informative experimental studies of the 1

impact of conifer plantation and harvesting undertaken by the NERC Centre for Ecology and Hydrology in mid-wales, our understanding of the effects of upland deciduous plantations and any guidelines designed to limit the hydrological impacts are rather lacking. This paper considers just one feature of the upland drainage system that currently contributes to biodiversity and may both help and hinder in the rehabilitation of acidified surface waters: natural soil piping. Natural soil pipes are the largest and most highly connected form of macropore and as such are potentially a major source of bypass flow within the soil body and a contributor to quickflow of comparable importance to classical overland flow in stream hydrographs. Jones et al. (1997) estimated that up to 30% of the land surface of Britain may be susceptible to pipe development (Figure 1). Almost all of the 70 catchments that displayed piping in their survey were in the uplands; the only exception was a pelosol in East Anglia with a high cracking potential. The piped basins are predominantly covered by podzolic soils (especially ferric stagnopodzols and brown podzolics) or peaty soils (especially raw oligo-fibrous peats) with rock exposures, mean annual rainfall in the range 1500-2000 mm, altitudes peaking around 500 m OD, and with mainstream slopes of the order of 10 o. A considerable amount of evidence has been published in recent years which supports the view that natural pipeflow can be an important pathway for hillslope drainage and a significant process in streamflow generation, especially in headwater catchments in the uplands. The evidence was first collected in catchments in the Welsh mountains, but this has been supplemented by a number of field monitoring experiments in a variety of climatic regimes, notably during the last decade. In comparison with the study of pipeflow regimes, research into the chemistry of pipeflow, especially its role in the acidification of surface waters, effects on aluminium concentrations and the movement of plant nutrients, remains very limited and the evidence available to date is highly complex. This paper attempts to review the evidence for the effects of pipeflow on streamflow response and water quality, and considers the possible implications for the management of piped catchments. 2. The Maesnant experiments Most of the detail in the following review is based on research in the University of Wales experimental catchment on the Maesnant, a headwater of the River Rheidol in mid-wales. By chance, this basin appears to be typical in all respects of the average piped basin in Britain described above. The Maesnant stream is a second order stream draining a 0.54 km 2 basin on the western flank of Plynlimon (Pumlumon) in the Cambrian mountains. The catchment ranges from 752 m O.D. at the peak of Plynlimon to 465 m O.D. at the combined v-notch gauging station. It is underlain by Ordovician greywacke, mudstone and grits, with soliflucted drift largely forming a small river terrace. Annual rainfall is around 2200 mm, with an annual water surplus of 1800 mm. The gauged section of the stream is 750 m long between upper and lower weirs (Figure 2). Intensive field monitoring was undertaken in the Maesnant catchment for a total of 8 years during the 1980s and 1990s (e.g. Jones and Crane, 1984; Jones, 1987; Hyett, 1990; Richardson, 1992; Connelly, 1993; Jones, 1997b). This involved continuous data logging at 10 minute intervals at up to 17 pipeflow sites, 3 riparian 2

seepage zones, 2 stream weirs and 2 tipping bucket raingauges. Initially, monitoring at the basin outfall was based on a Welsh Water weir that was enlarged for this research. This was subsequently replaced by an Institute of Hydrology weir. Water quality was monitored over a 3-year period using auto-samplers covering up to 13 pipeflow sites and at stream sites above and below the main inputs from the perennial pipes, plus spot sampling of soil water at the dipwell sites marked in Figure 2 (Hyett, 1990). This was supplemented by a single basin-wide survey of surface soil water extracts taken from 234 sites (Richardson, 1992). The most recent research has concentrated on the factors controlling the development of the pipe networks (Jones et al., 1997) and on devising a physically-based hydrological simulation model for pipeflow (Jones and Connelly, 2002). 3. Effects of pipeflow yields on streamflow response 3.1. Evidence from Maesnant and other sites in the British uplands The British uplands have been the principal world source of information on pipeflow. The experimental evidence comes mainly from three catchments in Wales and three in northern England: in Wales, the Maesnant (Jones, 1978, 1981, 1982; Jones and Crane, 1984; Jones, 1987, 1988; Jones et al., 1991; Jones, 1997a, c, d; Jones and Connelly, 2002), the adjacent Centre for Ecology and Hydrology Upper Wye catchment on the eastern slope of Plynlimon (Gilman and Newson, 1980; Chapman, 1994; Sklash et al., 1996; Chapman et al., 1993, 1997), and the Nant Llwch basin in the Brecon Beacons of South Wales (Wilson and Smart, 1984); in England all the basins lie within the Pennines, the Slithero Clough, Derbyshire (McCaig, 1983, 1984), Shiny Brook, near Huddersfield (Gardiner, 1983; Burt et al., 1990) and the Little Dodgen Pot Sike (LDPS) in the Moor House National Nature Reserve (Holden and Burt, in press). The earliest published study of piping in Britain comes from Jones (1971) on the Burbage Brook tributary of the River Derwent in Derbyshire. These observations generally indicate that pipeflow can be a very important contributor to streamflow in many upland headwater basins, although there are wide differences in the amount contributed in both space and time. Two important results from the Maesnant have been (1) proof that the response time for pipeflow can be quick enough to contribute storm runoff to the surface streams and (2) proof that the volumes of pipe discharge can be sufficient to supply a significant and on occasions a dominant proportion of flow to the rivers. The first of these disproves the earlier view of Whipkey and Kirkby (1977) that rain infiltration would take too long to reach the depth of the pipes to contribute to storm runoff in a basin. Their hypothesis would be more likely to be true if conventional diffuse infiltration were the only process, but it is now clear that bypass flow is more important, feeding the pipes directly through cracks or blowholes in their roofs or indirectly via crack flow feeding the phreatic surface and raising it to pipe level (Jones and Connelly, 2002). The detailed monitoring shows that there is a wide range of response times within the pipes, some before the stream, some after, but centring around the average response time of the stream in terms of both start of response and peak flows (Jones, 1988). Indeed, there is some evidence, comparing flows at the upper and lower stream weirs, that the pipe discharge forces stream response to be earlier as it passes the main outlets of the pipes. The Maesnant pipeflow may reach peak runoff rates that are comparable to saturation overland flow in small basins of 3

0.001 up to 0.2 km 2, although on average they are only about one-fifth those of saturation overland flow (Jones, 1997a). The second result counters the suggestion by Gilman and Newson (1980) that although pipes may be important as drains for hillslopes, they do not contribute much to streamflow because they end at the edge of the valley bottom. The Maesnant pipes generally discharge on the edge of the river terrace and dye tracing experiments indicate lags of only around 10 minutes for the outflows to reach the stream. In other basins, like the Burbage Brook, around 150 pipes issue directly from the banks into the stream. In the Maesnant, pipeflow contributes around 49% of stormflow and 46% of baseflow to the stream, with the figure rising to over 50% and even 70% in individual storms when moderately heavy rain (c. 10-60 mm) falls on a moderately wet catchment (with seven-day antecedent rainfall between c. 30 and 160 mm). In wetter conditions, saturation overland flow contributions reduce the percentage coming from the pipes. In drier conditions, the percentage contributed by riparian seepages rises. These results are corroborated by Wilson and Smart s (1984) calculation that ephemeral pipeflow contributes an average of 68% of streamflow in the Nant Llwch basin in the Brecon Beacons, South Wales, although this figure was based on indirect estimates, partly based on artificial pumping experiments on pipe flow capacities, rather than on direct measurement of natural pipeflow. The Maesnant research has also shown that there is great spatial diversity in pipe yields. There is nearly an order of magnitude difference between yields from individual perennially-flowing pipes in the basin in both mean stormflow discharge and peak discharges, and a 60-fold difference between the smallest ephemerallyflowing and the largest perennial pipe. In contrast, pipeflow has been reported to be rather less significant in a number of other upland basins. In a highly eroded peat bog in English Peak District, Gardiner (1983) and Burt et al. (1990) estimated that only about 1% of streamflow was derived from pipeflow. The density of piping was certainly lower there than on Maesnant. However, the study omitted to monitor flows from the larger pipes. More convincing evidence from a deep blanket peat catchment in northern England comes from Holden and Burt (in press) who monitored only a 10% contribution, plus about 0.5% from hand-sampled pipes (Holden, pers. comm., 2001). However, the percentage contributions from Holden and Burt s pipes rose to 30% during stream recession, suggesting that the real difference between these pipes and those of Maesnant lies more in the timing of contributions: flows from the pipes in the deeper blanket peat of their LDPS catchment are more delayed and miss the peak streamflow. Chapman (1994) and Chapman et al. (1997) also report only a 10% contribution from shallow ephemeral pipes in a 4 ha headwater study site within the Centre for Ecology and Hydrology s Upper Wye catchment, adjacent to the Maesnant basin. In this case, the difference is that the Maesnant basin contains more large perennially-flowing pipes, and though these exist in parts of the Upper Wye, they have not been monitored and included as pipeflow. Interestingly, the direct contribution to streamflow from ephemeral pipes on Maesnant is very comparable to this amount, if the large proportion of ephemeral pipeflow that feeds through the perennial pipes is excluded. Also, although the early work in the Nant Gerig basin in the Upper Wye catchment reported by Gilman and Newson (1980) did not record flow in the perennial pipes either, nor, more importantly, the flow in the adjacent 4

stream, it was estimated that an average of 34% of rainfall drained through the ephemeral pipes. It is also notable that contributions from these ephemeral pipes rise markedly during storm runoff and Chapman (1994) and Chapman et al. (1997) reported levels reaching 32% of streamflow in peak flow, or 38% in the table presented in Chapman et al. (1993). 3.2. Comparative measurements from elsewhere Field monitoring experiments on pipeflow in a wide variety of environments broadly reveal a similar range in pipeflow contributions, strongly weighted towards the higher end. This evidence comes from Japan (Yasuhara, 1980; Tanaka, 1982; Tsukamoto et al., 1982; Sidle et al., 1995; Terajima et al., 1996, 1997; Uchida et al., 1999; Uchida, 2000; Terajima et al., 2000), from Canada, in the arid badlands of Alberta (Bryan and Harvey, 1985), in southern Quebec (Roberge and Plamondon, 1987), and the subarctic tundra (Woo and dicenzo, 1988; Carey and Woo, 2000), from the Loess Plateau of Shanxi, China (Zhu, 1997; Zhu et al., 2002), and in the Western Ghats in India (Putty and Prasad, 1999). The highest percentage contributions, over 75% of basin runoff, come from small headwaters in the Tama Hills near Tokyo (Yasuhara, 1980; Tsukamoto et al., 1982). In the same area, Tanaka (1982) produced estimates almost identical to Maesnant. The monitoring programmes in China, Canada and India have all produced estimates in the range 20-35%. Again, there is abundant evidence of wide differences between mean and peak contributions, from highs of 76% during snowmelt in Quebec (Roberge and Plamondon, 1987) and 59% in India (Putty and Prasad, 1999) compared with 78% on Maesnant. Other aspects of pipeflow are reviewed in Jones (1990, 1994, 1997b) and Bryan and Jones (1997). 3.3 Conclusions on the significance of pipeflow contributions The important point for this paper is that the majority of monitoring programmes have concluded that pipeflow is a significant contributor to streamflow and that average contributions are commonly in excess of 40%. The response patterns in terms of peak lag times and peak runoff rates per unit of drainage area also tend to fall in-between saturation overland flow and matrix throughflow (Jones, 1997a, b, c). Pipes are by no means present in all basins and even where pipes are present they may not flow in all storms. Also, the large contrasts in both yields and regimes between adjacent pipes complicates the task of obtaining representative samples and of extrapolating the results from basin to basin (cp. Jones, 1997c). Nevertheless, the fact that nearly 30% of the land area of Britain is susceptible to piping (Jones et al., 1997) does suggest that current physically-based models of basin hydrology that are used in upland catchments and do not attempt to model pipeflow are ignoring a potentially important process. 4. Effects of pipeflow on streamwater quality At the last Celtic Hydrology conference, Colin Neal gave us an excellent analysis of the inscrutability of hydrochemistry in the CEH Plynlimon catchments (Neal, 2000). Among the major conclusions of his review, he found (1) that storm rainfall tends not to pass directly through the system as quickflow (as indicated by the dampening of the chloride signal as it passes from rainfall to runoff), (2) that this implies that there 5

is a major body of groundwater that contributes significant amounts of mainly neutral to alkaline water to baseflow in catchments previously considered impermeable, (3) that this contribution comes predominantly via fissure flow within the bedrock, (4) that soil water is the main source of acid waters and this contribution increases in storm flow, (5) that both soil water and groundwater exhibit large variations in chemistry that can overlap, and (6) that chemical equilibrium is rarely reached and that there are large spatial and temporal fluctuations in water quality, which complicate the selection of an appropriate scale for monitoring and modelling and make explanation difficult, if not impossible, to pin down in a detailed mechanistic sense. Neal (2000) was not explicitly referring to soil piping, but it is clear that pipeflow acts within the soil like fissure flow in the bedrock, and can be an important source of spatial and temporal heterogeneity in water quality. Studies of the hydrochemistry of pipeflow, streamflow, rainfall and soil moisture in the Maesnant basin largely corroborate Neal s conclusions (e.g. Jones and Hyett, 1987; Hyett, 1990; Richardson, 1992; Jones, 1997b). Water quality is highly variable, but it is possible to make a number of generalisations. 4.1. Acidity The most significant effect of piping appears to be in the acidification of surface streams. Piping reduces the buffering of acid rainfall by reducing residence times and by directing flow through the upper organic horizons, reducing contact with weathering mineral surfaces (Jones and Hyett, 1987; Gee and Stoner, 1989). It may also encourage the release of sulphates and organic acids from the peaty horizons by draining and aerating sections of the hillside (Jones, 1997b). Flows in the ephemeral pipes are more acid than the rainfall. Hyett (1990) reported an average ph of 4.8 for the rainfall and mean phs of 3.8 to 4.3 in the ephemeral pipes. These pipes typically flow at a depth of 150 mm near the base of the peat O horizon and, like those described in the Nant Gerig basin by Gilman and Newson (1980), are principally formed by desiccation cracking in the peat. The remnants of these cracks offer surface inlets allow rapid infiltration of rainwater into the pipe network. Although recent modelling simulations suggest that the storm response in these ephemeral pipes can best be modelled by assuming that pipeflow is initiated by a rising phreatic surface (Jones and Connelly, 2002), the chemistry suggests that they derive a significant amount of acidity from the thin peaty cover. These apparently conflicting pieces of evidence can be resolved by the fact that pipe response is relatively rapid. During average stormflows lasting 25.5 h, the mean lag time between start of rain and start of pipeflow at the outfalls of the ephemeral pipes is 10.5 h, but they reach their peak flow at about the same time as the rainfall peak and cease flow 8 h before the end of the rain (Jones, 1988). Peak lag time at the head of the ephemerals is slightly longer at 4 h. These statistics suggest that residence times are short and that pipeflow is initiated by rapidly infiltrating rainfall that raises the phreatic surface above the pipe beds within a few hours. Yields at the outfalls of the ephemeral pipes are also more acid than the discharges at the outfalls of the perennial pipes (Table 1), because the perennial pipes derive a certain amount of water from resurgent groundwater. Acidity tends to increase downstream along the perennial pipes (Figure 3), partly because the ph is raised to around 5.0 at the head of the perennial pipes by the resurgence of deep groundwater (Table1). This groundwater is the source of most baseflow in the pipes 6

and has passed through the Ordovician bedrock, albeit a weathering source low in base cations. The pipes subsequently collect water from effluent seepage along the pipe walls and from whole tributaries that originate within the blanket peat. These perennial pipes flow below a peat cover of around 500 mm and although they tend to run along the peat/drift interface, they derive relatively little discharge from the clayey drift. The overall pattern along the pipes tends to remain the same during baseflow (Figure 4), but there are considerable differences between storms in the level of acidity and concentration of solutes. Consistent patterns tend to disappear during stormflow as water arrives from different sources. There is also considerable variation in solute concentrations during individual stormflow events. Much of this can be explained by variations in the quality of water entering the pipes from the surface, either from rainfall or from washing off of dry deposition. The evidence collected by Hyett (1990) suggests that this variation in the quality of surface water inputs severely limits the ability to distinguish separate sources of pipe waters on the basis of chemistry. Jones and Crane (unpub.) measured the highest concentration of solutes based on the electrical conductivity of pipeflow during the first major rainstorms in September, following the summer dry period. Monitoring at other times also seemed to show a flushing effect during the initial stages of stormflow response, followed by lower solute concentrations at peak and during recession, which may also suggest the removal of a limited supply of weathering products. Overall, the pipes provide a major source of stream acidity even under baseflow conditions, but during storms the ph of pipeflow drops further. Despite this, the average stream acidity remains lower than the mean rainfall acidity by only a small margin. The basin therefore displays a small acid neutralising capacity, which is due to the amount of groundwater entering the stream directly from the riparian zone. The purest contributions from groundwater come from riparian seepage zones, which contribute around 36% of stream quickflow in an average 30 mm storm. Most of these zones yield diffuse seepage, predominantly from bedrock in the upper half of the monitored reach. Some are also fed by ephemeral pipes, which would lower the ph during the stormflows when these pipes are active; approximately 1 in every 3 storms that yield storm runoff from the perennial pipes. 4.2. Aluminium concentrations The pipes are also a significant source of aluminium in the streamwater. Table 1 shows that the ephemeral pipes tend to display the highest mean levels of aluminium. This is partly due to the fact that these pipes only flow under storm conditions. It is also partly due to the fact that these pipes run wholly within the peaty surface horizon where Richardson, 1992) and colleagues measured the highest levels of aluminium within the soil. Indeed, both total aluminium and the levels of the more toxic monomeric form are highest in the ephemeral pipes. Nevertheless, Table 2 shows that even the perennial pipes can be a significant source of aluminium for streamwater, with higher concentrations than either matrix throughflow or overland flow. Levels of monomeric aluminium in the perennial pipes also frequently exceed the toxic threshold for fish (Figure 4). Although Hyett (1990) found wide variations in aluminium concentrations between pipes, and frequently a poor relationship with ph, Figures 4 and 5 clearly show a reasonable overall correlation. Figure 5 also shows how a small storm 7

following a larger storm can result in a disproportionately high response once the hillslope system is thoroughly wetted. 4.3. Overall effects on streamwater quality Pipes are a particularly important source of dirty water, i.e., brown-stained water in the stream, which occurs after heavy rains and especially during the first rains of autumn following a dry summer. Hyett found the following average ranking in the effects of pipeflow on streamwater quality: colour > ph > conductivity > aluminium. Pipeflow is not always the single most important control on streamwater quality, but it is when the volumetric contributions from pipeflow are greatest, e.g. 50-70% of streamflow. 4.4 The old or new water controversy Conflicting views have been expressed as to whether pipe quickflow is predominantly derived from the current storm or from previous storms. Sklash et al. (1996) report what seems to be overwhelming evidence from hydrogen isotope analysis that the ephemeral pipes in the Upper Wye are draining predominantly old water from previous storms. The fact that the Maesnant pipes respond to short-term, within storm variations in rainfall quality and that a significant proportion of acidity is derived from the peaty surfaces and the deeper peat layers suggests otherwise: they suggest short residence times and limited contact with the mineral layers (Jones, 1997b). Hyett (1990) concluded that the main controls on Maesnant pipewater quality are primarily the amount, intensity and quality of storm rainfall, plus antecedent soil moisture. Short residence times are also supported by the purely hydrological observations of (1) the large proportion (averaging 68%) of total storm rainfall in the catchment that seems to appear as stream quickflow (Jones, 1997b), (2) the high levels of volumetric flow contributions to stream quickflow (averaging 49%) from the pipes, (3) the close correlation between flow patterns and rainfall patterns shown throughout the ephemeral network (Figure 6) and (4) the successful modelling of pipeflow assuming only surface and shallow groundwater sources (Jones and Connelly, 2002). Sklash et al. (1996) suggested that the rapid response of their ephemeral pipes might be explained by piston flow. Piston flow might well be an element in the hydrological response. However, the hydrochemical evidence collected by Hyett (1990) seems to marry well with the purely hydrometric data and together they point to a substantial proportion of quickflow being provided by new water from the current storm on Maesnant in both the ephemeral and perennial pipes. Baseflow in the perennial pipes is, of course, a different case. 5. Effects of piping on the hillslope environment The Maesnant pipes have a number of important effects on the hillslopes as well as the stream water. Surveys initiated by Richardson (1992) have shown effects on soil profile development, the distribution and diversity of moorland plant communities and the lateral redistribution of plant nutrients. Measurements of electrical conductivity in the top 150 mm of soil suggest that the ephemeral pipes have a marked effect on the distribution of plant nutrients. The highest levels of electrical conductivity in topsoil extracts were found below the 8

outfalls of the ephemeral pipes, where the pipes issue onto the surface and create a zone of overland flow during heavy storms before re-entering the pipe network at the head of the perennial pipes. This is counterbalanced by a zone of low electrical conductivity around the head of the ephemeral pipes, which suggests that the pipes are pathways for a nutrient depletion and enrichment process running down the hillslope, depleting the nutrient status of the surface soil over the main upslope area of piping and enriching the area downslope of the pipe resurgences (Richardson, 1992; Jones, 1997b). The pipes also drain and aerate sections of hillslope. This affects the plant communities and soil development, so that linear patterns are found in both following the lines of the pipe networks down the hillside, as illustrated in Figure 7 (Jones et al., 1991; Richardson, 1992; Jones, 1997b). Drainage and aeration around the pipes accelerates the decomposition of the peat, creating oligo-amorphous peat soils around the perennial pipes leaving the deep peat on the micro-interfluves. These bands of oligo-amorphous peat along the perennial pipe are up to 500 mm lower than the deep peat and mixed grassland heath interfluves between the pipes (Jones, 1997b). The direction of causality is perhaps less clearcut with the ephemeral pipes, which are associated with areas of stagno-podsol within a sea of blanket peat: was the peat always thinner there and desiccation cracking and pipe development has caused a positive feedback? Attempts to improve the upland grazing by ploughing and reseeding in the Upper Wye basin have resulted in waterlogging through destruction of this natural pipe drainage (Gilman and Newson, 1980). Vegetationally, both types of pipe create belts of dry grassland associations surrounded by mixed heath of heather and bilberry (Figure 7). 6. Implications for catchment management Sensitive management of piped catchments should recognise both the hydrological and the ecological role of pipeflow. This may well involve value judgements, since the pipes have both positive and negative effects on their environment. 6.1. Recognising the extent of the stormflow contributing area Accurate delimitation of the dynamic contributing area can be important for many purposes. It may be needed to determine the optimal area for liming in order to neutralise runoff (cf. section 6.2). It could also help in planning afforestation (cf. section 6.3). The Maesnant pipes carry the contributing area well beyond what would be recognised by topographic indices such as a/s (area drained per unit contour length divided by slope angle) used in TOPMODEL (Jones, 1986). They approximately double the quickflow contributing area. Moreover, this extension occurs in an irregular way rather than a simple expansion of the riparian contributing area. The longest network of combined ephemeral and perennial pipes extends 750 m from the stream. The longest single ephemeral pipe feeding directly into the stream extends some 300 m away from the stream. When the ephemeral pipes switch into the flow network, they can link remote contributing areas to the stream that may be regarded as disjunct from the stream if looked at simply from the point of view of overland connections. 9

Overland flow plays a relatively small part in quickflow generation. Sets of crest stage gauges installed in the most likely places to develop overland flow recorded very little, and visual observations confirm the general lack of overland flow contributions. This is despite the fact that surveys of infiltration capacities across the basin suggest that overland flow should in fact occur over large areas of the Maesnant basin. Comparing the frequency distribution of rainfall intensities in the basin with infiltration measurements made with a standard double-ring infiltrometer and a Guelph permeameter suggests that overland flow should occur on 75% of hillslopes in 1 in every 2 rainstorms. Nothing like this occurs in practice. This is because the standard methods of measuring infiltration capacity ignore the cracks and the blowholes and inlets in the roofs of the pipes, which have near infinite capacity to drain depression storage and overland flow. However, pipes do have a limited capacity to transmit flow and some clearly reached that capacity during monitoring, whereas overland flow only increases in efficiency as flow increases. 6.2. Controlling acidification and stream aluminium levels The pipes are a significant source of acid waters and aluminium for the stream, especially during stormflow. Gee (pers.comm., 2002) has admitted that one of failings in the Welsh Acid Waters Programme was lack of understanding of the role of pipes and identifying the sources of streamflow solely on the basis of the a/s index in TOPMODEL (cp. Edwards et al., 1990). Jones (1986; 1997a) found only limited correlation between pipe location and discharge and a/s. With this in mind, liming of pipeflow source areas, as identified by Jones and Connelly (2002), should be more cost-efficient than liming whole catchments and more effective than simply relying on traditional source areas. It is worth noting that basins are likely to be more sensitive to acidification where more ephemeral pipes feed the stream directly, as opposed to issuing onto the hillside or into perennial pipes. 6.3. Implications for forestry, biodiversity and moorland management As illustrated in section 5, piping has developed its own environment of moorland soils and plant associations on Maesnant, adding to the biodiversity and developing an increasingly marked micro-topography towards the lower end of the network. The pipes also have implications for afforestation. Conifer afforestation has been an important accelerator for stream acidification in Wales (Edwards et al., 1990; Neal, 1997). Planting coniferous trees on piped areas of hillslope is likely to aggravate the effect by speeding runoff, even without prior ploughing and ditching. This suggests that piped areas should be mapped as no-go zones for planting. Less information is available on the effects of deciduous trees. However, although the effects may be less due to the reduced interception of acid aerosols during the leafless period and to the lower release of organic acids from the trees, they are still likely to be efficient interceptors during foliage. This clearly has implications for the newly proposed tree planting schemes. Alternatively, if piping is not to be conserved, then the most acid pipes could be switched out of the drainage system, for example, by blocking and diverting to overland flow, which is a possible solution for the dirty water problem as well. 7. Conclusions 10

Pipeflow can be a substantial contributor to catchment quickflow in upland moorland basins in Britain, as demonstrated in the results of monitoring by a number of research teams. On Maesnant, pipes are equally important as sources of baseflow. Pipes increase acidification of surface waters, especially where ephemeral pipes issue directly into stream courses. Yet they also encourage diversity in plant communities and soils within moorland habitats. Landuse planning and management strategies should be sensitive to the possible role of piping processes within moorland basins and not rely solely upon delimiting the surface drainage networks and surface topography. Acknowledgments I would like to thank former research assistants and students Francis Crane, Glyn Hyett, Mark Richardson and Liam Connelly. We would also like to thank the Natural Environment Research Council (Grants GR/3683, GR3/6792 and studentship) and the University of Wales (studentship) for supporting this research. References Bryan, R.B. and Harvey, L.E., 1985. Observations on the geomorphic significance of tunnel erosion in a semiarid ephemeral drainage system. Geographiska Annaler, series A, 67: 257-273. Bryan, R. B. and Jones, J. A. A., 1997. The significance of soil piping processes: inventory and prospect. Geomorphology, 20(3-4): 209-218. Burt, T.P., Heathwaite, A.L. and Labadz, J.C., 1990. Runoff production in peat-covered catchment. IN Anderson, M.G. and Burt, T.P. (eds) Process studies in hillslope hydrology, Wiley, Chichester: 463-500. Carey, S.K. and Woo, M-K., 2000. The role of soil pipes as a slope runoff mechanism, Subarctic Yukon. Canada. J. Hydrol., 233: 206-222. Chapman, P.J. 1994. Hydrochemical processes influencing episodic stream water chemistry in a small headwater catchment, Plynlimon, Mid-Wales. Unpub. PhD thesis, University of London, 416pp. Chapman, P.J., Reynolds, B. and Wheater, H.S., 1993. Hydrochemical change along stormflow pathways in a small moorland headwater catchment in Mid-Wales, U.K. J. Hydrol., 151: 241-265. Chapman, P.J., Reynolds, B. and Wheater, H.S., 1997. Sources and controls of calcium and magnesium in storm runoff: the role of groundwater and ion exchange reactions along water pathways. Hydrology and Earth Systems Science, 1(3): 671-685. 11

Connelly, L. J., 1993. Modelling stormflow in natural subsurface pipes. Unpub. PhD thesis, University of Wales, Aberystwyth, U.K., 419pp. Dunne, T., 1978. Field studies of hillslope processes. In Kirkby, M. J. (ed.) Hillslope hydrology, Wiley, Chichester: 227-94. Edwards, R., Gee, A. and Stoner, J., 1990. Acid waters in Wales. Monographie Biologie 66, Kluwer Academic Publishers, Dordrecht, 337pp. Gardiner, A. T., 1983. Runoff and erosional processes in a peat-moorland catchment. Unpub. M.Phil. thesis, Huddersfield Polytechnic, Huddersfield, U.K.. Gee, A.S. and Stoner, J.H., 1989. A review of the causes and effects of acidification of surface waters in Wales and potential mitigating techniques. Archives of Environmental Contamination and Toxicology 18, 121-30. Gilman, K. and Newson, M. D., 1980. Soil pipes and pipeflow - a hydrological study in upland Wales. Geobooks, Norwich, 114 pp. Holden, C. and Burt, T.P., in press. Piping and pipeflow in a deep peat catchment. Catena. Hyett, G.A., 1990. The effect of accelerated throughflow on the water yield chemistry under polluted rainfall. Unpub. Ph.D. thesis, University of Wales, Aberystwyth. Jones, J.A.A., 1971. Soil piping and stream channel initiation. Water Resour. Res., 7(3): 602-610. Jones, J. A. A., 1978. Soil pipe networks - distribution and discharge. Cambria, 5: 1-21. Jones, J.A.A., 1981. The nature of soil piping: a review of research. GeoBooks, Norwich, 301 pp. Jones, J.A.A., 1986. Some limitations to the a/s index for predicting basin-wide patterns of soil water drainage. Z. Geomorph. Suppl. Bd., 60: 7-20. Jones, J. A. A., 1987. The effects of soil piping on contributing areas and erosion patterns, Earth Surf. Proc. and Landf., 12(3): 229-48. Jones, J. A. A., 1988. Modelling pipeflow contributions to stream runoff, Hydrol. Proc., 2: 1-17. Jones, J. A. A., 1990. Piping effects in humid lands. In Higgins, C.G. and Coates, D.R. (eds) Groundwater geomorphology: the role of subsurface water in earth-surface processes and landforms, Geological Society of America, Boulder, Special Paper, 252: 111-38. Jones, J.A.A., 1994. Soil piping and its hydrogeomorphic function. Cuaternario y Geomorfologia, 8(3-4): 77-102. Jones, J. A. A., 1997a. Pipeflow contributing areas and runoff response. Hydrol. Proc., 11(1): 35-42. 12

Jones, J. A. A., 1997b. Subsurface flow and subsurface erosion: further evidence on forms and controls. In Stoddart, D. R. (ed.) Process and form in geomorphology, Routledge, London and New York: 74-120. Jones, J. A. A., 1997c. The role of natural pipeflow in dynamic contributing areas and hillslope erosion: extrapolating from the Maesnant data. Physics and Chemistry of the Earth, 22(3-4): 303-308. Jones, J. A. A. and Connelly, L. J. 2002. A semi-distributed simulation model for natural pipeflow. J. Hydrol. Jones, J. A. A. and Crane, F. G., 1984. Pipeflow and pipe erosion in the Maesnant experimental catchment. In Burt, T.P. and Walling, D.E. (eds) Catchment experiments in fluvial geomorphology, GeoBooks, Norwich: 55-72. Jones, J.A.A. and Hyett, G.A., 1987. The effect of natural pipeflow solutes on the quality of upland streamwater in Wales. Abstracts, 19th General Assembly of the International Union of Geodesy and Geophysics, Vancouver, 3, 998. Jones, J. A. A., Richardson, J. M. and Jacob, H., 1997. Factors controlling the distribution of piping in Britain: a reconnaissance survey. Geomorphology, 20(3-4): 289-306. Jones, J. A. A., Wathern, P., Connelly, L. J. and Richardson, J. M., 1991. Modelling flow in natural soil pipes and its impact on plant ecology in mountain wetlands. In Nachtnebel, P. (ed.) Hydrological basis of ecologically sound management of soil and groundwater, International Association of Hydrological Sciences Publication, 202: 131-42. McCaig, M., 1983. Contributions to storm quickflow in a small headwater catchment: the role of natural pipes and soil macropores, Earth Surf. Proc. and Landf., 8(3): 239-52. McCaig, M., 1984. The pattern of wash erosion around an upland streamhead. In Burt, T.P. and Walling, D.E. (eds) Catchment experiments in fluvial geomorphology, GeoBooks, Norwich: 87-114. Neal, C. 1997. Introduction to the special issue of Hydrology and Earth System Sciences, the water quality of the Plynlimon catchments. Hydrology and Earth System Sciences 1(3), 385-388. Neal, C. 2000. Catchment science in a heterogeneous world: reflections on the water quality functioning of the Plynlimon catchments, mid-wales. In J. A. A. Jones, Kevin Gilman, Alain Jigorel and John Griffin (eds) Water in the Celtic World: managing resources for the 21 st century. British Hydrological Society Occasional Paper No. 11, 189-196. 13

Putty, M.R.Y. and Prasad, R., 2000. Runoff processes in headwater catchments - an experimental study in Western Ghats, South India. J. Hydrol., 235: 63-71. Richardson, J.M., 1992. Catchment characteristics and the distribution of natural soil piping. Unpub. MPhil dissertation, University of Wales, Aberystwyth. Roberge, J. and Plamondon, A. P., 1987. Snowmelt runoff pathways in a boreal forest hillslope, the role of pipe throughflow. J. Hydrol., 95: 39-54. Sidle, R.C., Kitahara, H., Terajima, T. and Nakai, Y., 1995. Experimental studies on the effects of pipeflow on throughflow partitioning. J. Hydrol., 165(1-4): 207-219. Sklash, M. G., Beven, K. J., Gilman, K. and Darling, W. G., 1996. Isotope studies of pipeflow at Plynlimon, Wales. Hydrol. Proc., 10(7): 921-944. Tanaka, T., 1982. The role of subsurface water exfiltration in soil erosion processes. International Association of Scientific Hydrology Publication, 137: 73-80. Terajima, T., Kitahara, H., Sakamoto, T., Nakai, Y. and Kitamura, K., 1996. Pipe flow significance on subsurface discharge from the valley head of a small watershed. J. Jpn. For. Soc., 78: 20-28. (English abstract.) Terajima, T., Sakamoto, T., Nakai, Y. and Kitamura, K., 1997. Suspended sediment discharge in subsurface flow from the head hollow of a small forested watershed, Northern Japan. Earth Surf. Proc. and Landf., 22: 987-1000. Terajima, T., Sakamoto, T. and Shirai, T., 2000. Morphology and flow state in soil pipes developed in forested hillslopes underlain by a Quaternary sand-gravel formation. Hydrol. Proc., 14: 713-726. Tir Coed, 2001. Woodland, water and flooding. Expert workshop held at the National Assembly for Wales, 5 th July, unpub. Report of Proceedings, New Woodlands for Wales. Tsukamoto, Y., Ohta, T. and Nogushi, H., 1982. Hydrological and geomorphological studies of debris slides on forested hillslopes in Japan. Int. Ass. Hydrol. Sci. Publ., 137: 89-98. Uchida, T. 2000. Effects of pipeflow on storm runoff generation processes at forested headwater catchments. Unpub. Ph.D. thesis, Kyoto University, Japan, 118pp. Uchida, T., Kosugi, K. and Mizuyama, T., 1999. Runoff characteristics of pipeflow and effects of pipeflow on rainfall-runoff phenomena in a mountainous watershed. J. Hydrol., 222(1-4): 18-36. 14

Walsh, R. P. D. and Howells, K. A., 1988. Soil pipes and their role in runoff generation and chemical denudation in a humid tropical catchment in Dominica. Earth Surf. Proc. and Landf., 13(1): 176-202. Whipkey, R. Z. and Kirkby, M. J., 1978. Flow within the soil. In M.J. KIRKBY (ed.) Hillslope hydrology, Chichester: Wiley, 121-44. Woo, M.-K. and dicenzo, P., 1988. Pipe flow in James Bay coastal wetlands. Can. J. Earth Sci., 25: 625-629. Yasuhara, M., 1980. Streamflow generation in a small forested watershed. Unpub. MSc thesis, University of Tsukuba, 55pp. Zhu, T.X., 1997. Deep-seated, complex tunnel systems - a hydrological study in a semi-arid catchment, Loess Plateau, China. In Jones, J. A. A. and Bryan, R. B. (eds) Piping Erosion, Special Issue, Geomorphology, 20 (3-4): 255-267. Zhu, T.X., Luk, S.H. and Cai, Q.G., 2002. Tunnel erosion and sediment production in the hilly loess region, North China. J. Hydrol., 257(1-4): 78-90. 15

Table 1 Mean water quality parameters for rainfall, streamflow and pipeflow sites on Maesnant Rainfall Stream Perennial pipe outfalls Heads of perennial pipes Ephemeral pipe outfalls Pipe mean (n=63) pipe 2 pipe 3 pipe 4 pipe 5 pipe 9 pipe 13 pipe 14 ph 4.84 5.16 4.90 4.48 4.52 5.20 5.50 4.26 4.10 4.58 conductivity μs cm -1 at 20 o C dissolved aluminium mg l -1 dissolved organic carbon mg l -1 39.5 34.1 35.8 41.9 40.0 39.9 33.4 57.4 47.0 41.5-0.211 0.162 0.238 0.295 0.104 0.208 0.524 0.37 0.271-2.69 4.11 4.95 3.19 1.60 1.40 2.20 15.6 3.81 Pipe reference numbers as in Figure 2. 16

Table 2 Contemporaneous water quality samples from 6 drainage flowpaths, Maesnant, during storm runoff Determinand Stream 2 Pipes 3 4 Matrix Overland flow ph 5.5 5.3 4.7 4.7 4.3 4.3 Conductivity μs cm -1 20C 32.0 32.0 34.0 38.0 39.0 36.0 Total hardness 4.7 4.4 3.7 4.9 2.4 2.0 mg l -1 CaCO 3 Chloride 5.0 6.0 6.0 7.0 6.0 5.0 mg l -1 Dissolved silicate 1.5 1.8 1.9 1.3 0.3 0.3 mg l -1 Dissolved 0.18 0.19 0.16 0.19 0.83 0.35 potassium mg l -1 Dissolved 0.077 0.118 0.167 0.257 0.109 0.122 aluminium mg l -1 Dissolved organic 1.7 3.6 3.3 5.4 10.6 8.8 carbon mg l -1 17

Figure 1. Distribution of piped catchments in Britain in relation to soil type. Winter Rainfall Acceptance Potential (WRAP) class 5 soils have the lowest infiltration capacity, class 2 have moderately high throughflow potential. 18

Figure 2. The main area of piping in the Maesnant catchment, showing gauging and sampling sites. 19

Figure 3. Trends in water chemistry down pipe 2 under baseflow conditions. Weir 15 is at the outfall of the ephemeral pipe section and W5 the head of the perennially-flowing section. See Figure 2 for location of these weirs. 20

Figure 4. Trend in acidity and aluminium concentrations down pipe 4 Figure 5. Rainfall, pipeflow and pipewater quality in a sequence of storms. Note the rise in aluminium levels during storms and the larger response to the second storm when the hillslope system is wetter. 21

Figure 6. Close correlation between rainfall and pipeflow response in two ephemeral pipes (pipes 14 and 15, Figure 2). 22

Figure 7. The pattern of major plant associations around piping on Maesnant. 23

Application of sequential optimisation for flood control Nysa Reservoir System case study TOMASZ DYSARZ Gdansk Technical University, ul. Narutowicza 11/12, 80-952 Gdańsk, Poland todys@pg.gda.pl JAROSLAW J. NAPIÓRKOWSKI Institute of Geophysics, Polish Academy of Sciences, ul. Ksiecia Janusza 64, 01-452 Warszawa, Poland jnn@igf.edu.pl Abstract We present an application of the controlled random search method for real time operation of the Nysa Klodzka reservoir system in Poland. To improve efficiency and accuracy of the used optimisation technique we tested a sequential optimisation and suggest a particular modification of the standard controlled random search. We show that the introduced concept considerably improves the performance of the control structure by reducing the dimensionality of the sub-problems. This allows to handle the optimisation problems that seemed to be unsolvable due to the so called course of dimensionality. Resume Une application de la méthode de la recherche aléatoire commandé pour commander en temps réel d un système des réservoirs de la rivière Nysa Klodzka en Pologne est présentée. Pour améliorer l eficacité de la téchnique d optimisation appliqué, l optimisation séquentielle était testé et une modification de la méthods classic de la recherche aléatoire est proposée. Il est montré que cette amélioration changé remarquable l eficacite, de la méthode de commande par la réduction du nombre des dimensions de sou-problèmes. Ce permettait résoudre le problème d optimisation qu il parait inrésoluble à cause de l imprécation de dimensions. INTRODUCTION Decision Support System for flood control for the Nysa Klodzka Reservoir System includes modules responsible for precipitation forecast for a catchment, rainfall-runoff transformation, unsteady flow routing for Nysa Klodzka and selected reach of Odra River as well as operational control. The main goal of the paper is to present a control structure and control mechanisms for the cascade of Otmuchow and Nysa reservoirs - problems related to the last module. The catchment of the Nysa Klodzka River as shown in Fig.1 is located in the southern part of Poland.

Fig. 1 Catchment of Nysa Klodzka River (south-west Poland) The hydrological features of the upper part of this catchment are: massive rocky underground covered only by small layer and an average yearly precipitation of about 900 mm. An inability of storing water underground leads to dangerous floods. To handle this problem two reservoirs were built, and two more are under construction. The paper discusses the management of two existing reservoirs governing the discharges in the city of Nysa, which lies just below the second reservoir. We propose the new control algorithm that makes use of characteristic features of the system and global optimisation methods. Relations resulting from the system dynamic equations allow to perform calculations separately for each particular reservoir in the cascade and to propagate the results to other system components. This decomposition makes the computational costs to depend linearly on the number of reservoirs. In this paper we apply the global optimisation technique elaborated by Price (1983), later developed by Ali and Storey (1994) (Controlled Random Search method), and finally modified by the authors. PROBLEM FORMULATION The considered system that consists of two reservoirs in series is schematically shown in Fig.2. Management and control of flooding generally require the use of forecasting techniques. At this stage we assume that inflows I ( t ) and I ( t) 1 2 to the system represent one of many possible scenarios taken into account by a decision maker. The scenarios considered could be based on rainfall-runoff prediction models, or recorded historical data. Otmuchów reservoir Nysa reservoir storage V 1 (t) storage V 2 (t) Brzeg I 1 (t) u 1 (t) u 2 (t) Q(t) Nysa Klodzka I 2 (t) I 3 (t)

Odra Fig.2 Schematic representation of the system Retention in each reservoir V j ( t) is described by the dynamics of a simple tank, with I t and one controlled output u j ( t), j = 1, 2 one forecasted inflow () j dv dt 1 = I () t u () t (2.1) 1 1 dv dt 2 = u () t + I () t u () t (2.2) 1 2 2 The following constraints on the reservoir storage and releases (given in Table 1) are taken into account: V1( 0) = V1 0 V 2( 0) = V 20 (initial condition) V min j [ 0 H V j () t Vmax U ( ) j min u j j t U max j (2.3) for j = 1, 2 and for any t,t ], where V min denotes dead storage, V max denotes total storage, and T is optimisation time horizon. H Table 1. System parameters Reservoir V min [mln m 3 ] V max [mln m 3 ] U min [m 3 /s] U max [m 3 /s] Upper (no 1) 19,38 124,66 0,0 1363,0 Lower (no 2) 20,29 113,60 0,0 1960,0 To simplify the optimisation problem the dynamics of flow in the reach between the reservoirs is omitted and flood routing in Nysa River below Nysa Reservoir is described by means of so called linear channel (pure delay) with time constant T 0, so the flow at Nysa Klodzka outlet Q(t) is ( ) ( ) Q t u t T = 2 0 The main goal of this system is the protection of the user located below the cascade of reservoirs against flooding by minimizing the peak of the superposition of waves Q ( t ) + I ( t 3 ) on Nysa and Odra rivers, respectively. This can be achieved by desynchronization of the flow peaks via accelerating or retarding flood wave on Nysa River. The second objective is storing water for future needs after flood. Hence the objective function of the optimisation problem under consideration can be written in the form of a penalty function: (2.4) 2 2 min β1 max ( Q() t + I3() t ) + β2 Vj( TH ) V max u1, u j 2 t [ 0, TH ] (2.5) j= 1

where symbols β 1 and β 2 denote appropriate weighting coefficients and optimisation time horizon. SEQUENTIAL OPTIMISATION T H is the In this section we describe the application of the particular optimisation procedure for ( k 1) two reservoirs in series. Let us assume that for k-th iteration step the control u1 = uˆ1 and the retention ˆ ( k 1) V1 = V1 of the of upper reservoir are specified for any t [ 0,T H ]. Then one ( k ) ( k ) has to determine the control value u 2 = u2 and retention value V 2 = V2 of the lower reservoir. The optimisation problem for lower reservoir takes form: under constraints min ( ) ( ) 2 ( k ) ( k ) { β1 max Q + I3 + β 2 V2 TH V max2 [ 0, ] } H u2 t T (3.1) ( k ) 2 dv dt ( k 1) ( ) uˆ1 I2 u = + 2 k (3.2) V min 2 ( k ) ( ) ( k ) ( ) Q t = u t T 0 (3.3) 2 ( k ) ( k ) V2 () t Vmax U ( ) 2 min u 2 2 t U max 2 (3.4) ( ) ( ) After solving optimisation problem (3.1), i.e. after solving for u2 and Vˆ k 2, the control and retention of the upper reservoir are modified so that to improve the primary objective function ( ) (2.5), while maintaining Vˆ k 2 ; all modifications of the control function u1 are directly transferred to the control function u 2 that describes the outflow from the reservoir system. The optimisation problem for the upper reservoir takes form: under direct constraints min ( k ) u1 ˆ k 2 ( k1 ) ( k ) { β1 max ( Q + I3) + β 2 V1 ( TH ) V max1 t [ 0, T ] } H (3.5) ( k ) 1 dv dt ( k ) 1 u1 = I (3.6) V min1 and indirect constraints resulting from eq.(2.2) ( k ) ( k ) V1 () t Vmax U ( ) 1 min u 1 1 t U max1 u ( k1 ) ( k ) ( k ) ( k 1) ˆ ˆ 2 = u2 + u1 u1 ( k ) ( ) 1 1 2 ( k ) ( ) (3.7) (3.8) Q t = u t T 0 (3.9) U min 2 ( k ) 2 ( t) max 2 1 u U (3.10) ( 2 where u k 1 ) is improved outflow from the lower reservoir.

The solutions of the optimisation problem (3.5) (3.10) are the optimal trajectories of ( ) ( ) u = u, V = Vˆ k ( ) and u = uˆ k1. Therefore the solutions of the primary optimisation ˆ k 1 1 1 1 2 2 ( ) problem at k -th iteration step are these three functions and the V 2 = Vˆ k 2 trajectory determined at the previous stage. After solving the problems for lower and upper reservoirs, i.e. completing k-th step of sequential optimisation, one can go to the next step, k + 1, once more solving the optimisation problem (3.1) (3.4) using the values obtained at the step k. Calculations terminate when the stopping rule is met. In our case the chosen criterion is the difference ε between outflows from the system at first and second stage for the current iteration step k. T H ( k 2 1) ( k) uˆ ˆ 2 u 2 dt ε (3.11) 0 The essential problem related to the described algorithm is the selection of initial values ( k= 0) approximation of u and ˆ ( k= 0) ˆ1 V1. Among different options two cases can be intuitively justified. In the first case one can assume a constant retention of the upper reservoir: ˆ ( k= 0) 1 = dv dt 0 ( 0) u k = 1 = I1 ˆ (3.12) It means that at the first stage of the next step, all inflows to the system must pass through the lower reservoir. This requirements can negatively affect the performance of the algorithm due to constraints (3.4). In some cases the better initial approximation is given by: ˆ ( k = 0) ( k= 0) dv1 uˆ 1 = 0 = I1 dt (3.13) In the majority of cases, the above formula does not guarantee that initial approximation meets the constraints(2.3) imposed on retention of upper reservoir. However, violation of the constraints resulting from (3.13) will be corrected at the second stage. CONTROL RANDOM SEARCH METHOD () The functions u j t, j = 1, 2 were represented by a train of rectangular pulses and the time horizon was divided into L unequal time intervals. The parameters to be determined were values of pulses û l and time instances of switching the control function u(t). This type discretisation, denoted as TD-RP (Time Dependent Rectangular Pulses) was described in detail by Dysarz and Napiórkowski (2002). The optimisation problems for the lower reservoir (3.1) and the upper reservoir (3.5) were solved by means of the global random search procedure, namely the following version of Controlled Random Search (CRS2) described in details in Dysarz and Napiórkowski (2002). The CRS2 algorithm starts from the creation of the set of points, many more than n + 1 points in n -dimensional space, selected randomly from the domain. Let us denote it as S. After evaluating the objective function for each of the points, the best x L (i.e. that of the minimal value of the performance index) and the worst x H (i.e., that of the maximal value of the performance index) points are determined and a simplex in n -space is formed with the best point x L and n points ( x2,..., x n + 1) randomly chosen from S. Afterwards, the centroid x G of points, x,, x is determined. The next trial point x is calculated, x 2x x. x L 2 n Q Q = G n+1

Then, if the last derived point x is admissible and better (i.e., Qx ( ) Qx ( )), it replaces Q the worst point x H in the set S. Otherwise, a new simplex is formed randomly and so on. If the stop criterion is not satisfied, the next iteration is performed. In the CRS2 version applied in the tests, the worst point of the current simplex will be the reflected point x = 2x x, rather than the arbitrary chosen one (Dysarz and Napiórkowski, 2002). RESULTS OF TEST FOR HISTORICAL DATA The described sequential optimisation was tested and verified on a number of historical and synthetic flood events. Results for two of them, namely for the historical floods in Nysa catchment in 1965 and 1997, are presented in Fig.4 and Fig.5, respectively. The floods in 1997 were caused by the most disastrous recent abundance of water in the region. During the first stage of a disaster, a rapid increase in runoff was noted after intense and long lasting rains in the 4-10 July period in the highland tributaries. Yet, a few days later, from 15 to 23 July, another series of intensive rains occurred. The highest precipitation in the Klodzko valley reached 100-200 mm. The flood virtually ruined the town of Klodzko (Kundzewicz et al., 1999), and the historic stage record was exceeded by 70 cm. During the 1985 flood, daily precipitation maxima were significantly (two to three times) lower than in 1997. Several all-time maximum stages recorded in 1985 were largely exceeded by the 1997 flood. Fig.4a and 5a show the performance of the Otmuchow (upper) Reservoir, Fig.4b and Fig.5b show the performance of Nysa (lower) Reservoir, and Fig.4c and 5c show the flow at the cross-section below the junction of Nysa and Odra Rivers. As one can see, by an appropriate choice of the control functions the peaks of the waves on Nysa Klodzka and Odra rivers were desynchronised and the culminations did not overlap. Q H Q G H

discharge [m 3 /s] 900 750 600 450 300 150 inflow outflow maximum admissible storage storage minimum admissible storage 120 90 60 30 storage [mln m 3 ] 0 1 61 121 181 241 301 361 421 481 541 601 661 721 781 time [h] Fig. 4a Performance of Otmuchów reservoir data from 1965 0 discharge [m 3 /s] 900 750 600 450 300 150 maximum admissible storage storage inflow outflow minimum admissible storage 120 90 60 30 storage [mln m 3 ] 0 1 61 121 181 241 301 361 421 481 541 601 661 721 781 time [h] 0 Fig. 4b Performance of Nysa reservoir data from 1965 discharge [m 3 /s] 1200 900 600 300 0 Odra River flow outflow from reservoirs superposition of waves without retention Nysa River flow obtained superposition of waves 0 100 200 300 400 500 600 700 800 time [h] Fig. 4c Flow below the junction of Nysa and Odra Rivers 1965 data

discharge [m 3 /s] 2000 1500 1000 500 inflow outflow storage maximum admissible storage minimum admissible storage 120 90 60 30 storage [mln m 3 ] 0 1 101 201 301 401 501 601 701 801 901 1001 1101 1201 time [h] 0 Fig. 5a Performance of Otmuchów reservoir data from 1997 2000 discharge [m 3 /s] 1500 1000 500 storage inflow maximum admissible storage outflow minimum admissible storage 120 90 60 30 storage [mln m 3 ] 0 1 101 201 301 401 501 601 701 801 901 1001 1101 1201 time [h] 0 Fig. 5b Performance of Nysa reservoir data from 1997 3200 superposition of waves without retention discharge [m 3 /s] 2400 1600 800 outflow from reservoirs obtained superposition of waves Odra River flow Nysa River flow 0 0 200 400 600 800 1000 1200 time [h] Fig. 5c Flow below the junction of Nysa and Odra Rivers 1997 data

CONCLUSIONS It is necessary to take into account the uncertainty of the inflows forecast in operation control of reservoirs system during flood. Hence the optimisation problem has to be solved repetitively for many scenarios using actual measurements and updated forecasts. Therefore, from the decision making point of view, the access to a quick and reliable, especially designed for the particular system optimisation module, is very important. The approach presented in the paper makes a decomposition of the general problem possible, so that computational costs grow linearly with the number of reservoirs. Hence, more complex representation, than that described by Niewiadomska-Szynkiewicz et al. (1996) and Niewiadomska-Szynkiewicz and Napiórkowski (1998), of the control functions u j ( t) can be adopted. Because of nondifferentiability of global and two local performance indices, the global optimisation technique CRS is used. The authors have not proved the convergence of the proposed method yet, however convergence was observed in all carried out tests. The results from applications of the sequential optimisation by means of control random search methods to determine the reservoir decision rules during flooding are encouraging. Accuracy of the proposed method is satisfactory. The initiation procedure and the stop criterion were cautiously investigated, so high efficiency does not cause losses in accuracy. As a result, the described control structure of Nysa Kłodzka reservoirs system can be easly extended to include transformation by means of hydrodynamic flood routing model, because the proposed technique guarantees that the solution of the optimisation problem can be obtained in reasonable time. Acknowledgement: This work was partially supported by Polish Committee for Scientific Research under grant 6 P04D 032 19 REFERENCES Ali, M.M., & Storey C. (1994): Modified controlled random search algorithms, Intern. J. Computer Math., Vol. 53, pp. 229-235 Dysarz T., & Napiórkowski J.J. (2002): Determination Of Reservoir Decision Rules During Flood, Acta Geophysica Polonica, Vol. 50, No. 1, 135-149. Kundzewicz, Z.W, Szamalek, K, & Kowalczak, P. (1999) The great Flood of 1997 in Poland. Hydrol. Sci. J. 44(6), 855-870. Niewiadomska-Szynkiewicz, E; Malinowski, K.; & Karbowski, A. (1996). Predictive Methods for Real-Time Control of flood operation of a Multireservoir System: Methodology and Comparative Study, Water Resources Research, Vol. 32, No. 9, 2885-2895. Niewiadomska-Szynkiewicz E., & Napiórkowski J.J., 1998. Application of Global Optimization Methods to Control of Multireservoir Systems. Proc. Advances in Hydro-Science and Engineering, Cottbus/Berlin Germany. http://www.bauinf.tucottbus.de/iche98/proceedings Price, W.L (1987): Global optimization algorithms for CAD Workstation, J. Optimiz. Theory Appl., Vol. 55, No.1, pp. 133-146.

MULTIFRACTAL MODELING OF THE BLAVET RIVER DISCHARGES AT GUERLEDAN MODELADUR MULTIFRAKTEL SKORVOÙ AR BLAVEZH E GWERLEDAN MODELISATION MULTIFRACTALE DES DEBITS DU BLAVET A GUERLEDAN P. Hubert (1), I. Tchiguirinskaia (2), H. Bendjoudi (3), D. Schertzer (4), S. Lovejoy (5) (1) UMR Sisyphe, CIG, Ecole des Mines, Paris, hubert@cig.ensmp.fr (2) UMR Sisyphe, LGA, Université Paris VI, tchigin@biogeodis.jussieu.fr (3) UMR Sisyphe, LGA, Université Paris VI, Hocine.Bendjoudi@ccr.jussieu.fr (4) Météo-France and LMM, Université Paris VI, schertze@ccr.jussieu.fr (5) Physics Department, McGill University, Montreal, lovejoy@physics.mcgill.ca Diverrañ Mont en-dro kas ar stêr zo anavezet eveit bezañ bet muzuliet abaoe hanter-kant bloavezh da vihanañ [Hurst, 1951]. Nevez zo eo bet kavet eo multifraktel rummadoù skorvoù ar stêr e meur a vuzuliadenn etre div sizhun ha seizh miz da vihanañ [Tessier et al., 1996; Pandey et al., 1998; Labat et al., 2002]. Diouzh ar muzulioù-se e voe priziet ar parametroù multifraktel hollvedel (H, C 1, α) [Schertzer and Lovejoy, 1987] kenkoulz hag ezponant war var (q D ) al lezenn a wana gwirheñvelder an dasparzh. An nevesañ eo ivez urzh diforc'hañ ar prantadoù stadegel, o klotañ gant urzh multifraktel kentañ ur prantad treuziñ [Schertzer and Lovejoy, 1992] ha gant dont war-wel strukturioù war var emaozet [Schertzer et al., 1993, Chigirinskaya et al., 1994]. Ar parametroù-se a hañval bezañ sichennet en tu all d'ar muzulioù, met ivez bezañ ur seurt ment diazad bras (eus un nebeud km² betek milionoù a km²). Er studiadenn-mañ e tielfennomp an dastumadoù pemdeziek graet etre 1939 ha 1999 diwar skorvoù ar Blavezh e Gwerledan. Evidomp ez eo q D = 3 urzh war var al lezenn a wana gwirheñvelder an dasparzh. Diouzh un tu all, diwar un dielfennañ spektrel o tiskouez o deus skorvoù ar stêr un ingalded kreñv diouzh ar c'houlzoù, e arguzennomp diwar-benn kemer kement-mañ e kont. Evel just, kementmañ a dorr amzer (stadegel) digemmusted an treuzkas, ma talc'her kont anezhi en dielfennañ multifraktel boas. Hervez an hent-se e arguzennomp penaos ober gant ur modeladur multifraktel ingal hag eeun-tre [Tchiguirinskaia, 2002] diazezet war lammoù stokastek lieskementiñ gant unanoù adaozet. Dont a ra dimp diouzh un tu muioc'h a briziadurioù eus ar parametr hollvedel ha diskouez e adkrou ar modeladur kas ur stêr e- keñver koulzoù muzuliet, met ivez e tegas kemmoù e reizhiadoù muzuliañ evit amzer hiroc'h. Abstract River flow phenomena have been known to be scaling for at least fifty years [Hurst, 1951]. More recently, river discharge series have been found to be multifractal over a

range of scales spanning at least from 2 weeks to 7 months [Tessier et al., 1996; Pandey et al., 1998; Labat et al., 2002]. On this range of scales the universal multifractal parameters ( H,C 1,α ) [Schertzer and Lovejoy, 1987] were estimated as well as the critical exponent (q D ) of the power-law fall-off of the probability distribution. The latter is also the order of the divergence of the statistical moments, which corresponds to a first order multifractal phase transition [Schertzer and Lovejoy, 1992] and the appearance of selforganized critical structures [Schertzer et al., 1993, Chigirinskaya et al., 1994]. These parameter estimates seem not only to be robust over this range of time scales, but also over a wide range of basin sizes (from a few km 2 up to millions of km 2 ). In the present study, we analyse daily records from 1939 to 1999 of the Blavet river discharges at Guerledan (Brittany). We estimate as q D =3 the critical order of the powerlaw fall-off of the probability distribution. On the other hand, since a spectral analysis shows that the river discharges have a strong seasonal periodicity, we discuss how to take it into account. Indeed, it breaks the (statistical) time translation invariance, which is implicitly assumed in the usual multifractal analysis. In this perspective, we discuss the applicability of a rather simple seasonal periodic multifractal model [Tchiguirinskaia et al., 2002] based on stochastic multiplicative cascades with re-ordered singularities. We obtain on the one hand more reliable estimates of the universal parameter and show that this model adequately reproduces a river flow not only at seasonal time scales, but also changes in scaling regimes for longer time scales. Résumé Le caractère scalant du débit des rivières est reconnu depuis au moins une cinquantaine d années [Hurst, 1951]. Plus récemment, on a pu montrer que les séries de débits étaient multifractales sur une gamme d échelles allant au moins de 2 semaines à 7 mois [Tessier et al., 1996; Pandey et al., 1998; Labat et al., 2002]. Sur cette intervalle, les paramètres multifractals universels ( H,C 1,α ) [Schertzer and Lovejoy, 1987] de même que l exposant critique (q D ) de décroissance algébrique de la distribution de probabilité ont pu être estimés. Ce dernier paramètre est aussi l ordre de divergence des moments statistiques, qui correspond à une transition de phase multifractale de première espèce [Schertzer and Lovejoy, 1992] et à l apparition de structures auto-organisées [Schertzer et al., 1993, Chigirinskaya et al., 1994]. Ces paramètres semblent non seulement robustes sur la gamme d échelles sur laquelle ils ont été estimés, mais se retrouvent aussi sur une large gamme de tailles de bassins (de quelques km 2 à plusieurs millions de km 2 ). Dans la présente étude, nous avons analysé les débits journaliers du Blavet à Guerledan (Côtes d Armor) de 1939 à 1999. Nous avons estimé à q =3 D l exposant critique ( q ) de D décroissance algébrique de la distribution de probabilité. Par ailleurs, comme l analyse spectrale montre que les débits présentent une forte périodicité saisonnière, nous avons cherché à en tenir compte. Cette périodicité rompt en effet l invariance (statistique) d échelle au cours du temps, hypothèse implicite des analyses multifractales habituelles. Dans cette perspective, nous avons proposé l application un modèle multifractal périodique simple [Tchiguirinskaia et al., 2002] basé sur des cascades stochastiques multiplicatives dont on ordonne les singularités. Nous obtenons ainsi une estimation plus fiable des paramètres universels et nous avons pu montrer que ce modèle reproduit correctement le débit de la rivière à toutes les échelles de temps et rend compte du changement de régime de la scalance observé pour les grandes échelles de temps.

a) b) Figure 1a,b: 52 years time series (01/01/1948-31/12/1999) of daily discharges at Guerledan (a) and annual maxima of these daily discharges (b). Figure 2: Spectra (from top to bottom) of annual maxima of daily discharges and of daily discharges themselves (log-log plot). Spectral analysis of Blavet discharges at Guerledan The scaling range may be assessed with the help of classical spectral analysis in determining the frequency bands where the spectrum displays a power-law. When studying the spectra of thirty French rivers, [Tessier et al., 1996] observed the existence of a scale break that appears roughly between 16 and 30 days. The ensemble averaged spectral exponent was estimated as β=1.3 for the 1 to 16 days regime and β=0.52 for the 1 month to 30 years regime of river runoff. The authors pointed out that the period associated with the break was not significantly correlated neither with the size of basin nor with the geology. Therefore, they argued that this 16-day period was associated to the atmospheric synoptic maximum [Kolesnikova and Monin, 1965], which is the typical lifetime of planetary scale atmospheric structures. As a consequence, the multifractal analysis was performed over two distinct frequency bands, respectively the high frequencies of 1-16 days using daily data and the low frequencies of 1-360 months using monthly averages of daily records. [Pandey et al., 1998] obtained similar results for daily river flow data from 19 river basins of varying watershed areas in the continental USA. For most of the rivers, the authors observed a break in the scaling regime, which was associated to half of the atmospheric synoptic maximum. The spectral exponent for low frequency region was estimated as β=0.72 and the multifractal analysis was only performed for the time-scales longer than 8 days. Although the basin areas varied over nearly six orders of magnitude, the scaling results were independent of the basin size and geology. In a recent study [Labat et al., 2002] of

three karstic springs located in the French Pyrenées Mountains, a change of scaling behavior was noted on the spectra and was also attributed to the synoptic maximum. In spite of it, a unique multifractal analysis was performed over about 11 years of daily records that displays a convincing unique multifractal regime over the range from 1 to 512 days. It is important to note that no upper time scale (in particular of 512 days) was found for the lower frequency scaling regime months [Tessier et al., 1996; Pandey et al., 1998; Labat et al., 2002]. On the other hand, the influence of annual cycle was disregarded in all three studies. Figure 3: A log-log plot of the DTM as a function of the scale ratio λ for q = 1. 5 and various values of η. The straight lines indicate scaling of moments of filtered daily discharges over time-scales from 1 to 512 days. Figure 4: A log-log plot of empirical scaling function K(q,η) vs. η for (from top to bottom) q =2.0; 1.5 and 0.8. The straight lines have a unique slope which corresponds to α=1.62. For each q, the intersection of such curve with the axis Logη=0 gives the corresponding LogK(q). Then the universal multifractal expression for K(q) is used to compute C1=0.14. The time series of daily discharges of Blavet river at Guerledan as well as of their annual maxima are displayed on Fig.1 for the period from 1939 to 1999. It is notable that this Celtic river visually exhibits much stronger intermittent spikes than those observed on other river discharges. Furthermore, the intermittency of the annual maxima and that of the full time-series are rather the same. The corresponding spectra are presented in Fig.2. The spectral exponent of daily discharges may be estimated as β=1.3 for time-scales shorter than a year. Since spectral analysis decomposes the statistics according to frequencies, the seasonal periodicity of Blavet discharges corresponds to a prominent annual spectral spike. Before this annual spike, the spectral slope is rather flat and the spectrum exponent β is close to 0.3. The annual maxima have the same spectral exponent. In contrast to months [Tessier et al., 1996; Pandey et al., 1998; Labat et al., 2002], we do not observe any peculiarity of the spectrum behavior on time-scales of the order of two weeks. In the following section we will proceed to a more involved scaling analysis of daily discharges keeping in mind three main remarks based on the spectral results: since no unique power-low exponent was found for the entire spectrum, we may face similar changes in scaling regimes within multifractal analysis; the annual spike may pose significant problems for a multifractal analysis performed in the physical space, whereas it does not in the Fourier space; since β=1.3 for high frequencies, the discharge statistics can not be directly in agreement with the statistics of a multifractal field produced by a multiplicative

cascade (which yields β<1), therefore a spectral filtering will be required to meet these two exponents. The last remark is not only relevant for the simulation of multifractal discharges, but also for an accurate empirical estimation of multifractal parameters. Figure 5: A log-log plot of the TM as a function of the scale ratio λ for different moments q. The straight lines indicate scale invariance of filtered daily discharges over time-scales from 1 to 512 days. Figure 6: Empirical scaling moment function K(q) of daily discharges (open dots) superposed to the theoretical one (α=1.62, C1=0.14, continuous curve). These curves are undistinguishable up to the critical order q D =3. How many parameters account for multifractals? Multifractals constitute a very convenient framework to analyze and simulate intermittent field (for a review [Schertzer et al., 1997]). In general, a multifractal field is obtained by a cascade process, whose paradigm can be traced back to the famous the Richardson's poem [Richardson, 1922]. An elementary process is repeated scale by scale, randomly transmitting a fraction of a given flux (e.g. energy flux for fluid turbulence) from a parent structure to its children structures. In the simplest case, the scale ratio λ = L/ l (L is the n external scale of the cascade, l the scale of observation) takes discrete values: λ = λ 1 (λ 1 is the scale ratio for the discrete elementary step of the cascade, usually λ 1 = 2 ). The resulting flux, n F = = n ( x) μ λ 2 i i= 1, rather corresponds to the multiplication of identically independently distributed (i.i.d.) random variables, μ (e.g. [Monin and Yaglom, 1975]). i As λ, the flux F λ is becoming no longer a point-wise function but a multi-singular measure. It means that the non-trivial flux limit F F has values defined on any arbitrary (small) neighborhood of a point (x), but not on this precise point. So, this flux depends on the scale of the point neighborhood. Usually, the flux over a given set should be conserved at all scales (e.g. its ensemble average should be strictly scale invariant in the case of a 'canonical conservation'). The probability distribution of the flux spikes may be defined in a scaling manner with the help of the co-dimension function c (γ ) of the singularitiesγ 's: γ c( γ ) Pr( Fλ λ ) λ (1) It is rather straightforward that the co-dimension function c(γ ) is a non-increasing function. This merely corresponds to the fact that high intensity events are less frequent than low intensity events, and therefore the most intense regions occupy a smaller

fraction of the probability space. A multifractal field corresponds to an infinite hierarchy of singularitiesγ 's and corresponding codimensions c (γ ). Therefore, its definition in general requires an infinite number of parameters. Hence, let us emphasize the possibility of having universal multifractals that are defined with the help of few robust and relevant parameters [Schertzer and Lovejoy, 1987]. The general idea of universality is that interactions between rather similar processes may lead them to converge to some attracting process. This fact has enormous consequences: only a few relevant parameters may define a stochastic process, whereas it could result from very complex interactions a priori requiring numerous parameters. It has been shown that by adding more and more intermediate levels in cascades or by multiplying i.i.d. cascades, one converges to a universal multifractal process [Schertzer and Lovejoy, 1997]. Figure 8: A log-log plot of the DTM as a function of the scale ratio λ for q = 1.5 and different values of η. The straight lines indicate scale invariance of moments of non-filtered annual maxima of daily discharges over the full range of time-scales. Figure 7: Probability that a discharge exceeds a given threshold for (from top to bottom) annual maxima of daily discharges and for daily discharges themselves (log-log plot). One may fit tails of these probabilities by straight lines having slope 3, which confirms q D =3 of Fig. 6. For the universal multifractals, only three parameters H, α and C 1 are of the fundamental importance. The first parameter ( H 0) indicates how differ the experimental data from a conservative flux. Only conservative fluxes can be directly modeled with the help of a multiplicative cascade. The spectral slope β=2h+1 gives a rough estimate of H=0.15 for Blavet discharges. As we already mentioned it, when β>1 a special filtering process of the data is required as a preliminary step of the multifractal analysis. Therefore, the timeseries of discharges must be first passed through a filter that weighs their Fourier H components by k, with k being a wave number. Next section illustrates the multifractal techniques that accurately estimate on the filtered data the Lévy parameter of multifractality α (0 α 2) and the parameter C 1 (> 0), which corresponds to the codimension of the mean singularity.

Universal multifractal analysis of Blavet discharges Two mean techniques of multifractal analysis correspond respectively to the estimation of trace moments (TM, [Schertzer and Lovejoy, 1987]) and double trace moments (DTM, [Lavallée et al., 1992]). The main idea of DTM is to compute the η -th power of the data (at their highest resolution) and to degrade this field at smaller and smaller resolution λ. The simple trace moment (TM) merely corresponds to η=1. As explained below DTM (with η 1) is rather indispensable to perform a direct parameter estimation. Then one may study the scaling behavior of the statistical moments of order q of the resulting field ε λ (η) : < (ε λ (η) ) q > λ K(q,η) (2) Figure 3 displays the scaling behavior of DTM curves, i.e. Log < ( ε λ ) > vs. Log(λ), for Blavet filtered discharges over the time-scales from 1 to about 512 days. Corresponding slopes determine the scaling function of the double trace moments. In the case of universal multifractals this scaling function K ( q, η) has simple relation with the single moment scaling function ( K(q) K(q,1)): K(q,η) = η α K(q) (3) Correspondingly, Fig. 4 displays the estimated K(q,η) (obtained with the help of Fig. 3) vs.η in logarithmic coordinates for three different orders q. On Fig. 4, the common slope of the curves yields an estimate of the parameter α, whereas the intersections of these curves with axis Logη=0 yield an estimate of the corresponding K(q). Then the parameter C 1 may be computed according to the following equation for the universal multifractal scaling function [Schertzer and Lovejoy, 1987]: K(q) + Hq = C 1 (q α q)/(α 1) (4) Applying this technique to Blavet data, we obtain the following fits:α = 1.62, C 1 = 0.14. It is now interesting to evaluate the quality of these estimates by comparing the empirical TM curve with the theoretical one defined by the empirical estimates of the universal parameters. Figure 5 displays the scaling behavior of TM for Blavet filtered discharges. Parameter α Parameter C 1 Parameter q D Tessier et al., 1996 1.45±0.25 0.2 ±0.1 2.7± 1.0 (<16 days) 3.2 ±1.5 (>16 days) Pandey et al., 1998 1.65± 0.12 0.13± 0.05 3.37± 0.86 Labat et al., 2002 0.78± 0.1 0.83± 0.1 0.76± 0.1 0.26± 0.1 0.27± 0.1 0.26± 0.1 ( η) 4.44± 0.2 5.0± 0.3 4.95± 0.62 Present study 1.62± 0.15 0.14± 0.03 3.0± 0.5 Table 1: Multifractal parameter estimates obtained on daily discharges in earlier studies For the range of time-scales from 1 to about 512 days one may determine the scaling function K(q) for different orders q. Figure 6 displays comparison between the empirical single moment scaling function that corresponds to the results of Fig.5 and the theoretical one that is obtained with the help of Eq.4 with DTM parameter estimates: q = 1. 5

α = 1.62, C 1 = 0.14. These scaling functions are in a perfect agreement up to the critical moment q D 3, which might corresponds to a multifractal phase transition. The latter is confirmed with the help of Fig. 7 that displays the probability of exceeding a threshold for annual maxima of daily discharges and monthly discharge data themselves. The asymptotic power-law for both series yield a common exponent estimate q D 3. Let us remind that an important consequence of this exponent is that the probability of having -qd river discharge 10 times larger than given large discharge will be only 10 times smaller, which is in sharp contrast with estimates based on exponential fall-off [Hubert et al., 2001]. Figure 9: Example of S-shaped DTM curves of non-filtered daily discharges. Straight lines indicate scales for which multifractality was confirmed by all former studies. Figure 10: Example of S-shaped DTM curves of synthetic discharges obtained by re-ordering multipliers at two consecutive steps of the cascade (α=1.6, C1=0.14). Results of the present study, as well as those of previous studies of daily discharges, are displayed in Table 1 for comparison. It is notable that our present estimate q D 3 is on the lower bound of earlier results and significantly smaller than the estimateq D 6 obtained for monthly discharges with the help of 1502 Russian gauges [Tchiguirinskaia et al., 2002]. This is in agreement with the visual observation, which we discussed above on Fig.1, that Blavet river intermittency is rather strong. More precisely, the frequency of large events for the Blavet river decreases only as the square root of the frequency decrease of Russian rivers. Influence of seasonal periodicity on a multifractal runoff The fact that daily discharges and corresponding annual maxima exhibit the same critical order q D 3 gives some credence that the observed seasonal periodicity (discussed above in relation to Fig.2) does not introduce a scale break, although there is no straight line fits for the whole range of time scales on Fig.3. The quality of the scaling for periods of time larger than a year can be checked on Fig. 8 that displays straight DTM curves of annual maxima. It is important to note that these annual maxima were not filtered before DTM analysis (since β<1). In comparison to Figs. 3, 8, Fig.9 S-shaped DTM curves of the original, i.e. non-filtered, daily discharges. Similar S-shaped curves for DTM curves were systematically observed for monthly discharges in Russia [Tchiguirinskaia et al., 2002], whereas it had been overlooked in earlier multifractal river flow studies months [Tessier et al., 1996; Pandey et al., 1998; Labat et al., 2002]. The generality of this phenomenon for discharge data called for a generalization of the cascade processes.

Indeed, to model river runoff with a seasonal periodicity by multiplicative cascades we may start with a usual cascade process down to the year-scale. Then, on the next level of this cascade (i.e. scale of six months) we re-order random multipliers [Tchiguirinskaia et al., 2002] in order that among the two next random variables the largest one will be used first. Therefore, the first six-month singularity will be larger than the second sixmonth singularity. It is important to note that this re-ordering will not modify the essential statistics, but breaks the time translation invariance cascade. If this reordering process is repeated for the next levels of cascade process, then we obtain a stronger periodicity. Once the cascade was developed up to a sufficiently small scale, one may upscale the resulting multifractal field up to the scale of interest. Summation over many realizations of such multifractal field could be interpreted as summing over various basin contributions and corresponds to a simplified multifractal runoff. The corresponding synthetic multifractal discharge has both the seasonal periodicity and S- shaped DTM curves (Fig. 10). Conclusions We proceeded to a detailed multifractal analysis of the Blavet river discharges at Guerledan (Brittany). We discussed in details the possible influences of the synoptic maximum and of the annual cycle with respect to scaling. We showed that high frequencies (time-scales shorter than a year) and low frequencies (time-scales larger than a year) of Blavet discharges have a common critical order of divergence of moments q D 3, which quantifies how much Blavet has a higher intermittency than other rivers. We also showed that high and low frequencies of suitably filtered discharges have common multifractal parameters α 1.6,C 1 0.14. Furthermore, we showed that the difference of scaling behaviours for high and low frequencies of the non-filtered discharges are related to seasonal periodicity. Indeed, we show that a simple model, based on singularity re-ordering, yields both of the seasonal periodicity and the S-shaped DTM curves, i.e. an apparent non unique scaling behaviour for high and low frequencies. Acknowledegments Discharges data of the Blavet River for the 1948-1999 are coming from the HYDRO data base and we sincerely thank Mr Scherer for providing us with these data. Discharges data for the 1939-1947 period have been collected from the hydrological yearbooks then published by the Société hydrotechnique de France. Thank you to Hervé Le Bihan for the translation of the abstract into Breton. References Chigirinskaya, Y., D. Schertzer, S. Lovejoy, A. Lazarev, and A. Ordanovich (1994) Multifractal Analysis of Tropical Turbulence, part I: horizontal scaling and self organized criticality, Nonlinear Processes in Geophysics, 1 (2/3), 105-114. Hubert, P., H. Bendjoudi, D. Schertzer, and S. Lovejoy (2002) Multifractal Taming of Extreme Hydrometeorological Events, in The Extremes of the Extremes, IAHS publication n 271, 51-56. Hurst, H.E. (1951) Long-term storage capacity of reservoirs, Transactions of the American Society of Civil Engineers, 116, 770-808.

Kolesnikova, V.N., and A.S. Monin (1965) Spectra of meteorological field fluctuations, Izvestiya, Atmospheric and Oceanic Physics, 1, 653-669. Labat, D., A. Mangin, and R. Ababou (2002) Rainfoll-runoff relations for karstic springs: multifractal analyses, J. Hydrology, 256, 176-195. Lavallée, D., S. Lovejoy, D. Schertzer, and F. Schmitt (1992) On the determination of universal multifractal parameters in turbulence, in Topological aspects of the dynamics of fluids and plasmas, edited by K. Moffat, M. Tabor, and G. Zalslavsky, Kluwer, 463-478. Monin, A.S., and A.M. Yaglom (1975) Statistical Fluid Mechanics, MIT press, Boston Ma. Pandey, G., S. Lovejoy, and D. Schertzer (1998) Multifractal Analysis Including Extremes of Daily River Flow Series for Basins one to a million square kilomemeters, J. Hydrology, 208, 62-81. Richardson, L.F. (1922) Weather prediction by numerical process, Cambridge University Press republished by Dover, 1965. Schertzer, D., and S. Lovejoy (1987) Physical modeling and Analysis of Rain and Clouds by Anysotropic Scaling of Multiplicative Processes, Journal of Geophysical Research, D 8 (8), 9693-9714. Schertzer, D., and S. Lovejoy (1992) Hard and Soft Multifractal processes, Physica A, 185, 187-194. Schertzer, D., and S. Lovejoy (1997) Universal Multifractals do Exist!, J. Appl. Meteor., 36, 1296-1303. Schertzer, D., S. Lovejoy, and D. Lavallée (1993) Generic Multifractal phase transitions and self-organized criticality, in Cellular Automata: prospects in astronomy and astrophysics, edited by J.M. Perdang, and A. Lejeune, World Scientific, 216-227. Schertzer, D., S. Lovejoy, F. Schmitt, I. Tchigirinskaya, and D. Marsan (1997) Multifractal cascade dynamics and turbulent intermittency, Fractals, 5 (3), 427-471. Tchiguirinskaia, I., P. Hubert, H. Bendjoudi, and D. Schertzer (2002) Multifractal Modeling of River Runoff and Seasonal Periodicity, Preventing and Fighting Hydrological Disasters, Timisoara, Romania, November 2002. Tessier, Y., S. Lovejoy, P. Hubert, D. Schertzer, and S. Pecknold (1996) Multifractal analysis and modeling of Rainfall and river flows and scaling, causal transfer functions, J. Geophy. Res., 31D, 26,427-26,440.

Flood Alleviation Planning in a Virtual Water World De-watering the Lower Feale Catchment A. Creating a Virtual Water World Abstract 1 Martin, J. (1,2), Migliori, L. (1), Smyth, J.A. (3), & O Kane, J. P. (1) (1) Department of Civil & Environmental Engineering, National University of Ireland, Cork, Ireland (2) Institute for Space Sensor Technology & Planetary Exploration, German Aerospace Center, Berlin (3) Office of Public Works, Dublin 2, Ireland In Part A of this paper, we present the creation of the computer models generated during the Lower Feale Catchment Flood Study (UCC, 1997-2002 ), describe the calibration process and demonstrate the agreement achieved between water level measurements and computer predictions. In the next part of the paper, Part B, we present a detailed evaluation of various degrees of engineering intervention for the alleviation of flooding: dredging and pumping. In doing so, we have considered two very different hydrological events. The first event is a series of minor floods, with return periods of roughly two years and less, occurring close together during December and January 1998. The second is a single isolated large flood in March 1998 with a return period of roughly 9 years. Both floods produce roughly the same maximum flooded areas, but the first lies longer on the land. The effectiveness of each engineering alternative is found by comparing its predicted maximum area of flooding with the reference case of no engineering intervention for both hydrological events. The corresponding flooded areas are weighted by their probabilities of exceedance to yield an annual expected or average flooded area for each alternative. The paper shows that pumps provide a very flexible solution to the flooding problem, tuned to each individual polder, but give a rise to a series of questions appropriate to a follow-on study. Some of these are: What energy source should be used to drive the pumps, individually and collectively? Wind, electricity, diesel, peat? What water level should be maintained in each individual polder? What will be the impact on the degraded remnants of raised peat bogs between the Brick and the Feale? Such questions are now being addressed in a follow-up study. 1. Infrastructure - The Arterial Drainage Solution The principle of lowland arterial drainage involves the construction of riverside embankments on tidal rivers. This allows: - Passage of rainfall run-off from higher catchment areas out to the sea through the lowlands, without flooding onto them. - The containment of water levels within the embanked waterway at high spring tides, which are often at a level greater than that of the surrounding lowlands. (Candy, 1938, p.5) Rainfall run-off from the low-lying areas obviously cannot drain directly into these embanked rivers. Networks of interconnected drains, (small, un-embanked channels or ditches) dug in these low-lying areas, collect and store rainfall run-off. Land-drains collect rainfall run-off water from the land, and lead it to drains that lie adjacent to the embanked rivers on both sides, called back-drains. Culverts, with steel sluice gates (see fig. 1) at their downstream end (called sluiced culverts), link these back-drains to the embanked river at regular intervals of 1,000-2,000 metres (see fig. 2). These provide a means of gravitational drainage from the back-drain into the embanked rivers when the water level in the tidal river drops below the drain water level. When high water levels prevail in the embanked river, these sluiced culverts close and rainfall run-off is stored in the backdrains (see fig. 2). Under normal circumstances in Ireland, two periods of drainage occur in each 24-hour period, during low tide. 1 This computer model is called Virtual Water World based on a) content, and b) display. The models database contains all data and all physical relationships necessary to describe the catchments hydrology, hydraulics and hydrodynamics. These are displayed clearly and realistically, as both input and results. The photo-realistic display of the integrated models dynamic floodmaps illustrates the power of these tools to simulate and display real data in a realistic manner earning it the name Virtual Water World.It does not claim to represent a three dimensional, walk-around Immersed Virtual Reality facility.

In some embanked tributaries, sluiced barrages at the mouth minimise the tidal effect. A barrage built across the mouth of the tributary (see fig 3) with large culverts running through it, controlled at their downstream end by wooden sluice gates (see fig. 4), has the following effects: - Allows high capacity discharge from the tributary into the larger river. - Prevents upstream propagation of the tide, reserving the volume in the tributary for rainfall run-off storage. - Under normal circumstances, maintains lower water levels in the tributary for the maximisation of sluiced culvert drainage from back-drains into the barraged tributary. Figure 1 - Steel flap valve used at downstream end of sluiced culverts. This flap gate has a seal of Ertalon Bush, and the edges of the circular gate fit flush into the steel ring surround, which is fixed to the concrete outlet structure. Figure 2 - Cross section of embankment showing sluiced culvert link between back-drain and main embanked river. Figure 3 - The sluiced barrage shown in the centre of this image minimises the upstream propagation of the tide from the tidal river on the right of the image (the Brick) to the tributary on the left of the image (the Crompaun).

The existence of different types of watercourse within the arterial drainage network allows the establishment of a hierarchical drainage network, consisting of drainage channels and drains (or ditches), as follows: - Primary Drainage Channels (1 st degree drainage) all rivers (usually tidal and embanked), which run directly (un-obstructed) to the sea. - Secondary Drainage Channels (2 nd degree drainage) larger tributaries (usually embanked), which drain through gravitational drainage structures (sluiced barrages at their mouth) into primary drainage channels. - Secondary Drains (2 nd degree drainage) Small Back-drains or Land-drains that drain through gravitational drainage structures (sluiced culverts) into primary drainage channels. - Tertiary Drains (3 rd degree drainage) Small Back-drains or Land-drains that drain through gravitational drainage structures (sluiced culverts) into secondary drainage channels (water from these drains must pass through two gravitational discharge structures before meeting the open sea). - Floodplain Drainage or Flooding (4 th degree drainage) Flow of rainfall run-off over the land (or through the surface) into back-drains or land-drains, and flooding of excess drain-water out over the floodplain. Figure 4 Typical Wooden 5 x 4 Hinged Gate use in Sluiced Barrages. The gates are made from Teak, and are hinged horizontally, half way up the gate. Drainage channels (both primary and secondary) drain rainfall run-off from areas outside (upstream of) the floodplain, and drains (back-drains and land-drains) and the floodplain drainage network stores and carries rainfall run-off from within the floodplain. Figure 5 - This schematic diagram shows a typical polder arrangement and illustrates the effects that lowland infrastructure, such as a road network, has on drain networks. Land-drains are interconnected to as great an extent as possible. Each network of land-drains leads rainfall runoff to a back-drain, from where gravitational drainage into primary (or secondary) drainage channels occurs. However, the topography and infrastructural arrangement of the lowlands force the existence of multiple landdrain/ back-drain networks (see fig. 5). All drains in the floodplain cannot be completely interconnected to form

a single network, due to restrictions such as primary/ secondary drainage channels (which are by definition embanked), road embankments etc. It is important here to define the notion of polders. Oosterbaan (Oosterbaan, 1994, p.636) defines polders as areas with both flood protection and drainage systems. Bos et al. (Bos et al., 1994, p.25) describe a polder as a low-lying area surrounded by a dike, in which the water level can be controlled independently of the outside water. This study uses a combination of these definitions. Here, a polder is defined as an area of land that is drained by a hydraulically independent network of back-drains and land drains, linked to a primary or secondary drainage channel by sluiced culverts only. Floodplain controls such as road embankments or embanked rivers force differentiation between drain networks, and prevent overland flow/ flood propagation from one network to another, thus keeping each network hydraulically independent of other networks. When considering flood alleviation measures, it is important to analyse the effect of such solutions in every polder individually, as a reduction in flood water level in one polder will often have no implication on the flood water level in other polders. Table 1 Arterial Drainage Network in the Lower Feale catchment NUMBER LENGTH (Km) STORAGE (km3) MAIN TIDAL CHANNELS & TIDAL TRIBUTARIES 6 49 - NON TIDAL (SLUICED) TRIBUTARIES 6 16 0.55 BACK-DRAINS & LAND DRAINS 45 66 0.53 EMBANKMENTS - 107 - SLUICED CULVERTS 55 1.3 - SLUICED BARRAGES 4 - - WEIRS 4 - - The Lower Feale catchment consists of a full arterial drainage infrastructure. The Lower Feale River and the Cashen Estuary (23km), and its two main tributaries, the Brick River (14km) and the Galey River (7.5km) are tidal. A number of small tributaries/large drains (4.7km), which are also embanked, flow directly into these. Others (15.6km) drain into them through sluiced culverts and sluiced barrages. Waters from the extensive network of back-drains and land-drains (66km in total) flow into this embanked channel network through 55 sluiced culverts. Table 1 above summarises this extensive hydrodynamic network. 2. The Problem Flooding in the Lower Feale Catchment During and after heavy rainfall, the water levels in the primary drainage channels (tidal, embanked rivers) and secondary drainage channels (non-tidal tributaries) stay higher for longer as they drain rainfall runoff from upstream hilly areas of the catchment. As the lowland (floodplain) becomes saturated, the rate of rainfall run-off into the land-drains and back-drains increases and they fill very quickly (see fig 6). Arterial drainage through sluiced culverts is reduced for 2 reasons: (a) Since the water level in the embanked rivers is higher for longer, the period of time for which there is a positive drainage head from the back-drain into the embanked river is shortened. (b) Because the level in the embanked river at low tide does not drop as low during heavy rainfall, the magnitude of the drainage head is reduced, in turn reducing the magnitude of the discharge through the open sluiced culvert. (c) These two effects result in insufficient volumes of drainage from the back-drains into the drainage channels during the opening of each sluiced culvert. This reduces the available storage volume for further runoff in the drain system, so drains fill up and overtop onto the surrounding land, causing flooding (see fig. 7). This is the problem that is addressed primarily in the course of this study. In a PhD thesis, Martin (Martin, 2002) examines the problem of insufficient drainage volumes from back-drains, through sluiced culverts into embanked rivers, in detail its causes, effects and potential solutions.

Figure 6 - This back-drain is full and will shortly flood onto the surrounding field. Figure 7 - This photograph shows the effects of back-drain and land-drain overtopping on the lowlands of the Lower Feale catchment. (Office of Public Works, 1998) Classification of three types of hydrodynamic flow assists in the understanding of how the drainage and flooding processes occur: (a) Tidal Channel Flow: Flow in the primary drainage channels, determined by the tide at their downstream end and rainfall run-off from upstream hilly catchments at their upstream end. (b) Drain Flow (i) Flow in the secondary drainage channels (mouth-barrage, embanked channels) determined by sluiced barrage drainage at their downstream end and by rainfall inflow from upland hilly areas at their upstream end. (ii) flow in secondary/ tertiary drains, determined by sluiced culvert drainage along their length and rainfall inflow along their length. (c) Floodplain Drainage/ Flood Propagation: Flow over (or through) the land into and out of drains. It is typical in the infrastructure described here, and certainly the case in the Lower Feale catchment, that the vast majority of flooding results from land-drain/ back-drain overtopping. Embankment overtopping from main drainage channels contributes to flooding only in very severe events. Theoretically, there is no limit on the

elevation of an embankment drainage performance of an infrastructure is independent of this parameter. If embankment overtopping is a significant contributor to flooding, a simple solution is to build the embankment to a higher level. Land-drain and back-drain spilling are not as simple to solve! 3. The Computer Models The Virtual Water World The Lower Feale catchment flood study uses an integrated computer model to analyse the performance of the hydraulic infrastructure and to predict how potential flood alleviation measures will perform. This computer model contains five parts. Figure 8 Software Integration Model (i) (ii) Feale_NET A hydrodynamic computer model that simulates the flow of water through the channels and hydraulic structures in the catchment. It includes embanked rivers, back-drains, landdrains, sluiced culverts, sluiced barrage storm gates, bridges and overflow spillways. High frequency time-series data of discharge provide the upstream boundary conditions at all upstream boundary points in the model. High frequency time-series of tide-plus-surge water levels provide the downstream boundary conditions at the models downstream point. Feale_RR (Rainfall Runoff) - Simulates the hydrological processes in the catchment. The input data into Feale_RR include daily rainfall averages, monthly evaporation averages & parameters relating to the soil conditions. It predicts the rainfall runoff into each embanked tributary and backdrain/land-drain. The results of the Feale_RR model provide rainfall runoff input to the Feale_NET model.

(iii) (iv) (v) Feale_DEM - a Digital Elevation Model (DEM) of the topography of the Lower Feale catchment with a resolution of 12 metres. It resolves features of the floodplain, which act as flood controls, such as hilly areas, local depressions, road embankments, ditches, buildings etc. Feale_GIS - Joins the results of Feale_NET with Feale_DEM for the automatic generation of dynamic floodmaps. The results of Feale_RR (the hydrological model) and Feale_DEM (the topographical model) provide input into Feale_NET and Feale_GIS respectively. This integrated software system is shown in figure 8. Feale_EVAL The Feale_EVAL model is the analytical component of the model, which uses the results of the Feale_NET model (as time series of water levels) to analyse the existing infrastructure and to compare the results of simulations of the modified infrastructure (such as dredging, pumping, additional sluiced culverts, etc.) with the existing scenario (reference situation). This part of the model deals with the statistics of the model, such as: - the Highest water levels over a given time period in a given polder (the Peak Flood Level ), - the Reduction in Peak Flood Level during a given simulation period as a result of a modification to the model, - The Maximum reduction (or increase) in water level (at any time step) due to modifications to the reference simulation. - In combination with Area Elevation Curves, Feale_EVAL provides quick estimates of maximum inundated area for a given flood event, for a given simulation. Before detailing the particular software packages that we used in the course of this study, it is important to mention other software tools that are available for such applications. The choice of software tools is not a reflection of the research teams opinion of it in relation to other tools, but merely a statement of its suitability for the project in its own right. The following packages could alternatively have been used in the creation of the model: ISIS (Wallingford Software), Hec-Ras (US Corps of Engineers), SOBEK & Delft_FLS (WL Delft Hydraulics) or Flo_2D (J. S. O Brien). Table 2 below summarises the five models that combine to make the Virtual Water World, the software packages that were use to create the models, the operation that each model performs and the data that is required to drive the models. Table 2 Summary of Model Components MODEL SOFTWARE PACKAGE FUNCTION DATA REQUIRED Feale_NET Mike11 (2000b) To compute Hydrodynamics and Water Levels Discharge Data, Water Level Data, Cross-section Data, Sluiced Culvert & Hydraulic Structure Data, Feale_RR Results Feale_RR Mike11 NAM (2000b) To compute magnitude of Rainfall Data, Soil data Feale_DEM Feale_GIS Rainfall Run-off Flux ERMapper5.5, ArcView 3.2a Digital Representation of Topography of the Floodplain Mike11, ArcView 3.2a, Mike11-GIS (2000b) Feale_EVAL Microsoft Excel 97 To integrate Hydrodynamic Water Level Data with Topography to generate Floodmaps Analysis of Feale_NET Model Results to compute benefits of Flood alleviation measures (in terms of reduction in water level and flood inundated area Digital Image Data, Results of Processing of Digital Image Photogrammetry Digital Elevation Model (Feale_DEM), Results of Feale_NET, Data about Floodplain Control Features Results of Feale_NET, Area Elevation Curves 4. Calibrating the Models Establishing Credibility 4.1 Locating Calibration Points There are six calibration points in the model. Figure 9 shows the position of these 6 calibration points. HD-calibration gauges (autographic gauges or automatic data loggers) sample water levels at these points at a frequency of 15 minutes. Two autographic tidal gauges, one inside the mouth of the estuary (Moneycashen) and

one at Ferry Bridge, provide calibration data for the hydrodynamic model within the estuary. Two water level data loggers, one in the main tidal river (Feale Railway Bridge) and one in a large tidal tributary (Brick Main Channel), provide calibration data for the tidal rivers (primary drainage channels). A water level data logger situated in a tributary with a sluiced barrage at its downstream end (Lixnaw Canal) provides calibration data for the non-tidal tributaries (secondary drainage channels). The model representation of the sluiced barrage structures is calibrated using data from this gauge. A water level data logger in a back-drain (Sleveen backdrain) provides calibration data for the network of back-drains. The model representation of the sluiced culverts is calibrated using data from this gauge together with coefficients found by experimental methods. 4.2 Corresponding Time-step Comparisons The Feale_NET model outputs water level data at a frequency of one hour. HD-calibration water level gauge data is sub-sampled to a frequency of 1 hour for all six calibration points. Observed and computed water level data are available for corresponding historical periods at this time interval. For a completely accurate model, the corresponding water level values should be identical. The Residual Error for each time step is the difference between the observed water level and the computed water level at corresponding times. These values can be positive or negative. Average Residual Error (ARE) The Average Residual Error (ARE) at a H-point over the full calibration period (average of all the residual errors) is indicative of the bias of the model, computed by the equation: ARE = (E 1 + E 2 + E 3 + +E n ) / n where E is the error at each time-step, and n is the number of time-steps used for the calibration at that point. ARE shows if the model is computing water levels with a bias that is above or below the observed values in general. The ARE (bias) of the model is below 6cm at all 6 points. Mean Absolute Error (MAE) The Mean Absolute Error (MAE) over the full calibration period (average of the absolute values of the residual error) is indicative of the magnitude of the errors in the model. This is represented by the equation: MAE = ( E 1 + E 2 + E 3 + + E n ) / n where E is the absolute value of the error at each time-step, and n is the number of time-steps used for the calibration at that point. The MAE (average magnitude of the model error) is below 20cm at all 6 points. Figure 9 Calibration Points for Feale_NET Model

Table 3 Results of Calibration at Six Calibration Points. Calibration Point ARE MAE Comparison Period Correlation Coeff. Ferry Bridge: -3 cm 12 cm 01 Jan. 98 27 May 98 0.968 Moneycashen: 3 cm 16.5 cm 01 Jan. 98 31 Dec. 98 0.965 Feale Railway Bridge: -5.5 cm 9.5 cm 31 Mar. 98 31 Dec. 98 0.982 Sleveen Main Channel: 0 cm 15 cm 13 May 98 31 Dec. 98 0.947 Sleveen Back-drain: 4 cm 8 cm 27 Mar. 98 31 Dec. 98 0.928 Lixnaw Channel: 5.5 cm 12.5 cm 19 Mar. 98 23 Dec. 98 0.915 Table 3 shows the values of ARE and MAE for each of the six calibration points. It also gives the calibration period i.e. the period of time from which results from the model are compared to real measured data. Even though the model simulates data for the full year of 1998, calibration data (reliable measured data) is only available from each calibration point for relatively short periods of time. This duration determines the comparison period for each calibration point. Correlation Co-efficient There is a high correlation between the time series of computed water levels and the observed water levels (correlation coefficient over 0.9, see table 3), for periods when there are no known errors in the calibration data. These values are shown for each calibration point in the table above. Graphical Check A plot of both observed water level and computed water level against time illustrates the accuracy of the model graphically. These plots show periods of mismatch. A plot of the residual error against time illustrates the bias of the model, maximum discrepancies and periods of mismatch. This plot lies (for the most part) within the +/- 25cm envelope of the MAE. Calibration data errors, such as scale errors and summertime time fix errors, can explain errors in excess of this envelope. Figure 10 shows the comparison water level time series plots, and the plots of residual error versus time at Ferry Bridge. The period corresponds to the best two weeks of calibration, but is typical of the results for the full calibration period. Physical Parameter Adjustments The calibration process provides estimated values for bed roughness coefficients (Mannings n). They vary, not only spatially, but also in time (Leendertse, 1988, p.28). The model does not account for temporal variation, but does account for spatial variation along the channel. Small changes in Mannings n, for stretches of the estuary, main rivers and tributaries, bring the model-computed values closer to observed values. A global value of 0.045 is applied around the full system, which relates to the roughness of the rough, often heavily grown back-drains and land-drains. Specific values are given to the larger channels. Calibrating The Flood Mapping (Feale_GIS) Model Without continuous video footage of flood events, it is difficult to calibrate the flood-mapping model both spatially and temporally. There are two principle methods of flood mapping model calibration. Calibration by Photography: Comparisons of instantaneous floodmaps with photographs taken at the corresponding time (Fig 11) give a good indication of the accuracy of the model. This method of calibration is based on visual inspection. The remote sensing survey also shows areas of flood inundation, as it was conducted at a time of moderate flooding. Social Calibration: Those who frequently witness flood events on the Lower Feale catchment have verified the models predictions of areas that are inundated, the lengths of time for which they were inundated and the speed of flood propagation (and recession). Small changes to the DEM (such as extra Fault Theme Lines, which allow the definition of narrow features in the model) bring the predictions of the model closer to those observed in the catchment by eyewitnesses. Social calibration results in flood control features such as additional embankments, road causeways and boundary walls to be included in the Feale_GIS model in the form of Fault Theme lines. The example of social calibration (fig. 12,13) illustrates the functionality of Fault Theme Lines clearly. Figures 12 and 13 highlight areas that are protected from flood inundation by such controls, but which are shown to be flooded in the un-altered model. Such areas are subsequently corrected in the Feale_GIS model, as shown in these figures.

Ferry Bridge Water Level (m.od) 5.5 5 4.5 4 3.5 3 2.5 2 15.3.98 17.3.98 19.3.98 21.3.98 23.3.98 25.3.98 27.3.98 29.3.98 Date Ferry Bridge Difference In Water Level (m) 2.25 1.75 1.25 0.75 0.25-0.25 15.3.98 17.3.98 19.3.98 21.3.98 23.3.98 25.3.98 27.3.98 29.3.98-0.75-1.25-1.75-2.25 Figure 10 - Calibration of Feale_NET model at Ferry Bridge (Comparison of simulated versus measured water levels for a two week period). Date Figure 11 Geometrically rectified flood event photography for flood-map model calibration.

Figure 12 In this figure, we see an area of flooding. An elevated roadway runs through this area. Local farmers (and engineers) confirm that the area to the right of the roadway (between the road and the embanked river) does flood, but that the road acts as a flood barrier and protects the area to the left of the road. Figure 13 - By representing the roadway as a fault theme line, we simulate the flood protection of the road embankment, keeping the area to the left of the roadway flood free. 5. Conclusion This paper describes components of the computer model that integrate together to form the digital replica of the Lower Feale catchment. Each piece of the model uses algorithms that accurately represent the reality of the situation. This Virtual Water World is fully credible (i.e. it simulates what happens in nature very closely) because we feed with accurate data and calibrate it in order to ensure its accuracy, as described earlier.

References Bos, M. G., & Boers, Th. M., 1994, Land Drainage: Why and How? Drainage Principles and Applications (Ed.- in-chief: Ritzema, H. P.), ILRI No.16, Wageningen, The Netherlands. Candy, J. P., 1938, Proposed Comprehensive Scheme of Proposed Arterial Drainage No2 Area - South-West, Office of Public Works, Dublin, Ireland. Leendertse, J. J., 1988, A Summary of Experiments with a Model of the Eastern Scheldt, Rand R-3611-NETH, The Netherlands. Martin, J., O'Kane J.P. and Javan, M. 2000. Computer Modelling For Flood Alleviation in the Lower Feale Catchment, Water In The Celtic World: Managing Resources for the 21 st Century, British Hydrological Society 2000. ISBN 0-948540-97-4, Aberystwyth, Wales. Martin, J. and O'Kane, J.P. 2000. Integration of Physically Based Computer Models With High Resolution Digital Elevation Models for the generation of Flood Maps, Water In The Celtic World: Managing Resources for the 21 st Century, British Hydrological Society 2000. ISBN 0-948540-97-4, Aberystwyth, Wales. Martin, J. and O'Kane, J.P. 2000. Development Of A High Resolution Hydrodynamic Flood Mapping Model Using One-Dimensional Hydraulic Models Integrated With High Resolution Digital Elevation Models, GIS-1, Hydroinformatics - 4th International Conference, IAHR ~ AIRH, 2000 (Abstract: P.163, Proceedings: CD-ROM: GIS-1\365.pdf), Iowa, USA. Martin, J., 2002, De-watering the Lower Feale A Virtual Water World, Ph.D. Thesis, Department of Civil & Environmental Engineering, National University of Cork, Ireland. Oosterbahn, R. J., 1994, Agricultural Drainage Criteria, Drainage Principles and Applications (Ed.-in-chief: Ritzema, H. P.), ILRI No.16, Wageningen, The Netherlands.

Rainfall-runoff modelling of two Irish catchments (One Karstic and one non-karstic) Goswami, M. 1, O Connor, K.M. 1 and Shamseldin, A.Y. 2 1 Department of Engineering Hydrology, National University of Ireland, Galway, Galway, Ireland 2 Department of Civil Engineering, The University of Birmingham, Edgbaston, Birmingham B15 2TT, U.K. Abstract Flow simulation studies were carried out for two Irish rivers, namely, the Fergus, at the Balleycorey Bridge gauging site, and the Brosna, at the Ferbane gauging site. The Fergus catchment is predominantly karstic, while the Brosna, although characterised by underlying carboniferous limestone, is considered to be non-karstic. Both the non-parametric (NP) and parametric (P) forms of two black-box-type rainfall-runoff models, namely, the Simple Linear Model (NP-SLM and P-SLM) and the seasonally-based Linear Perturbation Model (NP-LPM and P-LPM), the non-parametric wetness-index-based Linearly Varying Gain Factor Model (LVGFM), the nonlinear black-box Artificial Neural Network (ANN) Model, and the conceptual Soil Moisture Accounting and Routing (SMAR) Model, were used for this study. In addition, three Model-Output Combination Techniques (MOCTs), namely, the Simple Average Method (SAM), the Weighted Average Method (WAM) and the Neural Network Method (NNM), have also been applied for simulating discharge. Within this group of models, there is a considerable range of variation in the degree of complexity of structure and associated parameter parsimony, and correspondingly in the degree of intricacy of the objective function surface used for calibrating the models. A number of performance evaluation criteria have been used to comparatively assess model efficiency. Some limitatations and the uncertainties involved in the application of these models, in the Irish context, are discussed. The software package named the Galway Real-Time River Flow Forecasting System (GFFS), recently developed at the Department of Engineering Hydrology, of the National University of Ireland, Galway, has been used to obtain all the results presented in this study. Keywords : Karstic, Black-box model, Conceptual model, Rainfall-runoff modelling, Simulation, River flow forecasting system, Neural network Modélisation du passage précipitations/ruissellement sur deux bassins versants irlandais ( dont un bassin versant karstique) Résumé Des études de simulation de crues ont été menées sur deux rivières irlandaises, à savoir : la rivière «Fergus», au niveau de la station de jaugeage «Balleycorey Bridge» et la rivière«bosna» à la station de jaugeage de Ferbane. Le bassin versant de la rivière «Fergus» est de type karstique, tandis que celui de la Brosna n`est pas considéré comme tel. Quatre modèles de type «boite noire», à savoir, le «Simple Linear model» (SLM), le «linearly Perturbation Model» (LPM), le «Lineary Varyaing Gain Factor Model» (LVGFM), the «Artificial Neural Network», ainsi que le modèle conceptuel «Soil Moisture Accounting and Routing Model (SMAR), ont été utilisés dans cette étude. Pour les modèles SLM et LPM, les formes paramétriques (P) et non paramétriques (NP) ont été utilisées. En plus de cela, trois techniques (MOCTs) de combinaison des résultats de ces modèles ont été mises en œuvre pour estimer les débits. Ces techniques sont «the Simple Average Method» (SAM), «the weighted Average Method» (WAM) et le «Neural Network Method» (NNM). Entre ces modèles, il existe une assez grande variation du degré de complexité de leur stucture, du nombre de paramètres à utiliser, et également une grande variation du degré de complexité de la fonction objectif surface, utilisée pour l`étalonnage du modèle. Des critères d`évaluation de performance ont été utilisés pour juger de l`efficacité des modèles. Les conditions d`utilisation, et les incertitudes, liées à l`utilisation de ces modèles en Irlande sont commentées. Le programme nommé «Galway Real-Time River Flow Forcasting System (GFFS), récemment mis au point par le Département d`hydrologie, de la «Nationnal University of Galway», a été utilisé pour l`obtention de tous les résultats présentés dans l`étude. Mots clef : karstique, modèle boite noire, modèle conceptuel, modélisation de la transformation précipitationruissellement, simulation, système de prévision des crues, réseau neuronal. 151

1. Introduction As almost one-third of the world s population have been affected by natural disasters in the last decade of the 20 th century, about 86% of which were accounted for by either floods or droughts (Bartram (WHO), 2002), it is hardly surprising that rainfall-runoff modelling continues to be a major research activity in deterministic hydrology. Much of this research centres on efforts to understand the dynamics of the hydro-meteorological variables and the nature of the transformation mechanisms involved. The availability of increasingly efficient computing technologies and the continuous evolution of applied mathematics have provided hydrologists with new tools and techniques to carry out such modelling exercises. As a result, a plethora of such models are now available for river flow forecasting and real-time flow forecasting applications. However, despite continuous progress in research, in devising highly efficient and versatile models, no model developed thus far has succeeded in fulfilling all of its objectives uniquely and universally. So, pragmatically, the trend has been to develop or adapt catchment-specific models, to be operated either individually or in combination, in order to obtain the best individual or consensus flow forecast. Over the last three decades, the Department of Engineering Hydrology, at the National University of Ireland, Galway, has been actively involved in deterministic hydrological modelling and has acquired considerable experience in this field. Starting from very simple system-theoretic model development, the department has, in recent years, produced studies involving relatively sophisticated computing tools and techniques, including artificial neural networks, fuzzy systems etc. A large number of models of different categories are therefore available within the department. Recent efforts to create a comprehensive flow forecasting system, based on a selection of these models, have resulted in the windows-based computer package, called the Galway Flow Forecasting System (GFFS) (O Connor et al., 2001). A brief introduction to the GFFS package is provided in Annex-1. In this study, seven substantive rainfall-runoff models and three techniques for the combination of model outputs, all components of the GFFS, have been applied, in non-updating mode, to two Irish catchments, namely, the Brosna and the Fergus, both tributaries of the river Shannon. Five model performance evaluation criteria are used to assess model forecast efficiency. The primary objective is to identify an appropriate mathematical tool for modelling the rainfall-runoff transformation process in these two catchments, with the specific purpose of obtaining forecasts of the discharge at particular gauging sites. 2. The Models Used Five system-theoretic models namely, the non-parametric Simple Linear Model (NP-SLM), the parametric Simple Linear Model (P-SLM), the non-parametric Linear Perturbation Model (NP-LPM), the parametric Linear Perturbation Model (P-LPM), and the Linearly Varying Gain Factor Model (LVGFM) have been used in this study. In the context of the SLM and the LPM, the term non-parametric is used in this paper to indicate the absence of a mathematical-equation form for the series of pulse response ordinates involved in the convolutionsummation transform, the number of such independent ordinates being equal to the number of data intervals in the user-selected memory length of that response. This is in contrast to the parametric forms of these models in which the actual number of coefficient parameters required to define the transfer function (linear difference equation) form of the model, which has to be specified a priori by the user, is much smaller than the number of ordinates of its corresponding response series. A black-box type Artificial Neural Network Model (ANNM) and a physically-inspired conceptual-type model, namely, the Soil Moisture Accounting & Routing (SMAR) Model, have also been used. Results of combination of outputs from these seven basic rainfall-runoff models, in different combinations, using three techniques, namely, the Simple Average Method (SAM), the Weighted Average Method (WAM) and the Neural Network Method, are also reported. Detailed descriptions of these models and techniques are widely available, e.g. in O Connell et al., 1970; Nash and Foley, 1982; Nash and Barsi, 1983; Khan, 1986; Kachroo et al., 1988; Kachroo, 1992, a & b; Liang, 1992; Ahsan and O Connor, 1994; Liang et al., 1994; Tan and O Connor, 1996; Shamseldin, 1997; Shamseldin et al., 1997; Xiong et al. 2001. For completeness, very brief descriptions of these models and techniques are given below. 2.1 Non-Parametric Simple Linear Model (NP-SLM): Regarded as a naïve/primitive model, the intrinsic hypothesis of the SLM, introduced by Nash and Foley (1982), is the assumption of a linear time-invariant relationship between the total rainfall R i and the total discharge Q i. In its discrete non-parametric form, as the NP-SLM, it is expressed by the convolution summation relation (Kachroo and Liang, 1992), m m m Q= i Ri-j+ 1 h j+ei = G Ri-j+ 1B j, where B j = 1, and (1) j= 1 j= 1 j= 1 152

Q i and R i are the discharge and rainfall respectively at the i th time-step, h j is the j th discrete pulse response ordinate or weight, m is the memory length of the system, G is the gain factor, and e i is the forecast error term. It can be viewed simply as a multiple linear regression of the observed discharges on the current and previous observed rainfall values. Hence, for a selected memory length, estimates of the unit pulse response ordinates can be obtained directly by the method of ordinary least squares (OLS) (Nash and Foley, 1982; Kachroo and Liang, 1992). In the SLM, when the rainfall and discharge are expressed in the same units of measurements, the arithmetic sum of the discrete pulse response ordinates h j defines the gain factor G of the model, which may also be considered as the long term coefficient of runoff approximately reflecting the ratio of the total volume of the observed discharge hydrograph to that of the observed rainfall input (Kachroo and Liang, 1992). 2.2 Parametric Simple Linear Model (P-SLM): In parametric form, the Linear Transfer Function representation of the Simple Linear Model, for discrete input and output series, is given by the linear difference equation r s j yt j = j= 0 j= 1 α ω x (2) j t b j+ 1 where α j are the autoregressive parameters, with α 0 = 1, ω j are the exogenous input parameters and b is the pure time delay restricted to integer values only (Kachroo and Liang, 1992), r and s being the orders of the autoregressive and the exogenous-input parts respectively. Written explicitly for y t, with the addition of an error term, the above expression becomes t r j t j j= 0 j= 1 s y α y + ω x + e (3) = j t b j+ 1 t where α j and ω j are now the autoregressive and exogenous-input parameters respectively adjusted for the error term. Thus, the current value of y depends linearly on previous values of y and x. For a selected order of the model (r, s) and a value of the pure lag parameter b, these parameter sets α j and ω j are estimated by the OLS method. It is seen that, for given model parameter sets, the corresponding pulse response ordinates can be obtained as the output produced by applying a single pulse of unit input, giving h j = 0 j < b h j = α 1 h j-1 + α 2 h j-2 + + α r h j-r + ω j-b+1 j = b, b + 1, b + 2,, b + s 1 (4) h j = α 1 h j-1 + α 2 h j-2 + + α r h j-r j > b + s - 1 If the resulting shape of the pulse response is unsatisfactory, i.e. physically unrealistic, then appropriate changes need to be made to the model structure and the model recalibrated. In non-updating mode, past computed values of y are used in the first term on the right hand side of the transfer function equation (3), instead of the past observed values. However, the model can be more naturally used in updating mode, with the recent past observed values of y being used in that first term on the right hand side of equation (3). 2.3 Non-Parametric Linear Perturbation Model (NP-LPM): In the LPM, originally introduced in the context of rainfall-runoff modelling by Nash and Barsi (1983), it is assumed that, during a year in which the rainfall is identical to its seasonal expectation, the corresponding discharge hydrograph is also identical to its seasonal expectation. However, in all other years, when the rainfall and the discharge values depart from their respective seasonal expectations, these departures series are assumed to be related by a discrete linear time invariant system. The relation between the departures/perturbation series of the non-parametric NP-LPM, incorporating an output error term e i, may be represented algebraically by the convolution summation equation m Q i = R i j h - +1 j + ei (5) j=1 (similar in structure to the NP-SLM), where R i and Q i are the rainfall departures and the corresponding discharge departures from their seasonal expectations, respectively. As in the case of the NP-SLM, for a selected memory length, the OLS method is used to estimate the pulse response ordinates of the NP-LPM, provided that the values of these departures are known. Estimates of the seasonal expectation values are determined directly from the total rainfall and discharge series, these estimates being smoothed by harmonic analysis (by omitting the high-frequency harmonics) before subtracting them from their respective series to obtain the corresponding 153

departures series. Model estimated departure values are finally added to the seasonal expectations to provide the estimated discharge series. Like the NP-SLM, a suitable value of the memory length has to be determined by trial and error for the NP-LPM as part of the calibration procedure. 2.4 Parametric Linear Perturbation Model (P-LPM): The mathematical form of the linear component of the parametric Linear Perturbation Model (P-LPM) is identical in structure to that of the parametric Simple Linear Model (P-SLM). The only difference is that the series of departures of inputs and outputs from their seasonal mean values are used in the transfer function equation for the P-LPM instead of the actual recorded series as used in the P-SLM. The estimated departures are finally added to the seasonal means of the output series to obtain the estimated discharge values. The procedure for deriving the smoothed seasonal mean values is identical to that described in section 2.3 for the NP-LPM. 2.5 Linearly-Varying Gain Factor Model (LVGFM): The LVGFM, proposed by Ahsan and O'Connor (1994) for the single-input to single-output case, involves only the variation of the gain factor with the selected index of the prevailing catchment wetness, without varying the shape (i.e. the weights) of the response function. The model output has the familiar convolution summation structure (based on the concept of a time-varying gain factor G i ), i.e. Q m m i = G j Ri-j+ 1 B j, wh ere B j = j= 1 j= 1 1 (6) The multiple-input to single-output form of this model was investigated by Liang et al. (1994). In its simplest form, G i is linearly related to an index of the soil moisture state z i of the catchment by the equation G i = a + bz i where a and b are parameter constants. The values zi are conveniently obtained from the outputs of the naïve SLM, operating as an auxiliary model (Ahsan and O Connor, 1994), according to the relation Gˆ z i= Q m j= 1 R h ˆ i-j+ 1 j (7) where Ĝ and $ h j are estimates of the gain factor and the pulse response ordinates respectively of the SLM and Q is the mean discharge in the calibration period. The weighting sequence for the LVGFM is estimated directly by the OLS method. Like the SLM and the LPM, the LVGFM also requires the estimation of the memory length for calibration. 2.6 Artificial Neural Network Model (ANNM): The multi-layer feed-forward network type of artificial neural network, used in this study, consists of an input layer, an output layer and only one hidden layer located between the input and the output layers (Shamseldin, 1997). Each neuron of a particular layer has connection pathways to all the neurons in the following adjacent layer, but none to those of its own layer or to those of the previous layer (if any), i.e. nodes within a layer are not inter-connected. Likewise, nodes in non-adjacent layers are unconnected. There is only one neuron, for the single output, in the output layer. As the neural network itself does not incorporate storage effects, storage is implicitly accounted for by using the most recent output of one of the rainfall-runoff models as input to the ANNM, as well as the current and (in this case two) recent rainfalls, that of the NP-SLM being a convenient choice for this study. For a neuron either in the hidden or in the output layer, the received inputs y i are transformed to its output y out by a mathematical transfer function of the form M y = f ( w y + w ) out i=1 i i o (8) where f() denotes the transfer function, w i is the input connection pathway weight, M is the total number of inputs (which usually equals the number of neurons in the preceding layer), and w o is the neuron threshold (or bias), i.e. a base-line value independent of the input. The non-linear transfer function adopted for the neurons of the hidden layer and also that of the output layer is the widely-used logistic function, i.e. a form of sigmoid function, given by 154

f( M i=1 w y + w ) = i i o 1 M -σ wi yi + wo i =1 1+ e which is bounded in the range [0,1]. The weights w, the threshold i (9) w o and the σ of the different neurons can all be interpreted as the parameters of the selected network configuration. If l is the total number of neurons in the input layer and m is the total number of neurons in the hidden layer, then the total number of weights to be estimated for the ANN model adopted for the present study, is [(l+1)m + (m+1)]. The Simplex method is used for automatic optimisation of the parameters. 2.7 Soil Moisture Accounting and Routing (SMAR) Model: The Soil Moisture Accounting and Routing (SMAR) Model is a development of the Layers conceptual rainfallrunoff model introduced by O Connell et al. (1970), its water-balance component being based on the Layers Water Balance Model proposed in 1969 by Nash and Sutcliffe (Clarke, 1994). Using a number of empirical and assumed relations which are considered to be at least physically plausible, the non-linear water balance (i.e. soil moisture accounting) component ensures satisfaction of the continuity equation, over each time-step, i.e. it preserves the balance between the rainfall, the evaporation, the generated runoff and the changes in the various elements (layers) of soil moisture storage. The routing component, on the other hand, simulates the attenuation and the diffusive effects of the catchment by routing the various generated runoff components through linear time-invariant storage elements. For each time-step, the combined output of the two routing elements adopted (i.e. one for generated surface runoff and the other for generated groundwater runoff ) becomes the simulated discharge forecast. For the Brosna catchment, which is non-karstic, the variant of SMAR model incorporating the suggested modifications of both Khan (1986) and Liang (1992) is used. This variant of SMAR model, generally referred to as SMAR-G model form has nine parameters, five of which control the overall operation of the water-budget component, while the remaining four parameters (including a weighting parameter which determines the amount of generated groundwater runoff ) control the operation of the routing component. For the Fergus catchment, which is karstic and characterised by the loss of flow in underground channels, the model structure incorporates an additional parameter to facilitate the abstraction of a loss component from any excess runoff produced after all the soil layers have been filled to their combined capacity depth. This loss component is interpreted as a separate outflow function from the catchment system, defining that part of the rainfall which does not evaporate and yet will never subsequently contribute to the discharge at the outflow gauging station. This variant of the SMAR model, generally referred to as SMAR-K model form, has therefore an additional parameter, thereby raising the total number of parameters in the SMAR-K to ten. In this study, both variants of the SMAR model have been calibrated using the Simplex method of parameter optimisation. 2.8 Model Output Combination Techniques (MOCT): Although any basic model may be used independently as a flow forecasting model for a catchment, based on its own performance achieved during calibration, it will not always prove to be satisfactory in simulating hydrological events such as a flood, both in terms of the magnitude of the peak and the time to peak. However, even inferior models, having low individual performance, may be significant in improving the overall performance of the simulation, when included in a MOCT (Shamseldin et al, 1997). The Simple Average Method (SAM), the Weighted Average Method (WAM) and the Neural Network Method (NNM) are used as MOCTs in the present study. While the SAM simply performs the arithmetic average of the estimated discharge values obtained from the component basic models, the WAM assigns different but constant weights to each of the component model outputs. In the NNM, the estimated discharge output of each the basic models is assigned, at each time-step, to one (and only one) neuron in the input layer. The structure of the neural network for combination, adopted for the study, is identical to that described in section 2.8 for the ANNM. 3. The Model Efficiency Evaluation Criteria Used: Five model performance evaluation criteria have been used in the study [Kachroo, 1992a; Legates and McCabe, 1999; Beran, 1999] The coefficient of efficiency [Nash and Sutcliffe, 1970], is defined by the dimensionless expression 2 MSE = 1 N R 1 F, with o = ( ) Q 0 N Q o c 1 i [ ] 2 [ ] 2 1 N F and = ( Q o) ( Q e) MSE N (10) 1 i i 155

MSE being the mean square error. In expressions (8) and (9), (Q o ) i is the observed discharge and (Q e ) i the estimated discharge at the i th time step, N is the total number of discharge values, and Q c the mean of the (Q o ) i series over the calibration period. The index of agreement, IoA, is defined as [Willmott, 1981] N 2 [ ( Q o) ( Q )] 1 i e i IoA= 1.0 N (( ) ( ) ) 2 Q o Q c + Q e Q c i= 1 i i (11) in which the numerator is N times the MSE and the denominator is called the potential error. The other symbols have the same meaning as for R 2. The coefficient of determination, r 2, is given by r 2 where = N i= 1 N 1 [( Q ) Q ]( Q ) [ Q ] [( ) ] 0. 5 N Q Q ( Q ) Qo and o i o i o o i= 1 o i [ Q ] e i e e 0. 5 2 (12) Qe are the mean of the observed and the estimated discharge data series over the data period considered, and the other symbols have the same meanings as given above. The index of volumetric fit, IVF, the ratio of the total volume of (Q e ) i to the total volume of (Q e ) i, is given by N IVF = Q e i = 1 i N ( ) / ( ) Q o i = 1 i (13) The relative error of the peak (RE) is defined as Qp e Qp o RE = (14) Qp o where, (Q p ) o and (Q p ) e are the observed and estimated peak flows respectively. While the coefficient of efficiency R 2 is the measure of the relative improvement of the model under study over the performance of the naïve model whose forecast for all times is simply the mean of the flows in calibration, the coefficient of determination r 2 describes the proportion of the total variance in the observed data that can be explained by the model. The index of agreement IoA seeks to overcome the insensitivity of the correlation-based measures to differences in the observed and the model-simulated means and variances. Values of these three criteria range between 0 and 1, higher value indicating better model forecast efficiency. For the index of volumetric fit IVF, the value of unity indicates a perfect volumetric match of the observed flows with the estimated flows over a certain (e.g. calibration) period, indicating water balance. For the relative error of the peak RE, the lower the relative error the higher is the performance of the model. 4. The Test Catchments The Brosna catchment, located in the centre of Ireland, drains 1207 km 2 area, as measured up to the gauging site at Ferbane, at the outfall of the Brosna. The catchment is very flat, except for some undulations caused by glacial deposits. There are two lakes at the upper end of the catchment, covering about 2% of the total area. Boulder clay covers underlying carboniferous limestone and provides support for grazing ground. Some peat bogs with little woodland also exist. There is no noticeable evidence of substantial groundwater movement across the topographical boundary of the catchment. The data of the Brosna catchement are fairly typical of those generally available in the British Isles (O Connell et al., 1970). Daily data for ten years, from 1969 onwards, have been used in the study. The mean daily rainfall is 2.25 mm, the corresponding figures of pan evaporation being 1.31 mm and discharge 0.99 mm [(0.99 10-3 ) (1207 10 6 ) (24 60 2 )=13.83 cumecs]. The highest and the lowest values of flow in the data record are 97 cumecs and 2.3 cumecs respectively. 156

For the Fergus catchment, the contributing area at the Ballycorey Bridge gauging site, just upstream of Ennis town, is 562 sq. km. The catchment is predominantly flat, with stream networks and lakes concentrated mainly in the lower half of the catchment. In respect to land use, most of the catchment is farmland with some proportion of scrubland, coniferous plantation, natural and mixed woodland around the outlet. The Burren National Park, with its spectacular bare limestone landscapes, is located at the centre of the catchment. Lying in the western lowlands of Ireland, the underlying karstic limestone bedrock and thin subsoil combine to produce caves, a shallow network of swallow-holes, springs, scattered ponds, and turloughs (seasonal lakes). These features, along with some artificial channels, constitute a classic karstic hydro-geological environment, having a complex and dynamic relationship between rainfall, river flow and groundwater movement. The direction of the regional groundwater flow in the limestone is towards the river Fergus. There are a number of important groundwater features, such as streams flowing underground for part of their course through natural conduits formed in the limestone, springs, etc., most of which drain to the Fergus river at Ennis town. Daily data for six years, from 1994 onwards, have been used in this study. The mean daily rainfall is 3.94 mm, the corresponding figures of pan evaporation being 1.55 mm and discharge 1.72 mm (11.19 cumecs). The highest and the lowest values of flow in the data record are 53.51 cumecs and 0.31 cumecs respectively. For both catchments, the climate is described as being temperate. The locations of the catchments are shown in Figure 1 and the respective graphical plots of the seasonal mean variations in evaporation, rainfall and discharge for the two catchments are given in Figure 2. Fergus Brosna Fig 1: Map of Ireland showing location of the Fergus and the Brosna catchment 4 Catchment : Brosna (1207 sq.km), Ireland 8 Catchment : Fergus (562 sq.km), Ireland 3 6 2 4 1 0 0 50 100 150 200 250 300 350 Day 2 0 0 50 100 150 200 250 300 350 day Fig 2: Seasonal mean variation of evaporation, rainfall and discharge over the year (January to December) smoothed by fourier harmonic analysis (6 harmonics) Evaporation Rainfall Discharge 157

5. Methodology Each of the seven basic substantive models was applied to the test catchments, for split-record evaluation, involving the use of calibration and verification periods (about two-thirds for calibration and one-third for verification). Each model was run with the data for each catchment a number of times to determine the optimum set of values of the coefficients or the parameters for obtaining the model forms. Calibration using the whole series is required for final estimation of the response function or the parameters of the models, which are finally adopted for operational use. However, the objective of this study being the comparative assessment of performance of the different models in the GFFS, the results for split-record evaluation, rather than that of the final estimation, is presented. For the NP-SLM, the NP-LPM, and the LVGFM, the ordinary least squares (OLS) solution is used for estimation of the pulse response function. Recalling that calibration involves determining the suitable value of memory length by trial and error, by running the model a number of times and noting each time the shape of the pulse response, until a satisfactory shape of the pulse response is obtained for a particular value of the memory length and near-maximum value of R 2. The same procedure of OLS solution is adopted for estimation of the Linear Transfer Function parameters of the P-SLM and the P-LPM. The value of pure lag was chosen as 0 for both the P-SLM and the P-LPM. The SMAR model has a fixed number of parameters, namely, nine for the SMAR-G variant used for the Brosna catchment, and ten for the SMAR-K variant used for the karstic Fergus catchment, these parameters being estimated using the simplex method of optimisation. The selected memory lengths for the system-theoretic models and the SMAR model for the Brosna and Fergus catchments were 36 and 40 days respectively. For the ANNM, the number of weights depends on the number of neurons chosen for the input layer and the hidden layer. Four neurons for the input layer and three neurons for the hidden layer were selected for the Fergus catchment, whereas for the Brosna catchment the corresponding numbers were four and two respectively. The number of neurons adopted was selected by trial and error tests, with the objective of achieving the highest possible values of the performance evaluation criteria. The simplex method of automatic optimisation was also used for estimation of the ANNM weights. For the MOCT, the results obtained by combination of the outputs from all the seven models, together with those obtained by considering the best 6, the best 5, the best 4, the best 3, and the best 2, and using all three techniques namely, SAM, WAM and NNM, are presented. These results are shown in the bottom half of Table 1, in which the values of the five chosen performance evaluation criteria, obtained for the calibration and the verification periods are given. Representative graphical plots of the estimated discharge series, together with the corresponding observed discharge and rainfall series for both catchments, for the last year of data record, are shown in Figure 3. Cat chment : Brosna (1,207 sq. km), Ireland Year : 1978 100 80 60 40 20 0 0 50 100 150 200 250 300 350 Day 0 10 20 30 40 50 60 Cat chment : Fergus (562 sq. km), Ireland Year : 1999 60 0 20 50 40 40 30 20 60 80 100 120 10 140 0 160 0 50 100 150 200 250 300 350 Day Fig 3: Observed Discharge, Estimated Discharge (estimated by combining outputs of individual models by Neural Network for the test corresponding to the best performance, as shown in Table 2) and Observed Rainfall series for the last calendar year of the data record,for the Brosna and the Fergus catchments. Observed rainfall Observed discharge Estimated discharge 6. Results and Discussions: In Table 1, the performances of the models are ranked primarily according to the R 2 index. However, when the values of the R 2 index for two tests were found to be similar, the other indices in the order shown in Table 1 were used to differentiate the performance levels. The R 2 index generally showed a consistent pattern of comparison with the IoA and the r 2 index, i.e. a high value of R 2 corresponded to high values of both IoA and r 2. 158

In the columns displaying the ranks of the models according to their performances, the ranks of the basic individual rainfall-runoff models are shown left-justified, whereas the ranks of the three combination techniques are shown right-justified. For both catchments, the values of the performance indices of the highest ranking individual model and the highest ranking combination technique are shown in bold font. It is clear from these results that, for both catchments, the performances of both the parametric and nonparametric forms of the SLM are generally inferior to those of all the other GFFS models. This stems from the very crude and simplified SLM representation of the outflow series as the convolution summation of the response function with the inflow series, with the restrictive assumption of constancy of the gain factor. As expected, the LVGFM, which is an elaboration of the P-SLM, incorporating an element of linear variation of the gain factor with the catchment wetness index at each time-step, performs consistently better than the SLM. Recalling that the structure of the ANNM is characterised by a lack of parsimony in model parameters, and that the output from the naïve NP-SLM is used as input to one of the neurons in the input layer to take into account the storage effects of the system, it is not really surprising that its performance, although higher than those of the NP-SLM and the P-SLM, is hardly better than that attained by the LVGFM. The performances of both the nonparametric and the parametric forms of the LPM are seen to be significantly higher than those of all other models of the black-box type i.e. the NP-SLM, the P-SLM, the LVGFM, and the ANNM, for both catchments. This too is perhaps to be expected, considering the strong seasonality exhibited by the hydro-meteorological data series, as displayed in Figure 2. Nevertheless, none of the black-box models can be said to perform satisfactorily on either catchment. Overall, the performance of the conceptual SMAR model is seen to be consistently the best, for both catchments. For the Brosna catchment in the calibration period, both the Nash-Sutcliffe R 2 and the r 2 values for SMAR are about 85% and the value of the index of agreement (IoA) is around 96%. For the Fergus 159

Table 1 Calibration and verification results from seven models, applied individually and in combination using three combination techniques to two Irish catchments: the Brosna and the Fergus Model Brosna (1,207 sq. km) Ireland Fergus (562 sq. km) Ireland Calibration Verification Calibration Verification R 2 IoA r 2 IVF RE Rank R 2 IoA r 2 IVF RE Rank R 2 IoA r 2 IVF RE Rank R 2 IoA r 2 IVF RE Rank NP-SLM 0.410 0.748 0.409 1.007 0.500 5 0.462 0.722 0.561 0.867 0.544 7 0.696 0.896 0.707 1.075 0.241 7 0.769 0.920 0.796 1.004 0.233 7 P-SLM 0.381 0.737 0.391 1.085 0.526 7 0.493 0.756 0.556 0.959 0.544 6 0.727 0.902 0.780 1.133 0.334 6 0.788 0.923 0.863 1.047 0.261 6 NP-LPM 0.709 0.911 0.710 1.015 0.447 3 0.782 0.920 0.838 0.874 0.492 2 0.878 0.967 0.878 1.010 0.168 3 0.910 0.974 0.917 0.957 0.169 2 P-LPM 0.725 0.915 0.725 0.997 0.478 2 0.730 0.900 0.766 0.870 0.490 3 0.886 0.967 0.891 1.007 0.267 2 0.876 0.960 0.901 0.940 0.255 3 LVGFM 0.428 0.772 0.428 0.989 0.297 4 0.501 0.759 0.582 0.858 0.466 5 0.737 0.922 0.736 1.002 0.047 5 0.805 0.944 0.807 0.969 0.102 5 ANNM 0.410 0.746 0.401 1.001 0.505 6 0.525 0.777 0.596 0.879 0.489 4 0.743 0.743 0.743 1.004 0.089 4 0.817 0.945 0.818 0.945 0.076 4 SMAR 0.845 0.958 0.846 1.016 0.263 1 0.866 0.958 0.881 0.951 0.344 1 0.979 0.978 0.978 1.014 0.075 1 0.981 0.995 0.981 1.002 0.027 1 MOCT All 7 MOCT All 6 MOCT All 5 SAM 0.683 0.883 0.704 1.026 0.469 18 0.690 0.865 0.820 0.894 0.497 18 0.881 0.963 0.899 1.035 0.181 17 0.915 0.974 0.943 0.981 0.141 17 WAM 0.877 0.968 0.880 1.002 0.231 8 0.898 0.969 0.910 0.910 0.303 6 0.981 0.995 0.981 1.005 0.032 4 0.981 0.995 0.981 0.996 0.046 6 NNM 0.900 0.974 0.900 1.012 0.126 4 0.910 0.974 0.916 0.930 0.210 2 0.986 0.996 0.986 1.000 0.032 1 0.983 0.996 0.983 0.979 0.022 1 SAM 0.715 0.897 0.737 1.011 0.451 17 0.717 0.880 0.843 0.883 0.487 17 0.901 0.970 0.917 1.028 0.166 16 0.930 0.978 0.954 0.977 0.123 16 WAM 0.880 0.968 0.880 0.992 0.525 7 0.891 0.966 0.912 0.906 0.312 7 0.980 0.995 0.980 1.013 0.057 8 0.981 0.995 0.981 1.000 0.004 5 NNM 0.903 0.974 0.903 1.000 0.154 2 0.906 0.972 0.918 0.924 0.223 3 0.982 0.995 0.982 1.001 0.120 3 0.980 0.995 0.980 0.986 0.095 7 SAM 0.747 0.912 0.768 1.007 0.440 16 0.742 0.894 0.860 0.884 0.480 16 0.914 0.975 0.922 1.008 0.130 15 0.939 0.982 0.953 0.963 0.095 15 WAM 0.876 0.966 0.976 0.990 0.243 9 0.883 0.963 0.906 0.911 0.299 9 0.979 0.995 0.980 1.021 0.073 9 0.982 0.995 0.982 1.006 0.022 3 NNM 0.903 0.974 0.903 1.003 0.157 3 0.905 0.972 0.917 0.926 0.224 5 0.981 0.995 0.981 1.000 0.120 5 0.978 0.994 0.978 0.988 0.097 9 MOCT Best 4 MOCT Best 3 MOCT Best 2 SAM 0.792 0.932 0.806 1.007 0.406 15 0.787 0.918 0.880 0.888 0.454 15 0.933 0.981 0.941 1.009 0.156 14 0.948 0.985 0.964 0.961 0.139 14 WAM 0.864 0.962 0.865 1.012 0.3107 10 0.864 0.955 0.894 0.927 0.394 10 0.979 0.995 0.980 1.021 0.081 10 0.982 0.995 0.982 1.005 0.034 2 NNM 0.900 0.973 0.900 1.000 0.151 5 0.905 0.972 0.917 0.930 0.236 4 0.980 0.995 0.980 1.002 0.134 6 0.977 0.994 0.977 0.984 0.116 11 SAM 0.823 0.947 0.825 1.014 0.387 14 0.829 0.940 0.880 0.898 0.449 14 0.944 0.985 0.948 1.011 0.172 13 0.949 0.985 0.960 0.966 0.159 13 WAM 0.864 0.961 0.865 1.013 0.311 11 0.863 0.955 0.895 0.928 0.393 11 0.979 0.994 0.980 1.017 0.082 11 0.982 0.995 0.982 1.002 0.036 4 NNM 0.903 0.974 0.903 1.000 0.139 1 0.910 0.973 0.922 0.926 0.219 1 0.980 0.995 0.980 1.000 0.136 7 0.977 0.994 0.977 0.985 0.118 10 SAM 0.849 0.956 0.850 1.013 0.345 13 0.840 0.945 0.882 0.910 0.422 13 0.976 0.988 0.962 1.011 0.171 12 0.956 0.988 0.967 0.971 0.147 12 WAM 0.864 0.961 0.864 1.012 0.308 12 0.862 0.954 0.893 0.927 0.392 12 0.979 0.994 0.980 1.017 0.082 11 0.982 0.995 0.982 1.002 0.036 4 NNM 0.897 0.972 0.897 1.000 0.152 6 0.890 0.966 0.907 0.925 0.254 8 0.982 0.995 0.982 1.000 0.105 2 0.979 0.995 0.979 0.986 0.082 8 160

catchment, the SMAR values of these three indices, in the calibration period, is around 98%. The index of volumetric fit (IVF) for SMAR is around 1.02 for the Brosna catchment and 1.00 for the Fergus catchment. The relative error (RE) of peak, for both catchments, is at least as good as for the other GFFS models. So, while the performance levels attained by the black-box models are unacceptably low, those attained by the SMAR model are sufficiently high, and may therefore be accepted for flow forecasting applications. These results indicate the ability of the conceptual SMAR model to adequately simulate the rainfall-runoff transformation process in these small catchments, despite its lumped-input and constant-parameter restrictions. From the calibration results obtained by applying the MOCT to the outputs of the substantive models, it is seen that the performance of the Neural Network Method (NNM) of combination is generally the best, followed by the Weighted Average Method (WAM). The Simple Average Method (SAM) of combination, being a special case of the WAM, with equal weights, always performs worse than the other two MOCTs, as expected. It should also be noted that, while the combination of outputs of all of the seven basic rainfall-runoff models tested were necessary to achieve the best simulation of the observed flow series for the Fergus catchment, the combination of only the best three model outputs achieved performance levels similar to that of all seven models for the Brosna catchment. For the Brosna catchment, the application of the best model output combination technique (MOCT) increased the value of the R 2 index for the best individual model by 6%, whereas for the Fergus catchment the corresponding increase using the best MOCT was a modest 1%. Although not displayed here, the application of the best MOCT also produced an improvement in simulating the observed flow values both in terms of the magnitude of the peak flow and the time to peak. 7. Conclusions And Recommendations The performances of both the NP-SLM and the P-SLM, which are very crude and simplified forms of the actual input-output transformation process, are found to be generally inferior to those of all other GFFS models tested. For both catchments, characterised by strong seasonality, both the NP-LPM and the P-LPM outperform the LVGFM, the ANNM and the SLM, the performance of the ANNM being either higher or almost at the same level as that of the more parsimonious LVGFM. Overall, for both catchments, the SMAR conceptual model performs consistently better than all other individual models tested. Outputs from all of the basic rainfall-runoff models, including even the naïve SLM, in different combinations for different catchments, are useful in obtaining the best combined (MOCT) simulation of the observed outflow series, reflecting the attempt to combine the strengths and cancel out the perceptible weaknesses of the different GFFS models. The simulated flow series obtained by applying the model output combination techniques (MOCTs) to the outputs from the basic rainfall-runoff models shows significant improvement in simulating the observed flow values, the performance of the ANNM form of MOCT being generally the best. Having selected the final model structure, whether in the form of an individual rainfall-runoff model or as a combination of the outputs of a number of such models using a MOCT, the forecast efficiency can generally be further enhanced by using an appropriate forecast-updating technique. Although four different model-output updating techniques are included in the GFFS package, namely, auto-regressive (AR) models, Linear Transfer Function (LTF) models and two forms of neural networks, space limitations precluded the application or even a description of these updating procedures in the present paper. Such an exercise would facilitate a comparative assessment of performances of these different updating procedures for the test catchments, and thereby determine the best procedure for real-time flow forecasting applications. Considering the limitations and uncertainties involved in the application of these GFFS models in the wider Irish context, the present authors feel that although tested on only two Irish catchments, the results presented have relevance to other catchments in Ireland and perhaps in the British Isles generally, considering that the climatological and geographical features and the data of the Brosna catchment is considered to be fairly typical of the region (O Connell, 1970). It is recommended, however, that more extensive tests be performed with data from other such small catchments in Ireland and in Britain to determine if one or other variant of the SMAR model performs consistently better than the other substantive rainfall-runoff models in the GFFS and to investigate the consistency in the improvements afforded by the MOCTs and, in the real-time forecasting context, by the forecast updating options of the GFFS. Acknowledgements The first author (Monomoy Goswami) gratefully acknowledges the support provided from the National University of Ireland, Galway, through its Millennium Research Fund, for his research on the development of the GFFS package used to obtain all the results presented in this paper. 161

References Ahsan, M. and K.M. O'Connor, A simple non-linear rainfall-runoff model with a variable gain factor. Journal of Hydrology, 155: 151-183, 1994. Bartram, J. (WHO). Water too much or too little. WATER, Issue 14, Page 3, Dept. for International Development, HR Wallingford, Howberry Park, Wallingford, Oxon. OX10 8BA, U.K., May 2002. Beran, M., Hydrograph Prediction How much skill? Hydrology & Earth System Sciences, 3(2), 305-307, 1999. Clarke, R.T., Statistical Modelling in Hydrology, Page 309: Section 8.2.2: The Nash-Sutcliffe Layer Model. John Wiley and Sons, 1994. Kachroo, R.K., G.C. Liang, and K.M. O Connor, Application of the linear perturbation model (LPM) to flood routing on the Mekong River, Hydrological Sciences Journal, 33, 2-4, 193-214, 1988. Kachroo, R.K., River flow forecasting, Part 1, A discussion of principles. Journal of Hydrology, 133: 1-15, 1992a. Kachroo, R.K., River Flow forecasting, Part 5, Applications of a conceptual model. Journal of Hydrology, 133: 141-178, 1992b. Kachroo, R.K. and G.C. Liang, River flow forecasting, Part 2, Algebraic development of linear modelling techniques. Journal of Hydrology, 133: 17-40, 1992. Khan, H., Conceptual Modelling of rainfall-runoff systems, M. Eng. Thesis, National University of Ireland, Galway, 1986. Legates D.R. and G.J. McCabe Jr., Evaluating the use of goodness-of-fit measures in hydrologic and hydroclimatic model validation. Water Resources Research, 35(1): 233-241, 1999. Liang, G.C., A note on the revised SMAR model. Workshop Memorandum, Department of Engineering Hydrology, National University College of Ireland, Galway (Unpublished), 1992. Liang, G.C., K.M. O Connor, and R.K. Kachroo, A multiple-input single-output variable gain factor model. Journal of Hydrology, 155, 185-198, 1994. Nash, J.E. and J.V. Sutcliffe, River flow forecasting through conceptual models, Part 1, A discussion of principles. Journal of Hydrology, 10: 282-290, 1970. Nash, J.E. and J.J. Foley, Linear models of rainfall-runoff systems. In: Rainfall-Runoff Relationship, Proceedings of the International Symposium on Rainfall-Runoff modelling. Mississippi State University, May 1981, USA. Edited by V.P. Singh, Water Resources Publications pp.: 51-66, 1982. Nash, J.E. and B.I. Barsi, A hybrid model for flow forecasting on large catchments. Journal of Hydrology, 65: 125-137, 1983. O Connell, P.E., J.E. Nash, and J.P. Farrell, River Flow forecasting through conceptual models. Part 2. The Brosna catchment at Ferbane. Journal of Hydrology., 10: 317-329, 1970. O Connor, K.M., The Education and Training of Hydrologists, A Discussion Paper submitted to the Unesco Steering Committee of IHPV Project 8.1 in the context of its Draft Education Policy Document.(unpublished), 1998. O Connor, K.M., M. Goswami, G.C. Liang, R.K. Kachroo, and A.Y. Shamseldin, The Development of The Galway Real-Time River Flow Forecasting System (GFFS). The proceedings of the 19 th European Regional Conference of ICID, Czech Republic, 2001. Shamseldin, A. Y., Application of a neural network technique to rainfall-runoff modeling. Journal of Hydrology, 199, 272-294, 1997. Shamseldin, A.Y., and K.M. O Connor, 1999. A real-time combination method for the outputs of different rainfall-runoff models. Hydrological Sciences Journal, 44 (6): 895 912. Shamseldin, A.Y., K.M. O Connor, and G.C. Liang, Methods for combining the outputs of different rainfallrunoff models. Journal of Hydrology, 197: 203-229, 1997. Tan, B.Q. and K.M. O Connor, Application of an empirical infiltration equation in the SMAR conceptual Model. Journal of Hydrology, 185: 275-295, 1996. Willmott, C.J., On the validation of models. Phys. Geogr., 2, 184-194, 1981. Xiong, L., K.M. O Connor, and M. Goswami, Application of the artificial neural network (ANN) in flood forecasting on a karstic catchment. The XXIX IAHR Congress, Beijing, 2001. ANNEX 1: FEATURES OF THE GFFS SOFTWARE PACKAGE The Windows-based software package called the Galway Flow Forecasting System (GFFS) includes a range of models of both the system-based Black-Box type as well as the quasi-physical Conceptual type. The basic models included are the non-parametric and the parametric forms of the Simple Linear Model, the non-parametric and the parametric forms of the Linear Perturbation Model, the Linearly Varying Gain 162

Factor Model, the Artificial Neural Network Model, and the Soil Moisture Accounting and Routing Model. Various options are provided, wherever applicable, to facilitate the use of different variants of a model, different routing procedures, different optimisation schemes, etc. The models may be used for the simulation of river flow at a gauging site, purely in the context of rainfall-runoff transformation processes, or of routing processes, and also for processes involving a combination of both scenarios. An effective tool, in the form of the Model Output Combination Technique (MOCT), is also included to facilitate the combination of the performances of a number of different models (including perhaps one or more of the user s own models) in an efficient way, in order to better simulate the observed discharge. Options for updating the model outputs by four different updating procedures are included (e.g. Autoregressive, Linear Transfer Function and two forms of Neural Network). A separate facility for real-time application of the GFFS package is provided. For application in real time, an individual model or a combination of models, with or without updating option, has to be calibrated first. The calibrated values of the parameters or coefficients are used by the package, in addition to real-time inputs to be provided by the user for issuing real-time flow forecasts. There is provision for the display of inundation maps for floods, to be provided by the user, corresponding to various flow thresholds, to be specified by the user, for instantaneous display of maps along with the linked flow values, for the specified lead-times, in real time. Other useful utilities, such as making data files in standard format for use in the package, viewing the contents of a hydrological data file both in summary description as well as in graphical presentations, basic file handling operations, etc. are also included. The package is characterised by full functionality, a user-friendly Graphical User Interface (GUI), flexibility, adaptability, ease of operation, high level of visual displays, and considerable ability to recognise user s mistakes in entering data or keying in instructions. In the GUI environment, the user interfaces with the GFFS package using the keyboard, the mouse, the windows, the menus, and the toolbar buttons. The Multiple Document Interface (MDI) style of the package facilitates the display of a number of windows containing a wide variety of information simultaneously and easy negotiation of transfer from one window to another. The general look/feel of the package is very similar to other standard windows-based application programs. The following figures are intended to display some of the features contained in the package. The text boxes facilitate entering the input information while the check boxes and the option buttons present choices to the user. Graphics are displayed in a highly interactive environment. The graphs are accompanied by tabular display of information. If the user wants to use data for making customized graphs, according to the choice or the need, tables can be copied into the clipboard for pasting in Excel or other applications. Message boxes are used to prompt the user to perform an action, to provide the user with required information, to warn the user of unacceptable (fatal) errors, or to ask a question to the user. Facilities are provided for easy browsing for the files to be used in the package. Wherever applicable, sensible default values are provided, which may, however, be altered or overwritten by the user according to the choice or the need. Six representative snapshots of the windows from the GFFS package are given below in Figure 4. Start-Up Window of the GFFS Graphical display of seasonal expectations for the Linear Perturbation Model (LPM) Fig. 4 : Snapshots of windows from the GFFS package (continued on next page) 163

Continuation of Fig. 4 Data-entry Window for running Linearly Varying Gain Factor Model (LVGFM) Data-entry Window for running Soil Moisture Accounting and Routing (SMAR) Model Window showing outputs from the Auto- Regressive (AR) updating Model Window for choosing an option for real time flow forecasting Fig 4 : Snapshots of windows from the GFFS package Help files are included in the package, which provide detailed and hopefully understandable on-line documentation aimed at familiarising the user with the various features of the package and providing the user with answers to queries relating to its use. 164

Predicting biological and chemical recovery in the Galloway region of SW Scotland R.C. Helliwell 1 1 Macaulay Institute, Craigiebuckler, Aberdeen, Scotland, AB15 8QH, United Kingdom S. Juggins 2 2 Department of Geography, University of Newcastle, Newcastle upon Tyne, NE1 7RU, United Kingdom Abstract Historically, the Galloway region of SW Scotland has been subject to chronic acid deposition that has resulted in acidification of soils and surface waters and subsequent damage to salmonid fish and other freshwater biota. The dynamic model, MAGIC777, was used to simulate the current, reconstructed past, and projected future biochemical status of seven sites in the region. This application of MAGIC777 evaluates the degree of compliance with respect to restoration of acidified waters by the year 2016, as stipulated by the Water Framework Directive (WFD). Long-term biochemical trends are assessed in terms of catchment attributes (land cover, soil, geology, and topography), catchment management and emission reductions under the Gothenburg Protocol. Significant differences are observed between the temporal trends in surface water chemistry and biology of the seven sites. In particular, the Grannoch catchment is dominated by base poor granitic geology and extensive forestry, and is the most acidified site in this study with an acid neutralising capacity (ANC) 13.98 µeq l -1, ph 4.57 and Ca 50.69 µeq l -1. Application of biological models suggests that these conditions have a deleterious impact on the acid sensitive diatoms (Achnanthes minutissima) and macro-invertebrates (Baetis rhodani) at this site. In contrast, greywackes and shales with higher weathering rates underlie the catchment of Loch Knockstring, and as a consequence, the surface water chemistry is less acid (ANC 301.87 µeq l -1, ph 6.8, Ca 294.50 µeq l -1 ). Under these conditions, the probabilities of occurrence of Achnanthes minutissima and Baetis rhodani are 0.89 and 0.71 respectively. The relationships identified in this study between these taxa and ANC are applied to the output of a MAGIC simulation from 1857 to 2016. The mean ANC value for the seven sites declined from 133.89 µeq l -1 in 1857 to 61.61 µeq l -1 in 1997, in response to the large accumulated flux of sulphur and nitrogen deposition which occurred from the start of the Industrial Revolution to the mid 1970s. Biological predictions during this period also show a loss of the most sensitive taxa and a reduction in healthy salmonid populations. With the implementation of the Gothenburg Protocol (reduced sulphur and nitrogen deposition) an improvement in surface water ANC relative to present day is predicted, with a mean ANC increase of 24.11 µeq l -1 in 2016. Fish stocks are predicted to completely recover at two of the seven lochs as a result of the improved water quality in 2016. Populations of Achnanthes minutissima and Baetis rhodani also show a considerable improvement between present and projected future conditions at all sites. Further reductions in acid deposition and tighter controls on afforestation in this region are required if the biochemical status of these lochs is to return to pre-acidification levels. Résumé Historiquement, les dépositions acides chroniques, dans la région Galloway au sud-ouest de l écosse, ont entrainé une acidification des sols et eaux de surfaces qui conduit à l altération des salmonidés et autre biota d eau douce. Le modèle dynamique, MAGIC777, a été utilisé pour simuler les conditions biochimiques passées, présentes et futures de la région à sept localités. L application de MAGIC777 évalue le degré de conformité avec la restauration des eaux acidifiées d ici l année 2016, comme stipulé par la loi cadre sur l eau (2000/60/EC). Les tendances biochimiques sont estimées en fonction des caractéristiques du bassin versant 1

(paysage, sol, géologie, topographie), de la gestion du bassin et de la réduction des émissions sous le protocole de Gothenburg. Des différences significatives sont observées entre les tendances temporelles de la chimie des eaux de surface et la biologie des sept stations. En particuliers, le bassin versant Grannoch est dominé par une géologie granitique pauvre en base et une vaste sylviculture, et est le site le plus acidifié dans cette étude avec une capacité de neutralisation des acides (ANC) de -13,98 μeq.l -1, ph 4,57 et Ca 50,69 μeq.l -1. L application de modèles biologiques suggère que les conditions de ce site ont un effet nuisible sur les diatomées (Achnanthes minutissima) et macroinvertébrés (Baetis rhodani) sensibles à l acidité. En revanche, le plus haut taux d altération des schistes gréseux et argileux du bassin versant du Loch Knockstring génère une chimie moins acide des eaux de surface (ANC 301,87 μeq.l -1, ph 6,8 et Ca 294,50 μeq.l -1 ). Sous ces conditions, les probabilités de présence d Achnanthes minutissima et de Baetis rhodani sont respectivement de 0,89 et 0,71. Les rapports identifiés dans cette étude entre ces taxa et l ANC sont reliés aux résultats générés par une simulation de MAGIC de 1857 à 2016. La valeur moyenne de l ANC des sept sites déclina de 133.89 μeq.l -1 en 1857 à 61.61 μeq.l -1 en 1997, en réponse aux larges accumulations des dépositions de soufre et azote depuis la révolution industrielle jusqu à la moitié des années mille neuf cent soixante-dix. Les prédictions biologiques pendant cette période montrèrent aussi une perte des taxa les plus sensibles et une réduction des population de salmonidés en bonne santé. Avec l accomplissement du protocole de Gothenburg (déposition de soufre et azote réduites) une amélioration de l ANC des eaux de surface, par rapport aux conditions présentes, est prédite avec une augmentation moyenne de l ANC de 24.11 μeq.l -1 en 2016. Les prédictions du modèle montrent que les stocks halieutiques peuvent récupérer dans deux des sept lochs en conséquence de l amélioration de la qualité de l eau en 2016. Les populations d'achnanthes minutissima et de Baetis rhodani montrent également une amélioration considérable entre les conditions présentes et projetées dans le futur à tous les sites. Des réductions accrues de dépôt acide et un meilleur contrôle de reboisement dans cette région sont requis si le statut biochimique de ces lochs doit retourner au niveau précédant l acidification. Introduction Acidification of surface waters in the UK uplands is linked to the emission and subsequent deposition of oxides of sulphur and nitrogen from the atmosphere. Historically, deposition of non-marine sulphate increased to peak levels in the late 1970 s and then declined dramatically in the 1980 s, whilst nitrogen deposition increased during the 1950 s, 1960 s and 1970 s and has remained at high levels since (Ferrier et al. 2001; Harriman et al. 2001). The decline in S emissions in the 1980s was the result of a series of protocols that were initiated by the UN- ECE Convention on Long-Range Transboundary Air Pollution (CLRTAP). The most recent EN-ECE agreement, the Multi-Pollutant Multi-Effect Protocol (Gothenburg Protocol), was signed in 1999 with the aim of limiting emission of sulphur, nitrogen oxides, ammonia, and volatile organic compounds. The degree of biochemical recovery of surface water is dependent not only on the extent and nature of deposition, but also on the ability of catchments to buffer incoming acidity. Previous studies have shown how soil, geology, land cover and catchment hydrology influence the chemical composition of surface water (Kernan & Helliwell, 2001; Monteith & Evans, 2000). In the Galloway region, palaeoecological records and bio-monitoring provide evidence that surface water acidification during the past 150 years has resulted in a detrimental change to aquatic biota through the loss of acid sensitive taxa (Flower et al. 1987). More recent evidence of a link between reductions in S and N emissions as a consequence of emission reduction protocols, and the reversibility and recovery of surface water acidification and biology in the Galloway area was highlighted by Harriman et al. 2001. International efforts to protect and manage acid sensitive ecosystems have led to the implementation of the EC Water Framework Directive. An evaluation of the degree of 2

compliance with respect to restoration of acidified waters by 2016 (as specified under the WFD), with the emission reductions proposed under the Gothenburg Protocol is presented for seven sites in the Galloway region. Site description The seven loch catchments selected for this study are the Round Loch of Glenhead, Loch Enoch, Loch Grannoch, Loch Riecawr, Loch Macaterick, Loch Knockstring, and Loch Cornish (Fig. 1). Selected physical and chemical characteristics of the seven drainage basins are given in Table 1. Lochs range in altitude from 210 m (Loch Grannoch) to 490 m (Loch Enoch) and vary in character from small mountain lochs 0.06 km 2 (Loch Cornish) to large lochs 1.11 km 2 (Loch Grannoch). The bedrock geology of the Riecawr, Macaterick, Knockstring and Cornish catchments are characterised by organic soils overlying slates, shales and greywackes (Lower Palaezioc type, chiefly Ordovian and Silurian) which have metamorphosed into the harder quartzite and mica schists in the areas close to two dominant granitic plutons. To the north of the region the Enoch and Round Loch of Glenhead catchments are located on the granite pluton of Mullwharcher (692 m NGR NX 454867). In the south of the region the Loch Grannoch catchment is underlain by the granite pluton of the Cairnsmore of Fleet (711 m, NG NX 502672). Catchment soils are essentially blanket peat, peaty podzols and sub-alpine developed directly on bedrock or on coarse drift, and drainage is poor. Catchment vegetation consists mainly of acid molinietum heath with Calluna vulgaris (L.) Hull and Myrica gale L. locally abundant. This vegetation is partly replaced in the four afforested catchments (Table 1) by conifer plantations, mainly of sitka spruce (Picea sitchensis Carriere). Afforestation of the Loch Riecawr catchment occurred in 1965-1969 whilst the Loch Grannoch catchment was afforested in two phases in 1962 and 1977-1985. A small area (7%) of the Macaterick catchment was planted between 1965-74 and a further 10% was planted during 1975-1979. More recently (1990-1994), 60% of the Knockstring catchment was planted with commercial forestry. The region receives some of the highest rainfall in Scotland with annual average rainfall exceeding 2000 mm, and relatively high acid deposition (c. 32 kg ha -1 yr -1 S and c. 33 kg ha -1 yr -1 N (Monteith & Evans, 2000). 3

Hydrochemical and Biological Modelling Background to the MAGIC Model The process-based dynamic model, MAGIC (Model of Acidification of Groundwaters In Catchments) describes long-term changes in soil and surface water acidification status at the catchment scale (Cosby et al., 1985a,b). The model is applied to seven sites to predict the potential biochemical recovery under the latest agreements controlling sulphur emissions throughout Europe, namely the Gothenburg Protocol. Predications are also made of the possible effects of plantation forestry on acidification status in combination with changing acidic deposition. The forestry scenario selected to represent the most likely land-use management practice by the forestry Commission is to cut and replant areas of forest at 50 years of age. The site-based application of MAGIC requires data for each site on a range of physical and chemical properties. Input data to the model includes annual rainfall and runoff volumes, deposition chemistry, and soil depth, bulk density, porosity and cation exchange capacity (CEC). The model is calibrated to eight measured target variables ; surface water concentrations of base cations calcium (Ca), magnesium (Mg), sodium (Na), and potassium (K), and soil exchange fractions of the same ions. A detailed account of the model parameterisation and calibration procedures is given in Evans et al. (2001). At Loch Enoch, Loch Grannoch, and Round Loch of Glenhead sediment cores have been collected and palaeoecological reconstructions of ph have been inferred. The MAGIC model has been shown to closely match the historical build up of acidification (Jenkins et al. 1990). Background to Biological Modelling Incorporation of trout population status assessment into MAGIC The relationship between water chemistry and the survival of brown trout was investigated through chemistry and fish data from the Norwegian 1000-lake survey (Henriksen et al. 1989). The status of brown trout in these lakes differed significantly according to the concentrations of Al, NO 3, SO 4, Ca, ph and ANC. The fish in lakes with an ANC <20 µeq l -1 were generally damaged, and in those lakes with an ANC <0 µeq l -1 fish were generally barren. A simple logistic algorithm based on the probabilities of a lake within a given ANC falling into each of three trout status categories (healthy, marginal, barren) was incorporated into MAGIC (Wright et al. 1994). The application of the fish model was considered appropriate in this study since the region of Galloway has similar environmental characteristics to Norway in terms of geological, physiographical, and hydrochemical attributes. Diatom and invertebrate biological status Changes in the biological status of diatom and invertebrate assemblages are investigated through the use of simple logistic regression models relating the probability of occurrence of the diatom Achnanthes minutissima and the invertebrate Baetis rhodani to surface water ANC. The models are derived from modern taxon distributions in 358 acid to circumneutral streams across the UK (Juggins, 2001). Both organisms are common in streams above ph 6.0 and are among the first early-warning indicator taxa to decline in response to lowered ph. These models thus provide a tool to convert MAGIC chemical predictions into measures of biological status of loch outflow streams. 4

Results and Discussion Assessment of Water Chemistry and Catchment Characteristics The chemical composition of the seven lochs reflects the nature of the contributing catchment areas. The bedrock geology has a strong influence on surface water ANC and ph with the most acid lochs (Loch Grannoch, Loch Enoch and the Round Loch of Glenhead) being situated on the granitic plutons in the region. The ANC of these sites (-13.98, -6.33, and 2.42 µeq l -1 respectively) are considerable more acid than those sites that are underlain by greywackes (Table 1). The highest Ca concentration (294.50 µeq l -1 ) was recorded at Loch Knockstring, a catchment underlain predominantly by greywackes, the soils developed on this geology are generally base rich. Enhanced inputs of dry and occult deposition at the forested sites (Loch Knockstring, Loch Grannoch, Loch Riecawr and Loch Macaterick) result in the greatest concentration of SO 4 * and Cl. The relationship between the soil C pool and DOC in surface waters has been well-documented (Hope et al. 1997). In this study the greatest loch DOC concentration was observed in the predominantly peaty catchment of Loch Macaterick (5.66 mg l -1 ), and the lowest concentrations were observed in the high altitude lochs (Loch Enoch 1.79 mg l -1 and Round Loch of Glenhead 3.22 mg l -1 ) where the catchment soils are thin and mineral. There is no obvious relationship between concentrations in toxic labile Al and other catchment characteristics, although the greatest concentrations were observed at Loch Grannoch (184 µg l -1 ). MAGIC Modelling Model Calibration Simulated present day surface water chemistry is in good agreement with observed values. For the seven sites, the mean deviation of observed and simulated ANC, ph, SO 4, and Ca was 6.9%, 3.2%, 1.9%, 0.4% respectively (Table2). Long term hydrochemical trends and environmental change MAGIC simulated ANC for 1857 for Loch Knockstring, Riecawr, and Cornish (Fig. 2) illustrates that, prior to industrialisation, surface waters were not acidic. These surface waters can be considered reasonably well buffered, and would probably have supported a relatively diverse acid sensitive ecosystem at this time. The remaining four sites however have a simulated pre-industrial ANC between 50 and 94 µeq l -1, indicating sensitivity to acidification. With the onset of industrialisation in the mid to late nineteenth century increased emissions of acidic oxides and subsequent deposition of S and N resulted in a mean decline in ANC of 72 µeq l -1 from 1857 to 1997 at all sites. With the implementation of the Gothenburg Protocol the predicted acidification status of all lochs, (except Loch Knockstring), within 19 year period from present day (1997) to the 2016 is for a considerable improvement in water quality with ~49% of recovery towards preacidification ANC. ( Recovery is defined as the forecast ANC recovery to 2016 as a percentage of the net ANC decline from pre-industrial conditions to present day, i.e. [ANC 2016 -ANC 1997 ]/[ANC 1997 -ANC 1857 ] x 100 (Evans et al., 2001). As a consequence of recent forest practices (1990-94) at the Loch Knockstring site, the soil base cation pool is predicted to decrease and total inputs of acid deposition are predicted to increase resulting in a gradual decline in the loch ANC until the forest is felled 50 years after planting in 2040 (Fig. 2). Projected trends in surface water quality beyond 2016 indicate that recovery continues at the moorland sites at a much slower rate, and at the forested catchments the surface waters begin to re-acidify (Fig. 2). 5

If surface waters are to be restored to good status by 2016 then recovery towards simulated pre-industrial (1857) is predicted to be slow, implying that greater reductions in both sulphur and nitrogen will be required if full recovery is to be achieved. Validation of MAGIC hindcasts The validation of acidification models such as MAGIC is limited to sites where long-term monitoring exists. Moreover, these data must cover a period of changing acidic deposition to allow for rigorous model validation (Wright et al. 1994). While MAGIC was calibrated to the mean chemistry from surveys conducted in 1997-1998, historical data exists for all sites from 1979, 1988, 1993, 1994, and 1996. A good correspondence between historical observations and simulated ANC, ph, and Ca was achieved at the most acid sites, and MAGIC successfully reproduced SO 4 trends at all sites. The records of diatom floras taken from loch sediment cores from Loch Grannoch, the Round Loch of Glenhead, and Loch Enoch provide evidence of a longer term record of the timing, rate and magnitude of biological and chemical change at these acid sensitive sites (Juggins et al. 1996) and provide a means to validate the historical ph trends in MAGIC (e.g. Jenkins et al. 1990). At Loch Grannoch the diatom reconstruction indicates a ph of 5.4-5.6 for the period prior to 1925, suggesting that MAGIC is over-estimating the pre-acidification ph of this site (Fig. 3a). Between 1925 and 1950 MAGIC shows good agreement with the rapid acidification inferred from the diatom reconstruction. However, after 1950 MAGIC predicts continued low ph (4.6-4.8) whereas the diatom data suggests a recovery followed by renewed acidification in the early 1970s. At the Round Loch of Glenhead there is generally poor agreement between the diatom and MAGIC inferred ph, with the diatom data suggesting a relatively gradual decline in ph from 5.3 in 1850 to 4.7 in 1980 (Fig. 3b). The core from Loch Enoch covers the period from the early 1920s to 1990 and the diatom data suggests that loch has had a relatively stable ph of 4.6-4.7 for the whole of this period (Fig. 3c). This agrees well with the MAGIC simulations for the period after 1950, but there is a discrepancy between inferences for the period 1925-1950 when MAGIC predicts a rapid acidification from 5.3 to 4.7 ph units. The comparison of MAGIC reconstructed ph from pre- and early-acidification (1850-1950) conditions with independent diatom evidence represents the only means of validation of both techniques since no time series of water chemistry exists for such a long period. Although the uncertainties in both approaches are large, close matches between diatom-inferred ph and MAGIC-simulated ph during the mid and late twentieth century increase the confidence in the predictive capacity of the MAGIC model. Biological Modelling The Water Framework Directive states that the ecological status of surface waters are defined on the basis to which they differ from reference conditions (defined as pristine conditions largely unaffected by human influences (Scottish Executive, 2001)). The biological status of seven sites was modelled by coupling the logistic regression equations to the MAGIC ANC predictions for fish, diatoms and macro-invertebrates from 1857 to 2016 and beyond. Biological predictions using MAGIC Historical data on trout populations is scarce for the majority of lochs, although there is anecdotal evidence of brown trout deformities and approximate dates of fish extinction for some of the lochs in Galloway that coincide with peak deposition measurements (Wright et al. 1994). Historical references confirm that Loch Enoch was frequented by anglers in the early twentieth century, but in 1978, 1984, and 1991 evidence of trout disappearing from the 6

loch were reported (Collen, et al. 2000). These records are consistent with the historical model predictions of fish status in Loch Enoch (Fig. 4a) where the healthy fish population declined from 0.75 in 1857 to extinction in 1970. The successful reintroduction of trout into a fishless Loch Enoch in 1995 followed by the capture of juvenile trout in 2000, indicate that recovery has proceeded to the point where a sustainable trout population has been reestablished (Collen et al. 2000; Harriman et al. 2001). By 2000, the modelled ANC ( 3.66 µeq l -1 ) is indicative of acid conditions and as a consequence, the probability of a healthy fish population is 0.31 (Fig. 4a). Future projection shows that the fish status at this site will not be restored to reference conditions by 2016. Similar trends are seen in the diatom and invertebrate predictions, with only partial recovery indicated by 2016. A similar pattern of biological loss and recovery is predicted for the Round Loch of Glenhead (Fig. 4c). The predicted historical fish status at Loch Riecawr was compared with a unique, 80 year angling catch record (1918-1998) (Fig. 4d). MAGIC simulates a rapid decline in the probability of a healthy trout population from 1950, which parallels the decline in angling success. More recently reductions in S deposition has lead to a recovery in the predicted fish status and catch records. By 2016, model predictions indicate that Loch Riecawr will support a totally healthy fish population, and since 48% of the catchment is afforested at this time, this recovery may be attributed to solely to the reductions in acid deposition (Fig. 4d). Although there is a strong relationship between angling success in the loch and modelled fish probabilities it is important to note that a healthy fish population is a function of egg survival and hydrochemistry of the inflow streams. With the implementation of the Gothenburg protocol a marked improvement in the biological status of all sites is predicted in 2016. Beyond 2030 a reversal in this trend is forecast at the afforested sites (Figs. 4b,d,e) as a consequence of base cation uptake and enhanced interception of acid deposition during the second rotation of forest, resulting in the renewed acidification of soils and surface waters. Acknowledgements This project was financially supported by the Commission of the European Communities RECOVER:2010 project (ENK1-CT-1999-00087), the Scottish Executive Environment and Rural Affairs Department (SEERAD) and the Department for Environment, Food and Rural Affairs (DEFRA) Freshwater Umbrella: Recovery of Acidified Waters (EPG 1/3/183). The authors thank M. Coull for creating the map of the study area. References Collen, P., Harriman, R., Morrison, B.R.S., Keay, E. & Watt, A.W. (2000) Restoration of a brown trout (Salmo Trutta.L) population to Loch Enoch, an acidified loch in Galloway, south-west Scotland. Freshwater Forum 14, 3-14. Cosby, B.J., Hornberger, G.M., Galloway, J.N. & Wright, R.F. (1985a) Modelling the effects of acid deposition: assessment of a lumped 3parameter model of soil water and stream-water chemistry. Wat. Resour. Res. 21, 51-63. Cosby, B.J., Hornberger, G.M., Galloway, J.N. & Wright, R.F. (1985b) Time scales of acidification: a quantitative model for estimating freshwater acidification. Environ. Sci. and Technol. 19, 1144-1149. Evans, C., Jenkins, A., Helliwell, R., Ferrier, R. & Collins, R. (2001) Freshwater acidification and recovery in the United Kingdom. Centre for Ecology and Hydrology- Wallingord/Macaulay Institute Report. ISBN 1 903741 01 7. Ferrier, R.C., Helliwell, R.C., Cosby, B.J., Jenkins, A. & Wright, R.F. (2001) Recovery from acidification of lochs in Galloway, south-west Scotland, UK: 1979-1998. Hydrol. and Earth System Sci. 5(3), 421-431. 7

Flower, R.J., Battarbee, R.W. & Appleby, P.G. (1987) The recent palaeolimnology of acid lakes in Galloway, south-west Scotland: Diatom analysis, ph trends, and the role of afforestation. J. Ecol. 75, 797-824. Harriman, R., Watt, A.W., Christie, A.E.G., Collen, P., Moore, D.W., McCartney, A.G., Taylor, E.M. & Watson, J. (2001) Interpretation of trends in acidic deposition and surface water chemistry in Scotland during the past three decades. Hydrol. and Earth System Sci. 5(3), 407-420. Henriksen, A., Lien, L., Rosseland, B.O., Traaen, T.S. & Sevaldrud, I.S. (1989) Lake acidification in Norway: present and predicted fish status. Ambio 18, 314-321. Hope, D., Billet, M., Milne, R. & Brown, T.A.W. (1997) Exports of organic carbon in British rivers. Hydrol. Proc. 11, 325-344. Jenkins, A., Whitehead, P.G., Cosby, B.J. & Birks, H.J. (1990) Modelling long term acidification: a comparison with diatom reconstructions and the implication for reversibility. Phil. Trans. Royal Soc. Lon. B327, 435-440. Juggins, S. (2001) The CLAM biological-chemical database: the development and application of biological models to predict taxon distributions from SSWC and MAGIC hydrochemical models. In: Summary of Research under DETR Contract "Acidification of freshwaters: the role of nitrogen and the prospects for recovery", (ed. By C.Curtis & G.Simpson), 123-146. ECRC Research Report No.79, Environmental Change Research Centre, University College London, EPG1/3/117. Juggins, S., Flower, R. & Battarbee, R. (1996) Palaeolimnological evidence for recent chemical and biological changes in UK Acid Waters Monitoring Network sites. Freshwater Biol. 36, 203-219. Monteith, D.T. & Evans, C. (Eds) (2000) The UK Acid Waters Monitoring Network:10-year Report. Analysis and interpretation of results, April 1988-March 1998. ENSIS Publishing, London. Scottish Executive (2001) Rivers, Lochs, Coasts: The future for Scotland s waters. Crown copyright 2001, ISNB: 0 7559 0143 6. Wright, R.F., Cosby, B.J., Ferrier, R.C., Jenkins, A., Bulger, A.J. & Harriman, R. (1994) Changes in acidification of lochs in Galloway, southwestern Scotland, 1979-1988: The MAGIC model used to evaluate the role of afforestation, calculate critical loads, and predict fish status. J.of Hydrol. 161, 257-285. 8

Table 1 Physical and chemical characteristics for seven sites in the Galloway region Loch Enoch Loch Grannoch Round Loch of Loch Riecawr Loch Macaterick Loch Knockstring Loch Cornish Glenhead Catchment information Grid Reference NX445 855 NX548 715 NX448 803 NX440 937 NX444 912 NX696 883 NX407 943 Catchment area 2.58 14.87 1.08 14.0 10.33 0.88 2.8 km 2 Loch area km 2 0.5 1.11 0.12 0.94 0.76 0.07 0.06 Altitude of 490 210 300 280 280 300 400 outflow m Forestry % 0 63 0 50 17 60 0 Geology Granite Granite Granite Granite 46% Greywackes 20% Organics 34% Organics 60% Granite 23% Greywackes 17% Greywackes 98% Organics 2% Greywackes Soil types PP 88% R 12% PP 38% PG 39% P 19% R 4% R 50% PP 50% PP 35% P 46% R 17% PG 2% P 63% PG 3% PP 9% R 25% PP 79% NCG 19% P 2% PP 80% SAP 20% Loch chemistry 1996-1997 ph 4.83 4.57 4.79 5.56 4.97 6.80 5.88 ANC (µeq l -1 ) -6.33-13.98 2.42 56.75 17.62 301.87 60.59 *SO 4 (µeq l -1 ) 42.32 74.33 55.34 59.05 57.24 108.63 48.34 Ca (µeq l -1 ) 22.83 50.63 34.14 72.88 43.08 294.50 55.69 NO 3 (µeq l -1 ) 17.95 30.66 21.70 11.81 12.00 18.17 8.21 Cl (µeq l -1 ) 152.15 213.9 168.10 204.75 216.30 259.40 160.30 DOC (mg l -1 ) 1.79 5.14 3.22 4.43 5.66 5.52 5.31 Total Al (µg l -1 ) 69.00 239.00 110.00 90.00 118.00 21.00 65.00 Labile Al (µg l -1 ) 54.00 184.00 79.00 57.00 78.00 2.00 4.00 Non labile Al (µg l -1 ) 16.00 55.00 31.00 33.00 40.00 19.00 61.00 PP-Peaty podzol R-Ranker PG-Peaty gley P-Peat DOC-Dissolved Organic Carbon *SO 4 -Non marine sulphate ANC-Acid Neutralising Capacity 9

Table 2: Observed (mean 1996-1997) and simulated (1997) surface water and soil chemistry (All units in µeq l -1 except for ph). ANC ph SO 4 Ca Obs. Calib. Obs. Calib. Obs. Calib. Obs. Calib. Loch Enoch -6.33-7.33 4.83 4.85 58.0 57.75 22.83 22.77 Round Loch of 2.42 2.38 4.79 4.87 72.65 72.43 34.14 34.12 Glenhead Cornish 60.59 59.65 5.88 5.31 64.86 64.70 55.69 55.80 Macaterick 17.62 17.86 4.97 4.88 79.52 78.78 43.08 43.32 Riecawr 56.75 59.42 5.56 5.49 80.14 78.49 72.88 74.09 Loch Knockstring 301.87 309.28 6.80 7.13 135.35 128.30 294.50 294.98 Loch Grannoch -13.98-10.42 4.57 4.65 96.37 92.98 50.63 50.69 List of Figures Figure 1. The study area in Galloway, south-west Scotland, showing sample sites and the Cairnsmore of Fleet (south) granite pluton. 10

a) 6.5 6 5.5 ph 5 4.5 4 1850 1875 1900 1925 1950 1975 2000 2025 2050 MAGIC simulated ph Observed ph Inferred ph b) 6.5 6 5.5 ph 5 4.5 4 1850 1875 1900 1925 1950 1975 2000 2025 2050 MAGIC Simulated ph Observed ph Inferred ph c) 6.5 6 5.5 ph 5 4.5 4 1850 1875 1900 1925 1950 1975 2000 2025 2050 MAGIC Simulated ph Observed ph Inferred ph Figure 2: MAGIC simulated ph and diatom inferred loch ph for a) Loch Grannoch, b) R.L. Glenhead and c) Loch Enoch. 11

ANC (µeql -1 ) 400 350 WFD 300 250 200 150 100 50 0-501850 1875 1900 1925 1950 1975 2000 2025 2050-100 Loch Enoch Loch Grannoch R.L.Glenhead Loch Riecawr Loch Macaterick Loch Knockstring Loch Cornish Critical ANC for fish survival Figure 3. Long-term reconstruction of surface water ANC at the seven sites. 12

)a Loc h Enoc h )b Loch Grannoch 1 1 100 0.8 WFD 0.8 80 0.6 0.6 60 0.4 0.4 40 0.2 0.2 20 0 1850 1900 1950 2000 2050 Healthy Fish A. minutissima Baetis 0 0 1850 1900 1950 2000 2050 Healthy Fish A. minutissima Baetis Forest area )c 1 0.8 0.6 0.4 R.L.Glenhead )d 1 0.8 0.6 0.4 Loch Riecawr 300 250 200 150 100 0.2 0.2 50 0 1850 1900 1950 2000 2050 0 0 1850 1900 1950 2000 2050 Healthy Fish A. minutissima Healthy Fish A. minutissima Baetis Baetis Forest area Trout catch )e Loch Macaterick )f Loch Knockstring 1 100 1 100 0.8 80 0.8 80 0.6 60 0.6 60 0.4 40 0.4 40 0.2 20 0.2 20 0 1850 1900 1950 2000 2050 0 0 1850 1900 1950 2000 2050 0 Healthy Fish Baetis A. minutissima Forest area Healthy Fish Baetis A. minutissima Forest area )g Loch Cornish 1 0.8 0.6 0.4 0.2 0 1850 1900 1950 2000 2050 Healthy Fish A. minutissima Baetis Figure 4. Probability of occurrence for an acid sensitive diatom (Achnanthes minutissima), acid sensitive macroinvertebtrate (Baetis rhodani), and fish population status as reconstructed by MAGIC from 1857 to 2047. (Source of trout catch data, Harriman et al. 2001). 13

Flow characteristics of natural soil liners for the containment of landfill leachate and agricultural slurries Purcell P. 1, Scully, H 1 and Gleeson, T 2 1 Civil Engineering Department, University College, Earlsfort Terrace, Dublin 2. 2 Teagasc, Kinsealy, Co. Dublin. Abstract The paper describes a laboratory study of the suitability of Irish soils for use in the lining of landfill sites and animal waste storage tanks. The soils were taken from a number of locations and are typical of those found throughout Ireland. In the paper, laboratory investigations of the flow characteristics of the soils are presented. Basic soil tests, such as, for example, particle size distribution, were found to give a good indication of the suitability of the soils for lining municipal leachate or agricultural slurry containment facilities. In addition, agricultural slurries were found to cause a very significant reduction in the effective soil permeability due to the deposition of solids on the soil surface and within the pores of the soil. Some field results showing the performance of an earth-bank tank for the storage of agricultural slurry and the effects of seepage from the tank on groundwater quality are presented. Résumé Le présent article présente une étude sur le potentiel d utilisation des sols d Irlande pour imperméabiliser les sites de décharge de déchets et les réservoirs de stockage des déchets animaliers. Les sols ont été tirés d un certain nombre de lieux et sont représentatifs des sols que l on trouve à travers le pays. On présente ici les expériences portant sur les caractéristiques des sols relatives à leur perméabilité. Des tests, par exemple sur la distribution de la taille des particules, indiquent la possibilité d utiliser les sols pour imperméabiliser les produits contaminés des décharges municipales ou les fumiers. En outre, on a remarqué que les fumiers ont une capacité à réduire la perméabilité effective des sols par un méchanisme de déposition de particules solides à la surface et dans les pores des sols. On présente également les résultats d une étude de terrain sur l aptitude d un réservoir de terre pour le stockage de boues d origine agricole et sur les effets du réservoir sur la qualité des eaux souterraines. INTRODUCTION As part of a study into the suitability of Irish soils for use in the lining of landfill sites and animal waste storage tanks, four typical soils were selected and subjected to a series of laboratory tests. The soils were taken from four separate locations and are typical of those found throughout Ireland. In the paper, laboratory studies relating to (a) the flow of water and (b) the flow of slurry through soils are presented. SOIL-WATER EXPERIMENTS 1

Some of the salient details of the soils and the sites studied are given on Table 1. Soil Glacial origin Natural water content (%) Liquid limit (%) Plasticity index (%) Carlow Alluvial 20 34 16 Moorepark Fluvio-glacial 25 24 11 Grange Fluvio-glacial 14 17 15 Dublin till Lodgement till 10 26 13 Table 1 Details of soils and sites studied Particle size distribution curves for the materials are shown on Figure 1. All are reasonably well graded (a typical characteristic of Irish glacial soils), with the Carlow alluvial material having the highest fines content. The Moorepark and Grange fluvio-glacial deposits are somewhat gap-graded in the sand range with the Moorepark material having the greatest coarse content. Percentage passing 100% 75% 50% 25% Carlow Moorepark Grange Dublin 0% 0.0001 0.001 0.01 0.1 1 10 100 Particle size (mm) Figure 1 Particle size distribution chart No specific standard exists for the assessment of soils suitable for use in animal waste storage tanks. However reference can be made to a standard for landfill sites. Typical suitable ranges of parameters for use as landfill clay liners are shown on Table 2 (EPA, 2000, McCullen and Long, 1999). Table 2 Range of acceptable soil properties for use as clay liner in landfill sites Number Property Units Acceptable range 1 Permeability m/s < 1 x 10-9 2 Clay content (< 2 μm) % 10 3 Fines content (< 63 μm) % 20-30 4 Gravel content (> 2 mm) % 30 5 Maximum particle size mm 25-50 6 Plasticity index % 10-30 7 Liquid limit % 90 2

All of the soils pass criteria numbers 2 to 4, albeit with the Moorepark and Grange soils being borderline on one case and the Dublin till being borderline in two cases. All the soils pass criterion 5 as these particles were removed prior to testing. This obviously also could be achieved during pre-treatment on site. All the soils pass criteria 6 and 7, with again the Moorepark and Dublin till being borderline in one case. A series of laboratory triaxial permeability (BS1377) tests were then carried out on the soils. The specimens were prepared by standard Proctor compaction to approximate in-situ density with 2.5 kg ram. The objective of the tests was to check whether the method of placement influenced the permeability. In order to do this either the bulk density or moisture content values were systematically varied while keeping the other parameter constant. 1.0E-07 Dublin Moorepark Grange Carlow Permeability (m/s) 1.0E-08 1.0E-09 1.0E-10 Target value 1.8 1.9 2 2.1 2.2 2.3 2.4 Bulk density (Mg/m 3 ) Figure 2 Variation of soil permeability with bulk density Carlow Dublin Moorepark Grange 1.0E-07 Permeability (m/s) 1.0E-08 1.0E-09 1.0E-10 Target value 5 7.5 10 12.5 15 17.5 20 22.5 25 Moisture Content (% ) Figure 3 Variation of soil permeability with moisture content The variations in permeability values with bulk density are shown on Figure 2. The soil specimens were compacted at their natural moisture contents, increasing bulk density being achieved for each soil specimen by increasing the number of layers compacted. Regardless of the bulk density, the fine Carlow material, not surprisingly, is acceptable in all cases. The coarser Grange material is also insensitive to bulk density but has a permeability value closer to the target value. All tests on the coarse Moorepark material fail the criterion. Only the Dublin till material appears to have a permeability which is sensitive to bulk density (this is for the reason that the material is well graded). Beyond a bulk density of about 2.3 Mg/m 3 the 3

material has an acceptable permeability value. However, in order to achieve density values such as this in the field, very heavy compaction plant would be required. In the case of the permeability/moisture content experiments shown in Figure 3, the natural moisture content of each soil specimen was altered by adding varying amounts of water and the soil was then compacted in three layers. As only relatively few data points are available, it is difficult to make any definitive conclusions regarding the relationship between moisture content and permeability. For the range of data available, again, only the Carlow material passes the acceptability criterion. The values obtained for the other material show little sensitivity of permeability to moisture content and generally fail the criterion. Slurry-Soil Experiments Researchers (Culley et al.,1982) have established, in full-scale field trials, that animal slurries in contact with soil do cause a progressive sealing of the soil with time, thereby significantly reducing the effective permeability of the soil. The laboratory measurement of the permeability of porous media at very low values presents significant difficulties for conventional laboratory geotechnical equipment, such as the permeameter or manually operated triaxial apparatus, and requires the use of sophisticated computer-controlled apparatus such as that used in the soil-water experiments described above. However, animal slurries are unsuitable for use in such sophisticated apparatus due to problems with clogging of lines and potential damage to sensors. To overcome these difficulties, the specific resistance to filtration (srf) apparatus (Coackley and Jones, 1956) used in the laboratory assessment of the dewaterability of water and wastewater treatment sludges was adapted to study the flow of animal slurry through porous media (flow through a sludge cake and the filter medium is analagous to flow through a porous medium). In the srf test (see Fig. 4(a)), an aliquot of the sludge is placed onto a filter paper (Whatman no. 1,) in a Buchner funnel (70 mm diameter) and drawn through by vacuum pump; the cumulative filtrate volume is recorded as a function of time. In adapting the srf test to the case of animal slurry flowing through soil, it was considered to be inappropriate to draw the slurry through the soil by suction, because of the potential problem of air binding within the porous medium at subatmospheric pressures. Instead, the srf test apparatus was modified to enable a positive pressure to be applied to the upper surface of soil specimen, while maintaining the lower surface at approximate atmospheric pressure (Figure 4 (b)). 4

Applied pressure 1 bar PVC pipe Slurry Soil specimen 100 mm deep x 100 mm diameter Slurry Buchner funnel 70 mm diameter Filter paper Vacuum pressure -0.5 bars Filtrate Filtrate Figure 4 (a) SRF apparatus (b) Modified SRF apparatus A typical SRF result in the case of a cattle slurry with a solids concentration of 1.5% is presented in Figure 5. Cumulative filtrate volume (ml) 20 15 10 5 0 0 20 40 60 80 Time (minutes) 100 Figure 5 Slurry flow through laboratory filter paper 5

A typical result for the flow of the same cattle slurry through 100 mm deep by 100 mm diameter specimen of Leighton-Buzzard sand in the modified SRF apparatus is presented in Figure 6. Cumulative filtrate volume (ml) 1000 750 500 250 0 Fitted curvefitted curve 0 20 40 60 Fitted 80 100 120 140 curve Time (minutes) Figure 6 Slurry through Leighton Buzzard sand Examination of Figures 5 and 6 shows that, although there is only a difference of a factor of two in plan areas of the respective test specimens, there is a considerably larger volume of filtrate in the case of the slurry flow through the sand specimen. Clearly, the filter paper and the slurry cake formed on the surface of the filter offers much greater resistance to the flow than does the much more porous sand. In the case of the filter paper, all the suspended solids in the slurry were retained on the surface of the filter, whereas in the case of the sand, some penetration of the slurry solids into the pores of the sand occurred and resistance to flow was a combination of both the surface cake and sand bed. The sand specimen was housed in a clear walled pvc pipe and visual inspection corroborated this observation (i.e. deposition of solids on sand surface and within the sand bed). Examination of Figure 6 would appear to suggest an exponential flow decay through the sand bed, and a best-fit curve of this form is indicated on the figure, which is described by the following equation: V = 885(1 - e -0.08t ) where: V = cumulative filtrate volume (ml) t = time from start of filter run (minutes). dv The volumetric flow rate at any time t ( ) is therefore: 70.8e -0.08t (ml/minute) dt The corresponding computed effective permeability (k) of the sand specimen at any time t is presented in Figure 7. 6

Effective permeability (m x10-9 ) 1500 1000 500 0 0 20 40 60 80 100 120 140 Time (minutes) Figure 7 Effective permeability of Leighton-Buzzard sand due to progressive sealing by slurry Field Measurements Some preliminary results of a field investigation into the suitability of using an earth-bank tank for the storage of animal slurry is presented in this section. A typical cross-section through an earth-bank tank is presented in Figure 8. Original ground level Slurry level 1 m Figure 8 Typical cross-section through an earth-bank tank The experimental results presented below relate to an earth-bank tank constructed at the Teagasc Grange research station. A plan view showing the locations of groundwater monitoring boreholes is presented in Figure 9. The soil characteristics at this site have previously been presented in Table 1 and Figures 1, 2 and 3. Borehole 8a Borehole 4b 30 m Direction of groundwater flow 7

Figure 9 Plan view of earth-bank tank at Grange and locations of boreholes The earth-bank tank was filled with slurry over the three-day period 14 February to 16 February 2002. Following this date, the rate of infiltration of slurry through the bottom and sides of the tank was monitored by using a hook gauge to record the drop in the slurry level in the tank and the readings were corrected for rainfall (measured using standard gauge) and evaporation losses (measured using a Class A pan). There are no specific international standards in respect of permissible infiltration from earth-bank tanks to the adjacent groundwater. For example, a number of states in the United States specify maximum seepage rates ranging from 0.42 mm/day to 6.3 mm/day (Parker et al, 1999). Typical infiltration data, following filling of tank with slurry, is presented in Fig. 10 and the groundwater levels indicated in the boreholes 8a and 4b shown in Fig. 9 are presented in Fig. 11. 1.5 Infiltration (mm/day) 1 0.5 0 05/03/02 10/03/02 15/03/02 20/03/02 25/03/02 30/03/02 04/04/02 09/04/02-0.5 Date -1-1.5 Figure 10 Infiltration data for earth-bank tank Filling of tank 0 Water Level (m BOGL) 0.5 1 1.5 2 4B 8A 2.5 18/12/01 01/01/02 15/01/02 29/01/02 12/02/02 26/02/02 12/03/02 26/03/02 09/04/02 23/04/02 Figure 11 Groundwater levels for two boreholes (4b and 8a) [m BOGL = metres below original ground level] 8

The normal groundwater quality parameters (BOD 5, nitrate, phosphate, ammonia, conductivity etc.) were monitored before and after filling the tank with slurry. Some sample data are presented in Fig. 12. Once again, there is no specific standard relating to groundwater quality adjacent to earth-bank tanks but, for example, the E.U. Nitrates Directive stipulates that the nitrate concentration in freshwater and groundwater should not exceed the limit of 50 mg NO 3 per litre (11.3 mg NO 3 -N per litre). Examination of Fig. 12 shows that the NO 3 -N levels are less than 1 mg per litre approximately two weeks after tank filling and therefore well within the Nitrates Directive limit. 1 Filling of tank NO 3 -N (mg/l) 0.1 0.01 0.001 05/12/01 15/12/01 25/12/01 04/01/02 14/01/02 24/01/02 03/02/02 13/02/02 23/02/02 05/03/02 8A 4B Filling of tank P (mg/l) 1 0.1 0.01 8A 4B 0.001 05/12/01 15/12/01 25/12/01 04/01/02 14/01/02 24/01/02 03/02/02 13/02/02 23/02/02 05/03/02 1 Filling of tank NH 4 -N (mg/l) 0.1 0.01 8A 4B 0.001 05/12/01 15/12/01 25/12/01 04/01/02 14/01/02 24/01/02 03/02/02 13/02/02 23/02/02 05/03/02 Figure 12 Groundwater quality adjacent to earth-bank tank (boreholes 8a and 4b) 9

Conclusions Basic soil tests such as, for example, particle size distribution give a good indication of the suitability of the soils for lining municipal leachate or agricultural slurry containment facilities. However, if a soil is only marginally within the specification in respect of such basic soil parameters it may not be acceptable from a permeability point of view, unless very large compaction plant is available. Agricultural slurries cause a very significant reduction in the effective soil permeability due to the deposition of solids on the soil surface and within the pores of the soil. Preliminary results for an earth-bank tank constructed for the temporary storage of agricultural slurry show that there has been no significant deterioration in the quality of adjacent groundwater. Bibliography Coackley, P. and Jones, B.R.S (1956) Vacuum Sludge Filtration, Sewage and Industrial Wastes, 28, 8, pp.963-976. Culley, J.L.B. and Phillips, P.A. (1983) Sealing of soils by liquid cattle manure. Canadian Journal of Agricultural Engineering, 24: 87 90. EPA Ireland (2000), Landfill site design, ISBN 1 84095 026 9. McCullen, P and Long, M (1999) Arthurstown landfill facility - geotechnical site characterisation. Proc Seminar. on Arthurstown Landfill held at the Institution of Engineers of Ireland 9/3/99. Also published in Trans. IEI 1998 / 99. Parker, D.B., Schulte, D.D., Eisenhauer, D.E. (1999) Seepage from earthen animal waste ponds and lagoons an overview of research results and state regulations. Vesilind, P.A. (1975) Treatment and disposal of wastewater sludges, Ann Arbor Science, Ann Arbor, Michigan. 10

Water quality modelling of an agricultural field: from soil surface to groundwater Alaa El-Sadek a, Ingeborg Joris b and Jan Feyen c a Dr., Institute for Land and Water Management, K.U.Leuven, Vital Decosterstraat 102, B-3000 Leuven, Belgium b Ph.D. student, Institute for Land and Water Management, K.U.Leuven, Vital Decosterstraat 102, B-3000 Leuven, Belgium c Prof., Institute for Land and Water Management, K.U.Leuven, Vital Decosterstraat 102, B-3000 Leuven, Belgium Abstract In this study the water flow and nitrate transport to a subsurface drain, using a simplified and detailed model, are simulated for the specific hydro-geological conditions of Elverdinge and Assenede, Belgium. Previously, the DRAINMOD-N model proved to be able to simulate nitrate concentrations and drainage well for an in-situ leaching experiment, the 'Hooibeekhoeve' in the community of Geel (north-eastern part of Belgium), conducted in 1992-1995 (El-Sadek et al., 2002). In this study, the calibrated model is used to simulate the nitrate leaching for the winter period 2000-2001 in Elverdinge and Assenede and is compared to a model with a simplified nitrate transport description. The comparative analysis between both model approaches reveals that the simplified model is able to predict sufficiently accurate the observed nitrate leaching. The detailed approach however has the advantage of given a more accurate estimate of the nitrogen mineralization, N deposition and denitrification, resulting in a more precise modeling of the nitrate leaching to surface waters and groundwater. Keywords: subsurface drainage, nitrogen cycle, nitrate leaching, water quality, 1. Introduction Computer simulation models are useful tools to evaluate the complex mechanisms of nitrogen transport and transformation in agricultural fields (Brevé et al., 1997). In the fall-winter-spring period in Belgium, significant NO 3 -N losses can occur due to leaching of nitrate that remains in the soil after harvest. Mineralization of organic nitrogen in soil, organic material, plant residue or manure, in combination with the rainfall excess increases in this period the leaching of NO 3 -N. In order to meet the EU-norm of 11.3 mg NO 3 -N l -1 in surface and groundwater, the Flemish Government in Belgium states that the residual mineral nitrogen, measured in the soil profile of cropland (0-90 cm), may not exceed 90 kg N ha -1 between the 1 st of October and the 15 th of November (Geypens et al., 2001). The value of this residual mineral nitrogen content is still under debate. The objective of this article is to evaluate the performance of a simplified and detailed model approach to predict the nitrate transport to a subsurface drain, with application to the specific hydro-geological conditions of Elverdinge and Assenede, Belgium in the fall-winter period 2000-2001. 1

2. Field data Elverdinge and Assenede are experimental fields situated in Flanders, Belgium. Soil physical properties were determined for each soil horizon, using undisturbed core samples. The soil types in the fields are classified as a sandy loam and clay soils respectively. The groundwater level fluctuates between 40 and 80 cm below surface. The fields were left fallow during the fall-winter period. Soil samples were taken with three weeks intervals in layers of 30 cm, up to a depth of 90 cm. Nitrate concentration of the soil water was determined using suction cups at a depth of 90 cm. Samples of drainage water and groundwater were also analyzed for nitrate. Also, the mineralization rate and denitrification capacity of the soils were measured to get a better estimation of the nitrate leaching. The soil moisture content was measured at several depths in the soil profile, taking with an auger soil samples in the layers 0-30 cm, 30-60 cm and 60-90 cm. From the wet and dry weight of the soil samples the water content was derived. The fields are equipped with a subsurface drainage system consisting of parallel, 10 cm diameter, corrugated plastic drains, placed at a depth of 70 cm below surface, and spacings of 7 m in Elverdinge and 12 m in Assenede. The soil profile was assumed to have a depth of 4 m, on top of an impermeable layer. The properties of the two fields are shown in Tables 1 and 2. Table 1: Drain properties of the fields in Elverdinge and Assenede Elverdinge Assenede Drain depth (cm) 70 70 Drain spacing (cm) 700 1200 Effective diameter (cm) 2.5 2.5 Depth of soil profile (cm) 400 400 Soil physical properties were determined for each distinguishable soil horizon, using undisturbed soil samples taken with Kopecky rings. van Genuchten-Mualem parameters for describing the hydraulic functions (van Genuchten and Nielsen, 1985) were fitted on both water retention and multi-step outflow data, using the multi-step outflow program (van Dam et al., 1990). Basic water retention and hydraulic conductivity curves were established by averaging individual curves for each soil layer. Table 2: Soil hydraulic properties for the studied horizons θ r (cm 3 cm -3 ) θ s (cm 3 cm -3 ) α (cm -1 ) Elverdinge n (-) K s (cm d -1 ) 0-30 cm 0.0001 0.3811 0.0067 1.2726 5.76 0.5 30-60 cm 0.0001 0.3891 0.0045 1.2381 22.62 0.5 60-400 cm 0.0001 0.3756 0.0106 1.2457 36.42 0.5 l (-) Assenede 0-30 cm 0.0001 0.3746 0.0043 1.2837 425 0.5 30-60 cm 0.0524 0.3554 0.0058 2.2185 393 0.5 60-400 cm 0.0001 0.3565 0.0070 1.5984 25 0.5 2

3. Theory 3.1 Simplified model HYDRUS-2D (Simúnek et al., 1999) was used in this study as a simplified model to describe NO 3 -N transport and leaching without any production, transformation or exchange taking place. The HYDRUS-2D model numerically solves water flow and solute transport equations in a twodimensional flow domain in the unsaturated-saturated zone using the Galerkin finite element method. Water flow is calculated solving the Richards' equation, which can be written as: θ =.( K h + K z) S t (1) where θ is the volumetric water content [L 3 L -3 ], t is time [T], a vector differential operator [L - 1 ], K the hydraulic conductivity [L T -1 ], h is pressure head [L], z is gravitational head [L], and S a sink term accounting for root water uptake [L 3 L -3 T -1 ]. K and S can be functions of position, θ or h, and time. In this work, K is assumed to be isotropic. For a conservative solute, transport is described by the advective-dispersive equation as: ( θc) t =.( qc θd c) S c (2) where c is solute concentration [M L -3 ], q is volumetric water flux density vector [L 3 L -2 T -1 ], D is the dispersion tensor [L 2 T -1 ] and S c a sink term accounting for root solute uptake [M L -3 T -1 ]. The components of D are given by (Bear, 1972): qiq j θ D = θdij = DT qδ ij + ( DL DT ) + θddτδ ij (3) q with D ij components of the dispersion tensor [L 2 T -1 ], D L the longitudinal dispersivity [L], D T the transversal dispersivity [L], D d the molecular diffusion coefficient in free water [L 2 T -1 ], q i the i- component of q [L 3 L -2 T -1 ], q the absolute value of the water flux density [L 3 L -2 T -1 ], τ the tortuosity factor [-], and δ ij the Kronecker delta function. Equations (1) and (2) can be solved if initial and boundary conditions for a given problem are known. In this work, a fall-winter leaching period was studied so no root water uptake or plant NO 3 -uptake were considered (S=0; S c =0). 3.2 Detailed model DRAINMOD (Skaggs, 1981) was used as a detailed model to simulate nitrogen transformation within the soil profile. DRAINMOD is a computer model that was developed to simulate the performance of drainage and related water table management systems. DRAINMOD-N (Brevé et al., 1997) is an add-on module to DRAINMOD for simulating the nitrogen dynamics in artificially drained soils. Nitrate-nitrogen (NO 3 -N) is the main N pool considered. The 3

ammonium-nitrogen pool is ignored because in most soils ammonium nitrifies quickly or stays fixed to the soil; thus ammonium losses in subsurface drainage can be neglected. The controlling processes considered by the model (Brevé et al., 1992) are rainfall deposition, fertilizer dissolution, net mineralization of organic nitrogen, denitrifcation, plant uptake, and surface runoff and subsurface drainage losses. Assuming one-dimensional (vertical) flow processes in the unsaturated zone the nitrogen cycle can be represented by the advective-dispersive-reactive (ADR) equation: ( θc ) C ( qc) = θd + Γ t z z z (4) Where: C is the NO 3 -N concentration [M L -3 ], θ is the volumetric water content [L 3 L -3 ], q is the vertical water flux [L T -1 ], D is the coefficient of hydrodynamic dispersion [L² T -1 ], Γ is a source/sink term [M L -3 T -1 ] used to represent additional processes (plant uptake, transformations, etc.), z is the coordinate direction along the flow path [L], and t is the time [T]. The coefficient of hydrodynamic dispersion is defined as follows: q D = λ + τd* (5) θ Where λ is dispersivity [L], τ is a dimensionless tortuosity factor, and D* is the molecular diffusion coefficient [L² M -1 ]. Assuming z is positive in the downward direction and water flows downward in the soil profile, in DRAINMOD-N, functional relationships are used to quantify processes other than NO 3 -N transport (the source/sink term in Eq. 4), as follows (Brevé et al., 1998): Γ =Γ +Γ + Γ Γ Γ Γ (6) dep fer mnl rnf upt den Where: Γ dep stands for rainfall deposition [M L -3 T -1 ], Γ fer for fertilizer dissolution [M L -3 T -1 ], Γ mnl for net mineralization [M L -3 T -1 ], Γ rnf for loss [M L -3 T -1 ] in surface runoff, Γ upt for plant uptake [M L -3 T -1 ], and Γ den for denitrification [M L -3 T -1 ]. A detailed description of each functional relationship is given by Brevé et al. (1997). 4. Materials and methods 4.1 Model input Water flow and nitrate leaching are modeled in the flow domain of the drain spacing, and a depth of the soil profile of 4 m. The bottom of the soil profile was assumed to be impermeable. The drain was located at 70 cm depth, and was described as a half circular hole with real physical dimensions. The inner wall of the drain was described as a seepage face, implying that the drain is always practically empty. The models were applied to simulate the lateral subsurface drainage, 4

groundwater level and nitrate leaching for the fall-winter period 2000 2001 in Elverdinge and Assenede. In the simulations no ponding at the soil surface was allowed. As initial condition the measured nitrate concentrations were used (Table 3). For both fields, the flow domain is defined as a 400 cm deep soil profile with a width of one half drain spacing. Since flow on both sides of the drain is symmetrical, describing flow from the middle of the drain to one half of the drain spacing is sufficient. The lower, left and right boundaries are defined as no flow-boundaries, except for the position of the half drain tube at the left boundary. The drain tube is defined as a seepage face, through which water and solutes can leave the flow domain when saturated conditions occur (de Vos, 1997). The simulation period ranged from October 1, 2000 to March 31, 2001. Daily rainfall and evapotranspiration were used for the upper boundary. Table 3: Initial conditions for groundwater and nitrate concentrations Elverdinge Assenede Groundwater level -60-77 NO 3 (kg N/ha) 0-30 cm 70.0 54.8 30-60 cm 77.8 40.7 60-90 cm 20.1 10.7 90-120 cm 60.2 10.7 120-150 cm 100.5 10.7 150-400 cm 65.3 10.7 4.2 Model parameters for HYDRUS-2D In HYDRUS-2D, the flow domain is divided into a network of triangular elements that are smaller close to the surface. The denser grid close to the surface will allow more rapid changes in water and solute content, which can be expected under natural meteorological conditions. The flow domain is divided in subdomains so separate water and NO 3 N mass balances could be calculated for the layers 0-30 cm, 30-60 cm and 60-90 cm below surface to obtain water content and NO 3 -N content values that could be compared to the measurements. Groundwater levels were calculated at different distances of the drain and then averaged over the width of the flow domain to compare to measured levels. The transport parameters used in HYDRUS-2D are shown in Table 4. Table 4: Transport parameters used in HYDRUS-2D calculations Elverdinge Assenede longitudinal dispersivity D L (cm) 10 10 transversal dispersivity D T (cm) 10 10 diffusion coefficient D d (cm 2 d -1 ) 4.32 4.32 5

4.3 Model parameters for DRAINMOD-N Nitrogen movement parameters in DRAINMOD-N Nitrogen-related parameters required in DRAINMOD-N include standard rate coefficients for denitrification (K den ) and net mineralization (K min ), soil dispersivity (λ), the nitrogen content in rain and other inputs: Dispersion parameters - Soil dispersivity: 10 cm - Tortuosity factor: 1 - Diffusion coefficient: 0.01 Process rate constants - Rate coefficient of net mineralization: Elverdinge: 0.000480 d -1 Assenede: 0.000434 d -1 - Rate coefficient of denitrification: Elverdinge: 0.01 d -1 Assenede: 0.03 d -1 - Continuous days of saturation for denitrification: 3 days Model input - Damping depth: 50 cm - Nitrate concentration of rain: 0.8 mg l -1 There is no yield reduction by soil water stress. 5. Results and discussion 5.1 Water flow To ensure a good modeling of the nitrate leaching a good water table prediction is a necessity (Amatya et al., 1999). Therefore in the first step of the analysis the subsurface drainage discharge and the related groundwater level were modeled for the simulation period October 1, 2000 to March 31, 2001, being in Belgium the leaching period. Comparison of measured and simulated results for water table level in Elverdinge field is shown in Fig. 1. Assuming that the water is the vehicle needed to carry nitrate within the soil profile, the results of the water quantity modeling as presented in Fig. 1 ensure a good nitrate transport and leaching as a second step. During the simulation period in Elverdinge field, the total amount of rainfall, subsurface drainage and evapotranspiration were 61.3, 50.6 and 11.3 cm, respectively. As expected, there is a high subsurface drainage and a low ET because the simulation period was the fall-winter period. Under extremely wet conditions, the major part of the drainage water originated from the topsoil. Under dry conditions with a relatively deep phreatic surface drainage water mainly originated from the zone close to the drain depth. In general both models (simplified and detailed) simulated quite accurately the hydrological variables. 6

Date (day) 10/01/00 10/31/00 11/30/00 12/30/00 01/29/01 02/28/01 03/30/01 0-20 Ground w ater level (cm) -40-60 -80-100 -120-140 Measured DRAINMOD-N HYDRUS_2D Figure 1: Simulated and measured groundwater level in Elverdinge 5.2 Nitrate transport and leaching Groundwater levels simulated with both approaches are well. Simulated and measured drain discharge rates in time correspond well for both model approaches. Nitrate concentrations in the soil profiles as shown in Figs. 2 and 3 for Elverdinge and Assenede respectively match well with the measurements. For the Assenede field, the DRAINMOD-N model predicts higher nitrate concentrations in the soil profile. This is probably due to N-mineralisation, which is not calculated by the HYDRUS-2D model. Nitrate concentrations in the drainage water simulated with HYDRUS-2D (simplified model to calculate the nitrate leaching) are initially higher for both fields. This could be an overestimation because denitrification is not taken into account. The changes in NO 3 -N concentrations in the drainage water in time are simulated well. Nitrate concentrations in the drainage water simulated with DRAINMOD-N (detailed model to calculate the nitrate leaching) and measured correspond very well. At high infiltration rates the NO 3 -N just above the drain is transported downwards rapidly. The measured, simplified simulated (HYDRUS-2D) and detailed simulated (DRAINMOD-N) are show in Figs. 4 and 5 for Elverdinge and Assenede respectively. The results indicate that to improve the description of N leaching, N mineralization, N deposition and denitrification have to be accounted for to allow a better description of N leaching to surface waters and groundwater. 7

120 100 80 HY DRUS_2D Measured DRA INMOD 60 40 20 0 10/01/00 10/31/00 11/30/00 12/30/00 01/29/01 02/28/01 03/30/01 NO 3 -N (kg ha -1 ) 120 100 80 60 40 20 0 10/01/00 10/31/00 11/30/00 12/30/00 01/29/01 02/28/01 03/30/01 120 100 80 60 40 20 0 10/01/00 10/31/00 11/30/00 12/30/00 01/29/01 02/28/01 03/30/01 Date (day) Figure 2: NO 3 -N content in the soil profile in different layers (0-30, 30-60 and 60-90 from top) in Elverdinge 8

120 100 80 HY DRUS_2D Measured DRA INMOD 60 40 20 0 10/01/00 10/31/00 11/30/00 12/30/00 01/29/01 02/28/01 03/30/01 NO 3 -N (kg ha -1 ) 120 100 80 60 40 20 0 10/01/00 10/31/00 11/30/00 12/30/00 01/29/01 02/28/01 03/30/01 120 100 80 60 40 20 0 10/01/00 10/31/00 11/30/00 12/30/00 01/29/01 02/28/01 03/30/01 Date (day) Figure 3: NO 3 -N content in the soil profile in different layers (0-30, 30-60 and 60-90 from) in Assenede 9

100 90 80 70 DRAINMOD Measured HYDRUS NO 3 -N (mg l -1 ) 60 50 40 30 20 10 0 1/10/00 31/10/00 30/11/00 30/12/00 29/01/01 28/02/01 30/03/01 Date (day) Figure 4: Simulated and measured nitrate-nitrogen (mg l -1 ) in shallow groundwater during the fall-winter season in Elverdinge 100 90 80 70 Measured DRAINMOD-N HYDRUS_2D NO 3 -N (mg l -1 ) 60 50 40 30 20 10 0 10/01/00 10/31/00 11/30/00 12/30/00 01/29/01 02/28/01 03/30/01 Date (day) Figure 5: Simulated and measured nitrate-nitrogen (mg l -1 ) in shallow groundwater during the fall-winter season in Assenede 6. Conclusions The simulation results indicate that both simplified and detailed approaches described the dynamics of the water balance well. The layering of the soil profile had a pronounced effect on the flow paths to the drain under different atmospheric conditions. The nitrogen dynamics play an 10

important role in determining actual nitrate concentrations in the soil profile. Considering NO 3 -N as a conservative solute (as in HYDRUS-2D) leads to a higher estimate of NO 3 concentrations in the drainage water. Field data of the N balance show that also during winter periods considerable increases in NO 3 -N discharge occurred due to net mineralization and deposition especially during relatively dry periods. Other processes like immobilization and denitrification make the description of the N dynamics even more complex. The simplified approach is acceptable to give a general overview to predict the nitrate leaching. The detailed approach is required to give the decision maker the opportunity to take the right decision where nitrogen mineralization, N deposition and denitrification have to be accounted for to allow a better description of nitrate leaching. 7. References Amatya, D., Chescheir, G., Fernandez, G., and Skaggs R. (1999). Testing of watershed scale hydrologic/water quality model for poorly drained soils. Mini-conference, advances in water quality modeling, ASAE international meeting, Toronto, Ontario, Canada. Brevé, M.A., Skaggs, R.W., Kandil, H., Parsons, J.E., and Gilliam, J.W. (1992). DRAINMOD-N, a nitrogen model for artificially drained soils. Proceedings of the Sixth International Drainage Symposium. ASAE, St. Joseph, MI 49085-9659: 327-344. Brevé, M.A., Skaggs R.W., Parsons, J.E., and Gilliam J.W. (1997). DRAINMOD-N, a nitrogen model for artificially drained soils. Trans. ASAE, 40 (4): 1067-1075. de Vos J.A. (1997). Water flow and nutrient transport in a layered silt loam soil. Ph.D. thesis. Wageningen Agricultural University, Wageningen, The Netherlands, 287p. El-Sadek, A., Feyen, J., Skaggs, W., and Berlamont, J. (2002). Economics of nitrate loss from drained agricultural land. Environmental Engineering, 128 (4):376-383. Geypens, M., Feyen, J., Hofman, G., Merckx, R., Van Cleemput, O., and Van Orshoven, J. (2001). Mineral nitrogen in the soil as a policy instrument to reduce N-leaching from agricultural soils. Proceeding of 11 th Nitrogen Workshop, Reims, France, 9-12 September, book of abstracts: 451-452. Šimúnek, J., M. Šejna and van Genuchten M.Th. (1999). The HYDRUS-2D software package for simulating water flow and solute transport in two-dimensional variably saturated media (version 2.0). U.S. Salinity Laboratory, Riverside, CA, USA. Skaggs, R.W. (1981). Methods for design and evaluation of drainage water management systems for soils with high water tables, DRAINMOD. North Carolina State University, Raleigh, North Carolina, USA. 11

van Dam, J.C., Stricker, J.N.M., and Droogers, P. (1990). From one-step to multi-step. Determination of soil hydraulic functions by outflow experiments. Rep. 7, De. of Water Resour., Agric. Univ., Wageningen, The Netherlands. van Genuchten, M.Th., and Nielsen, D.R. (1985). On describing and predicting the hydraulic properties of unsaturated soils. Annales Geophysicae, 3 (5): 615-628. 12

Evaluation des dépôts sédimentaires dans les retenues : mesures directes et indirectes dans le barrage de Kerne Uhel sur le Blavet (Bretagne). A. Jigorel 1, et J. P. Morin 2. 1 INSA de Rennes, GRGCR Géologie, Rennes, France. 2 Conseil Général des Côtes d'armor, DAE, St Brieuc, France Abstract The sedimentary deposits in reservoir can be determined in a direct manner by bathymetry or indirectly by the use of sediment traps or sedimentometers. The advantage of the latter methods is that they provide, at a given moment, information describing the dynamics of sedimentation, the nature and the origin of the deposits. The quantitative evaluation of deposits, made by this method, is however often questioned. A comparative study of the two methods was carried out at the reservoir of Kerne Uhel constructed in 1981 on the River Blavet. The sedimentary deposits were measured over 8 years (1990 1997) with the aid of 5 sedimentometers, of an area of 0,1 m 2, placed on the bed and recovered every 2 months. In parallel, the quantity of sediment was estimated by bathymetry (12 profiles with a measurement point every 5 metres), and an analysis of their characteristics, on the occasion of the 10 year drain-down in 1998. The results show that the sediments are of mixed detrital and biogenic origin, they are concentrated in the bottom of the basin. The silted area represents 54% of the total surface area. After 17 years, the volume of deposits is 93 500 m 3 where 1/3 is localised in the bed of the submerged principal tributary. The average rate of sedimentation calculated from the bathymetry results (16 mm.year -1 ) is identical to that measured with the sedimentometers. The rate of sedimentation determined by sedimentometer is all the more precise in reservoirs since resuspension is small and endogenic biogenic material is dominant. Résumé Les dépôts sédimentaires dans les retenues peuvent être évalués de manière directe par bathymétrie ou de manière indirecte, par l'emploi de pièges à sédiment ou sédimentomètres. L'intérêt de ces derniers est de fournir à un moment donné, des informations précises sur la dynamique sédimentaire, la nature et l'origine des dépôts. L'évaluation quantitative des dépôts, faite selon cette méthode, est toutefois souvent discutée. Une étude comparée des deux méthodes, a été réalisée dans la retenue du barrage de Kerne Uhel créée en 1981 sur le Blavet (Bretagne). Les dépôts sédimentaires ont été mesurés pendant 8 ans (1990 1997) à l'aide de 5 sédimentomètres d'une surface de 0,1 m 2, posés sur le fond et relevés tous les deux mois. Parallèlement, les quantités de sédiments ont été évaluées par bathymétrie (12 profils avec un point de mesure tous les 5 mètres) et analyse de leurs caractéristiques, à l'occasion de la vidange décennale en 1998. Les résultats montrent que les sédiments de type vase (origine mixte détritique et biogène) sont concentrés sur le fond de la cuvette. La surface envasée représente 54 % de la surface totale. 17 ans après la création de la retenue, le volume des dépôts est de 93 500 m 3 dont 1/3 est localisé dans le lit ennoyé du tributaire principal. L'épaisseur moyenne annuelle des dépôts mesurée par bathymétrie est identique (16 mm.an -1 ) à celle mesurée à l'aide des sédimentomètres. Les taux de sédimentation déterminés à l'aide de sédimentomètres sont d'autant plus précis dans les réservoirs que les remises en suspension des sédiments sont faibles et que la fraction biogène endogène est dominante. 1 Introduction Toutes les retenues de barrage sont soumises à des apports sédimentaires qui réduisent progressivement leur volume utile et peuvent à terme, les combler totalement. Les sédiments ont également un impact sur la qualité de l'eau brute, lorsqu'ils constituent un stock important de nutriments. Dans le massif armoricain, l'envasement excessif des réservoirs contribue à l'altération de la qualité de l'eau, (Forsberg 1989, Dillon et al 1990, Jigorel et Morin 1994), réduit la biodiversité et accroît les risques de colmatage du lit de la rivière en aval du barrage, pendant les vidanges (Poirel 1994, Merle et al 1996). Aussi la gestion des retenues eutrophisées soumises à une

sédimentation biogène endogène importante (Jigorel et Bertru 1993) nécessite de bien connaître la nature et l'origine des sédiments, les taux annuels de sédimentation et le volume total des dépôts. Dans tous les cas, l'évaluation des apports sédimentaires est utile pour programmer une gestion à moyen ou long terme des réservoirs. Le volume des dépôts peut être évalué de manière directe en associant topographie et bathymétrie et/ou de manière indirecte, dans les contextes favorables, par sismique réflexion et/ou réfraction, ou encore à l'aide de sondes gammadensimétriques et échosondeurs (Andrieu et al 1997). Dans la plupart des retenues, quelle que soit leur profondeur, il est possible d'utiliser des pièges à sédiments ou sédimentomètres (Baker et al 1988, Håkanson et al 1989, Bloesh 1995). L'emploi de sédimentomètres permet en outre de connaître, à un moment donné, les caractéristiques des dépôts ainsi que la dynamique sédimentaire dans la retenue. La précision des mesures quantitatives, faites selon cette méthode, est toutefois souvent discutée (Gardner 1980, Baker 1988, Bloesch et Burns 1980). La présente étude montre les résultats obtenus par deux méthodes (sédimentomètres et bathymétrie par enfoncement à la perche), dans la retenue de Kerne Uhel (Bretagne) qui est gérée par le Conseil Général des Côtes d'armor (France). 2 Site et méthodes 2.1 La retenue de Kerne Uhel sur le Blavet. La retenue de Kerne Uhel a été créée en 1981 par la construction d'un barrage en béton dans la haute vallée du Blavet. L'ouvrage est de type voûte unique cylindrique à seuil libre. Sa hauteur totale est de 14 m par rapport aux fondations et sa longueur en crête est de 120 m (figure 1). La capacité du réservoir est de 2,3 millions de m 3 pour une superficie correspondante de 74 ha. Sa longueur totale est de 3,5 km, sa largeur maximale n'excède pas 200 m, et sa profondeur maximale est voisine de 10 mètres à la cote de retenue normale (220 m NGF). Le plan d'eau comporte deux bassins secondaires qui peuvent être isolés du bassin principal en période de basses eaux. Le bassin versant, d'une superficie de 90 km 2 à l'amont du barrage, est entièrement situé dans le massif granitique de Quintin. La retenue constitue essentiellement une réserve d'eau potable qui assure la desserte du Sud-Ouest du département des Côtes d'armor mais elle est également un lieu de loisirs notamment pour la pêche, la voile et la randonnée pédestre (sentier périphérique). 2.2 Evaluation de la sédimentation à l'aide de sédimentomètres. L'emploi de pièges à sédiments ou sédimentomètres s'est beaucoup développé depuis les années 60. Les appareils ont des volumes et des formes très variées et ils sont généralement faciles à mettre en oeuvre, à partir d'un bateau. Tantôt maintenus en suspension par des bouées dans la colonne d'eau, tantôt posés sur le fond, ils ont pour objectif de recueillir le flux primaire des matières en suspension qui décantent dans la colonne d'eau. Les sédiments transportés par saltation et dans la suspension de fond ne sont pas piégés. Pour éviter de collecter le flux secondaire généré par les bioturbations, les appareils en suspension sont disposés au dessus du fond. Pour la même raison, les appareils posés sur le fond doivent avoir une hauteur minimale de 50 cm environ. Le fonctionnement des très nombreux modèles utilisés dans les lacs et réservoirs, a fait l'objet d'examens critiques (Bloesch et Burns 1980, Gardner 1980, Baker et al 1988, Koresky 1994). La validation du fonctionnement des appareils de collecte, par des méthodes indirectes n'est pas aisée, car il est difficile de reproduire en laboratoire ou de modéliser toutes les conditions naturelles d'un milieu lacustre profond (gradient de vitesse de courant, stratification, turbulence, apports sédimentaires détritiques et biogènes...). Aussi nous avons choisi de comparer les résultats obtenus à l'aide des sédimentomètres sur une longue période (8 ans) avec ceux fournis par la bathymétrie, 17 ans après la création de la retenue. Dans la retenue de Kerne Uhel, les mesures de la sédimentation ont été faites à l'aide d'un sédimentomètre posé sur le fond (modèle original conçu à l'insa et déposé à l'institut National de la Propriété Industrielle). Cet appareil présente une surface de collecte importante (0,1 m 2 ) qui permet de recueillir des quantités de sédiments généralement suffisantes pour effectuer de nombreuses analyses. Etant posé sur le fond, il n'est pas soumis à des déplacements horizontaux qui peuvent modifier de façon significative les conditions de dépôt des MES à proximité immédiate de l'ouverture. Il est toutefois nécessaire de le mettre en place sur un fond plat afin d'avoir une surface de collecte constante.

b a Figure 1 a : La retenue sur le Blavet. Situation des profils de mesure et des sédimentomètres. = Profils = Sédimentomètres b : Vue du barrage c : Sédimentomètre et fût de rechange c L appareil est constitué d un cadre en acier inoxydable d'une hauteur totale de 50 cm, dans lequel est placé un fût à large ouverture d une contenance de 22 litres et lesté à sa base par une dalle de mortier. Tous les sédiments qui décantent dans la colonne d eau sont recueillis dans le cône en acier inoxydable fixé sur le bâti et entrant dans le fût. L ensemble est repéré à l aide d une bouée de surface. Lors des relevés bimestriels, le fût contenant les sédiments est remplacé par un fût vide (figure 1). L eau est soutirée au laboratoire, après décantation, puis les sédiments sont séchés au déshydrateur par insufflation d air sec. Les résultats sont exprimés en kg.m -2 de sédiment sec. Les épaisseurs des dépôts sont ensuite évaluées à partir de la masse volumique apparente (ou densité sèche) des sédiments. La sédimentation a été mesurée en continu pendant la période 1990 1997 dans cinq stations réparties selon le profil en long dans les différentes unités sédimentologiques (figure 1). Le relevé des appareils a été effectué régulièrement tous les deux mois par la Direction de l'agriculture et de l'environnement du Conseil Général des Côtes d'armor et les analyses ont été réalisées par l'insa de Rennes. Deux sédimentomètres (S1 et S2) ont été posés dans le bras du Blavet et un seul (S3) dans le bras du Loc'h. Les stations 1 et 3 sont directement soumises aux apports respectifs du Blavet et du Loc'h tandis que la station 2 située à l'aval de la route D 50 est partiellement protégée des apports du Blavet par un seuil fixe surmonté de clapets mobiles. En été, les clapets levés isolent totalement le plan d'eau de tous les apports sédimentaires du Blavet, tandis qu'en hiver, seuls les sédiments les plus fins transportés en suspension sont

susceptibles de transiter vers l'aval au-delà du seuil fixe, les clapets étant alors abaissés. Enfin la station 5 n'est pas directement soumise aux apports du Dourdu qui débouche en aval. 2.3 Bathymétrie. Le volume des vases a été déterminé par des mesures bathymétriques à l'occasion de la vidange décennale (1998). Les mesures ont été réalisées sous eau par l'emploi combiné d'une sonde de profondeur et d'une perche graduée avec embout de 30 ou 60 mm, selon la nature des dépôts. La sonde donne la hauteur d'eau tandis que le refus d'enfoncement de la perche ou l'accroissement de l'effort, indique la cote du fond originel de la retenue. Les épaisseurs de vases et sédiments correspondent à la différence entre la cote de l'interface eau-sédiment et la cote du fond originel. La précision de la mesure varie de 2 à 5 cm et est d'autant plus grande que le contraste de compacité entre les vases et le substrat sous-jacent est élevé. La limite entre la vase et les formations colluviales sous-jacentes est marquée par un refus net d'enfoncement, tandis que la limite entre la vase et les formations limoneuses ou argilo-tourbeuses correspond à un accroissement important de l'effort d'enfoncement, car les masses volumiques des formations qui comblent le fond de la vallée du Blavet sont nettement supérieures à celles des vases fluides accumulées dans le réservoir. Les mesures ont été faites pendant la baisse du plan d'eau, à partir d'un bateau déplacé le long d'une corde bien tendue entre les deux rives. L'emplacement des points de mesure est déterminé par des repères tous les 5 mètres, sur la corde. La situation des 13 profils transversaux est précisée sur la figure 1. 3 profils (1-2-3) sont situés dans la queue de retenue du Blavet, en amont du pont de la route D50. 2 profils (4-5) dans la branche du Blavet, 2 profils (6-7) dans la branche du Loc'h, 6 profils dans le réservoir lui-même, en aval de la confluence des deux queues de retenue. Les échantillons de sédiments ont été prélevés sous eau ou immédiatement après la mise à sec, à l'aide d'une drague manuelle et d'un carottier. Les analyses (50 échantillons) ont porté sur la granulométrie, la minéralogie, la teneur en eau et la teneur en matière organique. 3 Résultats des mesures. 3.1 Mesures faites à l'aide des sédimentomètres. Les bilans annuels rassemblés dans le tableau 1 donnent les taux moyens de sédimentation mesurés dans chacune des stations. Les taux moyens annuels de sédimentation ont été calculés pour la totalité de la retenue à partir de la moyenne arithmétique des mesures effectuées avec les 5 sédimentomètres. L'épaisseur correspondante des dépôts a été évaluée à partir de la masse volumique apparente moyenne des sédiments qui est de 0,35 pour les faciès de type vase (42 échantillons). L'analyse des données bimestrielles montre que la sédimentation est maximale au printemps (marsavril) et au début de l'automne (septembre-octobre), périodes très propices au développement des diatomées. Les apports détritiques des rivières accroissent les taux de sédimentation en période de crue notamment à l'exutoire des rivières, mais ils ne jouent pas le rôle majeur dans le processus global d'envasement de la retenue qui est dû pour l'essentiel aux fortes productions phytoplanctoniques induites par l'état d'eutrophisation. Les taux annuels de sédimentation montrent dans les 5 stations, un gradient décroissant de l'amont vers l'aval. Ils sont maxima dans les stations 1 et 2 situées dans le bras amont du Blavet. La sédimentation y est alimentée à la fois par les apports détritiques de la rivière (S1) et des fossés qui bordent la route départementale D 50 (S2) et par les fortes productions phytoplanctoniques favorisées ici par un renouvellement continu des nutriments. Ils varient peu dans le centre de la retenue, mais ils diminuent d'un facteur 2 à l'amont immédiat du barrage (S5). Ce phénomène observé dans toutes les retenues résulte de l'entraînement continu vers l'aval des matières en suspension sous l'action du courant engendré par le débit réservé et par les prélèvements d'eau brute pour la production d'eau potable.

Tableau 1 Taux de sédimentation et épaisseurs de vase. Période 1990 1997 N Station Année 1990 Année 1991 Année 1992 Année 1993 Année 1994 Année 1995 Année 1996 Année 1997 Période 90-97 Moyenne annuelle P E P E P E P E P E P E P E P E E. T. P E 1 7,6 22 7,8 22 3,5 10 7,4 21 9,4 27 10,9 31 6,0 17 6,4 18 168 7,4 21 2 10,6 30 6,0 17 6,9 20 7,3 21 6,5 19 6,7 19 7,1 20 5,4 15 161 7,1 20 3 4,3 12 5,4 15 2,7 08 6,9 20 4,9 14 4,7 13 6,1 17 4,2 12 111 4,9 14 4 8,8 25 7,3 21 6,8 19 5,4 15 5,0 14 4,5 13 3,9 11 5,5 16 134 5,9 17 5 3,3 09 2,7 08 2,3 07 6,5 19 2,8 08 2,4 07 2,2 06 2,7 08 72 3,1 09 Moyenne 6,9 20 5,8 17 4,4 13 6,7 19 5,7 16 5,8 17 5,1 14 4,8 14 130 5,7 16 P : Poids sec en kg.m -2 E : Epaisseur de vase en mm.an -1 E. T. : Epaisseur totale des dépôts de vase en mm Les taux moyens de sédimentation dans la retenue montrent pendant la période 1990 1997 de faibles variations relatives. L'envasement a été minimal en 1992 (13 mm.an -1 ) et maximal en 1990 (20 mm.an -1 ). L'épaisseur moyenne des dépôts est de 16 mm.an -1 pour une période de 8 ans. 3.2 Bathymétrie. La retenue est soumise à un double phénomène : une érosion des berges et un envasement important du fond de la cuvette. Le batillage responsable de l'érosion des berges s'exerce sur une hauteur de 1,5 à 2 m en relation avec les variations saisonnières de la cote du plan d'eau. Les limons superficiels et les arènes granitiques qui recouvrent le substratum rocheux sont érodés sur des épaisseurs qui varient de 40 à 80 cm. Ce phénomène s'observe pratiquement sur toute la périphérie du plan d'eau et dans tous les cas, l'horizon supérieur des sols (au sens pédologique) a disparu tandis que l'horizon inférieur plus compact est, tantôt érodé, tantôt préservé et forme alors une dalle résiduelle indurée qui restitue bien la morphologie originelle des pentes. Les sables et graviers qui constituent la fraction résiduelle des arènes érodées sur les berges, forment sur les pentes latérales, des rides et des petits cordons parallèles aux isobathes du réservoir. Ces dépôts détritiques présentent un gradient granulométrique décroissant de la berge vers le bas de pente (figure 2). Les vases recouvrent uniformément le fond relativement plat de la retenue qui correspond au lit majeur ennoyé du Blavet. Les dépôts de type vase sont également présents sur la partie inférieure des pentes latérales lorsque celles-ci n'excèdent pas 12 à 15 % ainsi que sur les "banquettes" qui sont localement présentes à mipentes (ex : anciens chemins). En raison de leur fluidité, les vases ont tendance à combler les petites dépressions, aussi en fin de vidange, le fond envasé de la cuvette apparaît uniformément plat et seuls quelques chaos granitiques, émergent au dessus des vases (figure 2). a Figure 2 : a : vue des sables grossiers sur les pentes latérales b : vue de l'érosion des berges et de l'envasement du fond de la cuvette b

Les sédiments fins de type vase ont toujours une couleur gris sombre,une texture fine, une forte teneur en matière organique (15 à 20 %) et une teneur en eau élevée (100 à 400 %). En bas de pente ils ont une texture plus sableuse du fait des apports détritiques de la pente. Enfin des sables propres interstratifiés avec des vases sont rencontrés dans le chenal et à l'exutoire des ruisseaux où ils et peuvent former des microdeltas (ex : Dourdu). Les sédiments fins présentent un double gradient granulométrique (longitudinal et transversal) qui est déterminé par l'importance relative des apports détritiques du Blavet et de l'érosion des berges. Les épaisseurs de vase varient sur une faible distance, surtout en relation avec la morphologie originelle du fond de vallée. Elles sont importantes, voisines ou supérieures à 50 cm dans les dépressions et en bas de la pente. Les hauts fonds sont peu ou pas envasés. En fin de vidange, l'envasement est minime sur une largeur de 10 mètres de part et d autre du lit de la rivière car les vases peu compactées ont été érodées par les courants pendant la vidange. Le lit originel de la rivière est rempli au 2/3 par des sables fins et de la vase sur une épaisseur moyenne de 50 à 75 cm (figure 3). 4 - Estimation des volumes de sédiments. Les volumes de sédiment ont été déterminés dans chaque secteur à partir des épaisseurs moyennes et de la surface envasée au niveau du fond de la cuvette. Les dépôts d'arènes granitiques sur les pentes, n'ont pas été pris en compte car ils ne participent pas directement au comblement du réservoir. La fraction sableuse arrachée aux berges se redépose sur la pente à une faible distance de la zone érodée. Il s'agit donc d'un déplacement de matériaux "in situ" et non d'un apport allochtone. Par contre, les arènes apportées par les tributaires et déposées dans les micros deltas et sur le fond, notamment dans le chenal, participent bien avec les vases organiques au comblement progressif de la retenue. Les surfaces envasées ont pu être déterminées avec une assez bonne précision à partir des relevés de terrain et du plan topographique du réservoir. Dans un secteur donné, la limite externe du secteur envasé suit assez bien une isobathe de la retenue. La surface envasée représente 53 % de la surface totale. Cette valeur, strictement identique à celle obtenue pour la retenue sur le Gouet (Jigorel et Morin, 1994) est surtout déterminée par la morphologie des vallées. Pour le calcul des volumes de vases, nous avons distingué le lit des rivières, du fond plat de la cuvette. Le lit des rivières est encore comblé au 2/3 à la fin de la phase de vidange et les épaisseurs moyennes de vase y sont très fortes (50 à 75 cm), nettement supérieures à celles mesurées sur le fond. Le lit des rivières fonctionne donc comme un piège à sédiment. Les débits de crues ont été insuffisants pour remettre en suspension pendant l'assec, la totalité des sédiments qui les comblait. Seule la partie supérieure des dépôts a été érodée et a été transportée dans la partie aval de la retenue où elle a décanté. Les volumes totaux de sédiment ont été évalué à 16 500 m 3 en queue de retenue (amont D50) et à 77 000 m 3 dans la retenue elle-même (tableau 2). Les sédiments accumulés dans le lit des rivières représentent environ 30 % du volume total des vases. Dans l'hypothèse d'une répartition uniforme des vases sur le fond de la cuvette, l'épaisseur moyenne est de 28,6 cm, ce qui correspond à une épaisseur d'envasement de 16 mm.an -1. Tableau 2 Evaluation des volumes de vases E. en cm Secteur Profils Maximum Moyenne N. de mesures S. du secteur en ha V. de vase en m 3 E. dans le lit en cm V. dans le lit en m 3 V. total en m 3 Queue de retenue P1 à P3 60 20 50 6,4 12 800 50 3 700 16 500 Bras du P4 à P5 40 24 25 3,75 9 000 60 4 500 13 500 Blavet Bras du P6 à P7 40 26 25 4,5 11 700 50 2 100 13 800 Loc h Retenue aval confluence P8 à P13 85 22 105 15,5 34 100 75 15 500 49 600 Total aval D 50 54 800 22 100 76 900 E : Epaisseur N : Nombre S : Surface V : Volume

Rive gauche Profil 7 Rive droite hauteur en m 220 219 218 217 216 Sables grossiers Vases sableuses Vases Sable et vase 215 214 0 20 40 60 80 100 200 Profil 11 100 Sable et vase Hauteur en cm 0-100 -200 Vases Vases -300-400 0 20 40 60 80 100 200 Profil 13 100 Sables grossiers Hauteur en cm 0-100 -200-300 Vase Sable et vase -400 0 10 20 30 40 50 60 70 80 90 100 Distance en m Figure 3 Répartition des sédiments sur le fond de la retenue. - Profil 7 : altitude en m (IGN 69). Profils 11 et 13 : altitudes relatives. Le niveau 0 correspond à la côte du plan d'eau au moment des mesures

5 Examen comparé des résultats obtenus par bathymétrie et à l'aide des sédimentomètres. Les mesures effectuées à l'aide des sédimentomètres montrent que les taux de sédimentation varient sensiblement d'une station à une autre : ils sont moindres dans le secteur situé à l'amont immédiat du barrage et à l'inverse, plus importants au débouché des rivières (tableau 1). Ce phénomène est fréquent dans les lacs et les étangs (Bloesch et Uelhinger 1995, Koren et Klein 1994, Banas 2001, Maleval 2002). Les aires de très forte sédimentation détritique (microdeltas) sont dans le cas présent peu étendues si l'on considère la surface totale du réservoir. Sur une période pluriannuelle, la succession de fortes crues et des vidanges décennales contribuent à une répartition plus homogène des dépôts dans la retenue. Les remises en suspension ont lieu dans les secteurs amont les plus envasés puis elles décantent, lors des vidanges partielles, à l'amont du barrage dans un secteur moins envasé. De ce fait, les épaisseurs de vase déduites de la mesure des taux de sédimentation à l'aide des sédimentomètres ne correspondent pas strictement dans chaque station aux épaisseurs mesurées par bathymétrie. Lors des vidanges décennales, la répartition amont-aval des dépôts sur le fond apparaît plus uniforme. L'examen comparé des mesures bathymétriques faites avant et après vidange partielle, montre que 20 à 25 % des sédiments ont été déplacés dans la retenue, pendant la vidange. La nouvelle répartition des dépôts qui en résulte ne modifie en rien leur volume total dans la mesure où la fraction évacuée pendant la vidange au delà du barrage est insignifiante. Celle-ci a été évaluée à 35 tonnes (100 m 3 ) pendant la vidange de l'automne 1998, par la mesure en continu des MES et des débits au barrage. Les deux méthodes d'évaluation de l'envasement (mesures à l'aide des sédimentomètres et mesures directes des épaisseurs de vases sur le fond), peuvent être comparées sous réserve : - de considérer les dépôts (vases biogènes, vases sableuse fines, vases et sables interstratifiés) qui recouvrent uniformément le fond de la cuvette, - de retenir pour les calculs la surface envasée, - de connaître la masse volumique apparente des sédiments sur le fond, - d'avoir une durée de suivi suffisamment longue pour prendre en compte les variations interannuelles des apports des rivières, - de disposer de suffisamment de données (points de mesures bathymétriques et nombre de sédimentomètres) pour caractériser chaque unité sédimentologique représentative. Vu la morphologie de la retenue (réservoir long établi dans une vallée encaissée) et la dynamique sédimentaire pendant les plus fortes crues et pendant les vidanges, il n'est pas apparu opportun de faire l'évaluation des volumes de dépôts par unités sédimentologiques à partir des mesures faites à l'aide des sédimentomètres. Les sédimentomètres étant bien répartis dans la retenue (figure 1), nous avons choisi de retenir la moyenne arithmétique des taux de sédimentation pour le calcul des volumes totaux de sédiments déposés pendant la période de suivi (1990 1997). Le taux moyen de sédimentation mesurés à l'aide des sédimentomètres sur une période de 8 ans est de 5,7 kg.m -2 ce qui correspond à une épaisseur de sédiment de 16 mm.an -1. Parallèlement les mesures bathymétriques in situ donnent une épaisseur de 16 mm.an -1 pour une période de 17 ans. Cette convergence remarquable des résultats montre qu'un suivi régulier à l'aide de sédimentomètres permet d'avoir une bonne estimation des taux d'envasement des retenues. Il convient toutefois de veiller à la répartition et à l'emplacement des appareils dans la retenue pour avoir une bonne estimation du taux moyen d'envasement. L'expérience montre que les résultats sont excellents quand les apports détritiques des tributaires et les remises en suspension sont faibles et lorsque parallèlement les dépôts ont une origine biogène dominante. 6 Conclusion. Les taux de sédimentation mesurés à l'aide des sédimentomètres sont très précis dans les retenues profondes (plus de 10 mètres) et soumises à une sédimentation biogène endogène dominante (plus de 70 %). En Bretagne, les teneurs excessives en nutriments des eaux des rivières induisent des taux de sédimentation très élevés dans tous les réservoirs. Les sédiments fins de type vase s'accumulent sur le fond des cuvettes, en particulier dans le lit ennoyé des rivières. Les épaisseurs moyennes des dépôts varient de 7 à 70 mm.an -1 (Jigorel et Morin 1994, Jigorel et al 1996, Andrieu et al 1997) en fonction de l'état trophique du milieu, de la présence ou non de

macrophytes, de l'importance des apports allochtones, de la nature des sédiments et de la morphologie des retenues. Les données recueillies à l'aide de sédimentomètres implantés dans les principales unités sédimentologiques d'une retenue permettent d'évaluer la perte de volume utile à moyen terme. La mesure en continu des taux de sédimentation fournit non seulement une aide pour la gestion des vidanges décennales mais elle permet également d'évaluer l'efficacité des mesures préventives et curatives prises dans les bassins versants pour restaurer la qualité de l'eau. Les aménagements mis en oeuvre aujourd'hui, notamment la plantation de haies traditionnelles et la mise en place de bandes enherbées le long des cours d'eau, devraient faire baisser à terme les apports détritiques dans les rivières et dans les retenues. Enfin toute évolution favorable des flux de nutriments devrait également entraîner une diminution corrélative de la sédimentation biogène. L'évolution pluriannuelle des taux de sédimentation traduit globalement, l'évolution de la qualité des milieux aquatiques dans le massif armoricain. Références Andrieu, J.P., Cault, J.B., Bouchard, J.P, Comtet, A., Cottin, L., Garros-Berthet, H., Guillemot, B., Jigorel, A., Levenq, J., Lurin, P. et Morin, J.P. 1997. Expérience française récente dans le domaine de la gestion des sédiments dans les réservoirs. Actes du 19ème Congrès des Grands Barrages. Florence, Italie, 26-30 Mai 1997.Q 74-R 23 : 365-383. Baker, E.T., Milburn, H.B. et Tennant, D.A. 1988. Field assessment of sediment trap efficiency under varying flow conditions. J. Mar. Res. 46, 573-592 Banas, D. 2001. Flux de matière en étangs piscicoles extensifs : Rétention, Sédimentation, Exportation. Thèse de Doctorat. Université de Metz. France. 237 p. Bloesch, J., Burns, N.M. 1980. A critical review of sedimentation trap technique. Schweiz. Z. Hydrol. 42 : 15-55. Bloesch, J. 1995. Mechanism, measurement and importance of sediment resuspension in lakes. Mar. Freshwater Res. 46, 290-304 Bloesch, J. et Uelhinger, U. 1986, Horizontal sedimentation differences in an eutrophic Swiss lake. Limol. Oceanogr. 31, 1094-1109 Dillon, P.J., Evans, R.D. et Molot L.A. 1990. Retention and resuspension of phosphorus, nitrogen and iron in a central Ontario lake. Can. J. Fish. Aquat. Sci. 47, 1269-1281 Forsberg, C. 1989. Importance of sediments in understanding nutrient cycling in lakes. In Sly P.G. and Hart, B.T., (eds) Sediment/Water Interaction. Hydrobiologia 176/177, 263-277 Gardner, W.D. 1980. Sediment trap dynamics and calibration : a laboratory evaluation. J. Mar. Res; 38, 17-39 Håkanson, L., Floderus, S. et Wallin, M. 1989. Sediment trap assemblages, a methodological description. Hydrobiologia. 176-177. 481-490 Jigorel, A. et Bertru, G. 1993. Endogenic development of sediments in a eutrophic lake. Hydrobiologia. 268, 45-55. Jigorel, A. et Morin, J.P. 1994. Eutrophisation et sédimentation dans les retenues départementales des Côtes d'armor, France. Journées nationales d'étude AFEID-CFGB "Petits Barrages", Bordeaux 2-3 Février 1993, Cemagref-Editions. 431-448. Jigorel, A. et Morin, J.P. 1994. Bilan de la sédimentation dans un retenue eutrophisée, quinze ans après sa création. Actes du 7 éme Congrès International de Géologie de l'ingénieur, Lisbonne, Portugal, 5-9 Sept 1994. A.A. Balkema Edition, Rotterdam. Volume IV. 2667-2674. Jigorel, A., Morin, J.P. et Bertru, G. 1996. Eutrophisation et sédimentation dans les retenues départementales des Côtes d'armor, France. Hydrologie dans les pays celtiques. Actes du 1 er Colloque Interceltique d'hydrologie et de Gestion des Eaux. Rennes, France, 8-11 juillet 1996. INRA Editions, Paris. 79, 203-214. Jigorel, A., Derville I., Bonenfant M., et Lepetit D. 1997. Démantèlement d'un barrage comblé par des sédiments fins détritiques et biogènes : impact sur le milieu. Actes du Symposium International de la Géologie de l'ingénieur et de l'environnement. Athènes, Grèce, 1997. Volume 3. 2733-2738

Jigorel, A., Morin, J.P. et Hébert, M. 2000. Impact sur les sédiments des épandages de sulfates de cuivre dans les retenues. Water in the Celtic World : managing resources for the 21 st century. Proceedings of the 2 nd Inter-celtic Colloquium. Aberystwyth, University of Wales, U.K., 03-07 juillet 2000. BHS Occasional Paper n 11 :279-286. Koren,N. et Klein, M. 2000. Rate of sedimentation in Lake Kinneret, Israel : Spatial and temporal variations. Earth Surface Process and Landforms. 25, 895-904 Koresky, H.P.1994. Possibilities and limitations of sediment traps to measure sedimentation and resuspension. Hydrobiologia. 284, 93-100 Maleval, V. 2002. Nature, origine et taux de sédimentation dans un lac oligotrophe : Le lac de Saint Pardoux, France. Celtic water in a european framework. Pointing the way to quality. Proceedings of the third Inter-Celtic Colloquium on Hydrology and Management of Water Resources. Galway, Ireland, 8 th -10 th july 2002 Merle, G., Nihouarn, A., Daligault, P. 1996. Opérations de restauration et recolonisation naturelle sur la Sélune (Manche) après une opération de vidange de barrages. Hydrologie dans les pays celtiques. Actes du 1 er Colloque Interceltique d'hydrologie et de Gestion des Eaux. Rennes, France, 8-11 juillet 1996. INRA Editions, Paris. 79, 275-282. Poirel, A., Vindimian, E. et Garric, J. 1994. Gestion des vidanges de réservoirs, mesures prises pour préserver l'environnement et retour d'expérience sur une soixantaine de vidanges. 18 ème Congrès des Grands Barrages, Commission Internationale des Grands Barrages, Q.69-R.9.Durban 1994. 321-349

Périmètres de protection des captages d eau souterraine dans le massif armoricain. Effets sur la qualité des eaux G Marjolet 1, A Artur 2 et M Freslon 3 1 Conseil Général des Côtes d Armor, Saint-Brieuc 2 Mission InterServices de l Eau, Préfecture du Finistère, Quimper 3 Direction Départementale de l Agriculture et de la Forêt de la Manche, Saint-Lô Résumé La réglementation française impose la mise en place de périmètres de protection autour des points de prélèvement d eau publics qui fournissent l'eau potable. Leur établissement est effectué selon une méthodologie adaptée au contexte du socle armoricain. Elle tient compte des caractéristiques particulières des aquifères, mais aussi de la pression exercée par l agriculture locale, souvent responsable de teneurs en nitrates élevées. Les mesures de protection portent sur les modalités de captage des eaux souterraines, et sur la restriction des activités agricoles dans les aires d alimentation. L'analyse des résultats permet de hiérarchiser les mesures à préconiser pour préserver et reconquérir la qualité des eaux souterraines du socle armoricain. Abstract The French legislation imposes the installation of protection zones around catchments for the human water supply. For their establishment in the context of the armorican substratum, a specific methodology is used. It takes into account the particular characteristics of the aquifers, but also the pressure exerted by local agriculture, often responsible for high nitrates ratios in groundwater. The protection measures take which aim at the methods of collecting of groundwater, and the restriction of agricultural activities. One can deduce a classification in the measurement effectiveness recommended for the preservation and restoration of quality for the armorican substratum groundwater. 1 La protection des captages destinés à la production d eau potable en France 1.1 Contexte réglementaire En complément aux règlements généraux relatifs à la protection des eaux contre les risques de pollution, suite à l'application des directives de l Union Européenne, la réglementation française impose des mesures supplémentaires pour la protection des points de prélèvement d'eau publics destinés à l'alimentation en eau potable (cf. Lois «sur l eau» du 16 décembre 1964 et du 3 janvier 1992). Ces mesures, obligatoires pour les captages sans protection naturelle, consistent dans la mise en place de périmètres de protection autour des points de prélèvement, alimentés par des eaux souterraines ou des eaux superficielles. Ces périmètres sont au nombre de trois : périmètre immédiat, périmètre rapproché (obligatoires) et périmètre éloigné (facultatif). Ils sont établis par arrêté préfectoral, après enquête d utilité publique (DUP) et avis d'un hydrogéologue agréé en matière d hygiène publique. Le périmètre immédiat, a une étendue généralement limitée, rarement supérieure à un hectare. Il est destiné à empêcher la détérioration des ouvrages de production d eau et à éviter les déversements et les infiltrations de substances polluantes à proximité immédiate du captage. A l intérieur de ce périmètre, propriété de la collectivité publique, toutes les activités autres que celles nécessaires à la production d eau potable sont généralement interdites. Le périmètre rapproché s étend au delà du périmètre immédiat, sur des terrains privés soumis à des servitudes de protection. D une superficie très variable (quelques hectares à plusieurs centaines d hectares), selon le contexte hydrogéologique, ce périmètre vise à interdire ou à réglementer les activités proches qui peuvent contaminer le captage et le rendre impropre à la production d eau potable. Un périmètre éloigné peut être instauré, au delà du périmètre rapproché lorsqu'il est nécessaire de compléter le dispositif de protection du captage par des réglementations particulières. Ainsi des mesures spécifiques peuvent être imposées pour le stockage de produits pouvant détériorer la qualité des eaux. Des actions complémentaires de protection telles que le conseil agronomique et la mise aux normes des bâtiments d'élevage peuvent également être mises en œuvre.

1.2 Situation de la protection des captages Une enquête sur la situation de l instauration des périmètres de protection en France, portant sur 94 départements, a été réalisée, en 1997 par la Direction Générale de la Santé (Carré et Howard, 1999). Elle indique, qu à cette date, moins d un tiers (31 %) des captages bénéficiait d une déclaration d utilité publique établissant les périmètres de protection. Malgré une progression (+ 10 % de périmètres) par rapport à la situation décrite par l enquête précédente de 1991 (Godet, 1992), le faible taux de périmètres ne manque pas d interpeller. Les principales raisons évoquées pour expliquer l absence de ces périmètres de protection, concernent: - l inadaptation globale de la démarche pour des contextes hydrogéologiques particuliers (aquifères karstiques par exemple) ; - la lourdeur de la procédure, peu adaptée aux petits captages desservant un faible nombre d habitants ; - les conflits liés à l usage de l eau (concurrence des prélèvements privés, effectués pour l irrigation par exemple) et aux servitudes de protection interdisant ou limitant certaines activités (cas notamment des activités agricoles) ; - le manque de moyens humains consacrés à la mise en œuvre de ces opérations (Marjolet, 1992 - Conseil national de l évaluation, 2001). En revanche, les coûts résultant de l instauration de ces périmètres ne constituent, généralement pas, un obstacle majeur, car les subventions attribuées par les acteurs publics (Etat, Régions, Départements et Agences de l eau) sont particulièrement incitatrices. 1.3 Politiques locales menées L enquête menée par la Direction Générale de la Santé montre que l état d avancement des procédures d instauration des périmètres de protection est très variable selon les départements. Ainsi, en 1997, 14 départements (15 % des 94 départements enquêtés) présentaient un taux d instauration supérieur à 50 %. Les départements les plus avancés (Agences de l Eau, 1999) sont ceux où des politiques locales en faveur de l établissement des périmètres de protection ont été appliquées. Les mesures les plus efficaces concernent la mise en place d accords-cadres entre les différents partenaires et la création de cellules techniques pour réaliser les opérations. 2 La protection des captages d eau souterraine dans le massif armoricain 2.1 L utilisation de l eau souterraine pour l alimentation en eau potable Le socle armoricain (fig. 1) couvre une superficie d environ 65000 km 2, et concerne une population d environ six millions d habitants, répartis sur quatre Régions administratives : la totalité de la Bretagne, une partie des Pays de Loire et de la Basse Normandie (ces trois Régions forment le «Grand ouest»), et une partie du Poitou Charente. Treize départements sont concernés (six en totalité et sept en partie), pour près de 80 % de leur superficie. Contrairement à la situation généralement observée en France, où l utilisation de l eau souterraine pour l alimentation en eau potable est largement majoritaire : 11.4 milliards de m 3 (64 %) pour un total de 17.8 milliards de m 3, en 2000 (Conseil national de l évaluation, 2001), le recours à ces ressources est minoritaire dans le socle armoricain, en particulier en Bretagne : 50 millions de m 3 produits (22%), pour un total de 220 millions de m 3, en 1995 (Région Bretagne, 1996). Cette faible proportion est due, au contexte géologique du massif armoricain, défavorable à la présence d'aquifères importants. Le développement insuffisant et trop tardif des connaissances sur ces ressources est aussi responsable de cette situation (Marjolet, 2001). Dans de nombreux secteurs ruraux, où les besoins sont modestes, des eaux souterraines auraient pu être utilement mobilisées. 2.2 Le contexte agricole du socle armoricain et ses conséquences sur la qualité des eaux Le développement de l'agriculture est très important dans le «Grand ouest», surtout dans le domaine des productions animales (lait, viandes, œufs). Celles-ci représentent plus de 45 % du total national (pour 15 % de la superficie). L'agriculture intensive et en particulier les productions hors sols sont, en Bretagne, à l origine d une dégradation de la qualité des eaux superficielles et souterraines, depuis une trentaine d années.(marjolet, 1999) Les conséquences les plus néfastes pour l alimentation en eau potable concernent les teneurs en pesticides et nitrates. Cette situation a conduit les Pouvoirs Publics à mettre en place une politique de maîtrise de la qualité des eaux, fondée sur des mesures réglementaires (en application de la Directive nitrates), accompagnées de mesures incitatives (Programme Bretagne Eau Pure). Pour l'instant, aucun effet positif n'a été mis en évidence, pour le paramètre nitrates (Mérot, 2000). 2.3 Principales caractéristiques des ressources en eau souterraines Types de captages d eau souterraine Avant 1975, les eaux souterraines du socle armoricain étaient captées uniquement par des ouvrages«traditionnels» peu profonds, le plus souvent implantés dans les altérites : puits, sources. Généralement de

Figure 1 Le massif armoricain situation des périmètres de protection des captages. 1 : Guébeurroux (Plémy Côtes d'armor) ; 2 : l'hôpital (Rospez Côtes d'armor) ; 3 : Bois Daniel (Elliant Finistère) ; 4 : la Gilberdière et le Piro (Sartilly Manche). faible débit, ces ouvrages répondaient aux besoins locaux des secteurs ruraux. Pour les besoins plus importants des secteurs urbains, les eaux souterraines ont aussi été captées par des réseaux de drains superficiels de plusieurs kilomètres, tels ceux réalisés par la Ville de Rennes et la Ville de Saint-Brieuc, au 19 ème siècle, et qui sont toujours en service (Marjolet, 2001). Au milieu des années soixante-dix, l arrivée dans la région de la technique du forage au «marteau fond de trou», a permis de mettre en évidence des ressources en eau plus profondes, qui sont à présent captées par des forages profonds (100 m et au delà). Principaux types d aquifères Si l on excepte quelques formations géologiques particulières (bassins tertiaires et quaternaires, plaines alluviales), de faibles étendues, les aquifères du socle armoricain sont principalement situés dans deux horizons bien distincts : les formations superficielles d'altération (altérites) et le substratum fissuré sous-jacent. Les altérites résultent de l histoire continentale du massif armoricain, marquée par l importance des phénomènes d altération, notamment à l'ère tertiaire mais aussi à l'ère quaternaire. Les formations d'altération qui en résultent sont plus ou moins meubles et de nature différente, selon la roche mère : arènes, pour les granites, sols argilo-limoneux, pour les schistes. On distingue deux niveaux superposés : - dans la partie supérieure, les allotérites, où la structure originelle de la roche a disparu, avec perte de masse et de volume ; - dans la partie inférieure, les isaltérites, où la structure de la roche est conservée, avec perte de masse sans perte de volume (BRGM, 1997). Les altérites sont caractérisées par une porosité d interstices permettant un stockage de l eau parfois important. Les perméabilités sont souvent faibles, sauf dans certaines formations comme les arènes sableuses. Elles donnent naissance à de nombreuses sources, de faible débit. Ces nappes superficielles sont captées par les ouvrages «traditionnels» : puits fermiers, anciens captages communaux. Le substratum fissuré est généralement situé sous les altérites, mais peut également affleurer. La présence d eau est liée aux discontinuités présentes dans les formations géologiques : contacts entre deux formations, alternances de faciès pétrographiques (schistes et grès), joints de stratification, plans de schistosité, diaclases, fractures, filons (pegmatites, dolérites, par exemple). Les forages d eau profonds montrent que les circulations d eau peuvent être présentes à - 300 m., dans la plupart des contextes géologiques. Les roches du substratum sont caractérisées par une porosité de fissures, avec des capacités de stockage généralement faibles. Les perméabilités sont très variables. Les débits, très rarement nuls, sont souvent de l'ordre de 5 à 10 m 3 /h. Ils peuvent atteindre ponctuellement des valeurs plus importantes, au delà de 100 m 3 /h. La superposition de deux formations géologiques : altérites au dessus d'un substrat fissuré, constitue la caractéristique principale des systèmes aquifères du socle armoricain. La fonction capacitive est assurée par les

altérites, et la fonction conductrice remplie par le substrat fissuré. Le comportement hydraulique s apparente à un système bicouche, avec un phénomène de drainance des altérites par le substrat sous-jacent. Compartimentation des aquifères Une autre caractéristique, souvent rencontrée, dans le socle armoricain, est la compartimentation des aquifères. Celle ci peut correspondre à une extension limitée d'un secteur fissuré aquifère ou à des vraies limites imperméables, liées à des changements de faciès ou à des fractures argilisées. Il en résulte une restriction des possibilités d exploitation, du fait de l exiguïté induite des aires d alimentation des captages, quelque soient les débits instantanés obtenus. On peut citer le cas du site du Syndicat d eau de Kerjaulez, de Launay, à Pommerit- Jaudy dans les Côtes d Armor, implanté dans des volcanites vacuolaires très aquifères, où les débits ponctuels sont de l ordre de 100 à 300 m 3 /h par forage, mais où la prise en compte de l aire d alimentation limitée à 200 hectares, dans un contexte de pluviométrie efficace de 250 mm par an en moyenne, conduit à des possibilités réelles d exploitation de l ordre de 60 m 3 /h, en moyenne annuelle. Qualité des eaux La qualité «naturelle» des eaux souterraines résulte de la nature pétrographique des roches les plus fréquentes : granites, schistes, grès. Les eaux sont généralement peu minéralisées et agressives. Des variations apparaissent en relation avec la proximité de la mer (chlorures), et dans des contextes géologiques particuliers. Dans la majorité des cas, les eaux souterraines superficielles, captées par des ouvrages «traditionnels» peu profonds (10-15 m au maximum), présentent des teneurs en nitrates élevées, induites par les activités agricoles pratiquées à proximité. En revanche, les eaux souterraines profondes, captées par des forages, à plus de 50 m de profondeur, présentent souvent des teneurs en nitrates nulles, même dans les secteurs d agriculture intensive. Ceci résulte d'un phénomène de dénitrification, très fréquent dans le socle armoricain (BRGM, 1997) qui détermine une zonation verticale des eaux. Dans le niveau supérieur (au dessus de 30-50 m), correspondant principalement aux altérites, les conditions sont oxydantes, avec présence d oxygène dissous et de nitrates ; dans le niveau inférieur (en deçà de 30-50 m), correspondant principalement au substrat fissuré, les conditions sont réductrices, avec disparition de l oxygène dissous et des nitrates, et présence, à l état dissous de fer et de manganèse, qu il faut éliminer par traitement, avant utilisation pour l eau potable. 2.4 Principaux objectifs recherchés par la mise en place des périmètres de protection dans le socle armoricain En France, les périmètres de protection des captages d eau souterraine visent essentiellement les risques de pollutions ponctuelles et accidentelles, à proximité des points de prélèvement. Ils ne sont pas adaptés à la protection de la ressource en eau, ni aux risques de pollution diffuse, en particulier ceux d origine agricole qui sont à appréhender par la réglementation générale (Conseil national de l évaluation, 2001). Les principales méthodes utilisées pour leur détermination correspondent d ailleurs à ces risques (Lallemand-Barrès et Roux, 1989). Dans le socle armoricain, les objectifs recherchés concernent aussi bien la protection du captage que la protection de la ressource : risques ponctuels et accidentels, mais aussi risques diffus (Carré et Marjolet, 1998). Ceci est rendu possible par la faible extension des aires d alimentation des captages, contrairement aux grands aquifères des autres régions. Les risques de pollutions d origine agricole sont particulièrement importants et peuvent mettre en péril la pérennité du captage. Ce sont, bien souvent les seuls risques recensés. Dès lors, une démarche de protection qui n intégrerait pas les activités agricoles serait vaine. C est pourquoi, dans le massif armoricain, les accords-cadres départementaux visent tous un objectif de maîtrise des pollutions diffuses d origine agricole. Ils sont d ailleurs signés par les Chambres d Agriculture (Département des Côtes d Armor, 1984 et 1997 ; Département du Finistère, 1993 et 2001 ; Département de la Manche, 1999). Une méthodologie spécifique a de ce fait été mise au point pour leur établissement, à la fin des années quatre-vingt (Marjolet et Burlot, 1992). 2.5 Méthodologie d'établissement des périmètres de protection dans le massif armoricain Détermination de l aire d alimentation La détermination de l aire d alimentation du captage est l'un des principaux problèmes à résoudre. Il faut définir son étendue et sa localisation. Les captages «traditionnels», implantés principalement dans les altérites, ont des aires d'alimentation, le plus souvent, assez proches du bassin versant topographique du captage (Marjolet et al., 2000). Dans une première approche, l aire d alimentation est assimilée à ce bassin topographique. On applique alors la méthode du bilan

hydrique, en comparant le débit moyen annuel mesuré du captage au débit moyen annuel résultant du calcul de l infiltration des pluies efficaces dans le bassin topographique. L'examen comparé des données et la prise en compte de paramètres, tels que la qualité des eaux et l environnement, conduisent à retenir ou à rejeter cette approche. Dans ce dernier cas, on est alors amené à la réalisation de piézomètres pour déterminer l aire d alimentation. Pour les forages profonds, qui exploitent le substrat fissuré sous-jacent, l'importance et les limites de l aire d alimentation sont précisées par les études hydrogéologiques préalables, en particulier les résultats des pompages d'essai et le suivi des piézomètres. Etude de l environnement Cette étude consiste à recenser et à décrire, dans l aire d alimentation définie précédemment, les éléments principaux de l environnement : écoulements des eaux superficielles, inventaire des points d eau, occupation des lieux, ainsi que les situations et activités présentant des risques de dégradation de la qualité des eaux : habitations (assainissements collectifs et individuels), activités industrielles et agricoles, stockages de produits polluants (hydrocarbures, notamment), voies de communication, etc. Elle comporte un volet agricole important, car l agriculture est souvent la seule activité présente. L'étude, dite «étude agropédologique» comprend une description du parcellaire agricole (cultures), parfois accompagnée d une cartographie pédologique, et des bilans de fertilisation établis à partir d'enquêtes auprès des agriculteurs. Les bâtiments agricoles, présents dans cette zone, et en particulier les ouvrages de stockage des déjections animales, sont également décrits. Cette étude conduit à des préconisations pour limiter les risques inventoriés de pollution. Exemples de mesures de protection préconisées Les mesures de protection préconisées qui prendront, après l enquête de DUP, la forme de servitudes publiques, sont de deux types : - des mesures " sévères " d interdictions d activités ; - des mesures "légères" de réglementation d activités. Elles s appliquent sur l ensemble du périmètre rapproché, souvent subdivisé en plusieurs zones correspondant aux différents niveaux de servitudes appliquées aux activités agricoles. Dans le Département des Côtes d Armor, le premier protocole établi en 1984 (Département des Côtes d Armor,1984) était relativement peu sévère. Il visait principalement une optimisation des pratiques agricoles. Le deuxième protocole, signé en 1997, est nettement plus contraignant. Il prévoit quatre niveaux de contraintes : R1, R2 (zone sensible), R3, R4 (zone complémentaire). Le niveau R1 interdit tout type de fertilisation et conduit à la mise en place d un couvert végétal permanent (bois ou prairies permanentes avec possibilité de pâturage extensif). Le niveau R2 conduit également à la mise en place d un couvert végétal, avec possibilité de retournement des parcelles en herbe (dans la limite de 20% de la surface de la zone R2), et limite la fertilisation azotée à 120 kg N/ha/an, pour les prairies. Les niveaux R3 et R4, moins «sévères» autorisent les cultures annuelles, avec des limitations de la fertilisation qui correspondent sensiblement, aujourd hui, à la Réglementation générale découlant de la directive nitrates, en zone vulnérable. Dans le Département de la Manche, le protocole prévoit trois niveaux : NP1, NP2, NP3. Le niveau NP1 correspond à la zone sensible et conduit à l'interdiction des labours, au maintien des parcelles en herbe, avec une possibilité de pâturage extensif et une fertilisation limitée (100 kg N/ha/an, avec épandage de déjections animales interdit). Le niveau NP2 correspond à la zone moyennement sensible et conduit au maintien des prairies permanentes et à la réglementation des cultures annuelles (mise en place d une interculture en hiver) et de la fertilisation (maximum : 170 kg N/ha/an). Le niveau NP3 correspond à la zone complémentaire, soumise à des réglementations et à des préconisations. Dans le Département du Finistère, le protocole prévoit deux types de périmètres rapprochés : le périmètre A et le périmètre B. Dans le périmètre A, les contraintes sont fortes et faciles à contrôler : le couvert végétal permanent est obligatoire (bois ou prairies permanentes fauchées), le pâturage est interdit et la fertilisation, uniquement minérale optimisée. Le périmètre B correspond à des réglementations qui, à présent, sont proches de la Réglementation générale. Dans les trois départements, l acquisition, par la Collectivité, des terrains compris dans les zones les plus sensibles, est fortement recommandée, et le plus souvent réalisée. Par ailleurs, des mesure alternatives conduisant à une protection de la ressource (mesures agri-environnementales des contrats territoriaux d'exploitation, transferts d'éligibilité, ) sont privilégiées. 3 Les résultats des périmètres de protection sur la qualité des eaux souterraines 3.1 Le captage de Guébeurroux (commune de Plémy, département des Côtes d Armor) (fig.2) Le captage de Guébeurroux contribue à l alimentation en eau potable des communes de Plémy et Moncontour, au sud-est de Saint-Brieuc. Il est situé dans une zone d agriculture intensive (maïs, céréales, prairies temporaires, élevages bovins, porcins et avicoles).

Le captage originel était «traditionnel», peu profond (6 m), et implanté sur une source drainant les arènes du granite hercynien de Moncontour. Il fournissait un débit moyen annuel de l ordre de 200 m 3 /jour, pour une aire d alimentation théorique estimée à 20-30 hectares. Les teneurs en nitrates des eaux captées ont connu dans les années quatre-vingt, une progression continue, pour atteindre 80 mg/l, à la fin des années quatre-vingt. En 1989, un forage de 104 m de profondeur a été réalisé à 30 m du captage. Il a traversé jusqu'à une profondeur de 30 m, les niveaux altérés superficiels du granite, et a indiqué un débit instantané de 20 m 3 /h, avec une teneur en nitrates proche de celle du captage. Dans le granite massif peu fissuré, sous-jacent, le forage est non productif, sur environ 60 m. A une profondeur de 90 m, le forage a atteint un niveau très fracturé, portant le débit total à 90 m 3 /h, avec une teneur en nitrates moyenne de l ordre de 18 mg/l. Un colmatage au ciment après réalésage de la partie superficielle, sur une hauteur de 40 m, a été réalisé pour empêcher les arrivées d eau superficielles. Le forage a été mis en production en 1989. Depuis lors, il fournit un débit moyen annuel de 400m 3 /jour, avec une teneur en nitrates fluctuant entre 0 à 5 mg/l. Le pompage sur le forage conduit à l assèchement de l'ancien captage en 24 heures. Son arrêt voit la remontée du niveau de l eau du captage. Il s agit en fait du même aquifère, présentant une zonation verticale de la qualité des eaux, caractérisée par un phénomène de dénitrification. La mise en service du forage a été accompagnée de la mise en place, en 1989, d un périmètre de protection d une superficie de 55 hectares, établi selon le premier protocole de 1984. Ce périmètre, relativement peu contraignant ne prévoit pas de contraintes de cultures particulières, mais uniquement des mesures de restriction d épandage des déjections animales liquides (lisiers). Une action de suivi et de conseil aux agriculteurs a été mise en place de 1990 à 1993. Elle a permis une diminution notable des excédents d azote autrefois épandus, sans toutefois atteindre l équilibre recherché, entre fertilisation et besoins des cultures. 3.2 Le captage de l Hôpital (commune de Rospez, département des Côtes d Armor) (fig.3) Le captage de l Hôpital fournit de l'eau au Syndicat d Alimentation en Eau Potable de Kreis Treger qui regroupe huit communes, dans la région du Trégor, au nord-ouest des Côtes d Armor. Il est implanté à la limite d une zone humide et d une zone d agriculture intensive (maïs, céréales, élevages bovins, porcins et avicoles), en mutation actuelle vers une zone à dominante naturelle et boisée, suite aux acquisitions foncières effectuées progressivement par le Syndicat. Le captage originel était constitué d un puits «traditionnel» peu profond (7 m), réalisé, à la fin des années soixante, à l emplacement d une ancienne source. Il exploitait une nappe superficielle dans un horizon d'altération d une formation volcanique cadomienne (tufs de Tréguier) et fournissait un débit moyen annuel de 200 m 3 /jour. L'aire d'alimentation théorique est de 30 à 40 hectares. Ce captage a connu, au cours des années soixantedix-quatre-vingt, une progression des teneurs en nitrates, qui ont atteint 63 mg/l, en 1983. Les recherches en eau, menées à partir de 1979, sur ce site (Dheilly-Carn, 1983), ont permis de mettre en évidence la présence d'eau souterraine profonde. Les débits ponctuels obtenus ont été importants (30 à 70 m 3 /h), avec des teneurs en nitrates de 0 à 35 mg/l, selon les forages. Le captage ancien et les forages sont établis dans le même aquifère (un pompage prolongé, à débit élevé, sur un forage, provoque l assèchement du puits). Ces ressources profondes ont été progressivement mise en exploitation. A partir de 1983, le forage F1 a été exploité, en mélange avec les eaux de l'ancien captage (300 à 400 m 3 /jour, en moyenne). A partir de 1988, les deux forages F1 et F2 ont fourni 500 à 600 m 3 /jour. Puis, les deux forages F1 et F3 ont permis de porter la production à 700 à 900 m 3 /jour, en moyenne. Le prélèvement, sur ces derniers, a été réduit, à partir de 1996, à 400 à 500 m 3 /jour, suite à la mise en service d un autre site d exploitation. Le périmètre de protection d'une superficie de 39 hectares a été mis en place, en 1988, selon le premier protocole de 1984. Une parcelle de 9 hectares a été acquise puis boisée. Le reste du périmètre a été soumis à des contraintes légères. Une action de suivi et conseil agricole a également été engagée. La mise en service du premier forage (F1) s est traduite par une baisse des teneurs en nitrates de 63 mg/l (puits seul) à 0-20 mg/l, dans les eaux du mélange puits - forage. Cette baisse a atteint 0 15 mg/l, après l'abandon de l exploitation du puits. Le forage F2, qui présentait des teneurs en nitrates élevées (40 mg/l) a été peu utilisé, et a été abandonné, en 1993. L'exploitation, à partir de 1993, du forage F3, en complément du forage F1 a été marquée par une augmentation des teneurs en nitrates qui sont passées de 5-15 mg/l, en 1994, à 32 mg/l, en 2000. Celles-ci semblent aujourd'hui diminuer (20 mg/l, fin 2001)..Durant les années quatre-vingt dix deux mille, le Syndicat d'eau a poursuivi les acquisitions de parcelles dans le périmètre de protection. Elles concernent aujourd'hui une trentaine d hectares. L'étude engagée, en 1999 (Sicard, 1999), pour rechercher les causes de la progression des nitrates a mis en évidence une modification de l'aire d'alimentation. Celle-ci résulte de l'augmentation des débits de pompage. Elle s'est ainsi étendue, dans la partie est, dans un secteur où l'activité agricole avait été maintenue et où les teneurs en nitrates des eaux souterraines de subsurface, mesurées sur des piézomètres, sont comprises entre 50 et 120 mg/l. Ces résultats ont conduit le Syndicat d eau a décider de deux types de mesures : - le maintien de l exploitation du site à

un débit limité à 500 m 3 /jour au maximum, et la réalisation d un quatrième forage d exploitation, à l ouest, dans un secteur protégé ; - la poursuite des acquisitions de parcelles dans le secteur agricole. 3.3 Le captage de Bois Daniel (commune de Elliant, département du Finistère) (fig.4) Le captage de Bois Daniel (Artur, 2001) contribue à l'alimentation en eau potable de la commune d'elliant, à l'est de Quimper. Il s'agit d'un captage "traditionnel", réalisé, dans les années cinquante et constitué d'un puits profond de 6 mètres, alimenté par des drains latéraux. Il exploite, à un débit moyen de 300 m 3 /jour, les eaux d'une nappe superficielle présente dans les arènes granitiques d'un massif granitique hercynien. L'aire d'alimentation de ce captage a été estimée à 80 hectares. Elle était occupée, initialement, par des cultures (maïs, céréales, légumes de conserve, pâtures). La teneur moyenne en nitrates, qui était de 10 mg/l, en 1950, a d'abord progressé lentement jusqu'en 1970 pour atteindre 25 mg/l, puis beaucoup plus rapidement jusqu'à un maximum de 88 mg/l, atteint en 1986. Pour remédier à cette dégradation la commune s'est engagée, en 1980, dans la mise en place d'un périmètre de protection qui a été instauré en 1985. Dans un premier temps, les mesures ont été relativement peu contraignantes : recommandation du maintien des surfaces en herbe et implantation d'une couverture hivernale des sols après maïs, limitation des périodes d'épandage des lisiers, raisonnement de la fertilisation, interdiction des cultures de légumes (épinards) et suppression des pollutions ponctuelles provenant des élevages. Devant l'absence d'amélioration de la qualité des eaux, la commune a procédé, à partir de 1990, à l'acquisition progressive de 30 hectares des terres les plus proches du captage, soit 37.5 % de l'aire d'alimentation. Ces terres ont dans un premier temps été maintenues en herbe, puis boisées en 1996. Après une période de stabilisation, les teneurs en nitrates ont commencé à diminuer nettement et régulièrement, à partir de 1993. En 2001, elles sont, en moyenne, inférieures à 50 mg/l, soit une diminution de 27 mg/l, en huit ans. D'autres exemples d'amélioration de la qualité des eaux de captages, suite à la maîtrise foncière, suivie d'un boisement d'une partie de l'aire d'alimentation, ont été constatés dans le département du Finistère (Préfecture du Finistère, 2000 et 2001) 3.4 Les captages de La Gilberdière et du Piro (commune de Sartilly, département de la Manche) (fig.5) Les captages de La Gilberdière et du Piro (Freslon et Chauvière, 2001 - Freslon, 2001), distants d'environ 500 mètres, ont été réalisés en 1963, pour l alimentation en eau potable de la commune de Sartilly, située entre les villes de Granville et d'avranches. Il s agit d ouvrages peu profonds (4 m) exploitant gravitairement des petits aquifères libres dans les arènes granitiques. Les aires d alimentation sont voisines et comprises entre 20 et 25 hectares. Les débits d'exploitation sont compris entre 100 à 150 m 3 /jour, selon la saison. De 1985 à 1995, les teneurs en nitrates ont progressé de 20 à 60 mg/l, pour le captage du Piro, et de 30 à 55 mg/l, pour le captage de La Gilberdière. Cette progression résulte d une intensification de l activité agricole, à la fin des années quatre-vingt. Les surfaces en herbe ont régressé dans les deux bassins versants au profit des cultures de maïs et de céréales. En 1994, la commune de Sartilly a acquis une grande parcelle cultivée de près de 8 hectares, dans l aire d alimentation proche du captage de La Gilberdière et a procédé à la mise en herbe de cette surface. Dès l'année suivante, la concentration en nitrates de ce captage a diminué régulièrement, passant de 54 mg/l, en 1995 (maximum) à 36 mg/l, en 2001. Les pollutions bactériologiques, liées aux ruissellements, fréquents en période pluvieuse, ont quant à elles disparu. En revanche, pour le captage du Piro, où aucune action n'a été engagée, les teneurs en nitrates ont augmenté de façon continue, la progression étant toutefois aujourd'hui moindre. Les teneurs étaient supérieures à 60 mg/l, en 2001. Parallèlement aux actions de prévention, la commune de Sartilly a réalisé, dans le même environnement hydrogéologique (granite de Carolles) un forage profond, pour exploiter une eau ferrugineuse mais très peu nitratée (< 5mg/l), à un débit proche de 150 m 3 /jour, afin de produire une eau conforme aux normes relatives à la qualité de l'eau distribuée. 2 Conclusions La mise en place des périmètres de protection des captages d eau souterraine du socle armoricain est menée, dans les différents départements, selon des démarches très proches. Celles-ci visent la protection des captages contre les pollutions directes et accidentelles, mais aussi la protection des ressources qui les alimentent, contre les pollutions diffuses, et en particulier celles d origine agricole. Les résultats des mesures de protection prises dans ces périmètres, peuvent être regroupées, pour les nitrates, dans trois catégories, classées selon une efficacité décroissante (GEOARMOR, 1997 - Burlot et al., 1997 Marjolet et Burlot, 2001). - une amélioration immédiate. Elle peut être constatée dans les cas de modification du mode de captage : remplacement d'un ouvrage "traditionnel" peu profond, par un forage. C'est le cas du captage de Guébeurroux et du

captage de l'hôpital. Cette amélioration peut être pérenne (captage de Guébeurroux) ou temporaire (captage de l'hôpital). - une amélioration progressive à court et moyen terme. Elle peut être constatée lorsque la mise en place du périmètre entraîne une modification importante de l'occupation des parcelles et des pratiques agricoles dans l'aire d'alimentation du captage. C'est le cas du captage de Bois Daniel et du captage de La Gilberdière. - pas d'amélioration sensible à court ou moyen terme. Ceci est constaté lorsqu'il n'y a pas eu de modification du mode de captage, de l'occupation des parcelles et des pratiques agricoles dans l'aire d'alimentation du captage. C'est le cas du captage du Piro. Ces trois catégories résultent des différentes situations de vulnérabilité et de risques de pollution de la ressource en eau souterraine. Le concept de vulnérabilité correspond à l aptitude ou à l inaptitude du milieu physique à empêcher, limiter, dégrader ou retarder la pollution de la ressource en eau utilisée. C'est ainsi que la vulnérabilité est moins importante lorsqu'une couche de terrains imperméables protège l aquifère capté. De même, vis à vis du paramètre nitrates, la dénitrification naturelle en profondeur constitue un élément favorable. La vulnérabilité peut être aggravée par les modalités du prélèvement. Ainsi l absence de cimentation dans les forages peut favoriser la communication entre des niveaux de qualité distincte. Le surpompage peut également diminuer la dénitrification des eaux (rabattements en dessous d un niveau imperméable de surface) Les risques de pollution résultent d'activités et de situations susceptibles d entraîner la pollution du captage et de la ressource en eau exploitée, de façon permanente ou accidentelle, ponctuelle ou diffuse. La combinaison de la vulnérabilité et des risques de pollution permet de définir, pour les aquifères du socle armoricain, quatre types de situation qui déterminent les mesures à prendre pour la protection des captages d'eau souterraine : - 1 : vulnérabilité et risques de pollution faibles. C est le cas, par exemple, d un forage profond, avec présence d un phénomène de dénitrification et situé dans une zone naturelle (bois par exemple). La démarche de protection, consistera, alors, à conserver ces deux caractéristiques favorables. - 2 : vulnérabilité faible et risques de pollution forts. C est le cas, par exemple d un forage profond, situé au sein d une zone d agriculture intensive. La démarche de protection, dans cette situation, consistera, d abord, à conserver le caractère de faible vulnérabilité, lié à la dénitrification en profondeur, et donc à empêcher les venues d'eau superficielles riches en nitrates dans les eaux profondes, par une cimentation sur une hauteur suffisante (30-50 m). Elle sera complétée par une action visant à réduire les pollutions à partir de la surface. - 3 : vulnérabilité forte et risques de pollution faibles. C est le cas, par exemple d un captage «traditionnel», peu profond, situé dans une zone naturelle (bois, par exemple). La démarche de protection, dans cette situation, consistera à conserver le caractère naturel de l environnement du captage. - 4 : vulnérabilité et risques de pollution forts. C est le cas le plus fréquent du captage «traditionnel», peu profond, situé dans une zone d agriculture intensive. La démarche de protection, dans cette situation, consistera à jouer, en fonction du contexte local, sur les deux paramètres. La vulnérabilité pourra être diminuée par une modification du mode de captage : remplacement du puits par un forage. Les risques de pollution par les nitrates d origine agricole pourront être réduits par l acquisition de terrains par la collectivité, suivie d un gel des activités, complété, par une optimisation des pratiques de fertilisation des cultures, sur les autres terrains... Références Agences de l Eau 1999. Mise en place des périmètres de protection des captages, Bilan et analyse d expériences positives. Les études des Agences de l Eau n 67, août 1999, 59p. Artur A. 2001. L instauration des périmètres de protection dans le Finistère : des résultats encourageants. In Périmètres de protection des captages : les conditions de la réussite, dossier des résumés et des interventions du colloque de Saint-Brieuc, 24-26 octobre 2001, 84-85. BRGM 1997. Qualité des eaux souterraines du massif armoricain, fréquence du phénomène de dénitrification naturelle en sous sol, équipement des forages et contamination de l eau exploitée, septembre 1997. Rapport BRGM n R39357, 46 p. Burlot T., Gallat G., Lucas G., Marjolet G., Roussel G. 1997. Bilan des périmètres de protection des captages d eau souterraine dans les Côtes d Armor, action sur les teneurs en nitrates. In Colloque : la protection régionale des eaux souterraines ; Document du BRGM n 275, 21-22.

Carré J., Marjolet G. 1998. Des périmètres de protection pour quoi faire? cas de la Bretagne. In Hydrogéologie n 4, 1998, 17-21. Carré J., Howard C. 1999. Mise en place des périmètres de protection des points de prélèvements d eau destinée à la consommation humaine, état des lieux en France. Ministère de l Emploi et de la Solidarité et Ministère de l Aménagement du Territoire et de l Environnement, 44 p. Conseil national de l évaluation. Commissariat général au plan 2001. La politique de préservation de la ressource en eau destinée à la consommation humaine. Rapport de l instance d évaluation présidée par Franck Villey- Desmeserets. La Documentation Française, 402p. Département des Côtes d Armor 1984. Protocole d accord relatif à la protection des captages, 00 janvier 1984, Préfecture, Chambre d Agriculture des Côtes d Armor, 00 p + annexes. Département des Côtes d Armor 1997. Protocole d accord relatif à la protection des points d eau publics destinés à l alimentation en eau potable dans les Côtes d Armor, 17 mars 1997, Préfecture, Conseil Général, Chambre d Agriculture des Côtes d Armor, Agence de l Eau Loire Bretagne, Syndicat départemental d alimentation en eau potable, Association des Maires des Côtes d Armor, 24 p. Département des Côtes d Armor 1997. Agir dans le bassin versant. Plan Départemental Environnement, décembre 1997, Conseil Général et Préfecture, 67 p +annexes. Département du Finistère 1993. Protocole relatif à l établissement des périmètres de protection des captages d eau potable, 2 juin 1993, Préfecture, Conseil Général, Chambre d Agriculture du Finistère, Agence de l Eau Loire Bretagne, Association des Maires du Finistère, 12 p + annexes. Département du Finistère 2001. Avenant n 1 au protocole relatif à l établissement des périmètres de protection des captages d eau potable, 17 avril 2001, Préfecture, Conseil Général, Chambre d Agriculture du Finistère, Agence de l Eau Loire Bretagne, Association des Maires du Finistère, 9 p + annexes. Département de la Manche 1999. Accord cadre périmètres de captages ; accord cadre départemental relatif à la mise en place des périmètres de protection des points d eau utilisés pour l alimentation en eau potable et aux prescriptions liées aux activités agricoles, 29 janvier 1999, Préfecture, Conseil Général, Chambre d Agriculture de la Manche, Agence de l Eau Seine Normandie, Association des collectivités gestionnaires de l eau potable et de l assainissement, Association des Maires de la Manche, 21 p + annexes. Dheilly-Carn, 1983. Contribution à l étude hydrogéologique des volcanites du Trégor. Thèse 3 ème cycle Université des Sciences et Techniques du Languedoc, Montpellier II, 172 p. Freslon M., Chauvière O. 2001. Suivi et évaluation des périmètres de protection, synthèse. Effets de mesures complémentaires à l arrêté de DUP sur la qualité de l eau. Rapport n 01/DDAF/10/HYD de la Direction Départementale de l Agriculture et de la Forêt de la Manche, 13 p + annexes. Freslon M. 2001. Mise en place des périmètres de protection dans le département de la Manche : bilan partiel et effets sur la qualité de l'eau. In Périmètres de protection des captages : les conditions de la réussite, dossier des résumés et des interventions du colloque de Saint-Brieuc, 24-26 octobre 2001, 82-83. GEOARMOR 1997. Bilan des périmètres de protection des captages d eau souterraine du département des Côtes d Armor, Conseil Général des Côtes d Armor, 81 p + annexes. Godet J.L. 1992. Le bilan de la mise en place des périmètres de protection en France. In Actes du colloque Périmètres de protection : mode d emploi, Saint-Brieuc (Côtes d Armor), 14-15 octobre 1992, 32-38 Lallemand-Barrès A., Roux J.C. 1989. Guide méthodologique d établissement des périmètres de protection des captages d eau souterraine destinée à la consommation humaine. BRGM, manuels et méthodes n 19, 219 p. Marjolet G., Burlot T. 1992. Les périmètres de protection des captages d eau publics : la démarche du département des Côtes d Armor. In Pen ar bed n 139, décembre 1990, La qualité de l eau en Bretagne, 38-44.

Marjolet G ; 1992. Synthèse des débats et des propositions. In Actes du colloque Périmètres de protection : mode d emploi, Saint-Brieuc (Côtes d Armor), 14-15 octobre 1992, 151-156. Marjolet G. 1999. Qualité des eaux et pollution la crise nitrates. In Géologues n 121, spécial Grand ouest, juin 1999, 72-76. Marjolet G., Faillat J.P., Sicard T. 2000. Contribution des eaux souterraines au débit et aux teneurs en nitrates des eaux dans un bassin versant granitique breton. In Water in the celtic world : managing ressources for the 21 st century, 2 nd inter-celtic colloquium, University of Wales Aberysthwyth 3-7 july 2000, British Hydrological Society occasional paper n 11, 303-309. Marjolet G. 2001. Les périmètres de protection dans le contexte du socle armoricain. In Périmètres de protection des captages : les conditions de la réussite, dossier des résumés et des interventions du colloque de Saint-Brieuc, 24-26 octobre 2001, 30-37. Marjolet G., Burlot T ; 2001. Bilan des périmètres de protection dans les Côtes d Armor et effets sur la qualité des eaux. In Périmètres de protection des captages : les conditions de la réussite, dossiers des résumés et des interventions du colloque de Saint-Brieuc, 24-26 octobre 2001, 76-78. Mérot P. 2000. Eau et agriculture en Bretagne : bilan d une politique incitative de maîtrise de la qualité des eaux. In Water in the celtic world : managing ressources for the 21 st century, 2 nd inter-celtic colloquium, University of Wales Aberysthwyth 3-7 july 2000, British Hydrological Society occasional paper n 11, 295-301. Préfecture du Finistère 2000. La qualité de l eau dans le Finistère. Cahier de la MISE (Mission InterServices de l Eau) n 3, avril 2000, 18p. Préfecture du Finistère 2001. La qualité de l'eau douce dans le Finistère. Cahier de la MISE (Mission InterServices de l'eau) n 4, mai 2001, 20p. Région Bretagne 1996. Schéma régional d'alimentation eau potable, 4 ème réunion ordinaire, novembre 1996, 57 p + annexes. Sicard T. 1999. Captage de l Hôpital (Rospez, Côtes d Armor), étude hydrogéologique complémentaire, rapport Université de Bretagne Occidentale, 39 p.

Nature, origine et taux de sédimentation dans un lac oligotrophe : le lac de Saint-Pardoux, France. Véronique Maleval Département de Géographie de l Université de Limoges - UMR 6042-CNRS Clermont-Ferrand. INSA de Rennes, GRGCR Géologie, Rennes, France. Résumé L utilisation de 4 pièges à sédiments a permis d estimer les taux de sédimentation dans le lac de St Pardoux (France). Ceux-ci évoluent selon un gradient amont-aval décroissant. Ils varient de 1063 g.m -2.an -1 à l'exutoire du tributaire à 442 g.m -2.an -1 à l'aval. Les épaisseurs correspondantes des dépôts sont respectivement de 5,3 mm.an -1 et de 2,3 mm.an -1. Ces faibles taux relatifs doivent être attribués aux apports réduits des tributaires et au bon état trophique du lac. Les relevés saisonniers montrent que la dynamique sédimentaire évolue en relation avec le débit des rivières et les cycles biologiques du réservoir. Les productions phytoplanctoniques participent de façon prépondérante à l'envasement. Dans un contexte hydrogéologique comparable à celui de la Bretagne, les taux de sédimentation sont ici 10 fois plus faibles. Ces différences importantes reflètent l'intensité relative des activités agricoles dans les bassins versants. Le bassin du lac de Saint-Pardoux étant largement boisé, le lac est bien préservé des pollutions d'origine agricole. Abstract The use of 4 sediment traps allow us to estimate the sedimentation rates in Saint-Pardoux lake (France). These change according to a gradient which decrease from upstream to downstream. They vary from 1063 g.m -2.y -1 where the river flows into the lake, to 442 g.m -2.y -1 downstream. The thickness of the deposits are respectively 5.3 mm.an -1 and 2,3 mm.an -1. These little rates are due to the reduced input sediment of the rivers and the good trophic quality of the lake. Seasonal statements show that the evolution of the sedimentary dynamic is linked to the river s flow and to the biologic cycles of the basin. Phytoplankton participates in a dominating way to the mud accretion. Comparatively to Brittany, which has a similar hydrogeological context, the sedimentation rates are ten times weaker here. These important differences show the relative intensity of agricultural activities in the drainage basin of the catchment areas. On the other hand, the drainage basin of Saint-Pardoux lake is widely afforested and well protected from agricultural pollutions. 1 Introduction Tout plan d eau créé par le barrage d une rivière joue le rôle d'un bassin de décantation pour les matières particulaires fournies par l érosion du bassin versant et des rivages. A ces dépôts détritiques s ajoutent les éléments biogènes fournis par la vie biologique du lac. Une partie de ces particules décante dans la colonne d eau et se dépose sur le fond. Dans les lacs soumis à une sédimentation biogène endogène importante, il n'est pas possible d'évaluer les taux de sédimentation à partir du bilan sédimentaire entrées - sorties du lac. Aussi, pour cette étude, des pièges à sédiment ou sédimentomètres ont été utilisés. Ils permettent de connaître la nature des dépôts, la dynamique sédimentaire et le taux moyen annuel d envasement du lac. 2 Site d étude Le lac de Saint-Pardoux créé en 1976, pour développer une activité de loisirs en zone rurale, est situé sur le piémont NO des Monts d Ambazac, à 25 km au nord de Limoges (figure 1a). Une digue en terre à noyau étanche de 19,40 m de haut barre la vallée encaissée de la Couze, rivière s écoulant vers la Gartempe (bassin de la Loire). La superficie de la retenue à la cote 360 NGF est 324 ha, son volume, 22 millions de m 3. La profondeur maximale du lac à la digue atteint 16,70 m, sa profondeur médiane 12 m. La Couze et son affluent le Ritord sont des rivières au régime simple de type pluvio-évaporal océanique. Les précipitations moyennes annuelles atteignent 1000 mm dans le bassin versant, et même si elles sont assez bien réparties sur l ensemble de l année, les hautes eaux se produisent en hiver, les basses eaux, en été. A l entrée du lac, les modules de la Couze et du Ritord sont respectivement de 1 m 3.s -1, et 0,7 m 3.s -1 (1967-1997, DIREN Limousin.).

Figure 1 : Le lac de Saint-Pardoux a situation b morphologie du lac et situation des sédimentomètres 3 Méthodes Le suivi de la sédimentation a été réalisé en continu, durant une période de deux ans (mars 2000 à mars 2002), à l'aide de quatre sédimentomètres. Cette méthode est bien adaptée pour connaître les taux de sédimentation de ce lac récent, oligotrophe et peu envasé. L'appareil utilisé a une hauteur de 50 cm et une contenance de 22 litres (modèle INSA, déposé à l'inpi). Il est posé sur le fond et repéré par une bouée. La surface de collecte des matières en suspension est de 0.1 m 2. Les sédimentomètres ont été répartis en fonction de la morphologie lacustre (un dans chaque bassin) et de la bathymétrie (figure 1b). Ils ont été posés dans les zones les plus planes et les plus profondes des différents bassins. Les deux pièges placés en queue de retenue (bassin de La Ribière et bassin de Santrop) sont représentatifs de la plage 5 à 10 m de profondeur. Les deux autres situés plus en aval caractérisent la zone variant de 10 à 15 m. Vu sa hauteur (50 cm), l'appareil ne recueille pas les sédiments des courants turbides, ni les remises en suspension dues aux bioturbations (flux secondaire). Le relevé des fûts des sédimentomètres a été fait selon une périodicité saisonnière (21 juin, 21 septembre, 21 décembre et 21 mars). Les sédiments sont séchés au laboratoire, à température ambiante, par insufflation d'air sec (déshydrateur). Les taux de sédimentation sont exprimés en g.m -2.an -1. Les épaisseurs correspondantes des dépôts sont calculées à partir de la masse volumique apparente des sédiments déposés sur le fond. Elles sont exprimées en mm.an -1. Les pièges étant placés dans la partie centrale des différentes cuvettes, il est possible de calculer les volumes déposés sur le fond dans chacun des bassins en connaissant les surfaces réellement envasées au niveau du fond.

Les analyses des sédiments récoltés dans les sédimentomètres ont porté sur la granulométrie et la teneur en matière organique. La nature des sédiments a été précisée par un examen au microscope électronique à balayage (MEB). Toutes les vases fines organiques ont été analysées au microgranulomètre laser Cilas 1180 après destruction de la matière organique avec de l eau oxygénée (H 2 O 2 ) et dispersion mécanique dans une solution d'hexamétaphosphate de sodium. Les sables ont été tamisés mécaniquement (tamiseuse Rotap) sur une colonne de tamis Afnor et lorsqu'ils recelaient une fraction fine significative, celle-ci a été analysée au granulomètre laser. La classification des sédiments est celle habituellement utilisée en sédimentologie en France. La teneur en eau et la densité sèche des sédiments ont été déterminées sur des échantillons prélevés à la drague manuelle sur le fond. 4 Etude des sédiments recueillis dans les sédimentomètres. 4.1 Variations saisonnières des quantités de dépôts Les poids de matériaux recueillis dans les sédimentomètres pendant trois mois, sont présentés par station et par année sur la figure 2. Les variations saisonnières sont fortes dans les stations 1 et 2 situées dans la partie amont du lac, mais nettement plus faibles dans les stations aval. Figure 2 Variations saisonnières du poids des sédiments recueillis dans les sédimentomètres Durant l'année 2000, les résultats de la station 1 se distinguent très nettement de ceux relatifs aux 3 autres stations. Les dépôts y ont été nettement plus élevés, notamment pendant la période septembre décembre (736 g.m -2 ). Dans les 3 autres stations les fluctuations saisonnières ont été pratiquement identiques: les minima ont été mesurés pendant la période juin septembre (18 à 54 g.m -2 ) et les maxima en septembre décembre (144 à 286 g.m -2 ). Durant le premier trimestre de l'année 2002, la sédimentation a été très homogène et légèrement supérieure à 100 g.m -2 dans toutes les stations. Elle a ensuite augmenté très nettement (facteur 4) dans les stations amont 1 et 2 (295 à 377 g.m -2 ), mais elle est demeurée faible dans la station 3 aval (25 à 92 g.m -2 ). 4. 2 Estimation des taux annuels de sédimentation Les taux annuels de sédimentation ont été évalués à partir des mesures faites avec les sédimentomètres pendant la période de suivi (mars 2000 mars 2002). Quelques mesures saisonnières manquent dans les stations 2 et 4 car les sédimentomètres ont été détériorés ou déplacés. Les taux annuels de sédimentation dans la station SP2 ont alors été évalués en extrapolant les résultats de la période de suivi (9 mois) à une durée de 12 mois. L'épaisseur des dépôts annuels a été déterminée pour chaque station à partir de la valeur de la densité sèche des sédiments sur le fond. Les taux de sédimentation et les épaisseurs des dépôts sont portés dans le tableau 1.

Tableau 1 : poids des sédiments recueillis dans les pièges- taux de sédimentation et épaisseurs moyennes annuelles des dépôts poids annuel (g.m 2 ) total année 1 et 2 (g.m 2 ) taux annuel de sédimentation (g.m -2.an -1 ) densité sèche épaisseur des sédiments (mm.an -1 ) épaisseur sur le fond (1976 2002) (cm) station année 1 année 2 SP 1 1165 960 2125 1063 0.20 5.3 13.8 SP 2 462 941 1403 701 0.21 3.3 8.6 SP 3 424 386 811 405 0.22 1.8 4.7 SP 4 442-442 442 0.19 2.3 5.9 taux du lac 653 0.20 3.2 8.3 Les poids annuels montrent à l'évidence deux milieux sédimentaires bien distincts. Dans les deux stations amont les quantités annuelles de sédiments déposés fluctuent le plus souvent autour d'une valeur proche de 1 000 g.m -2. Elles sont nettement plus faibles et voisines de 400 g.m -2 dans les deux stations aval. Pendant l'année 1 de suivi, le poids de sédiment dans la station SP2 a été plus proche des valeurs mesurées dans les stations aval que de celui enregistré dans la station amont SP1 tandis que pendant l'année 2 les poids de sédiments ont été pratiquement identiques dans les stations SP1 et SP2. Ces résultats très cohérents mettent en évidence un gradient de sédimentation amont-aval décroissant. La densité sèche (ou masse volumique apparente) des dépôts sur le fond est très faible (0,19 à 0,22) et correspond à celles de vases organiques très fluides. Les concentrations (190 à 220 g.l -1 ) traduisent une texture fine, des teneurs en eau et en matière organique élevées, liées à la présence d'une fraction biogène dominante (phytoplancton). L'épaisseur des dépôts annuels varie de 5,3 mm.an. -1 dans le bassin de la Ribière (SP1) à 1,8 mm.an -1 dans le bassin de Chabannes (SP3). Les valeurs les plus fortes sont enregistrées dans les deux bassins amont (La Ribière et Santrop) qui sont respectivement situés à l'exutoire de la Couze et du Ritord dans le lac. Les épaisseurs totales des vases déposées sur le fond depuis la création du lac (1976) ont été évaluées dans chacun des bassins à partir des données fournies par les sédimentomètres. Elle varient de 13,8 cm dans le bassin de La Ribière à 4,7 cm dans le bassin de Chabannes. Pour la totalité du plafond du lac de Saint-Pardoux, le taux de sédimentation est de 712 g.m -2.an -1 et l'épaisseur correspondante des dépôts est de 3,2 mm.an -1 4.3 Texture des sédiments Pour bien rendre compte de la variation spatio-temporelle de la texture des dépôts, les résultats des analyses granulométriques ont été portés sur un diagramme de texture (figure 3). Les points représentatifs des échantillons apparaissent alignés sur un segment de droite. Cette disposition montre que le rapport argile sur limon est constant et la texture des dépôts varie essentiellement en fonction de la teneur en sable. La fraction argileuse (0 2 µm) varie de 3 à 20 %, la fraction limoneuse (2 20 µm) présente une grande amplitude de variation (35 à 80 %) et la fraction sableuse (20 200 µm) fluctue de 5 à 60 %. L'absence de points à proximité du pôle sableux (S) s'explique par le fait que les sédimentomètres recueillent exclusivement les matières en suspension qui décantent dans la colonne d'eau à 50 cm au dessus du fond. Les sédiments les plus grossiers transportés par saltation et par traction au niveau du fond ne sont pas collectés. Les sédiments les plus fins (points situés près de l'axe A L) ont été déposés pendant l'automne 2000 et l'hiver 2001. Les sédiments les plus grossiers, riches en sables fins ont été mis en place pendant les printemps 2000 et 2001, périodes pendant lesquelles les rivières ont connu des débits soutenus. Pendant toutes les autres périodes, les sédiments déposés dans le lac présentent un gradient granulométrique décroissant de l'amont vers l'aval. En règle générale, les sédiments déposés dans l'aire d'influence des 2 tributaires, la Couze et le Ritord sont plus grossiers que ceux décantés dans les parties centrale et aval du lac.

Figure 3 Texture des sédiments recueillis dans les sédimentomètres 4.4 Nature des sédiments Les sédiments sont de nature détritique et biogène. Leur couleur sombre, brun gris foncé (E 61 du code expolaire de A. Cailleux et G. Taylor, non daté), indique la prépondérance des éléments organiques. Les analyses montrent des teneurs en matières organiques importantes qui varient globalement de 19 à 38 %. Pendant une période donnée, les valeurs fluctuent dans les différentes stations notamment en relation avec l'importance relative de la fraction sableuse détritique. Parallèlement, les teneurs moyennes des 4 stations varient surtout en relation avec les saisons. Tableau 2 Teneurs moyennes en matières organiques dans les quatre stations printemps été automne hiver Année 1 (2000-2001) 35,6 29,3 26,4 28, 2 Année 2 (2001-2002) 32,7 31,2 23,1 Les teneurs maximales sont mesurées au printemps et les minimales en automne. Ces fluctuations saisonnières résultent à la fois de l'importance relative des apports organiques endogènes et exogènes et de l'intensité du processus de minéralisation à la belle saison. La matière organique comporte des éléments grossiers peu ou pas dégradés et de la matière fine souvent très dégradée. Les feuilles et les brindilles sont essentiellement allochtones et sont plus abondantes en automne et en hiver dans les deux stations amont. La fraction organique fine est surtout fournie par la microflore et à un degré moindre par la faune et la microfaune (fèces des poissons et du zooplancton). La légère diminution des taux de M.O. en automne est vraisemblablement due à une activité bactérienne plus importante. L'examen au microscope électronique à balayage (MEB) a permis de préciser la nature des constituants des sédiments (figure 4). Tous les échantillons observés apparaissent constitués d'un mélange de minéraux et d'éléments biogènes plus ou moins bien conservés. L'ensemble est disséminé dans une matrice fine argiloorganique. Les éléments biogènes nettement dominants sont les tests siliceux des diatomées. La composition de la flore diatomique varie en fonction de la situation des stations dans le lac et des saisons. - au printemps, les sédiments recueillis comportent quelques minéraux et diatomées : Tabellaria ventricosa (a), Navicula cryptocephala (b) (cliché 1) Cyclotella pseudostelligera. - en été les diatomées pennées et centriques printanières sont toujours présentes acssociées à Aulacoseira ambigua (c) et à des débris du zooplancton difficilement identifiables (cliché 2). - en automne, les diatomées dominantes sont Aulacoseira ambigua, Melosira granulata (d) mêlées à quelques minéraux qui forment des colonies filamenteuses (cliché 3). Les autres espèces reconnues sont, Eunotia pectinalis (e), des Achnantes minutissima (f) (cliché 4) et des Navicula. - la microflore de la période hivernale est très diversifiée : Navicula cryptocephala, Cyclotella pseudostelligera, Aulacoseira ambigua et Asterionella formosa (g), un mélange de Tabellaria ventricosa, Tabellaria floculosa et Asterionella formosa (h).

Figure 4 Vues au MEB des sédiments recueillis dans les sédimentomètres Ces observations montrent que les populations algales et notamment les diatomées sont présentes tout au long de l année. L'apparente richesse en diatomées des sédiments pendant la période hivernale est probablement attribuée au bon état de conservation des tests plutôt qu'à de fortes productions phytoplanctoniques. La diversité des espèces traduit le bon état trophique (oligotrophe) du plan d'eau. Le lac de Saint- Pardoux se différencie très nettement, par sa flore diatomique variée, des lacs eutrophisés armoricains qui connaissent d'intenses proliférations quasi monospécifiques en période estivale (Jigorel et Bertru 1993, Jigorel et Morin 1994). 5 Discussion Avec une sédimentation de 3,4 mm.an -1 le lac de Saint-Pardoux se situe dans la norme mondiale des lacs oligotrophes. Les épaisseurs de dépôts y varient généralement de quelques dixièmes de millimètres à quelques millimètres par an (Cyberski 1973 cité par Touchart 2000). Le taux de Saint Pardoux est comparable à celui du lac danois Esrom (3 mm.an -1, Jonasson, 1984, cité par L. Touchart, loc. cit.) ou du lac écossais du Haut Glendevon (3 à 4 mm.an -1, Duck et al, cité par Touchart loc. cit.). Avec un taux moyen de sédimentation de 653 g.m -2.an -1, il connaît une évolution naturelle similaire à celle du lac Léman (Vernet et al. 1983, cité par Touchart loc. cit.). Ce taux est dix fois plus faible que celui des lacs armoricains fortement eutrophisés (Jigorel et Bertru 1993, Jigorel et Morin 2002). Cette étude a permis également de préciser la dynamique spatio-temporelle de la sédimentation : les taux mesurés montrent un gradient décroissant amont aval. La sédimentation est plus forte dans les deux bassins des queues de retenue (5,3 mm.an -1 à la Ribière et 4,4 mm.an -1 à Santrop) que dans les bassins de Chabannes et Fréaudour qui ont des taux respectifs de 1,8 et 2,3 mm.an -1. Si certains auteurs n observent aucune variation spatiale en lac (Lawacz 1969, Moeller et Likens 1978 cité par Banas 2001), d autres ont mis en évidence de telles variations durant de courts laps de temps (Evans et Håkanson 1992). A Saint Pardoux les taux de sédimentation sont déterminés à la fois par l'origine mixte des dépôts (apports des rivières et sédimentation endogène) et par la dynamique hydrosédimentaire interne au lac. Les multiples détroits qui délimitent de vastes bassins bien individualisés constituent des obstacles au transit amontaval des sédiments. Chaque bassin fonctionne donc de manière relativement indépendante. Les mesures faites à l'aide des sédimentomètres ont permis de mettre en évidence un gradient amont-aval à l'échelle du lac tout entier, et parallèlement les mesures bathymétriques dans les deux bassins amont ont également montré un gradient de sédimentation décroissant de l'exutoire des rivières vers l'aval de chacun des bassins (Maleval et Jigorel, article soumis). Ce double gradient est lié à l'origine mixte détritique et biogène des apports et à la morphologie globale du lac. Dans les deux bassins amont, les apports détritiques et biogènes des rivières décantent rapidement et transitent peu vers l aval. Seules les fines migrent au-delà des détroits. Les apports détritiques des rivières varient en fonction de la pluviosité qui peut entraîner une augmentation du débit des cours d eau et de l'intensité de l érosion dans le bassin versant. Les apports de la Couze et du Ritord accroissent les taux relatifs de sédimentation, pendant les périodes de crues, dans les stations 1 et 2. Ces dépôts de crue sont caractérisés par

une fraction sableuse plus importante. Les interventions anthropiques dans le bassin versant peuvent modifier ponctuellement la dynamique des apports en queue de retenue. Ainsi la diminution des dépôts dans la station SP2 pendant la première année de suivi, est liée au faible niveau d'eau de l'étang de Gouillet, situé en amont de Saint Pardoux. Cet étang partiellement vidangé a joué le rôle d'un bassin de décantation pour les apports hivernaux de crue du Ritord, jusqu'à son remplissage total. Durant le suivi de l'année 2, les sédiments recueillis dans les sédimentomètres ne recelaient aucune fraction sableuse, alors que parallèlement les poids étaient plus importants que la première année. L'absence de matière détritique grossière s'explique par les faibles débits des rivières au début de l année 2002, tandis que l'accroissement significatif des taux de sédimentation pourrait résulter quant à lui du développement plus important du phytoplancton en «année sèche» ( Jigorel et Morin 1993, Jigorel 1998). Les quantités de sédiments biogènes déposés sur le fond varient selon les saisons en relation avec les cycles biologiques du lac : au printemps, la vie algale est à son maximum de développement et une partie du phytoplancton est mangée par le zooplancton. En période de broutage intense (mai-juin), le poids des sédiments déposés sur le fond diminue. En été, la stratification thermique des eaux peut entraver la décantation d une partie des particules. Néanmoins, la présence de matière organique grossière (des fragments de végétaux dans les fûts 1 et 3) témoigne de l'apport de certains macrophytes à cette saison. L automne est la saison la plus favorable à la sédimentation car les particules ne sont plus bloquées par la thermocline et parallèlement l activité biologique et le broutage sont faibles. En résumé les quantités relatives de dépôts et leurs caractéristiques granulométriques permettent de différencier deux domaines distincts : les deux bassins amont et les deux bassins aval. Dans tous les bassins, les taux de sédimentation ne sont pas réguliers mais varient à la fois selon les saisons et d'une année à l'autre. D une façon générale, la sédimentation est moins importante et plus régulière à l aval du lac qu à l amont. Les sédiments du lac sont majoritairement biogènes. Les teneurs élevées de matières organiques montrent le rôle primordial des productions phytoplanctoniques dans l envasement de la retenue (Jigorel 1998). Dans le lac de Saint-Pardoux, la sédimentation liée à la dynamique biologique du lac est plus importante que celle liée à la dynamique hydrologique. L épaisseur de sédiments évaluée à partir des mesures des sédimentomètres apparaît faible. En réalité, elle est peut être légèrement plus importante, car l érosion a été beaucoup plus active durant les dix premières années de vie du lac (Maleval 1999, Maleval et Astrade article soumis). Il faut toutefois considérer l'évolution des sédiments sur le fond après leur dépôt. L'activité bactérienne tend à réduire progressivement les teneurs en matière organique. La densité sèche augmente et parallèlement le volume des dépôts diminue du fait de la minéralisation et de la compaction. Au total le volume des dépôts estimés à partir des mesures faites à l'aide des sédimentomètres est proche des volumes réels dans les retenues profondes à sédimentation biogène dominante (Jigorel et Morin 2002). Le volume de sédiments accumulés en 26 ans a été estimé à 357 000 m 3. Il représente seulement 1,6 % du volume total. Au rythme actuel de la sédimentation le lac peut être comblé dans 1 573 ans. 6 Conclusion L'évolution intra-annuelle de la sédimentation dans le lac de Saint Pardoux connaît des fluctuations saisonnières. Les apports varient avec le débit solide des rivières mais surtout en relation avec les productions phytoplanctoniques internes au lac. Chaque cuvette a un fonctionnement pratiquement indépendant qui résulte de la morphologie du réservoir et des faibles débits des tributaires. Le lac de Saint-Pardoux se situe dans la norme mondiale des taux de sédimentation. La faible accumulation des dépôts dépend à la fois de son état trophique (oligotrophe) et des faibles activités anthropiques dans son bassin versant. Les terrains sont occupés à 64 % par des forêts, à 29 % par des prairies et à 5 % par des eaux continentales. L'absence d'industrie, d'agglomération importante et d'agriculture intensive constitue des facteurs favorables à la préservation de la qualité des eaux du lac. La sédimentation biogène endogène y est prépondérante, mais les taux de sédimentation sont très inférieurs à ceux mesurés dans les lacs fortement eutrophisés du massif armoricain. Si l'évolution d'un réservoir d'eau oligotrophe est de devenir eutrophe par phases successives (Dussart 1996), les conditions naturelles très favorables du lac de Saint-Pardoux, permettent d'espérer que ce milieu aquatique sera préservé pendant de nombreuses années. Remerciements. Je remercie Monsieur Alain Jigorel, ingénieur de recherche du Groupe de Recherche Génie Civil de l'insa de Rennes ainsi que tous les membres de l'équipe géologie qui ont réalisé les analyses. Merci également à Monsieur Le Lannic, ingénieur de recherche au Centre Commun de Microscopie Electronique à Balayage de l'université de Rennes 1. Merci à mon directeur de thèse, Monsieur Laurent TOUCHART, professeur à l Université de Limoges. Une recherche menée par Naoko Ishiguro, doctorante japonaise à l Université de Limoges, permettra de répondre en partie à cette question.

Références Banas, D. 2001. Flux de matière en étangs piscicoles extensifs : rétention, sédimentation, exportation. Thèse de l Université de Metz, mention Sciences de la vie, 237 p. Cailleux, A., Taylor, G. (non daté). Code expolaire et sa notice. Edition N. Boubée et C ie, Paris. 19 p. Dussart, B-H. 1966. Limnologie. L étude des eaux continentales. Edition Gauthiers-Villars, Paris. 678 p. Réédition 1972, Boubée. Evans, R. D., Håkanson, L. 1992. Measurement and prediction of sedimentation in small Swedish lakes. Hydrobiologia 235/236, pp. 143-152. Jigorel, A., Bertru, G. 1993. Endogenic development of sediments in a eutrophic lake. Hydrobiologia 268, 45-65. Jigorel, A., Morin, J-P. 1993. Eutrophisation et sédimentation dans les retenues départementales des Côtes d Armor, France. Journées nationales d étude AFEID-CFGB «Petits Barrages», Bordeaux 2-3 février 1993, Cemagref Edition 1994, 431-448. Jigorel, A., Morin, J-P. 1994. Bilan de la sédimentation dans une retenue eutrophisée, quinze ans après sa création. Proceedings 7 th International, Congrès International, Association of Engineering Geology. Lisbonne, Portugal. Balkema ed., vol. 4, 2667-2675. Jigorel, A. 1998. Bilan de la sédimentation dans la retenue sur l Arguenon et l étang de Jugon : période 1989-1997. INSA de Rennes, Département Génie Civil, 55 p. Jigorel, A., Morin, J-P. 2002. Evaluation des dépôts sédimentaires dans les retenues : mesures directes et indirectes dans le barrage de Kerne Uhel sur le Blavet (Bretagne). Celtic water in a european framework. Pointing the way to quality. Proceedings of the third Inter-Celtic Colloquium on Hydrology and Management of Water Resources. Galway, Ireland, 8 th -10 th july 2002 Maleval, V. 1999. L évolution des rivages des lacs de barrages artificiels. L exemple du lac de Saint-Pardoux en Limousin. Norois, Poitiers, 183, tome 46/3. pp 453-464. Maleval, V., Astrade, L. (soumis). La datation de l érosion des rivages du lac de Saint-Pardoux en limousin par l utilisation de la dendrochronologie. Revue de Géographie Alpine. Maleval, V., Jigorel, A. (soumis). La sédimentation dans un lac artificiel. L exemple de Saint-Pardoux (Massif d Ambazac, Limousin) - Revue Géomorphologie. Touchart, L. 2000. Les lacs, origine et morphologie. L Harmattan, Paris, 202 p.

Feeding the fish - a Celtic perspective Feeding the fish - a Celtic perspective Changing potentials for land-sea exchanges of organic matter and other materials in selected Scottish sea lochs. Abstract Laurence A. Boorman, L A B Coastal, The Maylands, Holywell, St. Ives, Cambs, PE27 4TQ, UK. Much of the coast of the Scottish Highlands is steep and rocky with very limited exchanges between land and sea but the abundant sea lochs provide a dynamic link between the two. Experimental work over the past ten years has shown that salt marshes can play a key role in linking a range of terrestrial and marine communities with up to 40% of their primary productivity being exported. Recently this work has been extended to selected sites in Lochaber and Skye. Comparative studies are in progress to compare the ecosystem dynamics in lochs with varying proportions of salt marsh and contrasting hydrodynamics. The sea lochs of the Western Highlands vary greatly in the relative size of the freshwater catchment, the proportion of intertidal and the mean depth at low water as well as the proportion of salt marsh. In contrast to southern salt marshes the sediment supply is often very limited and generally has a higher proportion of coarser particles. The high rainfall and humidity of the area facilitates the accumulation of organic matter in salt marsh soils. Together these factors have created areas of a unique habitat with a limited distribution along the coast of the Western Highlands, areas which nevertheless play an important role in linking terrestrial and marine communities. Increasing rates of sea level rise, as a result of global warming, are likely to overtake rates of isostatic adjustment and augment existing anthropogenic threats to these special areas. All the five marshes studied showed extensive high marsh and transition communities but rather restricted lower marsh and pioneer communities compared with those of salt marshes elsewhere. Nevertheless the productivity of these salt marsh plant communities and their vertical range in the marsh zonation are comparable. In more exposed areas, however, salt spray can result in the occurrence of salt marsh vegetation well above the levels normally expected. Preliminary results from flux studies at Loch Beag in Skye indicate that the system is particularly dynamic with very variable rates of exchange of organic matter and sediment. It would appear that sediment accretion rates are generally low compared with those in active marsh systems elsewhere. The overall situation is further complicated by the high degree of variation in the inputs and organic loading of fresh water inflow to the system. It seem likely, however, that invertebrate and fish communities in these lochs benefit significantly from the input of organic matter of terrestrial origin and that these intertidal plant communities themselves benefit from material of marine origin. The studies are continuing. 1

Feeding the fish - a Celtic perspective Gearrchuntas 'S ann anns na Gaidhealtachd na h-alba a tha mòr-chuid na h-oirthire gu math càs agus creagach agus tha na cothroman airson a' bhuinne-sruth rud beag cuingichte ach tha moran lochan-mara ann a' cruthaicheadh an co-cheangal beathachadh eadar talamh 's cuan. 'S e cocheangal beathachadh fiùghantach a tha ann. Air feadh na h-oirthire tha mi a' faicinn, tha mi a' sgrudadh lusan de gach seòrsa anns na làthach-shaillte agus tha mis air rianachd ghoireasan a sgrudadh. Tha mi a' toirt tuairmse na buidseat bhliadhnail mu dheidhinn sruthan eadar tir 's muir. 'S e co-cheangal anabarrach cudromach a tha ann gu dearbh agus 's e ionad-coinnich eadar uisge 's sàile a tha innte cuideachd. O chionn goirid tha mise a' sgrùdadh coig lochan-mara ann an Lochabar agus anns an t-eilean Sgitheanach. Tha mi a' rannsachadh an àrainneachd anns na lathach-shaillte, gu sònraichte, am prìomh-toradh de lusan agus an t-eòlas-mara anns na coig lochan. 'S e gu math eadardhealaichte a tha na lochan seo. 'S e meud an locha mòra no beaga a tha ann. 'S e na lathach-shaillte fharsaingeachd no cuingichte a tha ann. Agus 's e na lochan domhainn no eudomhainn a tha ann cuideachd. 'S e grùid gu math nas gainmheala a tha ann, bho aite gu aite co-dhiù. Chan eil e uiread grùid anns an uisge-mara nas motha. 'S e àrainneachd gu math sònraichte a tha ann gu deimhinn. Tha e co-cheangal beathachadh eadar talamh 's cuan cuideachd. 'S e an galar de dhearmaid agus neo-shuim a tha a' bagradh oirthir nan Gàidhealtachd. 'S e blàthachadh na cruinne leis a' còmhnard-mara nas airde agus buaidhean na daoine a tha a milleadh an oighreachd beò sin. Tha na lusan a fàs anns an làthach-shaillte, aig cinn nan loch ann an Lochabar agus anns an t- Eilean Sgitheanach, gu math coltach ris iad sin anns na lathachan-shaillte eile. Tha am prìomh-toradh nan lus co-ionnan gu leor cuideachd. Tha mi air tuairmse a thoirt seachad air a' bhuidseat bliadhnail airson sruthan de stuth fàs-bheairteach agus beathachadhmèinnearachd mu thimcheall am bhuinne-sruth eadar tir agus muir airson Loch Beag ri taobh Loch Bràcadail anns an t-eilean Sgitheanach. 'S e an àrainneachd gu math eadar dhealaichte 's atharrachail a tha ann. Bho àm gu àm bithidh tuiltean uisge ann, leis eabar agus stuth fàsbheairteach, agus, aig amannan eile, bithidh tuiltean sàile ann leis stuth-mara de gach seòrsa. Tha creideamh agam gu bheil na h-èisg, agus na beathaicean gun chnaimh-droma anns an loch, a' biathadh leis an stuth seargte bho am prìomh-toradh nan lus anns an làthach-shaillte eadar seòl-mara. 'S e caran cudromach airson iasg a' chuain fhèin a tha ann cuideachd, tha m a' smaoineacheadh. Bithidh mise a' deanamh moran rannsachaidh mu dheidhinn an gnothaich seo. 2

Feeding the fish - a Celtic perspective Introduction Wetland communities are widely recognised for their high primary productivity and tidal wetlands, i.e. salt marshes, are no exception. Salt marshes are described as areas of intertidal mud stabilised by a cover of vegetation. In contrast to non-tidal wetlands the regular ebb and flow of saline water imposes a range of special ecological features making the habitats of unique ecological importance. The key features were first described in Europe by Chapman (1960) and in the United States by Teal (1962). The special attributes of salt marshes include the provision of a vital source of organic matter and nutrients connecting adjacent marine and terrestrial communities and so forming a living link between land and sea, (Boorman, 2000). Salt marshes also provide a dynamic buffer between the land and the sea since they are able to absorb wave energy during storms and to rebuild during calmer periods (Pethick, 1992). It is this feature that enables salt marshes to make a major contribution to coast protection and defence against the sea by giving wave protection to otherwise vulnerable sea walls (Boorman, 1999). As well as all this salt marshes themselves support a unique flora (Chapman, 1977 and Rodwell, 1999) and a rich and varied fauna providing spawning sites and nursery areas for many fish species (Boorman, 1999). Pivotal to all the attributes and functions of salt marshes are the various fluxes and exchanges between the salt marsh and adjacent ecosystems mediated by the flow of water between them. This process was first recognised four decades ago (Teal, 1962) but it has only been over the past decade or so that the process has been studied in detail and over a wide range of different areas both in the United States (Dame, 1989) and in Europe (Lefeuvre and Dame, 1994). The studies in the United States were based on the concept of 'outwelling' whereby salt marshes produce more organic matter than can be stored in the ecosystem and the excess material is exported to adjoining coastal waters where it can increase marine productivity (Odum, 1980). Despite efforts to standardise on methodologies and approaches the interpretation of the concept in the European situation is still unclear (Lefeuvre and Dame, 1994). Experimental work over the past ten years has shown that salt marshes can play a key role in linking a range of terrestrial and marine communities with up to 40% of their primary productivity being exported (Lefeuvre, 1996). The results showed great variation both seasonal and in relation to individual plant species. The results from the studies in eastern England, however, clearly indicate that exports of organic matter do occur but that under certain conditions organic material can also be imported. The results also indicated that there was a complex relationship between exchanges of organic matter and exchanges of mineral nutrients and sediments and this led to the formulation of the concept of 'marsh maturity' (Hazelden and Boorman, 1999). It became clear that marshes had a functional age which was independent of their chronological age. 'Functionally young' marshes were actively extending their area (the vegetated areas extending by the plants colonising bare mud) and were net importers of sediment and organic matter while 'mature' marshes which were not extending their area were net exporters of organic matter and sometimes even sediment (Boorman, 1999). Just as the export of organic matter could be related to the ecology of particular species it was clear from the studies in Essex and Norfolk that marsh function, in its various aspects, was related to the detail of the particular plant communities involved and that evidence of the functional age of a marsh or part of a marsh could be deduced from the species composition 3

Feeding the fish - a Celtic perspective and structure of the ecosystem. There are clearly distinct plant communities which occupy physically distinct zones in the salt marsh and each of the zones has its own distinctive structure and function including productivity (Boorman and Ashton, 1997). The range of plant communities which make up a particular salt marsh system reflect the functional variation of that marsh. For example, the occurrence of pioneer communities in an otherwise mature marsh reflects a local area of rejuvenation within that marsh. Valuable information on the overall functional properties of a marsh can be deduced from the make-up of the component plant communities. In Great Britain these studies were exclusively based on the salt marshes of Essex, Suffolk and Norfolk on the south-east coast where there are extensive areas of salt marsh (9000 ha - Burd, 1989). The marshes are generally situated in front of sea walls which protect very large areas of agricultural land which was formerly salt marsh, although the north Norfolk marshes form a coastal fringe against gently rising ground. The marshes themselves are particularly subject to the effects of rising sea levels and it has been estimated that over a period of 60 years over 40% of the Essex marshes will be lost to erosion (Boorman, 1999). In southern Britain the effect of sea level rise is exacerbated by the sinking of the land relative to the sea (isostatic adjustment) at a rate of 3 mm per year (Boorman, 1992). The marshes of north Norfolk are less vulnerable to rising sea levels both because there is a more abundant supply of sediment and because the lack of the sea wall means that they can respond to rising sea levels by a landward migration towards higher ground. Overall the marshes of East Anglia are typified by their extent and by their close proximity to each other. It was unclear how typical these marshes are of British salt marshes as a whole and for this reason it was decided to extend the studies to include selected marshes on the west coast of Scotland with the selection of sites in Lochaber and Skye. Comparative studies are in progress to compare the ecosystem dynamics in lochs with varying proportions of salt marsh and contrasting hydrodynamics. Many of the sea lochs of the west of Scotland have areas of salt marsh at the head of the loch and unlike most of the marshes of southern Britain these marshes are isolated and form self-contained units making them ideal for studies on exchanges between the terrestrial and marine habitats. The coastal marshes of Scotland are largely unaffected by sea level rise as in general isostatic adjustment means that in the north of the British Isles the land is rising in relation to the level of the sea. The sea lochs of the Western Highlands vary greatly in the relative size of the freshwater catchment, the proportion of intertidal and the mean depth at low water as well as the proportion of salt marsh. In contrast to most salt marshes in southern England the sediment supply is often very limited and generally has a higher proportion of coarser particles. The high rainfall and humidity of the west of Scotland facilitates the accumulation of organic matter in salt marsh soils. Together these factors have created areas of a unique habitat with a limited distribution along the coast of the Western Highlands, areas which nevertheless play an important role in linking terrestrial and marine communities. Increasing rates of sea level rise, as a result of global warming, are likely to overtake the positive rates of isostatic adjustment and augment existing anthropogenic threats to these unique areas. The study and monitoring of the terrestrial-marine fluxes in the Scottish lochs is a key factor in the management and conservation of unique areas of biodiversity around the coast and islands of Western Scotland, which are threatened by various factors including changes in sea level, nutrient enrichment and other forms of pollution such as antifouling residues and agricultural chemicals, and even tipping of waste. 4

Feeding the fish - a Celtic perspective Study Sites Four contrasting sea lochs were selected in the Lochaber area (Fig. 1); these were the southern part of Kentra Bay (Traigh Ceantra), the northern section of Loch Moidart (Ceann Loch Mùideart), Loch Ailort (Ceann Loch Aileort) and Arisaig Bay (Loch nan Ceall). At each site the extent and nature of the salt marsh vegetation is being monitored and regular sampling of soils, vegetation and water takes place to determine the magnitude and direction of organic fluxes. Kentra Bay is the southernmost of the sites with low rocky hills interspersed with sand flats and sandy salt marsh (Fig. 2). In contrast Loch Moidart, some 6 km to the northeast, is surrounded by high rocky slopes and there are extensive intertidal mud flats, amounting to 75% of the total area of the loch, and at the head of the loch there are more than 10 hectares of salt marsh (Fig. 3). Loch Ailort, ten kilometres further north, is of similar size and in a very similar setting. It does however, show a strong contrast with Loch Moidart because only 25% of its area is intertidal and this loch is dominated by sandy sediments. Nevertheless there is quite extensive sandy salt marsh at its head amounting to over 6 hectares (Fig. 4). On the other hand Arisaig Bay twelve kilometres to the north-west provides a rather mixed environment with a relatively steep slope and a mixture of rocky sandy and muddy habitats but with the inter-tidal accounting for only 10% of the total area of the bay (Fig. 5). Although the four lochs in Lochaber are all affected to some extent by human impacts such as stock grazing, fish farming, fishing and boating there have been no direct impacts on the geomorphology. Loch Beag (part of Loch Bracadale) on the west coast of Skye (Figs. 1 and 6) has been changed by the construction of a road across the mouth of the loch some 40 years ago, and the bridge will inevitably have had some impact on the hydrodynamics of the loch itself. There are approximately 4.7 hectares of salt marsh at the head of what is a much smaller loch (area of the inter-tidal 36%) and comparisons are being made on the development and functioning of the salt marsh areas of Loch Beag and of the four larger Lochaber lochs. Loch Beag was chosen for the initial flux measurements because of the existence of the bridge providing both access and a sampling point. The five Scottish sea lochs chosen each have a well-defined area of salt marsh and they are relatively self-contained water bodies. In addition these sea lochs provide interesting parallels and contrasts in many other parameters. The organic exports from this area will be affected by the proportion of the various units within the inter-tidal, as will the import of inorganic nutrients from the main water body of the loch. Each loch also forms an essentially self-contained system regarding sediment transport. These five lochs were considered to provide the optimum combination of similarities and differences to enable meaningful comparisons of structure and function to be made. Methods General physiographic details of each of the five marsh sites was obtained from the Ordnance Survey 1:25,000 maps together with Admiralty Charts No. 2207 and 2208. UKD Digital map (Version 3.00) published by the British Oceanographic Data Centre. Details of the freshwater catchment area of each of the sites were determined, together with the high and low water tidal volume of the loch as well as general bathymetric features including the exposure/fetch at each site. Tidal details were obtained from Admiralty tide tables. During August 1999 the main plant communities were mapped at each of the sites together with the 5

Feeding the fish - a Celtic perspective species composition and vegetation structure. The cover/abundance of the plant species was recorded using the Domin scale. Samples were taken to determine the biomass of the vegetation at the four Lochaber sites and core samples were taken of the soils at all five sites at depths of 0 to 250 mm wherever possible (in some of the sites there was rock or rocky fragments at depths of less than 250 mm). The organic content of the soil was determined at intervals of 50 mm. The height levels of the vegetation zones was determined in relation to the tidal regime of the site. At each of the sites there was a clearly marked drift line of vegetation debris and as these related to the same previous high tide at each site it was considered that this made a satisfactory reference level between the sites. Details of land use (e.g. grazing, fish-farming etc.) on and around each of the sites was also recorded. At one of the sites (Loch Beag) more detailed recordings were made, during August 2001, of both vegetation composition and structure together with turbidity levels and sediment transport. The recording at Loch Beag covered both a spring and a neap tidal cycle and involved recording the turbidity and water level and the taking of water samples. The botanical composition of the vegetation at this site was recorded again and changes over the past two years were noted. Air photos dating from October 1988 were obtained from the Scottish Natural Heritage office at Portree and this provided further valuable information. Detailed oblique photos in colour of the whole surface area of the vegetated marsh were taken from the prominent hillocks to the south and north of the site (Creag Mhòr 60 m high, 63 m from the edge of the loch and Dun Garsin Broch 120 m high, 126 m from the edge of the loch) and these were used to improve the vegetation map and for comparison with the earlier vertical aerial photographs. Results General physiographic features of the study sites The whole of Kentra Bay consists of a relatively low-relief landscape of small islands and rocky inlets with extensive sand flats in between. The sand flats had a limited covering of fine material but it was essentially a sand based system and there appeared to be a significant chance of blown sand contributing to the surface of the salt marsh. The area of salt marsh in Kentra Bay amounts to over 40 ha or 12 per cent of the area of the loch, the highest proportion of marsh at any of the sites. With an apparent high degree of similarity across the whole area sampling was performed on the marshes to the south of the Eileanan nan Gad and to the north of Bruach na Maorach (Fig. 2). The area was heavily sheep grazed with a very short grass sward. The salt marsh vegetation at the head of Loch Moidart is concentrated around the mouth of a stream flowing into the loch at the south-east corner (Allt na Glaice Moire). There is a large island of marsh to the south-west of Cnoc Aird Molach (Fig. 3) with smaller marsh islands in the mouth of the stream and a fringe of salt marsh extending along the loch shore in both directions. There is 17.4 ha of salt marsh which amounts to under 3% of the loch area. Mud is the dominant sediment type all around the head of the loch, particularly where the vegetation has facilitated accretion, but under the mud at varying depths are mixtures of sand, shingle and rock. Further up the stream there is a transition to fresh water wetland vegetation. The area is regularly grazed by cattle in the winter but also by deer and grey geese. 6

Feeding the fish - a Celtic perspective Although the large scale topography was very similar, the head of Loch Ailort formed a complete contrast to Loch Moidart with coarse sediment dominating. The whole area is based on coarse sand and shingle although there are some finer sediments involved in the areas of high marsh with relatively cohesive marsh mud being exposed in places. There is 8.9 ha of salt marsh amounting to just over 1% of the area of the loch. The main area of salt marsh is situated to the south of the River Ailort (Fig. 4). There appeared to be light to moderate sheep grazing over the whole area. Arisaig Bay at the head of Loch nan Ceall is also based on shingle and coarse sand but there was a definite admixture of mud with a distinct and sometimes quite thick layer of mud in places. The whole area appeared to be eutrophic with vigorously growing vegetation presumably the result of the various sewage outfalls across the site. There is a total of 3 ha of salt marsh covering just over 1% of the area of the loch (Fig. 5). The salt marsh vegetation which formed a fringe along the shore appeared to be virtually ungrazed in the area with only the occasional plant showing signs of having been eaten. Loch Beag is set in a steep rocky inlet on the east side of the much larger Loch Bracadale and it shores are predominantly rock and shingle. The Amar River feeds into the head of the loch and there is an area of salt marsh above the old road bridge. There is extensive intertidal with coarse sand and shingle covered with layers of fine sediment in the more sheltered places. The intertidal forms the dominant habitat and it occupies two thirds of the total area of the loch. There is approximately 3 ha of salt marsh (Fig. 6) with a further 1.2 ha of transitional marsh dominated by Iris and Juncus. The salt marsh is said to be common grazing and it is intermittently grazed by up to 50 Cheviots. The five lochs chosen for this study have contrasting features in terms of the proportion of inter-tidal, proportion of salt marsh, average depth of water, and magnitude of the fresh water inputs (Table 1.). Loch Beag and Kentra are shallow with large areas of intertidal in contrast to the other three lochs. There are also large differences in the contribution the adjoining catchment areas make to the flow of fresh water to the lochs. Even in Loch Moidart which has the lowest relative runoff into the loch the figure is still equivalent to 7 m of rainfall. With the average annual rainfall in the area around 2m the flow of fresh water through the ecosystem is in marked contrast to the English marshes which receive virtually no fresh water runoff and have an annual rainfall of less than 0.3 m. Table 1. Main parameters of the lochs studied in Lochaber and Skye. Areas are expressed in km 2. Depths and tidal range are given in metres. Runoff is given in M m 3 yr -1 and runoff relative to the area of the loch at high water is expressed as m 3 km -2. Data are from various sources including the UKD Map, Version. 3.00 published by BODC. Parameter Kentra Bay Loch Loch Ailort Loch nan Loch Beag Moidart Ceall High Water Area 3.3 1.2 8.2 2.8 0.75 Low Water Area 0.6 0.9 6.2 2.5 0.28 Intertidal Area 2.7 0.3 2.0 0.3 0.47 Intertidal % 81 25 24 10 63 Depth Low Water 1.0 4.3 11.1 2.9 1.85 Tidal Range 4.3 4.3 4.3 4.3 4.3 Runoff 48.4 8.7 133.7 30.4 48.5 Relative Runoff 14.7 7.3 16.3 10.9 64.6 7

Feeding the fish - a Celtic perspective The sites also vary considerably in the degree of exposure to the prevailing westerly winds. The details of exposure are given in Table 2. It will be seen that Loch Moidart and Loch Ailort are the most sheltered although the 10m bathymetric contour (deep water) is nearer in the case of Loch Ailort. Kentra Bay and Loch nan Ceall are the most exposed particularly the latter. Loch Beag appears to be relatively sheltered in terms of distance to the mouth of the loch but south-westerly winds are funnelled up the loch by the steep hill slopes on each side, effectively increasing the effects of exposure in this direction. Table 2. Exposure characteristics of the five lochs under study. Data was taken from Admiralty Charts 2207 and 2208 and OS 1:25,000 maps. Distances to 10 m depth refer to Lowest Astronomical Tide (Chart Datum) which is approximately 0.5 m below low water spring tides. Site name Distance to 10m depth (km ) Distance to mouth of loch (km) Width of loch mouth (km) Kentra Bay 4.7 4.7 2.2 Loch Moidart 8.3 8.3 0.7 Loch Ailort 0.5 8.6 2.1 Loch nan Ceall 1.2 0.5 1.2 Loch Beag 1.5 6.2 1.1 Comparison of the salt marsh plant communities The salt marsh vegetation in these loch sites was generally most extensive at the higher levels where it would only be covered by the higher (spring) tides. Zones representing the transition between saline and fresh water marsh communities were often quite well developed. Vegetation was found less often in the lower marsh zones with pioneer communities, covered by all but the lowest tides, being particularly poorly developed. Colonisation of bare areas of mud and sand could be by any one of three species Puccinellia maritima, Plantago maritima, and Armeria maritima although occasionally Salicornia species were found in the pioneer situation. Low marsh being covered by all neap tides was rare and was limited to small areas in Loch Ailort dominated by Puccinellia. Middle marsh which is only covered by spring tides was quite commonly found especially at Kentra Bay, Loch nan Ceall and Loch Beag with Armeria, Puccinellia and Glaux maritima the commonest species. High marsh which was only covered by high spring tides formed the commonest vegetation type. There were two rather distinct sub-types; a relatively well-drained type with Festuca rubra and Armeria dominant while in poorly drained areas Juncus gerardii and other species of Juncus were dominant. The high marsh vegetation at Kentra and Loch Beag was mainly of the former type while the wetter version of high marsh was common at Loch Moidart, Loch Ailort and Loch nan Ceall. The vertical heights of the salt marsh vegetation zones are given in Table 3. The overall vertical range of the vegetation is thus quite similar in all five lochs. Where a particular vegetation type is described as being absent then the next lower vegetation type generally occupies the intervening areas although sometimes the vegetation characteristic of the higher zone can be found at lower levels than normally expected. There is virtually no middle marsh at Loch Ailort but the Juncus gerardii is found at lower levels in this site than elsewhere. It is also noteworthy that the pioneer marsh at Loch Ailort extends to lower levels than elsewhere. Both these features may be linked to the relatively good drainage and aeration of the substrate with the coarser particle size at this site. At most of the 8

Feeding the fish - a Celtic perspective sites studied and particularly in the more exposed locations within each site the salt marsh extends up well above the normal upper limit of high water spring tides (HWST). This is a result of wind and waves carrying salt water and salt spray landwards. However although the plant communities so formed appear similar to the truly intertidal high marsh their situation out of the reach of even the high spring tides effectively prevents a significant functional contribution to the ecosystem. Table 3. Vertical height range (m) of the salt marsh vegetation zones at the five study sites. The height of the top of each of the marshes is taken to be at approximately 2.4 m OD (Newlyn) for purposes of comparison but this may vary somewhat from site to site (see text). Area Top of Salt Marsh Bottom High Marsh Bottom Middle Marsh Bottom Low Marsh Bottom Pioneer Marsh Kentra Bay 2.4 2.0 1.4 Absent 1.0 Loch Moidart 2.4 2.1 1.6 Absent 1.1 Loch Ailort 2.4 1.4 Absent 1.0 0.7 Loch nan Ceall 2.4 1.8 1.5 Absent 1.2 Loch Beag 2.4 2.0 1.6 1.3 1.0 Variation in sediments and salt marsh soils Although the vegetation may appear, at least superficially, to be similar the salt marsh soils of the west of Scotland are very different from the deep relative uniform sediments to be found in southern salt marshes. As well as having very variable organic contents at the surface there is also considerable variation at different depths (Table 4). Table 4. Variation of organic content of salt marsh soils at different depths at the five study sites. Soil depths are given in mm and represent the mean organic content expressed in terms of the loss on ignition at 450 o C. Sites where no figures are given for the 160-200 mm layer had compacted gravel or solid rock at this depth. Soil Organic Content (% LOI) AREA 0-40 41-80 81-120 121-160 161-200 Kentra Bay 13.2 9.3 7.4 5.1 - Loch Moidart 6.4 4.7 4.7 4.5 4.1 Loch Ailort 27.8 8.1 5.1 4.4 - Loch nan Ceall 19.6 15.4 14.7 14.7 - Loch Beag 4.7 3.9 6.9 8.0 6.9 There are considerable differences between the sites with coarse sandy sediments at Kentra Bay and Loch Ailort where there are organic-rich, almost peaty, surface layers over sandy layers with organic contents of 5% or less and the muddy sites such as Loch Moidart and Loch nan Ceall where there is much less variation of organic content with depth although the average organic contents are relatively high and quite variable. Loch Beag is distinctive in having dispersed but organically quite rich layers. Although fine sediments are dominant at Loch Beag there is sufficient coarser material present for the soils to be better drained than those at Loch nan Ceall or Loch Moidart. 9

Feeding the fish - a Celtic perspective Biomass of the salt marsh vegetation The data obtained from the vegetation sampling at the four Lochaber study sites are presented in Table 5. together with an assessment of the grazing status of each of the sites. Table 5. Biomass of the standing crop of the main plant communities in the Scottish lochs. The dominant species of each vegetation type is named. Figures given in columns 4-6 are of the dry weight of the standing crop (gm -2 ). The final column (Gr.) gives the grazing status of the sites as a whole (+++ = heavily grazed, + = lightly grazed, - = little or no grazing detected). The components named are respectively; the living material, algae of various sorts, and standing dead and litter. Site Marsh Vegetation Green Algae Dead Gr. Kentra Bay Pioneer Marsh Puccinellia maritima 150.3 85.7 1.7 +++ Kentra Bay Middle Marsh Armeria maritima 75.8 48.4 41.3 +++ Kentra Bay High Marsh Festuca rubra 49.6 0.0 4.5 +++ Loch Moidart Middle Marsh Puccinellia maritima 261.2 57.6 69.6 + Loch Moidart High Marsh Juncus gerardii 323.6 1.4 41.9 + Loch Ailort High Marsh Juncus gerardii 323.4 3.7 44.2 + Loch Ailort Low Marsh Puccinellia maritima 233.4 50.8 1.4 + Loch 'n Ceall Middle Marsh Plantago/Triglochin 387.6 4.3 47.0 - Loch 'n Ceall High Marsh Juncus gerardii 597.5 13.8 40.0 - Results from the earlier salt marsh productivity studies in Essex (Boorman and Ashton, 1997) showed that the net annual primary productivity was approximately equal to the peak biomass of the standing crop. Assuming the same to be true for the Scottish sites this would indicate productivities of roughly the same order as the Essex sites and on this basis it is possible to infer that the total annual productivity of the marshes at south Kentra is of the order of 12 tonnes, that of the marshes of Loch Moidart approximately 30 tonnes, that of Loch Ailort approximately 20 tonnes and that of Loch nan Ceall 12 tonnes. These figures can be taken to indicate the materials potentially available for export. The data obtained also shows that the plant communities dominated by Juncus gerardii are the most productive of the various plant communities although some stands of Puccinellia maritima can be nearly as productive. Exchange studies at Loch Beag The data from the water analyses in the field and laboratory are summarised in the Table 6. The data shown have been taken from the spreadsheets of the field monitoring and from the analytical sheets of the laboratory analyses. These very limited figures can be regarded as giving no more than preliminary indications but they do suggest that the marsh is of the juvenile type importing sediment during spring tides but only to a limited degree as the turbidity values are very much lower than in comparable lowland marshes. During ebb tides there appear to be only very limited possibilities for exchange between the marsh and the sea. When the river feeding the loch is at a high rate of flow material is being brought in but although the water is highly coloured the actual turbidity is little higher than on a neap flood. Although quite large water volumes are involved at these times it would appear that the proportion of sediment brought in along this way is quite small compared with the spring tide import of sediment of marine origin. These studies are ongoing. 10

Feeding the fish - a Celtic perspective Calculations on the sediment flux are based on the net import of sediment during a single spring tide. The calculations from the estimated water volume over the marsh at high water and the corresponding sediment load spread evenly over the marsh indicate that the maximum sediment deposition was only 3.8 g m 2 and if it was assumed that half the 706 annual tides were similar (probably optimistic) this would still only indicate a rate of accretion of around 1.3 mm yr -1. More realistic would be the assumption that 25% of the tides were comparable which would indicate a reduction of a half in the accretion i.e. a figure of around 0.7 mm yr -1. These low rates of accretion accord entirely with the concept of "fossil" marsh systems. However, actual rates need to be recorded directly as the redistribution of sediment from the lower mudflats could considerably enhance this figure. At low water springs there is an estimated 50 ha of flats which is 10 times the marsh area so if 20% of the sediment deposited here was transferred to the marsh subsequently this would double the annual accretion rate. This figure of 1.5 mm yr -1 is, however, still low compared with that found in active marsh systems elsewhere. Table 6. Summary of the data from tidal flow studies and water analyses at Loch Beag, August 2001. Turbidities are expressed in mg per litre and chlorinity in parts per thousand. Imports are indicated by '+' and exports by '-'. Event Turbidity (Field) Turbidity (Lab.) Chlorinity Flood 1 11.7 11.0 8.7 Ebb 1 6.6 5.2 26.4 River Flow 1-5.2 0.1 Import +/Export - +5.0 +5.8 N/A Event Turbidity (Field) Turbidity (Lab.) Chlorinity Flood 2-7.2 32.8 Ebb 2-8.8 32.5 River Flow 2-4.0 0.0 Import +/Export - - -1.6 N/A Salt marsh changes at Loch Beag, Skye The resolution of the 1988 aerial photographs limited the observable detail of the salt marsh at that time. It was, however, clear that changes in courses of the main creeks has resulted in the loss of some 800 m 2 of salt marsh. It appears that the majority of the lost marsh would have been low and middle marsh in origin but the changes in the creek system would also have released significant quantities of sediment from parts of the high marsh as well. These losses have been completely offset by considerable gains along the seaward (western) end of the marsh. These gains were estimated as amounting to over 3000 m 2 giving a net gain of 2300 m 2 (0.23 ha) over a period of 13 years. It was noted that one third of the marsh which had been gained during this period has already developed to middle marsh. While there is insufficient evidence to make exact calculations it would appear that most of the changes could be explained by the recirculation of existing sediment from the areas of erosion and from the deepening of the major creeks. A significant proportion of the new areas colonised 11

Feeding the fish - a Celtic perspective involve the colonisation of superficial deposits of mud over shingle and only subsequently are larger quantities of fine sediment involved. Discussion Contrasts in salt marsh soils There are significant variations between the lochs under study with respect to the nature of the material on which the site is based. There is the clear physical distinction between those lochs based on a sandy substrate and those based on silts and muds. There is, however, much less difference between these sites as far as the vegetation itself is concerned. Probably the major difference is that on the sand-based sites such as Kentra Bay and Loch Ailort the marsh surface is not only built up by the deposition of sediment settling out from the water column at high water but also by the deposition of wind-transported sand particles at low water. In those sites where there is a significant river inflow such as Loch Ailort and Loch Beag at least locally sediment layers of terrestrial (fluvial) origin are also found. There is a much greater degree of variability in the salt marsh soils in all these highland loch sites than in the soils of typical lowland marshes. There are clearly more mechanisms and more complex mechanisms at work in the highland sites and it can be said that salt marshes occur in a wider range of situations than elsewhere. Even where there is only a very limited volume of new sediment available salt marshes still develop and survive by the maximum recycling of the existing sediments, both coarse and fine, on the site. In some cases the marshes achieve significant vertical growth as a result of accumulations of organic matter giving higher soil organic levels that in many other marsh sites. Comparisons in plant species distribution and biomass In general the variability of the salt marsh vegetation which occurs in highland lochs is far less than might be expected from the variation in soil and general site characteristics. The tidal range at the Lochaber sites is slightly lower than in the salt marsh sites studied in Essex but the vertical extent of the salt marsh vegetation is very similar in both situations (excluding the extra 'high' marsh above HWST which has developed as result of wave action). The amplitude of the salt marsh vegetation in Lochaber varied from 1.2 m at Loch nan Ceall to 1.7 m at Loch Ailort. This compares closely with the situation at Tollesbury, Essex, where the vertical range of the marsh vegetation is between 1.0 m and 1.8 m in different parts of the area. The less well drained areas appear to be less able to support vegetation at lower and more frequently flooded levels in the marsh. It would appear that plant growth is more successful on the better drained sites than on the sites dominated by fine sediments. Similar conclusions can be reached when the ranges of individual plant species are examined. For example the vertical range over which Puccinellia maritima will grow is 1.9 m in Essex and 1.7 m in Lochaber. The biomass and the inferred productivity for the plant communities in the highland lochs was similar to that of lowland marshes. It was not surprising that the notably eutrophic site, Loch nan Ceall, showed the highest standing crop (biomass) and that the site where there was abundant fine sediment, Loch Moidart, had the next highest standing crop. The productivity of the sites which were essentially sand-based, Kentra and Loch Ailort, nevertheless had significant levels of standing crop and this was despite the relatively high levels of grazing at 12

Feeding the fish - a Celtic perspective these sites, particularly Kentra. The removal of the vegetation by grazing rapidly reduces the biomass of the standing crop and conceals high levels of productivity. The volume of organic matter available for export to adjoining ecosystems is reduced by grazing but at the same time the quantity of partially degraded organic matter is probably increased if the grazing animals remain on the marsh surface for any length of time and thus deposit faeces on the marsh. At least some of the salt marsh and marine invertebrates would be able to benefit from this material and in this way exploit the primary productivity of the marsh ecosystem. Varying potentials for the exchange of nutrients and organic matter Any discussion on the exchange of nutrients and organic matter has to take into account both the availability of material, organic matter or mineral nutrients, for transport and also the existence and functioning of appropriate mechanisms for this to take place. While currently we have reasonably good data on the primary productivity of these salt marshes there is a still a scarcity of data relating to the possibilities for the exchange of mineral nutrients and organic matter. The concept of 'marsh maturity' was developed in relation to the English east coast marshes (Hazelden and Boorman, 1999) but the principles can also be applied more generally. The concept recognised that chronological age was a poor indicator of the probable state of development of a salt marsh and that marsh maturity could often be a cyclical process with the possibilities for regeneration and renewal. Particular reference was made to the existence of 'over-mature' marshes with signs of decay and degeneration and that these marshes could often be younger than the 'immature' marshes which were still at a stage of active growth or, in some cases, of regeneration and renewal. Despite probable low rates of accretion resulting from the generally low levels of suspended sediment in the water column all the highland loch sites the general impression of active marsh growth with very few signs of major marsh decay. Certainly it seemed highly likely that much of the active marsh growth was the result of the recirculation of material within the ecosystem itself or the possible involvement of closely associated ecosystems. This view appears to be supported by the changes over the last 13 years at Loch Beag although even with the small quantities of sediment available for accretion the concentration of this accretion at the lower edge of the marsh, the area most often covered by the tide, would have made a small but significant addition to the sediment available from internal recirculation. The implication that these marshes can be largely self-contained systems should not be taken to mean that the potential for exchanges with adjacent ecosystems does not exist. The primary productivity of the salt marsh is high enough for there to be significant exports of organic matter when the combination of wind and tide is right. It has already been shown that significant quantities of organic matter are available. Salt marshes have also been shown to function as sinks for both mineral nutrients and organic matter (Hazelden and Boorman, 1999) and this would seem to apply to these marshes whether they are functioning directly as water filters in the case of the sand-based marshes or indirectly by the surface properties of the clay particles in those marshes based on fine sediments. The occurrence of high levels of organic matter in certain of the marsh soils could have a special significance with regard to salt marsh function. Certainly in these marshes the below-ground productivity of organic matter is contributing directly to the 13

Feeding the fish - a Celtic perspective vertical growth of the marsh itself as well as contributing greatly to soil biochemical processes. Overall, as already demonstrated for lowland areas, highland salt marshes can and do function as a vital living link between terrestrial and marine communities (Boorman, 2000). While there is much more evidence to be collected with regard to the precise functional role of the highland lochs the evidence so far all points to the probability of them fulfilling a similarly important role. Habitat conservation and management - the way ahead The marshes of lowland Britain are under threat in many different ways and particularly from rising sea levels resulting from global warming (Boorman, 1992). This process is accentuated by the process of 'coastal squeeze' where rising sea levels and rigid sea walls are combining to reduce greatly the area available for intertidal habitats. Large areas of salt marshes are being lost to port developments and other similar activities. Salt marshes are also threatened by increasing levels of pollution in various forms. At the same time there is an increasing recognition of the key contribution salt marshes make to the biological diversity in many lowland coastal areas. The marshes of the highlands must also be considered to be under threat although in rather different ways. While the more obvious threats to salt marshes are widely recognised in lowland areas there seems to be a general assumption that there is an absence of threats to salt marsh, and other intertidal communities for that matter, in the highland situation. Nevertheless it is clear from these studies that this is not the case and that appropriate management measures are equally needed in both the highland and lowland situation. Salt marsh sites are being threatened by the building of new roads and bridges; water management is affecting the patterns of river flow and the associated processes; and rising sea level can induce physical, biological and biochemical changes in the marshes at a rate that may well be unsustainable even over relatively short periods of time. The one feature which was identified as making the highland marshes particularly suitable for detailed study is their individually isolated situation. This also means that they are individually more vulnerable particularly with regard to the maintenance of their rich biological diversity. Species lost for one site may not easily re-establish from neighbouring sites because of the degree of isolation. If the studies described here are to be summarised under two points these have to be firstly that there is a clear need for active conservation measures and secondly that there is equally a need to work to understand the role these marshes play in the complex links between the land and the sea. Both these points must be seen against a background of the overall biodiversity and cultural richness of the Scottish Highlands. Acknowledgements The author wishes to express his thanks to the many colleagues within the EUROSAM project in Britain and Europe for their stimulating discussions and their encouragement, and to thank Scottish Natural Heritage and the various landowners involved for their active support. The author gives special thanks to Neil Beaton of Struan, Skye, for his help and encouragement and Mary Boorman for editing the manuscript. The financial contribution of DG XII of the 14

Feeding the fish - a Celtic perspective Commission of the European Community received under Grant ENV4-CT97-0436 is gratefully acknowledged. References Boorman, L.A. 1992. The environmental consequences of climate change on British salt marsh vegetation. Wetlands Ecology and management 2: 11-21. Boorman, L.A. 1999. Salt Marshes - present functioning and future change. Mangroves and Salt Marshes 3: 227-241. Boorman, L.A. 2000. The functional role of salt marshes in linking land and sea. In: British Saltmarshes. Ed. B.R. Sherwood, B.G. Gardiner and T. Harris. Forrest Text for the Linnean Society. pp. 1-24. Boorman, L.A. and Ashton, C. 1997. The productivity of salt marsh vegetation at Tollesbury, Essex, and Stiffkey, Norfolk, England. Mangroves and Salt Marshes 1: 113-126. Burd, F. 1989. Salt marsh survey of great Britain. An inventory of British Salt Marshes. Research and Survey in nature Conservation No. 17. Nature Conservancy Council: Peterborough. Chapman, V.J. 1960. Salt marshes and Salt Deserts of the World. Hill, London. Chapman, V.J. 1977. Wet Coastal Ecosystems, In: van der Maarel, E. (ed.) Ecosystems of the World; Vol. 1. Elsevier: Amsterdam. Dame, R. 1989. The importance of Spartina alterniflora to Atlantic Coast estuaries. Aquatic Sciences 1: 639-660 Hazelden, J. & Boorman, L.A. 1999. The Role of Soil and Vegetation Processes in the Control of Organic and Mineral Fluxes in Some Western European Salt Marshes. Journal of Coastal Research 15: 15-31. Lefeuvre, J.C. 1996. The effects of environmental change on European salt marshes: structure, functioning and exchange potentialities with marine coastal water. Vols. 1-6. Univeristy of Rennes, France. Lefeuvre, J.C. and Dame, R.F. 1994. Comparative studies of salt marsh processes in the New and Old Worlds: an introduction. pp. 169-179. In: Mitsch, W.J. (ed.), Global Wetlands: Old World and New, Elsevier: Amsterdam. Odum, E.P. 1980. The state of three ecosystem level hypotheses regarding salt marsh estuaries: tidal subsidy, outwelling, and detritus-based food chains. In. Kennedy, V.S. (ed.), Estuarine Perspectives. Plenum: New York. Pethick, J.S. 1992. Saltmarsh Geomorphology. pp. 41-62. In: Allen, J.R.L. and Pye, K. 15

Feeding the fish - a Celtic perspective (eds), Saltmarshes: Morphodynamics, Conservation and Engineering Significance. The University Press, Cambridge. Rodwell, J.S. 1999. British Plant Communities Volume 5: Maritime Communities and Vegetation of Open Habitats. The University Press, Cambridge. Teal, 1962, Energy Flow in the salt marsh ecosystem in Georgia. Ecology 43: 614-624. Loch Beag Arisaig Bay Loch Ailort Loch Moidart Kentra Bay Figure 1. Map showing the location of the five sea lochs studied; four sea lochs were selected in the Lochaber area (from south to north Kentra Bay Traigh Ceantra, Loch Moidart Ceann Loch Muideart, Loch Ailort Ceann Loch Aileort, Arisaig Bay Loch nan Ceall) and one on the west coast of Skye (Loch Beag part of Loch Bracadale). 16

Feeding the fish - a Celtic perspective Figure 2. Kentra Bay (Traigh Ceantra), view from the south looking towards the Eileanan nan Gad showing the salt marsh vegetation extending across the sand flats. Figure 3 Loch Moidart (Ceann Loch Muideart) view of the head of the loch from the northeast showing the clumps of vigorous salt marsh vegetation growing on mud. 17

Feeding the fish - a Celtic perspective Figure 4. Loch Ailort (Ceann Loch Aileort), view from the A830 (The Road to the Isles) looking to the south-west showing the extensive salt marshes on a sand and shingle substrate. The river Ailort can be seen flowing across on the left hand side. 18

Feeding the fish - a Celtic perspective Figure 5. Arisaig Bay (Loch nan Ceall) a close-up view of typical salt marsh vegetation showing high marsh dominated by Juncus gerardii and Plantago maritime and salt pan showing the rocky nature of the substrate. Figure 6. Loch Beag (part of Loch Bracadale), view of the seaward end of the marshes taken from the top of Creag Mhór to the south east. Sheep grazing the marsh are visible as white 19

Feeding the fish - a Celtic perspective dots and the main channel of the Amar River is running from bottom right to the middle of the left of the picture. 20

The effects of land-use changes and arterial drainage on the runoff component of the hydrological cycle: in the Irish context Keshav P. Bhattarai* and Kieran M. O Connor* *Department of Engineering Hydrology, National University of Ireland, Galway. Abstract The most extreme flooding/drought events in the recent years are a result of both climate change and human influences on the catchment behaviour. In the Irish context, the notable man-made changes are land improvement works and arterial drainage schemes. These effects are being assessed using lumped conceptual and distributed physically based rainfall-runoff models and the results verified statistically. Résumé Les inondations et sécheresses les plus importantes de ces dernières années sont le résultat à la fois de changement climatique et des activités humaines sur le basin versant. En Irlande, celles-ci sont les travaux de défrichement et de drainage des sols. Les effets de ces aménagements sont évalués à partir de modèles pluie ruissellement conceptuels, et les résultats sont vérifiés statistiquement.

Describing actual and future flood hydrological regimes Pierre Javelle (1), Eric Sauquet (2), Jean-Michel Grésillon (3) 1 University College Dublin, Centre for Water Resources Research, Earlsfort Terrace, Dublin 2, Ireland ; email: pierre.javelle@ucd.ie 2 Cemagref, UR Hydrologie-Hydraulique, 3bis quai Chauveau, 69336 Lyon, cedex 09, France 3 Laboratoire d étude des Transferts en Hydrologie et Environnement, 38402 St Martin d'hères, France Abstract This paper presents a Flood Frequency Analysis (FFA) which takes into account the notion of duration: the flood-duration-frequency (QdF) method. Two applications are presented. The first one concerns its application to describing observed flood regimes in a case study of a hundred French catchments. Different types of floods are defined and represented on a map of France. The second application illustrates the potential for describing the impact of a presumed climate change, by analysing simulated discharges. Results for the whole Rhone basin (86 200 km 2 ) will be soon available in the framework of the GICC-Rhone project. This will allow impacts of climate change to be spatially represented and critical areas identified. Key words Flood Frequency Analysis (FFA), flood-duration-frequency approach (QdF), flood volume analysis, hydrograph shape, impact of climate change. Résumé Cet article présente une analyse fréquentielle des crues prenant en compte la notion de durée : l approche débit-durée-fréquence (QdF). Deux applications sont présentées. La première concerne la description du régime des crues actuel, sur une centaine de bassins français. Differents régimes ont ainsi pu être définis et cartographiés. La seconde concerne la description des évolutions futures du régime, cette fois ci en analysant des débits simulés. Les resultats pour l ensemble du bassin du Rhône (86 200 km 2 ) seront bientôt disponibles dans le cadre du projet GICC-Rhône. L application de la méthode proposée pourra permettre de représenter spatialement les impacts d un changement du climat, et ainsi d indentifier les zones les plus affectées. Mots clefs Analyse fréquentielle des crues, approche débit-durée-fréquence, volume de crue, forme d hydrographe, impact du changement climatique. 1

1. Introduction In these last decades, new challenging hydrological problems have emerged, such as predicting the impact of a predicted climate change or of human activity on the behaviour of rivers. More than ever, there is a need for tools to describe in a concise way the hydrological regime. Because of their devastating potential, this paper focuses more specifically on floods, but other aspects of the hydrological regime (such as low flows) can be treated in the same way. Flood Frequency Analysis (FFA) is a commonly used tool for describing the flood regime. However, most often, FFA characterises a flood event only by its instantaneous peak or its maximum daily flow. This information is essential but insufficient for many purposes. Flood severity is not only defined by its peak value, but also by its volume and duration. This paper presents a FFA method which takes into account the notion of duration: the floodduration-frequency (QdF) approach. It shows its application to describing observed flood regimes in a case study of a hundred French catchments. Its potential for describing the impact of presumed climate change on flood behaviour is also illustrated. 2. The flood-duration-frequency (QdF) approach 2.1 Brief review As mentioned in introduction, most often, FFA describes a flood event only by its peak. However, more complete methods exist. One has been initiated by Ashkar (1980) and called peak-volume analysis. For each observed flood, a starting and an ending date are defined. Flood peak, volume and duration are then determined and analysed as random variables. Using the Gumbel mixed model, Yue et al. (1999) give a joint distribution, a conditional probability functions and the associated return periods for these variables. Ouarda et al. (2000) develop a regional flood frequency procedure based on canonical correlation analysis which can give estimates for ungauged basins. Another similar approach, mentioned by Cunnane (1988) under the term Volume over threshold analysis, is based on partial duration series (PDS). The data involved consists of volume of flow exceeding a threshold Q 0, where Q 0 is the existing within-bank or within-levee flow or a proposed design channel flow. According to Cunnane (1988), theoretical derivations of distribution of flood volumes has been discussed by Todorovic (1978), Askar and Rousselle (1982) and Correia (1987). This present paper deals with a third FFA approach which takes into account the duration notion and is called QdF analysis. Unlike the peak-volume approach, the duration is not considered as a random variable, but as a fixed parameter. QdF analysis is similar to the intensity-durationfrequency (IdF) analysis commonly utilised for rainfall. First, averaged discharges are computed over different fixed durations d. Then, for each duration, a frequency distribution is fitted to the maxima of these averaged discharge values. This approach is similar to the Flood volume over different durations approach mentioned in the Flood Studies Report (NERC, 1975), but was abandoned in the recent Flood Estimation Handbook (Institute of Hydrology, 1999). Only a limited number of papers focus on QdF analysis: Sherwood (1994), Balocki and Burges (1994), Galéa and Prudhomme (1997). 2

This paper deals with the QdF formulation presented by Javelle et al. (1999). This model was tested with good results on more than two hundred catchments: in France (Javelle et al., 1999; Javelle et al., 2000), Martinique (Meunier, 2001), Guadeloupe (Galea and Javelle, 2001), Burkina-Faso (Mar et al. 2002), Romania (Mic et al., 2002) and Canada (Javelle et al., 2002). 2.2 Methodology and QdF curves formulation Instead of studying only the maximum peak flood (Qmax), the QdF approach considers maximal flood volumes over different durations noted d. Maximal volumes are divided by d in order to deals with m 3 /s and noted noted V d. In this way, V d represents a mean flow, and instantaneous peak flood is the V d limit case, when d tends to zero (Figure 1). Qmax Q(t) V d Maximal volume over d V d = d d Figure 1: Variables definition t V d values are sampled using a procedure described by Lang (1995), which is now completely automatic. A moving average with a window length d, is computed over the whole instantaneous streamflow time series Q(t), providing a new time series Q d (t): t+ d / 2 1 (1) Qd ( t) = Q( τ ) dτ d t d / 2 V d values are sampled by selecting maximal values of this new time series. Different method can be carried out, as in any Flood Frequency Analysis (i.e. Peak-Over-Threshold or Annual Maximum analysis). Repeating these steps for N different durations d, N samples of V d values and N corresponding distributions V d (T) are fitted. The limit case d=0 corresponds to the original time series of instantaneous values (if available). The aim of the QdF modelling is to describe these N mean flow distributions in one general framework, by fitting to them a unique formula V(d,T) as a function on d, the length of the averaging window, and T the quantiles return period. The sketch of Figure 2 summaries the two described steps: a) A frequency distribution is fitted to the maximal values of the Qd(t) (1) time series. Here, the sizes of the averaging window are d 1, d 2 and d 3 respectively and no assumption is made about a specific probability distribution; b) A V(d,T) formulation is fitted, allowing quantiles to be represented as a function of d, for different return period T, respectively T 1, T 2 and T 3. 3

V(d,T) a) d 1 V(d,T) b) d 2 d 3 T 1 T 2 T T 3 d T 1 T 2 T 3 d 1 d 2 d 3 Figure 2: Flood-duration-frequency curves We can establish the link with the IdF approach by indicating that the maximum runoff volume (V d multiply by d) and the maximum mean streamflow V d are respectively equivalent to the depth and the intensity of precipitation. This study uses the following V(d,T) formulation: V ( d = 0, T ) (2) V ( d, T ) = 1+ d / Δ Where V(d=0,T) is the distribution corresponding to instantaneous peak discharges, Δ is a parameter which determines the curvature of the hyperbolic relationship in Figure 2b. These parameter are fitted on sampled data (Javelle et al., 2002). This relationship concerns flood analysis, but a similar QdF approach exists for low flows (Galéa et al., 2000). 3. Value of the QdF approach and his Δ parameter The value of the QdF approach is briefly illustrated by the following case study. Both studied basins are part of the database presented in paragraph 4. : the Jabron River at Souspierre (85 km 2, located in South of France) and the Nied Allemande river at Faulquemont (187 km 2, located in the north). Instantaneous streamflow time series are available for a period of 27 and 14 years, respectively. Figure 3 shows the results of the peak flood analysis (peak-over-threshold values fitted by an exponential distribution). For both studied examples, distributions are similar. 4

The Jabron River at Souspierre (85 km²) 60 60 The Nied Allemande River at Faulquemont (187 km²) instanteneous peak flood (m 3 /s) 50 40 30 20 10 instanteneous peak flood (m 3 /s) 50 40 30 20 10 0 0 0.1 1 10 100 0.1 1 10 100 T (year) T (year) Figure 3: Instantaneous peak flood distributions Despite the similarity in peak values, the flood hydrograph shapes are completely different, as illustrated by Figure 4 which for each catchment shows an actual flood hydrograph with a peak of approximately 10-year return period. This result is confirmed by an analysis of all the sampled floods, Figure 5. Each flood is plotted as a point, with its peak discharge along the x- axis and a time parameter denoted as t 1/2 on the y-axis. t 1/2 corresponds to the time during which the discharge exceeds half the value of the peak. For the 10-year flood, t 1/2 is equal to 12 hours for the Jabron River and 4 days for the Nied Allemande (Figure 4). The Jabron River at Souspierre (85 km²) The Nied Allemande River at Faulquemont (187 km²) streamflow (m 3 /s) 40 35 30 25 20 15 Q max Q max /2 Flood of 14 th November 1969 t 1/2 =12 hours streamflow (m 3 /s) 40 35 30 25 20 15 t 1/2 = 4 days Flood of 21 th December 1993 10 10 5 5 0 0 1 2 3 4 5 6 7 8 9 10 11 0 0 1 2 3 4 5 6 7 8 9 10 11 time (day) time (day) Figure 4: Hydrographs of the observed 10-year flood 5

t1/2 (day) 12 10 8 6 4 Nied Allemande Jabron 2 0 0 10 20 30 40 50 60 Inst. peak flood (m 3 /s) Figure 5: Time during which the half peak is exceeded (t 1/2 ) for each observed flood In term of water resource management, it is clear that both catchments have completely different flood regimes. However, as illustrated by Figure 3, a classical flood frequency analysis, considering only peak floods, would miss this point. The same catchments are now studied according the QdF methodology described in paragraph 2, concentrating on mean discharges over durations d, successively equals to 0 (instantaneous streamflow), 1, 2 and 5 days. It can be seen that now, the analysis is able to make the distinction between the different behaviour of the two catchments (Figure 6). The Jabron River at Souspierre (85 km²) 60 60 The Nied Allemande River at Faulquemont (187 km²) 50 : Vd over d=0 (inst.) : Vd over d=1 day 50 : Vd over d=0 (inst.) : Vd over d=1 day mean flows (m 3 /s) 40 30 20 : Vd over d=2 days : Vd over d=5 days mean flows (m 3 /s) 40 30 20 : Vd over d=2 days : Vd over d=5 days 10 10 0 0.1 1 10 100 T (year) 0 0.1 1 10 100 T (year) Figure 6: Distributions of mean flows over different durations d In the case of the Jabron River, the whole set of distributions shows considerable spread, while for the Nied Allemande River they lie within a narrow sector. In the first case, floods contain relatively small volumes, while on the second case, high volumes are involved (Figure 4). Consequently, when studying mean flows over the same durations (i.e. 5 days), the distribution obtained for the Jabron River is lower than for the Nied Allemande. 6

The difference between both examples is also characterised when studying the fitted QdF curves, represented as a function on d (Figure 7). Widely spread distributions of Figure 6 leads to curved hyperbolas, while more concentrated distributions leads to flat hyperbolas. The Jabron River at Souspierre (85 km²) 60 60 The Nied Allemande River at Faulquemont (187 km²) mean flows (m 3 /s) 50 40 30 20 : V d (T) : V(d,T) T=20 years T=5 years T=2 years T=0.5 year mean flows (m 3 /s) 50 40 30 20 : V d (T) : V(d,T) T=20 years T=5 years T=2 years T=0.5 year 10 10 0 0 2 4 6 d (day) 0 0 2 4 6 d (day) Figure 7: V(d,T) formulation fitting According to the V(d,T) formulation (2), it appears that hyperbola shape is controlled by the Δ parameter: the flatter hyperbolas are, the higher Δ is. For example, a value of 1.3 days is obtained from fitting QdF curves to the Jabron River, and 11 days for the Nied Allemande River. Consequently, this parameter is a very good indicator of the flood dynamic behaviour. Compared with other characteristic durations manually obtained from hydrographs analysis, this parameter can be calculated in an objective and automatic manner. 4. Application in France for characterising different actual flood regimes The QdF approach described in the previous sections was applied to 103 French catchments. Their areas range from 7 to 9387 km 2, and the available length of streamflow records ranges from 9 to 33 years. Javelle et al. (1999) showed that the fitted V(d,T) formulation gives acceptable results. Two flood regime indicators are identified: - the Δ duration, - the 10-year-return period instantaneous peak flood, V(0,10), simply noted Q10 The Δ duration is a good indicator of the flood dynamic behaviour, as it has been explained in the previous section. Q10 gives an information about peak flood. The 10-year-return period is chosen because its seems to be a good compromise between the rarity of the event (we want to study extreme events), and the data availability (due to record length, longer return period, i.e. 100 years, would lead to larger uncertainties). Figure 8 shows Q10 and delta as a function of the area of each studied catchment (in logarithmic scales). Area explains 70% of the Q10 s variance, while only 17% of the Δ s variance is explained. This result is mainly due to the fact that delta reflects the transfer processes within catchments. For a same rainfall amount, these processes depends on the surface area, but also on many other characteristics, such as topography, soil type, land use, etc. 7

10000 1000 x = 0.99 y 0.83 R 2 = 0.70 100 10 x = 1.03 y 0.22 R 2 = 0.17 Q10 (m 3 /s) 100 10 Δ (day) 1 1 1 10 100 1000 10000 Area (km 2 ) 0.1 1 10 100 1000 10000 Area (km 2 ) Figure 8: Q10 and Δ as a function of the area In order to compare Q10 and Δ for catchments of different sizes, both regime indicators have been divided by the area raised to the power of a coefficient deduced from the slope of the loglinear regression line between the variables. (Figure 8). 0.83 Q 10* = Q10 / A (3) 0.22 Δ * = Δ / A (4) Figure 9 represent the 103 studied basins as a function of their two scaled regime indicators, Q10* and Δ*. Plots, each representing one catchment, seems to follow an imaginary hyperbola, from top-left to right-down part of the graphic. Catchments located in top-left have a regime characterised by very intense and short floods (flash floods) and conversely, catchments located in the bottom-right are characterised by less intense and slower floods. 9 intensity indicator : Q10* 8 7 Strong and quick floods 6 5 Intermediate floods 4 Weak and slow floods 3 2 1 0 0 1 2 3 4 5 dynamic indicator : Δ* 6 Figure 9: Definition of flood regime typologies Three groups have been graphically delineated on Figure 9: one group for strong and quick floods, one for weak and slow floods, and one intermediate group. Then, each catchment has been represented on a map of France, with a symbol depicting its flood regime. 8

(Figure 10). It can be seen that flash floods basins are mainly located in the south east France. This expected result is explained by the fact that the Mediterranean region is known for its heavy precipitation occurring on very small areas, especially in autumn. For example, these kind of events caused huge damages in the Aude region, in 1999. Legend : Strong and quick floods : Intermediate : Weak and slow floods Figure 10: French studied catchments location according to their flood typology 5. Characterising the impact of a climate change on the flood regime The QdF approach may also been used to characterise flood regime modifications, due to changes in climate, or land use. The idea is to carry out the same QdF analysis on streamflows generated by a rainfall-runoff model using climatic model outputs. 5.1 Data and methodology The presented application is based on results provided by the GICC-Rhone project (Leblois, 2002). This ongoing project studies the impact of the climate change on the French Rhone basin (86 200 km 2 ). The climate change scenario modelled is a doubling of the CO 2 concentrations, following the predictions of the Intergovernmental Panel on Climate Change (IPCC, 2001). Four Global Circulation Models (GCMs) have been tested. Simulations indicated that the doubling of CO 2 concentrations will induce an increase of rainfall total amount in winter. For being used by distributed hydrological models, these results were downscaled using the "perturbations" method. More details are provided by Boone et al. (2000). During a former project (Gewex-Rhone), two distributed hydrological models were calibrated and validated using atmospheric forcings derived from meteorological database over the period 1981-1995 (Golaz-Cavazzi et al., 2001). Sauquet and Leblois (2001) showed that models were able to reproduce observed hydrological regimes. The GICC-Rhone project applies these models with six different climate change scenarios with equal probability (all assume a CO 2 doubling, but results depend on GCMs and resolutions). In the following application, results are shown for the standard MODCOU hydrological model and for one climate scenario. In the 9

future, all available hydrological models and climate scenarios should be considered in order to have an idea of the possible variability in the possible responses. 5.2 Application The application is illustrated for one example: the Lanterne River at Fleurey-lès-Faverney, a tributary to the north part of the Rhone basin. The QdF model is used to describe the flood regime for this 1020-km² basin: - measured discharge time series over the period 1981-1995 (Figure 11a) - discharge time series simulated by the MODCOU model over the same period 1981-1995 for validation under past conditions (Figure 11b) - discharge time series simulated by the MODCOU model forced by a climate change scenario (Figure 11c). QdF curves derived from the reconstituted time series (b) are close to the observed ones (a). This indicates that the hydrological MODCOU model can correctly reproduce the actual flood regime. Comparison of graphs a and c indicates that the considered climate change scenario leads to a significant increase of the floods magnitude. mean flows (m3/s) 450 400 350 300 250 200 150 (a) observation : Vd over d = 1 day : Vd over d = 4 days : Vd over d = 6 days : Vd over d =12 days mean flows (m3/s) 450 400 350 300 250 200 150 (b) reconstitution : Vd over d = 1 day : Vd over d = 4 days : Vd over d = 6 days : Vd over d =12 days mean flows (m3/s) 450 400 350 300 250 200 150 (c) climate change scenario : Vd over d = 1 day : Vd over d = 4 days : Vd over d = 6 days : Vd over d =12 days 100 100 100 50 50 50 0 0.1 1 10 100 T (year) 0 0.1 1 10 100 T (year) 0 0.1 1 10 100 T (year) Figure 11: Comparison of QdF curves for the Lantern River at Fleurey-lès-Faverney, (a) derived observed streamflow measurements, (b) reconstitued times series estimated by MODCOU and (c) expected under climate change Figure 12 represents the flood regime of (a), (b) and (c) cases, as a function of the magnitude and the dynamic of floods: Q10 and Δ respectively (see Figure 9, section 4). This graph confirms that the observed flood regime (a) and thus simulated by the MODCOU model (b) are very close. Concerning the impact of a possible climate change (c), results indicate that floods magnitude (Q10) may increase significantly, while their dynamic behaviour (Δ) may remains constant. Several reasons may explain the stability of Δ: - a flood saisonality analysis under climate change indicates that the Lantern River still demonstrates rainfall-fed hydrological regime (i.e. floods are still generated by rain in winter); - the same runoff and sub-surface processes are expected due to the unchanged surface status of the basin. Results for the whole Rhone basin are soon expected in the framework of the on-going GICC- Rhone project. 10

Q10 (m3/s) 400 350 300 250 200 150 100 50 climate change (c) observation (a) reconstitution (b) 0 0 5 10 15 20 Δ (day) Figure 12: Characterisation of the climate change impact for the Lantern River at Fleurey-lès- Faverney 6. Conclusion This paper presented a FFA method which takes into account the notion of duration: the QdF approach. On this basis, different flood typologies have been defined for a hundred French catchments. The possibility to characterise the impact of a climate change has also been illustrated. Both applications, concerning actual and future flood regimes, have shown the potential of the QdF approach. However, different points need to be investigated further and research is ongoing. The flood typology presented in section 4 has been empirically derived and can be improved. Other catchments, under different climatic conditions, have to be studied. Furthermore, the relationship between flood hydrographs and QdF curves needs to be better understood. Concerning the climate change study, results for the whole Rhone basin area (86 200 km 2 ) will be soon available in the framework of the GICC-Rhone project. This will allow the flood regime map presented in section 4 to be established for future regimes. Climate change impacts could be spatially represented and critical areas (hot-spots) identified. Furthermore, work on the uncertainties of the involved models should be attempted. Indeed, this difficult task is essential for determining if observed modifications are significant or not. Acknowledgement The present paper is a methodological contribution to the "impact of climate change on discharges assessment" task of the on-going GICC-Rhone project funded by the French Ministry of the Environment (MATE). The financial supports provided by the European Union for the Post-doctoral Marie Curie fellowship of P. Javelle is also acknowledged. Authors would like also to thank Dr. Michael Bruen (UCD) for assistance in producing the paper. 11

References Ashkar, F. (1980), Partial duration series models for flood analysis, Ecole polytechnique de Montréal, PhD thesis, Montréal, Canada. Askar, F., Rousselle, J. (1982), A multivariate statistical analysis of flood magnitude, duration and volume, in V.P. Singh (Ed.), Statistical analysis of Rainfall and Runoff, Water Resource Publication, 651-668. Balocki, J. B., Burges, S. J. (1994), Relationships between n-day flood volumes for infrequent large floods, Journal of Water Resource Planning and Management, 120(6): 794-818. Boone A., Noilhan J., Etchevers P., 2000 : GICC-Rhone Climate Scenarios. Technical report, CNRM, 55 p. Correia, F.N. (1987), Multivariate partial duration series in flood risk analysis, in V.P. Singh (Ed.) «Hydrologic frequency modelling», 541-554, D. Reidel Publ. Co. Cunnane, C. (1989), Statistical distributions for flood frequency analysis, World Meteorological Organization, Operational Hydrology Report No 33. Galéa, G., Javelle, P. (2000). Modèles débit-durée-fréquence de crue en Guadeloupe, Rapport d'étude, protocole Cemagref-Lyon, DIREN Guadeloupe et Météo-France, Cemagref-Lyon: 25p+Annexes. Galéa, G., Javelle, P. Chaput, N. (2001), A discharge-duration-frequency model adapted for low flows (in French), Revue des Sciences de l'eau, 13(4):421:440. Galéa, G., Prudhomme, C. (1997), Basic notions and useful concepts for understanding the modeling of flood regimes of basins in QdF models (in French), Revue des Sciences de l'eau, 1: 83-101. Golaz-Cavazzi, C., Etchevers, P., Habets, F., Ledoux, E., Noilhan J. (2001), Comparison of two hydrological simulations of the Rhone basin, Physics and Chemistry of the Earth, Part B: Hydrology, Oceans and Atmosphere, Volume 26, Issues 5-6, 2001, Pages 461-466 IPCC (2001), Climate Change 2001: The Scientific Basis, Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change (IPCC), summaries available on: http://www.ipcc.ch/pub/reports.htm Institute of Hydrology (1999), Flood Estimation Handbook (5 volumes), Institute of Hydrology, Wallingford. Javelle, P., Galéa, G. et Grésillon, J.M. (2000), The flow-duration-frequency approach : former and new developments (in French), Revue des Sciences de l Eau, 13(3): 305-323 Javelle, P., Grésillon, J. M. et Galéa, G. (1999), Discharge-duration-Frequency curves modeling for floods and scale invariance (French and English) Comptes Rendus de l'académie des Sciences, Sciences de la terre et des planètes, 329: 39-44. Javelle, P., Ouarda T. B. M. J., Lang, M., Bobée, B., Galéa, G., Grésillon, JM. (2002), Development of regional flood-duration-frequency curves based on the index-flood method, Journal of Hydrology, 258(1-4): 249-259. Lang, M. (1995), Traitements de bases et intervalles de confiance des quantiles de crues ; Techniques d'échantillonnage par la méthode du renouvellement, PhD Thesis, Université Joseph Fourier: 296p. Leblois, E. (2002), Evaluation des possibles impacts du changement climatique par modélisation distribuée (projets Gewex-Rhone et Gicc-Rhône), proceedings of the SHF conference "variations climatiques et hydrologie" - Paris - décembre 2001 (in press) Mar L., Gineste P., Hamattan M., Tounkara A., Tapsoba L. et Javelle P., (2002), Flood-durationfrequency modeling applied to big catchments in Burkina Faso (in French), 4 th FRIEND international conference, 18-22 March 2002, Le Cap, Afrique du Sud. Mic, R., Galéa, G., Javelle, P. (2002), Floods regionalization of the Cris catchments: application of the converging QdF modeling concept to the Pearson III law. Conférence des pays Danubiens, 2-6 septembre 2002, Bucarest (Romania). 12

Meunier, M. (2001), Regional flow-duration-frequency model for tropical island of Martinique, Journal of Hydrology, 247(1-2): 31-53. NERC (1975), Estimation of flood volumes over different durations,. Flood Studie Report, Vol. 1, chap. 5: 243-264. Ouarda, T. B. M. J., Girard, C., Cavadias, G. S. et Bobée, B. (2000), Regional flood peak and volume estimation in a northern Canadian Basin, Journal of Cold Regions Engineering, American Society of Civil Engineering (ASCE), 14(4): 176-191. Sauquet E., Leblois E. (2001), Discharge analysis and runoff mapping applied to the evaluation of model performance, Phys. Chem. Earth (B), 26(5-6): 473-478. Sherwood, J. M. (1994), Estimation of volume-duration-frequency relations of ungauged small urban streams in Ohio, Water Resources Bulletin, 30(2): 261-269. Todorovic, P. (1978), Stochastic models of floods, Water Resource Research, 14(2): 345-356 Yue, S., Ouarda, T. B. M. J., Bobée, B., Legendre, P. et Bruneau, P. (1999), The Gumbel mixed model for flood frequency analysis, Journal of Hydrology, 226: 88-100 13

Communication of Flood Hazard to the Public Jerry Grant & David Wilson M.C. O Sullivan & Co. Ltd. Abstract There is a growing awareness of flood risk in Ireland, and as more and more studies are completed, one of the issues to be addressed is how this information might be disseminated in a more comprehensible way to the public. Both Historical Flood and Predicted Flood Hazard Maps have various uncertainties associated with them. If the information is gathered once the flood peak has passed, then Historical Flood Spread maps can be a reflection of how land drains, and may not explain the mechanisms of how it floods. Historical Flood Spread maps can also be greatly influenced by blockages and other external factors. Predicted Flood mapping area is also influenced by many factors, not least of which is input data uncertainties. Does the public need to know about these uncertainties and if so how should these uncertainties be portrayed? There is also the question of what sort of information are the public really interested in? Is it the 1 in 100 year flood risk. Or is it how many times will a property be flooded in the next 10 years and how deep is the flood likely to be? Or is it the extra cost of living within a floodplain? Through an examination of these issues and looking at examples of Flood Hazard maps from around the world, this paper aims to stimulate debate. The aim of this paper is not to try to provide definitive answers but rather to ask questions and introduce ideas. Léiriú Baol Tuile don Phobal Achomair Tá níos mó aithne ar an mbaol tuile in Éirinn sa lá atá inniú ann, agus mar a chríochnaítear níos mó staidéir ar an ábhar seo, is gá díriú ar conas is fearr an eolas astu a léiriú go cruinn don phobal. Tá rudaí áirithe nach bhfuil deimhin faoi léarscáileanna tuile stairiúil agus réamháisnéiseach. Má chruinnítear an eolas tar éis ard na tuile, léiríonn na léarscáileanna stairiúil conas mar a thaoscann an talamh, ach b fhéidir nach míníonn siad conas a tuileann an talamh. Freisin, is féidir le blocanna agus rudaí imeallach eile an-thionchar a bheith acu ar na léarscáileanna tuile stairiúil. Tagann léarscáileanna réamháisnéiseach freisin faoi tionchair éagsúla, ar nós neamhchinnteacht faoin eolas ar chóir a chur isteach iontu. Ar gá don phobal bheith ar an eolas faoin neamhchinnteacht seo, agus más gá, conas gur fearr é seo a léiriú dóibh? Freisin is gá a fhiafrú cén sórt eolas ina bhfuil fíor-shuim ag an phobal? An é an baol tuile gach céad bliain? Nó cé chomh minic a shroicfidh uiscí tuilte foirgneamh i dtréimhse deich mbliana agus cé chomh doimhin a bheith na h-uiscí? Nó an é an costas breise a bhaineann le maireachtáil í magh tuile? Is é cospóir an pháipéar seo ná diospóireacht a spreagadh, trí staidéar a dhéanamh ar na h- ábhair seo agus trí scrúdú a dhéanamh ar shamplaí de léarscáileanna baol tuile ó ar fud na cruinne. Ní hé cospóir an pháipéar seo freagraí cinnte a chur i láthair, ach ceisteanna a chur agus tuairimí a thabhairt chun cinn. 1

The Current Awareness of Flooding Issues in Ireland There is a growing awareness of flooding issues in Ireland. This is a result of recent extreme floods and may also be influenced to a small degree due to the European Union Water Framework Directive (WFD). While concentrating on water quality, one of the purposes of the European Union Water Framework Directive WFD is to establish a framework which contributes to mitigating the effects of floods and droughts on water quality. To be able to mitigate the effects of floods there has to be an awareness of the behaviour of the water body during a flood event. This is acknowledged in the Directive whereby temporary deterioration in the status of bodies of water shall not be in breach of the requirements of this Directive if this is a result of circumstances which are exceptional when the conditions under which circumstances that are exceptional are stated in the River Basin Management Plan. Therefore the River Basin Management plans will have to identify what an exceptional flood event is and then it will have to ensure steps are taken to ensure that unexceptional events do not breach the requirements of the WFD. This will have to involve assessing the possible impacts of an unexceptional event. Ironically a couple of weeks after the adoption of the WFD (20 October 2000) Ireland was experiencing extreme flooding. As a direct result of the floods of 5-7 November 2000 the then Minister of State at the Department of Finance, Mr Cullen TD, announced in February 2001 that a programme of flood risk mapping would be initiated 1. He announced that this would take place in stages, the first being the collection of historical data and the second being the exercise of producing flood risk mapping for the country. The purpose of these flood risk maps is to provide local authorities with the information they need for making planning decisions. Mr Cullen also stated that It is essential for all future planning decisions that local authorities take into account areas known to be at risk from flooding. However, later in that same month in response to a question about the interaction between existing planning legislation and flood risk mapping he stated that Once [the historical information] has been made available to the local authorities developing the information further will be considered and at that stage the implications for planning legislation will be addressed. 2 This shows that currently the Irish Government may not see a necessary link between flood risk mapping and planning legislation sufficient to require them to be dealt with together. There was also no indication that the National Spatial Strategy, which is designed to provide a twenty year spatial development framework for Ireland, would take into account flooding issues. In fact the final Spatial Planning Unit consultation publication Implications for the Way Ahead released in September 2001, does not have a single mention of flooding in the whole 30 page document. This seems to indicate that at the top level of Irish Government there is an awareness of the lack of flooding information in Ireland but there does not seem to be an awareness of how flooding issues will fit into the bigger picture of the future development of Ireland. Most importantly there has not been a firm commitment by the Irish Government to date to provide funding for a 1 Dail Debates Official Report 13 February 2001 2 Questions - 20 February 2001 Written Answers 2

longer term flood hazard mapping programme or to link the existing program into a review of planning legislation. There has also not been any indication as to how the number of flooding studies that have been and are being completed will fit into this flood risk mapping programme. There are now also a number of studies of flood impact that have been undertaken for major road projects. It would be appropriate to take full advantage of all flooding information contained in these studies in developing national datasets. Flood Mapping Flood mapping is not a precise science and there are several sources of information all with their own sets of inaccuracies and numerous ways of portraying this imprecise information. Types of Flood Mapping There are two principle generic types of flood mapping that are frequently used. These are defined below. - Predictive (Indicative) Flood Mapping is the mapping of flood levels and extents that are predicted to occur for a given single or range of selected frequencies or return periods. The prediction for fluvial flood maps is often based on analysis in the following format; o hydrological modelling to estimate design flows, o detailed channel and topographical floodplain survey, o hydraulic modelling, including calibration from recorded events, to predict flood levels, o mapping of extents according to predicted levels. Predictive tidal flood extent mapping requires o predicted tidal levels (based on tidal models and/or frequency analysis of recorded levels) for the selected return period and o topographical data to determine areas below peak levels. Predictive flood mapping is a multi-step process and there are uncertainties associated with all the stages. The significance of these uncertainties needs to be assessed at each stage of the flood mapping process to ensure that useful data is produced at the end of the project. - Historic Flood Mapping is the mapping of observed (as opposed to predicted) flood events and extents as recorded by survey, photography, video, the press, memory, etc. Data incorporated into historic flood maps could include recorded flood level, depth and flow as well as extent and incidence/occurrence, each of which can potentially be associated with a probability of occurrence or event return period. The data can be collected from a variety of sources, but requires interpretation and verification before use. Although in theory there should not be uncertainties about historical flood maps, these maps can be influenced by many factors including; how soon after the event was 3

the data gathered, flow obstructions and even waves caused by passing vehicles. The usefulness of historical data needs to be assessed as part of the mapping process. The production process for predictive maps normally involves a greater degree of analysis than that required for historic maps, and as such is more labour and cost intensive to produce. Predictive maps are therefore generally only produced for areas of specific interest, which have often been identified through an examination of the historic flood data. On the other hand, the level of detail, potential area of coverage and degree of applicability of predictive maps are significantly greater than that generally available from historic maps. The data that is collected and interpreted for historic mapping is often also necessary for the calibration of the hydrological and hydraulic modelling required for indicative mapping. The two types of mapping are not mutually exclusive, and the preferred methodology may incorporate both, with historic mapping undertaken as a precursor to predictive mapping. Limitation of Flood Maps Historic flood maps developed in a rigorous manner can portray objective factual flood data in that they convey recorded data, whereas predictive maps carry an inherent degree of uncertainty due to their dependence on the concepts of probability and inexact modelling techniques. The term indicative map is indeed often used as a reflection of this uncertainty. In contrast however, the usefulness of historic maps can be limited unless a frequency or return period is associated to the mapped data, which again will incorporate a degree of uncertainty. The success of a flood mapping programme is dependent in part on the confidence placed in the maps by the endusers, and it is essential therefore that the degree of uncertainty is minimised as far as reasonably possible by rigorous verification of historic data and thorough calibration of any modelling. The uncertainty inherent in flood mapping is sometimes covered using the concept of freeboard, freeboard being a factor of safety added onto the flood levels to cover any uncertainties in the method used to obtain the flood level. Another limitation of flood maps is derived from misunderstanding of the concept of probability (e.g. how can two separate ten or twenty year events occur in successive years?) and uncertainty. It is very important that the size of the event and what this represents is made clear to the end-user, who more than likely is not going to be an extreme event statistician. If the key event is the 1 in 50 year event is it to be called:- the 1 in 50 year event the flood event that has a 2% chance of occurring in a any one year or even the flood event that has a 75% chance of occurring in the average person s lifespan. Many of the determining factors in flood mapping are prone to change. Variable catchment characteristics such as land-use and the degree of urbanisation will have a significant impact on the rate and volume of runoff, while drainage and channel maintenance will impact on the attenuation of a flood event and on water levels for given flows. Climate change is another major factor that could significantly change the estimated design flows for predicted events or the return periods associated with historic events. The flood maps should take account of known past changes in an 4

appropriate manner, such as highlighting any of the above changes that have occurred since an historic event, or allowances made for sea level rise of increasing frequencies in predictive levels or flows. Dissemination of Information There is clearly little point in undertaking a flood mapping exercise without disseminating the information to the end-users. There are a number of issues that should be considered in how to manage the dissemination process. The first issue is to identify the end-users, as this will have significant bearing on the methodology adopted. If the data is only to be provided to a limited number of users for whom document control will not be a problem, then physical dissemination in hardcopy or CD/DVD might be appropriate, whereas for a large number of users, this might cause problems as discussed below. What sort of information is to be included on the flood maps is also dependent on the needs of the end-users. As noted above under Limitations of Flood Maps, flood data will change with time, new data will become available, and existing data may become incorrect following, for example, the construction of a flood mitigation scheme or major landuse change. The problems of revising the maps and disseminating updates to all end-users should be considered, particularly with reference to preventing the use of superseded information. Large scale map revision and re-production will also have cost and resource implications. A third major consideration in the methodology for dissemination is the type of data to be made available. If the data to be supplied is limited, e.g. finished maps only or summary reports, then the above issues become less significant. If on the other hand, support information such as copies of press reports or photographs are to be released, then the increased volume of data and media representation of it can exacerbate the potential problems. A final consideration that is worth reiterating at this point is the clear explanation required on the appropriate use of the maps, particularly in relation to the degree of reliability/uncertainty and the concepts of frequencies and return periods. While secondary to the flood data, adequate provision of this information is essential to the success of a flood-mapping program. International Examples of Flood Hazard Mapping Tasmania Australia Background In Tasmania, joint Federal, State and Local Government funding has allowed the development of floodplain maps in many flood prone urban areas of Tasmania. These arrangements have led to the development between 1990 and 1996 of a set of floodplain maps covering seven urban areas. The projects were funded under the Federal Water Resources Assistance Program, the National Landcare Program and the Natural Heritage Trust. 5

In addition to the inundation maps, comprehensive documentation has been prepared for each flood study. These reports provide a means of preserving technical results and model details for the review of flood inundation in the future. They are considered living documents and the public are actively encouraged to contribute to the on going development of the document. An example of this is the fact that a blank flooding record sheet is provided at the end of the document. The detailed information collected as part of these studies does allow extrapolation to more extreme flood events. However, it was considered that such estimates did not warrant the cost of the work involved during the current round of projects. Provided there is an adequate information collection system established for future floods, estimates of the effects of extreme events may be possible when floodplain maps are revised. While the current maps are based upon the best available data, new flood events and longer periods of record may lead to significant revisions to the estimates. To ensure the continuing applicability of the maps prepared under this program it has been deemed necessary to revise the estimates of flood probabilities every 15 to 20 years as well as after major flood events. Many of the studies were constrained by inadequate hydraulic information. This identified a clear need to ensure that the hydraulics of all significant floods are monitored in future and this is best completed by establishing cross-sections, across which velocity and depth measurements can be taken, together with documentation and photographs of floods. Information included in the Flood Maps Floodplain maps provide the best available estimates of the areas inundated with 1:20, 1:50 and 1:100 annual exceedance probabilities. The definition of a flood event used is those areas which might be expected to be inundated with a 1 in 20, 1 in 50 or 1 in 100 chance in any given year. The floodplain maps show the location of the normal channel of the watercourse, surrounding features and development, ground contours (at 0.5m intervals) areas inundated and river cross section locations. The water surface elevations of the flood shown are included in tables at the appropriate river cross section A table showing the probability that a design flow would be equalled or exceeded at least once in a 20 year and 50 year period is also included on the map. Purpose of the Flood Maps The purpose of the maps is to provide town planners and developers concerned with future development with an indication of those areas subject to flooding. The maps are also linked to the Tasmanian Building Regulations, 1994 which state that buildings containing habitable rooms must not be built unless the floor level is at least 300 millimetres above the designated flood level for that land. Disclaimers included with Flood Maps Disclaimers include:- 6

The maps do not deal with the flooding which could occur in smaller water courses as a result of heavy local rain Extent of flooding shown on the map is approximate only The limit of the 100 AEP flood is not a boundary between flood prone and flood-free zones. Areas outside the limit of 1:100 AEP could well be inundated by rarer events. How the Flood Maps are distributed. The maps can be downloaded from the Internet and they can also be viewed at the relevant local authorities and at libraries of the Department of Primary Industries, Water and Environment. British Columbia Canada Background In British Columbia (BC) the Floodplain Mapping Program is a joint initiative by the Federal and British Columbian governments to provide information which will help minimize flood damage in British Columbia. The program identifies and maps areas that are highly susceptible to flooding. These areas may be designated as floodplains by the federal and provincial Environment Ministers. Information included in the Flood Maps The flood maps in B.C. are called floodplain maps and they delineate the area that can be expected to flood, in the 200-year flood. Floodplain maps show the location of the normal channel of a water course, surrounding features or developments, ground elevation contours, flood levels and floodplain limits. Within the floodplain, flood level isograms are used to show the water elevation during a 200-year flood. (The maps may also include the 20-year flood level, which is used in applying Health Act requirements for septic tanks.) A flood level isogram is defined as a line which spans the floodplain, plotting the location at which the floodwater is expected to reach the indicated elevation. The elevation of floodwater between each isogram can be interpolated. Purpose of the Flood maps Designated floodplains are subject to development restrictions. Subdivisions within a floodplain require the approval of the regional water manager. Crown agencies such as the Canada Mortgage and Housing Corporation do not support development on designated floodplains unless adequate floodproofing measures are taken. Local governments may impose further restrictions. Floodplain maps are administrative tools which depict minimum elevations for floodproofing. Minimum floodproofing requirements can then be incorporated into building bylaws, subdivision approvals, and local government planning and regulations. 7

Disclaimers included with Flood Maps Disclaimers include:- A 200 year flood can occur at any time in any given year; the indicated flood level may be exceeded; and portions of the floodplain can flood more frequently. The accuracy of the location of a floodplain as shown on a map is limited by the base topography. It is generally assumed to be plus or minus one-half the increment of the ground contours. Floodplain maps do not provide information on site-specific flood hazards, such as land erosion or sudden shifts in the channel of the watercourse. Other sources of water, roads or other barriers can restrict water flow and effect local flood levels. As well, obstructions such as ice, debris, flooding in surrounding areas, groundwater or other phenomena can cause flood levels to exceed those indicated on the map. Land adjacent to a floodplain may be subject to flooding from tributary watercourses. Map users should note the dates of topographic mapping, aerial photography, river surveys and map issue, and dates of development in the map area. Subsequent developments within the floodplain (natural or construction) may effect flood levels and render site-specific map information obsolete. How the Flood Maps are distributed. Floodplain maps are distributed by Crown Publications Inc., the authorized distributor of B.C. legislative publications. Floodplain maps can also be viewed at the offices of the Regional Water Managers. Additionally most local authorities will have copies of mapsheets within their jurisdiction available for viewing. City of Cape Town South Africa Background In an effort to promote awareness about potential flood risks, Indicative Floodplain Maps have been prepared for the major river and canal systems within the City of Cape Town Area. These maps are based on the latest hydrological studies undertaken by the City of Cape Town and will be continually improved as new data becomes available. Urban development has in some cases, been allowed to encroach into floodplains placing life and property at risk. This has necessitated implementation of flood alleviation schemes to deal with areas subjected to frequent and repeated flooding. Information included in the Flood Maps Floodplains are defined as the natural "overspill" areas inundated during flood events. The Indicative Floodplain Maps show the location of the normal channel of a water course, surrounding features or developments, roads, suburb labels and floodplain limits. The interactive maps indicate floodplain areas (shaded orange) likely to be inundated during floods equal to or exceeding the 50-year recurrence flood. They also include the interesting definition of a 50 year flood being the flood 8

that has a 75% chance of occurring within the average person's lifetime. This is an interesting way of giving the users a relevant definition for the flood. Purpose of the Flood Maps The Indicative Floodplain maps are anticipated to be of particular interest to property developers, engineering, town planning and environmental consultants, emergency services, private property owners, estate agents, etc. Development Control Guidelines for Floodprone Areas have recently been formulated in an effort to guide future development adjacent to rivers and other major drainage corridors. These guidelines will require the 20 year, 50 year and 100 year floodlines and Danger Flood Level to be indicated on all development layout plans. The concept of a Danger Flood Level, based on flow velocity and depth criteria, has been introduced for application in specific scenarios. This allows rational assessment of risk to human life (e.g. wading depths) and stability of structures under flood conditions to be utilised for disaster management purposes. Disclaimers included with Flood Maps Disclaimers include:- It is important to note that no warranty can be given regarding the particular degree of flood risk for property located within the shaded floodplain areas, nor that flooding can be definitely anticipated. There are many variables which cannot be depicted at this course mapping scale. How the Flood Maps are distributed. There is an interactive map on the Internet which indicates floodplain areas (shaded red) likely to be inundated during floods equal to or exceeding the 50-year recurrence flood. Further information is available on request from the City of Cape Town, Catchment Management Department. Environment Agency United Kingdom Background Nearly two million homes and businesses lie in natural river and coastal floodplains in England and Wales. The Environment Agency's new Indicative Floodplain Maps provide a general overview of areas of land in natural floodplains and therefore potentially at risk of flooding from rivers or the sea. The maps are based on historical flood records and geographical models. They indicate where flooding from rivers, streams, watercourses or the sea is possible. However the maps do not show flood defences which offer vital protection in many areas. Environment Agency's argument has been that the defences are present because of a risk and that they can never eliminate that risk. The Agency makes these maps widely accessible to make it easier for people to find out if they are in a flood risk area and what local flood warning arrangements exist. 9

They also inform planners and developers and are a key tool in informing decisions on controlling development in floodplains. Information included in Flood Maps The map shows the natural river and coastal floodplains in England and Wales. A floodplain is defined as the natural 'overspill 'area when a river rises above its banks or when high tides or stormy seas cause flooding of low-lying coastal areas. River floodplains are coloured blue. Coastal floodplains are coloured green. Some areas are at risk of flooding from rivers and the sea. In this case the colour of the flooding with the greatest extent will be shown. They are based on computer models, survey data and historical records and will be updated as new information becomes available. For flooding from rivers the maps indicate the extent of flooding for a one in one hundred (or a one per cent) chance of flooding each year For flooding from the sea and tidal estuaries, the maps show a one in two hundred (or 0.5 per cent) chance of flooding each year Purpose of the Flood maps The maps are design to be used by the Environment Agency, local authority planners, the emergency services, insurance industry and the public to assist decision making on: control of development within the floodplain improving the local flood warning service and targeting flood risk awareness campaigns location, planning, design, construction and maintenance of flood defences emergency planning Disclaimers included with the Flood Maps Disclaimers include:- Flood forecasting is not a precise science and the maps can only give a general indication of risk areas. How the Flood Maps are distributed. The maps were put on CD at 10,000 scale and distributed to public bodies who needed to know the flood risk as part of their business. Due to the high level of public interest in the maps it was decided to also place them on the internet, although at a reduced scale. FEMA United States of America Background In 1968, Congress created the National Flood Insurance Program (NFIP) in response to the rising cost of taxpayer funded disaster relief for flood victims and the increasing amount of damage caused by floods. The NFIP makes Federally-backed flood insurance available in communities that agree to adopt and enforce floodplain management ordinances to reduce future flood 10

damage. National Flood Insurance is available in more than 19,000 communities across the United States and it s territories. The NFIP is managed by the Federal Emergency Management Agency s (FEMA) Federal Insurance and Mitigation Administration and Mitigation Directorate. The Federal Insurance and Mitigation Administration manages the insurance component of the NFIP, and works closely with FEMA s Mitigation Directorate, which oversees the floodplain management and mapping components of the Program. FEMA s Mitigation Directorate set standards for the production of Floodway maps that are used to produce Flood Insurance Rating Maps (FIRM) which are used by the Federal Insurance and Mitigation Administration as part of the NFIP. Information included in the Flood Maps Floodway maps show flood boundaries, the normal channel of a water course, surrounding features or developments, river cross section locations and limits of study. FIRMs have similar background mapping and show Special Flood Hazard Areas (SFHA) which are used in administering the NFIP, see Purpose of Flood Maps section below. Flood Hazard Area designations appear as dark and light tints. Dark tints indicate areas of increased flood hazard; light tints indicate areas of lesser flood hazard. Floodplain boundaries show the limits of the 100- and 500-year floodplains. Elevation reference marks are found on all flood maps. These marks identify points where a ground elevation is established by survey. These elevations are usually expressed in feet; for some communities, however, the elevations are shown in meters. Descriptions of the marks, including their elevations are provided; however, descriptions of locations appear in different places, depending on the format of the Flood Map. Most Flood Maps cover only one community. If that community is a county, flooding information is shown only for the areas under the jurisdiction of the county government. Thus, flooding information will not be listed for incorporated areas (e.g., towns and cities) on the Flood Maps produced for most counties. Separate Flood Maps are prepared for incorporated areas. Recently, however, FEMA has produced countywide Flood Maps. These maps show flooding information for all of the geographic areas of a county, including the towns and cities. Purpose of the Flood Maps The purchase of flood insurance is mandatory as a condition of receipt of federal or federally-related financial assistance for acquisition and or construction of buildings in SFHA of any participating community. Those communities notified as flood-prone which do not apply for participation in the NFIP within 1 year of notification are ineligible for federal or federally-related financial assistance for acquisition, construction, or reconstruction of insurable buildings in the SFHA. Conventional loans are available in the SFHA of non-participating communities for these purposes at the lender's risk. 11

Disclaimers included with the Flood Maps Disclaimers include:- This map does not necessarily identify all areas subject to flooding, particularly from local drainage sources of small size or planimetric features outside the Special Flood Hazard Area. How Flood Maps are Distributed Copies of Flood Maps are made available by FEMA, for a nominal fee. The Flood Maps are also on file at the local community repository. In 1997 FEMA developed a plan to modernize the flood mapping program. As a result the Flood Maps can now be viewed on the internet. Conclusion Through the examination of international flood hazard mapping programs two significant driving forces behind the programs emerge:- 1/. To raise public awareness of flood issues 2/. To control development within the floodplain In Ireland, individual Local Authorities and OPW are aware of the need to understand flood risk, to adopt mitigation measures in critical areas and to assist development planning by developing risk mapping. However, there is no universal application of these techniques, nor is there a standard approach being adopted. Many Local Authorities are aware of the sensitivity of river flooding to land-use development and there is increasing application of Sustainable Urban Drainage Systems (SUDS) to new urban development (e.g. Dublin City Council has formal policies in place). These provide for attenuation measures to be incorporated in all new development. Obviously, a co-ordinated national approach to the issue of Flood Risk Mapping would be the optimum way forward and is to be encouraged. 12

Representation of climate change in flood frequency estimation Solomon S. Demissie and Conleth Cunnane Department of Engineering Hydrology, National University of Ireland, Galway, Ireland Abstract This study presents a simplified method to account climate change impact in the conventional flood frequency estimation procedures. A stochastic cluster point process model known as the Modified Bartlett-Lewis Rectangular Pulses Model (MBLRPM) has reproduced the rainfall process satisfactorily. The evaporation series, which is less important in the context of flood in wet catchments with negligible snowfalls, is fitted to an Auto-Regressive Moving Average model, ARMA (1,1). The rainfall-runoff transformation of the catchments in this study is well captured by the revised Soil Moisture Accounting and Routing (SMARG) conceptual model. Monthly dimensionless precipitation scenario profiles are derived from ECHAM4/OPYC3 GCM output. Stochastic downscaling employed to represent precipitation scenarios in the MBLRPM. Then, a continuous simulation scheme was organized from components of the MBLRPM, the ARMA (1,1), the SMARG, the stochastic downscaling routine, and from flood frequency processor. Precipitation scenarios that cover the plausible ranges of the General Circulation Models (GCMs) estimate and some scaling concepts are hypothesized, and introduced in to the rainfall process by perturbing the MBLRPM parameters. In order to investigate the impact of climate change on the flood frequency, calibrated parameters of the MBLRPM, the ARMA (1,1) and the SMARG models, and the dimensionless precipitation scenario profiles are routed through the continuous simulation scheme. A climate change correction growth curve for flood frequency is then developed. Résumé Cette étude présente une méthode simplifiée pour intégrer les changements climatiques dans les procédures classiques d estimation des fréquences de crue. Un modèle stochastique, connu sous le nom de Modified Barlett- Lewis Rectangular Pulses Model (MBLRPM) a reproduit de facon satisfaisante le phénomène des précipitations. Les chroniques d évaporation, cette derniere étant moins importante pour les crues se produisant dans des bassins versants humides peu soumis aux chutes de neiges, ont été en adéquation avec un modèle nomme Auto-regressive Moving Average model, ARMA (1,1). La transformation des précipitations en ruissellement sur le bassin versant concerné par l étude est bien rendue par le modèle conceptuel appelé Soil Moisture Accounting and Routing (SMARG). Chaque mois, les profils de scénarii de précipitation, sans dimensions, sont calculés a l`aide du modèle ECHAM4/OPY3 GCM, qui fournit un résultat global. Ce résultat a éte adapté pour le modèle MBLRPM, fonctionnant à une échelle géographique plus locale.. Puis, un programme de simulation continue a été mis en oeuvre à partir des composants du MBLRPM, du ARMA (1,1) du SMARG, du programme stochastique et du programme d analyse de fréquences de crue. Nous avons supposé que les scénarios de précipitation couvrent l étendue plausible des estimations du General Circulation Models (GCMs) et que les concepts d échelle sont vérifiés, et les avons introduits dans le modèle de pluie en modifiant les paramètres du MBLRPM. Pour étudier l impact des changements climatiques sur les fréquences de crue, les paramètres étalonnés des modèles du MBLRPM, du ARMA (1,1) et du SMARG, ainsi que les profils de scénarios des précipitations sans dimension sont introduits dans le programme de simulation continue. Une courbe corrigeant les fréquences de crue en fonction des changements climatiques peut alors être définie. 1. Introduction Design flood estimation is a principal component in the study, design and management of water resource projects that are planned to achieve economical and safe water control and/or use. The most widely applied procedure for this purpose is an event-based flood frequency estimation in which future relationships between flood quantile and its frequency of occurrence is estimated from statistical analysis of observed historical events. The analysis might be done either at-site or at regional level (Cunnane, 1988; Hosking and Wallis, 1997). The second class of methodology for flood frequency estimation is a derived distribution approach, which is pioneered by Eagleson (1972). The probabilistic distribution of flood flows is determined from the density functions of climatic and hydrologic variables. This method suffers from the simplified assumptions made about the distribution and interaction of these variables for the purpose of mathematical tractability (Hebson and Wood, 1982; Diaz-Granados et al., 1984; 1

Bierkens and Puente, 1990). The most appealing but not fully investigated approach to flood frequency estimation is reported to be continuous simulation modeling (IH, 1999). The method simulates catchments water balance continuously by routing continuous climatic inputs through moisture accounting hydrological models. Different investigation carried out recently proved the promising potential of the method for land use and climate change impact assessment studies (Calver and Lamb, 1996; Blazkova and Beven, 1997; Lamb, 1999, Cameron et al., 1999; Demissie, 1999; Hashemi et al., 2000; Loukas et al. 2002; Yu et al. 2002). All the above procedures estimate the magnitude of flood flow expected to occur at a specified return period from the current hydro-meteorological variables. But Global Circulation Models (GCMs) are forecasting pronounced variation in climatic variables under different integrations of greenhouse gas emission scenarios. For example, the main flood producing variable, precipitation, shows a general increasing trend in autumn and winter in the Northern Hemisphere mid- and high- latitudes (IPCC, 2001). Therefore, it is evident that flood frequency would be affected by changes in annual precipitation variability and by changes in rainfall characteristics (intensity and duration). This calls for plausible and physically based approach to incorporate climate-induced changes in the conventional flood frequency estimation procedures. Assessment of the effect of climate change on the hydrologic systems has been broadly approached in three different ways. The first category directly simulates GCM outputs through watershed water balance models (Kamga, 2001; Loukas et al., 2002). But GCM results are too crude to apply at catchment level due to uncertainties involved in coarse spatial resolution and in simplified parameterization of cloud cover and rainfall process. The second approach identifies trends in climatic variables from long historical records and generates future time series based on the current trend, which is used as an input for hydrological models (Yu et al., 2002). The drawback behind this method is the expectation of similar trends in present and future climatic variables, which lacks an account of social, economical and environmental changes and interaction between elements of the climate. The third methodology applies a downscaling mechanism that can transfer high-resolution GCM simulated variables into watershed levels either by nesting regional climate models over GCM outputs, by establishing statistical relationships between largescale circulation patterns and local climatic variables (Brandsma and Buishad, 1997; Wilby et al., 1998; Conway and Jones, 1998), or by representing the changing pattern of the GCMs outputs in to the respective variables stochastic process at the watershed level (Burlando and Rosso, 1991). The current study applies stochastic downscaling to plausible precipitation scenarios, percentage change of monthly precipitation, and then develops the corresponding flood frequency scenarios using continuous simulation modeling. 2. The Study Area and Data Sets This study was conducted on Fergus river catchment at Ballycorey in County Clare, Ireland. It has a drainage area of 512 km 2. The catchment is mostly flat with karstic nature in the upper lands. The dominant land use patterns are more farmlands with some proportions of scrubland, coniferous plantation, natural woodland and mixed woodland around the outlet. The Burren National Park is located at the center of the catchment, which will preserve the land in its present state for long time. Small caves, scattered ponds, and minor lakes are the major hydro-geological features of the catchment. The stream networks and the lakes are concentrated at the lower half part of the catchment while the upper part is a karstic land. The areal mean daily rainfall of the catchment is computed from the daily rainfall data of four meteorological stations: Carron, Corofin, Crushen and Kilmaley. The mean annual rainfall of the catchment is 1425 mm. Monthly potential evapo-transpiration from the nearby Shannon Airport synoptic station and mean daily flow from Ballycorey hydrological station (the outlet) from 1975 to 1999 inclusive are also used for calibration and verification of the rainfall-runoff model. The mean annual daily flow for this period was 10.25 m 3 /s. Monthly mean total precipitation prediction [1950 2099] from the German Max-Planck Institute s GCM (HCHAM4/OPYC3) for constant CO2 (control) and for greenhouse gas integrations of IS92a emission scenario is downloaded from the DKRZ ftp database organized on behalf of the IPCC Data Distribution Centre. From the data at a grid point close to the catchment (8.4375W, 51.4162N), dimensionless monthly profile of the changing factors of precipitation is established as discussed in section 4.2. 3. Modeling of Meteorological and Hydrological Variables Continuous simulation study requires generation of input variables to the water-balance model and proper specification of the model. The common input variables are rainfall and potential evapo-transpiration. Therefore, it is necessary to model the rainfall process, the potential evapo-transpiration series and the rainfall-runoff transformation independently using the respective historical time series. 2

3.1. Stochastic rainfall modeling The best representation of rainfall process can be achieved through cluster point process models. The parameters of these models physically explain the rainfall process. Previous studies done by Khaliq (1995) and Khaliq and Cunnane (1996) showed that Modified Bartlett-Lewis Rectangular Pulses model (MBLRPM) performs well in Ireland. Following Rodriguez-Iturbe et al. (1988), the MBLRPM defined as follows. Storm origins arrive in a Poisson process of rate λ. Each storm origin is followed by a Poisson process of rate β of cell origins. The process of cell origins terminates after a time that is exponentially distributed with parameter γ. The durations of the cells are exponentially distributed random variables with parameter η and each cell depth is a random constant exponentially distributed with mean μ X. The η values for distinct storms are independent random variables having a common gamma distribution with index α and scale parameter ν. In order to change the mean inter-arrival interval of cells and the mean storm duration time randomly with cell duration, dimensionless parameters κ = β/η and φ = γ/η are introduced. The number C of cells per storm has a geometric distribution with mean μ C = 1 + κ/φ. Generally, the MBLRPM has six parameters: λ, κ, φ, ν, α and μ X. The full analytical expression of the first- and second-order properties for both depth and aggregated process are given in Rodriguez-Iturbe et al. (1987 and 1988). The MBLRPM is applied to the areal mean daily rainfall of the catchment on monthly basis so that the stationary assumption of the model could be satisfied. The monthly empirical estimates [1975 1999] of mean, variance, lag-1 autocorrelation and dry proportion at 1-day and 2-days levels of aggregation are matched with the respective analytical expressions of the aggregated process for parameter estimation. Rosenbrock s method of nonlinear unconstrained optimization algorithm is used to minimize a weighted least square objective function. The estimated and validated parameters are listed in Table 1 and the performance of the model is displayed in Figure 1. Table 1. Estimated parameter values of the MBLRPM for daily rainfall of Fergus Catchment. Month λ (day -1 ) κ φ ν (day) α μ X (mm.day -1 ) Jan 0.50636 0.13078 0.02489 1.86020 30.42782 25.36800 Feb 0.52266 0.14836 0.02323 1.13195 31.30844 27.01556 Mar 0.64982 0.08860 0.02279 1.64738 42.55167 31.89741 Apr 0.12901 0.12380 0.01599 5.25120 24.88441 10.94763 May 0.37741 0.15949 0.03353 1.51322 30.87696 22.05181 Jun 0.52997 0.05628 0.03030 1.01563 33.03481 56.42801 Jul 0.43008 0.06357 0.01907 1.24320 36.35539 41.04176 Aug 0.24244 0.06316 0.01080 1.76717 28.24481 37.46912 Sep 0.41144 0.04433 0.01253 1.21746 35.65247 59.20959 Oct 0.49870 0.04786 0.00838 0.79702 44.67371 79.01536 Nov 0.55715 0.05392 0.01359 0.95851 35.65328 59.14481 Dec 0.10594 0.17143 0.00784 3.37363 22.42182 13.52191 3

Rainfall Depth in mm 60 50 40 30 20 JANUARY Hist. Sim. Rainfall Depth in mm 60 50 40 30 20 MAY Hist. Sim. 10 10 0-2 -1 0 1 2 3 4 5 EV1 Reduced Variate 0-2 -1 0 1 2 3 4 5 EV1 Reduced Variate Figure 1. Comparison of historical and MBLRPM simulated annual maximum areal mean rainfall for heaviest (January) and the lightest (May) months. 3.2. Evapo-transpiration series modeling The autocorrelation and partial autocorrelation functions of the series suggest that first-order Auto-Regressive and first-order Moving Average process, ARMA (1,1), suffices to model this time series. If the evaporation series E t has a mean of μ, the ARMA (1,1) model can be expressed as: 2 ( 1 φb) ( E t μ) = ( 1 θb) σ aξt, (1) where, φ is the auto-regressive parameter, θ is moving average parameter, B is the backward shift operator, σ a 2 is the variance the white noise process, and ξ t is the standard normal ordinate. For the purpose of satisfying the stationary condition, the modeling exercise is performed on monthly basis. Table 2 portrays the estimated and validated model parameters. Table 2. Estimated parameter values of ARMA (1,1) model for daily evapo-transpiration. Month φ θ 2 σ a μ Jan 0.97163 0.03539 0.11246 0.280 Feb 0.96082-0.30986 0.14216 0.708 Mar 0.96533 0.08535 0.09385 1.169 Apr 0.96567-0.18083 0.11852 2.014 May 0.95325 0.14513 0.23423 2.765 Jun 0.97945-0.25724 0.25959 3.065 Jul 0.97720 0.06263 0.30748 2.803 4

Table 2 continued Month φ θ 2 σ a μ Aug 0.96528 0.08481 0.22504 2.364 Sep 0.96901-0.33162 0.13909 1.690 Oct 0.96899 0.03861 0.06239 0.804 Nov 0.96310-0.21100 0.06703 0.320 Dec 0.96594 0.16786 0.07963 0.140 3.3. Rainfall-runoff modeling Since the proposed continuous simulation modeling involves generation of lengthy time series as many realization as possible, the rainfall-runoff model should not be complicated and computationally costly. Therefore, a lumped conceptual model developed in this Department and whose performance in such climatic regions was found to be reliable is selected for this study. Its development starts from the original Layers Model of O Connell et al. (1970), which is modified to SMAR (Soil Moisture Accounting an Routing) model by Khan (1986). Then, Liang (1992) introduced the concept of separating quick and slow runoffs in the SMAR model and came up with the SMARG model. Tan and O Connor (1996) and Mingkai (1996) made further improvement on this model. But, the SMARG version that is available in the UCGMODEL package is adopted in the current study. The SMARG model, as most conceptual models, has two major components; water balance and routing components. The water balance component simulates non-linear processes involved in runoff generation on conceptual basis and hence generates surface and ground water runoffs from rainfall and evaporation input. Six parameters are involved in the process: (1) T converts evaporation input series to model-estimated potential evaporation series, (2) H determines the proportion of the generated direct runoff from excess rainfall, (3) Y is the maximum infiltration capacity depth, (4) Z is the soil moisture storage depth of the layers, (5) C is the evaporation decay coefficient of the soil moisture storage layers in such a way that less evaporation occurs from deeper layer, and (6) G determines the proportion of the ground water generated runoff from moisture in excess of soil storage capacity. The routing component transforms the generated surface and ground water runoffs into discharge at the catchment outlet through linear time-invariant storage systems of Nash cascade of N equal reservoirs and simple linear reservoir respectively. This attenuation and diffusion process adds three more parameters: (1) N is a shape parameter and (2) NK is a scale parameter of a gamma unit impulse response of the Nash cascade reservoirs, and (3) Kg is the storage coefficient of the ground water simple linear storage. All together, the SMARG model has nine parameters. Using the twenty-five years [1975 1999] of synchronous daily data of rainfall, evapo-transpiration and stream flow of the catchment; the SMARG model was calibrated and verified for all sets of parameters so that all features of the model can be used in the simulation study. After some preliminary modeling exercises, a memory length of 15 days, a warm up period of 60 days, a calibration period of 17 years and a verification period of 8 years and the Rosenbrock s optimization algorithm are used for parameter estimation. The Nash-Sutcliffe s efficiency criterion (R 2 in %) of 92.70 and 91.31 achieved during calibration and verification respectively. Table 3 shows the list and estimated values of the parameters. The graphical display of the observed and simulated flow series shown in Figure 2 confirms the good performance of the model for the Fergus catchment. 5

Table 3. Optimized parameters of the SMARG model for Fergus catchment. Water Balance Component Routing Component Parameter Estimate Parameter Estimate T 0.992929 N 2.18895 H 0.512684 NK 16.0486 Y 250.540 Kg 34.3327 Z 483.373 C 0.564493 G 0.699249 Flow in cumecs 70 60 50 40 30 20 10 0 Rainfall Obs. Flow Sim. Flow 0 10 20 30 40 50 60 70 01/01/83 01/02/83 01/03/83 01/04/83 01/05/83 01/06/83 01/07/83 01/08/83 01/09/83 01/10/83 01/11/83 Rainfall in mm 01/12/83 Figure 2. Comparison of observed and SMARG model simulated flow series for Fergus catchment. 4. Development of Climate Scenarios The main objective of the present study is to develop flood frequency scenarios that can represent the effect of climate change on flow extremes from plausible ranges of climate scenarios known from GCM experiments using stochastic downscaling and continuous simulation modeling. However, floods from catchments that are wet all over the year with negligible snowfalls are less sensitive to changes in evapo-transpiration. This was verified by simulation studies conducted on the Fergus catchment. It is well understood that precipitation is the main deriving element of the water balance and plays the major role in flood producing mechanisms. Therefore, the impact of climate change on flood flows could be investigated by downscaling precipitation scenarios into the rainfall stochastic process and then to watershed process. 4.1. Stochastic downscaling of precipitation scenarios Following the pioneer work of Burlando and Rosso (1990), precipitation scenarios in terms of percentage change or rate of change of mean monthly precipitation between climate-enhanced and control GCM simulations, stochastic 6

modeling and scaling concepts could be applied to predict future precipitation patterns at operational resolution level. Burlando and Rosso (1991) introduced the methodology using Neyman-Scott Rectangular Pulses model to investigate the pattern of extreme storm rainfall under climate change. In this study the theoretical approach is further developed for the Modified Bartlett-Lewis Rectangular Pulses model. The scaling relationship between the second-order properties of the aggregated rainfall depth process is used to find the rate of change of the variance from the mean, the adopted scenario. The scaling exponents of central moments (mean and variance) of the 12 months were not significantly different from each other. The non-seasonality nature of the scaling exponent is taken as sufficient condition to assume the same value during the current and future climate. From empirical analysis of the areal mean daily rainfall of the catchment at different levels of aggregation (h), the following scaling relationship was obtained (see Figure 3a): var[ Y h i ] = c h 2 { E[ Y ]} ψ, where c = 5.42599 and ψ = 0.68815. (2) i Consequently, changing factor of variance R 2 σ can be written as in terms of the precipitation scenario, changing factor of mean R m, as follow: 2 R R (3) = ( ) ψ 2 σ m (a) 10 (b) 6 LOG [Variance] 8 6 4 2 y = 1.3763x + 1.6912 R 2 = 0.9994 LOG [Mean Cell Intensity] 5 4 3 2 y = -0.7829x + 1.1438 R 2 = 0.8396 0 0 1 2 3 4 5 LOG [Mean Rainfall Depth] 1-5 -4-3 -2-1 LOG [Mean Cell Duration] Figure 3. Scaling relationship of (a) central moments and (b) rain cell characteristics for daily areal rainfall of Fergus catchment. Rainfall depth is a highly fluctuating process. Most hydrological applications like DDF use the aggregated (integrated) values of this fluctuating process. The scale of fluctuation gives the level of aggregation required to obtain stable (low variance) estimates of the mean of the fluctuating rainfall depth process (Vanmarcke, 1983). For the MBLRPM, the scale of fluctuation θ can be given by: 2ν( μc + 1) θ = (4) κ ( α 2) 2 + φ + 1 If we assume a linear transformation between the current, (.) o and the climate-induced, (.) e rainfall depth process, the following relationship would be satisfied as shown by Vanmarcke (1983). 2 2 σ0θ0 σ e θe = (5) 2 2 m0 me Combining equations 2 and 4, the changing factor of the scale of fluctuation can be expressed as: 2 (1 R R ψ (6) θ = ( ) ) m 7

When Redriguez-Iturbe et al. (1988) introduced the concept of randomness in cell duration parameter (η), the mean number of cells per storm, μ C = 1+κ/φ, is kept constant without losing generality. Similarly, we can assume that μ C will be the same in both the control and climate-induced systems. Then, it is easy to show that the changing factors of κ and φ are equal. R κ = R φ (7) The purpose of two of the six parameters of the MBLRPM (α and ν) is to randomly vary the gamma distributed cell duration parameter (η) from storm to storm. So, if we assume that the shape parameter α remains unchanged during climate enhancement, the cell duration will still vary due to the change in scale parameter ν. Moreover, the mathematical expression will be simplified to extractable level. Therefore, R α = 1.0 (8) Different empirical studies and physical interpretation of the rainfall process confirms that average storm intensity is inversely related to its duration of occurrence. Koutsoyiannis and Foufoula-Georgiou (1993) suggested power relationship so that both the inverse relation and the scaling property could be exploited. Consequently, we are hereby assuming the same scaling relationship could occur between mean rain cell intensity, μ x and mean rain cell duration, E[η -1 ] = ν/(α-1) in the MBLRPM structure. Figure 3b empirically verifies this assumption and yields a power relationship of: δ ν μ X = d, where d = 3.1387 and δ = -0.7829. (9) α 1 From equations (8) and (9), the changing factor relationship between μ X and ν can be written as: ( ) δ μ = R ν R X Now we had changing factors for three second-order properties, mean, variance and scale of fluctuation, and for three theoretically sound assumptions that gave equations (7), (8) and (10). Then, after algebraic manipulation and application of simple optimization technique, the changing factors by which the six model parameters should be perturbed to represent the precipitation scenario into the stochastic rainfall process could easily be obtained. (10) 4.2. Seasonal profile of precipitation scenarios As the main purpose of this study is to figure out the amount by which flood magnitudes at different frequencies of occurrence will change when the rainfall changed by certain factor, it becomes necessary to set a reference season and monthly profile of precipitation changes. Almost all annual maximum floods in this country were observed in winter season (DJF). Therefore, percentage change of winter precipitation during climate change is hereafter taken as precipitation scenario. Since rainfall modeling is taking place in monthly basis, we have to devise a means to distribute this precipitation scenario to each month. The monthly and seasonal percentage changes of precipitation from ECHAM4/OPYC3 GCM predictions of greenhouse gas and control integrations for different climatic times are computed for the nearby grid point as shown in Figure 4. Then, a dimensionless monthly precipitation scenario profile, whose ordinate values are listed in Table 4, is obtained from rescaling the mean monthly changing factors of precipitation by the mean winter changing factor. And this dimensionless profile is used to distribute the above defined precipitation scenario to the respective months in the continuous simulation experiment. Based on the precipitation changes observed from the GCM experiment shown in Figure 4, precipitation scenarios that include all possible future ranges are identified as 20%R, -15%R, - 10%R, 0%R, +10%R, +15%R, +20%R and +25%R, and applied in the foregoing experiment. Hereinafter the 0%R precipitation scenario is known as reference precipitation scenario. 8

Table 4. Dimensionless monthly precipitation scenario profile derived from ECHAM4/OPYC3 GCM data. Month Profile Ordinate Month Profile Ordinate Jan 1.01232 Jul 0.76435 Feb 1.00469 Aug 0.75891 Mar 1.04580 Sep 0.89647 Apr 0.79504 Oct 0.95380 May 0.85542 Nov 1.02349 June 0.83505 Dec 0.98428 60 40 1950-1979 1980-2009 2010-2039 2040-2069 2070-2099 Mean Percentage Change 20 0-20 -40-60 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Figure 4. Monthly percentage changes of precipitation from greenhouse gas integration to control integration of the ECHAM4/OPYC3 GCM experiment at (8.4375W, 51.4162N) grid point. 5. Organization and Application of Continuous Simulation Scheme The time series generator part of the MBLRPM and ARMA (1,1), the simulator component of the SMARG model, the precipitation downscaling routine, and flood frequency processor are combined together to form a continuous simulation scheme. The inputs to the scheme are the calibrated parameters of the three models, the dimensionless monthly precipitation scenario profile and the precipitation scenario as percentage change of winter rainfall. The number of simulations (realizations), the length of the time series and the starting year of simulation are off course part of the input. The scheme was tested and verified by its ability to reproduce the observed annual maximum annual flow series from 1000 simulations of 25 years each for the current climate. The result displayed in Figure 5 confirms the dependability of the scheme for the proposed climate change impact assessment study. 9

75 Ann. Max. Flow in cumecs 60 45 30 15 0 Hist. Sim_2.5%L Sim_50%M Sim_97.5%U -2-1 0 1 2 3 4 5 EV1 Reduced Variate Figure 5. Comparison of the historical and the continuous simulation scheme results of annual maximum flow series. Dashed lines show the 95% confidence intervals of the simulation results. For each of the above-hypothesized precipitation scenarios, 1000 simulations of 100 years continuous flow series are generated and the corresponding annual maximum flow (AMF) series are selected and analyzed nonparametrically. The percentage changes of the annual maximum flow series of a given precipitation scenario form the respective series of the reference precipitation scenario are computed for all of the above scenarios. These percentage changes can be interpreted as AMF scenarios at the specified frequency of occurrence. Further normalizing the AMF scenarios by the respective precipitation scenarios gives dimensionless factors that quantify the impact of climate change on flood flows in terms of precipitation changes. These factors for different frequency of occurrences produced growth curves, as shown in Figure 6, which can be applied as a climate change correction to flood frequencies estimated from the current climatic data by any conventional method. 6. Discussion of Results and Conclusions The continuous simulation scheme assembled for this study requires higher level of computing power and takes appreciable time to get the result. Therefore, it is not feasible to adopt such simulation schemes for practical design flood estimation works. Since precipitation scenarios for a certain region can be estimated from GCMs output, the developed climate change correction growth curves can be used with conventional flood frequency analysis methods for design flood estimation. Hence engineers and technicians engaged in the design and operation of storm sewers, culverts, rain water harvesting reservoirs, weirs and other water resources projects can apply climate change corrections to design floods estimated from monographs, charts and conventional methods using such curves developed for their region. For example, if the precipitation scenario is 20%R and the return period considered is 20 years (EV1 reduced variate of 1.97), one can read 0.95 from Figure 6 and get 19% (=0.95 20%) as the amount by which the 20-year flood should be increased. The simulation result summarized in Figure 6 shows that there is no linear relationship between precipitation and annual maximum flood scenarios. It also reveals that low return period floods are more sensitive to climate change than high return period floods. This might be due to the degree of wetness of the cathment at different flood frequencies. If the catchment is wetter, its dynamics in less affected by changes in precipitation and produces higher floods. So, the derived curves exhibit a physically possible pattern. Although some assumptions and monthly mean changes in GCM precipitation outputs are employed in the downscaling process, this preliminary result of our on going study gives an indicative framework for incorporating climate change in flood frequency estimation procedures. We hope that most of the limitations of this work will be clarified at the final stage of the study. 10

Ann. Max. Flow % Change Winter Rainfall % Change 1.5 1.4 1.3 1.2 1.1 1.0 0.9-20%R -15%R -10%R +10%R +15%R +20%R +25%R Linear 0.8-2 -1 0 1 2 3 4 5 6 EV1 Reduced Variate Figure 6. Climate change correction growth curves for flood frequency estimation. Acknowledgments This research has been supported by funding from Environmental Change Institute of the National University of Ireland, Galway as part of the Higher Education Authority Programme for Research in Third Level Institutions (HEA PRTLI) Cycle 2. References Bierkens, M.F. and Puente, C.E., 1990. Analytically derived runoff models based on rainfall point processes. Water Resour. Res., 26(11): 2653-2659. Blazkova, S. and Beven, K., 1997. Flood frequency prediction for data limited catchments in the Czech Republic using a stochastic rainfall model and TOPMODEL. J. Hydrol., 195: 256-278. Brandsma, T. and Buishand, T.A., 1997. Statistical linkage of daily precipitation in Switzerland to atmospheric circulation and temperature. J. Hydrol., 198: 98-123. Burlando, P. and Rosso, R., 1991. Extreme storm rainfall and climate change. Atm. Res., 27: 169-189. Calver, A. and Lamb, R., 1996. Flood frequency estimation using continuous rainfall-runoff modeling. Phys. Chem. Earth, 20(5-6): 479-483. Cameron, D.J., Beven, K.J., Tawn, J. Blazkova, S. and Naden, P., 1999. Flood frequency estimation by continuous simulation for a gauged upland catchment (with uncertainity). J. Hydrol., 219: 169-187. Conway, D. and Jones, P.D., 1998. The use of weather types and airflow indices for GCM downscaling. J. Hydrol., 212-213: 348-361. Cunnane, C., 1988. Methods and merits of regional flood frequency analysis. J. Hydrol., 100: 269-290. Demissie, S.S., 1999. Assessment of land use and climate changes on flood flows. M.Sc. Thesis, Department of Engineering, National University of Ireland, Galway. Diaz-Granodos, M.A., Valdes, J.B. and Bras, R.L., 1984. A physically based flood frequency distribution. Water Resour. Res., 20(7): 995-1002. Eagleson, P.S., 1972. Dynamics of flood frequency. Water Resour. Res., 8(4): 878-898. Hashemi, A.M., Franchini, M. and O Connell, P.E., 2000. Climatic and basin factors affecting the flood frequency curve: PART I A simple sensitivity analysis based on the continuous simulation approach. Hydrol. Earth System Sci., 4(3): 463-498. Hebson, C. and Wood, E.F., 1982. A derived flood frequency distribution using Horton order ratios. Water Resour. Res., 18(5): 1509-1518. 11

Hosking, J.R.M. and Wallis, J.R., 1997. Regional flood frequency analysis: an approach based on L-moments. Cambridge University Press. Institute of Hydrology, 1999. Flood estimation handbook. Wallingford, UK. IPCC, 2001. Climate Change 2001: impacts, adaptation and vulnerability. Downloaded from www.ipcc.ch. Kamga, F.M., 2001. Impact of greenhouse induced climate change on runoff of the upper Benue river (Cameroon). J. Hydrol. 252: 145-156. Khaliq, M.N. and Cunnane, C., 1996. Modeling point rainfall occurrences with the Modified Bartlett-Lewis Rectangular Pulses model. J. Hydrol., 180: 109-138. Khaliq, M.N., 1995. Stochastic modeling of rainfall occurrences and its use for modeling effects of climatic changes on rainfall extreme values. Ph.D. Thesis, Department of Engineering hydrology, U.C.G., National University of Ireland. Koutsoyiannis, D. and Foufoula-Georgiou, E., 1993. A scaling model of a storm hyetograph. Water Resour. Res., 29(7): 2345-2361. Koutsoyiannis, D. and Foulfoula-Georgiou, E., 1993. A scaling model of a storm hyetograph. Water Resour. Res., 29(7): 2345-2361. Lamb, R., 1999. Calibration of a conceptual rainfall-runoff model for flood frequency estimation by continuous simulation. Water Resour. Res., 35(10): 3103-3114. Liang, G.C., 1992. A note on the revised SMAR model. Workshop memorandum, Department of Engineering Hydrology, U.C.G., National University of Ireland (unpublished). Loukas, A., Vasiliades, L. and Dalezios, N., 2002. Potetial climate change impacts on flood producing mechanisms in Southern British Columbia, Canada using the CGCMA1 simulation results. J. Hydrol., 259: 163-188. Mingkai, Q., 1996. Further development of the SMAR model An improvement to the description of the overland flow process. M.Sc. Thesis, Department of Engineering Hydrology, U.C.G., National University of Ireland. O Connell, P.E., Nash, J.E. and Farrell, J.P., 1970. River flow forecasting through conceptual models, Part 2, The Brosna catchment at Ferbane. J. Hydrol., 10(4): 317-329. Rodriguez-Iturbe, I., Cox, D.R. and Isham, V., 1987. Some models for rainfall based on stochastic point processes. Proc. R. Soc. London, A, 410: 269-288. Rodriguez-Iturbe, I., Cox, D.R. and Isham, V., 1988. A point process model for rainfall: further developments. Proc. R. Soc. London, A, 417: 283-298. Tan, B.Q. and O Connor, K.M., 1996. Application of an empirical infiltration equation in the SMAR conceptual model. J. Hydrol., 185: 275-295. Vanmarcke, E., 1983. Random fields: Analysis and Synthesis. MIT Press, Cambridge, Massachusetts. Wilby, R.L., Hassan, H. and Hanaki, K., 1998. Statistical downscaling of hydro-meteorological variables using general circulation model output. J. Hydrol., 205:1-19. Yu, P., Yang, T. and Wu, C., 2002. Impact of climate change on water resources in Southern Taiwan. J. Hydrol., 260: 161-175. 12

An investigation of the use of partial L-moments for analysing censored flood samples Keshav P. Bhattarai* Department of Engineering Hydrology, National University of Ireland, Galway (*keshav.bhattarai@nuigalway.ie) Abstract A method of partial L-moments for analysing censored flood samples has been derived and investigated using Monte Carlo simulation. The results obtained from a simulation study show that censoring the annual maximum series from below, at or about 30%, gives the minimum value of bias without appreciably increasing the root mean square error. Comparison of efficiencies of the method of partial L-moments with the method of LH-moment shows that the standard error and the mean square error from both methods are similar and, in some cases, the former method gives a smaller value of bias than the latter. To investigate the use of the method of partial L- moments on real data, the series of annual maximum floods of 25 rivers in Nepal were fitted using this method as well as the methods of LH-moments and simple L-moments. The results show that, for the annual maximum floods of glacier fed rivers having a single segment on the probability plot, the method of simple L-moment is adequate. But for the annual maximum floods having two or more segments on the probability plot, the methods of partial L- moments and LH-moments are equally useful. It is further observed that, in general, the use of partial L-moments with F 0 =10% and L1-moments, partial L-moments with F 0 =20% and L2-moments and so on, give near-identical results. In some cases, the method of partial L-moments results in a better fit than the method of LH-moments. Résumé La méthode des L-moments partiels utilisée pour l analyse d échantillons de crue a été étudiée et définie à partir de la simulation «Monte Carlo». Les résultats obtenus par l étude d'une simulation montrent qu`en tronquant la chronique des maxima annuels, par suppression d à peu près 30% des données correspondant aux valeurs les plus faibles, nous obtenons un biais faible, sans augmentation significative de l`erreur quadratique moyenne. En comparant l efficacité entre la méthode des L-moments partiels et la méthode des LH-moments, nous voyons que, pour les deux méthodes, l erreur standard et l erreur quadratique moyenne sont les mêmes, et, dans certains cas, la première méthode donne une plus petite valeur de biais que la deuxième. Pour étudier l utilisation de la méthode des L-moments partiels sur des données réelles, cette méthode, ainsi que celle des LH-moments et des moments simples, ont été utilisées pour les chroniques des débits maximums annuels de 25 rivières du Népal. Les résultats montrent que, pour les débits maximum annuels des rivières alimentées par un glacier et caractérisées par un seul segment sur le graphique d ajustement à la loi de Gumbel, la méthode des L-moments simples convient. Par contre, pour les débits maximums annuels, avec deux ou plusieurs segments sur le graphique d ajustement à la loi de Gumbel, les méthodes des L-moments partiels et des LHmoments sont toutes deux pertinentes. Il a été en outre observé que, en général, l utilisation des L-moments partiels avec F 0 =10% et avec les L 1 -moments, l`utilisation des L-moments partiels avec F 0 =20% et avec les L 2 - moments, et ainsi de suite, donne des résultats à peu près identiques. Dans certains cas, la méthode des L-moments partiels aboutit à un meilleur résultat que la méthode des LH-moments. The Third Inter-Celtic Colloquium 2002 Page 1 of 10 An investigation of the use of partial L-moments..

1. Introduction In some water-quality investigations, a substantial portion of reported values of many contaminants lie below the limits of detection. Likewise, in the field of hydrology, low-flow and sometimes annual maximum flow observations are rounded down to, or reported as, zero. Such river discharges may have actually been zero or they may have been between zero and the measurement threshold, yet are reported as zero. Such data sets are called censored samples because all values of observations in a complete sample that fell below some threshold level were removed, or censored (Stedinger et al., 1993). Censored data are categorized as either type I censoring, where the measurement threshold is fixed and the number of censored data points varies, or type II censoring, where the number of censored data points is fixed and the implicit threshold varies (David, 1981). With type I censoring, the number of values censored is a random variable whereas in type II censoring, the censoring threshold is a random variable (Kroll and Stedinger, 1996). Values for data below the measurement threshold are generally reported as less than the detection limit, and data sets containing such points are referred to as left censored data (Koulouris et al., 1998). Left-censored environmental data typically follow type I censoring because the censoring threshold is fixed by the measurement technology. The purpose of analysing annual maximum series of floods is, in most cases, to predict the magnitude of flood of relatively large return periods. When analysing floods of arid or semiarid regions, many very low (or zero) annual maximum flows occurs. Cunnane (1987) suggested that, in such cases, it is better to analyze censored sample based only on those floods whose magnitudes have exceed a certain threshold (x 0 ). Wang (1990a, b, 1996a) suggested that sometimes it is actually advantageous to intentionally censor (or eliminate) observations in order to better understand the frequency and magnitude of floods of larger return periods. The process of censoring, from below, a fixed numbers of sample points from the annual maximum series is classified as type II left-censoring. Right-censored data sets can arise in hydrology when flood observations are reported as having occurred above some threshold (Stedinger and Cohn, 1986). The present study concentrates on type II left-censoring. Several approaches are available for the analysis of censored data, including probability plots and probability plot regression, weighted-moment estimators, maximum likelihood estimators, conditional probability models, and partial probability weighted moments. Methods for the maximum likelihood estimate of a generalized extreme value (GEV) distribution from censored samples have been suggested by Prescott and Walden (1983) and Phien and Fang (1989). Wang (1990a, 1990b, 1996a) introduced the concept of partial probability weighted moments (PPWMs) for the analysis of censored samples and a derived unified expression for the GEV distribution. Kroll and Stedinger (1996) used Wang s method of PPWMs along with the log-probability plot regression method (LPPR) and the lognormal maximum likelihood method (MLE) for the analysis of censored samples using simulation techniques and suggested that, regardless of the underlying distribution, the MLE generally performed better than the others. Hosking (1990) introduced the concept of L-moments as an extension of probability weighted moments and the same author (1995) extended the theory of L-moments to the analysis of censored data and derived equations for estimating a number of distributions from upper bound censored samples. Koulouris et al. (1998) introduced L-moment diagrams for the evaluation of the goodness of fit of alternative distributional hypotheses for left-censored data from partial probability-weighted moments. The purpose of the method of higher probability weighted moments (HPWMs) (introduced by Wang, 1997a) and LH-moments (introduced by Wang, 1997b) is somewhat similar to the analysis of censored annual maximum flood series, that is, to reduce undesirable influence that small sample events may have on the estimation of large return periods events. In fact, both the methods of higher probability weighted moments and LH-moments give the same quantile estimates; as in the case of the method of simple L-moments and unbiased probability weighted moments, which also give the same quantile estimates (Bhattarai, 1997). Wang (1998) formulated a procedure for goodness-of-fit tests for the use of the GEV distribution in conjunction with the LH-moment estimation and suggested that the formulated tests should alleviate the need for finding the correct distribution. The purpose of the present study is to investigate the method of partial L-moments to analyse the annual maximum series censored from below (type II left censoring) at the discretion of the analyst. In this method, the sample partial L-moments are estimated directly from uncensored data using Wang s (1996 b) method of direct sample estimates of L-moments but with some modifications. The efficiency of partial L-moments was tested by the Monte Carlo simulation method. Though the method of partial L-moments for the analysis of censored samples and the method of LH-moments are different, their objective is somewhat similar, i.e., to reduce the undesirable influence of small sample events on the estimation of large return period floods. Therefore, the efficiency of the method of partial L-moments was compared with that of the method of LH-moments. To investigate the use of the method of partial L-moments on real data, the series of annual maximum floods of 25 rivers in Nepal, with catchment area ranging from 17 to 42890 km 2 and data length ranging from 20 years to 32 years, were fitted using the method of partial L-moments as well as using the methods of LH-moments and simple L-moments. The fitted curves are then compared on probability plots. The Third Inter-Celtic Colloquium 2002 Page 2 of 10 An investigation of the use of partial L-moments..

2. Parameter estimation of the Generalized Extreme Value distribution using partial L-moment In the present study, the necessary equations for the estimation of the parameters of GEV distribution using sample partial L-moments have been derived from the concept of the method of partial probability weighted moments introduced by Wang (1990a, b, 1996a). The derivation of such equations is presented in Appendix A. The partial L-moments can be calculated directly from the complete sample arranged in ascending order x (1) x (2). x (n) and using Wang s (1996b) method of direct sample estimation of L-moments, with some modification, as: n ' 1 * λ = (1) where, and 1 x n ( i ) C1 i= 1 n ' 1 1 i 1 n i * 2 = ( C C x n 1 1) ( i) 2 C2 i= 1 n ' 1 1 i 1 i 1 n i n i 3 = ( C C C + C n 2 2 1 1 2 ) 3 C3 i= 1 λ (2) λ x (3) * x 0 for x (i) x 0 (4) ( i) = * ( i) x( i) x = for x (i) x 0 (5) m C k m m! = = (6) k k!( m k)! In mnemonic form, we can write the j th partial L-moment as 1 1 n ' j 1 * λ j = ( A B) x n ( i) (7) j C j i= 1 where, in the binomial expansion of (A-B) j-1, A k = i-1 C k and B k = n-i C k, for k = 0, 1, 2,., j and j = 1, 2, 3. In the present study, the level of censoring, F 0, is selected a priori, from which the number of the sample data points to be censored can be determined as: n 0 = n * F 0 (8) where, n is the length of the uncensored sample and n 0 is the number of censored data points The value of x 0 is determined by counting the n 0 th data point from below. The sample partial L-moments estimated above are utilized to estimate the parameters of the GEV distribution using equations (A.9), (A.10), (A.17), (A.18) and (A.19). * ( i) 3. Results and Discussion Monte Carlo simulations have been carried out to investigate the efficiency of the method of partial L-moments. Synthetic data of length n = 30 were generated for two shape parameter values, k = +0.2 and k = -0.2, with location parameter ζ = 0.0 and scale parameter α = 1.0. Different levels of censoring threshold, viz. F 0 = 0%, 10%, 20%, 30%, 40%, 50% and 60% are considered. When the level of censoring F 0 = 0%, the method of partial L-moments is equivalent to the method of simple L-moment. The number of replications (M) used in the simulation study for each case is 10,000. Bias, standard error and root mean square (in percentage of the true quantiles) for 4 quantile estimators x(f=0.98), x(f=0.99), x(f=0.995) and x(f=0.998), equivalent to 50, 100, 200 and 500 year return periods respectively, were considered. A corresponding simulation study has also been carried out for the method of LHmoments so as to compare and judge the efficiency of the method of partial L-moments with that of LH-moments. The outcome of the simulation study is presented in Tables 1-2 and in Figures 1-3. The Third Inter-Celtic Colloquium 2002 Page 3 of 10 An investigation of the use of partial L-moments..

Table 1 Bias, standard error and root mean square error (in %) in the quantile estimates using the methods of partial L-moments and LH-moments for fitting the GEV distribution for n = 30, ζ = 0.0, α = 1.0 and M = 1000 (for k = -0.2) Level of T = 50 yr., x(f=0.98) T = 100 yr., x(f=0.99) T = 200yr., x(f=0.995) T = 500 yr.,x(f=0.998) Censoring Bias Se RMSE Bias Se RMSE Bias Se RMSE Bias Se RMSE i. Method of partial L-moment F 0 =0% -1.374 34.89 34.92-0.044 42.3 42.3 1.967 51.42 51.46 5.832 66.68 66.93 F 0 =10% -1.534 34.82 34.85 0.027 42.72 42.72 2.426 52.68 52.74 7.11 69.68 70.05 F 0 =20% -2.103 34.69 34.75-0.755 72.78 42.78 1.722 53.17 53.2 6.617 71.21 71.52 F 0 =30% -3.08 34.81 34.95-2.409 42.88 42.95-0.695 53.38 53.39 3.434 71.75 71.83 F 0 =40% -4.28 35.31 35.57-5.153 43.18 43.49-4.92 53.46 53.69-2.683 71.46 71.51 F 0 =50% -5.315 36.44 36.82-8.553 43.95 44.77-10.57 53.6 54.63-11.22 70.29 71.18 F 0 =60% -5.926 38.2 38.65-12.61 44.85 46.59-17.74 53.22 56.1-22.21 67.63 71.18 ii. Method of LH-moments L1-mom -2.539 34.67 34.76-1.383 42.62 42.65 0.676 52.82 52.82 5.021 70.52 70.7 L2-mom -3.422 34.4 34.57-2.51 42.44 42.51-0.511 53.08 53.09 4.091 72.11 72.22 L3-mom -4.177 34.24 34.49-3.485 42.15 42.29-1.613 52.94 52.97 3.103 72.69 72.75 L4-mom -4.684 34.19 34.51-4.352 41.86 42.09-2.647 52.63 52.7 2.095 72.76 72.79 Table 2 Bias, standard error and root mean square error (in %) in the quantile estimates using the methods of partial L-moments and LH-moments for fitting the GEV distribution for n = 30, ζ = 0.0, α = 1.0 and M = 1000 (for k =+0.2) Level of T = 50 yr., x(f=0.98) T = 100 yr., x(f=0.99) T = 200yr., x(f=0.995) T = 500 yr.,x(f=0.998) Censoring Bias Se RMSE Bias Se RMSE Bias Se RMSE Bias Se RMSE i. Method of partial L-moment F 0 =0% 0.498 19.17 19.18 1.33 21.96 22 2.327 25.1 25.2 3.871 29.64 29.89 F 0 =10% 1.178 19.24 19.3 2.31 22.49 22.62 3.812 26.3 26.57 6.036 31.97 32.54 F 0 =20% 0.971 19.1 19.13 2.248 22.65 22.76 3.833 26.94 27.21 6.387 33.53 34.14 F 0 =30% -0.668 18.85 18.86-0.176 22.34 22.34 0.704 26.67 26.68 2.435 33.37 33.46 F 0 =40% -3.492 18.51 18.84-4.751 21.58 22.11-5.492 25.34 25.93-5.679 31.11 31.64 F 0 =50% -7.032 18.16 19.47-10.76 20.36 23.02-13.66 22.95 26.7-16.46 26.78 31.44 F 0 =60% -9.96 17.48 20.11-16.22 18.19 24.37-21.09 19.04 28.41-25.95 20.29 32.94 ii. Method of LH-moments L1-mom 0.126 18.68 18.68 0.961 21.53 21.55 2.021 24.89 24.97 3.767 29.87 30.11 L2-mom -0.188 18.48 18.49 0.702 21.52 21.53 1.904 25.25 25.33 3.942 31.03 31.28 L3-mom -0.484 18.36 18.36 0.446 21.52 21.52 1.788 25.6 25.67 4.155 32.15 32.42 L4-mom -0.772 18.26 18.28 0.18 21.5 21.5 1.647 25.88 25.93 4.344 33.2 33.48 The Third Inter-Celtic Colloquium 2002 Page 4 of 10 An investigation of the use of partial L-moments..

5 0 (a) 10 5 (b) 0-5 Bias (%) -10 x(f=0.98) x(f=0.99) Bias (%) -5-10 x(f=0.98) x(f=0.99) -15 x(f=0.995) -15 x(f=0.995) -20-20 0 10 20 30 40 50 60 Level of censoring (%) -25 0 10 20 30 40 50 60 Level of censoring (%) Figure 1 Bias (%) in the quantile estimators x(f) at various levels of censoring for n = 30, ζ = 0.0, α = 1.0 and M = 1000 (a) for k = -0.2 and (b) for k = +0.2 65 60 x(f=0.98) x(f=0.99) x(f=0.995) (a) 29 (b) x(f=0.98) x(f=0.99) x(f=0.995) 55 27 50 25 45 23 40 21 35 19 30 0 10 20 30 40 50 60 Level of censoring (%) 17 0 10 20 30 40 50 Level of censoring (%) 60 Figure 2 Root mean square error (%) in the quantile estimators x(f) at various levels of censoring for n = 30, ζ = 0.0, α = 1.0 and M = 1000 (a) for k = -0.2 and (b) for k = +0.2 8 5 6 4 F0=0% F0=30% L1-mom L2-mom L3-mom L4-mom 4 3 F0=0% F0=30% L1-mom L2-mom L3-mom L4-mom 2 2 0 1-2 0-4 (a) -1 (b) -6 3.5 4.0 4.5 5.0 5.5 6.0 6.5 EV1Reduced Variate, y T -2 3.5 4.0 4.5 5.0 5.5 6.0 6.5 EV1 Reduced Variate, y T Figure 3 Bias(%) in the quantile estimation of various return periods, plotted against the EV1 reduced variate, using the methods of partial L-moments (F0=30%) and LH-moments for n = 30, ζ = 0.0, α = 1.0 and M = 1000 (a) for k = -0.2 and (b) for k = +0.2 The Third Inter-Celtic Colloquium 2002 Page 5 of 10 An investigation of the use of partial L-moments..

From Figures 1a and 1b, it is observed that the bias will be minimum at about F 0 = 30% level of censoring for both values of k considered in this study, that is, +0.2 and -0.2. Similarly, from Figures 2a and 2b it can be seen that the root mean square error is more or less constant up to F 0 = 30% level of censoring, after which it shows increasing trend, especially for the positive value of the shape parameter, k. Similar results are obtained in the case of standard error also, and hence are not shown here. This means that, censoring from below at or about 30% will not appreciably increase the standard error and root mean square error of the quantile estimates and gives the minimum bias. The efficiency of the method of partial L-moments was compared with that of the method of LH- partial L-moments and LH-moments are quite similar, but the bias is different. Figure 3a shows that, for k = -0.2, the bias from the partial L-moment at the censoring level of F 0 = 30% lies between those from the L2-moments and L3-moments. From Figure 3b, it is observed that for k = +0.2, the bias is generally less from the method of partial moments. From Table 1, it is observed that the standard error and root mean square error from the method of L-moment (at F 0 = 30%) than from all orders of LH-moments. Annual maximum flood data from 25 rivers of Nepal were fitted using the methods of simple L-moments, partial L-moments and LH-moments. From the fitted plots, it is observed that, for the snow-fed perennial rivers showing only one segment on probability plots, all the methods i.e., simple L-moment, LH-moment and partial L- moments give virtually the same curve. Therefore, in such cases, application of simple L-moments seems to be adequate. However, for the rain fed rivers for which the annual maximum series shows two or more segments on a probability plot, use of both partial L-moments and LH-moments give a better fit than simple L-moments. In such cases, it was generally observed that partial L-moments with F 0 = 10% and L1-moments, with F 0 = 20% and L2- moments and so on, give equivalent results (see Figures 4a - d). In some cases, the use of partial L-moments, with censoring level of F 0 = 30%, resulted in the best fit (see Figure 4d). 7000 80 Q, cumecs 6000 5000 4000 3000 (a) Qobs L-mom Q, cumecs 70 60 50 40 30 Qobs L-mom L3-mom PL30% (b) 2000 L3-mom 20 1000 PL30% 10 0-2 -1 0 1 2 3 4 5 EV1 Reduced Variate, y T 0-2 -1 0 1 2 3 4 5 EV1 Reduced Variate, y T 1600 12000 Q, cumecs 1400 1200 1000 800 600 400 (c) Qobs L-mom L2-mom PL20% Q, cumecs 10000 8000 6000 4000 Qobs L-mom L3-mom PL30% (d) 200 2000 0-2 -1 0 1 2 3 4 5 EV1 Reduced Variate, y T 0-2 -1 0 1 2 3 4 5 EV1 reduced variate, y T Figure 4 Fitting the GEV distribution using the methods of simple L-moments, partial L-moments and LH- to the annual maximum flood series of different rivers in Nepal (a) Bheri River, Stn. No. 280, n=32, catchment area = 12290 km 2 (b) Bagmati River, Stn. No. 505, n = 31, catchment area = 17 km 2 (c) East Rapti River, Stn No. 460, n = 26, catchment area = 579 km2 (d) Babai River, Stn. No. 290, n = 20, catchment area = moments 3000 km 2 The Third Inter-Celtic Colloquium 2002 Page 6 of 10 An investigation of the use of partial L-moments..

4. Summary and conclusion The method of partial L-moments was derived from the concept of probability weighted moments introduced by Wang (1990a, b, 1996a) and was investigated using Monte Carlo simulation. Results show that censoring from below, at or about 30% level, gives minimum bias without appreciably increasing the standard error and root mean square error. For the analysis of annual maximum floods, which show two or more segments in the probability plots, the results of the method of partial L-moments are comparable with those of the method of LH-moments. If the lower segments are of much smaller values than the upper segments (such as for annual maximum floods from arid or semi-arid regions), the method of partial L-moments might be the most suitable option, as was seen in some cases of the Nepalese rivers considered in this study. It is also observed that the method of partial L-moments with the censoring level of F 0 = 10%, 20%, 30% and 40% give quite similar results to those from using the method of LH-moments of 1 st, 2 nd, 3 rd and 4 th order (i.e., L1-moments, L2-moments, L3-moments and L4-moments) respectively. The method of simple L-moments, derived directly from the sample, is currently used extensively in the field of hydrology. As the method of simple L-moments and simple probability weighted moments are found to give the same quantile estimates, the method of simple probability weighted moments are now rarely used. Similarly, the method of LH-moments and the method of higher probability weighted moments (HPWMs) are found to give the same quantile estimates. Therefore, with the use of the method of LH-moments, the method of HPWMs is also no longer necessary. The method of partial L-moments, with the sample partial L-moments calculated directly from the annual maximum flow data, investigated here, also gives the quantile estimates similar to those from the method of partial probability weighted moments. Therefore, this method can be used in place of the method of partial probability weighted moments for the analysis of censored samples. With this consideration, the methods of simple L-moments, LH-moments and partial L-moments can completely replace the corresponding methods of simple probability weighted moments, higher probability weighted moments and partial probability weighted moments respectively, thus effectively eliminating the need to use the method of probability weighted moments. 5. Acknowledgement The author gratefully acknowledges Prof. C. Cunnane for his valuable suggestions and Prof. K. M. O Connor for his valuable comments and general support. The author also gratefully acknowledges the Environmental Change Institute (ECI) of the National University of Ireland Galway, and the Higher Education Authority of Ireland for providing postgraduate fellowship, which facilitated this study. 6. References Bhattarai, K.P. 1997. Use of L-moments in flood frequency analysis. M.Sc thesis, National University of Ireland, Galway. Cunnane, C. 1987. Review of Statistical Models for flood frequency estimation. In Hydrologic Frequency Modelling, edited by V. P. Sing, published by D. Reidel Publ. Co., Dordrecht, Holland, PP. 49-95. David, H.A., Order Statistics, John Wiley, New York, 1981. Hosking, J.R.M. 1990. L-moments: Analysis and estimation of distributions using linear combinations of order statistics. Journal of Royal statistical Society, Series B, 52, 105-124. Hosking, J.R.M. 1995. The use of L-moments in the analysis of censored data. In Recent Advances in Life-Testing and Reliability, edited by N. Balakrishnan, pp. 544-564, CRS Press, Boca Raton, Fla. Koulouris, A.Z., Vogel R.M., Craig S.M. and Habermeier J. 1998. L-moment dieagrems for censored observations. Water Resour. Res., 34 (5), 1241-1249. Kroll, C.N. and Stedinger, J.R. 1996. Estimation of moments and quantiles using censored data. Water Resour. Res., 32 (4), 1005-1012. Phien, H.N. and Fang, T.S. E. 1989. Maximum likelihood estimation of the parameters and quantiles of the generalized extreme-value distribution from censored samples. J. Hydrology, 105, 139-155. Prescott, P. and Walden A.T. 1983. Maximum likelihood estimation of the parameters of the three-parameter generalized extreme-value distribution from censored samples, Journal of statistical computation and simulation, 16, 241-250. Stedinger, J.R., and Cohn, T.A. 1986. Flood frequency analysis with historical and paleoflood information, Water Resour. Res., 22 (5), 785-793, 1986. Stedinger, J.R., Vogel R. M. and Foufoula-Geourgiou E. 1993. Frequency analysis of extreme events, in Handbook of Hydrology, edited by D. A. Maidment, chap. 18, McGraw-Hill, New York. The Third Inter-Celtic Colloquium 2002 Page 7 of 10 An investigation of the use of partial L-moments..

Wang, Q.J. 1990a. Estimation of GEV distribution from censored samples by method of partial probability weighted moments, J. Hydrology, 120, 103-114. Wang, Q.J. 1990b. Unbiased estimation of probability weighted moments and partial probability weighted moments from systematic and historical flood information and their application to estimating the GEV distribution, J. Hydrology, 120, 115-124. Wang, Q.J. 1996a. Using partial probability weighted moments to fit the extreme value distributions to censored samples, Water Resour. Res., 32 (6), 1767-1771. Wang, Q.J. 1996b. Direct sample estimators of L-moments, Water Resour. Res., 32 (12), 3617-3619. Wang, Q.J. 1997a. Using higher probability weighted moments for flood frequency analysis. Journal of hydrology, 194, pp. 95-106. Wang, Q.J. 1997b. LH-moments for statistical analysis of extreme events, Water Resour. Res., 33 (12), 2841-2848. Wang, Q.J. 1998. Approximate goodness-of-fit tests of fitted generalized extreme value distributions using LHmoments, Water Resour. Res., 34 (12), 3497-3502. 7. Appendix The generalized extreme value (GEV) distribution combines into a single form the three possible types of limiting distribution for extreme value. The distribution function is k k F ( x) = exp exp 1 ( x ξ ) α for k 0 (A.1) F ( ) = exp 1 exp 1 ( ξ ) α x for k = 0 (A.2) The GEV distribution are classified into 3 types, viz. EV1 (Gumbel), EV2 and EV3 corresponding to k = 0, k < 0 and k > 0 respectively. The inverse form of the GEV distribution is given by k x ( F) = ξ + α{ 1 ( log F) }/ k for k 0 (A.3) x( F) = ξ + α log( log F) for k = 0 (A.4) where ζ, α and k are location, scale and shape parameters of the GEV distribution respectively. The partial probability moments (PPWMs) of the GEV distribution for k 0 are given by Wang (1990a, 1996a) as: r 1 Γ(1 + k) r ( 1 F0 ) α ' α 1 β = ξ + + P(1 + k, ( r + 1)log F0 ) 1+ k k r + 1 k ( r + 1) (A.5) where P(..,..) is an Incomplete Gamma function and F 0 = F(x 0 ), x 0 being the censoring threshold. Using r = 0, 1 and 2 in equation A.5, we can write, ' α α β 0 = ξ + ( 1 F0 ) Γ(1 + k) P(1 + k, log F0 ) (A.6) k k ' 2β1 1 2 F 0 ' β0 1 F 0 α P(1 + k, 2log F = k 2 k 2 (1 F0 ) 0 ) P(1 + k, log F ) 0 1 F0 (A.7) The Third Inter-Celtic Colloquium 2002 Page 8 of 10 An investigation of the use of partial L-moments..

2 1 1 P(1 + k, 2log F ' ' β1 β0 0 ) P(1 + k, log F0 ) 2 k 2 F0 1 F0 2 (1 F0 ) 1 F0 = ' ' 3β 3log F 2 β P(1 + k, 0 0) P(1 + k, log F0 ) 3 k 3 F 1 F 3 (1 F0 ) 1 F 0 0 0 (A.8) ' β ' b r When F0 is known, we can replace r by, the sample estimate of the PPWMs and estimate parameters ζ, α and k as solutions of equations (A.6), (A.7) and (A.8). The exact solution of equation (A.8) requires iterative methods which are cumbersome. Wang (1990a) has also proposed a simple method of solution using the following equation: Let z be the right-hand side of equation A.8, that is, z = P(1 + k, 2log F0 ) P(1 + k, log F0 ) k 2 2 (1 F0 ) 1 F0 P(1 + k, 3log F0 ) P(1 + k, log F0 ) k 3 (1 F ) 1 F 3 0 0 (A.9) When z is plotted versus k, for a fixed F 0, the curve is very smooth. The exact solution of the curve changes with the F 0 value. The curve can be accurately approximated by a quadratic function of the form, k = a 0 + a 1 z + a 2 z 2 (A.10) For fixed F 0, three z values can be calculated by equation (A.9), corresponding to three chosen k values, e.g. k = - 0.5, - 0.1 and + 0.5, avoiding the use of k = 0 as the limiting form of equation (A.9). Substituting these three values into equation (A.10), yield a set of linear equations. We can then find the solutions for a 0, a 1 and a 2 corresponding to that fixed value of F 0. Combining equations (A.8) and (A.9), we get, z = ' 2β β0 1 F F 3 β 1 F F ' 1 2 0 1 ' 2 3 0 1 0 ' 0 0 (A.11) The relationship between L-moments and PWMs is given by Hosking (1990) as: λ 1 = β 0 λ 2 = 2β 1 β 0 λ 3 = 6β 2-6 β 1 + β 0 λ 4 =20 β 3-30 β 2 +12 β 1 β 0 (A.12) (A.13) (A.14) (A.15) where β 0, β 1, β 2, β 3 are the probability weighted moments and λ 1, λ 2, λ 3 and λ 4 are the first, second, third and fourth L-moments respectively. Similar linear relation can be established between the partial probability moments and partial L-moments. ' ' ' Replacing the values of the partial probability weighted moments β, β and β in equation (A.11) by the partial L-moments z ' λ 0, ' λ1 and λ + λ λ 0 1 2 ' λ 2 using the relationship in equations (A.12, A.13, A.14 and A.15), we get, ' ' ' 2 1 1 2 1 F0 1 F0 = ( 1 ' ' ' ( λ3 + 3λ2 + 2λ1 ) ' 2 λ1 3 1 F0 1 F0 A.16) The Third Inter-Celtic Colloquium 2002 Page 9 of 10 An investigation of the use of partial L-moments..

The Third Inter-Celtic Colloquium 2002 Page 10 of 10 An investigation of the use of partial L-moments.. In equation (A.16), replacing z by its sample estimate and by its sample estimate, we can write, ẑ ' λ ˆλ ' 0 ' 1 3 1 1 2 F F 0 ' 1 ' 2 ' 3 0 2 0 ' 1 ' 2 ˆ ) ˆ 2 3 ˆ ˆ ( 1 1 1 ˆ ˆ ˆ F F z + + + = λ λ λ λ λ λ (A.17) ters can then ' ˆλ 1 Substituting equa parame equations A.6 and A.7 as tion A.17 in equation A.10, we can find the estimate for kˆ. The other two by estimated successively from 0 0 2 0 ˆ 0 1 ) log ˆ, (1 ) (1 2 ) 2log ˆ, (1 ) (1 F F k P F F k P k + + + Γ 0 ' 1 2 0 ' 1 ' 2 1 ˆ 1 ˆ ˆ ˆ ˆ ˆ F F k + = λ λ λ α k (A.18) 0 + + Γ + = 1 1 ) log ˆ, (1 ˆ) (1 ˆ ˆ 1 ˆ ˆ 0 0 ' 1 F F k P k k F α λ ξ (A.19) where and are the first, second and third partial L-moments and can be derived by using equations (1-8) given in the body of the text., ˆ' 1 λ ' ˆλ 2 ' ˆλ 3

Using automatic water quality monitoring systems to monitor the transport of suspended sediment from an upland catchment in the west of Ireland. D. G. George 1, M. A. Rouen 1, B. O Hea 2, P. McGinnity 3 and N. Allott 4 1 Centre for Ecology and Hydrology, Windermere, Cumbria. 2. Marine Institute, Newport, County Mayo. 3. Central Fisheries Board, Glasnevin, Dublin. 4. Trinity College, Dublin. Abstract In recent years, the combined effect of extreme weather events and changes in land-use have had a significant effect on the erosion rates recorded in many upland areas. Here we describe how a network of automatic monitoring stations is being used to quantify the impact of these factors on the transport of sediment from an upland catchment in the west of Ireland. The Burrishoole catchment in County Mayo is surrounded by high mountains and includes a large freshwater lake (Lough Feeagh) that supports an important salmonid fishery. The automatic monitoring systems deployed in the catchment included river stations to measure the downstream transport of sediment and a lake station to record the spatial dispersion of the sediment in Lough Feeagh. The river stations were equipped with sensors for measuring water temperature, water level, ph, conductivity and suspended sediment concentration. The lake station was rather more complicated and included sensors that recorded wind speed and direction, air temperature, humidity, solar radiation, PAR, UV-B, water temperature, water level, ph, conductivity, oxygen, underwater light (PAR), the stability of the water column and the concentration of suspended sediment. All the monitoring stations were equipped with a sophisticated control system to minimise the risk of failure and a telemetry system for relaying the acquired data to a remote site. Some example results included in the paper demonstrate that these instruments were capable of producing very high-resolution records of the downstream transport of the sediment and its subsequent dispersal in Lough Feeagh. Crynodeb Yn ystod y blynyddoedd diweddar, y mae effaith newidiadau eithafol yn y tywydd ynghyd a dulliau newydd o drin y tir wedi cael effaith andwyol ar erydiad pridd mewn manau mynyddig. Yma disgrifiwn rhwydwaith o orsafoedd awtomatig a ddyfeisiwyd i fesur effaith y ffactorau hyn ar gludiad gwaddod o ddalgylch mynyddig yng ngorllewin Iwerddon. Y mae dalgylch Burrishoole yn swydd Mayo wedi ei amgylchu gan fynyddoedd uchel ac yn cynhwys Llyn o ddwr crai (Llyn Feeagh) sy n cynnal pysgodfa bwysig o frithyll ag eog. Yn y rhwydwaith hyn, defnyddiwyd gorsafoedd i fesur crynodiad y gwaddod yn y prif afonydd a gorsaf i fesur gwasgariad y gwaddod yn Llyn Feeagh. Yr oedd gan y gorsafoedd a osodwyd ar yr afonydd sensorau i fesur, tymheredd y dwr, lefel y dwr, ph, dargludedd a chrynodiad y gwaddod. Yr oedd yr orsaf a osodwyd ar Lyn Feeagh yn fwy cymhleth ac yn cynhwys sensorau i fesur cyfeiriad a chyflymder y gwynt, tymheredd yr awyr, pelydredd yr haul, pelydredd UV-B, tymheredd y dŵr, lefel y dŵr, ph, dargludedd, ocsigen, golau tanddwr (PAR), sefydlogrwydd y llyn a chrynodiad y gwaddod. Yr oedd gan y gorsafoedd i gyd systemau soffistigedig i leihau y perygl o fethiant a radio neu ffôn i drosglwyddo r gwybodaeth a gasglwyd i safle cyfleus. Dengys yr enghreifftiau a gynhwysir yn y papur fod y rhwydwaith yn medru cynhyrchu cofnod manwl o gludiad y gwaddod yn yr afonydd a gwasgariad y gwaddod yn Llyn Feeagh. 1

1 Introduction In recent years, the combined effects of extreme weather events (Arnell, et al., 1990) and changes in land-use (Maitland, et al., 1990) have resulted in a marked increase in the quantity of suspended sediment transported from many upland areas. The problem is particularly acute in peat catchments that are steep, wet and heavily grazed (Mulqueen et al., 2001). Erosion of peat is minimal on land that has an intact vegetation cover but the loss of vegetation exposes more of the soil, increases the rate of runoff and leads to increased rates of erosion. These erosion rates are, however, highly variable and are strongly influenced by the frequency as well as intensity of heavy rain. Automatic instruments that can be deployed on site to monitor the effect of short-term changes in the weather on the downstream transport of sediment provide an ideal means of quantifying these complex responses. Instrumentation of this kind is now commonly used for river flow measurements (Carling, 1995) but most sediment flux studies are still based on the periodic collection of spot samples. In this paper, we describe how a network of custom-designed automatic monitoring stations incorporating newly available sensors and systems are being used to study the downstream transport of sediment in the Burrishoole catchment in the west of Ireland. The study forms part of a wider EU funded project on the applications of automatic monitoring and dynamic modelling for the active management of lakes and reservoirs (LIFE98 ENV/UK/000607). The data acquired by these instruments are currently being used to validate a GISbased model of peat erosion and to assess the impact of year-to-year variations in the weather on the management of fish stocks in the Burrishoole system (Poole et al, 1997). 2 Description of Site The Burrishoole catchment is situated a few kilometres from the town of Newport on the west coast of Ireland (53º 57 N; 9º, 35 W). The catchment is surrounded by the Nephin Beg mountain range, which rises to an altitude of 698 m and includes a large brackish lagoon (Lough Furnace) and a freshwater lake (Lough Feeagh). The main rivers flowing into Lough Feeagh are the Black River and the Glenamong. The Black River flows into the lough from the north and subdivides into five main tributaries: the Maumaretta, the Altahoney, the Gaulaun, the Rough (or Srahrevagh) and the Lodge. The Glenamong flows into Lough Feeagh from the northwest from an upland valley that is surrounded by high mountains. Most of the rivers in the catchment can be classified as spate rivers (Muller, 1997) since the average rainfall is high and the land relatively steep. The annual rainfall typically varies between 1400 mm per year near the coast to more than 2000 mm per year on high ground. The spatial pattern of the rainfall can, however, be quite variable and is usually closely correlated with the direction of the prevailing wind. A very large proportion of the catchment (c. 75%) is covered with blanket peat that can be up to 6 m deep in some sheltered areas (Hammond, 1981). The natural vegetation of the area includes a wide range of peatland species but large areas have recently been denuded by the grazing activities of sheep. In recent years, there has been a significant expansion of forestry in some parts of the catchment. Approximately 30% of the catchment is currently covered by coniferous forest where the main species planted are Pinus contorta (Lodge Pole Pine), Picea sitchensis (Sitka Spruce), Picea abies (Norway Spruce) and Larix decidua (Larch). Some of the erosion and siltation problems encountered in the catchment have been connected with the draining and clear-felling of these forested areas. Others appear to be associated with increased grazing pressures, particularly where sheep are managed under the commonage system of land tenure. No separate statistics are held on the number of sheep held in the Burrishoole catchment but the national flock in Ireland is known to have increased from 3.3 million in 1980 to 8.6 million in 1993. During the course of this study, four automatic monitoring stations have been deployed in the Burrishoole catchment. Here we provide operational details and some example results from the Automatic River Monitoring System (ARMS) installed on the Rough River and Automatic Water Quality Monitoring Station (AWQMS) deployed in Lough Feeagh. Additional data were acquired using a network of twelve Davis Instruments recording rain gauges, seven OTT Orphimedes water level recorders and a number of Sigma samplers that were used to collect water samples for sensor calibration. 2

3 The network of automatic monitoring stations deployed in the Burrishoole catchment GSM telemetry link GSM (cellphone) telemetry Rivers: ARMS Public telephone network Radio telemetry link Shore Station Landline modem and uhf radio telemetry Lakes: AWQMS Remote Station Communications Monitoring Stations Figure 1 The principal features of the instrument network deployed in the Burrishoole catchment. The schematic diagram in Figure 1 shows the principal features of the instrument network deployed in the Burrishoole catchment. In functional terms, the network includes four principal sub-systems: 1. The river monitoring stations (ARMS) used to quantify the downstream transport of sediment in the main rivers. 2. The lake monitoring station (AWQMS) used to monitor the spatial dispersion of sediment in Lough Feeagh. 3. The communication systems (either via a Shore Station using a telephone landline and uhf radio telemetry or direct to the monitoring station via GSM telemetry) that relay data from the river and lake stations. 4. The remote station that retrieves the acquired data for analysis by specialists working at different sites in Ireland and the UK 3.1 The Automatic River Monitoring Station (ARMS) The Automatic River Monitoring Stations (ARMS) deployed on the main rivers utilise elements of a stand-alone profiling system designed by Rouen (1989). Table 1 lists the five environmental parameters recorded by the stations: water temperature; water level; ph; conductivity and an optical estimate of the suspended sediment concentration. All the sensors used were obtained commercially and were selected to provide the best combination of accuracy and proven reliability. The sensor used to measure the temporal variations in the suspended sediment is relatively new, but tests showed that it could detect suspensions of peat silt at concentrations below 0.02 g.l -1. Table 1 Specification of the sensors fitted to the River Monitoring Stations in the Burrishoole catchment. Parameter Sensor Sensor Type Range Resolution Accuracy Water Level Water Temperature ph Conductivity Suspended sediment Druck pressure transducer Platinum resistance AmpHel LTH Electronics Chelsea Instruments Semiconductor strain gauge Stainless steel sheathed PRT Combination electrode 3 Graphite electrode As required 0.01% full scale 0.1% full scale -5 C to +40 C 0.01 C 0.05 C 2 to 14 ph units 0.001 ph units 0.1 ph units 0 to 2000mS.cm -1 0.1µS.cm -1 1µS.cm -1 Nephelometer g.l -1 sediment dependent on composition dependent on sediment composition 3

Support for terrestrial sensors Data Logger Flow Cell Sensors (temperature, ph, conductivity, nephelometer) Hydraulic- Ram Pump Main Outflow Level Sensor Stream Figure 2 Schematic diagram showing the Automatic River Monitoring Station. The schematic diagram in Figure 2 shows how the different components of the River Monitoring Station are linked to form an integrated system. The sensors for measuring water temperature, ph, conductivity and suspended sediment are housed in a flow-cell installed on the river bank. This reduces the risk of flooding and makes it much easier to clean the sensors and replace any faulty transducers. The sensor for measuring the depth of water is mounted in a stilling well that is fixed to the river bed at a point where the level measurements can be converted to an average flow by in situ measurements. River water is delivered to the flow-cell either by an electrically powered peristaltic pump or a hydraulic ram pump that is driven by the head of water along a convenient section of the river. The electronic components of the river monitoring station are mounted in waterproof cases that are then housed within a second weatherproof box fabricated in stainless-steel. This stainless-steel box also provides some screening against possible electromagnetic emissions from the station and reduces its susceptibility to any external electromagnetic interference. A commercial logger that can either be downloaded on site or connected to a remote station by a telemetry link logs the data acquired by the system. The data logger is a Campbell CR10X that stores the recorded data in two output areas. The data logger is programmed so that Area 1 is used to store hourly and daily summary data, while Area 2 is used to record high resolution minute-by-minute data. Both output areas operate as circular buffers so the oldest data in each area is overwritten when that area is full. All cables are connected using high-specification, environmentally sealed, military specification connectors. The system is powered by two sealed lead-acid batteries which can either be exchanged periodically or recharged on-site by a wind generator and /or a solar panel. 3.2 Automatic Water Quality Monitoring Station (AWQMS) The Automatic Water Quality Monitoring Station (AWQMS) deployed on Lough Feeagh has been designed to monitor the thermal characteristics of the lake and the effect of wind-mixing on the horizontal distribution of the suspended sediment. Table 2 lists the different meteorological and limnological parameters measured by the lake station. 4

Table 2 Specification of the sensors fitted to the Lake Monitoring Station deployed on Lough Feeagh MEASURAND SENSOR TYPE RANGE Suspended sediment Nephelometer dependent on sediment composition Dissolved Oxygen Concentration Rapid Pulse Polarographic Cell 0 to 200% air saturation Conductivity 4 Electrode Cell 0 to 100mS.cm -1 Water Quality ph Glass combination electrode 2 to 14 ph units Water Temperature Platinum resistance -5 to +40 C Water column stability Platinum resistance Light Extinction (PAR) Pair of photodiode based 0 to 2500µE.m -1.s -2 quantum light cells (surface and 0 to 100% submerged) Water Level Semiconductor strain gauge 0 to lake depth Incident UV-B Light UV-B cell, photodiode based - Wind Speed Cup anemometer 0 to 50m.s -1 Wind Direction Wind vane (Potentiometer) 2 to 358 Meteorological Atmospheric Pressure Semiconductor strain gauge 0 to 2000mBar Air Temperature Platinum resistance sensor -40 C to +60 C Relative Humidity Semiconductor sensor 0 to 100% Solar Radiation Pyranometer 0 to 1000W.m -2 The sensors for measuring wind speed, wind direction, air temperature, relative humidity, solar radiation and incident UV-B light are all standard commercial units. Most of the aquatic sensors are integrated in a commercial sonde (a YSI6920) but the system also includes surface and sub-surface PAR (photosynthetically active radiation) sensors and a nephelometer to measure the concentration of suspended sediment. A magneticflux gate electronic compass is used to record the alignment (stability) of the buoy and provide the directional information necessary to correct the wind direction measurements. A high-resolution absolute pressure sensor is mounted above water on the buoy to measure barometric pressure. A second absolute pressure sensor is fixed to the lakebed and used to measure the lake level by subtracting the surface pressure from the readings taken in deep water. The other important sensor is the array of platinum resistance thermometers suspended from the buoy. This is used to record the vertical variations in water temperature and provide indirect estimates of the physical stability of the water column. Wind Instruments Incident Radiation (PAR and UV-B and Solar Radiation) Air Temperature and Relative Humidity Data Logger Radio Telemetry Batteries Lake surface Lake surface Submerged quantum sensor (south side of buoy) Pressure Transducer (measuring Lake Level) Sonde Thermistor Chain Mooring Figure 3 Schematic diagram showing the general configuration of the Automatic Water Quality Monitoring Station (AWQMS) 5

The schematic diagram in Figure 3 shows the general configuration of the lake monitoring station. The torroidal buoy is constructed of closed-cell foam and is fitted with a 2 m high mast to support the meteorological instruments. Two steel boxes mounted on the buoy house the control electronics, the data logger, the communication systems and the batteries. The data-logger is a Campbell CR10X unit with six differential analogue channels and eight digital input/output channels. When coupled to optional expansion modules, it is capable of accommodating a vast array of analogue and digital inputs. All the electronic components are modular and are housed in waterproof cases that can easily be exchanged on site. Six sealed lead-acid batteries provide the primary source of power for the system. These can be supplemented by solar cells, although none are fitted as standard. The sealed lead-acid batteries, which are selected for their suitability to operate over a wide range of temperatures, are exchanged at each service visit. The state of charge of the sealed lead-batteries is monitored by the data logger. They are exchanged at the normal service interval (monthly) with a second set of batteries, so a total of twelve are required at each site. Batteries are recharged either at the laboratory or at the shore station. Primary cells can be used to provide an emergency backup to the buoy station in case of an unexpected drain of the main rechargeable batteries or should the rechargeable batteries suffer reduced capacity. 3.3 The Communication Systems Two different methods are used to transfer data from the lake and river stations to the Marine Institute laboratories on the shores of Lough Feeagh. With the lake station, a low power uhf radio transmitter is used to transfer data from the buoy to a shore station housed in the laboratory. At the shore station, a microcontroller linked to a transceiver handles all communications with the buoy and serves as a continuous point of contact with the instruments in the field. A PC can be connected directly to the shore station or it can be contacted by telephone at any time to establish communications with the lake monitoring station to view data in real-time or download stored data. The operator can also communicate via the shore station to reconfigure the sensors or reprogram the data logger on the buoy station. The maximum data transfer rate from the buoy to the shore station is limited to 9600 baud but this still gives satisfactory download times for the quantity of data acquired. This telemetry configuration provides low operating costs (free running costs when a PC is connected directly to the shore station), but line-of-sight conditions from the shore station antenna and buoy station antenna are required for reliable operation at a maximum distance of 10 km. With the river stations, the data is downloaded using a telephone link since all the units were installed at locations that are beyond the range of the uhf transmitter. While the use of one or more uhf repeater stations could have provided a solution, GSM (cellphone) links were chosen instead as they offered a more cost effective way of local communication. The three river monitoring stations had adequate coverage from at least one GSM network and at one location (the Rough River), it was possible to use a conventional land-line connection. 3.4 The Remote Station The remote station consists of a PC that can download data from the river and lake stations via a modem connected to the appropriate telephone network. An easy-to-use Windows program on this PC allows the operator to view and retrieve data and acquire diagnostic information on the performance of the stations. This software allows a remote operator to: 1. Connect to the shore station. 2. Establish a link to the river and lake stations to view or download data. 3. Establish a link to the sub-systems included in the river and lake stations. 4. Set up an automatic download schedule for the river and lake monitors. 6

4 Some example outputs from the river and lake stations a) Rainfall (cm) 16 14 12 10 8 6 4 2 0 29/03/02 30/03/02 31/03/02 01/04/02 02/04/02 Date b) 0.10 80 Suspended Solids Water Level 0.08 60 Suspended Solids (g.l -1 ) 0.06 0.04 40 Water Level (cm) 20 0.02 0.00 29/03/02 00:00 29/03/02 12:00 30/03/02 00:00 30/03/02 12:00 31/03/02 00:00 31/03/02 12:00 01/04/02 00:00 01/04/02 12:00 02/04/02 00:00 02/04/02 12:00 0 03/04/02 00:00 Date / Time Figure 4 Some example results from a) an automatic rain-gauge in the Rough River catchment and b) the ARMS station on the Rough River Figure 4 shows some example results from the ARMS station located on Rough River together with rainfall data from the corresponding period. The water level and suspended sediment data were acquired between 30 March and 3 April 2000 and are the hourly averages stored by the data-logger. The rainfall values are daily averages and were measured by a recording rain gauge located about 30 m away from the river station. The Rough River catchment drains the hillsides on the eastern side of the Burrishoole system. The upper reaches are quite steep and the river usually responds quite quickly to rainfall events. Prior to 1996, 95% of the catchment was forested, but since then approximately 50% of the trees have been harvested and replanted. The results show the effect that a relatively modest increase in the rainfall had on the water level in the river and the downstream transport of sediment. At the beginning of the period selected, very little rain had fallen during the previous 18 days. The first day of heavy rain (30 March) produced a sharp increase in the water level and a brief pulse of sediment but the most pronounced increase in the sediment load was not recorded until 1 April. Delayed responses of this kind are commonly recorded in afforested catchments when the weather has been dry for some time (Mills, 1991). Peaty soils are very difficult to drain gravitationally, but once the soil is saturated substantial quantities of water and soil can be channelled though any drainage ditches present. The impact of this threshold effect on the downstream transport of sediment becomes even clearer when we examine the high-resolution measurements recorded by the ARMS on the 1 April. 7

0.16 80 Suspended Solids Water Level 0.12 60 Suspended Solids (g.l -1 ) 0.08 40 Water Level (cm) 0.04 20 0.00 0 23:00 00:00 01:00 02:00 03:00 04:00 Time Figure 5 The high-resolution measurements recorded by the ARMS station on the night of 31 March / 1 April 2002 Figure 5 shows the minute by minute variations in the water level and suspended sediment recorded by the logger over this critical period. Although the average concentrations of sediment recorded in the river during this event was rather low (c. 0.04 g.l -1 ) there is a very close temporal correlation between the recorded concentrations and the flood hydrograph. 0.11 25 0.10 Suspended Sediment Wind Speed 20 Suspended Sediment (g.l -1 ) 0.09 0.08 0.07 0.06 15 10 Wind Speed (m.s -1 ) 0.05 5 0.04 0 19/05/02 00:00 20/05/02 00:00 21/05/02 00:00 22/05/02 00:00 23/05/02 00:00 24/05/02 00:00 25/05/02 00:00 26/05/02 00:00 27/05/02 00:00 28/05/02 00:00 29/05/02 00:00 Date/Time Figure 6 The temporal variation in the concentration of suspended sediment recorded by the AWQMS in Lough Feeagh in relation to the measured wind speed. Figure 6 shows some example results acquired by the AWQMS station deployed on Lough Feeagh between the 19 May and 29 May 2002. The broken line shows the recorded variation in the suspended sediment and the solid line the variations in the mean hourly wind-speed. Very little is known about the way in which short-term changes in the weather influence the dispersion of sediment in lakes. It is generally assumed that lakes act as large settling tanks where the concentration of suspended sediment decreases progressively with distance from the main inflow. These results, in contrast, suggest that the sediment was being carried down the lake as a welldefined patch followed by a mass of clearer water. The only other factor capable of producing such shortlived variations is the wind-induced re-suspension of the bottom sediment. On this occasion, there was no 8

evidence of any such process since the wind speed was relatively low and there was a weak negative correlation between the sediment concentrations and the measured wind speed. 5 Conclusions In this study, a combination of conventional sampling and automatic monitoring is being used to quantify the impact of anthropogenic and climatic effects on the flux of sediment in the Burrishoole catchment. Particular attention has been paid to the impact of short-lived extreme weather events on the downstream movement of sediment and the dispersion of suspended particles in Lough Feeagh. The results presented here demonstrate that automatic instruments can now be used to monitor these responses at frequencies that are very close to the time-frame of turbulent flow fluctuations. From a technical point of view, the most important features of the network were its overall reliability, the sophistication of the self-diagnostics and the versatility of the assembled components. A significant portion of the design effort was devoted to ensuring optimal reliability of the systems when exposed to hostile environmental conditions. For example, all the external electrical connectors are of military-quality and suitable for use in extreme conditions. Experience has shown these to be exceptionally reliable. All the electronic components are rated to industrial temperature specifications (i.e. rated for operation at temperatures down to -25ºC). Attention has also been paid to minimising the impact of potential faults that may occur in a particular sensor or sub-system. For example, supplies to each sensor are monitored and can be switched on or off either under automatic control of the data logger or through a command issued by an operator at a remote location. Both the river and lake monitoring stations incorporate sophisticated self-diagnostic features that provide quality assurance data as well as system performance records. Engineers at Windermere are able to contact the stations via the telemetry to investigate any suspected sensor or system malfunction and suggest remedial measures to local, non-technical staff. The lake and river monitoring stations are both designed with expansion in mind. The lake station can support almost any sensor that can be powered from a 12V supply. Similarly the river monitoring station has provision for additional sensors and supports a range of water delivery options. In their current configuration, the river and lake monitoring stations store the data in a circular buffers and the high resolution data is overwritten after periods of 1 to 3 weeks. The software could, however, be rewritten to enable these systems to operate in an event-driven mode (e.g. a rapid rise in water level). From a scientific point of view, the most important feature of the network was the simultaneous acquisition of data from the river and lake instruments. The results from the Rough River demonstrate that the movement of sediment down the main channel is strongly influenced by the variation in the rainfall experienced over several days. Although the drainage channels in the forest areas are designed to reduce the risk of flooding they can, during periods of prolonged rain, produce sudden floods that enhance the local rate of soil erosion. The data acquired from Lough Feeagh suggests that erosional events in the catchment also have a significant effect on the spatial distribution of suspended sediment in the lake. The horizontal dispersion of sediment in lakes is commonly viewed as an internal process governed by the mixing effect of the wind. The data presented here suggests that these patterns are also influenced by temporal variations in the quality of the inflows (i.e. the patterns are a relic of previous erosion events). Trout and salmon can tolerate very high concentrations of suspended sediment (Alabaster and Lloyd, 1980) but there is some evidence to suggest that they may delay their upstream movements if they experience a sudden increase in the concentration of particles in suspension. Acknowledgements This study was partly funded by the European Union LIFE Project (LIFE98 ENV/UK/000607). We are particularly grateful to David Aspinall, Jim Crompton, Brian Godfrey, Paul Hodgson, Paul Jones, Jack Kelly, Mike Lee and George Waimann, who have all contributed to the design, construction, testing and installation of the monitoring stations. We also wish to thank Mary Dillane, Marcus Muller, Davey Sweeney and Pat Nixon for help with fieldwork and Diane Hewitt for assisting with the collation and processing of data. References Alabaster, J. S. and Lloyd, R. 1980. Water Quality Criteria for Freshwater Fish. London: Butterworth and Co (Publishers) Ltd. Arnell, N.W., Brown, R.P.C. and Reynard, N.S. 1990. Impact of climatic variability and change on river flow regimes in the UK. Institute of Hydrology, Wallingford. Report No. 107. 154 pp. Carling, P. 1995. Implications of sediment transport for instream flow modelling of aquatic habitat. In: The Ecological Basis of River Management. D.M. Harper and A.J.D. Ferguson (eds). Wiley. Hammond, R.F. 1981. The Peatlands of Ireland. An Foras Taluntais, Dublin. Soil Survey Bulletin No. 35. 9

Maitland, P.S., Newson, M.D. and Best, G.A. 1990. The impact of afforestation and forestry practice on freshwater habitats. Nature Conservancy Council, Peterborough. Report No. 23. 80 pp. Mills, D. 1991. Ecology and Management of Atlantic Salmon. Chapman and Hall, London. 351 pp. Mulqueen, J., Rodgers, M., Marren, N. 2001. Erosion of hill peat in Western Ireland. Proceedings of Conference 32, International Erosion Control Association, February 5-9 2001, Las Vegas, Nevada, USA. Muller, M. 1997. Hydrogeographische Untersuchungen eines spate river systems in Nordwest Irland, Burrishoole catchment, Newport, Co. Mayo), M. Sc. Thesis, Institut fuer Physische Geographie, Johann Wolfgang Goethe Universitaet, Frankfurt am Main, Germany. Poole, W.R., Whelan, K.F., Dillane, M.G., Cooke, D.J., Matthews, M. 1996. The performance of sea trout, Salmo trutta L., stocks from the Burrishoole system, western Ireland, 1970-1994. Fisheries Management and Ecology, Vol.3, No. 1. Rouen, M.A. 1989. The design and development of the Windermere Profiler. Annual Report of the Freshwater Biological Association, 57, 93-106. 10

CRUES DE DECEMBRE 2000 ET JANVIER 2001 EN BRETAGNE ET REFERENCES HISTORIQUES Daniel DUBAND Société Hydrotechnique de France, 25 rue des favorites, 75012 Paris, France Résumé : Les crues du bimestre décembre 2000 - janvier 2001 dans la partie Sud de la Bretagne, supérieures à la crue décennale mais pas exceptionnelles (crues tricennales) si l on se réfère au bassin versant du Blavet à Guerledan, ont véhiculé un volume très rare d apports cumulés sur trois ou six mois, du fait de précipitations modérées mais régulières et répétitives pendant le semestre d automne-hiver sur l Ouest et le Nord Ouest de la France. December 2000 and January 2001 floods in Brittany and historical references. Abstract : The december 2000 - january 2001 floods in the southern area of Brittany, greater than the 10-year floods but not exceptional (30-year floods) as estimated from the Blavet basin at Guerledan, brought a very large cumulated amount of water during three or six months, due to moderate but regular and repetitive precipitations over the Western and North- Western part of France along the fall and winter period. Ar c'hreskoù-dour e Kerzu 2000 ha Genver 2001 e Breizh hag an daveennoù istorel Diverrañ : Kreskoù-dour an daouviziad Kerzu 2000 Genver 2001 e Su Breizh, uheloc'h eget an dek vloaz kresk hep bezañ divoutin (tregont vloaz kreskoù) ma seller ouzh diazaddoureier ar Blavezh e Gwerledan, o deus kaset gante ur c'hementad rouez-tre a zegaserezh berniet war tri pe c'hwec'h miz, abalamour da c'hlaveier bihan met ingal dre veur a wezh e- pad ar c'hwechmiziad diskaramzer-goañv war Kornôg ha Gwalarn ar Frañs. Introduction Cette étude de crues a contribué au rapport de la commission d Expertise Inter-ministérielle sur les inondations en Bretagne. Les crues d automne-hiver n ont pas été exceptionnelles en débits de pointe et en volumes individuels, mais leur répétition et l importance record des volumes cumulés pendant les six mois d octobre à mars est due à une succession d averses régulières d intensités modestes, dans un type de circulation météorologique persistant sur l ouest de l Europe, dont le cumul est le plus élevé observé au cours des 100 dernières années. Il n y a pas eu de victimes, mais des dégâts conséquents. Ce texte a pour objectif de comparer les crues actuelles aux crues passées des rivières pour lesquelles on dispose de séries de mesures de qualité et en nombre suffisant. Pour cela, il est nécessaire dans un premier temps de présenter les caractéristiques qui expliquent le débit des rivières, dont la surface des bassins versants varie de quelques centaines à quelques milliers de kilomètres carrés. Ensuite les pluies tombées sur le bassin versant sont comparées aux écoulements dans les rivières. On remarque que la réponse, des bassins aux pluies, en débits, se situe entre 12 et 30 heures, favorisant la prévision.

1 -Caractérisation des bassins versants étudiés, et étude des débits des rivières On considérera essentiellement les bassins versants de quelques centaines de km² à quelques milliers de km², équipés de stations hydrométriques avec courbes de tarage depuis une trentaine d années, représentant une surface témoin d environ 11000 km². Il s agit de : - l Aulne à Pont Pol Ty Glas (1230 km²) à l ouest, qui prend sa source dans les Monts d Arrée région la plus arrosée par les pluies - l Odet à Tréodet (205 km²) qui prend sa source dans les Montagnes Noires ainsi que l Isole à Quimperlé (224 km²) et l Ellé à Pont Ty Nadan (578 km²) - le Scorff à Pont Kerlo (300 km²) - le Blavet à Guerlédan (620 km²) et à Quellenec (1951 km²) - l Oust à Le Guélin (2465 km²) - la Vilaine à Malon/Guipry (4138 km²). Pour le Blavet à Guerlédan, les mesures de débits moyens journaliers remontent à 1931 (soit 70 ans) avec des observations depuis 1911. Pour les autres bassins, les séries de débits débutent entre1960 et 1980. Si l on considère la position de ces bassins géographiquement, les séries montrent que les apports annuels de pluie dans ces bassins décroissent d ouest en est comme les précipitations. Cette décroissance est assez constante au fil des années. Aussi, il existe de très bonnes corrélations entre les débits moyens mensuels de ces bassins versants pendant les mois d hiver, saison des plus grandes crues. C'est particulièrement le cas entre l Oust et le Blavet, l Ellé et le Blavet. La corrélation est moins élevée entre ces rivières et la Vilaine supérieure. Pour l ensemble de ces bassins, si l on analyse les apports de pluie mensuels, la préparation des sols se fait avec les pluies d octobre et particulièrement celles de novembre, qui conditionnent l état de saturation des mois suivants (même si des crues moyennes ont été observées en automne). Les mois de décembre - janvier - février sont particulièrement favorables aux crues importantes, les rapports écoulement de la rivière/pluie moyenne sur le bassin peuvent se situer entre 50 et 80 % 2 - Caractérisation des précipitations sur ces bassins versants, comparaison entre bassins Il existe un réseau de pluviomètres et de pluviographes relativement important depuis une trentaine d années, gérés par Météo France, les DDE, la DIREN et d autres. Les longues séries historiques remontant à la fin du 19ème siècle sont rares mais précieuses : Quimper, Feins, Laval, Nantes, Chateaulin, Mur de Bretagne depuis 65 ans et sans doute quelques autres, Brennilis.... Les normales (moyennes pluri annuelles) sont estimées par exemple à : -1060 mm à Quimper - 1500 mm à Brennilis - 940 mm à Mur de Bretagne - 980 mm à Lanrivain - 730 mm à Feins, au nord de Rennes Le coefficient de variation des pluies annuelles (écart type/moyenne) est de 0.17. Les corrélations sont élevées entre les pluies des mois d hiver observées à Quimper, Mur de Bretagne et Lanrivain sur les périodes communes. La station de Quimper paraît une bonne référence pour les pluies historiques (100 ans) dans l ouest de la Bretagne (jusqu au bassin de l Oust inclus) tant en valeurs annuelles qu en valeurs mensuelles pour les mois d automne et d hiver (Voir les figures 7, 8 et 9 des graphiques de distribution en fréquence en annexe).

L analyse des données sur 115 années met en évidence les points suivants : 1. En ce qui concerne la pluie cumulée de décembre-janvier-février, à Quimper : la moyenne calculée sur 100 ans est de 360 mm, l écart type calculé sur 100 ans est de 120 mm, soit un coefficient de variation de 0.33. Figure 1 : Pluviométrie et écoulements en Bretagne 2. il n y a pas de tendance significative, semble-t-il, d évolution des quantités de précipitations, en particulier pour le trimestre décembre-janvier-février considérées pour chaque décennie ; les pluies de la décennie 1930-1939 ont les mêmes caractéristiques que celles de la décennie 1990-1999, soit une moyenne de 392 et 394 mm respectivement et un écart type de 163 et 169 mm. Ce sont les décennies les plus humides sur 100 ans, à comparer avec 360 de moyenne et 120 d écart type calculés sur cette durée 3. Dans les distributions empiriques de pluie et d écoulement, les années anciennes et récentes sont équitablement réparties dans les extrêmes faibles et forts. 4. Les gradex des pluies journalières maximales annuelles se situent entre 12 et 6 mm/jour d Ouest en Est..

3 - Comparaison des crues de 1974-1995 - 2000 et 2001 Dans les paragraphes suivants, les paramètres pris en compte sont : Le débit de pointe en m³/s des rivières et le volume d apport en eau en hm³ (cumulé sur 48 h, 10 jours, 1 mois, 3 mois) et l écoulement (en mm) Pour être comparé à la pluie tombée (hauteur d eau en mm), l écoulement de la rivière est exprimé en mm 1. 1 Il est calculé de la manière suivante : le débit en m 3 /s est ramené en volume sur la durée (multiplié par 86400 x le nombre de jours) puis est divisé par la surface du bassin versant en m 2, le tout est divisé par 1000 pour ajuster les unités

0 10 20 30 40 Mur de Bretagne 50 60 70 Pluie (mm) 140 120 100 Débit (m 3 /s) Le Blavet à Guerledan 620 km 2 80 60 40 20 0 Jan/1 Jan/9 Jan/17 Jan/25 Feb/2 Feb/10 Feb/18 Feb/26 Janvier 1974 Février 1974 Fig.2 Pluie à Mur de Bretagne et débit du Blavet à Guerledan en 1974.

0 5 10 15 20 25 30 Mur de Bretagne 35 40 Pluie (mm) 140 120 Débit (m 3 /s) Le Blavet à Guerledan 620 km 2 100 80 60 40 20 0 Jan/1 Jan/9 Jan/17 Jan/25 Feb/2 Feb/10 Feb/18 Feb/26 Janvier 1995 Février 1995 Figure 3 : Pluie à Mur de Bretagne et débit du Blavet à Guerledan en 1995

0 10 20 30 40 Mur de Bretagne 50 Pluie (mm) 150 Débit (m 3 /s) Le Blavet à Guerledan 620 km 2 100 50 0 Dec/1 Dec/9 Dec/17 Dec/25 Jan/2 Jan/10 Jan/18 Jan/26 Décembre 2000 Janvier 2001 Figure 4 : Pluie à Mur de Bretagne et débit du Blavet à Guerledan en 2000-2001

La crue de février 1974 (figure 2) Elle figure parmi les grandes crues, elle est caractérisée par une montée rapide, un débit de pointe élevé (hors bassin complet de la Vilaine) et un volume en 48 heures et 10 jours très important, essentiellement dus à une intensité extrême de plus de 80 mm en 2 jours, soit près de la moitié de l épisode en 10 jours. Tableau 1 : La crue de Février 1974 (10-20/02) en sud Bretagne Rivière Station Surface du bassin versant km² Débit de pointe maximum m³/s Q moyen sur 10 jours m³/s Volume cumulé sur 10 jours hm³ Ecoulement mm AULNE Pt Pol Ty Glas 1224 345 172 149 122 ODET Tréodet 205 94 37.7 33 161 ELLÉ Pt Ty Nadan 578 119 61.4 (53) (92)? SCORFF Pt Kerlo 300 110 44.1 38 127 BLAVET Guerlédan 620 200 83.9 73 117 OUST Le Guelin 2456 268 209.2 181 74 VILAINE Cesson 877 37 28.1 24 28 VILAINE Malon 4138 236 176.4 152 37 La crue de janvier 1995 (figure 3) Elle est la conséquence de trois épisodes pluvieux successifs, dont le cumul entre le 16 et 30 janvier se situe entre 250 et 450 mm selon les stations de mesures. Ils ont engendré 3 crues consécutives chevauchantes sur les bassins supérieurs de moins de 600 km², qui se sont agrégées pour n en former pratiquement qu une seule crue à l aval de bassins, de surface comprise entre 2000 et 4000 km² (Blavet -Oust - Vilaine). Le volume de cette crue en 10 jours est de 290 hm³ pour le Blavet dont 94 à Guerledan (durée de retour entre 80 et 100 ans) et de 650 hm³ pour l Oust et la Vilaine avant leur confluent. Ce volume est considérable alors que les débits de pointe pour chacune des rivières ont une durée de retour de 20 à 30 ans. Tableau 2 : La crue de janvier 1995 (20-30/01) en Sud Bretagne Rivière Station Surface Débit Bassin Versant Maximum km² m³/s Débit moyen sur 10 jours m³/s Volume cumulé sur 10 jours AULNE Pt Pol Ty Glas 1224 400 258.5 223 182 ODET Treodet 205 90 50.1 43 211 ISSOLE Quimperlé 224 110 61.8 53 238 ELLÉ Pt Ty Nadan 578 100 116 100 173 SCORFF Pt Kerlo 300 93 56.4 49 162 BLAVET Guerlédan 620 108.4 94 151 BLAVET Porzo 867 185 139.1 115 133 BLAVET Quellenec 1951 450 333.5 288 148 OUST La Tertraie 929 235 148.5 128 138 OUST Le Guelin 2465 380 296.1 256 104 VILAINE Sevigné 854 125 88.8 77 90 VILAINE Malon/Guipry 4138 490 448 387 94 Total 11085 km 2 1400 hm³ hm³ Ecoulement 10 jours mm

Crues de 2000-2001 (figure 4) Les crues de 2000-2001 ont une étroite parenté avec celles de décembre 1935 et janvier 1936, pour le Blavet à Guerlédan (seuls chiffres dont on dispose actuellement). Individuellement elles sont inférieures à celle de 1995, mais en pluies cumulées et volumes écoulés sur le trimestre décembre - janvier - février, elles sont analogues pour 1936 et 2001, soit respectivement plus de 575 et 630 mm en pluie et plus de 325 et 365 hm³ en volume d apport comme le montrent les calculs réalisés à partir des tableaux suivants. Tableau 3 : Pluies mensuelles en Bretagne en mm Mois Quimper Mur de Rostronen Lorient Tremuson Rennes Vannes Bretagne (Guerledan) Octobre 2000 181 149 181 141 111 116 130 Novembre 2000 259 220 221 236 156 142 242 Décembre 2000 264 272 282 207 127 117 211 Janvier 2001 213 251 262 185 150 146 186 Février 2001 134 108 117 97 64 48 83 Mars 2001 240 193 On notera que la forte valeur à Mur de Bretagne s explique par la position particulière de la station située dans la zone à forte intensité pluvieuse. Le tableau suivant présente le cumul des précipitations des pluies P en mm pour deux périodes en 1936 et en 2000-2001 (P11/2 signifie cumul de pluie de novembre à février) Tableau 4 : cumul des précipitations des pluies pour deux périodes en 1936 et en 2000-2001 Mur de Bretagne P11/2 P10/3 Quimper P11/2 P10/3 en 1936 725 890 907 1098 en 2001 851 1193 870 1291 On peut aussi comparer la pluie tombée à Mur de Bretagne avec l écoulement du bassin du Blavet mesuré à Guerledan et étudier le rendement (c est à dire le rapport entre l écoulement et cumul des précipitations). Tableau 5 : Comparaison de l Ecoulement et du rendement du BLAVET à Guerledan (mm) comparé à la pluie à Mur de Bretagne( moyenne, 1936 et 2000-2001) 0ctobre Novembre Décembre Janvier Février Mars Ecoulement en mm Ecoulement moyen sur 19 40 73 100 94 75 53 ans Ecoulement mensuel 17 61 137 251 135 72 1936 Ecoulement mensuel 2000-2001 27 116 211 242 137 139 Rendement /cumul mensuel moyenne sur 53 ans 0,30 0.42 0.54 0.63 0.68 en 1936 0.32 0.49 0.68 0.74 0.76 en 2001 0.39 0.55 0.67 0.73 0.73

Une valeur supérieure à 0.5 atteinte dans la période novembre-décembre devrait être considérée comme un seuil atteint pour un indicateur de risque On notera que les valeurs extrêmes pour le semestre Octobre-mars sont de 0.35 pour les années très sèches et de 0.8 pour les années très pluvieuses. On notera aussi qu en décembre 2000 et janvier 2001 il y a eu deux crues séparées de 10 jours. La première engendrée par 180 mm en 10 jours véhiculant 70 hm³ en 10 jours (92 hm³ en 15 jours). La deuxième engendrée par 180 mm a véhiculé 75 hm³ en 10 jours (100 hm³ en 15 jours) dont 25 hm³ en 48 h, valeur très comparable à celle de la crue de 1995 en 48 h. 4 - Etude du Blavet à Guerlédan(figure 5) Figure 5 : Le bassin versant du Blavet à Guerledan

Figure 6 Le Blavet à Guerledan (620 km 2 ) Comme on a noté de bonnes corrélations entre les débits du Blavet, de l Oust et de l Ellé, il est intéressant d étudier les données du Blavet à Guerledan qui sont les plus nombreuses et précises. Ce bassin versant de 620 km², se situe entre les altitudes 310 m et 120 m. Le Blavet alimente un réservoir hydroélectrique de capacité totale 49 hm³ mais de capacité utile 31 hm³. La durée historique des observations journalières de débits en fait la seule référence

comparative des crues dans le sud de la Bretagne (hors le bassin de la Vilaine supérieure, à Malon). L apport annuel minimal a été observé en 1921 est de 4.3 m³/s soit 135 hm³, lors de la sécheresse européenne, le débit d été aurait été inférieur à 0.5 m³/s, à l opposé, l apport annuel maximal a été observé en 1951, soit 19.1 m³/s ou 602 hm³. Si l on considère le trimestre Décembre -Janvier- Février, le volume moyen d apports est de 168 hm³ avec un minimum de 40 hm³ en 1976 et un maximum de 320 hm³ en 1936, et probablement un peu plus en 2001 (figure 6). L écoulement moyen est de 270 mm, pour une pluie moyenne de 320 mm à Mur de Bretagne/Guerlédan qui est bien représentative de la pluie moyenne sur le bassin. Par ailleurs la corrélation entre l écoulement au cours du trimestre Décembre - Janvier - Février et la pluie cumulée des mêmes mois, se situe au voisinage de 0.90. Si on prend en compte l écoulement observé en novembre, il permet de calculer le rendement qui est représentatif de l état de saturation du bassin avant l hiver. Par les calculs, on obtient un coefficient de corrélation multiple supérieur à 0,950 (soit 91 % de variante expliquée), avec une relation linéaire simple où le terme prépondérant est la pluie dont l incidence est modulée par la capacité de rétention des sols : Ê12/2 = 0.85P12/2 + 1.25 E (novembre) -50 + & = 631mm+& (écart type erreur = 35 mm) Pour 2001 on trouve 632 mm calculés pour 590 mm observés. Ces calculs tiennent compte de 50 ans d observations). Si l on s intéresse à présent à la durée de retour de l épisode 2000-2001 à Guerledan, il faut auparavant s appuyer sur le tableau suivant qui fournit les éléments de comparaison des 10 plus fortes crues observées en 70 ans. Le tableau 6 ci-après donne pour le Blavet à Guerlédan des estimations de débits et de volumes selon différents modes de calcul : La colonne 1 représente la hauteur de pluie en mm (en 48 h) colonne 2, le Débit moyen journalier maximum en m³/s colonne 3, La hauteur de pluie en mm sur une période de 10 jours colonne 4, le volume de crue en 10 jours en hm³ (paramètre choisi pour le classement) colonne 5, la hauteur de pluie cumulée sur 3 mois en mm colonne 6, le volume mensuel dans le mois de l inondation (janvier ou février) colonne 7, le volume cumulé pour les mois de décembre janvier et février colonne 8, l écoulement calculé pour les cumuls de décembre, janvier et février colonne 9, caractérisation de la crue rapide ou lente (à partir de l analyse des colonnes 4 et 7) Ce tableau permet donc de comparer la nature des différentes crues (lentes ou rapides) Tableau 6 : Le Blavet à Guerlédan 1935-2001 (Bassin Versant 620 km²) Les 10 crues les plus importantes en volume sur 10 jours Date (1) mm (2) m³/s (3) mm (4) hm³ (5) mm (6) hm³ (7) hm³ jan 1995 57 146 189 94 557 153 320 jan 2001 70 139 180 73 631 164 365 fév 1974 82 135 167 73 449 131 217 jan 1936 47 100 129 69 575 156 323 fév 1988 47 87 140 63 514 121 257 fév 1950 88 122 154 61 294 114 180 fév1990 55 104 140 58 500 132 198 fév1935 52 100 162 54 479 87 210 fév1957 37 76 142 53 433 115 233 jan 1982 30 112 104 52 374 106 212 (8) mm 516 590 350 520 415 290 320 340 375 340 Crue 3 rapides lente rapide lente lente rapide rapide rapide rapide rapide

Pour calculer les périodes de retour, différents calculs ont été menés :sachant que la durée moyenne de l hydrogramme de ruissellement direct est comprise entre 24 et 48 h, les calculs par la méthode du gradex ont été réalisés sur la base de temps de 48 h : gradex des pluies extrêmes annuelles, 12.1 mm Tableau 7 durée de retour 10 ans 100 ans 1000 ans probabilité de dépassement 0.1 0.01 0.001 débit moyen en 48 heures m³/s 85-95 180-210 280-310 volume en 48 heures hm³ 14-16 31-36 48-54 débit instantané de pointe en m³/s 125-140 270-310 420-465 en l/s/km² 200-230 430-500 680-750 Estimation du volume en 10 jours en hm³ 50-60 90-100 120-135 Les paramètres de la crue de 2001, caractérisée d une part par un débit moyen de 140 m³/s et un volume en 48 h de 25 hm³ et d autre part un volume de crue sur 10 jours de 71 hm³, ont une durée de retour de l ordre de 25 à 35 ans. Seul, le volume en 3 mois (365 hm³) est le paramètre qui a la valeur la plus élevée depuis 70 ans, dépassant la valeur de 1936, qui a une durée de retour de l ordre de 50 à 70 ans. 5 - Remarques et recommandations Quelques pistes concernant l étude de la genèse des crues en Bretagne. Afin de mieux établir les paramètres importants explicatifs de la genèse des crues, il serait important que des chercheurs s efforcent de : - Tracer des isohyétes à différents pas de temps (1 jours, 2 jours, 10 jours, 15 jours, 1 mois, 3 mois) pour les principales crues observées en automne et surtout en hiver pour le sud de la Bretagne en tenant compte des principaux bassins versants (200 à 4000 km²). - essayer de mieux appréhender la succession temporelle des averses ainsi que leur répartition spatiale qui est importante dans la formation des crues, en calculant la pluie reçue par chaque bassin versant, - étudier parallèlement les causes météorologiques et leur origine : circulation d ouest, de sudouest, de sud, leur persistance chaque jour, - mieux comprendre le processus de saturation progressif qui préparent le rendement des pluies d hiver (décembre-janvier-février) pour des bassins d une certaine taille supérieure à 100 km² - mieux comprendre les processus d agrégation de crues amont avec celles de l aval pour les bassins versants importants en ne se perdant pas dans le détail spatial ou temporel Par ailleurs, la connaissance et les données de précipitations observées sont sous exploitées pour l hydrologie des crues supérieures à la décennale en Bretagne, alors que c est un facteur prépondérant, il faudrait mieux étudier les intensités journalières (et 6 h ou 12 h), surtout le volume, vue la modestie des rétentions. Pour se convaincre de la nécessité de telles études, il faudrait remettre à sa juste place l affirmation de changement climatique due à l effet de serre, qui reste prématurée tout en pervertissant les décisions de prévention et de gestion de crise par une fuite en avant occultant la mémoire historique par alibi, amnésie ou ignorance. En

s appuyant sur l histoire, il faut donc revitaliser la pédagogie de connaissance et la familiarisation aux risques naturels. (les fluctuations climatiques des pluies et températures air depuis 150 ans concernent autant la Bretagne qu une grande partie de la France en terme de séquences --10 à 50 ans humides et sèches). Ces études devraient bien sûr tenir compte des cohérences importantes et aussi de quelques différences de l hydrologie des bassins d ouest en est de la Bretagne. Pour cela, il faut réaliser de véritables descriptions hydrologiques (avec hydrogrammes) et pluviométriques des crues importantes, avant de modéliser. Commentaires sur les actions possibles sur les crues, rétentions amont ou accélération aval La rétention de l eau dans des espaces d expansion naturelles ou par des bassins artificiels a pour effet d écrêter soit une pointe de crue soit un volume pour créer un effet retard dans les bassins supérieurs de fleuves ou de rivières ( ceci plutôt en cas de crues rapides), le but étant d éviter la conjonction naturelle avec des affluents à l aval. Il est alors aussi essentiel de vérifier que l on ne provoque pas artificiellement une conjonction d affluents, ce qui suppose de bonnes prévisions en pluies-débits. L accélération vers l aval, dans le cas de bassins versants à faible pente soumis à des crues lentes successives et volumineuses engendrées par des pluies longues et régulières (plusieurs semaines et mois), peut éviter de constituer des zones d accumulations extensives. En effet, il ne faut pas oublier que de telles zones favorisent la saturation par «en dessous», et que, soumises à des pluies mêmes modérées, elles ont un rendement élevé pour l écoulement qui n arrive plus à s évacuer. Il faut donc retenir qu il n y a pas de solutions miracles, mais des solutions adaptées à la morphologie d un bassin, à la météorologie et à la pluviométrie auxquelles il est sujet, particulièrement pour les événement rares ( de fréquence supérieure à la décennale) Annexe : Analyse statistique des pluies à Quimper Figure 7 Quimper Cumuls mensuels d Octobre à Septembre de 1886-1887 à 1999-2000

Figure 8 Quimper Cumuls mensuels d Octobre à Mars de 1886-1887 à 1999-2000 Figure 9 : Quimper Cumuls mensuels de Décembre à Février de 1886-1887 à 1999-2000

An introduction to the Galway RiverFlow Forecasting System (GFFS) Keshav P. Bhattarai*, Monomoy Goswami* and Kieran M. O Connor * *Department of Engineering Hydrology, National University of Ireland, Galway. Abstract The sequential structure of the GFFS is presented, with a summary of its main features. It provides the un updated river flow forecasts of its suite of rainfall-runoff models, the corresponding consensus forecasts, and the updated forecasts. Brief descriptions of the models are given and also selected snapshots of windows of the package. Resumé La totalité de la structure du GFFS est présentée, ainsi qu un résumé de ses principales caracteristiques. Il fournit les previsions non mises a jour des differents modeles, les prévisions combinées et les prévisions mises a jour. Des descriptions succintes de ces modeles sont fournies, et également des illustrations par impression-ecran du GFFS.

Addresses of Authors Abesser, Corinna ca22@st-andrews.ac.uk Department of Geography & Geosciences, University of St. Andrews, St. Andrews, KY1 69AL, Scotland, UK. Allott, N. nallott@tcd Trinity College, Dublin. Aquilina, L. Luc.aquilina@univ-rennes1.fr Laboratoire de Ge Geosciences Universite de Rennes 1 Campus de Beaulieu F-35042 Rennes, France. Artur, A. Mission Inter Services de l Eau, Préfecture du Finistère, Quimper, France. Bhattarai, K. P. keshav.bhattarai@nuigalway.ie Department of Engineering, Hydrology, National University Of Ireland, Galway, Ireland. Bendjoudi, H. Hocine.Bendjoudi@ccr.jussieu.fr UMR Sisyphe, LGA, Université Paris VI, France. Bertru, G. georges.bertru@univ-rennes1.fr U.M.R. Ecobio Université de Rennes I, France. Bird, E. Camborne School of Mines, University of Exeter, Redruth, Cornwall TR15 3SE, GB. Boorman, Laurie. laurie.boorman@btinternet.com, L A B Coastal, The Maylands, Holywell, St. Ives, Cambs. PE27 4TQ, UK. Brient, L. Luc.Brient@univ-rennes.1fr U.M.R. Ecobio Université de Rennes I, France. Cawley, Tony hydroenvironmental@eircom.net, Hydro environmental Ltd., Unit 3, Campus Innovation Centre, Newcastle, Galway, Ireland. Chalmers, Neil NA_Chalmers@compuserve.com The Macaulay Institute, Craigiebuckler, Aberdeen, AB15 8QH, Scotland. Cullen, Joanne BANGOR1.Cullej@environment-agency.gov.uk, Environment Agency Wales, Parc Menai, Bangor, LL57 4DE.Wales. Cunnane, Conleth Conleth.Cunnane@nuigalway.ie, Department of Engineering, Hydrology, National University of Ireland, Galway, Ireland. Crabtree, Bob r.crabtree@zetnet.co.uk The Macaulay Institute, Craigiebuckler, Aberdeen, AB15 8QH, Scotland. Duband, Daniel shf@shf.asso.fr Societe Hydrotechnique de France, 25 rue des favorites, 75012 Paris, France. Demissie, Solomon S. Solomon.Seyoum@nuigalway.ie, Department of Engineering, Hydrology, National University of Ireland, Galway, Ireland. Dreis, Gerda g.dreise@wilckenwiericke.nl Head Dept. Integrated Water Management & Planning, Water Board Wilck and Wiericke, The Netherlands.

Dunn, S. M. s.dunn@macaulay.ac.uk, The Macaulay Institute, Craigiebuckler, Aberdeen, AB15 8QH, Scotland. Dysarz, Tomasz todys@pg.gda.pl Gdansk Technical University, ul. Narutowicza 11/12, 80-952 Gdansk, Poland. El-Sadek, Alaa Alaa.El-Sadek@agr.kuleuven.ac.be, Institute for Land and Water Management, K.U.Leuven, Vital Decosterstraat 102, B-3000 Leuven, Belgium. Feyen, Jan Email? Institute for Land and Water Management, K.U.Leuven, Vital Decosterstraat 102, B-3000 Leuven, Belgium. Frengstad, Bjørn Bjorn.Fregstad@ngu.no Norwegian University of Science and Technology/Geological Survey of Norway, Trondheim, Norway. Freslon, M. Direction Départementale de l Agriculture et de la Forêt de la Manche, Saint-Lô, France. Gascuel-Odoux, C. cgascuel@roazhon.inra.fr UMR INRA Sol Agronomie Spatialisation 65, Route de St Brieuc- CS 84215 F-35042 RENNES France. George, D.G. dgg@ceh.ac.uk Centre for Ecology and Hydrology, Windermere, Cumbria, UK. Gleeson, T. TGleeson@Kinsealy.Teagasc.ie, Teacgasc, Kinsealy, Co Dublin Goswami, M. Monomoy.Goswami@nuigalway.ie Department of Engineering Hydrology, National University of Ireland, Galway, Galway, Ireland. Grant, Jerry jgrant@dublin.mcos.ie M.C.O'Sullivan & Co. Ltd, Ashurst, Blackrock, Co Dublin. Grésillon, Jean-Michel Jean-Michel.Gresillon@hmg.inpg.fr Laboratoire d étude des Transferts en Hydrologie et Environnement, 38402 St Martin d'hères, France. Haria, Atul atu@ceh.ac.uk,centre for Ecology and Hydrology (CEH), Crowmarsh Gifford, Wallingford, Oxfordshire OX10 8BB, UK. Helliwell, Rachel R.Helliwell@macaulay.ac.uk, Integrated Catchment Management, Macaulay Institute, Craigiebuckler, Aberdeen, AB15 8QH, UK. Hubert, Pierre iahs@ensmp.fr, Secrétaire général de l'aish, Ecole des Mines de Paris 35 rue Saint Honoré, 77305 Fontainebleau, France. Janman, S. email Camborne School of Mines, University of Exeter, Redruth, Cornwall TR15 3SE, GB. Javelle, Pierre pierre.javelle@ucd.ie University College Dublin, Centre for Water Resources Research, Earlsfort Terrace, Dublin 2, Ireland. Jenkin, Loveday E.T. L.Jenkin@csm.ex.ac.uk Camborne School of Mines, University of Exeter, Redruth, Cornwall TR15 3SE, GB. Jigorel, Alain Alain.Jigorel@insa-rennes.fr GRGC Rennes Géologie, INSA de Rennes, 20 Av des Buttes de Coesmes, CS 14315, 35043 Rennes Cedex, France.

Jones, J. A. A. Jaj@aber.ac.uk Reader, Institute of Geography and Earth Science, University of Wales, Aberysrwyth, SY23 3DB, UK. Janssen, Geert G.Janssen@fugro.nl Dept. Hydrology, FUGRO, The Netherlands. Joris, Ingeborg email? Institute for Land and Water Management, K.U.Leuven, Vital Decosterstraat 102, B-3000 Leuven, Belgium. Juggins, S. Stephen.Juggins@ncl.ac.uk Department of Geography, University of Newcastle, Newcastle upon Tyne, NEI 7RU, UK. Kilcullen, E. Edward.kilcullen@nuigalway.ie, Department of Engineering, Hydrology, National University of Ireland, Galway, Ireland. Krijgsman, B. baskrijgsman@hotmail.com Free University Amsterdam, Van Tuyll van Serooskerkenweg 101-II, 1076 JJ Amsterdam. Long, P., Environment Agency, Bodmin, Cornwall, UK. Lobo-Ferriera, J.P. lferreira@lnec.pt. Laboratorio Nacional de Engenharia Civil, Hydraulics Department, Groundwater Research Group Head, Avenida do Brasil, 101, P-1700-066 Lisboa, Portugal. Lovejoy, S. lovejoy@physics.mcgill.ca Physics Department, McGill University, Montreal, Canada. Maleval, Véronique email? Département de Géographie de l Université de Limoges -UMR 6042- CNRS Clermont-Ferrand, INSA de Rennes, GRGCR Géologie, Rennes, France. Marjolet, G. MARJOLETGilles@cg22.fr Conseil Général des Côtes d Armor, Saint-Brieuc, France. Martin, Charlotte martin@roazhon.inra.fr, UMR INRA Sol Agronomie Spatialisation 65, Route de St Brieuc- CS 84215 F-35042 RENNES France. Martin, J. john.martin@epost.de Institute for Space Sensor Technology & Planetary Exploration, German Aerospace Center, Berlin, Germany. Mc Dermott, D. Dermot.Mcdermott@nuigalway.ie, Department of Engineering, Hydrology, National University of Ireland, Galway, Ireland. McGinnity, P. Central Fisheries Board, Glasnevin, Dublin. Mc Garrigle, M. m.mcgarrigle@epa.ie Environmental Protection Agency, St Martin s House, Waterloo Road, Dublin 4, Ireland. Merot, Philippe, UMR Sol Agronomie et Spatialisation INRA-ENSAR, 65 route de Saint-Brieuc, CS84215, 35042 Rennes cedex, France. Misstear, B. D.R. bmisster@tcd.ie, Department of Civil, Structural and Environmental Engineering, Trinity College Dublin 2, Ireland. Migliori, L. Department of Civil & Environmental Engineering, National University of Ireland, Cork, Ireland.

Molenat, J. Jerome.molenat@roazhon.inra.fr UMR INRA Sol Agronomie Spatialisation 65, Route de St Brieuc- CS 84215 F-35042 RENNES France. Morin J P. MorinJeanPierre@cg22.fr Conseil General des Cotes d Armor, DAE, St Brieuc, France. Napiorkowski, J. Jaroslaw jnn@igf.edu.pl, Institute of Geophysics, Polish Academy of Sciences, ul. Ksiecia Janusza 64, 01-452 Warszawa, Poland. O Connor, K.M. Kieran.OConnor@nuigalway.ie, Department of Engineering Hydrology, National University of Ireland, Galway, Galway, Ireland. O Hea, B. Brendan.ohea@marine.ie Marine Institute, Newport, County Mayo, Ireland. O Kane, J.P. p.okane@ucc.ie, Department of Civil & Environmental Engineering, National University of Ireland, Cork, Ireland. Purcell, P. J. PJ.Purcell@ucd.ie, Department of Civil Engineering, University College, Earlsfort Terrace, Dublin 2, Ireland. Rouen, M.A. Centre for Ecology and Hydrology, Windermere, Cumbria, U.K Robins, N. S. nsro@bgs.ac.uk British Geological Survey, Crowmarsh Gifford, Wallingford Oxfordshire OX10 8BB,UK. Ruiz, L. ruiz@rennes.inra.fr UMR INRA Sol Agronomie Spatialisation 65, Route de St Brieuc- CS 84215 F-35042 RENNES France. Ryan, M. Department of Civil & Environmental Engineering, National University of Ireland, Cork, Ireland. Sauquet, Eric eric.sauquet@cemagref.fr Cemagref - Groupement de Lyon, Unite de Recherche Hydrologie-Hydraulique, 3 bis quai Chauveau, CP220, 69336 LYON Cedex 09, France. Soulsby, Chris, Department and Environment, University of Aberdeen, Aberdeen AB24 3UF, Scotland,UK. Schertzer, D. schertze@ccr.jussieu.fr Météo-France and LMM, Université Paris VI, Farnce. Scully, Heather heatherscully@hotmail.com Civil Engineering Department, University College, Earlsfort Terrace, Dublin 2, Ireland. Shamseldin, A.Y. a.shamseldin@bham.ac.uk Department of Civil Engineering, The University of Birmingham, Edgbaston, Birmingham B15 2TT, UK. Shand, Paul. psd@bgs.ac.uk, British Geological Survey, Maclean Building, Crowmarsh Gifford, Wallingford Oxfordshire, OX10 8BB, UK Smyth, J. A. tony.smyth@opw.ie" Office of Public, 51 St. Stephen's Green, Dublin 2, Ireland.

Stalham, Mark ms123@cus.cam.ac.uk Cambridge University Farm Potato Agronomy Unit, Cambridge, England. Tchiguirinskaia, I. tchigin@biogeodis.jussieu.fr UMR Sisyphe, LGA, Université Paris VI, France. Terwey, Leonard l.terwey@fugro.nl Water & Environment, FUGRO, The Netherlands. Tollow, A. J. tollow@civil.ntech.ac.za, Director of Studies (Water), Technikon Natal, Box 684, KLOOF 3640, KZNatal South Africa. Viaud, Valérie Viaud@roazhon.inra.fr UMR Sol Agronomie et Spatialisation INRA-ENSAR, INRA, Unite SAD Armorique, Rennes 65 route de Saint-Brieuc, CS84215, 35042 Rennes cedex, France. Wilson, David dwilson@galway.mcos.ie M.C.O'Sullivan & Co. Ltd, The Grainstore, Abbeygate St. Lower, Galway, Ireland.