Doctorat en sciences forestières Philosophiæ doctor. Québec, Canada

Save this PDF as:
 WORD  PNG  TXT  JPG

Dimension: px
Commencer à balayer dès la page:

Download "Doctorat en sciences forestières Philosophiæ doctor. Québec, Canada"

Transcription

1 Doctorat en sciences forestières Philosophiæ doctor Québec, Canada Martin Barrette, 2015

2

3 Résumé La surabondance de cervidés représente une nouvelle perturbation des écosystèmes forestiers qui interagit avec les perturbations préexistantes des forêts naturelles. Les interactions entre de multiples perturbations peuvent altérer les mécanismes de résilience des forêts et engendrer de nouvelles trajectoires successionnelles menant à des assemblages d espèces n ayant jamais coexisté historiquement. L occurrence de tels nouveaux écosystèmes soulèverait des enjeux de conservation de la biodiversité et de maintien des services écosystémiques fournis par la forêt naturelle (c.-à-d. préindustrielle). Il devient donc important d identifier d éventuels mécanismes de perte de résilience qui pourraient empêcher la forêt naturelle de récupérer des interactions entre les perturbations préexistantes et celles occasionnées par les cervidés. L île d Anticosti qui abrite des cerfs de Virginie (Odocoileus virginianus) surabondant, donne l occasion d identifier de tels mécanismes. L identification de mécanismes de perte de résilience nécessite d abord de reconstituer la variabilité naturelle que la forêt préindustrielle devrait retrouver après perturbation, et de comprendre sa dynamique des perturbations naturelles. Le paysage naturel de l île d Anticosti était caractérisé par une matrice de sapinières surannées principalement dynamisées par des petites trouées et où les perturbations sévères générant de nouveaux peuplements ont été rares au cours des derniers ~160 ans. Le broutement préférentiel de la régénération par les cerfs a interrompu les processus de régénération des trouées des sapinières surannées de l île d Anticosti. La perte de ce mécanisme de résilience a engendré la dégradation du couvert et de nouvelles trajectoires successionnelles menant à des pessières blanches et des forêts-parcs d épinettes blanches. La coupe totale historique ( ) des sapinières surannées a formé des pessières blanches de seconde venue maintenant matures. Nous avons démontré que l interaction entre le broutement préférentiel de la régénération par le cerf et la coupe totale de ces pessières blanches altère les processus de régénération des forêts et engendre une trajectoire successionnelle menant à la formation de forêts-parcs d épinettes blanches. iii

4 Un plan d aménagement vise actuellement la restauration de l habitat du cerf sur l île (c.-àd. les sapinières). Lors d une éventuelle révision de ce plan, les aménagistes devraient considérer trois enjeux importants de l aménagement écosystémique, soit le maintien de sapinières surannées, l occurrence de pessières blanches en tant que nouvel écosystème et le maintien de l état forestier. iv

5 Abstract Overabundance of cervids represents a new disturbance of forest ecosystems which interacts with preexisting disturbances of natural forests. Interactions between multiple disturbances can alter resilience mechanisms, thereby triggering alternative successional pathways that move the system toward assemblages of species that have not co-occurred historically. The occurrence of such novel ecosystems would raise issues concerning conservation of biodiversity and ecosystem services that are provided by natural (i.e., preindustrial) forests. Hence, it is important to identify eventual mechanisms of resilience loss which could prevent natural forests recovery from interactions between pre-existing disturbances and disturbances caused by cervids. Anticosti Island which shelters overabundant white-tailed deer (Odocoileus virginianus), provides an opportunity to identify such mechanisms. To identify mechanisms of resilience loss, it was first necessary to describe the variability which the natural forest should recover after disturbances and to understand its disturbance dynamics. The natural landscape of Anticosti Island was characterized by a forest matrix of overmature gap-driven balsam fir stands in which severe stand-initiating disturbances were rare over the last ~160 years. Preferential browsing by deer has disrupted regeneration processes of gap-driven balsam fir stands on Anticosti Island. The loss of this resilience mechanism triggered forest degradation and alternative successional pathways toward white spruce stands and parklands. The historic clear-cutting ( ) of overmature balsam fir stands has formed mature second-growth white spruce stands. We have shown that interactions between preferential deer browsing and clear-cutting of these white spruce stands have altered regeneration processes and triggered an alternative successional pathway toward the formation of parklands. v

6 A management plan aims to restore deer habitat on the island (i.e., balsam fir stands). In the eventual revision of this plan, managers should consider three important ecosystem management issues: the maintenance of overmature balsam fir stands, the occurrence of white spruce stands as a novel ecosystem, and the maintenance of the forest state. vi

7 Table des matières Page Résumé... iii Abstract...v Listes des tableaux... xi Liste des figures... xiii Remerciements... xvii Avant-propos... xix Introduction générale...1 La surabondance de cervidés et la résilience de la forêt naturelle...1 Connaissance de la sapinière préindustrielle et de sa dynamique des perturbations naturelles...2 Interactions entre la dynamique des petites trouées et le broutement préférentiel de la régénération...4 Interactions entre la coupe totale et le broutement préférentiel de la régénération...6 Chapitre 1 : Preindustrial reconstruction of a perhumid mid-boreal landscape, Anticosti Island, Quebec Résumé Abstract Introduction Materials and methods Study area Historical data Landscape analysis Field sampling and laboratory procedures Dendrochronological analysis Results Preindustrial landscape structure and composition Landscape metrics and distributions of patches of young stands in the overmature forest matrix Growth reduction periods within the time span of the historical data Discussion...31 vii

8 Preindustrial landscape composition, spatial structure, and natural disturbances Implications for conservation of forest ecosystems Acknowledgements References...38 Chatpitre 2: Preferential deer browsing disrupts regeneration processes of gapdriven balsam fir stands triggering forest degradation and alternative successional pathways Résumé Abstract Introduction Materials and methods Study Area Field sampling Data analysis Gap regeneration processes and forest degradation processes Spatial pattern of forest degradation and pathways of succession Results Gap regeneration processes Forest degradation processes Spatial patterns of forest degradation Pathways of succession Discussion Disruption of gap regeneration processes and forest degradation Spatial pattern of forest degradation and pathways of succession Implications for resilience of gap-driven forests Acknowledgements References...71 Chapitre 3: Cumulative effects of chronic deer browsing and clear-cutting on regeneration processes in second-growth white spruce stands Résumé Abstract Introduction Materials and methods Study area Field sampling Data analysis Results...88 viii

9 Stand characteristics and seed availability Seedling density and stocking Substrate suitability and availability for seedling establishment Discussion Regeneration processes in mature second-growth white spruce stands Regeneration processes in recent clear-cuts of mature second-growth white spruce stands Conclusion Implications for management of white spruce stands Acknowledgements References Conclusion générale Variabilité et dynamique naturelle de la sapinière préindustrielle de l île d Anticosti Dégradation du couvert des sapinières préindustrielles et nouvelles trajectoires successionnelles Processus de régénération défaillant dans les pessières blanches de seconde venue matures et nouvelle trajectoire successionnelle Conséquences pour l aménagement des forêts de l île d Anticosti Références de l introduction et de la conclusion générale ix

10

11 Listes des tableaux Page Table 1-1. Range of natural variability and natural tendency for each stand age class and type calculated from the variability observable at the scale of the historical management compartments used to describe the preindustrial landscape of Anticosti Island, Quebec,...25 Table 1-2. Landscape metrics by age class and size class of patches describing the preindustrial landscape of Anticosti Island, Quebec, Canada...27 Table 2-1. Description of decay class for snags and logs...53 Table 2-2. Mean (±1 SE) number of stems per hectare of understory regeneration and canopy trees in 9 transects located in old balsam fir stands of Anticosti Island, Quebec, Canada...58 Table 2-3. Mean (± 1 SE) tree (DBH 10 cm) density per hectare of snags and logs according to decay stages and mortality process in 9 transects and estimated ( X ± 1 SE) time since death of balsam fir trees according to decay stages, in 8 old balsam fir stands (dendrochonological plots; Fig. 2-2 ) of Anticosti Island, Quebec, Canada...59 Table 3-1. Mean (±1 SE) density (D, seedling ha 1 ) and stocking (S, %) of seedlings in the four ecosystem types...90 Table 3-2. Mean percentage (±1 SE) of the area covered by seedling substrates and mean heights of logs and live individuals of Recalcitrant Layer Species by ecosystem type...93 xi

12

13 Liste des figures Page Fig (A) Location of the study area by historical forest management blocks and of the dendrochronological study plots;. (B) Reproduction of the map of the ecoclimatic regions of Canada located around the Gulf of St. Lawrence. (GTE 1989). Reproduced with the permission of the Minister of Public Works and Government Services Canada, Fig Scanned historical forest map of Shallop Forest management block with an expanded view of the legend...15 Fig Stand age class distribution by type in the area covered by 14 historical management blocks of Anticosti Island, Quebec, Canada...21 Fig Distribution of the relative abundance of each stand age class in 305 historical management compartments (mean size: 23 km²) of Anticosti Island, Quebec, Canada Fig Proportion of the total area by age class in 14 historical management blocks (mean size: 494 km²) of Anticosti Island, Quebec, Canada...23 Fig Distribution of the relative abundance of each stand type in 305 historical management compartments (mean size: 23 km²) of Anticosti Island, Quebec, Canada Fig (A) Distribution of the number of patches by class of patch size of 100 ha for stands in ages classes 100 years. (B) Expanded view showing the distribution of the number of patch by class of patch size of 1 ha for patches 100 ha Fig Distribution of the area covered by class of patch size of 100 ha for stands in ages classes 100 years Fig Mean standardized tree-ring chronologies of host species (black lines for (aa) balsam fir (Abies balsamea) and (bb) white spruce (Picea glauca)) and non-host species (grey line for jack pine (Pinus banksiana)), with associated histogram illustrating the proportion of stems showing significant growth reductions and sample depth. Grey bands indicate significant growth reduction periods. Percentages in the grey bands indicate the proportion of stems showing a significant growth reduction in each period for each plot...30 Fig Diagram illustrating predictions of process triggered by a shift in the composition of understory regeneration of gap-driven balsam fir forests...49 Fig Location of the transects and of the dendrochronological plots in balsam fir stands of Anticosti Island, Quebec, Canada...52 xiii

14 Fig Distribution of mean (± 1 SE) stem density by diameter at breast height (DBH) classes in the 9 transects Fig Distribution of number of sections of transects (100 m2; n = 351) by density classes of understory white spruce regeneration (A) and canopy trees (B)...61 Fig Wavelet scale analysis of spatial patterns of forest degradation processes calculated from the spatial distribution of canopy tree basal area (DBH 10 cm) in the 9 transects (see Fig. 2-2 for reference number). Black line (wavelet variance) above grey line (confidence interval >90%) shows significant scales of spatial patterns, identified SG (small gap 200 m 2 ) or LG (large gap >200 m 2 )...63 Fig Transect length by canopy gap criteria, spatial pattern of forest degradation and white spruce stand regeneration criteria...64 Fig Spatial distribution (histograms) of canopy trees (A; diameter at breast height (DBH) 10 cm) and understory regeneration (B; saplings (1 cm < DBH < 10 cm) and tall seedlings (height cm) along contiguous 25 m sections (100 m 2 ) of 2 representative transects (see Fig. 2-2 for reference number). The edge of a canopy gap is encountered (A) when the following criteria are met: 1) transect section is under or next to a significant peak in wavelet position variance of canopy tree basal area, i.e.the black line (wavelet variance) is above dotted line (confidence interval >90% of wavelet variance), 2) basal area of live canopy trees is 25% (i.e. canopy density class D: MRNQ 2011) of a fully stocked balsam fir stand (i.e. 13 m 2 /ha: Zarnovican and Vézina 1985) and 3) basal area of dead canopy trees is above range of natural variability (12-23 m 2 /ha), or 4) the openness index is 60%. Transect sections meeting these gap criteria are identified as SG (i.e. small gap 200 m 2 ) if they covered one or two neighbouring sections or LG (i.e. large gap >200 m 2 ) if they covered more than two neighbouring sections. A section was considered regenerated enough to fill gaps and eventually recreate a closed canopy forests (B) if tall seedling and sapling density was over 1500 stems/ha (Greene et al. 2002) Fig Photographs showing the four ecosystem types of the study Fig Location of regeneration grids and transects in the four ecosystem types that were studied on Anticosti Island, Québec, Canada Fig Distribution of mean (±1 SE) stem density by diameter at breast height (DBH) classes and mean (±1 SE) seed density in 4 stands by cover type...89 Fig Estimated ( X ± 1 SE) index of the suitability of nine substrates for establishment of white spruce seedlings. The index ranges from 0 (low suitability) to 1 (high suitability) xiv

15 À Mathilde, Félix et Christine xv

16

17 Remerciements Je tiens tout d abord à remercier mon directeur, Louis Bélanger, d avoir su être suffisamment intéressé, patient et loyal pour réaliser mon éducation scientifique et forestière. Je remercie mon codirecteur, Louis De Grandpré, pour son sens critique qui a contribué à améliorer la qualité de ma rédaction en vue de la publication. Je remercie également Jean-Claude Ruel pour sa précieuse collaboration dans le troisième chapitre de ma thèse. Merci à Jean Huot, Jean-Pierre Tremblay et Mathieu Bouchard pour les discussions éclairantes, William F.J. Parsons pour la révision de la langue anglaise, Denise Tousignant pour la révision de la langue française et Nathalie Langlois pour la mise en page. Merci aux étudiants chercheurs et professionnels de recherche du laboratoire d aménagement intégré et de la Chaire de recherche industrielle CRSNG-Produits forestiers Anticosti pour leurs interactions enrichissantes. Je remercie aussi Caroline Boyau, Mélanie Veilleux-Nolin, Francois Lebel et Christine Marquilly pour leur dévouement lors de la récolte de données sur le terrain et des analyses en laboratoire. Cette étude n aurait pas été possible sans le soutien financier de la Chaire de recherche industrielle CRSNG- Produits forestiers Anticosti et la contribution financière du Ministère des Forêts, de la Faune et des Parcs du Québec, de l Ordre des ingénieurs forestiers du Québec et de la Fondation de la faune du Québec. J aimerais remercier mes parents d avoir fait en sorte que je puisse choisir un parcours en toute liberté. Finalement, j aimerais offrir un merci spécial à Christine d avoir été compréhensive et de m avoir encouragé et soutenu durant toutes ces années, notamment lorsque la vraie vie a fini par nous rattraper. xvii

18

19 Avant-propos Ce document est présenté sous la forme d une thèse par articles. La thèse a été élaborée selon les critères du comité de programme de 2 e et 3 e cycles en sciences forestières de l Université Laval. Les articles suivants, rédigés en anglais, sont inclus dans la thèse. Chapitre 1: Barrette, M., Bélanger, L. et De Grandpré, L Preindustrial reconstruction of a perhumid midboreal landscape, Anticosti Island, Quebec. Canadian Journal of Forest Research 40: Chapitre 2: Barrette, M., Bélanger, L. et De Grandpré, L. Preferential deer browsing disrupts regeneration processes of gap-driven balsam fir stands triggering forest degradation and alternative successional pathways. Sera soumis sous peu à Journal of Applied Ecology. Chapitre 3: Barrette, M., Bélanger, L., De Grandpré, L. et Ruel, J-C Cumulative effects of chronic deer browsing and clear-cutting on regeneration processes in secondgrowth white spruce stands. Forest Ecology and Management. Volume 329, 1 Octobre 2014, Pages 69-78, ISSN , En tant qu étudiant au doctorat et premier auteur de ces articles, j ai déterminé les questions et hypothèses de recherche, j ai effectué les revues de littératures, j ai planifié et réalisé la récolte de données sur le terrain et les analyses en laboratoire, j ai analysé les données et rédigé les manuscrits. Louis Bélanger (directeur), Louis De Grandpré (codirecteur), et Jean-Claude Ruel (collaborateur), m ont conseillé durant mon cheminement et ont commenté les manuscrits lors du processus de rédaction. xix

20

21 Introduction générale La surabondance de cervidés et la résilience de la forêt naturelle Une cause récente d inquiétude est la surabondance de cervidés dans plusieurs forêts du monde causée par de nouvelles introductions, la diminution des prédateurs et l augmentation de la nourriture disponible (Danell et al 2003; Côté et al. 2004; McLaren et al 2004; Chollet et Martin 2012). Les cervidés peuvent être jugés surabondants lorsque qu ils causent des dysfonctionnements de l écosystème (Caughley 1981; Côté et al. 2004). La surabondance de cervidés représente habituellement une nouvelle perturbation qui interagit avec les perturbations préexistantes des forêts naturelles (Chapin et al. 2004; Frelich et Reich 2010; Royo et Carson 2010). Les interactions entre de multiples perturbations peuvent altérer les mécanismes de résilience des forêts et engendrer de nouvelles trajectoires successionnelles (Paine et al. 1998; Tremblay et al. 2007; Buma and Wessman 2011). La résilience est la capacité d un système d absorber une perturbation et de se réorganiser de manière à retrouver essentiellement la même structure et les mêmes fonctions, p.ex. une forêt qui retourne à une forêt après perturbation (Holling 1973; Gunderson 2000). La dynamique par petites trouées et les coupes totales sont des perturbations importantes du couvert forestier (McCarthy 2001; Boucher et al. 2009; Lundmark et al. 2013). Les processus de régénération sont parmi les principaux mécanismes de résilience des forêts parce qu ils déterminent les trajectoires successionnelles et permettent à la forêt de récupérer de la perturbation de son couvert (Buma and Wessman 2011; Hidding et al. 2013). Le broutement préférentiel des semis par les cervidés peut perturber les processus de régénération en changeant la densité et la composition de la régénération préétablie (Tremblay et al. 2007; Nuttle et al. 2014). Les interactions entre la perturbation du couvert et le broutement préférentiel de la régénération pourraient donc altérer la résilience et engendrer de nouvelles trajectoires successionnelles vers des assemblages et des abondances relatives d espèces n ayant jamais coexisté historiquement. L occurrence de tels nouveaux écosystèmes soulèverait des enjeux quant à la conservation de la biodiversité et au maintien des services écosystémiques fournis par la forêt naturelle (Hobbs et al. 2006; Bridgewater et al. 2011). Il devient donc important d identifier d éventuels mécanismes de perte de résilience qui pourraient empêcher la forêt 1

22 naturelle (c.-à-d. préindustrielle) de récupérer des interactions entre la perturbation du couvert et le broutement préférentiel de la régénération par les cervidés. La surabondance de cervidés semble préoccupante dans l est du Canada, notamment autour du golfe du Saint Laurent. En effet, elle a été rapportée dans les forêts de Terre-Neuve (Gosse et al. 2011), de Saint-Pierre et Miquelon (Michallet et al. 2009), de l île du Cap-Breton (Latourelle et Bird 2010) et de la péninsule gaspésienne (Gingras et al. 2014). Au milieu de ce golfe, l île d Anticosti abrite des cerfs de Virginie (Odocoileus virginianus) surabondant qui altère la résilience de la forêt (Potvin et al. 2003; Tremblay et al. 2007; Hidding et al. 2013). Ce contexte crée donc l occasion d identifier d éventuels mécanismes de perte de résilience de la forêt due à l interaction entre la surabondance de cervidés et les perturbations préexistantes. Connaissance de la sapinière préindustrielle et de sa dynamique des perturbations naturelles À la suite de l introduction de 200 cerfs de Virginie sur l île d Anticosti à la fin du 19 e siècle, leur densité dépasse maintenant 20 individus km -2, en l absence de prédateurs (Rochette and Gingras 2007). Depuis les années 1920, le broutement chronique préférentiel par le cerf a changé la composition du sous-étage des sapinières, jadis dominé par le sapin baumier (Abies balsamea (L.) Miller); ce sous-étage est maintenant dominé par l épinette blanche (Picea glauca (Moench) Voss) et d autres espèces pouvant former une strate récalcitrante de sous-étage comme le calamagrostis du Canada (Calamagrostis canadensis (Michauxx) P.Beauv.), la fougère aigle (Pteridium aquilinum (L.) Kuhn) et le chardon (Cirsium spp.), lesquelles ne sont pas broutées par le cerf (Potvin et al. 2003; Tremblay et al. 2007; Hidding et al. 2013). Les sapinières de l île d Anticosti fournissent donc un contexte pour identifier d éventuels mécanismes de perte de résilience de la forêt naturelle due aux interactions entre les perturbations du couvert et le broutement préférentiel de la régénération par les cervidés. Afin de permettre l identification de mécanismes de perte de résilience, il est nécessaire de décrire la variabilité naturelle que la forêt préindustrielle devrait retrouver après perturbation et de comprendre sa dynamique des perturbations naturelles. Ce défi représente l objectif du chapitre 1 de la thèse. Pour les sapinières de l est du Canada, Baskerville (1975) a proposé un modèle cyclique qui s autorégule dans lequel 2

23 les épidémies sévères récurrentes de tordeuse des bourgeons de l épinette (Choristoneura fumiferana (Clem.)) tuent les sapinières matures tout en favorisant leur renouvellement en libérant la banque de semis de sapins baumiers préétablis. Cette dynamique des perturbations produit une mosaïque fine dans laquelle se mélangent une majorité de peuplements équiens d âges et de tailles variés (Leblanc et Bélanger 2000; Belle-Isle et Kneeshaw 2007). Des études récentes dans les régions maritimes de l est du Canada (la péninsule gaspésienne et Terre-Neuve) ont révélé l existence de paysages où les peuplements équiens étaient peu fréquents et distribués dans une matrice de sapinières surannées irrégulières (Dallaire 2004; McCarthy 2004). La sévérité des épidémies d insectes semble avoir été beaucoup plus faible dans ces régions maritimes que dans les régions continentales (Blais 1983; Gray et al. 2000; Williams et Birdsey 2003). À l ouest de Terre-Neuve, Jardon et Doyon (2003) ont décrit l occurrence d épidémies d intensité faible à modérée, susceptibles de créer des peuplements multicohortes, alors que McCarthy (2004) a décrit un paysage constitué de vieilles sapinières où les perturbations sévères étaient rares. Ces travaux mettent en lumière le besoin de conceptualiser un nouveau modèle de dynamique forestière, différent du modèle de la mosaïque cyclique (Baskerville 1975; Leblanc et Bélanger 2000), qui décrirait mieux la dynamique des perturbations des sapinières autour du golfe du Saint- Laurent. Étant située dans le golfe du Saint-Laurent, l île d Anticosti offre un contexte pour vérifier le besoin de développer un nouveau modèle de dynamique forestière pour les sapinières des régions maritimes de l est du Canada. Nous prédisons que la sapinière de l île sera mieux décrite par un nouveau modèle de dynamique forestière où les perturbations sévères sont rares. Acquérir la connaissance sur la forêt préindustrielle et sur sa dynamique des perturbations naturelles représente un défi, non seulement en raison du broutement par le cerf, mais aussi parce que les coupes historiques ( ) et actuelles ont altéré la forêt préindustrielle (Potvin et al. 2003; Beaupré et al. 2004). En l absence de paysages forestiers préindustriels, l utilisation de méthodes reconstructives est reconnue comme une approche intéressante pour évaluer la variabilité des écosystèmes forestiers (Foster et al. 1996; Landres et al. 1999; Kuuluvainen 2002). Ainsi, avec l utilisation de cartes forestières historiques et de la dendrochronologie 3

24 (Frelich et Reich 1995; Weir et Johnson 1998; Axelsson et Östlund 2001; Etheridge et al. 2006), les marges de variabilité de la forêt préindustrielle de l île d Anticosti et la dynamique des perturbations naturelles ont été déterminées. Interactions entre la dynamique des petites trouées et le broutement préférentiel de la régénération En l absence de perturbations sévères récurrentes, les sapinières sont dynamisées essentiellement par des perturbations par trouées suivies de processus de régénération des trouées qui assurent le maintien de peuplements surannés de sapin baumier en mélange avec des espèces compagnes comme l épinette blanche et le bouleau à papier (Betula papyrifera Marshall; McCarthy et Weetman 2007). Le sapin baumier crée des banques de semis denses et persistantes en sous-étage. Les semis de ces banques sont très tolérants à l ombre, prompts à réagir à l ouverture du couvert et à remplir les trouées dans le couvert (Côté et Bélanger 1991; McCarthy 2001; McCarthy et Weetman 2007). En raison du broutement préférentiel par le cerf, l épinette blanche représente la seule espèce d arbre disponible pour remplir les trouées dans les sapinières surannées d Anticosti. L effet des interactions entre les perturbations par petites trouées et le broutement préférentiel de la régénération par le cerf sur les processus de régénération des trouées n a pas encore été évalué. Un tel changement dans la composition de la régénération en sous-étage pourrait interrompre les processus de régénération des trouées (Frelich et Lorimer 1985, Healy 1997, Pedersen et Wallis 2004) engendrant des processus de dégradation du couvert et de nouvelles trajectoires successionnelles. La dégradation du couvert forestier a d importantes répercussions sur les services écosystémiques (Foley et al. 2007; Sasaki et Putz 2009). L occurrence éventuelle de pessières blanches soulèverait des enjeux quant à la conservation de la biodiversité et au maintien des services écosystémiques fournis par la forêt préindustrielle, étant donné que ces peuplements sont des nouveaux écosystèmes qui ne se retrouvent pas habituellement dans l est de l Amérique du Nord (Stiell 1976; Bell et al. 1990; Lieffers et al. 2008; Saucier et al. 2009). L état forestier pourrait ne pas être la seule trajectoire successionnelle induite par les interactions entre les perturbations par petites trouées et le broutement préférentiel par les cerfs. Effectivement, l épinette blanche est une espèce secondaire dans le couvert sur l île alors que sa régénération est 4

25 intermédiaire quant à la tolérance à l ombre (Burns et Honkala 1990). Elle est vulnérable à la compétition par les espèces pouvant former une strate récalcitrante de sous-étage (Hogg et Lieffers 1991; Lieffers et al. 1993; Cole et al. 2013), nécessite des substrats spécifiques pour l établissement optimum des semis (Simard et al. 1998), forme rarement des peuplements purs en condition naturelles (Stiell 1976; Bell et al. 1990) et est considérée comme une espèce secondaire pour remplir les trouées (Pham et al. 2004, Brassard et Chen 2006). De nouveaux écosystèmes tels que des forêts-parcs (c.-à-d. une déforestation partielle), pourraient aussi résulter de cette interaction. L occurrence d une déforestation partielle soulèverait des enjeux encore plus grands quant à la conservation de la biodiversité et au maintien des services écosystémiques. Ainsi, l objectif du chapitre 2 de la thèse est de déterminer si le changement dans la composition de la régénération en sous-étage peut interrompre les processus de régénération des trouées engendrant des processus de dégradation du couvert des sapinières et de nouvelles trajectoires successionnelles, les empêchant de retourner vers leur variabilité naturelle. La première prédiction est que le changement dans la composition de la régénération en sous-étage interrompra les processus de régénération des trouées engendrant des processus de dégradation du couvert. La deuxième prédiction est que le patron spatial de la dégradation déterminera la nouvelle trajectoire de la succession. La dégradation du couvert par petites trouées qui maintiennent la pénombre (c.-à-d. 200 m 2 : McCarthy 2001) engendrera une nouvelle trajectoire successionnelle vers des forêts-parcs d épinettes blanches. Nous prédisons que la régénération d épinette blanche ne sera pas suffisante pour remplir ces petites trouées et recréer un peuplement, parce que l espèce est intermédiaire quant à la tolérance à l ombre, qu elle est vulnérable à la compétition par les espèces pouvant former une strate récalcitrante de sous-étage et qu elle est considérée comme secondaire pour remplir les trouées. Par contre, la dégradation du couvert par plus grandes trouées (> 200 m 2 ) engendrera une nouvelle trajectoire successionnelle vers des pessières blanches. Nous prédisons que la régénération d épinette blanche sera suffisante pour remplir ces trouées et recréer un peuplement parce que l espèce n est pas broutée, qu elle tolère la lumière directe et qu elle a formé des peuplements sur l île d Anticosti après de grandes perturbations sévères (c.-à-d. les coupes totales des années 1920 et l épidémie d insectes des années 1970: Jobin 1980, Beaupré et al. 2004). 5

26 Interactions entre la coupe totale et le broutement préférentiel de la régénération En raison du broutement préférentiel par le cerf, l épinette blanche représente la seule espèce d arbre disponible pour recréer une forêt après la coupe totale des sapinières surannées d Anticosti. Dans un tel contexte, la régénération d épinette blanche a formé des pessières blanches après la coupe totale historique ( ) des sapinières (Beaupré et al. 2004). Ces pessières blanches de seconde venue, maintenant matures, représentent des nouveaux écosystèmes. Plusieurs de ces pessières matures ont récemment été coupées (1999) dans un aménagement équien pour la production de matière ligneuse. Toutefois, l effet des interactions entre la coupe totale et le broutement préférentiel de la régénération par le cerf sur les processus de régénération des pessières blanches n a pas encore été évalué. Si les processus de régénération ne permettent pas aux pessières blanches de récupérer de l interaction de ces deux perturbations, une nouvelle trajectoire successionnelle pourrait mener les pessières vers des forêts-parcs. Ainsi, l objectif du chapitre 3 de la thèse est de déterminer si les processus de régénération permettent aux pessières blanches de récupérer des interactions entre les coupes totales et le broutement préférentiel de la régénération par les cerfs. La régénération a donc été étudiée en relation avec la disponibilité des semences, la disponibilité des substrats d établissement des semis, et la convenance des substrats dans les pessières blanches de seconde venue matures et dans les coupes récentes des pessières blanches de seconde venue matures. Les processus de régénération ont aussi été étudiés en sous-étage et dans les trouées de sapinières surannées pour obtenir une référence des caractéristiques de régénération qui ont permis aux sapinières de récupérer partiellement des coupes historiques et du broutement pour produire des pessières blanches. La première prédiction est que la régénération dans les pessières blanches de seconde venue matures ne sera pas suffisante pour qu elles récupèrent des interactions de la coupe et du broutement, en raison d un manque de substrats adéquats pour l établissement des semis. Effectivement, l épinette blanche nécessite des substrats spécifiques pour l établissement optimal des semis (p.ex. : des gros débris ligneux, le sol minéral exposé; Simard et al. 1998) et elle forme rarement des peuplements purs en condition naturelles (Stiell 1976; Bell et al. 1990). La deuxième prédiction est que la régénération ne sera pas suffisante dans les coupes récentes des pessières blanches de seconde venue matures pour que les peuplements retournent à un état forestier en raison de 6

27 l envahissement par les espèces pouvant former une strate récalcitrante de sous-étage, qui exacerbera ainsi le manque de substrats adéquats pour l établissement des semis. Cette strate inhibe la régénération, en altérant le taux de la succession forestière et changeant sa direction (Royo et Carson 2006), même après la réduction des densités de cerfs (Nuttle et al. 2014). 7

28

29 Chapitre 1 : Preindustrial reconstruction of a perhumid mid-boreal landscape, Anticosti Island, Quebec 1.1. Résumé La connaissance de la dynamique de perturbations naturelles et des paysages préindustriels est essentiel à la réalisation de l aménagement forestier durable. Des travaux récents ont identifié le besoin d un nouveau modèle de dynamique forestière, différent du modèle cyclique standard de Baskerville (1975. For. Chron. 51 : ), pour les sapinières maritimes de l est du Canada. Au moyen de l analyse de cartes forestières historiques et de la dendrochronologie, nous avons reconstruit les limites de variabilité du paysage préindustriel (6798 km²) et inféré sur la dynamique naturelle de perturbations des sapinières de l île d Anticosti. Le paysage préindustriel était caractérisé par une matrice de peuplements résineux surannés avec des inclusions de peuplements résineux plus jeunes, d une superficie variant de 0,1 à 7837 ha. L occurrence de perturbations initiatrices de nouveaux peuplements semble avoir été faible durant les derniers ~160 ans. Étant donné que les résultats ne sont pas bien représentés par le modèle cyclique, qui prédit l occurrence d une mosaïque de peuplements d âges différents, nous proposons un modèle alternatif de dynamique forestière pour les sapinières de l est du Canada localisées près du golfe du Saint-Laurent. L aménagement forestier inspiré par ce modèle alternatif pourrait être plus approprié pour maintenir ou restaurer les caractéristiques écologiques des sapinières de cette région à l intérieur de leurs limites de variabilité naturelle. 9

30 1.2. Abstract The knowledge of natural disturbance dynamics and preindustrial landscapes is essential to implement sustainable forest management. Recent findings identify the lack of a forest dynamics model, different from the standard cyclic model of Baskerville (1975. For. Chron. 51: ), for balsam fir (Abies balsamea (L.) Mill.) ecosystems of maritime eastern Canada. With the use of historical forest maps and dendrochronology, we reconstructed the range of variability of the preindustrial landscape (6798 km²) and inferred on the natural disturbance dynamics of the balsam fir forest of Anticosti Island. The preindustrial landscape was characterized by a forest matrix of overmature softwood stands with inclusions of younger softwood stands, ranging from 0.1 to 7837 ha in size. Widespread stand-initiating events were apparently rare in the preindustrial landscape over the last ~160 years. Since our results were not well represented by the cyclic model, which predicts the occurrence of a mosaic of stands in different age-classes, we proposed an alternative forest dynamics model for eastern balsam fir ecosystems near the Gulf of St. Lawrence. Forest management inspired by this alternative model may be more appropriate to maintain or restore ecological characteristics of balsam fir forests of this region within their range of natural variability. 10

31 1.3. Introduction Understanding natural disturbance dynamics and acquiring knowledge on the range of variability of preindustrial landscapes are important for the restoration or the management of forest ecosystems toward their natural variability (White and Walker 1997; Gauthier et al. 2008; Lindenmayer et al. 2008; Shorohova et al. 2009). In fire-prone boreal ecosystems, such reconstructions have been undertaken (Engelmark 1984; Bergeron et al. 2002; Gromtsev 2002). However, in regions where fire is not the sole ecological driver of forest dynamics, the development of forest dynamics models is still necessary. For balsam fir (Abies balsamea (L.) Mill.) ecosystems of eastern Canada, Baskerville (1975) proposed a self-regulating cyclic model where recurrent widespread stand-initiating spruce budworm (Choristoneura fumiferana (Clem.)) outbreaks kill mature balsam fir stands while promoting new even-aged stands by releasing a pre-established balsam fir seedling bank. This disturbance regime produces a fine-grained landscape mosaic where a majority of even-aged balsam fir stands of different size and age are intermixed (Leblanc and Bélanger 2000; Belle-Isle and Kneeshaw 2007). Recent studies in maritime areas of eastern Canada (Gaspé Peninsula and Newfoundland) revealed the occurrence of natural landscapes where even-aged stands were not frequent and scarcely distributed in a forest matrix of overmature balsam fir stands (Dallaire 2004; McCarthy 2004). There is evidence that the severity of insect disturbances in these maritime areas was much lower than in continental areas (Blais 1983; Gray et al. 2000; Williams and Birdsey 2003). In western Newfoundland, Jardon and Doyon (2003) described the occurrence of light to moderate severity outbreaks, which were prone to create multicohort stands, while McCarthy (2004) described a primary balsam fir ecosystem with rare stand-initiating events. These findings stress the need for a conceptual forest dynamics model that is different from the cyclic mosaic model (Baskerville 1975; Leblanc and Bélanger 2000), and better suited to describe the natural disturbance dynamics shaping balsam fir forests around the Gulf of St. Lawrence. On Anticosti Island, located in the Gulf of St. Lawrence, fires are historically rare events (Lavoie et al. 2009) and primeval balsam fir forests are still present. Primeval balsam fir forests are the prevalent forests of natural origins in which chronic browsing by deer has not altered the development of canopy trees. Hence, the island represents an opportunity to verify, on a large scale, the need for the development of an alternative forest 11

32 dynamics model for balsam fir forests located in the perhumid boreal ecoclimatic region of Canada (Groupe de travail sur les écorégions (GTE) 1989; Fig. 1-1A). Fig (A) Location of the study area by historical forest management blocks and of the dendrochronological study plots;. (B) Reproduction of the map of the ecoclimatic regions of Canada located around the Gulf of St. Lawrence. (GTE 1989). Reproduced with the permission of the Minister of Public Works and Government Services Canada,

33 However, Anticosti Island also holds a very large white-tailed deer (Odocoileus virginianus Zimmermann) population (>20 deer / km² locally; Rochette and Gingras 2007) originating from the introduction of deer between 1896 and Since the 1930s, the population of deer has been important enough for chronic browsing to inhibit the recruitment of balsam fir and deciduous species (Potvin et al. 2003). This browsing pressure threatens to eradicate the remaining primeval balsam fir stands within the next 50 years (Potvin et al. 2003). Consequently, before this forest is lost due to chronic browsing, the general objective of this study is to assess, on a large scale, the past natural disturbance dynamics. We predict that it will be better described by an alternative forest dynamics model than by the cyclic mosaic model. Acquiring knowledge on the natural disturbance dynamics and the range of natural variability was challenging, not only because of deer browsing, but also because historic ( ) and ongoing forest management has altered the preindustrial landscape of Anticosti Island (Potvin et al. 2003; Beaupré et al. 2004). In the absence of preindustrial forest landscapes, the use of reconstructive methods is recognized as a valid approach to assess the natural variability of forest ecosystems (Foster et al. 1996; Landres et al. 1999; Kuuluvainen 2002). Therefore, with the use of historical forest maps and dendrochronology (Frelich and Reich 1995; Weir and Johnson 1998; Axelsson and Östlund 2001; Etheridge et al. 2006), we aim to reconstruct the range of variability of the preindustrial landscape of Anticosti Island and infer on the natural disturbance dynamics. These results and their interpretations will be discussed in the context of conservation of forest ecosystems Materials and methods Study area Anticosti Island (7943 km²) is located in the Gulf of St. Lawrence (49 28 N, W) (Fig. 1-1) and is part of the perhumid mid-boreal ecoclimatic region of Canada (GTE 1989). The island is part of the eastern balsam fir paper birch (Betula papyrifera Marsh.) bioclimatic subdomain of Quebec (Saucier et al. 2009). Mean annual temperature is 2 C and mean annual precipitation is 907 mm (Environment Canada 1982). Large-scale natural disturbances comprise fire, blowdowns, and spruce budworm and hemlock looper 13

34 (Lambdina fiscellaria (Guen.) outbreaks (Jobin 1980; Martel 1999; Chouinard and Filion 2005). For the analysis of the historical data, the study area (6798 km²) included 14 historical forest management blocks (Figs. 1-1A and 1-2). Sampling of dendrochronological study plots was restricted to the western half of the island where the concentration of primeval balsam fir stands was highest Historical data To describe the structure and composition of the preindustrial landscape,14 historical forest maps (mean area = 494 km², 1: scale) (Figs. 1-1A and 1-2) of stand age class were analysed (Royer 1956a). The corresponding summary sheets of stand area records were divided by age class and stand type (Anctil 1957). Only the historical maps that had not experienced industrial logging for timber or pulpwood prior to their creation were included in the analyses (Fig. 1-1A) (Royer 1956a, 1956b; Lejeune and Dion 1989). The spatial distribution of logging operations was verified using historical maps of roadways and cutover areas (Royer 1956b). Each forest map corresponded to a management block, which was subdivided into management compartments (n = 305, mean area = 23 km²). Delimitation of the historical management blocks and compartments reflects the boundaries for the watersheds of the main rivers on Anticosti Island. Accordingly, the management blocks were designated by the names of these rivers. The historical maps included seven age classes groups that corresponded to successional stages (0 burn, 1 20, 21 40, 41 60, 61 80, , and 101 years) and three non-forest cover types (barren, muskeg, and water). Stands in ages classes 100 years had regular and homogenous stand characteristics (e.g., density and height: Ronald Dixon, personal communication (2006); Luc Généreux, personal communication (2009)). Hence, these younger stands were most probably generated by a main stand-replacing disturbance that occurred <100 years before the production of the historical maps. However, it was not specified how much older than 100 years were stands in ages classes 101 years. These older stands most probably have experienced secondary disturbances, since break-up of balsam fir stands begins at around the age of 90 years (McCarthy and Weetman 2007). In this region, insect defoliation can result in stand-replacing 14

35 Fig Scanned historical forest map of Shallop Forest management block with an expanded view of the legend. 15

36 disturbances that release pre-established balsam fir and white spruce (Picea glauca (Moench) Voss) regeneration (ages classes 100 years) or result in moderate to light disturbances that contribute to shape multicohort stands (age classes 101 years) (Jardon and Doyon 2003; McCarthy 2004; McCarthy and Weetman 2007). The three stand types found in the summary sheets were softwoods ( 75% softwood species), mixedwoods (26% 74% softwood), and hardwoods ( 75% hardwood). Consolidated Paper Corporation Ltd. originally made these forest maps for their general forest management plan of 1956 produced for all of their forest concessions. The plans were made to describe the existing forest landscape and determine the sustainable annual allowable cut to support the company s 10-year pulpwood cutting programs (Bubie 1958). The stand age class maps (1956) were generated following photointerpretation of aerial photographs (1:9600 scale) taken in Information gathered on field surveys, carried out in 1953 for the western part of the island and in 1956 for the eastern part, was used to help determine and validate stand age classes on the aerial photographs (Fortin 1954; Royer and Grondin 1960). In the process of photointerpretation, the main topographical features, such as lakes, rivers, creeks, heights of land, cliffs, etc., were outlined on the photographs. The photographs were then interpreted and the contours of forest stands, recent windfall, barren, recent burns, etc., were outlined on the photographs. Forest stands were photointerpreted for forest type and age and density classes. All of the information extracted from the photographs regarding topography and forest conditions was then transferred to the base map with the sketchmaster prior to the field operations (Royer and Grondin 1960). Up to 2% of the area of the compartments was sampled (Ronald Dixon, personal communication (2006); Luc Généreux, personal communication (2009)) in 1/5 acre (809 m²) rectangular cruise plots that were each 6 chains (120.7 m) long by 1/3 chain (6.7 m) wide (Fortin 1954). In each plot, the diameter at breast height (1.3 m) (DBH) and species of all trees with DBH 4 in. (10.2 cm) were tallied and cored at DBH for age determination (Fortin 1954). According to Fortin (1954), these plots have been laid out at preselected locations at irregular intervals along cruise lines of irregular bearings to provide representative tree tallies in a weighted distribution of forest stands for forest type and age class according to their importance. Information on the area that had been burned by a major fire, which occurred in 1955, was also available, since the contours of the fire were 16

37 superimposed on the stands that had already been mapped from the aerial photographs of 1945, the oldest set from the island (Luc Généreux, personal communication (2007)). The historical forest maps (Royer 1956a, 1956b), the corresponding summary sheets (Anctil 1957), and the explanatory documents (Fortin 1954; Bubie 1958; Royer and Grondin 1960) were all found in the private archives of Abitibi- Bowater (Grand-Mère, Québec). The usual high quality of historical forest data produce by forest companies has been recognized by many authors (Weir and Johnson1998; Axelsson and Östlund 2001; Etheridge et al. 2006). For instance, Boucher et al. (2006) used additional information (aerial photographs of 1941) to validate older historical forest maps (1930) and found a high similarity (77% 95%) between the two sources. In our case, we did not find additional information allowing further validations of the historical forest maps Landscape analysis Age structure and composition of the preindustrial landscape were described using stand age class by stand type distribution. To illustrate the variability in the occurrence of each stand age class and type, we also determined their relative abundance at two smaller scales, i.e., in each historical block and compartment. Based on the variability observed at the scale of the 305 compartments, we calculated ranges of natural variability for each stand age class and types (Barrette and Bélanger 2007). To describe the spatial structure of the preindustrial landscape, we georeferenced the scanned historical maps and digitized in vector format all polygons, which corresponded to stands belonging to age classes between years, using ARCMAP (ARCGIS version 9) (ESRI 2004). All polygons were analysed by age and size class with Patch Analyst version 3 (Elkie et al. 1999; Etheridge et al. 2006). In the ARCVIEW 3.1 interface (ESRI 1998), Patch Analyst produces spatial statistics generated by FRAGSTATS (McGarigal et al. 2002). We define a patch as a continuous polygon, which represents a stand of 100 years old or less, as drawn on the historical maps independently of the boundary of the maps. We generated one area metric, three patch density and size metrics, one shape metric, and two interspersion metrics. Interspersion metrics were calculated from a grid theme, with a pixel size of 1 ha, which was obtained from the conversion of the original vector theme. We used 17

38 the interspersion juxtaposition index to determine if patches were distributed evenly or unevenly in the landscape. To better illustrate area, patch density, and size metrics, we produced the distribution of number and area of patches by 100 ha size classes Field sampling and laboratory procedures To describe the natural disturbance dynamics, a field survey was conducted in the remaining primeval balsam fir stands of the island for subsequent dendroecological disturbance reconstruction (Fig. 1-1A). To select the stands, from all stands preliminary numbered, a stratified random sampling method was used. The method was based on age and composition of stands as well as on sector of study area, equally dividing the length of the western half of the island. Sampling of dendrochronological study plots was restricted to the western half of the island, where the highest concentrations of primeval balsam fir stands (age classes 90 years) could be found, according to the latest numerical forest maps of the Ministère des Ressources naturelles et de la Faune du Québec (MRNFQ). These maps were made following the interpretation of aerial photographs taken in We sampled mature and overmature stands to better link our field data to the two oldest historical stand age classes ( years, n = 2; >100 years, n = 3) and also because these stands allowed the longest dendrochronological reconstruction. Moreover, mature and overmature stands are the only ones on the island still containing canopy trees that have grown without the effect of deer browsing. A rectangular plot (400 m², 50 m 8 m) was positioned along a bearing that ran across the longest dimension of each sampled stand. In each plot, we felled all trees of DBH >9.1 cm. Cross sections from the boles of all live and dead trees were sampled at ground level (0.1 m), 0.5, 1.3, and 2 m and at each subsequent 1 m in the entire area of the plots older than 100 years and over half the area (50 m 4 m) of the 81- to 100-year-old plots. The cross sections were dried and finely sanded. They were then cross-dated using two radii from the lowest cross- section of each tree and one radius from higher cross sections with the pointer-year method (Yamaguchi 1991). We cross-dated our ring width series against known pointer years determined by Chouinard and Filion (2005) from samples 18

39 collected on Anticosti Island. We measured the tree-ring widths on the two radii from the lowest cross section with a Velmex (Bloomfield, New York) micrometer (0.001 mm precision) connected to a computer and validated the cross-dating with the program COFECHA (Holmes 1983) Dendrochronological analysis Since hemlock looper and spruce budworm outbreaks were identified as an important disturbance factor of the forest landscape of Anticosti Island (Jobin 1980; Martel 1999; Chouinard and Filion 2005), we used dendrochronology methods to identify tree-ring growth patterns that could be associated with insect defoliation within the five primeval balsam fir stands that we sampled. To remove trends in the radial growth patterns that were related to tree age and site, we standardized the raw tree-ring chronologies with the program ARSTAN (Cook and Holmes 1999). In ARSTAN, we fitted a cubic spline function with a length of 60 years to the data with a 50% frequency response cutoff (Bouchard et al. 2006). We produced a mean standardized chronology from all balsam fir (n = 251 radii from 140 trees) and white spruce (n = 14 radii from eight trees) of the five plots pooled together by averaging the tree-ring index chronologies using a biweighted robust mean (Cook and Holmes 1999). We also produced mean standardized chronologies by plot for balsam fir. Before standardization, we removed saplings and tree-ring series with a low correlation from the analysis to optimize the detection of landscape-scale disturbances. To identify tree-ring growth patterns that could be associated with insect defoliation within the time span of the historical data ( ), we compared our host (balsam fir and white spruce) mean standardized chronologies with a non-host (jack pine (Pinus banksiana Lamb.)) chronology. We constructed the non-host chronology from 11 jack pines in the same fashion as for the host species. The jack pine samples were randomly sampled in the Quebec North Shore region (50 00 N, W), located ~250 km from Anticosti Island, by Bouchard and Pothier (2010). We chose jack pine as a nonhost species because its growth is rarely affected by spruce budworm (Simard and Payette 2001), while it has a 19

40 response to climate similar to that of balsam fir and white spruce (Tardif et al. 2001). Moreover, 80% of the jack pines samples grew in the same ecoclimatic region as the balsam fir and white spruce sampled on Anticosti Island (MBp) (Fig. 1-1B). Accordingly, balsam fir and jack pine chronologies had a year to year agreement of 62% (Gleichlaufigkeit index of similarity) when periods of growth reduction potentially caused by insect defoliation were removed. This correlation indicated that radial growth of host and non-host species reacted similarly to climatic conditions (Jardon et al. 2003; Chouinard and Filion 2005). We used the host versus non-host chronologies to visually distinguish U- or V-shaped growth reductions in the host chronologies that were potentially caused by insect defoliation from those caused by climatic factors (Jardon et al. 2003). We further analysed tree-ring growth reductions using the program OUTBREAK (Holmes and Swetnam 1996) with the jack pine chronologies as a control. According to Martel (1999), tree mortality during a hemlock looper outbreak can occur after only 1 3 years of defoliation. Hence, to register growth reductions possibly caused by defoliation from hemlock looper or spruce budworm within OUTBREAK, we interpreted a significant growth reduction in the host chronologies when it lasted 3 years and included at least 1 year with an index value more than 1.28 standard deviations below the mean (Bouchard et al. 2006). We interpreted important growth reduction periods from the proportion of trees with significant growth reduction during a given 10-year period and chose the year where the growth was lowest as a reference for the growth reduction period (Bouchard et al. 2006) Results Preindustrial landscape structure and composition A matrix of stands that were >100 years old occupied 75% of the forest area or 63% of the total area in the preindustrial landscape (Fig. 1-3). Recent burns and stands in the 81- to 100-year-old age class both occupied 10% of the forest area, while stands in the other age classes collectively covered <5%. The resulting stand age class structure followed a positive exponential distribution (Fig. 1-3). Variability analysis revealed that stands >100 years old could occupy 5% 100% of the total area of the historical management 20

41 compartments (n = 305 watersheds, mean area = 23 km²), but they covered >60% of the area in 74% of the compartments (Fig. 1-4). Variability at the scale of historical management blocks (n = 14 watersheds, mean area = 494 km²) was lower and stands >100 years old always covered >50% of the total area, except for the Potatoe and Wickenden management blocks burned in part by the 1955 fire (Fig. 1-5). The occurrence of stands of years old also demonstrated variability; however, 85% of the compartments had <25% of their total area covered by stands in this age class. Stands in the other age classes (0 burn, 1 20, 21 40, 41 60, and years) showed little variability at the management compartment scale and all had steeply negative exponential distributions with a median of zero (Fig. 1-4). Muskegs occupied 13% of the study area, while water and barrens, respectively, covered 3% and <1% Area (ha) Recent Burn Hardwood Mixedwood Softwood burn Age class (years) Fig Stand age class distribution by type in the area covered by 14 historical management blocks of Anticosti Island, Quebec, Canada. 21

42 years years Number of historical management compartments Median = years Median = Median = Median = years years Median = Median = years 1955 fire Proportion of total area (%) Proportion of total area (%) Fig Distribution of the relative abundance of each stand age class in 305 historical management compartments (mean size: 23 km²) of Anticosti Island, Quebec, Canada. 22

43 A matrix of stands of the softwood type occupied 83% of the forest area or 70% of the total area (Fig. 1-3). Stands of the mixedwood type covered 6% of the forest area, while stands of the hardwood type were rare and primarily found within the 61- to 81- years-old age class. At the scale of the historical management compartment, variability in the relative abundance of stands of softwood and mixedwood types was moderate (Fig. 1-6). Stands of the softwood type occupied >70% of the total area in 76% of the historical management compartment, while stands of the mixedwood type covered <20% of the total area in 90% of the compartments (Fig. 1-6). Block area (%) Huile Jupiter McDonald Potatoe Wickenden Vauréal Galiotte Cap Pavillon Natiscotek Salmon Shallop Bell Fox Non forest 1955 fire Fig Proportion of the total area by age class in 14 historical management blocks (mean size: 494 km²) of Anticosti Island, Quebec, Canada. Based on the variability observable at the scale of the historical compartments, we calculated ranges of variability for each stand age class and type (Table 1-1). The range of variability included 66% of the observations and was centred on the median relative abundance (17th to 83rd percentiles) to better describe the skewness of the distribution of the 305 observations (Figs. 1-4 and 1-6). We interpret the median relative abundance as the natural tendency of the observations within the range of natural variability. The range of natural variability for overmature stands (>100 years old) and that for stands of the softwood type were both positioned above a threshold of 55% of the total area, while 23

44 ranges for all other stand age classes or types were found below a threshold of 15% of the area (Table 1-1) Softwood 250 Mixed Wood Number of historical management compartments Median = Median = 0.0 Hardwood Median = Proportion of total area (%) Proportion of total area (%) Fig Distribution of the relative abundance of each stand type in 305 historical management compartments (mean size: 23 km²) of Anticosti Island, Quebec, Canada. 24

45 Table 1-1. Range of natural variability and natural tendency for each stand age class and type calculated from the variability observable at the scale of the historical management compartments used to describe the preindustrial landscape of Anticosti Island, Quebec, Stand age-class (years) Proportion of total area (%) Range of natural variability * Overmature (> 100) Mature (61-100) Young (21-60) - 0 Regeneration (1-20) Stand type Softwood Mixedwood Hardwood - 0 Natural tendency * 66% of the observations, centred on the median relative abundance. Median relative abundance Landscape metrics and distributions of patches of young stands in the overmature forest matrix The distribution of the number of patches (i.e., continuous polygons representing stands 100 years old) in the overmature forest matrix followed a negative exponential curve, with 99% of the patches <1000 ha in size and 94% of the patches <100 ha (Fig. 1-7; Table 1-2). The distribution of patch areas would have roughly followed the same curve if we had removed an outlier patch ( ha) representing most of the area burned by the 1955 fire (Fig. 1-8; Table 1-2). If we include patches from the 1955 fire, patches that were 1000 ha occupied 11% of the total area, while patches ranging from 100 to 999 ha and ones from 0 to 99 ha covered 5% of the area. When calculated by patch age class, patches of the 81- to 100-year-old age class were the most abundant and comprised 59% of the total number of patches and 44% of the total area covered by patches (Table 1-2). When calculated by patch size class for all patches except those of the 1955 fire, large patches had a more complex 25

46 shape than did smaller patches (Table 1-2). In contrast, patches of the 1955 fire, which was mainly composed of one huge patch, had a simple shape. Finally, within a time frame of 100 years, the spatial distribution of large patches was irregular within the overmature forest matrix and the mean distance to the nearest neighbouring patch in the same size class was almost 10 km (Table 1-2). Patches within smaller size classes were nearer to one another and were distributed more evenly across the landscape. Number of patches (a) Mean = 62 ha Median = 9 ha (b) Mean = 16 ha Median = 8 ha Age class (years) burn Patch size (ha) Fig (A) Distribution of the number of patches by class of patch size of 100 ha for stands in ages classes 100 years. (B) Expanded view showing the distribution of the number of patch by class of patch size of 1 ha for patches 100 ha. 26

47 Table 1-2. Landscape metrics by age class and size class of patches describing the preindustrial landscape of Anticosti Island, Quebec, Canada Age-class (years) Area (ha) Patch size (ha) Shape Interspersion No. of MNN IJI patches Mean ± SD Range AWMSI (m) (%) ± ± ± ± ± fire ± Size class (ha)* ± ± ± ± ± * Includes only patches of the 1-20, 21-40, 41-60, and years age-classes. Calculated from a grid theme obtained from the conversion of the original vector theme. Calculated from the agglomeration of 3 of the 8 patches, less than 150 meters distant from each other. AWMSI: Area Weighted Mean Shape Index. AWMSI = 1 when all patches are circular. MNN: Mean Nearest Neighbour Distance. IJI: Interspersion Juxtaposition Index. IJI approaches zero when patch adjacency becomes uneven and 100 when all patch types are equally adjacent. 27

48 60000 Area (ha) Age class (years) burn Patch size (ha) Fig Distribution of the area covered by class of patch size of 100 ha for stands in ages classes 100 years Growth reduction periods within the time span of the historical data Using the mean standardized host versus non host chronologies, we identified seven growth reduction periods that could be associated with insect defoliation within the time span of the historical data ( ). These growth reduction periods occurred around 1854, 1880, 1894, 1905, 1915, 1928, and 1936 (Fig. 1-9). The seven growth reduction periods may coincide with the 81- to 100-, 61- to 80-, 41- to 60-, 21- to 40-, and 1- to 20-year-old historical age classes, which, respectively, occupied 10%, 2%, 0.4%, 0.1%, and 1% of the forest area (Fig. 1-3; Table 1-2). The intensity of the growth reductions seems to have been variable but generally of moderate to low intensity. Furthermore, growth reductions were mostly localized, since they were generally not observed in all plots. The growth reduction period of 1936 was observed in the chronology constructed from the trees of the five plots pooled together in which it affected 76% of all balsam fir trees (n = 140) and 88% of all white spruce trees (n = 8) (Fig. 1-9). This growth reduction period was visible in the 28

49 individual balsam fir chronologies of plots 1, 2, 4, and 5 (Fig. 1-9A). The growth reduction period of 1928 was apparent only in the white spruce chronology that had been constructed from all trees pooled together and affected 50% of them (Fig. 1-9A). The growth reduction period of 1915 was visible only in the balsam fir chronology where it affected 15% of all trees and was perceptible in the individual chronologies of plots 1, 3, and 4. The growth reduction period of 1905 was also only observable in the balsam fir chronology where it affected 75% of all trees and was apparent in all individual balsam fir chronologies except for that of plot 2 (Fig. 1-9A). The growth reduction period of 1894 was visible in both the balsam fir and white spruce chronologies where it, respectively, affected 57% and 25% of all trees. The 1894 growth reduction was mainly apparent in the individual chronologies of plots 1 and 5. The growth reduction of 1880 was also visible in both balsam fir and white spruce chronologies where it, respectively, affected 46% and 50% of all trees. This growth reduction period was noticeable in plots 2 and 5. Finally, the growth reduction of 1854 was again apparent in both host chronologies where it affected 75% of all balsam fir trees and 60% of all white spruce trees. However, only the individual balsam fir chronology of plot 5 dated back far enough to encompass this period (Fig. 1-9A). 29

50 1.5 (A) Balsam fir n= n= n=15 Tree-ring index n=33 n= % affected (B) White spruce Sample depth % affected Sample depth Fig Mean standardized tree-ring chronologies of host species (black lines for (aa) balsam fir ( ) and (bb) white spruce ( )) and non-host species (grey line for jack pine ( )), with associated histogram illustrating the proportion of stems showing significant growth reductions and sample depth. Grey bands indicate significant growth reduction periods. Percentages in the grey bands indicate the proportion of stems showing a significant growth reduction in each period for each plot.

51 1.6. Discussion Preindustrial landscape composition, spatial structure, and natural disturbances A forest matrix of overmature softwood stands with spaced inclusions of younger softwood stands, which ranged from 0.1 to 7837 ha in size, characterized the preindustrial landscape of Anticosti Island. Even if the species composition and age structure of stands were not specified in our historical data, we interpret that most overmature softwood stands were balsam fir stands (Potvin et al. 2003) of multicohort age structure, since stand breakup of balsam fir stands usually begins at around 90 years of age (McCarthy and Weetman 2007). Although growth reductions resulting from insect disturbances probably occurred, the landscape structure that we have described for Anticosti Island is not well represented by the cyclic model (Baskerville 1975), which is usually recognized to illustrate the disturbance dynamics taking place in the eastern balsam fir ecosystem (MacLean 1984; Morin 1994). According to the standard cyclic model, recurrent and widespread standinitiating spruce budworm outbreaks generate a landscape with a mosaic-type spatial structure where stands of different age and size ( 85 ha) are intermixed (Leblanc and Bélanger 2000; Belle-Isle and Kneeshaw 2007). Hence, our prediction was apparently verified, and therefore, we propose an overmature matrix forest dynamics model that would be better suited to describe the natural disturbance dynamics shaping perhumid balsam fir ecosystems located around the Gulf of St. Lawrence (Fig. 1-1B). According to this model, widespread stand-initiating insect outbreaks would be rare events, which would allow the occurrence of a matrix of overmature balsam fir stands. Stand-initiating disturbances would generate spaced inclusions of younger even-aged balsam fir stands in the matrix. Finally, the frequent occurrence of light to moderate insect outbreaks would contribute to shaping a multicohort structure in the balsam fir stands (Jardon and Doyon 2003) of the overmature matrix. Concordantly with the occurrence of a matrix of overmature softwood stands, widespread stand-initiating events did seem to have been rare events in the preindustrial landscape of Anticosti Island over the last ~160 years. From 1845 to 1957, 25% of the forest area was affected by stand-initiating disturbances, which corresponds to a yearly stand- initiating 31

52 disturbance rate of 0.22%, but only of 0.13% if we do not consider known fire disturbances. Based on our dendrochronological data, we determined the occurrence of seven growth reduction periods that could have been associated with insect defoliation around 1854, 1880, 1894, 1905, 1915, 1928, and However, according to our historical data, these probable insect outbreaks could not have initiated new stands on more than 15% of the forest area, which corresponds to the area covered by the historical age classes of 1 20, 21 40, 41 60, 61 80, and years. The seven growth reduction periods that we have recorded have been associated with insect outbreaks in other studies in regions bordering the Gulf of St. Lawrence. The occurrences on Anticosti Island of a severe hemlock looper outbreak (1320 km²) around 1936 and a moderate one (314 km²) between 1923 and 1928 have already been documented (Jobin 1980; Martel 1999; Chouinard and Filion 2005). In contrast, and according to our historical data, the 1936 hemlock looper outbreak did not seem to have been severe, since the corresponding historical age class (1 20 years) covered only 1% of the forest area. Before the 1920s, no other insect outbreak of any type had been identified on Anticosti Island. Nevertheless, a growth reduction period that occurred around 1915 was associated with a spruce budworm outbreak of moderate severity on the Quebec North Shore (De Grandpré et al. 2004; Bouchard and Pothier 2010) and of light severity in the Gaspé Peninsula (Blais 1983). It also was associated with a moderate hemlock looper outbreak in Newfoundland (Jardon and Doyon 2003). Regionally, a growth reduction period occurring around 1905 was recorded in Newfoundland (Jardon and Doyon 2003) and on the Quebec North Shore (De Grandpré et al. 2004). Jardon and Doyon (2003) observed a growth reduction period around 1892 and 1882 in Newfoundland, while Bouchard and Pothier (2010) associated a growth reduction period around 1880 to a spruce budworm outbreak on the Quebec North Shore. Similarly, the 1854 growth reduction period on Anticosti Island was also observed by Jardon and Doyon (2003) in Newfoundland. According to our historical data, only the 1854 growth reduction period seems to correspond to a widespread stand-initiating insect outbreak, since stands in the historical age class of years (i.e., originating between 1845 and 1864) covered 10% of the forest area, while stands of historical age classes of 1 20, 21 40, 41 60, and years pooled together covered <5% of the forest area. If we include a more recent period ( ), no other insect outbreak event comparable in severity with the one of

53 occurred on Anticosti Island before the hemlock looper outbreak of 1974 (Jobin 1980; Martel 1999). The 1974 outbreak initiated new stands on about 12% of the forest area (Beaupré et al. 2004). This gap of ~120 years, between the 1854 disturbance and the 1974 outbreak, was similarly described by Blais (1983) for the Quebec North Shore where the widespread budworm outbreak of the 1970s was the first one of such severity for a period of at least ~130 years. On a regional scale, the colder climate of the perhumid midboreal ecoclimatic region (GTE 1989) may explain the rare occurrence of widespread stand-initiating insect outbreaks within the balsam fir paper birch forest of Anticosti Island (Blais 1968; Candau and Fleming 2005). Cool wet weather can compromise insect development and contribute to population collapse (Blais 1968; Gray et al. 2000). In the perhumid midboreal ecoclimatic region, there is one less month without frost and the mean annual temperature is 3.6 C lower than in the perhumid low-boreal ecoclimatic region (GTE 1989). In the latter region, which is farther from the Gulf of St. Lawrence, widespread stand-initiating insect outbreaks are not rare events and the standard cyclic model applies. The forest matrix of overmature balsam fir stands described by Dallaire (2004), which partly inspired our overmature matrix model, was located in the perhumid low-boreal ecoclimatic region. However, the study sites of Dallaire (2004) were located at high altitudes ( m) where the climate is colder, which seemed to have contributed to lowering the severity of spruce budworm outbreaks (Dansereau 1999). The colder climate of the perhumid midboreal ecoclimatic region (GTE 1989) may also explain the higher outbreak recurrence of hemlock looper compared with spruce budworm on Anticosti Island since the 1920s (Martel 1999; Chouinard and Filion 2005). Hence, the growth reduction periods that we have associated with the occurrence of insect defoliation before the 1920s may be attributable to hemlock looper instead of spruce budworm. Hemlock looper infestations are generally associated with regions with maritime climates (Jobin 1980), while the climate of Anticosti Island could be considered marginal for the occurrence of recurrent widespread stand-initiating spruce budworm outbreaks (Blais 1985; Gray et al. 2000). Growth reduction periods in paper birch chronologies, which is a host of hemlock looper but not of spruce budworm, from Anticosti Island (Martel 1999) and from Newfoundland (Jardon and Doyon 2003) 33

54 occurred synchronously with the growth reduction periods that we have identified around 1854, 1880, 1893, and In both of our data sources, we were not able to distinguish disturbances specifically caused by blowdown alone. Considering that balsam fir stands are more vulnerable to blowdowns following insect outbreaks (Leblanc and Bélanger 2000), it is inevitable that some of the 0- to 100-year-old stands found in the historical maps originated from blowdown alone or in interaction with insect defoliation. Stand-replacing blowdowns do occur on Anticosti Island, notably in 1996 when a blowdown disturbed 6070 ha of forest in patches ranging from 0.1 to 171 ha in size. In the eastern region of Gaspé Peninsula, blowdowns were the determinant natural disturbance shaping the preindustrial landscape (Lévesque 1997), while they are second in importance to insect outbreaks in Newfoundland (Thompson et al. 2003). A forest matrix of overmature softwood stands and a forest age class structure following a positive exponential distribution suggest the occurrence of a long fire cycle (Bergeron and Dansereau 1993). Furthermore, the area occupied by small (1 99 ha) and medium-sized inclusion patches ( ha) was equivalent to the area occupied by large inclusion patches ( 1000 ha), which contrasts with a fire-mediated landscape where the vast majority of the disturbed area is burned by a few large fires (Johnson et al. 1998). Lavoie et al. (2009) proposed that fires were rare local events on Anticosti Island during the Holocene. Accordingly, the fire cycles that were determined in regions near our study area ranged from 250 to >500 years (Foster 1983; Lévesque 1997; Bouchard et al. 2008). Moreover, the fact that the younger inclusions in the forest matrix were mainly stands of the softwood type is a strong indication that the landscape was generally perturbed by disturbance types other than fire. Effectively, fire is more likely to kill advance conifer regeneration and promote the establishment of hardwood pioneer species (Bergeron and Dansereau 1993) in balsam fir forests. 34

55 In conclusion, our methodological approach can also be useful for the reconstruction of the preindustrial variability of other ecosystems. Notably, we have exemplified the use of historical forest data coming from pulp and paper forest companies to describe preindustrial ecosystems on wide areas coupled with the use of dendrochronological data of actual stands that have experienced few or no logging to cross-validate the historical records Implications for conservation of forest ecosystems Forest management inspired by the cyclic model will generate an important conservation issue in balsam fir ecosystems located around the Gulf of St. Lawrence by reducing the occurrence of overmature primeval balsam fir stands below their range of natural variability. This issue is magnified on Anticosti Island, since the remaining primeval balsam fir forest, an essential winter deer habitat on the island, may be eradicated in <50 years because of chronic deer browsing (Potvin et al. 2003). To preserve moderate to high deer densities needed for high-quality hunting, an island-wide deer habitat restoration plan, inspired by the standard cyclic model plan, has been implemented (Beaupré et al. 2004). According to the restoration plan, around 150 fenced cut-blocks (3 15 km²) are planned over the next 70 years, which should generate a landscape spatial structure of mosaic type where even-aged balsam fir stands of different size and age are intermixed (Beaupré et al. 2004). However, following the new knowledge acquired from our study, it may appear more appropriate to review the deer habitat restoration plan and to base it on the overmature matrix model to restore balsam fir forests toward their range of natural variability. Moreover, the implementation of the restoration plan, inspired by the cyclic model, accelerates the deer-induced eradication process of the primeval balsam fir forests because of the clearcutting rate contributing to the inversion of the overmature matrix to one constituted by young stands. Hence, the occurrence of overmature primeval balsam fir stands below their range of natural variability generates an important conservation issue because they are distinct from silviculturally mature second-growth balsam fir stands (Thompson et al. 2003; Desponts et al. 2004). Overmature primeval balsam fir stands are a distinct ecosystem because they represent an ecological entity with a specific structure, and function that results in the occurrence of a higher species diversity of mosses, liverworts, lichens, and saprophytic fungi (Thompson et al. 2003; Desponts et al. 2004). To meet the 35

56 criteria of sustainable forest management for the conservation of biological diversity (Food and Agriculture Organization 2001), the restoration plan has to address the conservation issue of overmature primeval balsam fir stands. The retention of 30% of habitat could be viewed as a prudent threshold to minimize local extinction of specialized species (Fahrig 2003). In this regard, we recommend maintaining 22% of the forest area (30% of the natural tendency of overmature stands, i.e., 73%) in patches of overmature primeval balsam fir stands. These forest refuges should be large enough and of a form that minimizes edge effects to constitute a core area of forest interior habitat able to maintain or restore the diversity and abundance of species usually harboured by overmature primeval stands (Desponts et al. 2004; Ries et al. 2004). Since the deer habitat restoration plan aims to maintain a moderate to high deer density, it is probably necessary to fence these refuges of overmature primeval stands permanently within a network of small-scale protected areas. Forest dynamics on Anticosti Island share some similarities with maritime moist forests of boreal Eurasia less prone to fires and where small-scale disturbances can be a major ecological driver (Gromtsev 2002; Caron et al. 2009; Shorohova et al. 2009). For instance, both Shorohova et al. (2009) and Caron et al. (2009) acknowledged the existence of forests in quasi-equilibrium driven by continuous small-scale disturbances and recognized these natural dynamics as a major ecological driver of pristine Eurasian boreal forests. Such matrix disturbances, as conceptualized by Lewis and Lindgren (2000), slowly modify the forest matrix over a long time period. In our study, we have proposed that matrix disturbances shapes 75% of the forest area on Anticosti Island. Caron et al. (2009) specified that quasi-equilibrium forest dynamics could be destabilized by infrequent (~200 years) large-scale disturbances, which is also similar to the occurrence of widespread standinitiating insect outbreaks on Anticosti Island (~120 years). The importance of insect disturbances is, however, less explicit in boreal Eurasian studies, possibly in part because knowledge on their influence on forest landscapes is still in acquisition (Gromtsev 2002; Shorohova et al. 2009). In conclusion, the overmature matrix model that we have proposed could also have management implications in parts of maritime moist forests, less prone to fires, of boreal Eurasia (Gromtsev 2002; Caron et al. 2009; Shorohova et al. 2009). 36

57 1.7. Acknowledgements We wish to thank C. Boyau, M. Veilleux-Nolin, F. Lebel, and C. Marquilly for field and laboratory assistance. We also thank A. Delwaide and M. Bouchard for guidance in the dendrochronological analysis together with the loan of jack pine non-host chronologies from M. Bouchard. This study was supported by the Natural Sciences and Engineering Research Council of Canada Produits forestiers Anticosti Inc. Industrial Chair, the Ministère des Ressources naturelles et de la Faune du Québec, the Ordre des ingénieurs forestiers du Québec, and the Foundation de la faune du Québec. We are thankful to William F.J. Parsons for English revision as well as to Jean Huot, Jean-Pierre-Tremblay, Steeve Côté, and Pierre Beaupré for enlightening discussions. 37

58 1.8. References Anctil, L Summary of stand records. Anticosti Management Unit No. 16. Forestry Department. Consolidated Paper Corporation Ltd., Grand-Mère, Qué. Axelsson, A.L., and Östlund, L Retrospective gap analysis in a Swedish boreal forest landscape using historical data. For. Ecol. Manage. 147(2 3): doi: /S (00) Barrette, M., et Bélanger, L Reconstitution historique du paysage préindustriel de la région écologique des hautes collines du Bas- Saint-Maurice. Can. J. For. Res. 37(7): doi: /X Baskerville, G.L Spruce budworm: super silviculturalist. For. Chron. 51: Beaupré, P., Bédard, C., Dufour, C., Gingras, A., Malenfant, C., and Potvin, F Plan général d aménagement intégré des ressources du milieu forestier de l île d Anticosti. Produits forestiers Anticosti, Québec. Available from carrefour.cegep-baie-comeau. qc.ca/plans_amenagement/ile_anticosti/index.php [accessed 30 September 2008]. Belle-Isle, J., and Kneeshaw, D A stand and landscape comparison of the effects of a spruce budworm (Choristoneura fumiferana (Clem.)) outbreak to the combined effects of harvesting and thinning on forest structure. For. Ecol. Manage. 246(2 3): doi: /j.foreco Bergeron, Y., and Dansereau, P.R Predicting the composition of Canadian southern boreal forest in different fire cycles. J. Veg. Sci. 4(6): doi: / Bergeron, Y., Leduc, A., Harvey, B.D., and Gauthier, S Natural fire regime: a guide for sustainable management of the Canadian boreal forest. Silva Fenn. 36(1): Blais, J.R Regional variation in susceptibility of eastern North American forests to budworm attack based on history of outbreaks. For. Chron. 44(3): Blais, J.R Trends in the frequency, extent, and severity of spruce budworm outbreaks in eastern Canada. Can. J. For. Res. 13(4): doi: /x Blais, J.R The ecology of the eastern spruce budworm: a review and discussion. In Recent advances in spruce budworms research. Edited by C.J. Sanders, R.W. Stark, E.J. Mullins, and J. Murphy. Proc. CANUSA Spruce Budworms Res. Symp., Bangor, Maine, September, Canadian Forestry Service, Ottawa, Ont. pp Bouchard, M., and Pothier, D Spatio-temporal variability in tree and stand mortality caused by spruce budworm outbreaks in eastern Quebec. Can. J. For. Res.40: doi: /X

59 Bouchard, M., Kneeshaw, D., and Bergeron, Y Forest dynamics after successive spruce budworm outbreaks in mixedwood forests. Ecology, 87(9): doi: / (2006)87[2319:FDASSB]2.0.CO;2. PMID: Bouchard, M., Pothier, D., and Gauthier, S Fire return intervals and tree species succession in the North Shore region of eastern Quebec. Can. J. For. Res. 38(6): doi: /X Boucher, Y., Arseneault, D., and Sirois, L Logging-induced change ( ) of a preindustrial landscape at the northern range limit of northern hardwoods, eastern Canada. Can. J. For. Res. 36(2): doi: /x Bubie, W.L Ten year pulpwood cutting program. Anticosti Management Unit No. 16. Forestry Department. Consolidated Paper Corporation Ltd., Forestry Department, Grand-Mère, Qué. Candau, J.-N., and Fleming, R.A Landscape-scale spatial distribution of spruce budworm defoliation in relation to bioclimatic conditions. Can. J. For. Res. 35(9): doi: /x Caron, M.N., Kneeshaw, D.D., De Grandpre, L., Kauhanen, H., and Kuuluvainen, T Canopy gap characteristics and disturbance dynamics in old-growth Picea abies stands in northern Fennoscandia: is the forest in quasi-equilibrium? Ann. Bot. Fenn. 46: Chouinard, A., and Filion, L Impact of introduced white-tailed deer and native insect defoliators on the density and growth of conifer saplings on Anticosti Island, Quebec. Ecoscience, 12(4): doi: /i Cook, E.R., and Holmes, R.L Program ARSTAN chronology development with statistical analysis (users manual for program ARSTAN). Laboratory of Tree-Ring Research, University of Arizona, Tucson, Az. Dallaire, S Bases écologiques pour l aménagement de l habitat hivernal du caribou de la Gaspésie: le cas d une sapinière boréale inéquienne. M.Sc. thesis, Université Laval, Québec, Qué. Dansereau, M.-C Mortalité des arbres dans les sapinières matures du Parc de la Gaspésie, Québec. M.Sc. thesis, University of Quebec at Montreal, Montreal, Qué. De Grandpré, L., Simard, M., and Gauthier, S La dynamique de développement de peuplements résineux dans la forêt boréale de l est du Québec: une analyse dendroécologique. Chap. 6. In Amélioration de la précision du calcul de la possibilité forestière par une meilleure connaissance de la dynamique naturelle de la forêt boréale de la Côte-Nord. Projet #no présenté au Fonds Forestier, Direction de la recherche forestière, Ministère des Ressources naturelles de la Faune et des Parcs. 39

60 Desponts, M., Brunet, G., Belanger, L., and Bouchard, M The eastern boreal oldgrowth balsam fir forest: a distinct ecosystem. Can. J. Bot. 82(6): doi: / b Elkie, P.C., Rempel, R.S., and Carr, A.P Patch analyst user s manual: a tool for quantifying landscape structure. NWST Technical Manual TM-002. Ontario Ministry of Natural Resources, Thunder Bay, ON, Ont. Engelmark, O Forest Ffires in the Muddus National Park (northern Sweden) during the past 600 years. Can. J. Bot. 62(5): doi: /b Environment Canada Canadian climate normals, temperature and precipitations, , Quebec. Atmospheric Environment Service, Ottawa, Ont. ESRI Arcview 3.1. Environmental Systems Research Institute Inc., Redlands, Calif. ESRI ArcGis version 9/ArcMap. Environmental Systems Research Institute Inc., Redlands, Calif. Etheridge, D.A., MacLean, D.A., Wagner, R.G., and Wilson, J.S Effects of intensive forest management on stand and landscape characteristics in northern New Brunswick, Canada ( ). Landsc. Ecol. 21(4): doi: /s Fahrig, L Effects of habitat fragmentation on biodiversity. Annu. Rev. Ecol. Evol. Syst. 34(1): doi: /annurev.ecolsys Food and Agriculture Organization Criteria and indicators for sustainable forest management: a compendium. Paper compiled by F. Castañeda, C. Palmberg- Lerche, and P. Vuorinen, May Forest Management Working Papers, Working Paper 5. Forest Resources Development Service, Forest Resources Division, FAO, Rome. Fortin, F Field work of Anticosti summer survey Forestry Survey Department. Consolidated Paper Corporation Ltd., Grand-Mère, Qué. Foster, D.R The history and pattern of fire in the boreal forest of southeastern Labrador. Can. J. Bot. 61(9): doi: /b Foster, D.R., Orwig, D.A., and McLachlan, J.S Ecological and conservation insights from reconstructive studies of temperate old-growth forests. Trends Ecol. Evol. 11(10): doi: / (96) Frelich, L.E., and Reich, P.B Spatial patterns and succession in a Minnesota southern-boreal forest. Ecol. Monogr. 65(3): doi: /

61 Gauthier, S., Vaillancourt, M.-A., Leduc, A., De Grandpré, L., Kneeshaw, D.D., Morin, H., Drapeau, P., and Bergeron, Y Aménagement écosystémique en forêt boréale. Presses de l Université du Québec., Québec, Qué. Gray, D.R., Regniere, J., and Boulet, B Analysis and use of historical patterns of spruce budworm defoliation to forecast outbreak patterns in Quebec. For. Ecol. Manag.127(1 3): doi: /S (99) Gromtsev, A Natural disturbances dynamics in the boreal forests of European Russia: a review. Silva Fenn. 36(1): Groupe de travail sur les écorégions Les régions écoclimatiques du Canada. Série de la classification écologique du territoire, no. 23. Première approximation. Environment Canada, Canadian Wildlife Service, Sustainable Development Branch, Ottawa, Ont. Holmes, R.L Computer-assisted quality control in tree-ring dating and measurement. Tree-Ring Bull. 43: Holmes, R.L., and Swetnam, T Detecting outbreaks of spruce budworm and tussock moth in annual tree-ring growth, and distinguishing between the insect species. Laboratory of Tree-Ring Research, University of Arizona, Tucson, Az. Jardon, Y., and Doyon, F Balsam fir stand dynamics after insect outbreak disturbances in western Newfoundland ecoregion (Corner Brook Subregion). Western Newfoundland Model Forest, Inc., Corner Brook, Nfld. Jardon, Y., Morin, H., and Dutilleul, P Périodicité et synchronisme des épidémies de la tordeuse des bourgeons de l épinette au Québec. Can. J. For. Res. 33(10): doi: /x Jobin, L L arpenteuse de la pruche: un insecte insulaire. Racine, 2: 5 7. Johnson, E.A., Miyanishi, K., and Weir, J.M.H Wildfires in the western Canadian boreal forest: landscape patterns and ecosystem management. J. Veg. Sci. 9(4): doi: / Kuuluvainen, T Natural variability of forests as a reference for restoring and managing biological diversity in boreal Fennoscandia. Silva Fenn. 36(1): Landres, P.B., Morgan, P., and Swanson, F.J Overview of the use of natural variability concepts in managing ecological systems. Ecol. Appl. 9(4): Lavoie, M., Filion, L., and Robert, E.C Boreal peatland margins as repository sites of long-term natural disturbances of balsam fir/spruce forests. Quat. Res. 71(3): doi: /j.yqres

62 Leblanc, M., and Bélanger, L La sapinière vierge de la forêt Montmorency et de sa région: une forêt boréale distincte. Mémoire de recherche forestière no Ressources naturelles du Québec, Qué. Lejeune, L., and Dion, J.-N Anticosti: l époque de la console Les Éditions JML, Saint-Hyacinthe, Qué. Lévesque, F Conséquence de la dynamique de la mosaïque forestière sur l intégrité écologique du Parc National Forillon. M.Sc. thesis, Université Laval, Québec, Qué. Lewis, K.J., and Lindgren, S.B A conceptual model of biotic disturbance ecology in the central interior of B.C.: how forest management can turn Dr. Jekyll into Mr. Hyde. For. Chron. 76(3): Lindenmayer, D., Hobbs, R.J., Montague-Drake, R., Alexandra, J., Bennett, A., Burgman, M., Cale, P., Calhoun, A., Cramer, V., Cullen, P., Driscoll, D., Fahrig, L., Fischer, J., Franklin, J., Haila, Y., Hunter, M., Gibbons, P., Lake, S., Luck, G., MacGregor, C., McIntyre, S., Nally, R.M., Manning, A., Miller, J., Mooney, H., Noss, R., Possingham, H., Saunders, D., Schmiegelow, F., Scott, M., Simberloff, D., Sisk, T., Tabor, G., Walker, B., Wiens, J., Woinarski, J., and Zavaleta, E A checklist for ecological management of landscapes for conservation. Ecol. Lett. 11(1): PMID: MacLean, D.A Effects of spruce budworm outbreaks on the productivity and stability of balsam fir forests. For. Chron. 60(5): Martel, M Analyse dendroécologique des effets de la défoliation par l arpenteuse de la pruche [Lambdina fiscellaria (Guen.)] sur les populations forestières de l île d Anticosti, Québec. M.Sc. thesis, Université Laval, Québec, Qué. McCarthy, J.W Natural disturbance and structure in two primary boreal forests of western Newfoundland. Ph.D. dissertation, Faculty of Forestry, Department of Forest Sciences, University of British Columbia, Vancouver, B.C. McCarthy, J.W., and Weetman, G Self-thinning dynamics in a balsam fir (Abies balsamea (L.) Mill.) insect-mediated boreal forest chronosequence. For. Ecol. Manag. 241(1 3): doi: /j.foreco McGarigal, K., Cushman, S.A., Neel, M.C., and Ene, E FRAGSTATS: spatial pattern analysis program for categorical maps. Computer software program produced by the authors at the University of Massachusetts, Amherst, Mass. Available at Morin, H Dynamics of balsam fir forests in relation to spruce budworm outbreaks in the boreal zone of Quebec. Can. J. For. Res. 24(4): doi: /x

63 Potvin, F., Beaupré, P., and Laprise, G The eradication of balsam fir stands by white-tailed deer on Anticosti Island, Québec: a 150-year process. Ecoscience, 10(4): Ries, L., Fletcher, R.J., Battin, J., and Sisk, T.D Ecological responses to habitat edges: mechanisms, models, and variability explained. Annu. Rev. Ecol. Evol. Syst. 35(1): doi: /annurev.ecolsys Rochette, B., and Gingras, A Inventaire aérien du cerf de Virginie de l île d Anticosti été Ministère des Ressources naturelles et de la Faune, Direction de l aménagement de la faune de la Côte-Nord. Royer, R. 1956a. Book III: maps of stand age classes. Anticosti Forest Management Unit No. 16. Forestry Department. Consolidated Paper Corporation Ltd., Grand-Mère, Qué. Royer, R. 1956b. Book III: map of cut-over areas. Anticosti Forest Management Unit No. 16. Forestry Department. Consolidated Paper Corporation Ltd., Grand-Mère, Qué. Royer, R., and Grondin, M Management plan report: general information. Consolidated Paper Corporation Ltd., Grand-Mère, Qué. Saucier, J.-P., Grondin, P., Robitaille, A., Gosselin, J., Morneau, C., Richard, P.J.H., Brisson, J., Sirois, L., Leduc, A., Morin, H., Thiffault, É., Gauthier, S., Lavoie, C., and Payette, S Écologie forestière. In Ordre des ingénieurs forestiers du Québec. Manuel de foresterie. 2 éd. Ouvrage collectif. Éditions MultiMondes, Québec, Qué. pp Shorohova, E., Kuuluvainen, T., Kangur, A., and Jogiste, K Natural stand structures, disturbance regimes and successional dynamics in the Eurasian boreal forests: a review with special reference to Russian studies. Ann. For. Sci. 66(2): article 201. doi: /forest/ Simard, M., and Payette, S Black spruce decline triggered by spruce budworm at the southern limit of lichen woodland in eastern Canada. Can. J. For. Res. 31(12): doi: /cjfr Tardif, J., Conciatori, F., and Bergeron, Y Comparative analysis of the climatic response of seven boreal tree species from northwestern Québec, Canada. Tree-Ring Res. 57(2): Thompson, I.D., Larson, D.J., and Montevecchi, W.A Characterization of old wet boreal forests, with an example from balsam fir forests of western Newfoundland. Environ. Res. 11(Suppl. 1): S23 S46. doi: /A

64 Weir, J.M.H., and Johnson, E.A Effects of escaped settlement fires and logging on forest composition in the mixedwood boreal forest. Can. J. For. Res. 28(3): doi: /cjfr White, P.S., and Walker, J.L Approximating nature s variation: selecting and using reference information in restoration ecology. Restor. Ecol. 5(4): doi: / j x x. Williams, D.W., and Birdsey, R.A Historical patterns of spruce budworm defoliation and bark beetle outbreaks in North American conifer forests: an atlas and description of digital maps. U.S. For. Serv. Gen. Tech. Rep. NE-308. Yamaguchi, D.K A simple method for cross-dating increment cores from living trees. Can. J. For. Res. 21(3): doi: /x

65 Chatpitre 2: Preferential deer browsing disrupts regeneration processes of gap-driven balsam fir stands triggering forest degradation and alternative successional pathways 2.1. Résumé Le broutement par les cervidés peut altérer les mécanismes de résilience des forêts engendrant des trajectoires successionnelles alternatives. Les processus de régénération sont un des principaux mécanismes de résilience des forêts parce qu ils déterminent les trajectoires successionnelles. Sur l île d Anticosti, le broutement chronique par le cerf de Virginie (Odocoileus virginianus), une espèce introduite, a changé la composition de la régénération des sapinières dynamisées par les petites trouées vers une dominance de l épinette blanche (Picea glauca). Un tel changement de composition pourrait interrompre les processus de régénération des trouées engendrant des trajectoires successionnelles alternatives. Avec une méthodologie novatrice intégrant les ondelettes, nous montrons que l épinette blanche n a pas permis aux sapinières de récupérer complètement des perturbations par touées. Effectivement, lorsque les espèces d arbres non broutées sont des remplisseurs de trouées inefficaces, le broutement préférentiel par les cerfs altère les mécanismes de résiliences des forêts dynamisées par les petites trouées en interrompant les processus de régénération. Nous avons aussi montré qu une telle perte de résilience engendre des processus de dégradation du couvert et des trajectoires successionnelles vers de nouveaux écosystèmes. Il a été démontré que la dégradation du couvert forestier et les forêts-parcs peuvent avoir d importantes répercussions sur la biodiversité et les services écosystémiques dans d autres écosystèmes. Notamment, nos résultats envoient un avertissement aux aménagistes forestiers des forêts dynamisées par les petites trouées aux prises avec de hautes densités d orignaux (Alces alces) et des populations grandissantes de cerfs de Virginie. Pour tenir compte de ces nouveaux enjeux, les aménagistes forestiers devront s assurer que les initiatives d aménagement écosystémique sont efficaces pour maintenir les processus de régénération des forêts dynamisées par les petites trouées. 45

66 2.2. Abstract Cervid browsing can alter resilience mechanisms of forests, thereby triggering alternative successional pathways. Regeneration processes are an important mechanisms of forest resilience because they drive successional pathways. On Anticosti Island, chronic browsing by introduced white-tailed deer (Odocoileus virginianus) shifted composition of understory regeneration of gap-driven balsam fir (Abies balsamea) forests toward the dominance of white spruce (Picea glauca). Such a shift in the composition of understory regeneration could disrupted gap-regeneration processes triggering forest degradation processes and alternative successional pathways. With an innovative methodological approach using wavelets, we show that white spruce has not enabled balsam fir forests to recover completely from gap disturbances because of its poor aptitude in filling canopy gaps. Effectively, when the unbrowsed tree species are ineffective gap filler, preferential browsing by introduced deer alters resilience mechanisms of gap-driven forests by disrupting regeneration processes. We also show that such a loss of resilience triggers forests degradation processes and alternative successional pathways toward white spruce stands and parklands. Forest degradation and parklands (i.e. partial deforestation), have been shown to be a threat to biodiversity and ecosystem services in other ecosystems. Notably, our findings send a warning signal to forest managers of gap-driven boreal forests with already high densities of moose (Alces alces) and growing populations of white-tailed deer. To deal with these novel issues, forest managers will have to make sure that ecosystem management initiatives are efficient in maintaining regeneration processes of gap-driven forests. 46

67 2.3. Introduction Cervid browsing can alter resilience mechanisms of forests, thereby triggering alternative successional pathways (Tremblay et al. 2007; Hidding et al. 2013; DiTommaso et al. 2014). Resilience is the capacity of a system to absorb disturbance and reorganize so that the same structure and functions are essentially recovered, e.g., forest recovering to forest following perturbation (Holling 1973; Gunderson 2000). Regeneration processes are an important mechanisms of forest resilience because they drive successional pathways (Buma and Wessman 2011; Hidding et al. 2013). In gap-driven forests, deer browsing could disrupt regeneration processes by shifting composition of understory regeneration and causing tree recruitment failure (Frelich and Lorimer 1985, Healy 1997, Pedersen and Wallis 2004). Such a disruption of gap regeneration processes could trigger forest degradation (i.e. loss of forest canopy cover density; Foley et al. 2007; Sasaki and Putz 2009) and alternative successional pathways that move the system toward assemblages and relative abundances of species that have not co-occurred historically. The occurrence of such novel ecosystems can represent a threat to biodiversity and ecosystem services that are provided by preindustrial forests (Hobbs et al. 2006; Bridgwater et al. 2011). Hence, a recent cause for concern is the increasing deer density in gap-driven forests of North America (Côté et al. 2004; McLaren et al 2004; Chollet and Martin 2012). Following the introduction of 200 white-tailed deer (Odocoileus virginianus) into the forests of Anticosti Island (Quebec) at the end of the 19th century, their population densities have now reached more than 20 individuals/km² in the absence of predators (Rochette and Gingras 2007). Since the 1920 s, chronic deer browsing has shifted the understory composition of balsam fir forests on the island from regeneration that is dominated by balsam fir (Abies balsamea (L.) Miller) towards one that is dominated by white spruce (Picea glauca (Moench) Voss), together with a recalcitrant ground cover that includes bluejoint (Calamagrostis canadensis (Michauxx) P.Beauv.), bracken fern (Pteridium aquilinum (L.) Kuhn), and thistle (Cirsium spp.), which are species that are not browsed by deer (Potvin et al. 2003; Tremblay et al. 2007; Hidding et al. 2013). Historically, the unaltered forests of Anticosti Island were driven by gap disturbance and gap regeneration processes that assured the maintenance of overmature balsam fir stands, 47

68 mixed with white spruce and paper birch (Betula papyrifera Marshall; Fig. 2-1; Barrette et al. 2010). Balsam fir regeneration created dense and persistent understory seedling banks, which were highly tolerant of shade and that were prompt to react and fill canopy gaps (Côté and Bélanger 1991; McCarthy 2001; McCarthy and Weetman 2007a). From the 1920's onwards, white spruce has represented the sole remaining tree species that is available to fill gaps and recreate a forests after canopy gap disturbances. It has not yet been evaluated if such a shift in the composition of understory regeneration can disrupt gap regeneration processes triggering forest degradation processes and alternative successional pathways. Forest degradation has important repercussions on ecosystem services (Foley et al. 2007). Eventual white spruce stands would also represent a threat to biodiversity and the ecosystem services that were provided by the preindustrial forest (Barrette et al. 2010), since they are largely novel ecosystems which are not commonly found in eastern North America (Stiell 1976; Bell et al. 1990; Lieffers et al. 2008; Saucier et al. 2009). Stands may not represent the sole successional pathway induced by preferential browsing from introduced deer. Effectively, white spruce is a secondary canopy species on the Island (Barrette et al. 2010) while its regeneration is intermediate in shade tolerance (Burns and Honkala 1990), is vulnerable to overtopping from understory recalcitrant layer (Hogg and Lieffers 1991; Lieffers et al. 1993; Cole et al. 2013), needs specific substrates for optimum seedling establishment (Simard et al. 1998), rarely forms pure even-aged stands under natural conditions (Stiell 1976; Bell et al. 1990) and is considered a secondary gap filler species (Pham et al. 2004, Brassard and Chen 2006). Alternative novel ecosystems, such as 48

69 Balsam fir White spruce Paper birch Recalcitrant understory layer species Gap disturbance Balsam fir stand Gapregeneration process Preferential browsing of main gap filler ( ) by introduced deer Composition shift of understory regeneration Gap disturbance Forest degradation Disruption of gap-regeneration process Usual small gap pattern Large gap pattern Insufficient regeneration Sufficient regeneration Pathways of alternative succession Recalcitrant understory layer White spruce parkland White spruce stand Fig Diagram illustrating predictions of process triggered by a shift in the composition of understory regeneration of gap-driven balsam fir forests 49

70 parklands (i.e. partial deforestation), may thus result from preferential deer browsing. The occurrence of partial deforestation would represent an even greater threat to biodiversity and ecosystem services. Hence, the objective of this study is to determine if the shift in the composition of understory regeneration has disrupted regeneration processes of gap-driven balsam fir forests triggering forest degradation processes and alternative successional pathways. We predict that composition shift of understory regeneration following preferential browsing of main gap filler species will disrupt gap regeneration processes triggering forest degradation processes (Fig. 2-1). We predict that the spatial pattern of forest degradation will determine the pathway of the alternative succession. Forest degradation in usual small gap patterns (i.e. 200 m² : McCarthy 2001) will trigger an alternative successional pathway toward white spruce parklands (Fig. 2-1). White spruce regeneration will not be sufficient to fill these small gaps and recreate a closed canopy forest because its intermediate in shade tolerance, its vulnerable to overtopping from understory recalcitrant layer and its considered a secondary gap filler species. Alternatively, forest degradation in larger gap patterns (>200 m²) will trigger an alternative successional pathway toward white spruce stands. White spruce regeneration will be sufficient to fill larger gaps and recreate a closed canopy forest because it is not browsed, tolerates direct sunlight and has formed closed canopy forests on Anticosti following large-scale disturbances (i.e s clearcuts and 1970 s hemlock looper outbreaks: Jobin 1980, Beaupré et al. 2004) Materials and methods Study Area Anticosti Island (7943 km²) is located in the Gulf of St. Lawrence (49 28 N, W; Fig. 2-2). The island is part of the eastern balsam fir-paper birch bioclimatic subdomain of Quebec (Saucier et al. 2009). Mean annual temperature is 2 C and mean annual precipitation is 907 mm (Environment Canada 1982). The preindustrial forest was dominated (75 %) by softwood stands that were about 100-years-old (Barrette et al. 2010). Natural disturbances include blowdowns, outbreaks of spruce budworm (Choristoneura 50

71 fumiferana (Clem.)) and hemlock looper (Lambdina fiscellaria (Guen.) defoliation (Jobin 1980; Martel 1999; Chouinard and Filion 2001) and, to a much lesser extent, fire (Lavoie et al. 2009). Sampling of old balsam fir stands was restricted to the western half of the island where the concentration of such stands was highest (Fig. 2-2). The eastern part of the island has a higher proportion of unbrowsed black spruce (Picea mariana (Miller) BSP) forests and wetlands. In old balsam fir forests, chronic deer browsing has not altered development of canopy trees because their crowns were beyond the reach of deer when browsing began to inhibit tree recruitment (Chouinard and Filion 2001; Potvin et al. 2003; Barrette et al. 2010) Field sampling Sample site selection was based on forest inventory maps that were obtained from the Ministère des Ressources naturelles du Québec (MRNQ). We selected balsam fir forest zones (>75% of total basal area composed of balsam fir), of a minimal size of 2 km² and of an estimated photo-interpreted age equal or greater to 90 years. Only stands 90 years were considered because they correspond to the age-class when canopy mortality starts occurring in balsam fir stands, thus associated with the beginning of the old-growth phase (Kneeshaw and Gauthier 2003). Nine balsam fir forests were randomly selected from all available forests. In each of these forests, a 4 m wide and of length varying between m transect was sampled, for a total length of 8775 m and a total area of 3.51 ha (Fig. 2-2). Transect orientation followed the longest axis crossing each forest. Limits of individual gaps were hard to identify in the field and we could not distinguish adjacent or overlapping gaps one from the other, because the gap-filling process was hindered by deer browsing. Hence, we mapped understory regeneration and canopy trees for subsequent identification of canopy gaps, statistically with wavelets (Bradshaw and Spies 1992; Rouvinen et al. 2002, Pham et al. 2004, Aakala et al. 2007). Along each transect, tall seedlings (height cm) of all tree species were counted in contiguous 5 m X 4 m sections (20 m²), for subsequent spatial analysis. To complete the inventory of likely established regeneration (i.e. >30 cm in height; Côté and Bélanger 1991, Ruel. et al 1995, Westerberg 1995), the density of small seedlings (height <61 cm) was measured in a subsample of six clusters of 51

72 Anticosti Island Balsam fir stands (age-class 90 years) Dendrochronological plots Transects (1 to 9) Roads Province of Quebec N Eastern balsam fir-paper birch bioclimatic subdomain Kilometers Fig Location of the transects and of the dendrochronological plots in balsam fir stands of Anticosti Island, Quebec, Canada. five circular subplots (4 m²) by transect. Clusters were placed 150 m apart along the central line of the transect while subplots were placed 10 m apart within a cluster. Small seedlings were counted by tree species and by height class (0-10, and cm) in each subplot. The percent cover of species forming recalcitrant understory layers was also estimated in these subplots by 5% cover classes. The diameter at breast height (DBH) of live and dead saplings (DBH cm) and canopy trees (DBH 10 cm) of all tree species was measured along each transect. Live and dead saplings and trees were positioned within the transect with a precision of 0.1 m. A decay class (Table 2-1) and a cause of mortality (suppression, root rot, butt rot, stem breakage, uprooted) were assigned to all dead 52

73 stems (Kuuluvainen et al. 2001, Campbell and Laroque 2007, Taylor and MacLean 2007). A total of 104 dead canopy trees were collected from eight dendrochronological plots (50 m X 8 m) in old balsam fir stands located near a transect (Fig. 2-2) to evaluate time since death of trees in different decay classes and ultimately to verify if forest degradation expands gradually along transects. We sampled two cross-sections from 6 snags and 6 logs (>50 year rings) of balsam fir, in each of the first four decay classes (Table 2-1). Dead trees in decay classes 5 to 7 were not sampled since they were too decayed for dendrochronological analysis. Dried cross-sections were finely sanded (220 to 400 grit). Ring width was measured (0.01 mm) on two radii on each cross-section of dead trees (n = 208 from 104 trees) with a Velmex (Bloomfield, New York, USA) micrometer connected to a computer. The year of last ring formation was visually cross-dated on each tree using the pointer-year method (Yamaguchi 1991). We used pointer-years and the master chronology of Barrette et al. (2010), which was built from 140 live balsam fir trees (n = 251 radii) located in the same plots (Fig. 2-1) for visual cross-dating. We validated cross-dating of year of death of each tree with the COFECHA program (Holmes 1983). Table 2-1. Description of decay class for snags and logs. Decay class Description of snags Description of logs 1 Presence of needles and twigs. Presence of needles and twigs. Wood still solid. 2 Needles absent, twigs may be present and bark still attached. Wood still solid. Needles absent, twigs may be present and bark still attached. Tree elevated on branches. Knife penetrates up to 1 cm. 3 Main branches present only. Bark begins to loosen. Knife penetrates up to 1 cm. 4 More than 20% of bark has fallen. Knife penetrates between 1 and 3 cm. 5 Knife penetrates between 3 and 5 cm. Broken top. 6 More than 80% of bark has fallen. Knife penetrates all the way. Broken top. Knife penetrates between 1 and 3 cm. Presence of moss. More than 20% of bark has fallen. Moss covers more than 20% of log. Knife penetrates between 3 and 5 cm. Moss covers more than 50% of log. Knife penetrates all the way. Moss covers most of log. Woody material crumbles between fingers. 7 Flattened log incorporated into the humus layer and appears as red woody material without structure. 53

74 Data analysis Gap regeneration processes and forest degradation processes Tall seedlings, saplings and trees along all transects were grouped in contiguous 25 m sections (100 m²; n = 351). This size was chosen as it corresponds to mean gap size observed in many boreal forests (McCarthy 2001). A section was considered regenerated enough to fill gaps and eventually recreate a closed canopy forests if tall seedling and sapling density was over 1500 stems/ha (Greene et al. 2002). Similarly, a section was considered in a forest degradation process if density of live canopy trees was below the range of natural variability ( live trees/ha; m²/ha) and density of dead canopy trees was above the range of natural variability ( dead trees/ha; m²/ha) of gap-driven balsam fir stands without forest degradation (Ker 1976, Zarnovican and Vézina 1985, McCarthy and Weetman 2007, Barrette et al. 2010). We used tree density has a proxy of canopy cover density to analyse forest degradation processes, because limits of individual gaps were hard to identify in the field and density, notably basal area, is highly correlated to canopy cover density (Mitchell and Popovich 1997). To provide an integrated view, we calculated an openness index for each transect section by dividing density of dead canopy trees by density of live and dead canopy trees. We then produced DBH distributions and distributions of number of transect sections by basal area, density and openness index classes. We tested differences between means with the DIFF ADJUST (Tukey-Kramer) option of the LSMEANS statement, within a mixed linear model (PROC MIXED; SAS Institute Inc. 2003) with transects as a random effect. From time since death data of the eight dendrochronological plots, we estimated mean time since death by decay class with a mixed linear model (PROC MIXED; SAS Institute Inc. 2003) with decay classes as a fixed effect and plots as a random effect. To verify if forest degradation expands gradually along transects, we then attributed a time since death to the dead canopy trees (i.e. gap makers) found along the transects according to their corresponding decay class. 54

75 Spatial pattern of forest degradation and pathways of succession To identify canopy gaps statistically we analyzed the spatial distribution of canopy trees basal area along transects with wavelets (Bradshaw and Spies 1992, Dale and Mah 1998, James et al. 2010). The wavelet is a scalable windowing function that can be used to characterize the spatial patterns in canopy structure (Bradshaw and Spies 1992, Rosenberg and Anderson 2011). Where the template of the wavelet function matches the original transect data, the value of the wavelet variance at this position is high (Rosenberg and Anderson 2011). Where the two do not match, the value of the wavelet variance is low. Therefore, where there is an abrupt change in adjacent canopy tree basal area, there will be a significant increase (i.e. peak) in wavelet position variance indicating the encounter of the edge of a canopy gap (Bradshaw and Spies 1992, Camarero et al. 2006, James et al. 2010). To identify significant peaks in wavelet position variance, we determined confidence intervals by generating wavelet analysis on 999 randomizations of the original transect data and considered the 90th highest value of position variance for each position (Camarero et al. 2006). A maximum scale of 20% of the length of the transect (i.e. 200 m) was used for the analysis with the wavelet functions (M.S. Rosenberg, personal communication). We used Mexican Hat wavelets to analyze distribution of spatial patterns and Morlet wavelets to analyze scale of spatial patterns (Keitt and Urban 2005, Mi et al. 2005). The spatial distribution of canopy tree basal area was analyzed along contiguous 25 m sections (100 m²; n = 351) of transects with both wavelet types. Basal area was used instead of number of trees because it is more correlated to canopy cover density than number of trees (Mitchell and Popovich 1997) and basal area is a measure of density that integrates tree size. Separate analyses were performed for living and for dead canopy trees. In part from this wavelet analysis, we identified the occurrence of canopy gaps. To be considered as being part of canopy gap, a transect section had to meet the following criteria: 1) be under or next to a significant peak in wavelet position variance of canopy tree basal area, 2) basal area of live canopy trees had to be <25% (i.e. canopy density class D: MRNQ 2011) of a fully stocked balsam fir stand (i.e. 13 m²/ha: Zarnovican and Vézina 1985) and 3) basal area of dead canopy trees had to be above range of natural variability (12-23 m²/ha), or 4) the openness index had to be 60%. Independently from the wavelet scale analysis of 55

76 spatial patterns of forest degradation, gap size was also determined from the number of consecutive neighbouring sections meeting gap criteria. Gaps covering one or two neighbouring sections ( 200 m²) were considered as small gaps while ones covering more than two neighbouring sections (>200 m²) were considered as large gaps (McCarthy 2001). In each gap, we described dead canopy trees as gap makers and tall seedlings and saplings as gap fillers. Minimal regeneration criteria, described earlier, were used to determine if these gaps had sufficient or insufficient regeneration to fill gaps and eventually recreate a closed canopy forest, which determined pathways of succession.. The software PASSAGE 2 was used (Rosenberg and Anderson 2011) to perform all spatial analyses Results Gap regeneration processes Composition of understory regeneration was dominated by white spruce, which comprised 90% of all live saplings and likely established seedlings, i.e. height >30 cm (Table 2-2 and Fig. 2-3). All other species including balsam fir were rare or absent in the regeneration strata >30 cm. Balsam fir and deciduous species were only present in the lowest regeneration strata, i.e. <30 cm. Two thirds of live saplings were found in the 2 cm DBH class (Fig. 2-3A; t 2,40 = 2.57; p = 0.014), in which white spruce comprised 99% of all stems. Larger live saplings (5 and 8 DBH class) were still predominated by white spruce but were two times less abundant than small live trees (11 and 14 cm DBH classes; t 5,88 = -2.14; p = 0.035) which were predominated by balsam fir. The low abundance of large live saplings of balsam fir and of white spruce generated a bimodal DBH distribution. Large dead saplings (DBH > 2 cm), mainly composed of balsam firs, were five times more abundant than large live saplings (Fig. 2-3B; t 1,24 = 5.09; p < 0.001), mainly composed of white spruce. Small dead saplings, again mainly composed of firs, were six times less abundant than large dead saplings (t 2,40 = -2.81; p = 0.008). 56

77 White spruce sapling density was insufficient to fill gaps and recreate a closed canopy forests (i.e. <1500 stems/ha), in 97% of transect sections and they were totally absent in 71% (Fig. 2-4A). Tall seedling density was insufficient in 63% of sections and they were absent in 36% of transect sections Forest degradation processes The forest canopy was composed of balsam fir (66%), white spruce (28%) and paper birch (5%) (Table 2-2 and Fig. 2-3A). Dead canopy tree composition was represented by balsam fir (69%), white spruce (13%) and paper birch (5%) and unidentified coniferous species (because of advanced decay; 13%). Identifiable causes of mortality of canopy trees were stem breakage (42%), stem breakage with butt rot (21%), uprooting (28%) and uprooting with root rot (6%) (Table 2-3). Almost two thirds of the snags were in intermediate decay stages and one third were in initial decay stages while very few snags where in an advanced decay stage. Conversely, there were as many logs in initial, intermediate and advance decay stages (t 2,16 = 2.92; p = 0.083). The number of live canopy trees was below the range of natural variability (i.e trees/ha) in 85% of the transect sections (Fig. 2-4B). Basal area of live canopy trees was below the range of natural variability (i.e m²/ha) in 72% of sections. Accordingly, the number of dead canopy trees was above the range of natural variability (i.e trees/ha) in 64% of sections, while basal area of dead canopy trees was above the range of natural variability (i.e m²/ha) in 63% of sections. In a more integrated analysis, there were more dead canopy trees than live ones in 54% of transect sections, i.e. in these sections, openness index was >50% (Fig. 2-4B) Spatial patterns of forest degradation Wavelet scale analysis of the spatial distribution of canopy tree basal area confirmed that small and large gaps represented a significant scale of spatial pattern of forest degradation process respectively in two thirds and in all transects (Fig. 2-5). 57

78 Table 2-2. Mean (±1 SE) number of stems per hectare of understory regeneration and canopy trees in 9 transects located in old balsam fir stands of Anticosti Island, Quebec, Canada. Small seedlings Tall seedlings Saplings (DBH cm) Canopy trees (DBH 10 cm) 0-10 cm cm cm cm Live Dead Live Dead Balsam fir (6383) 2970 (1350) 206 (206) 45 (42) 28 (12) 305 (82) 423 (53) 539 (70) White spruce 6699 (1161) 6234 (1372) 2940 (691) 1945 (153) 173 (34) 57 (28) 181 (67) 101 (29) Black spruce 16 (16) 206 (131) 168 (94) 84 (44) 10 (4) 1 (<1) 3 (1) 4 (1) Unidentified coniferous 30 (7) 98 (27) Paper birch 19 (19) 7 (5) 1 (1) 3 (1) 33 (6) 38 (8) Quaking aspen 344 (344) 172 (172) 20 (20) (1) <1(<1) Mountain ash 134 (71) (1) Measured in a subsample of 6 clusters of 5 circular subplots by transect 58

79 Table 2-3. Mean (± 1 SE) tree (DBH 10 cm) density per hectare of snags and logs according to decay stages and mortality process in 9 transects and estimated ( X ± 1 SE) time since death of balsam fir trees according to decay stages, in 8 old balsam fir stands (dendrochonological plots; Fig. 2-2 ) of Anticosti Island, Quebec, Canada. Decay stages Initial (Classes 1 and 2) Intermediate (Classes 3 and 4) Advanced (Classes 5, 6, and 7) Cause of mortality Snags Balsam and unidentified coniferous Spruces Others Time since death (n=69 firs) Logs Balsam and unidentified coniferous Spruces Others Time since death (n=35 firs) trees/ha years trees/ha years 51 (7) 25 (14) 4 (2) 4 (3) 146 (19) 15 (3) 2 (2) 6 (4) 107 (22) 24 (9) 8 (2) 18 (2) 190 (33) 32 (15) 10 (2) 16 (4) 12 (4) 1 (1) 1 (1) 132 (21) 8 (4) 11 (4) Suppression 10 (2) 4 (2) 0 8(4) 0 0 Root rot 1 (1) (10) 5 (2) 4 (1) Butt rot 9 (4) <1 (<1) <1 (<1) 145 (26) 13 (4) 3 (1) Stem breakage 8 (3) 1 (<1) 1 (1) 277 (32) 39 (15) 9 (2) Uprooted 2 (1) 0 1 (<1) 195 (30) 16 (6) 13 (4) Unknown 147 (26) 45 (15) 12 (2) 34 (6) 4 (2) 4 (1) A tree can qualify for multiple causes of mortality Samples too decayed to estimate time since death 59

80 sity (stems/ha) Den A B DBH classes (3 cm) Live stems Spruces Others Balsam fir Mean DBH: 17 cm Median DBH : 17 cm Dead stems Mean DBH : 15 cm Median DBH : 14 cm Fig Distribution of mean (± 1 SE) stem density by diameter at breast height (DBH) classes in the 9 transects. 60

81 A. Understory white spruce regeneration 250 M Saplings (DBH cm) tall seedlings (height cm) X Number of stems (500 stems/ha) M X Number of stems (500 stems/ha) B. Canopy trees (DBH 10 cm) 50 live M X live 40 M X 30 Number of sections of transect X M dead dead M X Number of trees (100 trees/ha) Basal area (4 m 2 /ha) Openness index (dead / live+dead) Openness index (dead / live+dead) X X M M Openness index from number of trees (5%) Openness index from basal area (5%) Fig Distribution of number of sections of transects (100 m2; n = 351) by density classes of understory white spruce regeneration (A) and canopy trees (B). 61

82 According to our canopy gap criteria, forest degradation covered close to half of the 8775 m of transect, 18% in small gaps (n = 47) and 29% in large gaps (n = 24; Fig. 2-6). The total gap number ranged from 5 to 11 by transect (Fig. 2-7A). Number of small gaps range from 2 to 11 and number of large gaps range from 0 to 4 by transect. Mean length of small gaps that intercepted the transect was 34 m (gap size 136 m²) while mean length of large gaps was 106 m (gap size 424 m²). There were four times more gap makers in large gaps than in small gaps (48 gap makers versus 11; t 1, 58 = -8.57; p < 0.001). Gap makers in all three decay stages (i.e. initial, intermediate and advanced) were present in all large gaps and in 85% of the small ones (Table 2-3 and Fig. 2-7A). According to our subsample of time since death of trees in different decay classes, ultimately to verify if forest degradation expands gradually along transects, we estimated that time since death of snags and logs of balsam fir in initial decay stage was 4 to 6 years and time since death of snags and logs in intermediate decay stage was 16 to 18 years (Table 2-3). Dead wood in the advanced decay stage was too decayed to estimate time since death Pathways of succession Density of white spruce high regeneration was insufficient to fill gaps and eventually recreate a closed canopy forest in 49% of the transect length in gaps (Figs. 2-6). Regeneration was insufficient to fill gaps mainly in small gaps (i.e.83% of transect length in small gaps (n = 37)) and to a much lesser extent in large gaps (i.e. 28% of transect length in large gaps (n = 6); Figs. 2-6, 2-7A and 2-7B). Markedly, white spruce seed trees (i.e. live and dead canopy trees) were present in 68% of these gaps (Fig. 2-7A). Furthermore, the density of regeneration in the cm height class was also insufficient in these gaps (small gaps: 639 ± 216 seedlings/ha; large gaps: 650 ± 371 seedlings/ha). Lastly, the cover of species forming recalcitrant understory layers was 6.5 times more abundant in gaps than under canopy, i.e. cover increased from 2% to 13% (t 10, 257 = -2.65; p = 0.009). In 51% of the transect length in gaps, there was sufficient white spruce regeneration to fill gaps and recreate a closed canopy forest (Figs. 2-6). This was mainly observed in large gaps (i.e. 72% of transect length in large gags), and to a much lesser extent in small gaps (i.e. 17% of transect length in small gags; Figs. 2-6, 2-7A and 2-7B). 62

83 Fig Wavelet scale analysis of spatial patterns of forest degradation processes calculated from the spatial distribution of canopy tree basal area (DBH 10 cm) in the 9 transects (see Fig. 2-2 for reference number). Black line (wavelet variance) above grey line (confidence interval >90%) shows significant scales of spatial patterns, identified SG (small gap 200 m 2 ) or LG 2 (large gap > 200 m ) Live canopy tree basal area SG LG LG LG Dead canopy tree basal area SG LG LG Wwavelet position variance 5 2 SG LG LG LG LG LG LG SG 9 SG LG LG 3 Gap SG LG LG 4 SG LG LG Scale (meters)

84 Canopy gap criteria Spatial pattern White spruce stand regeneration criteria Transect length (meters) Fig Transect length by canopy gap criteria, spatial pattern of forest degradation and white spruce stand regeneration criteria. 64

85 Canopy tree basal area (x 10 m 2 /ha) Live Dead Understory regeneration (x1 000 stems/ha) Transect 8 A white spruce all other species SG SG LG SG LG SG SG Decay stage initial intermediate advanced B White spruce saplings tall seedlings Wavelet variance Canopy tree basal area (x 10 m 2 /ha) Live Dead 6 Transect 3 A SG LG SG LG SG SG LG SG Wavelet variance Understory regeneration (x1 000 stems/ha) B Position (25 m) 200 Fig Spatial distribution (histograms) of canopy trees (A; diameter at breast height (DBH) 10 cm) and understory regeneration (B; saplings (1 cm < DBH < 10 cm) and tall seedlings (height cm) along contiguous 25 m sections (100 m 2 ) of 2 representative transects (see Fig. 2-2 for reference number). The edge of a canopy gap is encountered (A) when the following criteria are met: 1) transect section is under or next to a significant peak in wavelet position variance of canopy tree basal area, i.e.the black line (wavelet variance) is above dotted line (confidence interval >90% of wavelet variance), 2) basal area of live canopy trees is 25% (i.e. canopy density class D: MRNQ 2011) of a fully stocked balsam fir stand (i.e. 13 m 2 /ha: Zarnovican and Vézina 1985) and 3) basal area of dead canopy trees is above range of natural variability (12-23 m 2 /ha), or 4) the openness index is 60%. Transect sections meeting these gap criteria are identified as SG (i.e. small gap 200 m 2 ) if they covered one or two neighbouring sections or LG (i.e. large gap >200 m 2 ) if they covered more than two neighbouring sections. A section was considered regenerated enough to fill gaps and eventually recreate a closed canopy forests (B) if tall seedling and sapling density was over 1500 stems/ha (Greene et al. 2002). 65

86 2.6. Discussion Disruption of gap regeneration processes and forest degradation Preferential deer browsing has shifted the composition of understory regeneration of balsam fir stands towards a dominance of white spruce. White spruce is now the dominant species in the understory of these forests, while balsam fir and co-occurring species were rare or absent. Considering that balsam fir stands that are not under chronic preferential browsing present an understory dominated by balsam fir (> stems/ha), with white spruce as a secondary species (Côté and Bélanger 1991, Morin and Laprise 1997, McCarthy and Weetman 2007), it is undeniable to say that deer is the major driver of such changes. Even with its new competitive advantage, white spruce regeneration was generally insufficient to fill gaps and eventually recreate a closed canopy forest. Hence, preferential browsing has disrupted gap regeneration process triggering forest degradation processes. Effectively, density of canopy trees was below the range of natural variability of similar fir forests without degradation (Ker 1976, Zarnovican and Vézina 1985, McCarthy and Weetman 2007, Barrette et al. 2010) in all the sampled forests Spatial pattern of forest degradation and pathways of succession Forest degradation in small and large gap patterns covered 47% of the transect area, in an intermixed manner. There were twice as many small gaps as large gaps but small gaps covered 23% less area since they were three times smaller Gap fraction in similar boreal ecosystems without forest degradation is usually two times lower (McCarthy 2001). Forest degradation in small gap patterns ( 200 m²: McCarthy 2001) triggered an alternative successional pathway, mostly toward white spruce parklands. Regeneration was insufficient to fill these small gaps and eventually recreate a closed canopy forest. Forest degradation in larger gap patterns triggered an alternative successional pathway, mostly toward white spruce stands. We chose 1500 stems/ha as the minimal regeneration criterion to consider if gaps were regenerated enough to eventually be filled because it corresponds to the density necessary to reach 50% stocking, i.e. proportion of area actually occupied by trees (Greene et al. 2002). It is highly improbable that forests with 66

87 lower stocking levels will have more than half of their area covered by trees, especially considering the mortality risk of regeneration. We consider such semi-treeless ecosystems would correspond more to parklands than to forests. If so, half of the gap area will follow a successional pathway toward white spruce parklands, triggered by forest degradation primarily in small gap patterns. The other half of the gap area will follow a successional pathway toward white spruce stands, triggered by forest degradation primarily in larger gap patterns. Mid-tolerance of white spruce to shade (Burns and Honkala 1990) could explain why regeneration was usually insufficient in small gaps where the environment is shadier than in large gaps (McCarthy 2001). Notably, white spruce seedlings do not survive more than 2 to 4 years in a shady environment (Raymond et al. 2000). Larger gaps with sufficient density of regeneration lacked big saplings or small trees able to close gaps rapidly. White spruce is vulnerable to overtopping from understory recalcitrant layer which impairs their establishment (Hogg and Lieffers 1991; Lieffers et al. 1993; Cole et al. 2013). Competitive interactions with the recalcitrant understory layer may have contributed to the poor performance of white spruce to fill larger gaps since the cover of such a layer was 6.5 times more abundant in gaps than under canopy. Once established, this layer impairs regeneration, altering rate and shifting direction of succession (Royo and Carson 2006) even after reduction of deer density (Nuttle et al. 2014). Other authors have also recognized white spruce regeneration as an ineffective gap-filler species (Pham et al. 2004, Brassard and Chen 2006, Lieffers et al. 2008). Conversely, balsam fir and paper birch are effective gap filler species playing a crucial role in regeneration processes of forests. Balsam fir has the capacity to create dense and persistent understory seedling banks, highly tolerant to shade and prompt to react and fill gaps (Côté and Bélanger 1991, McCarthy 2001, McCarthy and Weetman 2007) while paper birch can establish itself and fill gaps rapidly from yearly seed production and preexistent seed banks (Foster and King 1986, Bélanger et al. 1993). Forest degradation can be considered as a gradual process since most gaps contained gap makers originating from three different time periods. Hence, gaps probably expanded gradually to actual size following coalescence of the original gap with new neighbouring gap events. Gap coalescence could occur because gap fillers failed to fill old gaps before newer ones occurred in their neighbourhood. Small gaps included the same number of gap makers than in other gap studies (Liu and Hytteborn 1991, Pham et al. 2004, St-Denis et al. 2010). However, large gaps 67

88 contained four times more gap makers, indicating they were probably formed from coalescence of smaller gap events. The gradual forest degradation process has passed a critical state and now is in a phase of acceleration. According to Taylor and MacLean (2007), when a critical state of decline is reached (<40% crown closure), the probability of mortality for remaining canopy trees is dramatically increased due to absence of protection from neighbouring trees and increased wind exposure. Wind-related damage is then increased, accelerating stand break-up. In our study, degradation covers 47% of total area and 70% of the mortality was directly related to wind damage, which is similar to wind-related mortality in other declining balsam fir stands (Ruel et al. 2000, Taylor and MacLean 2007). This phase of acceleration contrasts with the first phase of degradation, which most probably occurred with a time lag induced by the release of high balsam fir regeneration, out of reach of deer, usually present in similar fir stands (McCarthy and Weetman 2007). The rare occurrence of live fir saplings confirms that high regeneration is now ineffective in slowing down forest degradation Implications for resilience of gap-driven forests Chronic preferential browsing by introduced deer disrupted regeneration processes (i.e.resilience mechanism) of gap-driven balsam fir forest by eliminating all effective gap-filler species. White spruce, the only remaining tree species, has not enabled balsam fir forests to recover completely from gap disturbances. Effectively, forest degradation did occur and triggered alternative successional pathways toward white spruce stands and parklands, i.e. partial deforestation. White spruce contributed only partially to regeneration processes of gap-driven forests because of its poor aptitude in filling canopy gaps (Pham et al. 2004, Brassard and Chen 2006, Lieffers et al. 2008). We have shown that white spruce seedling bank was not dense or well distributed and when present, only filled gaps slowly. It was recognized that browsing by native cervids could shift composition of understory regeneration and sometimes cause tree recruitment failure (Gill 1992, Healy 1997, Pedersen and Wallis 2004, Côté et al. 2004). However, understanding of the processes induced by the disruption of gap regeneration process was very limited (van der Wal 2006, Frelich and Reich 2010, Gosse et al. 2011). We have shown that when the unbrowsed tree species are ineffective gap filler, preferential browsing by introduced deer alters 68

89 resilience mechanisms of gap-driven forests by disrupting regeneration processes. We have also shown that such a loss of resilience triggers forests degradation processes and alternative successional pathways toward novel ecosystems. Forest degradation and novel ecosystems, notably parklands (i.e. partial deforestation), have been shown to be a threat to biodiversity and ecosystem services in other ecosystems (Folke et al. 2004, Hobbs et al 2006; Foley et al. 2007). Such threats emanating from browsing by cervids may be exacerbated by climate changes (Frelich and Reich 2010). Climate change will probably allow cervids to go into as yet unoccupied gap-driven ecosystems of the boreal forest where they will contribute to altering its resilience (Miller et al. 2003, Parmesan and Yohe 2003, Côté et al. 2004). Novel conservation issues will also arise in these ecosystems and challenge forest managers implementing ecosystem management. Notably, our findings send a warning signal to forest managers of gap-driven boreal forests with already high densities of moose and growing populations of white-tailed deer (Côté et al 2004, Gosse et al. 2011). In such contexts, our methodological approach using wavelets will be useful for monitoring forest ecosystems subject to or already in a gradual degradation process. In conclusion, maintaining resilience will enable forests to recover or adapt more easily to climate change and cervid introductions, thus limiting forest degradation and alternative successional pathways (Millar et al. 2007, Thrush et al. 2009, Mori 2011). Hence, forest managers will have to make sure that ecosystem management initiatives are efficient in maintaining regeneration processes of gap-driven forests. Evaluating naturalness of managed forests is a promising avenue to ascertain if gap-driven forests are being pushed over resilience thresholds and toward novel ecosystems (Anderson 1991, Hobbs et. al 2006; Bridgewater et al. 2011, Rüdisser et al. 2012) Acknowledgements This study was supported by the Natural Sciences and Engineering Research Council of Canada (NSERC)-Produits forestiers Anticosti Inc. Industrial Chair, the ministère des Ressources naturelles du Québec, the Ordre des ingénieurs forestiers du Québec, and the Foundation de la faune du Québec. We wish to thank C. Boyau, M. Veilleux-Nolin, F. Lebel and C. Marquilly for field and laboratory assistance. We also thank Mathieu Bouchard for comments on a previous 69

90 version of this paper as well as Jean Huot, Jean-Pierre Tremblay and Steeve Côté for enlightening discussions. 70

91 2.8. References Anderson, J. E A conceptual framework for evaluating and quantifying naturalness. Conservation Biology 5: Aakala, T., T. Kuuluvainen, L. De Grandpré, and S. Gauthier Trees dying standing in the northeastern boreal old-growth forests of Quebec: spatial patterns, rates, and temporal variation. Canadian Journal of Forest Research 37: Barrette, M., L. Belanger, and L. De Grandpré Preindustrial reconstruction of a perhumid midboreal landscape, Anticosti Island, Quebec. Canadian Journal of Forest Research 40: Beaupré, P., C. Bédard, C. Dufour, A. Gingras, C. Malenfant, and F. Potvin Plan général d'aménagement intégré des ressources du milieu forestier de l'île d'anticosti. Produits forestiers Anticosti, Québec. Bélanger, L., D. Allard, and P. Meek Dynamique d établissement d un peuplement biétagé de bouleau blanc et de sapin baumier en zone boréale. The Forestry Chronicle 69: Bradshaw, G. A., and T. A. Spies Characterizing canopy gap structure in forests using wavelet analysis. Journal of Ecology 80: Brassard, B. W., and H. Y. H. Chen Stand structural dynamics of North American boreal forests. Critical Reviews in Plant Sciences 25: Bridgewater, P., E. S. Higgs, R. J. Hobbs, and S. T. Jackson Engaging with novel ecosystems. Frontiers in Ecology and the Environment 9:423. Burns, R. M., and B. H. Honkala Silvics of North America: 1. Conifers; 2. Hardwoods. Agriculture Handbook 654. U.S. Department of Agriculture, Forest Service, Washington, DC. vol.2, 877 p. Camarero, J. J., E. Gutierrez, and M. J. Fortin Spatial patterns of plant richness across treeline ecotones in the Pyrenees reveal different locations for richness and tree cover boundaries. Global Ecology and Biogeography 15: Campbell, L. J., and C. P. Laroque Decay progression and classification in two oldgrowth forests in Atlantic Canada. Forest Ecology and Management 238: Chollet, S., and J.-L. Martin Declining woodland birds in North America: should we blame Bambi? Diversity and Distributions 19:

92 Chouinard, A., and L. Filion Detrimental effects of white tailed deer browsing on balsam fir growth and recruitment in a second-growth stand on Anticosti Island, Quebec. Ecoscience 8: Contamin, R., and A. M. Ellison Indicators of regime shifts in ecological systems: What do we need to know and when do we need to know it? Ecological Applications 19: Côté, S., and L. Bélanger Variations de la régénération préétablie dans les sapinières boréales en fonction de leurs caractéristiques écologiques. Canadian Journal of Forest Research 21: Côté, S.D., T. P. Rooney, J. P. Tremblay, C. Dussault, and D. M. Waller Ecological impacts of deer overabundance. Annual Review of Ecology and Systematics 35: Cole, E. C., M. Newton, and A. Youngblood Effects of overtopping on growth of white spruce in Alaska. Canadian Journal of Forest Research 43 : Dale, M. R. T., and M. Mah The use of wavelets for spatial pattern analysis in ecology. Journal of Vegetation Science 9: Danell, K., R. Bergstrom, L. Edenius, and G. Ericsson Ungulates as drivers of tree population dynamics at module and genet levels. Forest Ecology and Management 181: DiTommaso, A., S. H. Morris, J. D. Parker, C. L. Cone, and A. A. Agrawal Deer browsing delays succession by altering aboveground vegetation and belowground seed banks. PLoS ONE 9: e Environment Canada Canadian climate normals, temperature and precipitations, , Québec. Atmospheric Environment Service, Ottawa, Ontario. Foley, J. A., G. P. Asner, M. H. Costa, M. T. Coe, R. DeFries, H. K. Gibbs, E. A. Howard, S. Olson, J. Patz, N. Ramankutty, and P. Snyder Amazonia revealed: forest degradation and loss of ecosystem goods and services in the Amazon Basin. Frontiers in Ecology and the Environment 5: Folke, C., S. Carpenter, B. Walker, M. Scheffer, T. Elmqvist, L. Gunderson, and C. S. Holling Regime shifts, resilience, and biodiversity in ecosystem management. Annual Review of Ecology and Systematics 35: Foster, D. R., and G. A. King Vegetation pattern and diversity in S.E. Labrador, Canada: Betula papyrifera (Birch) forest development in relation to fire history and physiography. Journal of Ecology 74:

93 Frelich, L. E., and C. G. Lorimer Current and predicted long-term effects of deer browsing in hemlock forests in Michigan, USA. Biological Conservation 34: Frelich, L. E., and P. B. Reich Will environmental changes reinforce the impact of global warming on the prairie forest border of central North America? Frontiers in Ecology and the Environment 8: Gill, R A review of damage by mammals in north temperate forests: 1. Deer. Forestry, 65: Gosse, J., L. Hermanutz, B. McLaren, P. Deering, and T. Knight Degradation of Boreal Forests by Non-native Herbivores in Newfoundland's National Parks: Recommendations for Ecosystem Restoration. Natural Areas Journal 31: Greene, D. F., D. D. Kneeshaw, C. Messier, V. Lieffers, D. Cormier, R. Doucet, K. D. Coates, A. Groot, G. Grover, and C. Calogeropoulos Modelling silvicultural alternatives for conifer regeneration in boreal mixedwood stands (aspen/white spruce/balsam fir). The Forestry Chronicle 78: Gunderson, L Ecological resilience--in theory and application. Annual Review of Ecology and Systematics 31: Healy, W. M Influence of deer on the structure and composition of oak forests in central Massachusetts. Pages in W. J. McShea, H. B. Underwood and J. H. Rappole, editors. The science of overabundance: deer ecology and population management. Smithsonian Institution Press, Washington, D.C. Hidding B., J.-P. Tremblay, and S. D. Côté A large herbivore triggers alternative successional trajectories in the boreal forest. Ecology 94: Hobbs, R. J., S. Arico, J. Aronson, J. S. Baron, P. Bridgewater, V. A. Cramer, P. R. Epstein, J. J. Ewel, C. A. Klink, A. E. Lugo, D. Norton, D. Ojima, D. M. Richardson, E. W. Sanderson, F. Valladares, M. Vilà, R. Zamora, and M. Zobel Novel ecosystems: theoretical and management aspects of the new ecological world order. Global Ecology and Biogeography 15: 1-7. Hogg, E. H., and V. J.Lieffers The Relationship between Seasonal-Changes in Rhizome Carbohydrate Reserves and Recovery Following Disturbance in Calamagrostis-Canadensis. Canadian Journal of Botany 69: Holmes, R. L Computer-assisted quality control in tree-ring dating and measurement. Tree-Ring Bulletin 43: Holling, C. S Resilience and stability of ecological systems. Annual Review of Ecology and Systematics 4:

94 James, P. M. A., R. A. Fleming, and M. J. Fortin Identifying significant scale-specific spatial boundaries using wavelets and null models: spruce budworm defoliation in Ontario, Canada as a case study. Landscape Ecology 25: Jobin, L L'arpenteuse de la pruche: un insecte insulaire. Racine 2:5 7. Keitt, T. H., and D. L. Urban Scale-specific inference using wavelets. Ecology 86: Ker, M. F Metric yield tables for the major forest cover types of Newfoundland. Information Report N-X-141. Newfoundland Forest Research Centre, St. John s, Newfoundland. 79 p. Kneeshaw, D. D., and S. Gauthier Old growth in the boreal forest: A dynamic perspective at the stand and landscape level. Environmental Reviews 11: Kuuluvainen, T., K. Syrjänen, and R. Kalliola Logs in a pristine Picea abies forest: occurrence, decay stage distribution and spatial pattern. Ecological Bulletins 49: Lavoie, M., L. Filion, and E. C. Robert Boreal peatland margins as repository sites of long term natural disturbances of balsam fir/spruce forests. Quaternary Research 71: Lieffers, V. I., S. E. Macdonald, and E. H. Hogg Ecology of and Control Strategies for Calamagrostis-Canadensis in Boreal Forest Sites. Canadian Journal Forest Research 23: Lieffers, V.J., G. W. Armstrong, K. J. Stadt, and E. H. Marenholtz Forest regeneration standards: are they limiting management options for Alberta's boreal mixedwoods? The Forestry Chronicle 84: Liu, Q. H., and H. Hytteborn Gap structure, disturbance and regeneration in a primeval Picea-Abies forest. Journal of Vegetation Science 2: Martel, M Analyse dendroécologique des effets de la défoliation par l'arpenteuse de la pruche [Lambdina fiscellaria (Guen.)] sur les populations forestières de l'île d'anticosti, Québec. Mémoire de maîtrise. Université Laval, Québec. McCarthy, J Gap dynamics of forest trees: A review with particular attention to boreal forests. Environmental Reviews 9:1 59. McCarthy, J. W., and G. Weetman Stand structure and development of an insect-mediated boreal forest landscape. Forest Ecology and Management 241:

95 McLaren, B. E., and R. O. Peterson Wolves, moose, and tree-rings on Isle Royale. Science 266: McLaren, B. E., B. A. Roberts, N. Djan-Chekar, and K. P. Lewis Effects of overabundant moose on the Newfoundland landscape. Alces 40: Mi, X., H. Ren, Z. Ouyang, W. Wei, and K. Ma The use of the Mexican Hat and the Morlet wavelets for detection of ecological patterns. Plant Ecology 179:1 19. Millar, C. I., N. L. Stephenson, and S. L. Stephens Climate change and forests of the future: Managing in the face of uncertainty. Ecological Applications 17: Miller, K. V., L. I. Muller, and S. Demarais White-tailed deer (Odocoileus virginianus).pages in G. A. Feldhamer, B. C. Thompson, and J. A. Chapman, editors. Wild mammals of North America: biology, management, and conservation. Johns Hopkins University Press, Baltimore, Maryland. Mitchell, E., and S. J. Popovich Effectiveness of basal area for estimating canopy cover of ponderosa pine, Forest Ecology and Management 95: Mori, A. S Ecosystem management based on natural disturbances: hierarchical context and non-equilibrium paradigm. Journal of Applied Ecology 48: Morin, H., and D. Laprise Seeding bank dynamics in boreal balsam fir forests. Canadian Journal of Forest Research 27: MRNQ Norme de stratification écoforestière. Quatrième inventaire forestier. Ministère des Ressources naturelles et de la Faune. Forêt Québec. Direction des inventaires forestiers. 101 p. Parmesan, C., and G. Yohe A globally coherent fingerprint of climate change impacts across natural systems. Nature 421: Pedersen, B.S., and A. M. Wallis Effects of white-tailed deer herbivory on forest gap dynamics in a wildlife preserve, Pennsylvania, USA. Natural Areas Journal 24: Pham, A.T., L. De Grandpré, S. Gauthier, and Y. Bergeron Gap dynamics and replacement patterns in gaps of the northeastern boreal forest of Quebec. Canadian Journal of Forest Research 34: Potvin, F., P. Beaupré, and G. Laprise The eradication of balsam fir stands by white-tailed deer on Anticosti Island, Québec: a 150-year process. Écoscience 10:

96 Raymond, P., J.-C. Ruel, and M. Pineau Effet d une coupe d'ensemencement et du milieu de germination sur la régénération des sapinières boréales riches de seconde venue du Québec. The Forestry Chronicle 76: Rochette, B., and A. Gingras Inventaire aérien du cerf de Virginie de l île d Anticosti Été Ministère des Ressources naturelles et de la Faune, Direction de l aménagement de la faune de la Côte-Nord. 19 p. + 1 annexe. Rouvinen, S., T. Kuuluvainen, and J. Siitonen Tree mortality in a Pinus sylvestris dominated boreal forest landscape in Vienansalo wilderness, eastern Fennoscandia. Silva Fennica 36: Rosenberg, M. S., and C. D. Anderson PASSaGE: Pattern Analysis, Spatial Statistics and Geographic Exegesis. Version 2. Methods in Ecology and Evolution 2: Royo, A. A., and W. P. Carson On the formation of dense understory layers in forests worldwide: consequences and implications for forest dynamics, biodiversity, and succession. Canadian Journal of Forest Research 36: Rüdisser, J., E. Tasser, and U. Tappeiner Distance to nature A new biodiversity relevant environmental indicator set at the landscape level. Ecological Indicators 15: Ruel, J.-C., R. Doucet, and J. Boily Mortality of balsam fir and black spruce advance growth 3 years after clear-cutting. Canadian Journal of Forest Research 25: Ruel, J.-C., C. P. Quine, S. Meunier, and J. Suarez Estimating windthrow risk in balsam fir stands with the ForestGales model. The Forestry Chronicle 76: Saucier, J.-P., P. Grondin, A. Robitaille, J. Gosselin, C. Morneau, P. J. H. Richard, J. Brisson, L. Sirois, A. Leduc, H. Morin, É. Thiffault, S. Gauthier, C. Lavoie, and S. Payette Écologie forestière. Pages in Ordre des ingénieurs forestiers du Québec, Manuel de foresterie, 2e éd. Ouvrage collectif, Éditions MultiMondes, Québec. SAS Institute Inc The SAS system for Windows, version 9.1. SAS Institute Inc. Cary, NC, USA. Sasaki, N., and F. E. Putz Critical need for new definitions of forest and forest degradation in global climate change agreements. Conservation Letters 2: Simard, M.J., Y. Bergeron, and L. Sirois Conifer seedling recruitment in a southeastern Canadian boreal forest: the importance of substrate. Journal of Vegetation Science 9:

97 St-Denis, A., D. D. Kneeshaw, and Y. Bergeron The role of gaps and tree regeneration in the transition from dense to open black spruce stands. Forest Ecology and Management 259: Taylor, S. L., and D. A. MacLean Dead wood dynamics in declining balsam fir and spruce stands in New Brunswick, Canada. Canadian Journal of Forest Research 37: Thrush, S. F., J. E. Hewitt, P. K. Dayton, G. Cocol, A. M. Lohrer, A. Norkko, J. Norkko, and M. Chiantore Forecasting the limits of resilience: integrating empirical research with theory. Proceedings of the Royal Society B: Biological Sciences 276: Tremblay, J.-P., J. Huot, and F. Potvin Density-related effects of deer browsing on the regeneration dynamics of boreal forest. Journal of Applied Ecology 44: van der Wal, R Do herbivores cause habitat degradation or vegetation state transition? Evidence from the tundra. Oikos 114: Walker, B., C. S. Holling, S. R. Carpenter, and A. Kinzig Resilience, adaptability and transformability in social ecological systems. Ecology and Society 9:5. Westerberg, D Profitable forestry methods maintaining biodiversity as an integral part of Swedish forestry. Pages in C. R. Bamsey, editor. Innovative silviculture systems in boreal forests, a symposium held in Edmonton Alberta, Canada, Oct. 2 8, Clearlake Ltd., Edmonton, Alberta. Yamaguchi, D. K A simple method for cross-dating increment cores from living trees. Canadian Journal of Forest Research 21: Zarnovican, R., and P.-E. Vézina Étude de la production totale : exemple d'une table pour le sapin baumier. Canadian Journal of Forest Research 15:

98

99 Chapitre 3: Cumulative effects of chronic deer browsing and clear-cutting on regeneration processes in second-growth white spruce stands 3.1 Résumé Les interactions entre de multiples perturbations peuvent altérer les mécanismes de résilience des forêts engendrant des trajectoires successionnelles alternatives. Les processus de régénération sont un des principaux mécanismes de résilience des forêts parce qu ils déterminent les trajectoires successionnelles. Sur l île d Anticosti, le broutement chronique par le cerf de Virginie (Odocoileus virginianus), une espèce introduite, a changé la composition de la régénération des sapinières surannées vers une dominance de l épinette blanche (Picea glauca). La coupe totale historique de ces sapinières altérées a formé des pessières blanches de seconde venue maintenant mature. Toutefois, l effet cumulatif de la coupe totale et du broutement préférentiel de la régénération par le cerf sur les processus de régénération de ces pessières blanches n a pas encore été évalué. Notre objectif est de déterminer si les processus de régénération permettent aux pessières blanches de récupérer de l effet cumulatif de ces deux perturbations. Nous avons étudié la régénération en relation avec la disponibilité des semences, la disponibilité des substrats d établissement des semis et la convenance des substrats dans les pessières blanches de secondes venues matures et dans les coupes récentes des pessières blanches de secondes venues matures. Nos résultats indiquent une défaillance de la régénération dans les deux types d écosystèmes, laquelle peut être expliquée par un manque de débris ligneux pourris adéquats pour l établissement de suffisamment de semis d épinette blanche. Ainsi, l effet cumulatif du broutement préférentiel de la régénération par le cerf et de la coupe totale de ces pessières blanches altère les processus de régénération des forêts et engendre une trajectoire successionnelle alternative vers la formation de forêts-parcs d épinettes blanches, i.e. une déforestation partielle. Nous proposons que la réalisation de coupes partielles créant des débris ligneux devrait être investiguée pour maintenir les pessières blanches sans avoir à planter. 79

100 3.2. Abstract Interactions between multiple disturbances can alter resilience mechanisms, thereby triggering alternative successional pathways. Regeneration processes are important mechanisms of forest resilience because they drive successional pathways. On Anticosti Island, chronic browsing by introduced white-tailed deer (Odocoileus virginianus) shifted composition of understory regeneration of overmature balsam fir (Abies balsamea) forests toward dominance by white spruce (Picea glauca). Historic clear-cutting of these altered forests generated mature secondgrowth white spruce stands. However, the cumulative effect of chronic deer browsing and recent clear-cutting on regeneration processes of mature second-growth white spruce stands has not yet been evaluated. Our objective is to evaluate if regeneration processes would enable white spruce stands to recover from the cumulative effects of these two disturbances. We studied regeneration in relation to seed availability, substrate suitability for seedling establishment, and substrate availability in mature second-growth white spruce stands and recent clear-cuts of mature secondgrowth white spruce stands. Our results indicate regeneration failure in both ecosystems, which can be explained by a lack of suitable rotten logs for sufficient establishment of white spruce seedlings. Hence, the cumulative effects of chronic deer browsing and clear-cutting of mature second-growth white spruce stands have altered regeneration processes and triggered an alternative successional pathway toward parklands, i.e., partial deforestation. We propose shelterwood cuttings that create nurse logs should be investigated to maintain white spruce stands without planting. 80

101 3.3. Introduction Interactions between multiple disturbances can alter resilience mechanisms of forests, thereby triggering alternative successional pathways (Paine et al., 1998; Tremblay et al., 2007; Buma and Wessman, 2011). Resilience is the capacity of a system to absorb disturbance and reorganize so that the same structure and functions are essentially recovered, e.g., forest recovering to forest following perturbation (Holling, 1973; Gunderson, 2000). Regeneration processes are important mechanisms of forest resilience because they drive successional pathways (Buma and Wessman, 2011; Hidding et al., 2013). For example, cervids such as white-tailed deer (Odocoileus virginianus) can shift the composition of understory regeneration by selectively browsing seedlings (Tremblay et al., 2007; Nuttle et al., 2014), while forest management practices can change the composition and density of regeneration through diverse silvicultural scenarios (Boucher et al., 2009; Lundmark et al., 2013; Rist and Moen, 2013). Interaction between these two disturbances could potentially alter regeneration processes, triggering alternative successional pathways that move the system toward assemblages of species that have not cooccurred historically. The occurrence of such novel ecosystems can represent a threat to biodiversity and ecosystem services that are provided by preindustrial forests (Hobbs et al., 2006; Bridgewater et al., 2011). In this context, a recent cause for concern is increasing cervid densities in the forests of North America (Côté et al., 2004; McLaren et al., 2004; Chollet and Martin, 2012), which increases the probability for potential interactions between cervid browsing and forest management. Following the introduction of 200 white-tailed deer into the forests of Anticosti Island (Quebec) at the end of the 19th century, their population densities have now reached more than 20 individuals km 2 in the absence of predators (Rochette and Gingras, 2007). Since the 1920s, chronic deer browsing has shifted the understory composition of balsam fir forests on the island from regeneration that is dominated by balsam fir (Abies balsamea (L.) Miller) towards one that is dominated by white spruce (Picea glauca (Moench) Voss), together with a recalcitrant ground layer that includes blue joint grass (Calamagrostis canadensis (Michaux) P.Beauv.), bracken fern (Pteridium aquilinum (L.) Kuhn), and thistle (Cirsium spp.), which are species that are not browsed by deer (Potvin et al., 2003; Tremblay et al., 2007; Hidding et al., 2013). Recalcitrant understory layers are dense and persistent monodominant strata that can impede regeneration, by 81

102 altering the rate and shifting the direction of forest succession (Royo and Carson, 2006). Such a layer can persist even after reductions in deer densities (Nuttle et al., 2014). Historically, the unaltered forests of Anticosti Island were driven by gap disturbance and gap-filling processes that assured the maintenance of overmature balsam fir stands, mixed with white spruce and paper birch (Betula papyrifera Marshall; Barrette et al., 2010). Balsam fir regeneration created dense and persistent understory seedling banks, which were highly tolerant of shade and that were prompt to react and fill canopy gaps (Côté and Bélanger, 1991; McCarthy, 2001; McCarthy and Weetman, 2007a). From the 1920s onwards, white spruce has represented the sole remaining tree species that is available to recreate a forest after disturbances. In such a context, white spruce regeneration formed white spruce stands following historic clear-cutting ( ) of altered balsam fir stands (Fig. 3-1; Beaupré et al., 2004). These now mature second-growth white spruce stands represent a threat to biodiversity and the services that were provided by the preindustrial forest (Barrette et al., 2010), since they are novel ecosystems which are not commonly found in eastern North America (Stiell, 1976; Bell et al., 1990; Lieffers et al., 2008; Saucier et al., 2009). Many mature second-growth white spruce stands have been recently clear-cut (1999) under a system of timber production that uses even-aged management. However, the cumulative effects of chronic deer browsing and clear-cutting on regeneration processes within white spruce stands has yet to be evaluated. If regeneration processes cannot enable the recovery of white spruce stands from these two disturbances, then alternative successional pathways could develop and lead towards deforestation, since white spruce is the only remaining tree species. The occurrence of a non-forest state would represent an even greater threat to biodiversity and ecosystem services. To avoid deforestation and meet sustainable forest management objectives, forest managers would then have to resort to plantation silviculture. 82

103 Understory of mature second-growth white spruce stand Recent clear-cut of mature second-growth white spruce stand Understory of overmature balsam fir stand Gap in overmature balsam fir stand Fig Photographs showing the four ecosystem types of the study. Consequently we investigated if regeneration processes would enable white spruce stands to recover from the cumulative effects of chronic deer browsing and clear-cutting. To meet this objective, we studied regeneration in relation to seed availability, substrate suitability for seedling establishment, and substrate availability in mature second-growth white spruce stands and in recent clear-cuts of mature second-growth white spruce stands. We also studied regeneration processes in the understory and within canopy gaps of overmature balsam fir stands to provide a baseline of regeneration characteristics which have enabled fir stands to recover partially from historic clear-cuts and reproduce white spruce stands. We predicted that regeneration in mature second-growth white spruce stands would not be sufficient for them to recover from clear-cutting because of a lack of suitable substrates for establishment. Effectively, white spruce requires specific substrates for optimum seedling establishment (e.g., large woody 83