Du processus de sélection d`habitat à la survie des proies dans le
Transcription
Du processus de sélection d`habitat à la survie des proies dans le
N° d’ordre _ _ _ _ Année 2010 THESE présentée devant l’UNIVERSITÉ DE SAVOIE pour l’obtention du DIPLOME DE DOCTORAT (arrêté du 7 août 2006) présentée et soutenue publiquement le 21 juin 2010 par Mr VINCENT TOLON Du processus de sélection d’habitat à la survie des proies dans le paysage du risque : implications pour l’exploitation de la faune sauvage. Directeurs de thèse : Anne Loison (Université de Savoie) & Stéphane Dray (UCB Lyon 1) JURY : Mr Claude Miaud (Université de Savoie) Mr Stefano Focardi (ISPRA) Mr Mark Hewison (INRA CEFS) Mr Dominique Allainé (UCB Lyon 1) Mr Eric Baubet (ONCFS CNERA-CS) Mlle Anne Loison (Université de Savoie) Mr Stéphane Dray (UCB Lyon 1) Président Rapporteur Rapporteur Examinateur Invité Directrice Co-directeur Laboratoires d’accueils UMR CNRS 5553 « Laboratoire d'Ecologie Alpine » Université de Savoie Bat. Belledonne 73376 Le Bourget-du-Lac FRANCE & UMR CNRS 5558 « Laboratoire de Biométrie et Biologie Evolutive » Université Claude Bernard - Lyon 1 43, bd du 11 Novembre 1918 69622 Villeurbanne Cedex Financements Office National de la Chasse et de la Faune Sauvage – CNERA Cervidés et Sanglier Fédérations Départementales des Chasseurs de l’Ain et de la Haute Savoie. 1 Du processus de sélection d’habitat à la survie des proies dans le paysage du risque : implications pour l’exploitation de la faune sauvage. Résumé : Face aux variations spatiotemporelles du risque de prédation, les proies peuvent moduler leur distribution, sélectionner leur habitat ou se déplacer à différentes échelles de façon à accroître leur survie. Ces réponses adaptatives devraient néanmoins être fortement conditionnées par les capacités des proies à percevoir le risque dans l’espace, ainsi que par les contraintes et processus de mouvement sous-jacents. Cette thèse propose d’étudier l’interaction des aspects ultimes et proximaux du comportement de sélection d’habitat dans le paysage du risque. Le modèle biologique étudié est le sanglier Sus scrofa L., une espèce chassée posant d’importants problèmes de gestion en Europe. En préambule de ce travail, nous rappelons à travers une réflexion méthodologique, la nécessité de l’adéquation entre les outils télémétriques, les analyses statistiques et les questions biologiques lorsque la sélection d’habitat est étudiée sur la faune sauvage. Pour répondre à la problématique énoncée, la suite des travaux se base sur des localisations télémétriques récoltées sur des sangliers vivant autour d’une zone protégée du Bassin Genevois (France, Suisse) ainsi que des localisations GPS de chiens de chasse effectuées lors de battues. Nous démontrons que les déplacements des centres d’activités, en réponse aux variations journalières et saisonnières du risque, sont plus importants à proximité de la réserve et reflètent donc l’échelle de structuration spatiale sous-jacente du risque. Nous mettons ensuite en évidence comment, à l’approche de la zone protégée, les patrons multivariés de sélection d’habitat des sangliers se dissocient progressivement de ceux des chiens de chasse locaux. Cette dissociation aboutit à un report de la mortalité adulte sur les classes les plus jeunes lorsque les adultes trouvent refuges dans la zone protégée. L’étude des mouvements inter-journaliers des sites de repos des sangliers en relation avec le paysage du risque, révèlent que les animaux sélectionnent directement, d’un site à l’autre, les zones les moins chassées même lorsque aucune activité de chasse n’est manifeste lors de ce choix. Ceci suggère l’utilisation de sources d’informations indirectes par le sanglier (ex: odeur, mémorisation, processus sociaux) afin à estimer le niveau de risque dans l’espace. A la lumière de ces résultats, nous explorons finalement différents scénarios théoriques de dynamique de population dans une optique appliquée. Nous montrons comment l’application d’une pression d’exploitation raisonnée, couplée à la création d’un biais de perception du risque chez les individus, permettrait de restaurer le contrôle des populations qui, en rejoignant des refuges, échappent à l’exploitation. Mots clefs : risque, mouvement, survie, habitat, chasse, réserve, sanglier. 2 From habitat selection processes to prey survival in the landscape of risk: implication for wildlife harvesting. Abstract: In response to spatial and temporal variations of the predation risk, prey can modulate their distribution, select their habitat or move at different scales to increase their survival. However, these adaptive responses might be strongly constrained by the ability of prey to perceive risk levels in space, and by the underlying movement constraints and processes. This work aims to study interrelations between ultimate and proximal aspect of habitat selection strategies in the landscape of risk of a game species, the wild boar Sus scrofa L., known to be source of numerous management troubles in Europe. To initiate this work, we highlighted, through a methodological reflection, the necessity of a good fit between telemetry tools, statistical analyses and biological question when studying habitat selection on free ranging animals. Our field study relies on telemetric locations dataset recorded on wild boar living in the vicinity of a protected area of the Geneva basin (France, Switzerland), and on GPS tracking of hunting dogs during drive hunts. First, we showed that shifts of activity centres, in response to daily and seasonal risk variations on the whole study site, were the highest in the vicinity of the protected areas and reflected therefore the underlying scaling structure of the landscape of risk. Then, we showed how the increased proximity of the protected area leads to the progressive dissociation of both wild boar and hunting dogs’ multivariate habitat selection patterns. This mismatch led to a switch of mortality from adults to younger age classes when adults found refuge in the protected perimeter. By studying the inter-day movements of wild boar resting sites, in relation to the structure of the landscape of risk, we showed that individuals selected directly the lowest hunted areas even if no hunting occurred at dawn when animals performed this habitat choice. This suggested the use of indirect information by wild boar (e.g. olfactory cues, memorisation, social processes) to asses risk levels in space. We finally studied, in an applied context, different theoretical scenarios of population dynamic through the light of above mentioned results. We showed how the use of a moderate harvesting pressure, combined with the creation of a perceptual bias about risk levels in animals, could help to restore the control of population escaping from harvesting in refuges. Key words: risk, movement, survival, habitat, hunting, reserve, wild boar. 3 Liste des publications réalisées pendant la thèse Articles scientifiques Tolon, V., S. Dray, A. Loison, A. Zeileis, C. Fischer, and E. Baubet. 2009. Responding to spatial and temporal variations in predation risk: space use of a game species in a changing landscape of fear. Canadian Jounal of Zoology 87(12):1129-1137. Tolon, V., J. Martin, S. Dray, A. Loison, C. Fischer, and E. Baubet. Impact of refuges on predator-prey spatial interactions and prey mortality: wildlife harvesting as a case study. Soumis dans Oikos. Tolon, V., B. van Moorter, S. Dray, A. Loison, C. Fischer, and E. Baubet. In prep. Moving in a landscape of risk: the role of directed and non-directed habitat selection processes. Saïd, S., V. Tolon, S. Brandt, and E. Baubet. In prep. Sex effect on habitat selection in response to hunting disturbance: the study of Wild boar. ∆ Richomme, C., E. Afonso, V. Tolon, C. Ducrot, L. Halos, A. Alliot, C. Perret, M. Thomas, P. Boireau & E. Gillot-Fromont. 2010. Seroprevalence and factors associated with Toxoplasma gondii infection in wild boar (Sus scrofa) in a Mediterranean island. Epidemiology and Infection, doi:10.1017/S0950268810000117* Chapitre d’ouvrage scientifique Martin, J., V. Tolon, B. Van Moorter, M. Basille, and C. Calenge. 2009. On the use of telemetry in habitat selection studies .in D. Barculo and J. Daniels, editors. Telemetry: research, technology and applications Nova Science Publishers Inc. Articles de gestion Tolon, V. & E. Baubet. 2008. L’effet réserve sur l’occupation de l’espace du sanglier. Foret Wallonne, 92: 15-25* ∆ publications présentées en annexe. publications non présentées dans ce manuscrit. * 4 5 6 À mon fils... 7 8 SOMMAIRE 1. INTRODUCTION ...................................................................................................... 13 1.1. Sciences du comportement et exploitation des especes......................................... 14 1.1.1. L’homme en tant que prédateur................................................................ 14 1.1.2. L’exemple de l’effet réserve ...................................................................... 15 1.1.3. L’étude du comportement : une clef pour la gestion de la faune sauvage.15 1.2. Des causes ultimes aux causes proximales du comportement.............................. 17 1.2.1. La théorie de l’évolution : un dogme ou une heuristique ? ...................... 17 1.2.2. Répercussions dans les sciences du comportement .................................. 18 1.2.2.a. L’écologie comportementale à l’épreuve................................................. 18 1.2.2.b. De la critique à l’émergence de nouveaux champs théoriques ................ 19 1.2.2.c. L’approche intégrative des comportements ............................................. 19 9 1.3. Survivre dans le paysage de la peur ....................................................................... 21 1.3.1. L’évitement comme premier rempart à la prédation ................................ 21 1.3.2. Le compromis énergie-risque : manger sans être mangé......................... 22 1.3.3. Stratégies de sélection d’habitat risque-dépendantes .............................. 23 1.3.3.a. Importance de l’échelle spatiale............................................................... 23 1.3.3.b. Echelle populationnelle : Distribution Libre Idéale................................. 24 1.3.3.c. Echelle individuelle.................................................................................. 25 1.3.4. Paysage de la peur vs. paysage du risque ................................................ 27 1.3.4.a. Le paysage de la peur............................................................................... 27 1.3.4.b. Le paysage du risque................................................................................ 30 1.4. Importance de la perception du risque .................................................................. 32 1.4.1. Rôle de l’information dans l’évitement du risque de prédation ............... 32 1.4.1.a. Valeur adaptative de l’information .......................................................... 32 1.4.1.b. Types d’information ................................................................................ 32 1.4.2. Les contraintes de l’information dans le paysage du risque .................... 33 1.4.2.a. Incertitude ou fiabilité.............................................................................. 33 1.4.2.b. Coût de l’information............................................................................... 34 1.4.2.c. Acquisition de l’information .................................................................... 34 1.4.2.d. Limitation................................................................................................. 35 1.4.2.e. Effet sur la performance........................................................................... 35 1.5. Le mouvement dans le paysage du risque ............................................................. 37 1.5.1. Le mouvement comme processus de sélection d’habitat .......................... 37 1.5.2. L’écologie du mouvement ......................................................................... 37 1.5.2.a. Les composantes du mouvement ............................................................. 38 1.5.2.b. Une approche intégrative ......................................................................... 40 1.5.3. Le mouvement comme processus d’évitement du risque .......................... 40 1.5.3.a. Direction des mouvements et sélection d’habitat..................................... 40 1.5.3.b. De la « fuite aveugle » à l’orientation vers les refuges............................ 41 1.6. Objectif de la thèse................................................................................................... 43 10 2. MODELE D’ETUDE : LE SANGLIER Sus Scrofa L............................................. 49 2.1. Une espèce problematique....................................................................................... 50 2.2. Le sanglier dans l’espace et le temps...................................................................... 51 2.2.1. Utilisation de l’espace et habitat .............................................................. 51 2.2.1.a. Influence du système d’appariement........................................................ 51 2.2.1.b. Habitat...................................................................................................... 52 2.2.2. Revue des paramètres d’utilisation de l’espace ....................................... 53 2.3. Le sanglier dans le paysage du risque.................................................................... 55 2.3.1. Variations spatiales et temporelles du risque........................................... 55 2.3.1.a. Temporalité : le sanglier: un nocturne récent?......................................... 55 2.3.1.b. Effet de la chasse sur le domaine vital..................................................... 55 2.3.1.c. Paysage du risque : l'effet réserve ............................................................ 56 2.3.2. Sélection d’habitat et distribution risque-dépendante.............................. 56 2.3.2.a. Evidence du compromis énergie-risque ................................................... 56 2.3.2.b. Importance de l’échelle spatiale et temporelle......................................... 57 3. MATERIEL ET METHODES .................................................................................. 61 3.1. Aire d’étude : le bassin Genevois............................................................................ 62 3.1.1. Une mosaïque de méthodes de gestion ..................................................... 62 3.1.1.a. Présentation .............................................................................................. 62 3.1.1.b. Modes de gestion ..................................................................................... 63 3.1.1.c. L’étude des sangliers du Bassin Genevois ............................................... 64 3.1.2. Récolte des données .................................................................................. 65 4. Partie 1 : CONSIDERATIONS METHODOLOGIQUES...................................... 69 5. Partie 2 : REPONSES AUX VARIATIONS SPATIALES ET TEMPORELLES DU RISQUE .................................................................................................................. 103 6. Partie 3 : RELATION ENTRE SURVIE ET SELECTION D’HABITAT ......... 129 11 7. Partie 4 : LE MOUVEMENT DANS LE PAYSAGE DU RISQUE..................... 155 8. Partie 5 : EXPLOITATION DES ESPECES ET EFFET REFUGE ................... 179 9. SYNTHESE ............................................................................................................... 201 9.1. Résumés étendus .................................................................................................... 202 9.2. Implications pour la gestion.................................................................................. 213 9.2.1. Synthèses des résultats............................................................................ 213 9.2.2. Réduire l’effet réserve : l’importance de l’information.......................... 214 9.2.2.a. Erreur de perception et système source-puits......................................... 214 9.2.2.b. Risque et dérangement........................................................................... 215 9.2.2.c. Gestion par l’information....................................................................... 216 9.2.3. Implications pour la gestion du sanglier ................................................ 217 9.2.3.a. Mesures de gestion potentielles ............................................................. 217 9.2.3.b. Importance des contextes écologique et humain ................................... 219 9.3. Perspectives ............................................................................................................ 222 9.3.1. La nécessité d’approches expérimentales............................................... 222 9.3.2. Etude des systèmes sans réserve ............................................................. 224 9.3.3. Utilisation des refuges et réponse fonctionnelle..................................... 225 9.4. Conclusion .............................................................................................................. 229 Remerciements .............................................................................................................. 237 REFERENCES.............................................................................................................. 237 ANNEXE 1..................................................................................................................... 237 12 1. INTRODUCTION 13 1.1. SCIENCES DU COMPORTEMENT ET EXPLOITATION DES ESPECES 1.1.1. L’homme en tant que prédateur. Si l’exploitation des espèces animales fût à l’origine uniquement pratiquée dans un but alimentaire (ex: sociétés de chasseurs-cueilleurs, Surovell et al. 2005), elle est aujourd’hui également réalisée dans une optique commerciale (ex: pêche, Gell and Roberts 2003, Neubert 2003), récréative (ex: chasse au trophée, Coltman et al. 2003) ou dans une nécessité de gestion de l’environnement (ex: espèces invasives, Carter and Leonard 2002). Quel que soit le but recherché, l’homme se trouve toujours dans la nécessité de disposer de moyens efficaces d’exploitation et de gestion de la faune sauvage. D’un point de vue biologique, l’exploitation de la faune sauvage peut néanmoins se résumer par la création d’une nouvelle source de mortalité chez les espèces concernées. L’homme peut tenir le rôle de prédateur pour l’espèce qu’il exploite, particulièrement lorsque ses prédateurs naturels sont absents dans l’environnement. Or, en réponse à la pression de sélection que constitue la prédation, les proies sont capables de développer une très grande diversité de stratégies anti-prédatrices leur permettant de réduire l’effet de ce risque sur leurs fitness individuelles. De façon analogue aux systèmes prédateur-proie, les exploitants de la faune sauvage entrent donc en profonde interaction avec les espèces exploitées, puisque chacun de ces acteurs devrait tendre à adopter de nouvelles stratégies en réponse à celles développées par l’autre. Si les capacités technologiques de l’homme lui confèrent souvent un avantage important aboutissant parfois à l’épuisement des ressources exploitées (ex: sur pêche, Jackson et al. 2001), il arrive à l’inverse que des espèces échappent à toute possibilité de gestion voire d’exploitation en adoptant une stratégie difficile à contrecarrer. 14 1.1.2. L’exemple de l’effet réserve Face à l’érosion de la biodiversité et à l’épuisement des ressources naturelles, le développement de réseaux de réserves assurant divers degrés de protection aux espèces a largement été préconisé durant ces dernières décennies (Diamond 1975, Simberloff and Abele 1982) et a souvent démontré son efficacité (Novaro et al. 2000, Gell and Roberts 2003). Néanmoins, ces réserves créent de nouvelles problématiques de gestion puisque bon nombre d’espèces exploitées sont capables de les utiliser afin d’échapper au risque de prélèvement (ex: Fox and Madsen 1997, Madsen 1998). Cet effet refuge peut entraîner une baisse des prélèvements et une augmentation de la densité d’espèces exploitées dans les réserves, ce qui n’est pas sans conséquences sur le fonctionnement de l’écosystème protégé (Lenihan et al. 2001, Ripple and Beschta 2004). Dans le cas d’espèces invasives, l’éradication ou la réduction de la densité peut devenir difficile à réaliser si elles s’établissent dans des zones protégées (Usher 1988, Pyšek et al. 2002). Il devient alors urgent de comprendre l’intérêt adaptatif que présentent ces aires protégées pour les espèces exploitées, les mécanismes permettant à ces dernières de les utiliser, ainsi que les conséquences de ce comportement sur les dynamiques des populations. L’écologie comportementale peut donc contribuer activement à l’amélioration de la gestion des espèces exploitées échappant, de façon problématique, au contrôle de l’homme lorsqu’elles utilisent des espaces protégés. 1.1.3. L’étude du comportement : une clef pour la gestion de la faune sauvage. D’un point de vue théorique, les comportements peuvent d’abord être abordés à travers leurs aspects proximaux, en s’approchant au plus près des causes immédiates de leur apparition (ex: neurobiologie, cognition, ontogenèse - c.f. la question du "comment" de Tinbergen 1963). Il est également possible d’aborder les comportements à travers 15 leurs aspects ultimes, en les étudiant dans une optique évolutive (ex: adaptation, optimalité, évolution, phylogénie - c.f. la question du "pourquoi" de Tinbergen 1963). Ces deux approches correspondent à deux facettes indissociables du comportement. Si une proie rejoint un refuge, c’est probablement qu’elle a perçu certains stimuli, qu’elle les a associés à un risque, que sa motivation à survivre est conséquente et qu’elle possède la capacité motile et cognitive de s’y rendre (mécanisme proximal). D’un autre point de vue, elle peut aussi rejoindre ce refuge en réponse à la pression de sélection crée par la prédation, qui tend à éliminer les individus n’adoptant pas ce comportement (cause ultime). Le lien entre ces deux approches apparaît si l’on considère que mécanismes proximaux de la proie (motivation, capacités sensorielles, motiles et cognitives) sont en partie conditionnés par la pression de sélection subie par ses ancêtres, au même titre qu’elles influencent en retour les possibilités d’adaptation de la proie sur le long terme. Dans une optique de gestion de l’effet réserve, l’utilisation des deux approches semble fondamentale. Si l’attractivité des aires protégées sur les espèces exploitées peut être expliquée par des causes ultimes (ex : bénéfice en termes de survie), l’approche proximale permet de comprendre comment les individus s’y rendent. Considérer les mécanismes proximaux et les causes ultimes des comportements devrait donc permettre de trouver les clefs permettant de sortir de cette problématique. 16 1.2. DES CAUSES ULTIMES AUX CAUSES PROXIMALES DU COMPORTEMENT 1.2.1. La théorie de l’évolution : un dogme ou une heuristique ? La publication de « L’origine des espèces » par Darwin en 1859 a confronté les biologistes à une des théories les plus productives et fondatrices que la discipline ait pu rencontrer depuis ses débuts : la théorie de l’évolution. Le principe de sélection naturelle y est en effet proposé comme mécanisme d’adaptation des organismes à l’environnement, comme moteur de l’évolution et finalement comme explication de la diversité et de l’organisation du vivant. En 150 ans d’existence, ce raisonnement a largement montré son formidable potentiel heuristique, prédictif et explicatif (c.f. "Nothing in biology makes sense except in the light of evolution", Dobzhansky 1973). Avec le principe de sélection naturelle, les biologistes ont été confrontés à la notion d’optimalité et d’adaptation. Parmi toute la variabilité existante sur un caractère, le processus de sélection naturelle devrait théoriquement faire émerger celui offrant le meilleur succès évolutif (optimalité) en réponse aux conditions environnementales (adaptation). La tentation fut forte de tomber dans le « tout adaptatif », en postulant que chaque caractère observé est une adaptation à l’environnement, ou est optimal, puisque soumis à la sélection naturelle. Sur la fin du XXème siècle, d’importantes critiques, dont le porte drapeau fut sans conteste Stephen Jay Gould (e.g. Gould and Lewontin 1979), sont apparues contre une certaine radicalisation de la pensée évolutionniste. Ces critiques insistent notamment sur l’importance de la dérive génétique (processus aléatoire) et la 17 corrélation entre traits (tous ne peuvent être optimaux) dans les processus évolutifs. Ce débat a alors poussé les évolutionnistes contemporains à recentrer l’optimalité sur son rôle heuristique, en rappelant que ce concept sert avant tout de référence théorique pour comprendre les mécanismes d’adaptation, et non à prouver que la sélection naturelle produit des solutions optimales (ex: Parker and Maynard Smith 1990). 1.2.2. Répercussions dans les sciences du comportement 1.2.2.a. L’écologie comportementale à l’épreuve Cet historique du raisonnement évolutionniste de la fin du XXème siècle a été particulièrement visible dans les sciences du comportement. Suite à l’essor de l’éthologie dont Konrad Lorenz, Nikolaas Tinbergen et Karl von Frisch furent les pionniers dans les années cinquante, certains biologistes ont commencé à considérer les comportements comme des traits pouvant être soumis à la sélection naturelle. De cette idée a émergé l’écologie comportementale, qui tend à confronter l’observation des comportements à des prédictions issues d’un raisonnement théorique réalisé sous l’hypothèse d’optimalité. Cette discipline proposa une explication évolutive pour beaucoup de comportements, comme les stratégies d’approvisionnement (e.g. Charnov 1976), les règles de distribution, le choix d’habitat (Fretwell 1972) ou encore les stratégies anti-prédatrices (e.g. Hamilton 1971). Néanmoins, la même déviance du « tout adaptatif » a provoqué de sévères critiques inspirées de celles de Gould et Lewontin (1979). La plus retentissante fut celle de Pierce et Ollason (1987) qui publièrent un article intitulé de façon provocatrice «Eight reasons why optimal foraging theory is a complete waste of time » (Traduction : « Huit raisons pour lesquelles la théorie d’approvisionnement optimal est une perte de temps totale »). Les auteurs insistent sur le fait que l’optimalité peut ne jamais être atteinte (maximum locaux du paysage adaptatif, Wright 1932) et que les animaux ne sont pas « manufacturés » (« designed » en anglais), c’est-à-dire qu’on ne peut pas les considérer 18 comme un ensemble de composants tous optimisés pour atteindre la meilleure performance possible. 1.2.2.b. De la critique à l’émergence de nouveaux champs théoriques Un constat, au premier abord assez sévère pour l’écologie comportementale, s’est imposé à l’issue de ce débat : le comportement des animaux ne peut être optimal. Toutefois, plutôt que de constituer un frein à la discipline, ce constat a été source de nouvelles réflexions et a amené les acteurs de l’écologie comportementale à prendre en compte, puis à étudier précisément cette « imperfection » des individus. Parmi les postulats les plus répandus en écologie comportementale, figurent l’hypothèse d’omniscience (les individus ont une connaissance parfaite de l’environnement) et l’hypothèse d’absence de coût des déplacements (e.g. Fretwell and Lucas 1969). Ces hypothèses furent souvent contestées (e.g. Kennedy and Gray 1993) et certaines modifications furent apportées dans les modèles afin de prendre en compte ces imperfections (e.g. Bernstein et al. 1991, Ranta et al. 1999). Mais, au-delà d’une complexification des modèles préexistants, c’est l’imperfection des individus elle-même qui devint l’objet d’intérêt le plus profond. Ainsi, certains écologues ont récemment mis en lumière l’importance de l’information dans les décisions prises par les animaux, et proposent de traiter l’utilisation de l’information dans une optique évolutive en considérant sa fonction adaptative (Danchin et al. 2004, Dall et al. 2005). En parallèle se développe une nouvelle discipline intégrative combinant l‘optimalité, la cognition, la biomécanique et les processus aléatoires, afin d’étudier le rôle du mouvement dans la réponse des animaux à leur environnement : l’écologie du mouvement (Holden 2006, Holyoak et al. 2008, Nathan 2008, Nathan et al. 2008). 1.2.2.c.L’approche intégrative des comportements Ainsi, de critiques faites sur des hypothèses irréalistes, l’écologie comportementale a fait de nouveaux paradigmes émergeant naturellement des réflexions 19 réalisées précédemment. Dans cette démarche, les causes ultimes (évolutives) et proximale (mécanistiques) des comportements tendent à se réunir pour former une approche probablement plus réaliste des comportements (Fig. 1.1) tout en conservant la puissance heuristique de la théorie de l’évolution. C’est fort de ces développements modernes que cette thèse propose d’aborder la question de l’influence du risque de prédation sur la distribution, le choix d’habitat et la survie des proies en milieu naturel. Environnement Comportements Mécanismes Proximaux Sélection naturelle Fonction Ultime Figure 1.1 : Approche intégrative du comportement schématisée d’après Tinbergen (1963). Les flèches pleines indiquent l’influence d’une entité sur une autre à l’échelle de l’individu. La flèche en pointillés symbolise l’influence de la sélection naturelle (échelle populationnelle). 20 1.3. SURVIVRE DANS LE PAYSAGE DE LA PEUR La peur du prédateur semble habiter tous les organismes capables de percevoir un tel risque, même lorsque la prédation a disparu depuis plusieurs générations (Byers 1997). L’être humain n’échappe pas à cette règle (Diamond 1997), alors que les réels cas de prédation sur l’homme ne représentent plus qu’une infime cause de mortalité depuis des millénaires, à l’exception de certaines zones très conflictuelles (e.g. Gurung et al. 2008). Cette inertie traduit à quel point la prédation a pu, et peut encore, être structurante pour les individus. Les variations naturelles des traits, ou phénotypes, des proies (ex : métabolisme, coloration, taille, comportement…) peuvent avoir un impact fort sur le nombre de descendants laissés dans l’environnement par un individu, surtout s’ils impliquent une survie différentielle à la prédation. Puisque la mort d’un individu signe l’arrêt de sa contribution à la production des générations suivantes, la prédation peut constituer une véritable pression de sélection pour les proies, aboutissant à l’émergence de stratégies anti-prédatrices. 1.3.1. L’évitement comme premier rempart à la prédation Le comportement des proies peut constituer un facteur-clef leur permettant de réduire le risque de prédation (Caro 2005). Eviter, détecter, fuir, tromper, dissuader un prédateur en adoptant le comportement adéquat sont autant de moyens pour une proie d’éviter d’être consommée, et d’accroître ainsi sa fitness. 21 Les stratégies anti-prédatrices comportementales peuvent être classifiées en s’inspirant de la séquence comportementale de prédation qui permet à un prédateur de passer de la recherche à la consommation d’une proie : rencontre, détection, identification comme proie profitable, attaque et capture, consommation (Endler 1991, Caro 2005) . Ainsi, pour faire échouer la séquence de prédation, les proies peuvent adopter des stratégies d’évitement, de détection anticipée, de regroupement, de dissuasion et de défense et de duperie (Caro 2005). Notons que le tout premier moyen de réduire le risque de prédation est tout simplement d’éviter de rencontrer un prédateur. Si le risque présente des variations spatiales et temporelles, les changements d’habitat et/ou de distribution constituent donc le premier rempart des proies contre la mortalité induite par la prédation. L’influence du risque sur les distributions ou les choix d’habitat des proies et sur leur dynamique temporelle a ainsi fait l’objet de très nombreuses recherches depuis que les biologistes se sont attelés à comprendre les stratégies anti-prédatrices. 1.3.2. Le compromis énergie-risque : manger sans être mangé Le nombre de descendants laissés dans l’environnement par un individu n’est pas uniquement dépendant de sa survie mais dépend aussi du gain énergétique lui permettant de maintenir son métabolisme et d’assurer sa reproduction. En l’absence de risque, les comportements des animaux sont théoriquement supposés être déterminés par le gain énergétique qu’ils assurent (Optimal Foraging Theory, MacArthur and Pianka 1966). Le risque de prédation et la réponse comportementale des proies impliquent cependant de s’écarter de ces stratégies d’approvisionnement optimales. Les comportements antiprédateurs représentent donc un coût énergétique pour les proies, pas seulement parce qu’ils nécessitent de l’énergie pour être effectués, mais parce qu’ils empêchent l’approvisionnement optimal d’être réalisé (Brown 1988, Houston et al. 1993). Pour déterminer la stratégie optimale au sein de ce compromis énergie-risque, certains auteurs ont d’abord considéré que les individus minimisaient le rapport du risque sur le gain énergétique (Houston et al. 1993, Werner and Anholt 1993). Un des 22 principaux résultats théoriques est qu’il existe une stratégie optimale unique, souvent intermédiaire entre la stratégie de maximisation du gain et celle de minimisation du risque. Néanmoins, lorsque la densité à un effet sur la fitness (ex : compétition alimentaire), il peut alors exister un continuum de stratégies permettant d’obtenir une fitness maximale, ou tout du moins égale aux autres (ex : une forte exposition au risque peut être récompensée par un fort gain dû à une faible densité). Cette situation aboutit théoriquement à l’égalisation des fitness entre individu (isofitness, Fretwell and Lucas 1969), alors qu’ils adoptent des stratégies différentes. Ces deux approches du compromis énergie-risque ne sont évidemment pas exclusives, même pour une espèce donnée : si une stratégie unique est optimale à faible densité, il est possible qu’un panel de stratégies différentes devienne optimal lorsque la densité est forte (Morris 1989, 2003). 1.3.3. Stratégies de sélection d’habitat risque-dépendantes 1.3.3.a. Importance de l’échelle spatiale Le concept d’habitat répond à différentes définitions suivant le contexte théorique dans lequel il est employé (Martin 2009). L’habitat peut premièrement être défini comme les ressources et conditions d’une aire qui sont à l’origine de son occupation – incluant la survie et la reproduction – par un organisme (Hall et al. 1997). Cette définition s’inscrit quasiment dans une perspective évolutive puisqu’elle considère la persistance ou la performance des individus comme un critère de définition de l’habitat. Dans une optique plus descriptive, l’habitat peut être défini comme la partie de l’hyperespace, défini par les variables environnementales d’une aire donnée, occupée par un organisme (Whittaker et al. 1973). Nous nous référons dans la suite du texte à cette deuxième définition, probablement mieux adaptée au concept d’habitat sur de fines échelles spatiales (voir cidessous). Dans des conditions naturelles, la sélection d’habitat intervient en effet à différentes échelles spatiales (échelles de Johnson 1980) : au sein de l’aire de distribution 23 d’une espèce ou d’une population (niveau 1), un individu peut réaliser un choix d’habitat pour établir son domaine vital (niveau 2) puis, à plus fine échelle, sélectionner certains îlots d’habitat (l’anglicisme « patchs » sera ensuite utilisé) au sein de son domaine vital (niveau 3) et enfin sélectionner certains items alimentaires au sein des patchs (niveau 4). Rettie and Messier (2000) ont proposé la théorie selon laquelle la hiérarchie spatiale et temporelle de la sélection d’habitat reflétait la hiérarchie des facteurs limitant la fitness individuelle, les facteurs les plus limitants déterminant la sélection aux plus grandes échelles. Concrètement, des individus peu limités par la ressource mais fortement limités par la prédation devraient privilégier l’évitement du risque au détriment du gain énergétique lors de l’établissement de leurs domaines vitaux (niveau 2). A plus fine échelle, ils devraient ensuite sélectionner les patchs les plus riches en ressource au sein de leurs domaines vitaux (niveau 3). Ainsi, suivant les traits d’histoire de vie des espèces (ex : longévité, nombre d’occasions de reproduction), les caractéristiques des individus (ex : sexe, âge), leur état interne (ex : réserves énergétiques) et l’environnement (ex : richesse, niveau du risque), les individus devraient choisir leur habitat à large échelle d’après le premier facteur limitant leur fitness (énergie ou risque), puis d’après le facteur suivant à l’échelle inférieure (Martinez et al. 2003, Anderson et al. 2005). Néanmoins certains auteurs observent que le compromis entre l’énergie et le risque est résolu au sein d’une même échelle, celle du domaine vital (niveau 3, Johnson et al. 2002, Dussault et al. 2005). Hebblewhite and Merrill (2007) suggèrent que la migration serait un moyen de résoudre le compromis énergie-risque entre les échelles alors que les individus nonmigrants devraient résoudre le compromis au sein d’une même échelle. 1.3.3.b. Echelle populationnelle : Distribution Libre Idéale Au sein des populations, une forte densité peut avoir un effet positif sur la fitness (ex: Allee et al. 1949) mais un effet négatif est théoriquement attendu à terme du fait de la capacité d’accueil limitée des habitats (ex: Fowler 1987, Stewart et al. 2005, Bonenfant et al. 2009). Ces effets, lorsqu’ils sont conséquents, peuvent aboutir à l’apparition de stratégies de sélection d’habitat densité-dépendantes (Haugen et al. 2006, McLoughlin et 24 al. 2006, Morris 2006), s’équilibrant autour de la Distribution Libre Idéale (Ideal Free Distribution, IFD, Fretwell and Lucas 1969). Puisque la prédation réduit la fitness individuelle, le risque devrait provoquer une réduction de la densité des individus à l’équilibre dans l’habitat où il existe, jusqu'à ce que la fitness des individus soit équivalente à celle des individus dans les habitats non risqués (Moody et al. 1996, Morris 2003). Dans cette situation d’équilibre théorique, la mortalité subie par les individus dans l’habitat risqué est compensée par la faible densité locale aboutissant à une fitness identique à celle des individus de l’habitat non risqué (Abrahams and Dill 1989, Abramsky et al. 1997, Grand and Dill 1997). Notons que l’effet du risque sur la distribution peut dépendre du degré de la compétition intraspécifique puisque, si celle-ci devient négligeable et que la mortalité n’est pas densité dépendante, la meilleure stratégie peut être simplement d’utiliser l’habitat le moins risqué quelle que soit la densité (Morris 1989, 2003). La densité peut néanmoins avoir un effet direct sur le risque de prédation, qui peut s’avérer positif en termes de gain de fitness (ex : dilution du risque). Dans ce cas, l’IFD prédit une tendance à l’agrégation des individus dans certains habitats en réponse à un haut niveau de risque (Grand and Dill 1999a, b). Ainsi, dans le cas où les individus se distribuent de façon à atteindre l’isofitness (prédiction générale de l’IFD), il est nécessaire de bien déterminer tous les effets de densité-dépendance intervenant sur la fitness des individus ainsi que leurs importances respectives afin de pouvoir prédire la distribution des proies. 1.3.3.c.Echelle individuelle Giving Up Density Les ressources utilisées par les animaux ont souvent une distribution morcelée, constituant des patchs riches parsemés au milieu d’une matrice pauvre. Lorsqu’un individu se nourrit, il épuise petit à petit le patch jusqu'à un certain seuil provoquant son départ du patch pour un autre. Le Théorème de la Valeur Marginale (Charnov 1976) permet de prédire le temps de résidence dans un patch de nourriture (Giving Up Time) ou 25 la densité de nourriture restante, à partir de laquelle l’individu a intérêt à changer de patch (Giving Up Density, GUD). Brown (1988) a intégré dans cette approche le risque de prédation et le gain de fitness apporté par une autre activité que le nourrissage. Il a déterminé que les individus devraient laisser plus de nourriture dans le patch au moment de leur départ (GUD élevée) lorsque le gain instantané y est faible, lorsque le coût métabolique de son utilisation est élevé, lorsqu’il est risqué en terme de prédation et lorsque les activités alternatives sont importantes pour la fitness (ex : territorialité, vigilance, reproduction, soins aux jeunes). Beaucoup d’auteurs ont utilisé cette méthode pour étudier l’influence du risque sur l’utilisation des patchs, avec comme résultat général une utilisation moindre des patchs risqués (Brown 1988, Kotler et al. 1991, Brown 1999, Hochman and Kotler 2007, van der Merwe and Brown 2008). Mais l’intérêt de cette théorie n’est pas véritablement de démontrer que les individus évitent la prédation, mais plutôt de fournir un moyen de quantifier le niveau de la « peur » dans une optique d’écologie évolutive (il s’agit d’un coût énergétique), et de comprendre son effet sur le comportement des individus, sur leur utilisation de l’espace et sur le fonctionnement des écosystèmes via les réseaux trophiques (Brown et al. 1999, Ripple and Beschta 2004). Utilisation temporaire d’un refuge Les refuges sont des espaces où le risque de prédation est nul (notion absolue) ou, tout du moins, significativement moindre qu’ailleurs (notion relative). Ces espaces peuvent être naturels (ex : fissures de roches, zones escarpées, végétation dense), être partiellement modifiés par les individus (ex : terrier) ou encore résulter de l’utilisation de l’espace par les prédateurs. Lorsque les refuges ne permettent pas d’acquérir les ressources nécessaires à la survie et la reproduction, les individus ont intérêt à choisir la meilleure stratégie d’allocation du temps à passer dans l’habitat refuge, ou le meilleur temps de réémergence après une attaque (Martin and Lopez 1999a, b, Martin et al. 2003, Cooper 2009). 26 Lima et Benekoff (1999) ont proposé un modèle permettant de prédire l’influence des variations temporelles du risque de prédation sur l’intensité des comportements antiprédateurs (l’utilisation d’un refuge dans notre cas): l’hypothèse de l’allocation au risque de prédation (Predation Risk Hypothesis). Ils démontrent que la stratégie optimale consiste d’abord à allouer le plus de temps dans le refuge lors des phases risquées (lorsque son utilisation est la plus bénéfique), et d’allouer le plus temps à l’extérieur lors des phases de faible risque. Plus le contraste entre phases risquées et non risquées est fort, plus le contraste en termes d’utilisation du refuge devrait l’être également. Leur seconde prédiction, moins intuitive, annonce que plus la phase de risque est fréquente ou longue, moins les animaux devraient utiliser le refuge, surtout lors des phases de faible risque. Ceci provient du fait que les animaux n’ont d’autre choix que de satisfaire leurs besoins énergétiques minimums, et ne peuvent rester indéfiniment dans le refuge. Ainsi, les prédictions sur l’allocation de temps optimale dans le refuge dépendent de l’état interne des individus et de leur métabolisme en interaction avec l’environnement (Cooper 1998 , Martin and Lopez 1999b, Cooper 2009), ainsi que de l’intensité et la fréquence des variations temporelles du risque de prédation (Lima and Bednekoff 1999, Sih et al. 2000, Cooper 2009, Martin et al. 2009a). 1.3.4. Paysage de la peur vs. paysage du risque 1.3.4.a. Le paysage de la peur La théorie de l’approvisionnement optimal (Optimal Foraging Theory, OFT, MacArthur and Pianka 1966, Charnov 1976, Stephen and Krebs 1986) peut constituer une référence théorique permettant d’estimer le coût du risque. Si aucun autre facteur que le risque ne vient s’ajouter au système, alors l’écart entre le comportement observé et le comportement prédit d’après l’OFT correspondra théoriquement à l’effet du risque, que l’on pourra traduire en coût énergétique (Brown 1988). Certains auteurs ont par exemple introduit expérimentalement un facteur de risque dans une population distribuée de façon 27 libre et idéale et ont ensuite estimé l’enrichissement nécessaire pour rétablir la distribution observée sans facteur de risque (Abrahams and Dill 1989, Grand and Dill 1997). Rapportée à un individu, cette valeur correspond à l’équivalence énergétique du risque de prédation. Plus récemment, d’autres auteurs ont quantifié dans l’espace le coût du risque sur le gain énergétique à partir de patchs de nourriture répartis dans l’espace (Shrader et al. 2008, van der Merwe and Brown 2008). Pour cela ils ont cartographié l’écart entre les GUDs observées et théoriques dans un système où le risque était le principal facteur capable de faire dévier les individus du comportement d’approvisionnement optimal. Ils ont ainsi pu décrire le « Paysage de la Peur » (« Landscape of Fear », Laundré et al. 2001) qui représente la variation spatiale du coût énergétique du risque dû à l’allocation au comportement anti-prédateur (Fig. 1.2). Figure 1.2 : Paysage de la peur de chèvres domestiques (Capra hircus) dans un milieu ouvert, en réponse à des stimuli odorants de lynx du désert (Cararcal cararcal). Sont indiqués par des points noirs les emplacements des bacs de nourrissage artificiels utilisés pour déterminer l’effort de nourrissage (Giving Up Densities, GUDs). Les lignes associées aux chiffres représentent les contours d’efforts de nourrissage identiques après lissage spatial. Les plus fortes GUDs (lignes les plus claires) indiquent une faible utilisation d’une aire donc une forte « peur ». D’après Shrader et al. (2008). 28 Certains auteurs proposent également de mesurer le paysage de la peur directement à partir des comportements anti-prédateurs, comme la vigilance ou les signaux d’alarmes. De cette manière il est parfois possible de quantifier des paysages de la peur spécifiques à chacun des prédateurs lorsque ceux-ci sont multiples (Willems and Hill 2009). Cette méthode ne produit pas directement une image du coût énergétique de la peur mais plutôt une cartographie de l’investissement dans le comportement antiprédateur (Fig. 1.3). Ces deux composantes sont néanmoins positivement corrélées en situation de compromis énergie-risque, ce qui permet alors de retomber sur la définition classique du paysage de la peur. Figure 1.3 : Paysages de la peur spécifiques aux prédateurs d’un groupe de singe vert (Cercopithecus aethiops) au sein de leur aire de distribution. Les prédateurs sont des aigles (Eagle), des babouins (Baboon), des léopards (Leopard) ou des serpents (Snake). Le risque perçu est mesuré à partir d’estimations de densité locale de comportement d’alarme et de l’identification du prédateur impliqué. D’après Willems and Hill (2009). 29 1.3.4.b. Le paysage du risque Le paysage du risque n’a pas été aussi clairement défini que le paysage de la peur (ex: Valeix et al. 2009a, Valeix et al. 2009b). Il pourrait correspondre au niveau du risque de prédation réel existant pour une proie en chaque point de l’espace (Fig. 1.4). Figure 1.4 : Risque de prédation des wapitis (Cervus elaphus) par les loups (Canis lupus) pendant l’été dans le Parc National de Banff (Alberta). Ce paysage a été dessiné à partir d’une combinaison de la densité spatiale, des tailles de meute et de la sélection d’habitat de loups (niveaux de gris). Les croix blanches localisent les wapitis tués par les loups. D’après Hebblewhite & Merrill (2007). 30 Il devrait théoriquement se mesurer à partir de la mortalité réelle des proies (McLoughlin et al. 2005). Néanmoins, un taux de prédation nul observé dans un habitat fréquenté par les proies peut provenir du fait que le risque est effectivement nul, mais peut aussi s’expliquer par un évitement très efficace de ce risque par les proies dans cet habitat. Il est donc nécessaire de soustraire l’effet des comportements anti-prédateurs pour mesurer le paysage du risque réel. Comme il quasiment impossible d’observer des proies sans aucun comportement anti-prédateur, il est finalement plus pratique de l’estimer à partir de la répartition spatiale et temporelle des prédateurs (Fig. 1.4, ex: Fortin et al. 2005, Hebblewhite and Merrill 2007, Kittle et al. 2008). La différence entre paysage de la peur et paysage du risque provient du fait que le premier est la version du second perçue par la proie. Le paysage de la peur est donc pleinement dépendant de la qualité de la perception par les proies du risque auquel elles sont soumises. L’optimalité prédit une parfaite concordance entre ces deux variables (hypothèse d’omniscience). Le paysage du risque et le paysage de la peur peuvent néanmoins se dissocier (Kristan 2000) du fait des erreurs de perception, ce qui peut avoir d’importantes conséquences sur le comportement et la fitness des individus ainsi que sur la dynamique spatiale de la population (Remes 2000, Gilroy and Sutherland 2007). Le rôle de l’information, de par sa valeur adaptative et les contraintes qu’elle induit, sera donc primordial pour comprendre l’effet du paysage du risque sur l’utilisation des habitats. 31 1.4. IMPORTANCE DE LA PERCEPTION DU RISQUE 1.4.1. Rôle de l’information dans l’évitement du risque de prédation 1.4.1.a. Valeur adaptative de l’information Considérer qu’un animal est capable de réaliser un choix optimal d’habitat en réponse à un risque de prédation, c’est supposer préalablement qu’il dispose de suffisamment d’informations sur le paysage du risque. Ainsi, parce qu’elle permet d’entrevoir les choix possibles et leurs conséquences en termes de fitness individuelle, l’information présente une valeur adaptative importante (Stephens 1989 , Koops 2004, Dall et al. 2005). Bon nombre d’espèces expriment leurs comportements anti-prédateurs simplement en présence de certains stimuli, sans qu’aucun risque de prédation n’existe réellement (stimuli acoustiques, Baxter et al. 2006, visuels, Cooper 2008, tactiles, Roberts and Liebgold 2008, chimiques, Fraker 2009). Le rôle de la perception du risque est donc un concept fondamental pour comprendre les réponses spatiales des proies dans le paysage du risque. 1.4.1.b. Types d’information Les individus peuvent premièrement estimer les niveaux de risque à partir de stimuli directement émis par les prédateurs, donc indiquant leur présence immédiate (e.g. Baxter et al. 2006). Il semble néanmoins plus intéressant pour les proies d’utiliser des 32 stimuli indirects permettant d’estimer rapidement, et sans que cela soit trop coûteux, le risque de prédation. Beaucoup de proies sont par exemple sensibles aux indices de présence des prédateurs (e.g. féces, Roth et al. 2008), à la structure de l’habitat (ex: ouverture de la végétation, Schmitz 2005, Whittingham et al. 2006), à la proximité de leurs refuges (Cowlishaw 1997) ou encore aux comportements de leur congénères (Wong et al. 2005, Webster and Laland 2008) et d’autres espèces soumises aux mêmes risques de prédation (Schmidt et al. 2008). Pour acquérir ces informations, plusieurs voies peuvent être utilisées (Danchin et al. 2004, Dall et al. 2005). Chaque individu peut acquérir de « l’information personnelle », issue de son interaction directe avec l’environnement. Mais lorsque les proies vivent en groupes, « l’information socialement acquise » peut constituer pour chaque individu un raccourci très intéressant pour obtenir les informations dont il a besoin. Les signaux d’alarme permettent par exemple une diffusion très efficace de l’information quant à la présence d’un prédateur (Sherman 1977). De plus, même en dehors de ces comportements sociaux les proies émettent des « informations publiques » (Danchin et al. 2004), notamment lorsqu’elle essaient d’échapper à la prédation (présence, fuite, vigilance, cris, mort…). Ces informations peuvent être, pour les autres individus, des raccourcis intéressants pour estimer le niveau du risque local.. 1.4.2. Les contraintes de l’information dans le paysage du risque 1.4.2.a. Incertitude ou fiabilité Chaque information comporte un certain degré d’incertitude permettant de l’associer de façon plus ou moins claire à la présence d’un prédateur (Bouskila and Blumstein 1992, Sih 1992, Koops 2004). Certains auteurs préfèrent parler de fiabilité pour décrire la probabilité qu’une information soit vraie (Koops 2004). L’ouverture du milieu peut par exemple expliquer une certaine part de la variation spatiale du risque mais ne fournira jamais une estimation parfaite de celui-ci. Les proies ont intérêt à utiliser des 33 informations fiables de façon à sélectionner l’habitat idéal. Néanmoins, lorsque les proies sont dans un refuge où leur sécurité est assurée, et font face à une incertitude sur les niveaux de risque seulement à l’extérieur de ce refuge, il semble qu’elles aient intérêt à surestimer le niveau du risque à l’extérieur du refuge (Bouskila and Blumstein 1992, Sih 1992, Martin et al. 2009a). Ceci proviendrait du sévère coût de la sous-estimation en termes de fitness (mort) par rapport au coût de la surestimation (réduction du gain énergétique). 1.4.2.b. Coût de l’information Pour obtenir une information, les individus doivent allouer du temps et de l’énergie (ex : comportement d’exploration) et, dans le cas de la perception de la prédation, peuvent être amenés à prendre des risques. Les individus sont supposés utiliser les sources d’information de façon optimale d’après leurs coûts et les bénéfices qu’ils en retirent (Dall et al. 2005). Si le coût d’accès à l’information en terme de fitness devient trop important, on peut s’attendre à ce que les individus ignorent la source d’information (Koops 2004, Dall et al. 2005) et utilisent une source moins coûteuse. Lorsqu’ il s’agit pour l’animal d’estimer les niveaux de risque, le coût de l’information peut rapidement devenir démesuré si les individus meurent pour avoir simplement expérimenté le risque dans une zone particulière. Les informations permettant une estimation indirecte du risque devraient donc être fortement privilégiées. 1.4.2.c.Acquisition de l’information Un individu dispose initialement d’une certaine information (« priors ») lui permettant de réaliser le choix d’habitat qu’il perçoit comme optimal. Suite à ce choix qui a lui-même été source d’information, l’individu peut corriger ses estimations (« posteriors ») de façon à choisir l’habitat lui assurant un meilleur succès dans cet environnement. Ce processus inspiré de la théorie de la décision Bayesienne (McNamara and Houston 1980, McNamara et al. 2006) peut correspondre à de l’apprentissage au 34 niveau individuel mais peut également résulter d’un mécanisme de sélection à l’échelle de la population, si les informations sont héritables (Dall et al. 2005). L’acquisition de l’information implique donc que les individus se confrontent d’abord au milieu avant d’apporter une réponse adaptée (Welton et al. 2003). La nécessité d’acquisition de l’information pourra aboutir à une dynamique temporelle des comportements qui devrait converger progressivement vers le comportement optimal. Dans le cas où une erreur de perception se maintient, le comportement des individus à terme devrait dévier du comportement idéal (Gilroy and Sutherland 2007). 1.4.2.d. Limitation La limitation de l’information acquise résulte surtout du fait que, même avec des capacités mnésiques infinies, aucun individu ne pourrait obtenir l’intégralité des informations existantes simplement parce qu’il n’aurait pas l’occasion d’y accéder. Les proies ne peuvent percevoir les variations spatiales du risque que sur une zone restreinte qu’elles ont eu l’occasion d’explorer, définissant ainsi leur champ (ou fenêtre) de perception (Lima and Zollner 1996). La limitation du champ de perception des individus restreint donc leur choix d’habitats potentiels. Même si des proies avaient une perception exacte, non coûteuse et instantanée du risque de prédation, certains individus ne pourraient pas rejoindre les habitats refuges, simplement parce qu’ils sortent de leur champ de perception. 1.4.2.e.Effet sur la performance Les individus ne peuvent réellement disposer d’une perception parfaite de leur environnement. Les conséquences d’un choix à cours terme sur la fitness à long terme sont difficilement perceptibles (manque d’information, ex: Gilroy and Sutherland 2007). De plus, la nouveauté de l’environnement peut être trop soudaine ou trop importante pour que les individus puissent modifier suffisamment leur « priors » et répondre de façon adaptée (manque de capacité d’apprentissage ou d’adaptation, Remes 2000). Une erreur 35 de perception sur le risque implique une sur- ou une sous-fréquentation des habitats par rapport à la situation optimale, et donc une réduction de fitness. Si tous les individus réalisent la même erreur et vivent dans des environnements comparables, alors leurs fitness devraient être équivalentes. Des erreurs de perception du risque peuvent néanmoins aboutir à une différence de fitness entre les individus si un risque nouveau n’apparaît par exemple qu’à certains endroits (Remes 2000, Gilroy and Sutherland 2007). Dans le cas d’une IFD, une erreur d’estimation du risque aboutira à une fréquentation trop importante de certaines zones par rapport aux prédictions obtenues sous l’hypothèse d’omniscience (Remes 2000, Shochat et al. 2002). Si la densité réduit la fitness, les individus de l’habitat sur-fréquenté devraient alors présenter une fitness moindre que celle des individus de l’habitat sous-fréquenté. La limitation du champ de perception peut également aboutir à une hétérogénéité de fitness entre les individus simplement parce la qualité du meilleur habitat local pourra varier suivant les individus. Ainsi, les contraintes inhérentes à l’acquisition de l’information peuvent aboutir à une réduction et une hétérogénéité des performances individuelles. La réduction de l’incertitude, l’utilisation d’informations peu coûteuses, la rapidité d’acquisition d’informations nouvelles ainsi que l’élargissement du champ de perception constituent donc très probablement des avantages adaptatifs importants lorsqu’il s’agit d’éviter le risque de prédation. Toutes ces caractéristiques devraient aboutir au fait que la structure du paysage de la peur d’un individu tend à converger vers la structure du paysage du risque. Néanmoins, la qualité de la perception du risque n’est pas la seule contrainte agissant sur la sélection d’habitat : même si les individus disposent d’une perception parfaite du risque, ils doivent encore avoir les capacités de se mouvoir dans l’environnement et d’atteindre les habitats optimaux. 36 1.5. LE MOUVEMENT DANS LE PAYSAGE DU RISQUE 1.5.1. Le mouvement comme processus de sélection d’habitat Les patrons de sélection d’habitat des animaux ont été étudiés dans leur environnement naturel depuis plusieurs décennies (Johnson 1980), et les liens existant in situ entre ces patrons et la fitness individuelle commencent maintenant à être mis en évidence (McLoughlin et al. 2006, McLoughlin et al. 2007). A l’intérieur des domaines vitaux (troisième échelle de Johnson 1980), les processus sous-jacents à la sélection d’habitat ne sont néanmoins que rarement mis en évidence, surtout lorsqu’il s’agit d’espèces ayant d’importantes capacités de déplacement (Nathan et al. 2008). A cette échelle, la sélection d’habitat résulte de la modification des mouvements des individus (Morales et al. 2004, Frair et al. 2005) aboutissant à une allocation différentielle de temps parmi les différents habitats disponibles. Ainsi, l’étude des processus de mouvement et de leurs interactions avec l’environnement devient essentielle pour comprendre l’émergence des patrons de sélection d’habitat à l’échelle individuelle. 1.5.2. L’écologie du mouvement L’écologie comportementale ne semble pas pouvoir fournir à elle seule un cadre théorique adapté à la compréhension des déplacements des individus, tant ceux-ci sont soumis à des mécanismes pouvant les faire dévier de l’optimalité. Si les modalités de 37 mouvement peuvent effectivement être soumises à une certaine pression de sélection les faisant converger vers une stratégie optimale (e.g. Charnov 1976), ils dépendent aussi fortement de contraintes et processus cognitifs, biomécaniques et aléatoires (Nathan et al. 2008). L’écologie du mouvement (Holden 2006, Holyoak et al. 2008, Nathan 2008, Nathan et al. 2008) propose d’unifier tous les champs théoriques préalablement existants autour des processus de déplacement, dans un cadre conceptuel unique permettant d’expliquer tous types de mouvement observés dans la nature. 1.5.2.a. Les composantes du mouvement Ce cadre conceptuel s’articule autour de trois composantes individuelles que sont l’état interne, les capacités de navigation et les capacités de déplacement, auxquelles est ajoutée une quatrième composante correspondant aux facteurs externes (Fig. 1.5, Nathan et al. 2008). - L’état interne correspond à l’état multidimensionnel de l’individu qui influence sa motivation et sa propension à se déplacer (ex: états physiologique, énergétique, neurologique). - Les capacités de navigation sont l’ensemble des traits qui permettent à l’individu d’orienter son mouvement dans l’espace et le temps (ex : capacités cognitives et sensorielles permettant de traiter l’information). - Les capacités de déplacement correspondent à un ensemble de traits qui permettent à l’individu d’exécuter ou de faciliter les mouvements (ex : états biomécanique et bioénergétique). - Les facteurs externes sont l’ensemble des facteurs environnementaux biotiques et abiotiques (ex : structure de la végétation, température) qui affectent le mouvement de l’individu. Les mouvements observés devraient donc résulter des propriétés intrinsèques de ces quatre composantes ainsi que de leurs interactions (Fig. 1.5). L’environnement influence directement toutes les composantes individuelles pendant que l’état interne de l’individu joue sur les capacités cognitives et motiles de l’individu. Les capacités de 38 déplacement représentent l’ultime facteur limitant le mouvement observé. Ce fonctionnement comporte également une boucle rétroactive interne dans la mesure où le mouvement réalisé influence l’état interne des individus (ex : il coûte de l’énergie), et une boucle rétroactive externe puisque le mouvement modifie aussi les facteurs environnementaux agissant sur l’individu (ex : il permet d’éviter la prédation). Figure 1.5: Cadre conceptuel général de l’écologie du mouvement, composé de trois composantes reliées à l’individu (fond jaune ; internal state : état interne, motion capacity : capacités de déplacement, navigation capacity : capacités de navigation) et une quatrième composante représentant les facteurs externes (fond turquoise) affectant son mouvement. Les relations entre les composantes sont représentées par des flèches dont la direction représente le sens de l’impact. Le mouvement résultant (movement path) rétroagit sur l’état interne et sur les facteurs externes. D’après Nathan et al. (2008). 39 1.5.2.b. Une approche intégrative Le cadre conceptuel proposé implique une approche intégrative du mouvement puisque différents champs théoriques sont nécessaires pour comprendre le fonctionnement de chaque composante du mouvement ainsi que leurs interactions (Nathan et al. 2008). Le concept d’optimalité permet de comprendre les relations existant entre l’environnement et l’état interne des individus, en prédisant un mouvement qui permet de maximiser la fitness individuelle dans l’environnement donné. L’étude des processus cognitifs permet de comprendre comment les individus perçoivent et traitent les informations internes et externes, et par quels processus de navigation ils orientent leurs mouvements. Les approches biomécaniques et bioénergétiques permettent de comprendre en quoi les mouvements sont contraints par des facteurs intrinsèques aux individus (ex : type de locomotion, réserves énergétiques) et, de façon indirecte, par l’environnement si celui-ci conditionne les capacités des individus. Enfin, l’étude des processus aléatoires permet de comprendre la part résiduelle aléatoire du mouvement qui ne peut être expliquée par aucune des précédentes approches. 1.5.3. Le mouvement comme processus d’évitement du risque 1.5.3.a. Direction des mouvements et sélection d’habitat Deux types de mouvements peuvent être différentiés d’après les patrons, les processus sous-jacents et leur efficacité ultime en terme de maximisation de l’utilisation des habitats optimaux : les mouvements non-dirigés et les mouvements dirigés (Mueller and Fagan 2008). Les mouvements non-dirigés consistent en une modification des propriétés intrinsèques du mouvement (vitesse, angle de rotation, tortuosité…) d’après des conditions environnementales locales, sans que le mouvement résultant ne soit pour autant dirigé vers un habitat particulier (Benhamou and Bovet 1989). Un animal peut par exemple simplement réduire sa vitesse de déplacement lorsqu’il rencontre une zone de 40 faible risque (Lewis et al. 2001). Il devrait ainsi rester plus longtemps dans ces habitats, entraînant l’émergence d’un patron de sélection intra-domaine vital. Les mouvements dirigés impliquent un ajustement de la direction et la vélocité des mouvements en fonction d’habitats distants des individus. Les individus peuvent par exemple percevoir à distance des structures environnementales particulières, ou mémoriser leur environnement et naviguer jusqu’à l’habitat ciblé (cartes cognitive, O'Keefe and Nadel 1978, e.g. integration du trajet, Benhamou 1997, reconnaissance de route, Lipp et al. 2004). Dans ce cas, le processus de sélection provient du fait que les individus effectuent directement un choix informé en se déplaçant vers ou dans un habitat donné au sein d’un environnement perçu. Si ces deux types de mouvements peuvent faire émerger un patron de sélection d’habitat, les mouvements dirigés devraient néanmoins permettre une utilisation plus efficace des habitats optimaux que les mouvements non-dirigés, bien plus aléatoires (Armsworth and Roughgarden 2005, Brooks and Harris 2008). L’importance relative des mouvements dirigés vs. non-dirigés devrait dépendre des limitations de la perception individuelle, les individus adoptant des mouvements dirigés vers un habitat donné lorsqu’ils peuvent le percevoir à distance (Zollner and Lima 1999). 1.5.3.b. De la « fuite aveugle » à l’orientation vers les refuges De simples processus de mouvements non-orientés peuvent à eux seuls générer un évitement des zones de fort risque de prédation. En effet, un des comportements antiprédateurs les plus simples pour les proies est de se déplacer, à différentes échelles spatiales et temporelles, en réponse à la présence du prédateur (ex: fuite, Ydenberg and Dill 1986, abandon du patch, Brown 1999, migration, Hebblewhite and Merrill 2007, Stankowich 2008). De ce comportement devrait donc naturellement émerger un processus de sélection des refuges puisque les proies devraient rester plus longtemps dans les zones de moindre risque du fait de leur faible taux de rencontre avec les prédateurs (Lewis et al. 2001). 41 Les proies devraient néanmoins clairement bénéficier de mouvements dirigés vers les zones refuges qui réduiraient le temps d’exposition aux prédateurs par rapport à des mouvements non-dirigés (e.g. Cowlishaw 1997). Bon nombre de proies se comportent différemment suivant la proximité des refuges (Bonenfant and Kramer 1996). Ceci implique une perception à distance de ces espaces ou une mémorisation de leurs emplacements (Clarke et al. 1993). La perceptibilité des variables environnementales composant le paysage du risque devrait donc déterminer le type de mouvement exprimé par les proies. Lorsque des variables paysagères ou topographiques constituent une bonne estimation du risque pour les proies, il est très probable qu’une majorité d’animaux puissent exprimer des mouvements dirigés vers ces habitats. En revanche, lorsque les structures spatiales du risque ne proviennent que des variations de l’utilisation de l’espace par le prédateur, sans lien particulier avec des variables environnementales, les proies pourraient avoir plus de difficultés à se diriger directement vers les refuges et exprimeraient alors une réponse non-dirigée. 42 1.6. OBJECTIF DE LA THESE Afin d’étudier les stratégies de sélection d’habitat d’une proie dans le paysage du risque, cette thèse adoptera une démarche aussi intégrative que possible en abordant à la fois les aspects ultimes et proximaux de ce comportement ainsi que leurs interactions (Fig. 1.6). La question principale sera de comprendre pourquoi et comment les animaux sélectionnent leur habitat au sein de leurs domaines vitaux lorsqu’ils vivent dans un paysage du risque spatialement et temporellement structuré. Nous étudierons en quoi les processus et patrons de sélection d’habitat dans le paysage du risque constituent une réponse adaptative (l’ultime explique le proximal) et en quoi les contraintes de mouvement et de perception conditionnent la sélection d’habitat et la survie résultante (le proximal conditionne l’ultime). Cette approche est synthétisée par le schéma conceptuel présenté ci-dessous. 43 Paysage du risque Paysage de la peur Habitat Utilisation Barrières Mouvement Informations Accessibilité Individu Cognition Mobilité Fitness État interne Sélection naturelle Comportement de selection d’habitat Mécanisme proximal Echelle de temps courte Rétroaction proximale Consequence ultime Echelle de temps longue Rétroaction ultime Variations d’une composante Figure 1.6 : Schéma conceptuel intégrant les mécanismes proximaux (voie bleue) et les causes ultimes (voie rouge) du comportement de sélection d’habitat dans le paysage du risque. Les flèches indiquent le sens de l’impact d’une composante sur une autre. La sélection d’habitat résulte d’un mouvement (flèche verte) influencé par différentes composantes individuelles internes (cadres bleus) ainsi que par l’accessibilité de l’environnement et l’information disponible sur les habitats (flèches bleues pleines). Le choix d’habitat peut en lui même modifier la perception et l’accessibilité de l’environnement local (flèches bleues pointillées). Sur une échelle de temps plus long apparaissent les variations spatiales du niveau de la peur de l’individu, se traduisant par une utilisation différentielle des habitats et une certaines fitness individuelle en fonction des conditions rencontrées. La sélection naturelle exerce ensuite une pression de sélection, à l’échelle intergénérationnelle, sur le mécanisme individuel du mouvement permettant la sélection d’habitat. 44 Cette approche sera appliquée au cas d’une proie vivant dans un paysage du risque structuré et variable dans le temps. Après avoir présenté le modèle biologique utilisé dans ce travail, le site d’étude et les méthodes de suivis des individus, la thèse abordera différents aspects du cadre conceptuel présenté ci-dessus et sera structurée de la façon suivante : • Partie 1: Nous rappellerons d’abord certaines considérations méthodologiques relatives à l’utilisation d’outils télémétriques dans le cadre de l’étude de la sélection d’habitat. Cette partie prendra la forme d’un chapitre d’ouvrage scientifique rédigé en anglais (« On the use of telemetry in habitat selection studies ») et publié chez Nova Science Publishers (Martin et al. 2009b). • Partie 2: Nous mettrons ensuite en évidence le comportement de sélection d’habitat de la proie en étudiant l’impact des variations spatiales et temporelles du risque sur les patrons de distribution des individus (flèche verte, Fig. 1.6). Une attention particulière sera portée à l’effet de l’échelle de structuration spatiale du risque sousjacente aux domaines vitaux sur les types de réponse. Cette partie sera présentée sous la forme d’un article scientifique rédigé en anglais (« Responding to spatial and temporal variations in predation risk: space use of a game species in a changing landscape of fear ») et publié dans Canadian Journal of Zoology (Tolon et al. 2009). • Partie 3: Nous étudierons ensuite comment la présence d’un refuge influence les patrons de sélection d’habitat respectifs entre la proie et son prédateur et quelles sont ses conséquences sur la mortalité des proies (voie rouge, Fig. 1.6). Cette partie est destinée à montrer comment des contraintes proximales d’accessibilité des habitats influencent à terme la fitness des proies. Elle sera présentée sous la forme d’un article scientifique rédigé en anglais (« Impact of refuges on predator-prey spatial interactions and prey mortality: wildlife harvesting as a case study »), actuellement soumis dans Oikos. 45 • Partie 4: Nous étudierons ensuite comment les individus se déplacent dans le paysage du risque en fonction de la présence réelle ou non du prédateur dans l’environnement. Cela permettra de comprendre par quel mécanismes les individu perçoivent le risque de prédation dans l’espace et par quels types de mouvements ils y répondent (voie bleue, Fig. 1.6). Cette partie correspondra à un article scientifique rédigé en anglais (« Moving in a landscape of risk: the role of directed and non-directed habitat selection processes »), actuellement en préparation. • Partie 5: Nous proposerons finalement une exploration théorique de différents scenarii de dynamique de population, à la lumière de l’approche développée précédemment. Ce travail s’inscrira dans une optique appliquée et visera à montrer l’effet de la dissociation du paysage du risque et de la peur sur les dynamiques de populations. Cette partie sera rédigée en français. La thèse s’achèvera sur une discussion, rédigée en français, concernant les principaux résultats observés dans ce travail et leurs conséquences théoriques. Une attention particulière sera néanmoins portée aux enseignements et aux recommandations pour la gestion des espèces exploitées. Cette problématique sera abordée sur un cas d’étude assez particulier, puisque le prédateur en question sera l’homme, à travers ses activités de chasse. Nous étudierons la sélection d’habitat d’une espèce gibier, le sanglier Sus scrofa Linnaeus 1758, dans le cas d’une population chassée à proximité d’une zone de réserve. Les zones protégées jouent en effet fortement sur la structure spatiale du risque pour les espèces chassées (Fox and Madsen 1997, Madsen 1998), en créant de larges espaces où le risque est faible voire nul. Ainsi, le statut de protection s’ajoute aux autres variables environnementales et permet potentiellement de dissocier les habitats physiques (ex : forêt ou pente) du risque. De plus, la majorité des activités d’exploitation par l’homme comportent également une forte dynamique temporelle résultant des restrictions légales qui sont la plupart du temps en 46 vigueur autour de ces activités. A proximité des réserves les variations temporelles du risque s’associent aux variations spatiales, offrant ainsi un contexte semi-expérimental idéal pour étudier l’influence du risque sur les stratégies de sélection d’habitat ou de distribution des individus. 47 48 2. MODELE D’ETUDE : LE SANGLIER SUS SCROFA L. 49 2.1. UNE ESPECE PROBLEMATIQUE Le sanglier Sus scrofa L. constitue un gibier très prisé par les chasseurs de toute l’Europe. Depuis bientôt trois décennies, les populations de sangliers ne cessent pourtant d’augmenter en Europe (Saez-Royuela and Telleria 1986, Schley and Roper 2003), créant de nombreux problèmes de gestion aux conséquences économiques et écologiques parfois importantes. Les activités agricoles sont en effet particulièrement sensibles à la surabondance de sangliers puisque ceux-ci trouvent dans les cultures une importante source de nourriture (maïs, céréales, tubercules), causant au passage d’importants dégâts et de lourdes pertes financières pour les agriculteurs (Hone 1995, Geisser and Reyer 2004). Cette espèce possède en outre la capacité de fortement influencer l’écosystème dans lequel elle vit (Bratton 1975, Wright and Jones 2004, Heinken et al. 2006) et peut donc générer des troubles d’ordre écologique en cas de surabondance. Ainsi, lorsque des zones de réserve initialement destinées à la protection des habitats et de la biodiversité sont utilisées par les sangliers, les problématiques de gestion se multiplient en particulier lorsque ces zones protégées jouxtent des zones cultivées (Maillard 1998). Dans ce contexte, il paraît nécessaire de bien appréhender les mécanismes d’utilisation de l’espace par cette espèce, notamment lorsque celle-ci a la possibilité d’échapper à la pression de chasse en se réfugiant dans des zones à faible risque comme les réserves de chasse ou les réserves naturelles. Dans cette optique, l’étude de l’utilisation de l’espace par les individus et de la dynamique spatiale des populations de sangliers autour de zones de réserve constitue une clef pour établir des règles de gestion raisonnées et adaptées à cette espèce. 50 2.2. LE SANGLIER DANS L’ESPACE ET LE TEMPS 2.2.1. Utilisation de l’espace et habitat 2.2.1.a. Influence du système d’appariement Il est établi que les domaines vitaux des mâles sont de surface supérieure à ceux des femelles (Janeau and Spitz 1984, Saunders and Kay 1991, Boitani et al. 1994, Morini et al. 1995, Caley 1997, Baubet 1998) et apparaissent comme plus fragmentés (Boitani et al. 1994, Baubet 1998). Les domaines vitaux des groupes de femelles peuvent présenter, dans une même zone, un recouvrement quasi-total (Baubet 1998). A l'inverse, les domaines vitaux des mâles ne se recouvrent que faiblement et cette partie commune correspond généralement à l'emplacement des groupes de femelles (Boitani et al. 1994). Les distances journalières parcourues par les sangliers varient dans la littérature de 1 km à 16 km (Janeau and Spitz 1984, Janeau and Spitz 1990, Maillard 1996, Lemel et al. 2003). Elles apparaissent, comme pour les domaine vitaux, plus grandes chez les mâles que chez les femelles (Janeau and Spitz 1984, Janeau and Spitz 1990). La même observation peut être faite lorsque l'on s'intéresse à la distance inter-bauges (Baubet 1998). En milieu de garrigue méditerranéenne, Maillard (1996) observe néanmoins des distances journalières de déplacement quasi identiques entre les deux sexes. Janeau et al. (1995) ont montré que les femelles suitées présentent plus de trajets journaliers de type « stationnaire » et moins de « déplacements longs ». Lorsqu'elles ne sont pas suitées, la proportion de grands déplacements augmente. Les subadultes présentent une plus grande proportion de trajets du type « déplacements longs », et les mâles adultes présentent 51 quant à eux une forte proportion de « déplacements longs » mais surtout une absence totale de trajets de type « stationnaires ». Toutes ces observations vont dans le sens d’une influence du système d’appariement du sanglier (polygynie) sur la surface et la structuration des domaines vitaux ainsi que sur leurs déplacements : les mâles solitaires cherchent à se reproduire avec plusieurs femelles et se déplacent sur plusieurs groupes familiaux qu’ils défendent contres les autres mâles. Les femelles, plutôt dépendantes des ressources, concentrent leur utilisation de l’espace sur les zones les plus riches. Pour ces raisons, il sera nécessaire de contrôler l’effet éventuel du sexe sur les stratégies d’occupation de l’espace observées. 2.2.1.b. Habitat La sélection de l'habitat est étroitement liée à l'activité des sangliers, qui s’avère nocturne dans la majorité des cas (voir plus loin). En journée, la sélection du site de bauge est caractérisée par l'évitement du dérangement, qui se concrétise par la sélection des sites à fort couvert végétal (Saunders and Kay 1991, Boitani et al. 1994, Meriggi and Sacchi 2001 , Brandt et al. 2005), et par une nécessité de thermorégulation, qui aboutit à la sélection de zones fraîches, surtout lors des fortes chaleurs (Saunders and Kay 1991). Les animaux peuvent également montrer une occupation de l’espace indiquant une recherche de proximité des zones riches en nourriture, conduisant parfois à la sélection des zones cultivées au printemps pour l’établissement de leur site de bauge (Santos et al. 2004, Keuling et al. 2008a). L'habitat sélectionné pour la phase nocturne de nourrissage dépend des ressources que celui-ci fournit suivant les saisons, ainsi que des risques pour la survie qui lui sont associés (Janeau and Spitz 1984, Boitani et al. 1994, Spitz and Janeau 1995, Meriggi and Sacchi 2001 , Calenge et al. 2004, Wilson 2004). Ceci implique que le sanglier fréquente différents types d'habitat suivant la saison pour la prise alimentaire. Ainsi, de nombreux auteurs observent une fréquentation préférentielle des forêts en automne et hiver, lorsque la production en fruits forestiers (glands, faînes, châtaignes…) est la plus forte, puis un 52 transfert vers des espaces plus ouverts, notamment des zones cultivées, au printemps et en été (Janeau and Spitz 1984, Keuling et al. 2008a). Il sera donc nécessaire de considérer la temporalité de l’activité du sanglier en lien avec les variations spatio-temporelles de l’environnement, de façon à bien comprendre l’influence du risque sur les stratégies de sélection d’habitat. Les activités de chasse étant généralement diurnes et se produisant sur une saison particulière, il est notamment attendu que l’évitement du risque soit le principal facteur explicatif de la sélection du site de bauge lors de cette saison particulière (voir plus loin). 2.2.2. Revue des paramètres d’utilisation de l’espace Nous reportons ici des valeurs de tailles de domaines vitaux, de distances de retour de marque et de déplacements journaliers observés chez le sanglier en Europe (Tableau 1.1). De façon à pouvoir replacer la population étudiée parmi celles déjà référencées, nous fournissons également, de façon anticipée, ces valeurs calculées dans cette thèse sur les sangliers étudiés (voir matériel et méthode pour la présentation des données). La population de sanglier que nous allons étudier étudiée dans ce travail présente des tailles de domaines vitaux, des distances de retour de marque et des trajets journaliers plutôt inférieurs aux valeurs reportées initialement dans la littérature scientifique, mais restant néanmoins dans les ordres de grandeur habituel du sanglier (ce qui n’est pas le cas du sanglier en milieu de garrigue méditerranéenne qui présente une organisation spatial à plus grande échelle, Maillard 1996). 53 Etudes Brandt et al. (1998) Sodeikat & Pohlmeyer (2003) ■ Massei et al. (1997) ■ Boitani et al. (1994) ▲ Baubet (1998) Maillard (1996) Ce travail (moy+/-ecart.type) Dardaillion & Beugnon (1987) Truvé & Lemel (2003) Spitz (1989) ● Boisaubert & Klein (1984) ● Baubet (1998) Ce travail (moyenne) ♀ ♂ Domaine vital ou de bauge (km²) 11.2 19.9 12.1 5.2 11.7 11.1 13.4 8.3 26.3 31.5 54.9 10.1 +/-2.5 7.6 +/-2.9 Distance de retour de marque (km) 19.6 20.6 5.0 7.0 4.5 16.6 6.5 8.7 3.2 5.0 3.3 4.1 3.3 4.6 10.8 12.6 3.5 1.6 11.2 Médiane Quantile 95% ▲ Singer et al.(1981) Lemel et al. (2003) Janeau & Spitz (1990) Maillard (1996) Ce travail (moy +/- ec. type) 6.9 2.9 19.5 Déplacement journalier (km) 5.8 7.2 5.0 7.2 5.8 5.5 4.9 +/-0.3 6.0 +/0.5 Tableau 1.1: Paramètres moyens de l’occupation de l’espace du sanglier issus de la littérature scientifique. Nous avons conservé uniquement les valeurs de domaines vitaux représentant plus de 5 mois de suivi et estimées par un Polygone Convexe Minimum à 95 ou 100 %. Les études marquées par un (■) ne fournissent que les valeurs maximale et minimale et nous reportons simplement la valeur intermédiaire. Les études marquées par un (●) fournissent les données de dispersion sous forme de distribution d’individus dans différentes classes de distance. Nous calculons alors la moyenne avec la médiane de chaque classe pondérée par l’effectif associé. Les études marquées par un (▲) donnent seulement la vitesse horaire moyenne et nous effectuons la conversion en kilomètres par jour. 54 2.3. LE SANGLIER DANS LE PAYSAGE DU RISQUE 2.3.1. Variations spatiales et temporelles du risque 2.3.1.a. Temporalité : le sanglier: un nocturne récent? Chez cette espèce, la phase d'activité la plus intense, qui correspond principalement à celle de l'alimentation, est nocturne (Saunders and Kay 1991, Boitani et al. 1994, Caley 1997, Russo et al. 1997, Cahill et al. 2003, Lemel et al. 2003) excepté dans certaines zones où les activités diurnes peuvent prendre une place assez conséquente (Singer et al. 1981), voire la place principale (Kruz and Marchinton 1972). Les patrons d’activité les plus diurnes semblent apparaître dans des zones peu dérangées comme des parcs naturels ou des zones non chassées. Certains auteurs détectent un patron d'activité bimodal (Caley 1997) avec des pics d'activité en début et fin de nuit. Les perturbations anthropiques, notamment la forte pression de chasse exercée sur le sanglier (Toïgo et al. 2008), sont probablement une des causes de ce basculement nocturne (Saunders and Kay 1991, Keuling et al. 2008b). 2.3.1.b. Effet de la chasse sur le domaine vital Le dérangement généré par la chasse provoque généralement une augmentation de la surface du domaine vital des sangliers, encore plus forte s’il s’agit de mâles subadultes (Maillard and Fournier 1995, Calenge et al. 2002, Sodeikat and Pohlmeyer 2003, Keuling et al. 2008b) Les auteurs estiment que cet effet est le résultat d’une augmentation des déplacements inter-bauges ou de déplacement des centres d’activité. Néanmoins, des résultats contrastés peuvent apparaître dans une même étude, avec notamment des réductions de domaines vitaux (McIlroy and Saillard 1989, Baubet 1998). Il est probable que la structuration spatiale des zones refuges puisse expliquer cette diversité. En effet, des individus proches d’une zone refuge plus petite que leurs domaines vitaux et stable 55 dans le temps (ex : réserve) pourraient restreindre leurs domaines saisonniers au sein de celle-là pendant la chasse. A l’inverse, les individus ne possédant pas de refuges stables n’auraient d’autre choix que de fuir le dérangement quand il se produit, augmentant ainsi leurs déplacements et la taille de leurs domaines vitaux. 2.3.1.c.Paysage du risque : l'effet réserve L'absence des activités de chasse dans certaines zones, qu'elle soit due à la présence de réserves naturelles ou non, crée d’importantes zones refuges pour les sangliers. Il est fréquent d'observer les animaux rejoindre une réserve au moment de la chasse (Baubet 1998, Maillard 1998). Il est très probable que ces zones permettent aux individus d’accroître leur survie, mais leur utilisation peut également dépendre de la disponibilité alimentaire. Ainsi, Rosell et al. (2004) montrent que l’utilisation d’une réserve par le sanglier est maximale en automne et en début d’hiver, lorsque la pression de chasse est maximale à l'extérieur de la réserve et que les ressources sont abondantes à l’intérieur de celle-ci (fruits forestiers). Au printemps et en été, les animaux se déplacent quotidiennement entre l’intérieur et l’extérieur de la réserve de façon à profiter de sa quiétude pour établir leur site de bauge le jour, et à se nourrir à l’extérieur, notamment dans les cultures, la nuit. 2.3.2. Sélection d’habitat et distribution risque-dépendante 2.3.2.a. Evidence du compromis énergie-risque En catégorisant l'habitat de sangliers suivant la richesse et le degré de fermeture, Spitz & Janeau (1995) ont observé que les deux sexes sélectionnent les zones les plus riches mais que les femelles apparaissent comme beaucoup plus sensibles au degré de fermeture et évitent les milieux ouverts même s'ils s'avèrent riches. Les auteurs expliquent ces patrons différents par le caractère dimorphique et polygyne du sanglier. 56 Les mâles, devant croître rapidement et le plus possible pour être compétitifs vis-à-vis des autres mâles, sélectionneraient préférentiellement des zones offrant d'abondantes ressources. Les femelles, qui doivent en revanche maximiser le nombre et la survie de leurs jeunes, sélectionneraient plutôt les habitats sécurisés. L’analyse de la structuration spatiale des dégâts montre que les zones les plus proches des forêts sont plus sévèrement endommagées (Calenge et al. 2004, Wilson 2004). La proximité du couvert végétal semble donc être préférentiellement sélectionnée par les sangliers lorsqu'ils se nourrissent en milieux ouverts. Ces résultats laissent présager l'importance du compromis énergie-risque lors de la prise alimentaire chez le sanglier. 2.3.2.b. Importance de l’échelle spatiale et temporelle Un gradient d’échelle de réponse Suivant l'habitat disponible et la méthode de chasse, les zones potentiellement refuges pour le sanglier peuvent être simplement des zones de fourrés distribuées finement au sein d'un domaine vital (Saïd et al. in prep, voir annexe 1), ou à l'inverse de larges zones refuges de type réserves de chasse ou réserves naturelles (Baubet 1998, Maillard 1998). Suivant l’environnement, le niveau de dérangement et les méthodes de chasse, les animaux peuvent donc éviter le dérangement à différentes échelles spatiales et temporelles. Maillard & Fournier (1995) mettent en évidence que les animaux qui subissent des dérangements réguliers effectuent de longs déplacements de leurs centres d'activités vers des sites plus sécurisés où ils tendent à rester pendant la saison de chasse. Ils reviennent néanmoins hors saison de chasse sur les secteurs qu’ils avaient délaissés. Ces déplacements saisonniers peuvent notamment être observés à proximité des zones protégées (Baubet 1998, Maillard 1998). La réponse des sangliers face à la chasse peut néanmoins se produire à une échelle intermédiaire (au sein des domaines vitaux saisonniers) et pour de plus courtes périodes (durant de quelques jours à quelques 57 semaines). Les animaux peuvent alors exprimer des déplacements temporaires suite au dérangement et revenir ensuite sur leur lieu d’origine, surtout si le dérangement n’est pas très fréquent (Maillard and Fournier 1995, Sodeikat and Pohlmeyer 2003). Il arrive aussi que la réponse ne soit pas visible à l’échelle de l’occupation de l’espace, mais plutôt au niveau de l’utilisation des habitats, i.e. à plus fine échelle. Dexter (1996) observe par exemple que la taille et la position du domaine vital de sangliers ne sont pas influencées par des sessions de tirs par hélicoptère. L'auteur précise néanmoins que lors du passage de l'hélicoptère, les sangliers trouvent rapidement refuge dans la végétation dense abondamment répartie sur la zone d'étude. Ceci illustre bien que les animaux peuvent encore, à très fine échelle, modifier leur sélection d’habitat lorsque le risque est finement structuré dans l’espace, sans qu’une réponse apparaisse à large échelle. Influence sur la sélection d’habitat La végétation dense est souvent considérée comme un habitat potentiellement refuge et les zones très ouvertes sont généralement évitées pendant toutes les saisons (Boitani et al. 1994, Caley 1997, Santos et al. 2004). Dans un milieu forestier, Saïd et al. (in prep, Annexe 1) observent effectivement avant l'ouverture de la chasse une sélection majoritaire par les sangliers des parcelles de fourrés les plus denses. Lorsque la chasse intervient, les auteurs observent, contre toute attente, une réduction partielle de cette sélection au profit des parcelles de taillis, qui offrent pourtant moins de couvert végétal au niveau du sol. Après la chasse, la sélection pour la végétation dense se rétablit. Ainsi, contrairement aux attentes, les sangliers n’intensifient pas leur sélection pour la végétation dense mais tendent à la réduire en réponse au dérangement créé par la chasse. Cela pourrait traduire une réponse adaptative des animaux si la pression de chasse était effectivement plus importante dans les zones de fourrés denses. Néanmoins, cette réponse pourrait simplement provenir du fait que la chasse intervient comme un élément perturbateur d’un système à l’équilibre. En effet, le dérangement tend à augmenter le déplacement des sangliers (Maillard and Fournier 1995, Calenge et al. 2002, Sodeikat 58 and Pohlmeyer 2003) et pourrait les pousser à quitter les zones de fourrés plus fréquemment pendant la chasse. 59 60 3. MATERIEL ET METHODES 61 3.1. AIRE D’ETUDE : LE BASSIN GENEVOIS 3.1.1. Une mosaïque de méthodes de gestion 3.1.1.a. Présentation Le Bassin Genevois est un milieu semi urbanisé à tendance agricole contenant néanmoins encore 30 % de forêts fragmentées. Il est délimité par une ceinture montagneuse plus boisée sur trois de ses cotés, et parcouru dans son axe Nord-Est SudOuest par l'extrémité du Lac Léman et le Rhône (Fig. 3.1). Une réserve naturelle (réserve de l'Etournel) de 318 ha, est présente sur la limite interdépartementale Ain-Haute Savoie (Fig. 5.1, voir plus loin). Elle est constituée pour l'essentiel de terrains marécageux et comprend de petites îles dans le Rhône. Dans la partie sud, d’importants problèmes de gestion se concentrent autour de la réserve naturelle de l’Etournel (318 ha). Il s’agit d’un marais constitué de roselières et de forêts hygrophiles, sans grand intérêt alimentaire pour le sanglier mais apparemment très favorable pour l’établissement de sa bauge. La périphérie de la réserve est constituée de cultures, de prairies et de forêts de feuillus contenant notamment du chêne et du hêtre. La réserve fut créée en 1962 initialement dans un but de conservation de l’avifaune locale et de la zone marécageuse. Sa composition correspond à de la forêt hygrophile (Alnus spp, Salix spp; 44%), des roselières (Phragmites spp; 30%), des surfaces aquatiques (16%), des hêtraies ou chênaies (Quercus spp, Fagus spp; 6%) et à quelques cultures à proximité de la bordure (maïs, céréales; 4%). Le reste de la partie française du bassin genevois est constitué de prairies (29%), de hêtraies ou de chênaies (27%), de forêts de feuillus indifférenciés (25%) et de cultures (12%; maïs, céréales). Entre mars et août, de l’agrainage dissuasif est pratiqué (avec du maïs) dans les massif forestier à l’extérieur de la réserve entre mars et août (environ 200-300 kg/an pour un km² de foret). De fortes productions de fruits forestiers (glands, faines) ont été reportés non loin de la zone d’étude pour les années 2004 et 2006 62 (respectivement 836 et 386 kg.ha-1 de forêt, contre 2 à 64 kg.ha-1 de forêt pour les autres années, National Forestry Office 2008). Réserve de l’Etournel Frontière franco-suisse N Zone d’étude France Suisse Genève 10 km Figure 3.1 : Localisation (à gauche) et cartographie générale (à droite) du Bassin genevois. L’essentiel des animaux étudiés dans ce travail ont été localisés dans le polygone pointillé rouge. Le fond de carte à droite représente le relief ombré issu d’un modèle numérique de terrain (source © 2004. CGIAR - Consortium for Spatial Information). 3.1.1.b. Modes de gestion De part et d'autre des frontières, les modalités de gestion de l’espèce sanglier varient fortement. Le canton de Vaud pratique la chasse en battue de début septembre à fin janvier. Le Canton de Genève propose une gestion entièrement assurée par l'état via le Domaine Nature et Paysage. Les gardes faunes sont chargés de réguler les effectifs, uniquement en milieux ouverts, par des tirs nocturnes de mi-juillet à fin février. L'optique de cette gestion est de disposer de moyens de régulation sans avoir à déranger les animaux dans les forêts afin de les cantonner le plus possible dans ce milieu. Dans le département de l'Ain la chasse en battue se pratique de mi-septembre à fin janvier, le fond de chasse reposant sur des chasses privées. Récemment, dans ce département, des ouvertures anticipées ont permis de débuter la chasse du sanglier dès la mi-août. En 63 Haute Savoie, si la chasse se pratique également de mi-septembre à fin janvier, le système de chasse est basé sur le principe des A.C.C.A. La principale source de mortalité du sanglier dans le bassin genevois est d’origine anthropique (chasse ou tir de régulation : 81.6%, collisions routières et ferroviaires: 15.3%, autre: 3.1%; n=196 retours de marques entre 2002 et 2007). La saison de chasse débute sur la partie française entre le 8 et le 14 septembre et s’achève entre le 25 et le 31 janvier de l’année suivante. Les horaires de chasse correspondent aux horaires de coucher et de lever du soleil mais la majorité des battues se réalisent entre 8h et 19h. Les territoires de chasse s’étendent sur des surfaces allant de 3 à 20 km² suivant les communes. La chasse se pratique en battue à l’aide de petits chiens et de chiens courants, entre une à trois fois par semaines sur des zones de quelques dizaines d’hectares dans les massifs forestiers. 3.1.1.c.L’étude des sangliers du Bassin Genevois Lors de la dernière décennie, les gestionnaires de la faune sauvage du Bassin genevois ont fait face à une forte augmentation des dégâts causés par le sanglier sur les cultures. Face aux difficultés de coordination de la gestion entre les trois zones (France, Canton de Genève, Canton de Vaud) les gestionnaires locaux se sont regroupés de façon à échanger leurs connaissances, coordonner leurs actions et lancer un programme d’étude des populations locales de sanglier. En effet, aux vues des concentrations de sanglier observées dans certaines zones, la compréhension des mouvements et de la dynamique des populations locales est apparue comme une des clés permettant de résoudre la problématique de gestion. Ainsi, les Fédération Départementales des Chasseurs de l’Ain et de la Haute Savoir (France), le Domaine Nature et Paysage du Canton de Genève (Suisse), l’Office Fédérale de l’Environnement (Suisse) ainsi que le Service des Forêts, de la Faune et de la Nature du Canton de Vaud (Suisse) se sont associés avec l’Office National de la Chasse et de la Faune Sauvage (CNERA Cervidés et Sanglier) pour lancer, en 2002, une étude transfrontalière sur les déplacements et la dynamique des populations de sanglier du Bassin Genevois. Une importante campagne de capture et de suivis du déplacement des individus s’est déroulée de part et d’autre de la frontière jusqu’à la fin 64 de l’année 2007. Deux thèses ont été initiées dès 2005, l’une via l’Université de Lausanne (Hebeisen 2007), l’autre via l’Université de Savoie et l’Université Lyon 1 (ce travail) afin de tirer des enseignements, d’un point de vue plus théoriques, de cette étude dans l’optique d’améliorer la gestion du sanglier. 3.1.2. Récolte des données Entre 2002 et 2008, 548 sangliers ont été capturés, dans des cages pièges ou des corrals installés sur la partie ouest du bassin genevois (partie française : N=317, partie suisse : N=230). Lors des captures, les animaux ont été pesés puis marqués d’une boucle auriculaire, et pour 161 d’entre eux (animaux de plus de 30 kg) équipés d’un système permettant leur localisation à distance. La plupart ont été équipés de boucles auriculaires ou de collier VHF (partie française : N=55, partie suisse : N=96) mais 10 individus ont également été équipés, sur la partie française, de collier GPS (LOTEK®). Pour les animaux équipés d’émetteur VHF, le protocole de suivi consistait en (1) une localisation par jour de façon à identifier leur site de repos, et (2) de deux à dix suivis continus nocturnes (environ un point par demi heure, variable suivant le terrain) de façon à mesurer les déplacements lors de leur phase active. Le protocole de suivis par GPS consistait en une double localisation toutes les 4 heures au minimum avec pour certaines nuits des suivis continus (1pt/15min) et pour certains jours de chasse des suivis intensifs diurnes (1pts/5min). Les animaux ont été suivis, en majorités, sur moins d’une années (durée des suivi : médiane = 120 jours, quantiles 95% = 590 jours). Une base de données cartographiques a également été construite pour le site d’étude, et comprenait notamment : - un modèle numérique de terrain (source © 2004. CGIAR - Consortium for Spatial Information) - la couverture du sol via CORINE LANDCOVER (France) et la DNP de l’Etat de Genève (photo satellites). 65 - le réseau de sentiers, de routes, de voies ferrées, le réseau hydrographique et le cadastre des bâtis via l’Institut Géographique National (France) et la DNP de l’Etat de Genève. - les périmètres et le type des forêts sur la partie française (travail réalisés par des stagiaires ONCFS et des Fédération Départementales des Chasseurs de l’Ain et de la Haute Savoie). - Les contours des îlots de cultures et leur type pour 2004, 2005 et 2006 uniquement sur la partie française (source : Chambre d’Agriculture de l’Ain et de la Haute Savoie). Parallèlement, une quantification de l’effort de chasse dans l’espace a été réalisée pendant les saisons de chasse 2006 et 2007 sur les sept communes situées à proximité de la réserve naturelle de l’Etournel. Cette quantification s’est faite via l’étude des carnets de battues d’une part, et par des suivis GPS des chiens de chasse locaux lors des battues d’autre part. Pour 105 battues, un ou deux chiens ont été équipés de collier GPS GEOPOINTER® et localisés toutes les minutes du début de l’action de chasse jusqu’à leur récupération par leurs maîtres (durée de suivis = 133 +/-79 minutes, moyenne +/écart-type). 66 Ces données nous permettent donc de répondre aux objectifs de la thèse exposés précédemment, puisqu’elles permettent de mesurer les déplacements, l’utilisation de l’espace et la survie des individus. Nous pouvons ensuite mettre en relation les patrons observés avec la disponibilité des habitats physiques et biotiques locaux ainsi que les variations spatiales et temporelles du risque. De cette manière, il devient possible d’explorer à la fois la voie proximale du processus de sélection d’habitat en étudiant le mouvement des individus et l’hétérogénéité de leur réponse, et la voie ultime en étudiant les conséquences de leur choix d’habitat sur leur survie. La suite du manuscrit expose les différents points abordés dans la partie Objectif sous la forme de chapitres ou d’articles scientifiques rédigés en anglais (excepté pour la partie 5). Les résumés étendus, traduits en français, des chapitres rédigés en anglais sont néanmoins présentés avant le paragraphe discussion. 67 68 4. PARTIE 1 : CONSIDERATIONS METHODOLOGIQUES 69 ON THE USE OF TELEMETRY IN HABITAT SELECTION STUDIES Jodie Martin1,2,3, Vincent Tolon1,3,4, Bram Van Moorter1, Mathieu Basille1,5 and Clément Calenge3 1 Université de Lyon, F-69000, Lyon ; Université Lyon 1 ; CNRS, UMR5558, Laboratoire de Biométrie et Biologie Evolutive, F-69622, Villeurbanne, France. 2 Department of Ecology and Natural Resource Management, Norwegian University of Life Science, P.O. Box 5003, NO-1432 Ås, Norway 3 Office National de la Chasse et de la Faune Sauvage, 85 bis avenue de Wagram, BP 236, F75822 Paris Cedex 17, France 4 Laboratoire d’Ecologie Alpine CNRS UMR5553, Université de Savoie, Bâtiment Belledonne, F-73376 Le Bourget-du-Lac, France. 5 Centre for Conservation Biology, Department of Biology; Norwegian University of Science and Technology; NO-7491 Trondheim, Norway Published in: Barculo, D. and Daniels, J. (eds), Telemetry: research, technology and applications Nova Science Publishers Inc, 37-55 70 Abstract Understanding the relationships between organisms and their habitat is a central question in ecology. The study of habitat selection often refers to the static description of the pattern resulting from the selection process. However the very nature of this habitat selection process is dynamic, as it relies on individual movements, which are affected by both internal components (i.e. related to the animal itself, such as its behavior; foraging, resting, etc.) and external components (i.e. related to the composition of the environment). Coupling habitat selection and movement analyses should thus provide new insights into the proximal mechanisms and evolutionary causes of animals' space use. To achieve this, the introduction of GPS technology in the early 1990s showed great promise, as it facilitates tracking of animals with high fix frequency over long time periods. From a statistical point of view, this led to an increased temporal autocorrelation in the positions of successive locations. Whereas classic approaches of habitat selection often relied on the assumption of statistical independence between relocations, the development of newer methods has made possible the use of autocorrelation for more dynamic approaches. As several statistical tools are now available for researchers, autocorrelation can be incorporated successfully into the analysis, instead of being removed or even ignored. We emphasize the need to integrate individual behavioral mechanisms in habitat selection studies. The use of GPS technology in wildlife management issues is, however, often motivated by its technological advantage to produce large amounts of data, rather than biological questions. We warn users of GPS devices about the statistical and conceptual changes induced by this technology used for studying habitat selection. We encourage a solid biological reflection about the ecological characteristics of studied species and spatial and temporal scales considered, before deciding on which sampling protocol and which telemetry technology to use in accordance with the biological question of interest. 71 Introduction Understanding the relationships between organisms and their habitat is a central question in ecology. The habitat defines the available range of resources and living conditions for a species (Hall et al., 1997). Thus, the habitat potentially has an important impact on vital rates, such as survival and reproduction, which are directly related to population dynamics and evolution (Caswell, 2000). It is therefore crucial to study the mechanisms of habitat selection, i.e. the process by which animals actually choose specific habitat components within their environment. This process is recognized to be highly scale dependent; habitat selection at a given scale can be seen as the result of individual behavioral processes at finer scales (like movements), while at the same time it will be constrained by larger scale processes (like the geographical range of a species). In order to clarify the study of habitat selection, Johnson (1980) distinguished four levels or orders of selection, from the species' geographical range to the selection of food items, through individual home range establishment and patch selection within the home range. The higher levels, like the species' range, are governed mainly by population processes, whereas at lower levels individual choices are more important. To understand animal space use, ecologists have to investigate these multiple scales, often simultaneously (Johnson, 2002). It has been recognized that both spatial and temporal scales are positively correlated (Holling, 1992). For example, distributional ranges of populations are necessarily defined at large spatial scales and the study of factors affecting large scale distribution of a given population requires knowledge of long-term population dynamics at large temporal scales. On the contrary, individual movement processes at fine spatial scales (e.g. within a home range) are considerably faster than population processes, and therefore need to be studied at much finer temporal resolutions. Although “habitat selection” has been defined as a process, the study of habitat selection often refers to the static description of the pattern resulting from this process, i.e. “space occupancy”. For example, many individual-level studies dealt with the 72 characteristics of the environment within home ranges (e.g. Mc Loughlin 2002 and 2004; Mitchell and Powell 2007). However, the very nature of the habitat selection process is dynamic, as it relies on individual movements, which are affected by internal and external components (such as individual state or vegetation types). Hence, habitat selection and movement processes are intimately related, because movement partly is driven by habitat selection, whereas habitat selection is a consequence of movements. Coupling habitat selection and movement analyses should thus provide new insights into the proximal mechanisms and evolutionary causes of animals' space use. To study the mechanistic aspects of individual habitat selection, it is therefore advisable to obtain and use detailed information on movements of individuals, based on a high frequency of relocations. The transition from a static to a more dynamic approach to habitat selection therefore can be achieved by increasing serial (or temporal) autocorrelation in the positions of successive locations (Swihart and Slade, 1985). The level of serial autocorrelation in datasets describes the importance of temporal structure in individual movements. In a dynamic approach of habitat selection, the spatio-temporal structure of relocations arising from more or less complex movements should be considered and studied (Frair et al. 2005), whereas in the static approach all relocations are considered to be temporally independent (Otis and White 1999). The choice to analyze habitat selection from a static or a dynamic point of view should be motivated by the biological questions of interest. Commonly, when studying habitat selection in a static way, autocorrelation is considered a nuisance that should be overcome (Morrison et al. 1998). The use of autocorrelation in habitat selection studies reflects a shift from the more classic static approach to a dynamic approach of habitat selection. To study mechanisms of individual habitat selection, one possibility is to record movements by directly observing animals in their environment (e.g. Shine et al. 2004; Klaassen et al. 2006). However, this method of tracking animals requires many hours spent in the field, and cannot be done for elusive species, which are difficult to observe in their natural environment. Since its appearance in the early 1960s (Figure 4.1), VHF technology facilitated tracking of wildlife by providing researchers with a new tool to remotely detect individuals and then directly measure their location (consisting of a set of 73 coordinates in the form of latitude x, longitude y, and time t) by homing or triangulation. This technology has been used successfully on elusive species. However, using this technology to estimate individual small-scale movements still requires a great deal of field effort (in terms of field work) to allow the collection of frequent relocations, and therefore often results in short-duration tracking (e.g. 1 location every 10 minutes for a couple of days, Nicholls and Racey 2006). The introduction of GPS technology in the early 1990s, and its generalization at the turn of the century (Figure 4.1), showed great promise for studying processes of habitat selection. Thousands of relocations can now be stored on-board or even directly transmitted to the user, without relying on much human intervention after an animal is equipped with the tracking device. GPS technology, therefore, facilitates fine-scale tracking (relocations can be delivered up to every second, Fritz et al. 2003) of animals over longer time periods (e.g. 1 location every 3 hours for a whole year, Johnson et al. 2002). Although the accuracy of individual relocations obtained by GPS does not equal direct observations and measurements in the field, the accuracy obtained with recent GPS remains acceptable for most research purposes (around 20 meters). Hence, GPS technology facilitates the study of animal movement both at small and large spatial scales during long periods. The purpose of this chapter is twofold. The first part is dedicated to the role of autocorrelation in habitat selection studies from the past till now. We demonstrate that, whereas temporal autocorrelation was first considered as a problem when studying habitat selection in a static way, the progressive combination of habitat selection and movement concepts have led to an increased use and interest in autocorrelation. 74 Figure 4.1: Trend in the number of publication including « radio telemetry » and « global positioning system ». The review was performed using the ISI Web of Knowledge with all data bases and was restricted to zoology, biodiversity and conservation, evolutionary biology and behavioral sciences. Dashed arrows represent the appearance of important concepts in habitat selection studies and black arrows the appearance of analytical methods. As several statistical tools are now available for researchers, autocorrelation can be incorporated successfully in the analysis, instead of being removed or even ignored. We emphasize the need to integrate individual behavioral mechanisms into habitat selection studies. In the second part, we stress the importance of thorough reflections about biological questions before heading out into the field to deploy GPS collars. We provide some guidelines for the choice of monitoring technology in the context of habitat selection, regarding considerations of biological questions, spatio-temporal scales, and research costs. 75 Towards the Use of Autocorrelation in Individual Habitat Selection Studies Johnson (1980) defined the selection of a habitat component as “the process in which an animal actually chooses that component”. Thus, tests of nonrandom habitat selection by individual animals usually compare used habitat components (the relocations of individuals, i.e. their actual habitat choice) with a null model describing the habitat components that could have been used alternatively by the individuals under the hypothesis of absence of habitat selection (i.e. random habitat use; Pendleton 1998, Millspaugh and Marzluff 2001, Boyce et al. 2002). The latter components are called “available” components (Manly 2002). Rejection of this null model for the observed pattern allows the conclusion that a nonrandom process generated the observed data. At the scale of individual habitat selection, it is therefore crucial to define precisely what is available to animals, because the null model will depend directly on it. Availability can be defined differently depending on whether the animals are identified or not (see Thomas and Taylor 1990, 2006, for an overview of the different study designs). In the case of identified animals (typically the case with telemetry), availability can be related to the selection order of concern: for second-order selection availability is defined at a population level, whereas for third-order selection availability is defined at an individual level (Aebischer et al. 1993). Several other factors may further affect availability. For example, movement capacity of animals is a factor that can restrict availability (Martin et al. 2008). The behavioral state of an animal is also a potential factor that can limit access to a particular area (e.g. resting animals). These factors therefore should be taken into account when determining which habitat components are available for use. Classic statistical methods used to test habitat selection (e.g. logistic regression, log-linear regression, χ²) assume independence between locations of a given individual (Johnson 1980, Thomas and Taylor 1990, Swihart and Slade 1997, Aebisher et al 1993, Alldredge and Ratti 1992, Pendleton et al 1998). In other words, individual relocations 76 must not be spatially or temporally correlated. Dependency between relocations produces more similar values than expected by chance; as such, positive autocorrelation should result in underestimating the true variance. This induces an increased probability of type I error by inflating the number of degrees of freedom (Legendre 1993, Lennon 1999, Diniz-Filho et al. 2003, Martin et al. 2008), i.e. the null hypothesis (random habitat use) is rejected too frequently. The independence between relocations is often ensured by adopting a sufficiently large time lag between successive relocations, which circumvents the problem of autocorrelation. However, the growing use of GPS technology has led to a decrease in time intervals, and resulted in an increased serial autocorrelation. FROM THE PAST. THE NULL MODEL AS RANDOM LOCATIONS: AUTOCORRELATION AS A PROBLEM Few statistical tools have been developed for individual data analyses; most of them being adapted from methods developed for second-order selection by unidentified individuals (Calenge 2005). For example, Manly et al. (2002) recommended estimating a Resource Selection Function (RSF) for each animal and then combining the results to infer conclusions at the population level. Means and variances for each individual are therefore estimated without considering autocorrelation between relocations and then averaged across animals to estimate population level selection parameters (Thomas and Taylor 2006). As a matter of fact, several authors considered autocorrelation not to be a concern if the statistical unit is the animal and not the animal’s relocations (Alldredge and Ratti 1992, Aebisher 1993, Otis and White 1999). Indeed, there are several ways to deal with pseudo-replication when pooling a collection of relocations from several animals in analyses. However, temporal autocorrelation between individual relocations still remains a problem. When using relocations as sampling units, autocorrelation in the data makes variances and hypothesis tests no longer valid. Therefore, tools that have been developed to test habitat selection at the population level still assume independence between 77 relocations, i.e. a comparison of independent use points and independent available points (that is, randomly sampled in the study area; Figure 4.4a). This problem is critically important in third-order selection, where the null model commonly is built using points randomly sampled within what is considered to be available. This will result, when important autocorrelation exists in the animal’s relocations, in the comparison of used locations containing this autocorrelation structure with random available locations lacking such structure. Therefore, comparing autocorrelated data with uncorrelated data is not valid (Martin et al. 2008). An empirical demonstration of the effects of not considering autocorrelation in habitat selection studies has been conducted by Martin et al. (2008) on brown bears (Ursus arctos). Figure 4.2: Martin et al. (2008) tested the marginality (deviation from the average conditions in the area) in habitat selection of female brown bears, following two tests. Both relied on the comparison of the observed dataset with datasets simulated under the hypothesis of random habitat use. The first analysis did not take movement constraints into account (simulations were carried out by randomly distributing a set of points in the home range), whereas the second analysis accounted for these constraints (simulations were carried out by building random trajectories within the home range). In the first case, the observation is out of the range of the simulations and would be considered significant, whereas it is not the case while taking into account constraints (after Martin et al. 2008). 78 They compared two extreme approaches to test third-order selection by individuals with autocorrelated data. Both approaches relied on the comparison of two datasets; an observed dataset of habitat used by individuals (through individual trajectories) and a simulated dataset (with the same number of relocations as the observed dataset) under the hypothesis of random habitat use. For each dataset, they estimated marginality (an index of the eccentricity of the used environmental conditions relative to the average available environmental conditions) and tolerance (an index of the range of used environmental condition compared to the range of available conditions), two measures of the strength of habitat selection (Doledec et al. 2000). In the first approach, simulated datasets did not take into account movement constraints affecting individuals, i.e. random relocations where points are randomly and independently sampled in the home range (corresponding to the classic approach). In the second approach, simulated datasets accounted for the observed shape of individual trajectories, i.e. simulations were carried out by randomly rotating the observed trajectories within the home range. The results obtained from the two approaches differed dramatically (Figure 4.2). The second approach (with unchanged trajectory shape) led to the conclusion that bears showed no selection of the variables considered in the analysis (i.e. the marginality of the observed dataset was not different from marginalities estimated from simulated datasets). On the contrary, the classic approach concluded that there was strong habitat selection (i.e. marginality of the observed dataset was significantly different from marginalities estimated using random locations from the home range) (see Figure 4.2). However, as noted by the authors, both approaches have their limitations, because neither of them took into account the nature of the autocorrelation between relocations. Indeed, the first one did not take any movement constraints into account and the second one kept the shape of the trajectory unchanged; in this later case both movement constraints and some actual habitat selection were included in the null model, which results in a too conservative testing of habitat selection. Therefore, this study highlights the risk of not accounting for autocorrelation in animal relocations by comparing them with an inappropriate null model. 79 As autocorrelation affects our ability to perform standard statistical procedures (Legendre 1993), it has often been advised to avoid autocorrelation in individual relocations (Morrison et al. 1998). Swihart and Slade (1985) developed a framework for analyzing independence between successive relocations in order to determine the time interval necessary to achieve statistical independence. They proposed using Schoener’s (1981) ratio statistic (t²/r²; where t is the average distance between successive observations and r the average distance to the center of activity) to estimate the Time To Independence (TTI), i.e. the smallest time lag necessary to consider successive relocations as independent. Often, the TTI between consecutive fixes is considered as the time lag required by an animal to cross its entire home range (Swihart and Slade 1985, White and Garrot 1990). The problem of autocorrelation can then be effectively circumvented by subsampling data (Boyce et al. 2002) or adopting a sampling regime that uses the TTI as a criterion for independence between relocations. Unfortunately, subsampling data (which is not a problem in itself, because it only removes redundancy) inherently brings with it the loss of data which are, as every field biologist knows, expensive to collect. Moreover, several authors have shown that attempting to obtain independent data is not always possible and can lead to a loss of biological meaning (Rooney et al. 1998, De Solla et al. 1999). Serial autocorrelation is linked intimately with the definition of availability for individuals. Hence, independence between individual relocations assumes that animals are free to move between two relocations across the area that the researcher considered available, meaning that this whole area (often the home range) is available at each step. But even if the time lag between two relocations is long enough for the animal to cross this area, behavioral constraints (e.g. need for rest, need for foraging, movement constraints) result in the fact that animals are not free to move everywhere every time in this area. An extreme example illustrates this; an animal never wakes up at the other side of its home range! Therefore, even if we statistically reach the TTI, we seldom reach the biological independence between relocations with telemetry data. Moreover, when there 80 is no stable home range, there is no TTI, or the TTI might become very large, in which case subsampling will not provide a solution to the problem. There is thus a need to create a statistical framework that allows taking into account spatio-temporal structures of individual relocations, i.e. explicitly incorporating autocorrelation into models (Legendre 1993). Today, with the increased use of intensive sampling protocols, we can no longer consider fixes as independent relocations, but instead should consider them as trajectories. Analyses of animal movements are therefore needed in order to proceed with habitat selection. THE PRESENT. THE NULL MODEL IS A RANDOM WALK: ACCOUNTING FOR SERIAL AUTOCORRELATION Analyses of Animal Movement An animal’s movement is a continuous path in space and time, but a discrete representation of the path facilitates its analysis (Figure 4.3a; Turchin 1998). Often, relocations are recorded with a fixed time interval and the straight-line moves between consecutive fixes are referred to as steps. The sequence of steps then provides the basic units for further analysis of the path (Turchin 1998). Several descriptors then can be used to describe the step series; most frequently used are step lengths and turning angles (see Figure 4.3b) and their distributions can be used to characterize animal movement paths. Accordingly, movement types can be identified based on these distributions. For example, intensive search movements can be characterized by short steps and low directionality of the turning angles, whereas exploratory movement steps will be long and have a high directionality in their turning angles. 81 Figure 4.3: a) Representation of a movement path. In gray: actual path; in black: discrete representation of the path. A movement path is defined by a set of successive relocations, characterized by their position (generally latitude and longitude). Each movement between two successive relocations (i.e. between time t and t+1) describes steps. b) Examples of movement path descriptors: d is the distance between 2 relocations; Ar is the turning angle (or relative angle), i.e. the angle between the direction of the previous step (small dashed line) and the actual one; Aa is the absolute angle, i.e. the angle between a given direction (gray line) and the direction of the actual step. The statistical framework for animal movement path analyses is based on the comparison of empirical data with a theoretical null model. Several theoretical movement models have been developed that could be used as reference. Each of these models has different statistical properties, and often different assumptions about the independence (i.e. absence of serial autocorrelation) of specific movement descriptors (like step length or relative turning angles). The observed movement characteristics are then compared with the theoretical ones and departure from these models allows inference of biological conclusions about animal behavior (Franke et al. 2004, see below). The mere random walk (RW) is the simplest null model of movement which can be used to model animal movement through a homogeneous environment (Turchin 1998; Figure 4.4b). It relies on the independence of all descriptors, and thus it assumes the absence of serial correlation between successive steps. Therefore a random walk does not take into account the natural tendency of animals to go forward and is therefore not very accurate to represent most of animal movement, especially at short time intervals (Turchin, 1998; Bovet and Benhamou, 1988). Today, the most widely used theoretical model is the correlated 82 random walk (CRW), which differs from the RW in that absolute angles are generally dependent, whereas the relative or turning angles are assumed to be independent (Turchin 1998; Bovet and Benhamou, 1988). The distribution of turning angles is centered on 0, resulting in a forward persistence in the direction of movements. The direction of the previous step thus influences the direction of the following step. Figure 4.4: a) The null model as a set of locations randomly sampled in the study area; b) The null model as a random walk, where the whole trajectory can be characterized by step length distribution (top right) and turning angle distribution (lower right); c) The null model as a mixture of two random walks. In gray, a mere random walk, characterized by a uniform distribution of turning angles and normally distributed step lengths; in black a correlated random walk characterized by a distribution of turning angles centered around 0 (forward persistence) and normally distributed step length. 83 DISCRETE-CHOICE MODELS Some authors have started to take serial autocorrelation into account by defining habitat availability separately for each relocation in the so-called “discrete-choice models” framework (Arthur et al. 1996, Hjermann 2000, Fortin et al. 2005, Rhodes et al. 2005). The test for nonrandom habitat selection is derived from a comparison of random locations from this fix-specific availability with the actual chosen locations. In this case, the null model consists of random locations with serial autocorrelation, similar to the class of random walk models. Different authors have used different methods to determine availability, corresponding to different types of random walks. The simplest approach involves determining a circle around a fix of available locations (e.g. the availability radius of Arthur 1996); in this case there is no directional persistence. Alternatively, the observed distributions of step lengths and turning angles have been used to define availability. Even more complex relationships are possible with a dependence of availability on the time interval between fixes or habitat characteristics (Hjermann 2000). It is therefore assumed that for a given time lag between two relocations, the animal has access only to areas close to the current position, and not to its complete home range (Arthur et al. 1996). In the same perspective, Cooper and Millspaugh (1999) adopted a statistical technique derived from the field of economics that allows researchers to define availability separately for each animal observation. At each relocation, a unique set of habitat or resources is available, called the “choice set”. Fortin et al. (2005) also developed a simple statistical approach that incorporates movement into a logistic regression framework. This method, called Step Selection Function, considers steps (displacement between two relocations) as sampling units, each of them being contrasted with n random steps, which are defined using the observed distributions of step lengths and turning angles. This approach partially resolves the statistical and biological issues of serial autocorrelation of relocation data. Indeed, this procedure only considers first-order autocorrelation, i.e. dependence between relocations at time t and t-1 only, and deems the successive steps as independent (Martin et al. 2008). However, the nature of the 84 dependence between all the relocations making up the whole trajectory is seldom analyzed, as noted by Calenge (2005). Martin et al. (2008) stressed that positions of individuals are the results of three effects: (i) intrinsic constraints (e.g. movement capacity, activity patterns, internal state), (ii) extrinsic constraints (e.g. environmental or artificial barrier) and (iii) habitat selection behavior (the animal is in a particular habitat because it is “suitable” for it at this moment). Therefore, testing differences between used and available points suppose knowledge of the processes that generate data without any habitat selection, i.e. null model under the hypothesis of no habitat selection, by taking into account the internal constraints of animals that partly shape the trajectory. However, habitat selection analyses rarely take into account intrinsic and extrinsic constraints (Martin et al. 2008; see Matthiopoulos (2003) for an example accounting for movement and extrinsic constraints). In general, researchers aggregate data from different behavioral states and, therefore, their conclusions on habitat selection result from the joint selection of both activity and habitat components (Cooper and Millspaugh 2001). As habitat selection and animal behavior are closely related, there is need for a statistical approach that includes spatio-temporal aspects of individual behavior (Thomas and Taylor 2006, Martin et al. 2008). INTO THE FUTURE. THE NULL MODEL IS A MIXTURE OF RANDOM WALKS: AUTOCORRELATION AS A PARADIGM Nonstationarity, an Interesting Property of Many Animal Trajectories As movements and activities are closely related, movement processes tend to be different according to the animal’s behavioral state. For instance, during foraging activity an animal may have shorter and more sinuous movements than during transitions 85 between patches of resources, where it should have directed and faster movements. As an illustration, consider a bee foraging on a patch of flowers. Movements between flowers will be short and sinuous compared with movement between patches of flowers or between the patch and the hive. Franke et al. (2004) used differences in movement characteristics to differentiate behavioral states of woodland caribou; they distinguished bedding, feeding and displacements. Indeed, for given spatial and temporal scales, we can consider most animal trajectories as a succession of different types of movement corresponding to specific activities, each of them being characterized by its own statistical properties (Figure 4.4c). However, trajectories can remain unchanged for different activities defined at very fine temporal scales (animals can switch quickly between foraging and vigilance while keeping the same type of movement). But at longer temporal scales, major activities, such as foraging, exploring, or resting, often correspond to specific movement types. A statistical process is said to be nonstationary if the statistical properties of the process generating the trajectory change over time. Therefore, a trajectory composed of different movement types may be considered to have been generated by a nonstationary process. This implies that the definition of availability should be different according to the state of the individual. Indeed, an animal that is foraging or resting does not have the same available habitat as when it is searching for mates or patrolling its territory. Therefore, this nonstationarity is of major interest, because it provides information on animal behavior and activities. This nonstationarity of the process is often the cause of the autocorrelation in the data. BUILDING MOVEMENT MODELS AS MIXTURES OF RANDOM WALKS Each movement type trajectory potentially can be represented by a different theoretical model based on its properties. The trajectory can then be modeled as a succession of these movement models (Figure 4.4c). It is therefore important to partition 86 the whole trajectory into different pieces of stationary paths, with stable mathematical properties. Typically, each of these stationary paths corresponds to a certain type of behavior. To date, several methods have been developed and used for this partitioning of trajectories, for instance First Passage Time (Fauchald and Tveraa 2003), fractal dimension (Nams and Bourgeois 2004), and State Space Modeling (Patterson et al., 2007). First Passage Time, for instance, has been used as a method to detect AreaRestricted Search (ARS) behaviors, which can occur when an animal encounters a foodrich resource patch. More recently, state-space models based on hidden Markov models have become more popular for extraction of behavioral states from movement paths (reviewed by Patterson et al. 2007). For example, Morales et al. (2004) employed statespace models to highlight a biphasic movement for elk (Cervus elaphus); the “encamped” movement with small movements and sharp turns and the “exploratory” movements with longer directed movements. They modeled elk movement by fitting a mixture of random walk models with different properties, each model corresponding to a different behavioral state. However, these approaches often assume a constant probability of animal behavior changes (“switching probability”), or at least assume prior knowledge of factors that could potentially affect this switching probability (e.g. constraints on hourly activity patterns or environmental features, Morales et al. 2004). Such methods therefore require prior exploratory analyses of factors potentially influencing the shape structure of the trajectory. SOME RECOMMENDATIONS In order to test third-order habitat selection with highly autocorrelated data, we stress that an in-depth analysis of the characteristics of individual trajectories is an important step towards a more accurate analysis of habitat selection process. Autocorrelation between relocations should not be removed or avoided, but rather integrated into a statistical framework. Discrete-choice models are a first step toward this 87 integration, but only consider the first degree of autocorrelation to create the null model. Therefore, they do not take into account the behavioral state of individuals, which potentially may affect habitat selection behavior. We emphasize the need to analyze the rules of animal movement using partitioning methods based on the division of the whole trajectory into homogeneous movement bouts. Each of these movement bouts can then be characterized by a probability distribution for each descriptor (e.g. step length, turning angles). Then, for each relocation belonging to a particular behavioral state, availability can be estimated more precisely using the corresponding movement characteristics for this state. These state-movement analyses are required to build more realistic null models of random habitat selection that take into account behavioral constraints. GPS Technology: A Double-Edged Sword The appearance of new technologies has resulted in important advances in many scientific fields, as it offers new opportunities to answer more questions. Since the early 1990s, GPS technology has facilitated the measure of fine-scale movements of elusive animals in their natural environments over long time periods. Especially, it aids in our understanding of the link between fine-scale behavioral movement mechanisms and the actual distributions of animals. However, scientists, wildlife managers, and conservationists should be careful before adopting this technology. Even though GPS technology is appealing, the choice of the monitoring tool should be the consequence of a well-defined biological question. Such careful planning could aid the avoidance of mismatches between the question of interest and the type of data collected using a particular tool. Compared to older technologies (especially VHF) the use of GPS technology offers many advantages, but also induces different constraints and is not necessarily the adequate tool for every question about animal movements or distributions. Above all, the choice of the appropriate tool to record an animal’s locations should be 88 directly dependent on the sampling protocol defined itself by the question of interest. Irrespective of the biological question or sampling protocol, it is an accepted fact that the number of equipped individuals should be as large as possible in order to increase the generality of the findings. Two other parameters can then vary according to the question of interest: the time lag between relocations and the study period. In the following, we will discuss the relationships between biological question, sampling protocols, and choice of the adequate tool to record animal locations. INDIVIDUAL VARIABILITY The ecological characteristics of the focus species and especially the ratio of inter/intra-individual variability in habitat selection can help determine the best sampling protocol and therefore the right tool to track animals in their environment. Girard et al. (2006) showed in their study that the number of animals is more important to assess habitat selection than is the number of fixes per animal. They obtained accurate habitat selection by moose with fix frequencies of only 1-7 per week, whereas often more than 15 individuals were needed to generalize the findings over the population. This suggests that, for generalist species with high inter-individual variability, it is important to obtain data from many individuals to assess habitat selection. On the contrary, if there is a strong intra-individual variability, the number of relocations obtained for each individual can become as important as the number of tracked individuals. Thus, knowledge about the biology of the species will influence the choice of sampling protocol (number of samples per individual versus number of sampled individuals) and, by consequence, the monitoring technology. 89 A MATTER OF SCALE The fix frequency of a telemetry protocol and the study period are generally a direct consequence of the scale of the study: questions regarding small-scale movements require high fix frequency (e.g. 1 point every second for a fractal analysis of albatrosses’ movements, Fritz et al. 2003). In the beginning of this chapter we discussed the interest to combine fine-scale movement analyses with habitat selection studies. In this context, increased fix frequency may facilitate a more profound investigation of animal behavior. Moreover, the study duration may be long enough to explore the link between fine-scale behaviors and habitat selection at longer time scales. However, many questions in ecology do not rely directly on fine-scale behaviors of animals. Large-scale distributional questions can be answered with a lower fix frequency (e.g. 1 point every 3 weeks for an analysis of home range composition and habitat use, McLoughlin et al. 2005). Important in the context of conservation and population dynamics in general is the linkage between habitat and animal performance (McLoughlin et al. 2005), for example to identify critical habitats for population viability in conservation biology (e.g. Akçakaya 1995). Mere occurrence has been shown to be misleading in some situations. For instance, an attractive sink is a habitat that is selected despite the lower performance experienced by the animals occupying it (Delibes et al. 2001). It can, therefore, be argued that to assess the existence of such attractive sinks, the measurement of performance is critical. Performance of animals, e.g. lifetime reproductive success, however is defined over quite long time scales and often is related to large-scale distribution patterns (e.g. habitat use, Conradt et al. 1999, home range composition, McLoughling et al. 2007). These measures of space occupancy do not require fine-scale measures of animals’ movements. Instead, a high number of individuals is often needed to highlight relationships between their performance and habitat. In this context it might be more interesting to invest less in fix frequency and more in number of tracked individuals. 90 MAIN COSTS AND BENEFITS OF GPS TECHNOLOGIES Tools and sampling protocols are intimately related, they often involve trade-offs between costs and perfect match. Prior to the selection of a telemetry technology, researchers should think about the associated costs and benefits (in terms of correspondence with the defined protocol) of the use of different tools within the biological context of the question. In theory, any study using VHF tracking by triangulation could be done with GPS tracking instead. That is, GPS collars potentially can deliver the exact same data (regarding frequency and time-lag) as VHF collars, whereas the reverse is not necessarily true over a long period. That being said, a study using telemetry will cover three main budget compartments; capture, equipment, and operation costs. We will consider only the case of nonlimiting captures, as it potentially involves limiting equipment and operating costs directly related to the monitoring technology to be chosen. Figure 4.5. The sampling strategy, at a fixed cost, is the result of a trade-off between the number of individuals marked (N) and the number of relocations per individual (i). Typically, VHF monitoring allows a large number of individuals with fewer relocations per individual (thus no autocorrelation), whereas GPS monitoring allows fewer individuals with more relocations (thus autocorrelation). 91 The cost of equipment to monitor one animal is considerably higher with GPS technology than with VHF technology; GPS-tracking devices are approximately 10 times more expensive than VHF devices. However, with GPS devices the collection of data is automated, whereas VHF devices require human intervention to be effective; relocations are usually collected by triangulation of the signal, which implies a relatively high time and financial investment in the field. This results in limited operation costs of GPS technology (even with the use of a GSM device to download the data) as compared to VHF technology, especially for high fix frequencies. At a fixed cost, GPS technology, therefore, leads to an increase in sampling intensity compared to VHF technology. However, when the budget of the study is limited, this is often at the cost of lower numbers of individuals monitored (Figure 4.5). It is thus important to take into account these considerations of costs before deciding upon the adequate tool to choose for the sampling protocol. SOME RECOMMENDATIONS When a large number of relocations per animal must be collected, GPS technology can provide adequate data, thanks to the automation of the process. This technology is particularly useful when fine-scale movements of animals need to be recorded, especially over long periods of time. This tool is, therefore, most appropriate to study dynamic aspects of habitat selection, as reported in this chapter. However, when the number of individuals is critical (as for generalist species, see above), and no intensive tracking is required, other tools can be more appropriate (e.g. VHF). For example, when ecologists must link habitat selection and animal performance on long time scales, it is better to invest in more individuals, tracking duration, and field personnel to obtain the essential animal performance measures (source of mortality, breeding, litter size, etc.). The GPS technology does not seem to be the best tool for this task and VHF tools combined with direct observations can provide sufficient locations per animals to 92 estimate their habitat selection at a large spatio-temporal scale. In this case, the savings made on equipment can be used to increase the number of tracked individuals. In general, at constant costs, what is gained in sampling intensity with GPS technology, on one hand is lost in generality on the other hand. It should be noted, however, that the increased fix frequency obtained by using GPS-tracking might prove useful in determining animal performance using their movement patterns. GPS-tracking can be used to assess foraging success in particular cases. For instance, it is now commonly used to determine kill sites by large predators (e.g. cougar (Felis concolor), Anderson and Lindzey 2003; wolf (Canis lupus), Sand et al. 2005). These applications require previous calibration and validation of the models on the field. These examples are likely only the top of the proverbial iceberg of potential applications; we can expect more applications with our increasing knowledge of how movement patterns change with specific factors, like the presence of offspring etc. Conclusion Despite some budget limitations, GPS technology offers interesting avenues for our understanding of the habitat selection process. The shorter time intervals between relocations allow for the study of more rapidly changing decisions of habitat selection at small spatio-temporal scales. The increased temporal autocorrelation also allows the detailed investigation of different behavioral states with different movement characteristics. Whereas static approaches of habitat selection often rely on the assumption of statistical independence of relocations, the development of newer methods, driven by the collection of relocations with a shorter time lag, now allow the use of autocorrelation for more dynamic approaches. We expect such dynamic habitat selection studies to become even more common in the near future, as ecologists become more 93 familiar with the use of methods from time-series analysis like state-space models (Patterson et al. 2008). Coupling habitat selection and movement analyses should provide new perspective to understand how individuals react to environmental heterogeneity during their lifetime. The use of GPS technology in wildlife management or conservation issues, however, often is motivated by technological advantage rather than biological questions. The appealing nature of GPS data (through higher precision and frequency, as well as automation of the data collection) often results in sampling strategies targeting large amounts of data. This often leads afterwards to data-dredging that can cause mismatches between the scale of the monitoring and the scale of the biological processes of interest. We warn every user of GPS devices about the statistical and conceptual changes induced by this technology used for studying habitat selection (Figure 4.5). In answer to the potential increase in autocorrelation due to higher fix frequency following technological developments in telemetry, we call for a paradigmatic shift in the study of animal habitat selection from static patterns towards dynamic processes. Especially with critical management or conservation issues, we stress that every ecologist should take care to start from the biological question at hand in making his choice of the most appropriate monitoring technology. Acknowledgements Financial support has been provided by the ANR (project “Mobilité” ANR-05BDIV-008). We thank Jean-Michel Gaillard (Laboratoire de Biométrie et Biologie Evolutive; Université Lyon1, France) and our referee Eloy Revilla (Estación Biológica de Doñana, Sevilla, Spain) for their useful comments on the manuscript. We also are grateful to the working group “Trajectometry” (Laboratoire de Biométrie et Biologie Evolutive; Université Lyon1, France) for earlier discussions on the topic. 94 References Aebischer, N.J., Robertson, P.A., and Kenward, R.E. (1993) Compositional analysis of habitat use from animal radio-tracking data. — Ecology, 74, 1313-1325. Alldredge, J.R. and Ratti, J.T. (1992) Further comparison of some statistical techniques for analysis of resource selection. — Journal of Wildlife Management, 56, 1-9. Akçakaya, H.R., McCarthy, M.A. and Pearce, J.L. (1995) Linking landscape data with population viability analysis: Management options for the helmeted honeyeater Lichenostomus melanops cassidix. — Biological Conservation, 73, 169-176. Anderson, C.R. and Lindzey, F.G. (2003) Estimating cougar predation rates from GPS location clusters. — Journal of Wildlife Management, 67, 307-316. Arthur, S.M., Manly, B.F.J., McDonald, L.L., and Garner, G.W. (1996) Assessing habitat selection when availability changes. — Ecology, 77, 215-227. Bovet, P. and Benhamou S. (1988) Spatial analysis of animals’ movements using a correlated random walk model. — Journal of Theoretical Biology, 131, 419-433. Boyce, M.S., Vernier, P.R., Nielsen, S.E., and Schmiegelow, F.K.A. (2002) Evaluating resource selection functions. — Ecological Modelling, 157, 281-300. Calenge, C. (2005) Des outils statistiques pour l'analyse des semis de points dans l'espace écologique., Université Claude Bernard Lyon 1, Lyon. Caswell, H. 2000 Matrix population models: construction, analysis, and interpretation, second edition. Sinauer, Sunderland, Massachusetts, USA Conradt, L., Clutton-Brock, T.H. and Guinness, F.E. (1999) The relationship between habitat choice and lifetime reproductive success in female red deer. — Oecologia, 120, 218-224. Cooper, A.B. and Millspaugh, J.J. (1999) The application of discrete choice models to wildlife resource selection studies. — Ecology, 80, 566-575. Cooper, A.B. and Millspaugh, J.J. (2001). Accounting for variation in resource availability and animal behavior in resource selection studies. In Radio tracking and 95 animal population (eds J.J. Millspaugh and J.M. Marzluff). Academic Press, San Diego, California, USA. Dahle, B. and Swenson, J.E. (2003) Seasonal range size in relation to reproductive strategies in brown bears Ursus arctos. — Journal of Animal Ecology, 72, 660-667. De Solla, S.R., Bonduriansky, R., Brooks, R.J.D.-S.S.R., R, B., and J, B.R. (1999) Eliminating autocorrelation reduces biological relevance of home range estimates. — Journal of Animal Ecology, 68, 221-234. Delibes, M., Gaona, P., and Ferreras, P. (2001) Effects of an attractive sink leading into maladaptive habitat selection. — The American Naturalist, 158, 277-285. Diniz, J.A.F., Bini, L.M., and Hawkins, B.A. (2003) Spatial autocorrelation and red herrings in geographical ecology. — Global Ecology and Biogeography, 12, 53-64. Doledec, S., Chessel, D., and Gimaret-Carpentier, C. (2000) Niche separation in community analysis: a new method. — Ecology, 81, 2914-2927. Erickson, W.P., McDonald, T.L., Gerow, K., Kern, J., and Howlin, S. (2001). Statistical issues in resource selection studies with radio-marked animals. In Radio Tracking and Animal Populations. (eds J.J. Millspaugh and J.M. Marzluff), pp. 209-242. Academic Press, San Diego, California, USA. Fauchald, P. and Tveraa, T. (2003) Using first-passage time in the analysis of arearestricted search and habitat selection. — Ecology, 84, 282-288. Ferreras, P., Delibes, M., Palomares, F., Fedriani, J.M.,, Calzada, J. and Revilla, E. (2004) Proximate and ultimate causes of dispersal in the Iberian lynx Lynx pardinus. — Behavoral Ecology, 15 31-40. Fortin, D., Beyer, H.L., Boyce, M.S., Smith, D.W., Duchesne, T., and Mao, J.S. (2005) Wolves influence elk movements: behavior shapes a trophic cascade in Yellowstone National Park. — Ecology, 86, 1320-1330. Frair, J.L., Merill, E.H., Visscher, D.R., Fortin, D., Beyer, H.L. and Morales, J.M. (2005) Scales of movement by elk (Cervus elaphus) in response to heterogeneity in forage resources and predation risk. — Landscape Ecology, 20, 273-287. 96 Franke, A., Caelli, T., and Hudson, R.J. (2004) Analysis of movements and behavior of caribou (Rangifer tarandus) using hidden Markov models. — Ecological Modelling, 173, 259-270. Fritz, H., Said, S., and Weimerskirch, H. (2003) Scale-dependent hierarchical adjustments of movement patterns in a long-range foraging seabird. — Proceedings of the Royal Society of London Series B-Biological Sciences 270 1143-1148. Girard, I., Dussault, C., Ouellet, J.-P., Courtois, R., and Caron, A. (2006) Balancing number of locations with number of individuals in telemetry studies. — Journal of Wildlife Management, 1249-1256. Hall, L.S., Krausman, P.R., and Morrison, M.L. (1997) The habitat concept and a plea for standard terminology. — Wildlife Society Bulletin, 25, 173-182. Harris, S., Cresswell, W.J., Forde, P.G., Trewhella, W.J., Woolard, T., and Wray, S. (1990) Home-range analysis using radio-tracking data - a review of problems and techniques particularly as applied to the study of mammals. — Mammal Review, 20, 97-123. Hirzel, A.H., Hausser, J., Chessel, D., and Perrin, N. (2002) Ecological-niche factor analysis: How to compute habitat-suitability maps without absence data? — Ecology, 83, 2027-2036. Hjermann, D.O. (2000) Analyzing habitat selection in animals without well-defined home ranges. — Ecology, 81, 1462-1468. Johnson, D.H. (1980) The comparison of usage and availability measurements for evaluating resource preference. — Ecology, 61, 65-71. Johnson, C.J., Parker, K.L., Heard, D.C., and Gillingham, M.P. (2002) Movement parameters of ungulates and scale-specific responses to the environment. — Journal of Animal Ecology, 71, 225-235. Klaassen, R.H.G., Nolet, B.A., and Bankert, D. (2006) Movement of foraging tundra swans explained by spatial pattern in cryptic food densities. — Ecology, 87, 22442254 Legendre, P. (1993) Spatial autocorrelation: trouble or new paradigm? — Ecology, 74, 1659-1673. 97 Lennon, J.J. (1999) Resource selection functions: taking space seriously? — Trends in Ecology and Evolution, 14, 399-400. Manly, B.F.J., McDonald, L.L., Thomas, D.L., MacDonald, T.L., and Erickson, W.P. (2002) Resource selection by animals. Statistical design and analysis for field studies Kluwer Academic Publisher, London. Martin, J., Calenge, C., Quenette, P.-Y., and Allainé, D. (2008) Importance of movement constraints in habitat selection studies. — Ecological Modelling, 213, 257-262. Matthiopoulos, J. (2003) The use of space by animals as a function of accessibility and preference. — Ecological Modelling, 159, 239-268. McLoughlin, P.D., Dunford, J.S. and Boutin, S. (2005) Relating predation mortality to broad-scale habitat selection. — Journal of Animal Ecology, 74, 701-707. McLoughlin, P.D., Gaillard, J-M. Boyce, M.S., Bonenfant, C., Messier, F., Duncan, P., Delorme, D., Van Moorter, B., Saïd, S., and Klein, F. (2007) Lifetime reproductive success and composition of home range in a large herbivore. — Ecology 88, 3192– 320. Millspaugh, J.J. and Marzluff, J.M. (2001) Radio tracking and animal populations. Academic Press, San Diego, California, USA. Mitchell M.S. and Powell R.A. (2007) Optimal use of resources structures home ranges and spatial distribution of black bears. — Animal Behaviour, 74, 219-230. Morales, J.M., Haydon, D.T., Frair, J., Holsiner, K.E., Fryxell, J.M., and Holsenger, K.E. (2004) Extracting more out of relocation data: Building movement models as mixtures of random walks. — Ecology, 85, 2436-2445. Morrison, M.L., Marcot, B.G., and Mannan, R.W. (1998) Wildlife-habitat relationships: concepts and applications. University of Wisconsin Press. Nams, V.O. and Bourgeois, M. (2004) Fractal analysis measures habitat use at different spatial scales: an example with American marten. — Canadian Journal of Zoology, 82, 1738-1747. Nicholls, B. and Racey, P.A. (2006) Contrasting home-range size and spatial partitioning in cryptic and sympatric pipistrelle bats. — Behavioral Ecology and Sociobiology 61, 131-142. 98 Otis, D.L. and White, G.C. (1999) Autocorrelation of location estimates and the analysis of radiotracking data. — Journal of Wildlife Management, 63, 1039-1044. Patterson, T.A., Thomas, L., Wilcox, C., Ovaskainen, O., and Matthiopoulos, J. (2008) State-space models of individual animal movement. — Trends in Ecology and Evolution, 23, 87-94. Pendleton, G.W., Titus, K., DeGayner, E., Flatten, C.J., and Lowell, R.E. (1998) Compositional analysis and GIS for study of habitat selection by goshawks in southeast Alaska. — Journal of Agricultural, Biological, and Environmental Statistics, 3, 280-295. Rhodes, J.R., McAlpine, C.A., Lunney, D., and Possingham, H.P. (2005) A spatially explicit habitat selection model incorporating home range behavior. — Ecology, 86, 1199-1205. Rooney, S.M., Wolfe, A. and Hayden, T.J. (1998) Autocorrelated data in telemetry studies: time to independence and the problem of behavioural effects. — Mammal Review, 28, 89-98. Sand, H., Zimmermann, B., Wabakken, P., Andren, H., and Pedersen, H.C. (2005) Using GPS technology and GIS cluster analyses to estimate kill rates in wolf-ungulate ecosystems. — Wildlife Society Bulletin, 33, 914-925. Schoener, T.W. (1981) An Empirically Based Estimate of Home Range. — Theoretical Population Biology, 20, 281-325. Shine, R., Lemaster, M., Wall, M., Langkilde, T. and Mason, R. Why did the snake cross the road? Effects of roads on movement and location of mates by garter snakes (Thamnophis sirtalis parietalis). — Ecology and Society, 9, 9. http://www.ecologyandsociety.org /vol9/iss1/art9. Swihart, R.K. and Slade, N.A. (1985) Testing for independence of observations in animal movements. — Ecology, 66, 1176-1184. Swihart, R.K. and Slade, N.A. (1997) On testing for independance of animal movements. — Journal of Agricultural, Biological, and Environmental Statistics, 2, 48-63. Thomas, D.L. and Taylor, E.J. (1990) Study designs and tests for comparing resource use and availability. — Journal of Wildlife Management, 54, 322-330. 99 Thomas, D.L. and Taylor, E.J. (2006) Study designs and tests for comparing resource use and availability II. — Journal of Wildlife Management, 70, 324-336. Turchin, P. (1998) Quantitative analysis of movement: measuring and modelling population redistribution in animals and plant. Sinauer Associates, Sunderland, MA. White, G.C. and Garott, R.A. (1990). Habitat analysis. In Analysis of Wildlife RadioTracking Data. (ed U.K. Edition), pp. 183-205. Academic Press, Inc. 100 101 102 5. PARTIE 2 : REPONSES AUX VARIATIONS SPATIALES ET TEMPORELLES DU RISQUE 103 RESPONDING TO SPATIAL AND TEMPORAL VARIATIONS IN PREDATION RISK: SPACE USE OF A GAME SPECIES IN A CHANGING LANDSCAPE OF FEAR V. Tolon 1, 2, 5, S. Dray 2, A. Loison 1, A. Zeileis 3, C. Fischer 4 and E. Baubet 5. 1 Laboratoire d’Ecologie Alpine CNRS UMR5553, Université de Savoie, Bâtiment Belledonne, F-73376 Le Bourget-du-Lac, France. E-mail: [email protected]. 2 Université de Lyon, F-69000, Lyon ; Université Lyon 1 ; CNRS, UMR5558, Laboratoire de Biométrie et Biologie Evolutive, F-69622, Villeurbanne, France. E-mail: [email protected]. 3 Department of Statistics and Mathematics, Wirtschaftsuniversität Wien, Augasse 2-6, A-1090 Wien, Austria. E-mail: [email protected]. 4 Ecole d’Ingénieurs de Lullier, S-1254 Jussy, Switzerland. E-mail: [email protected]. 5 Office National de la Chasse et de la Faune Sauvage, Centre National d’Etude et de Recherche Appliquée - Cervidés et sanglier, F-55 000 Bar-le-Duc, France. E-mail: [email protected]. Published in Canadian Jounal of Zoology 87(12): 1129-1137. 104 ABSTRACT Predators generate a “landscape of fear” within which prey can minimize the risk of predation by selecting low risk areas. Depending on the spatial structure of this “landscape”, i.e. whether it is coarse or fine grained, prey may respond to increased risk by shifting their home ranges or by fine-scale redistributions within these ranges, respectively. We studied how wild boar (Sus scrofa Linnaeus, 1758) responded to temporal changes in risk in hunted areas (risky habitat) surrounding a nature reserve (refuge habitat). Animals with home ranges “in contact” with the reserve during the low risk season were the only ones to shift toward the refuge when the risk increased. These shifts occurred at two temporal scales in response to the increased risk during the daytime and during the hunting season. Whereas animals not influenced by the reserve found food and shelter in forest during the hunting season, shifts to the refuge area were detrimental to the rather scarce forest areas in the reserve. This confirms that spatiotemporal changes in risk are major drivers of animal distribution when predation strongly limits their fitness. Their response is however scale-dependent, and reflects at the individual level the perceived structure of their “landscape of fear”. Key words: wild boar, Sus scrofa, reserve, risk, home range, landscape of fear. 105 INTRODUCTION By definition, the act of predation leads to the death of the prey, but predator-prey interactions can also have non-lethal or indirect effects (Creel and Christianson 2008), because prey often try to avoid the risk of predation by changing their behavior. For instance, animals can switch activity patterns (Lima and Bednekoff 1999), or modulate vigilance levels (Roberts 1996) or group size (Hamilton 1971) in response to increased predation risk, usually incurring energetic costs. The concept of the “ecology of fear” (Brown et al. 1999) concerns the behavioral responses of prey to stress and fear caused by predators, and their consequences for individual fitness and ecosystem functioning through trophic cascades (Ripple and Beschta 2004). Predation risk is often structured in space and time at multiple scales. Attack success or accessibility to predators can vary in different habitats (Cowlishaw 1997), but predators' activity can also be intrinsically structured in space (e.g. Fortin et al. 2005) and in time (Sih et al. 2000). For prey that perceive these heterogeneities, these spatiotemporal variations of predation risk generate a structured and dynamic “landscape of fear” (Laundré et al. 2001), within which they can minimize risk by relocating to lower-risk or safe areas. Changes in the use of space is one of the most widespread behavioral responses to spatiotemporal differences in predation risk over a given resource distribution area (Brown 1988, Abrahams and Dill 1989, van der Merwe and Brown 2008). According to Rettie and Messier (2000), the scale at which individuals respond to predation risk depends on the predominant factor limiting individual fitness (e.g. resources or predation). In a context of high predation pressure, the way animals adapt their use of space may depend on spatiotemporal variations of risk, rather than on other environmental factors, such as food abundance or thermal conditions. However, the risk-dependent distribution of animals can also be determined by their perception or knowledge of local risk levels and/or by their ability to move away from high-risk areas. Many adaptive distribution theories assume that individuals are omniscient and can move freely (e.g. Fretwell and Lucas 1970), however this is rarely the case in the natural environment (Kennedy and Gray 1993, Shochat et al. 2002). For 106 instance, the use of refuges by prey can depend on several environmental and behavioral factors, such as competition for refuges (Berryman and Hawkins 2006), their accessibility and knowing where they are located (Clarke et al. 1993), or social constraints (Banks et al. 2007). Individuals may therefore only react if differences in risk levels are perceived as spatially structured within the space available to them. Depending on the local spatial structure of risk, i.e. whether it is coarse or fine grained, animals may respond either by shifting their home ranges or by fine-scale redistribution within the home range, respectively. For a broad range of animal species, a considerable proportion of their mortality is anthropogenic (e.g. road accidents, Bruinderink and Hazebroek 1996), and especially by harvesting populations (e.g. hunting, Toïgo et al. 2008). Most harvested species face marked temporal variations in risk due to restricted seasons, specific days or times of day when hunting is permitted (e.g. Toïgo et al. 2008). Moreover, the existence in the vicinity of protected areas where human access is restricted induces marked spatial differences in risk (Madsen 1998). This means that hunting can shape the landscape of fear for the harvested species, as the level of risk differs considerably in terms of space (the existence of protected areas), and of time (periods when hunting is permitted). If individual fitness is mainly determined by the risk of being hunted, animals will strongly seek protected areas during pulses of risk. In the absence of risk (a “flat” landscape of fear), the distribution of the animals will be driven mainly by other factors (often the availability of resources). The wild boar (Sus scrofa Linnaeus, 1758) is a game species that often faces extreme hunting pressures (e.g. a 40% of chance of being harvested annually, and up to 70% for adult males, Toïgo et al. 2008). Consequently, its use of space changes markedly when the risk of being predated (through hunting) starts (Boitani et al. 1994, Sodeikat and Pohlmeyer 2003, Keuling et al. 2008b). In the study reported here, we investigated how wild boar responded to temporal changes in risk in hunted areas (risky habitat) surrounding a protected nature reserve (refuge habitat). The wild boar is known to shift its home range toward forest areas providing food and shelter during the hunting season, (e.g. Boitani et al. 1994), but we expected to find that home range shifts near the reserve 107 would also be driven by the local protection status. Individuals for whom the presence of a reserve creates widely contrasting risk levels within their locally-available environment could be expected to display the greatest shifts in their home ranges. We therefore hypothesized that there could be a threshold distance for this response, depending on how far from the reserve the wild boar range during the closed season (prediction 1). Controlling for differences in forest use, we expected diurnal pulses of risk to drive the animals back into the protected area every morning, creating a circadian pattern of space use (prediction 2). We also expected the marked increase in risk during the hunting season to draw animals deeper into the reserve than during the closed season (prediction 3). However these seasonal movements might not appear at night (prediction 4) because the nocturnal risk does not reflect this seasonal variation. METHODS Study area The Geneva basin is located on the French-Swiss border (46.135, 5.930, WGS84), and is surrounded by ranges of mountains and Lake Geneva. The reserve was set up in 1962, and occupies 318 ha (Fig. 5.1). Habitat types in the reserve correspond to hygrophilous forest (Alnus spp, Salix spp; 44%), reed beds (Phragmites spp; 30%), surface water (16%), oak or beech woods (Quercus spp, Fagus spp; 6%) and crops (maize, cereals; 4%). The area immediately around the reserve (the surrounding 4 km, equivalent to 9000 ha, Fig. 5.1) is composed of meadows (29%), oak and beech woods (27%), other types of forest (25%), and crops (12%; maize, other cereals). Supplemental feeding with maize is provided outside the reserve in the forest plots (around 200-300 kg year-1 km-2 forest surface from March to August). Evidence of high acorn production during the autumn and winter were reported in the region (60 km from the reserve) for 108 2004 and 2006 (836 and 386 kg.ha-1 forest surface respectively, compared to 2-64 kg.ha-1 forest area for other years, National Forestry Office 2008). Figure 5.1: The nature reserve in the Geneva basin and its surroundings. Numbers bordering the map show the latitude (x) and longitude (y) in decimal degrees. The value -450m shows the point with the minimum Relative Position from the Reserve. The main cause of wild boar mortality is anthropogenic (hunting or culling: 81.6%, road accidents: 15.3%, other: 3.1%; n=196 dead animals recovered from 2002 to 2007). The hunting season is open in the French part of the Geneva basin between September 8th and 14th, and closed at the beginning of the following year between January 25th and 31st. Hunting takes place from 8:00 a.m. to 5:00 p.m. or 7:00 p.m., with no hunting at night. Hunting territories range from 3 to 20 km². During the open hunting 109 season, hunters perform up to three drives per week (with dogs, beater and shooters) in the forest of each hunting territory, except in areas with special protection status, such as the focal nature reserve (Fig. 5.1). Road and rail traffic is mostly diurnal, and is especially dense around 7:00-10:00 a.m. and 5:00-8:00 p.m. due to cross-border workers commuting to and from Geneva. Trapping and tracking Wild boar were trapped, weighed and ear-tagged from 2002 to 2007 in the Geneva basin. Animals weighing more than 30 kg were fitted with VHF radio-transmitter collars (females: 8 juveniles, 15 yearlings, 6 adults; males: 3 juveniles, 5 yearlings, 3 adults). They were tracked by triangulation from tagging to the animal’s death or collar loss (survey duration = 222 +/- 169 days, mean+/-SD). The location of the animals was determined once a day between 8:00 am and 6:00 pm, while they were resting (Boitani et al. 1994, Keuling et al. 2008b). Animals were also tracked twice a month at night (with an interval of 30 minutes between successive locations) from 10:00 pm to 4:00 am during their main activity period (e.g. Keuling et al. 2008b). These time slots defined the “day” and “night” periods. On the seasonal scale, we also defined two seasons of risk: the “closed” hunting season (7 months) and the “open” hunting season (5 months), based on the official dates of the hunting season (see above). In the analyses that follow, we use the “individual-year” (i.e. a given individual observed during a given year) as the statistical unit, because 21 of the 29 females and 10 of the 11 males were only tracked for one year. Threshold effect In the following analyses, diurnal and nocturnal locations are considered separately. We aimed to estimate the distance from the reserve below which home range shifts of animals living in peripheral areas were influenced by the existence of the 110 reserve. Seasonal activity centers were defined as the medians of the x- and y-coordinates of the individual’s locations for each season (closed/open). Relative Positions from the Reserve (RPR) were computed as the distances in meters between the seasonal activity centers and the nearest boundary of the reserve. If a center was located within the reserve, then the RPR was negative (Fig. 51). The reserve’s attractiveness was investigated using a linear model using the RPR during the closed hunting season (RPRclosed) as the explanatory variable, and the RPR during open hunting season (RPRopen) as the response variable. The intercept described the tendency of movement relative to the reserve amongst individuals, whereas the slope measured the influence of the initial position of the individual on these movements. If no movement occurred relative to the reserve, the estimated slope is equal to 1, and intercept 0 (i.e. RPRopen = RPRclosed). Other combinations of values (for slope and intercept) could correspond to attraction effects (i.e. RPRopen<RPRclosed) or repulsion effects (i.e. RPRopen>RPRclosed). We expected the reserve to attract animals located nearby, but not those located further away from it (i.e. to identify a buffer zone around the reserve extending up to a specific threshold distance). The relationship between RPRclosed and RPRopen would then change beyond this threshold parameter. Technically, this implies that separate linear regression models of RPRopen on RPRclosed were fitted for animals below and above the threshold in RPRclosed (see Fig. 5.2 for an illustration). We estimated five parameters simultaneously (the threshold value, and two intercepts and two slopes of the linear models estimated below and above the threshold respectively) using weighted least squares. To find out whether the regression relationships below and above the threshold were significantly different, we used a supLM test (Andrews 1993, Zeileis 2005), which takes the supremum of LM statistics over all conceivable thresholds. Threshold estimation and testing were performed using R software (R Development Core Team 2008) using the "strucchange" package (Zeileis et al. 2003). As individuals were observed at different numbers of locations during and outside the hunting seasons, statistical procedures were adapted to allow for this when performing weighted estimations and inferences (weights = geometric means of the number of locations in the two seasons). 111 Space use We first determined annual home range size (with all locations) by measuring the 95% Minimum Convex Polygon area of each individual. Then we computed for each individual the distance of all locations with their barycenter and considered the 95% quantile of these distances as measure of the home range radius. Finally we computed daily travel distances by cumulating distance between successive relocations during a movement phase (at night in our case). Then for all individuals, the data set was split into four subsets: diurnal home range - closed season, nocturnal home range - closed season, diurnal home range hunting season, nocturnal home range - hunting season. Only individuals that had been located at least 20 times during periods of more than 20 days, and tracked on at least 3 different nights were included (28 female-years and 5 male-years were surveyed from 2004 to 2006). For each subset, the kernel utilization distribution (UD, Worton 1989) of each animal was estimated using the R package “adehabitat” (Calenge 2006). We chosed a single smoothing parameter value (h) for all UD estimates following the recommendations of Pellerin et al. (2008), so that any change in space use parameters (see below) could be clearly interpreted as a change in animal behavior and not as a potential change in the smoothing parameters. An h value of 150 was chosen after graphical explorations. 112 Several space-use parameters were computed for all individuals. Home Ranges sizes (HR: area of the 95% contour) were calculated in hectares. Mean Inter-individual Overlaps (MIO: percentage) were estimated by averaging for each individual, and “Volumes of Intersections” (the VI in Fieberg and Kochanny 2005) with all the other individuals tracked during the same year. Forest Overlaps (FO: percentage of the UD in forest areas) were measured over seasons and day periods. Finally, the reserve-utilization (RU: percentage), which was estimated only for animals that lived close to the protected area, was the proportion of the UD included in the protected perimeter (similar to the PHR in Fieberg and Kochanny 2005). Variations of these parameters were analyzed using linear mixed-effect models with the R “nlme” package (Pinheiro and Bates 2000), setting “individuals-year” as random (N=132). Three categorical variables with two states were considered as fixed effects: season (closed or open), period (day or night) and group (reserve or other). The group variable determines whether animals lived below the RPRclosed threshold estimated previously (reserve) or lived beyond this threshold (other). The group “other” is used as a comparative control in order to verify that space use patterns are specific to the reserve surrounding or are common to the entire study site. Analysis of the RU variation was restricted to the individuals with the status ‘reserve’ for the group variable RESULTS Threshold effect Significant structural changes were identified in the linear regression of RPRopen on RPRclosed (day: n = 52, supLM = 16.55, P = 0.0081; night: n = 50, supLM = 14.33, P = 0.016). 113 Transition from first to second model occurred within an interval of RPRopen = 1207-2123 m for diurnal data, and an interval of RPRclosed = 975-1087 m for nocturnal data (these intervals correspond to the maximum and minimum observed values of RPRclosed for the first and second model respectively, see vertical dotted lines in Fig. 5.2). Models below the threshold (on the left part of the graphs, Fig. 5.2A and 5.2B) had slopes of less than 1, and intercepts of less than 0 for the daytime, and around 0 for the nighttime (estimate [CI95%], day: intercept = -154 [-273;-35]; slope = 0.31 [0.02;0.59]; night: intercept = -7 [-164;151], slope = 0.26 [-0.16;0.68]). These models showed that animals moved closer to the reserve during the open season, especially individuals living just outside the reserve (up to 1.2 km from the edge) during the closed season (intercept<0, slope<1, Fig. 5.2C and 5.2D). Models beyond the threshold (on the right part of the graphs, Fig. 5.2A and 5.2B) had slopes slightly greater than 1 for both the daytime (slope = 1.06 [1.001;1.12]), and the nighttime (slope = 1.06 [1.02;1.11]), and a slightly negative intercept for the nighttime (intercept = -423 [-791;-56]) whereas, as expected, the intercept was not different from 0 for the daytime (intercept = -318 [702;65]). These models revealed a virtual absence of any attractive effect on diurnal space use (intercept=0, slope=1, Fig 5.2A), whereas a slight attractive effect appeared for nocturnal space use for some of the animals living around 2-4 km from the edge of the reserve (intercept<0, slope>1, Fig. 5.2B). 114 Figure 5.2: The threshold effect on the attraction of wild boar (S. scrofa) activity centers by the protected areas in response to hunting estimated by the segmented linear regressions of RPRhunting (y) or RPRclosed (x). (a) full plot with the diurnal dataset; (b) full plot with the nocturnal dataset; (c) zoom on the zone of influence using the diurnal dataset; (d) zoom on the zone of influence using the nocturnal dataset. Black crosses: individual-year. Solid lines: segmented linear models. Dotted-lines: transition zones delimited by maximum and minimum observed values of RPRclosed for the first and second model respectively. 115 Space use Annual home ranges were equal to 7.6 +/-2.9 km2 for females, and 10.1 +/-2.5 km2 (mean+/-sd) for males. Home range radii ranged around 1.8 +/-0.3 km and 2.3 +/-0.4 km (mean+/-sd) for females and males respectively whereas daily travel distances were equal to 4.9 +/-0.3 km.day-1 for females and 6.0 +/0.5 km.day-1 (mean+/-sd) for males. Period had a slight additive effect on changes in home range (HR), whereas season displayed only a non-significant tendency (intercept: F = 188.04, d.f. = 1,93, P < 0.0001; season: F = 2.80, d.f. = 1,93, P = 0.098; period: F = 6.02, d.f. = 1,93, P = 0.016). The effect of group (group: F = 0.51, d.f. = 1,31, P = 0.48), and interactive effects between season, period and group were not significant (season:period: F = 0.32, d.f. = 1,93, P = 0.57; season:group: F = 0.08, d.f. = 1,93, P = 0.78; period:group: F = 0.69, d.f. = 1,93, P = 0.41, season:period;group: F = 0.12, d.f. = 1,93, P = 0.73). Home ranges were smaller during the daytime than at night, and tended to be smaller during the open season than the closed season (Fig. 5.3A). Season, period and group had strong significant additive effects on differences in mean inter-individual overlap (MIO) (intercept: F = 78.75, d.f. = 1,93, P < 0.0001; season: F = 14.70, d.f. = 1,93, P = 0.0002; period: F = 7.81, d.f. = 1,93, P = 0.0065; group: F = 24.87, d.f. = 1,31, P < 0.0001). Interactions of season and period with group were significant or virtually significant (season:group: F = 7.93, d.f. = 1,93, P = 0.0059; period:group: F = 3.53, d.f. = 1,93, P = 0.063). No other interaction was significant (season:period: F = 0.79, d.f. = 1,93, P = 0.38; season:period:group: F = 0.41, d.f. = 1,93, P = 0.52). Inter-individual overlaps of animals living close to the protected area were higher during the day than at night, and higher during the open season than during the closed season, whereas it did not vary and remained lower for animals living far from the reserve (Fig. 5.3B). 116 Figure 5.3: Variations in several space-use parameters of wild boar depending on season (Closed/Open), period (Day/Night) and group (Reserve/Other). The group determines whether animals lived under the influence of the reserve (reserve) or out of this influence (other) according to the first analysis (see methods and Fig.5.2). Parameters (mean+/-se) were estimated on 132 kernel utilization distributions (UD) estimated for 33 individual-years. (a) Home range size: area of 95% contours of UDs in hectares; (b) Inter-individual overlap: mean Volume of Intersection (%) with UD of other individuals alive during the corresponding year; (c) Forest overlap: percentage of UDs included in forest areas. 117 Period had a strong additive effect on differences in forest overlaps (FO), whereas the additive effects of season and group were not significant (intercept: F = 742.15, d.f. = 1,93, P < 0.0001; season: F = 0.064, d.f. = 1,93, P = 0.80; period: F = 20.55, d.f. = 1,93, P < 0.0001; group: F = 0.11, d.f. = 1,31, P = 0.74). The interaction between season and group was significant (season:group: F = 7.71, d.f. = 1,93, P = 0.0066), unlike other interactions (season:period: F = 1.74, d.f. = 1,93, P = 0.19; period:group: F = 1.39, d.f. = 1,93, P = 0.24; season:period:group: F = 1.50, d.f. = 1,93, P = 0.22). Forest overlaps were generally smaller at night than during the day, and displayed contrasting seasonal variations depending on group: they decreased during the open season for animals living near the reserve, whereas they increased for individuals living further from the protected area (Fig. 5.3C). Finally, after restricting the analysis to animals influenced by the protected area, we observed strongly significant additive effects of season and day period on reserve-use (RU) variation (N=72; intercept: F = 57.96, d.f. = 1,57, P < 0.0001; season: F = 12.65, d.f. = 1,57, P = 0.0008; period: F = 16.31, d.f. = 1,57, P = 0.0002). The interactive effect of season and period was not significant (season:period: F = 1.37, d.f. = 1,57, P = 0.24). Reserve-use was lower at night than in the daytime, and more pronounced during the open season than the closed season (Fig. 5.4). DISCUSSION Our results support the hypothesis that there is a threshold distance effect beyond which animals do not respond by shifting their home ranges toward the reserve (prediction 1). Regarding diurnal space use patterns (during the highest risk conditions), animals living more than 2.1 km beyond the edge of the reserve during the closed season were not attracted into the reserve during the open season (Fig. 5.2A). Interestingly, this threshold corresponds to the mean radius of the home ranges of wild boar (see results). 118 This suggests that animals with a home range that was “in contact” with the reserve during the closed season were the only ones to enter the refuge during the open season. Daily trajectories higher than home range radii have been shown to be quite common for wild boar of our study site (see results, similar values were founded in Janeau et al. 1995) and annual home range sizes remained quite small regarding to daily movement capacities (see results, see also Boitani et al. 1994, Massei et al. 1997 for males, and Keuling et al. 2008a for an exhaustive review on females). This implies that moving over distances greater than the radius of the home range to seek refuge in the protected area might not be a very limiting factor (in terms of energy and time). The lack of response may then result from limitations of the range over which the individual perceives the risk, which is probably restricted for a given individual by its home range boundaries. Animals can benefit from having a good knowledge of refuge locations (Clarke et al. 1993) as this enables them to hide quickly, and so reduce the predation risk (Cowlishaw 1997). Individuals whose home ranges are at least partially within the reserve may therefore be the only ones to perceive its safety, and respond to increases in predation risk by shifting their home ranges toward this refuge. Near the edge of the reserve, the wild boar responded to pulses of risk by shifting their home ranges into the protected area. This occurred as expected on a daily scale (prediction 2) in response to increased risk during the daytime (Fig. 5.4A and 5.4C vs. Fig. 5.4B and 5.4D). During the closed season, road and railway traffic may in fact have induced a greater diurnal risk outside the reserve (Fig. 5.1), but wild boar could also perceive any human activity in general (e.g. forestry, tourism) as a predation risk (Frid and Dill 2002). These daily shifts coincided with the greater use of forests during the daytime than at night that we observed generally throughout the study site. Wild boar commonly use shelters while resting (Boitani et al. 1994), but in our case habitat protection status appeared to be an additional factor when selecting resting sites. A similar pattern also occurred on a seasonal scale in response to the marked increase of risk during the open season, which drove animals further into the protected areas than during the closed season (prediction 3, Fig. 5.4A and 5.4B vs. Fig. 5.4C and 5.4D). 119 Figure 5.4: Seasonal and daily impact of the presence of the reserve on wild boar space use. (a) day - closed season, (b) night - closed season, (c) day - hunting season and, (d) night - hunting season. Black contours: perimeter of the nature reserves. Transparent gray polygons: core areas (kernel 50%) of 20 individual-years in 2004, 2005, and 2006. Grey scale corresponds to the number of polygons overlapping a same area. Note that some overlaps are not real and come from animals observed during different year. The same individual-years are shown on all plots. These seasonal shifts reduced forest use during the open season (Fig. 5.3C). Moreover, even though seasonal variations in nocturnal risk were expected to be negligible, we actually found that the nocturnal use of space around the reserve was affected by hunting as much as diurnal space use (prediction 4 not supported, Fig. 5.4B and 5.4D). The effect of diurnal risk constraints was extended to nocturnal space use, probably because of the interdependency between nocturnal and diurnal locations (due to the need for refuge proximity while foraging, Wilson 2004). This led to an increased nocturnal use of the protected area, and decreased use of forest areas during open 120 seasons. Forests also provide abundant food (Schley and Roper 2003, Bieber and Ruf 2005) and shelter (Boitani et al. 1994) during autumn and winter, and so were preferentially used during the open season by animals outside the sphere of influence of the reserve (Fig. 5.3C). These findings confirm the predominance of the landscape of fear (shaped by the existence of protected areas and hunting activities) to explain wild boar space use patterns. In wetlands wild boar were shown to switch to eating bulbs and rhizomes of semi-aquatic plants to satisfy their food requirements during autumn and winter (Gimenez-Anaya et al. 2008). Moreover, the increased inter-individual overlaps during the hunting season (Fig. 5.3B and 5.4B vs 5.4D) could increase food competition within the refuge. This implies that the delayed concentration effect on nocturnal space use could have potential consequences for the diet and food intake of wild boar. Daily and seasonal variations of home range sizes (smaller during the daytime and during the hunting season) were however common to the entire study site (Fig. 5.3A). The nocturnal activity of wild boar (Keuling et al. 2008b), and heterogeneities of the sampling protocol (unequal season lengths) may explain these variations, as well as differences in spatial and temporal risk. Our study shows that restricted harvesting periods, combined with the presence of protected areas in the vicinity, create a structured and dynamic landscape of fear for the harvested species. The close parallel between the use of the reserve and pulses of risk occurring at various different temporal scales confirms that spatiotemporal variations of risk are major drivers of animal distribution when predation is the main factor limiting their fitness (Rettie and Messier 2000). Animals outside the influence of the reserve seemed to respond to risk variations on a finer scale by increasing their use of forest areas within their home range (third level of habitat selection, Johnson 1980). This suggests that the response of prey, in terms of space use, to spatiotemporal changed in risk is scale dependent, and varies with the local structure of the landscape of fear. Protected areas, such as hunting or nature reserves, can then be used as refuge by game species, and become a key environmental feature driving their distribution (Madsen 1998). The nature reserve studied partially dissociates safety and food abundance on the large scale (Fig. 5.1). Individuals consequently shifted their home ranges between these two resources 121 until the distance between the safe refuge and the rich habitat become limiting. These patterns could reveal a landscape complementation process (see an example on wild pig in Choquenot and Ruscoe 2003) resulting in an increase in the population growth rate around the reserve. Common or invasive species (species other than those targeted for conservation) can then benefit from the protection status of a reserve. The wild boar is considered to be an engineer species that can strongly modify species richness by its feeding, rooting, stalling, or wallowing activities in protected ecosystems (see a short review in Wright and Jones 2004). Our findings highlight how harvesting activities can have a major impact on natural ecosystems, even those protected from human activities, via the mobility of refugee animals (Lenihan et al. 2001, Ripple and Beschta 2004). We encourage future research into multi-scale habitat selection strategies, and their consequences for the individual survival probability of game or harvested species. The high spatiotemporal variations of the human-induced predation risk can result in highly contrasting environmental conditions, and help us to understand how animals respond within their landscape of fear. ACKNOWLEDGEMENTS This study was commissioned and funded by the Office National de la Chasse et de la Faune Sauvage (France), the Fédération Départementale des Chasseurs de l’Ain et de la Haute Savoie (France), the Chambre d’Agriculture de la Haute Savoie (France), the Direction générale de la Nature et du Paysage of Canton of Geneva (Switzerland), the Federal Office for the Envionnement (Switzerland), ECOTEC Environnement SA (Switzerland) and the Service des Forêts, de la Faune et de la Nature of Canton of Vaud (Switzerland). The authors thank the Conseil General de l’Ain (France) for additional funding. Many thanks to all the field personnel involved, especially to J.O. Chappuis, M. Comte, F. Corcelle and J. Vasse and to H. Fritz, D. Christianson and J-M Gaillard for their insightful comments on an earlier version of the manuscript. 122 REFERENCES Abrahams, M. V. and Dill, L. M. 1989. A determination of the energetic equivalence of the risk of predation. — Ecology 70: 999-1007. Andrews, D. W. K. 1993. Tests for parameter instability and structural-change with unknown change point. — Econometrica 61: 821-856. Banks, S. C. et al. 2007. Sex and sociality in a disconnected world: a review of the impacts of habitat fragmentation on animal social interactions. — Canadian Jounal of Zoology 85: 1065-1079. Berryman, A. A. and Hawkins, B. A. 2006. The refuge as an integrating concept in ecology and evolution. — Oikos 115: 192-196. Bieber, C. and Ruf, T. 2005. Population dynamics in wild boar Sus scrofa: ecology, elasticity of growth rate and implications for the management of pulsed resource consumers. — Journal of Applied Ecology 42: 1203-1213. Boitani, L. et al. 1994. Spatial and activity patterns of wild boars in Tuscany, Italy. — Journal of Mammalogy 75: 600-612. Brown, J. S. 1988. Patch use as an indicator of habitat preference, predation risk, and competition. — Behavioral Ecology and Sociobiology 22: 37-47. Brown, J. S. et al. 1999. The ecology of fear: Optimal foraging, game theory, and trophic interactions. — Journal of Mammalogy 80: 385-399. Bruinderink, G. W. T. A. G. and Hazebroek, E. 1996. Ungulate traffic collisions in Europe. — Conservation Ecology 10: 1059-1067. Calenge, C. 2006. The package adehabitat for the R software: a tool for the analysis of space and habitat use by animals. — Ecological Modelling 197: 516-519. Choquenot, D. and Ruscoe, W. A. 2003. Landscape complementation and food limitation of large herbivores: habitat-related constraints on the foraging efficiency of wild pigs. — Journal of Animal Ecology 72: 14-26. Clarke, M. F. et al. 1993. Site familiarity affects escape behaviour of the eastern chipmunk, Tamias striatus. — Oikos 66: 533-537. 123 Cowlishaw, G. 1997. Refuge use and predation risk in a desert baboon population. — Animal Behaviour 54: 241-253. Creel, S. and Christianson, D. 2008. Relationships between direct predation and risk effects. — Trends in Ecology and Evolution 23: 194-201. Fieberg, J. and Kochanny, C. O. 2005. Quantifying home-range overlap: The importance of the utilization distribution. — Journal of Wildlife Management 69: 1346-1359. Fortin, D. et al. 2005. Wolves influence elk movements: Behavior shapes a trophic cascade in Yellowstone National Park. — Ecology 86: 1320-1330. Fretwell, S. D. and Lucas, H. L. J. 1970. On territorial behavior and other factors influencing habitat distributions in birds. — Acta Biotheoretica 19: 16-36. Frid, A. and Dill, L. M. 2002. Human-caused disturbance stimuli as a form of predation risk. Ecology and Society. p 11. Gimenez-Anaya, A. et al. 2008. Food habits of wild boars (Sus scrofa) in a Mediterranean coastal wetland. — Wetlands 28: 197-203. Hamilton, W. D. 1971. Geometry for the selfish herd. — Journal of Theoretical Biology 31: 295-311. Janeau, G. et al. 1995. Daily movement patterns variations in Wild Boar (Sus scrofa L.). — Ibex J.M.E. 3: 98-101. Johnson, D. H. 1980. The comparison of usage and availability measurements for evaluating resource preference. — Ecology 61: 65-71. Kennedy, M. and Gray, R. D. 1993. Can ecological theory predict the distribution of foraging animals? A critical analysis of experiments on the ideal free distribution. — Oikos 68: 158-166. Keuling, O. et al. 2008a. Annual and seasonal space use of different age classes of female wild boar Sus scrofa L. — European Journal of Wildlife Research 54: 403-412. Keuling, O. et al. 2008b. How does hunting influence activity and spatial usage in wild boar Sus scrofa L.? — European Journal of Wildlife Research 54: 729-737. Laundré, J. W. et al. 2001. Wolves, elk, and bison: reestablishing the “landscape of fear” in Yellowstone National Park, U.S.A. — Canadian Journal of Zoology 79: 14011409. 124 Lenihan, H. S. et al. 2001. Cascading of habitat degradation: Oyster reefs invaded by refugee fishes escaping stress. — Ecological Applications 11: 764-782. Lima, S. L. and Bednekoff, P. A. 1999. Temporal Variation in Danger Drives Antipredator Behavior: The Predation Risk Allocation Hypothesis. — The American Naturalist 153: 649-659. Madsen, J. 1998. Experimental refuges for migratory waterfowl in Danish wetlands. II. Tests of hunting disturbance effects. — Journal of Applied Ecology 35: 398-417. Massei, G. et al. 1997. Factors influencing home range and activity of wild boar (Sus scrofa) in a Mediterranean coastal area. — Journal of Zoology 242: 411-423. National Forestry Office 2008. REseau National de suivi à long terme des ECOsystèmes FORestiers. Pellerin, M. et al. 2008. Roe deer Capreolus capreolus home-range sizes estimated from VHF and GPS data. — Wildlife Biology 14: 101-110. Pinheiro, J. C. and Bates, D. M. 2000. Mixed-effects models in S and S-PLUS. — Springer-Verlag. R Development Core Team 2008. R: A Language and Environment for Statistical Computing. — R Foundation for Statistical Computing, Vienna, Austria, <http://www.R-project.org/>. Rettie, W. J. and Messier, F. 2000. Hierarchical habitat selection by woodland caribou: its relationship to limiting factors. — Ecography 23: 466-478. Ripple, W. J. and Beschta, R. L. 2004. Wolves and the ecology of fear: Can predation risk structure ecosystems? — Bioscience 54: 755-766. Roberts, G. 1996. Why individual vigilance declines as group size increases. — Animal Behaviour 51: 1077-1086. Schley, L. and Roper, T. J. 2003. Diet of wild boar Sus scrofa in Western Europe, with particular reference to consumption of agricultural crops. — Mammal Review 33: 43-56. Shochat, E. et al. 2002. Density-dependent habitat selection in migratory passerines during stopover: what causes the deviation from IFD? — Evolutionary Ecology 16: 469-488. 125 Sih, A. et al. 2000. New insights on how temporal variation in predation risk shapes prey behavior. — Trends in Ecology and Evolution 15: 3-4. Sodeikat, G. and Pohlmeyer, K. 2003. Escape movement of family groups of wild boar Sus scrofa influenced by drive hunts in Lower Saxony, Germany. — Wildlife Biology 9: 43-49. Toïgo, C. et al. 2008. Disentangling natural from hunting mortality in an intensively hunted wild boar population. — Journal of Wildlife Management 72: 1532–1539. van der Merwe, M. and Brown, J. S. 2008. Mapping the Landscape of Fear of the Cape Ground Squirrel (Xerus inauris). — Journal of Mammalogy 89: 1162-1169. Wilson, C. J. 2004. Rooting damage to farmland in Dorset, southern England, caused by feral Wild Boar Sus scrofa. — Mammal Review 34: 331-335. Worton, B. J. 1989. Kernel methods for estimating the utilization distribution in homerange studies. — Ecology 70: 164-168. Wright, J. P. and Jones, C. G. 2004. Predicting effects of ecosystem engineers on patchscale species richness from primary productivity. — Ecology 85: 2071-2081. Zeileis, A. 2005. A unified approach to structural change tests based on ML scores, F statistics, and OLS residuals. — Econometric Reviews 24: 445-466. Zeileis, A. et al. 2003. Testing and dating of structural changes in practice. — Computational Statistics & Data Analysis 44: 109-123. 126 127 128 6. PARTIE 3 : RELATION ENTRE SURVIE ET SELECTION D’HABITAT 129 IMPACT OF REFUGES ON PREDATOR-PREY SPATIAL INTERACTIONS AND PREY MORTALITY: WILDLIFE HARVESTING AS A CASE STUDY. Vincent Tolon1, 2, Jodie Martin2, Stéphane Dray2, Anne Loison1, Claude Fischer3 &. Eric Baubet4. 1 Laboratoire d’Ecologie Alpine CNRS UMR5553, Université de Savoie, Bâtiment Belledonne, F-73376 Le Bourget-du-Lac, France. 2 Laboratoire de Biométrie et Biologie Evolutive CNRS UMR5558, Université de Lyon, Bâtiment Gregor Mendel, F-69622 Villeurbanne, France. 3 4 Ecole d’Ingénieurs de Lullier, S-1254 Jussy, Switzerland. Office National de la Chasse et de la Faune Sauvage, Centre National d’Etude et de Recherche Appliquée - Cervidés et sanglier, F-55 000 Bar-le-Duc, France. Submitted in Oikos 130 ABSTRACT To avoid mortality, prey generally tend to dissociate their long-term habitat use from predators, but predators tend to match prey habitat use at the same time to maximise their gain. In this spatial interaction, environmental constraints on predator movement offer the opportunities for prey to outdistance the risk source in a durable way (i.e. refuges) and increase their survival. If this spatial mismatch between prey and predator reduce the vulnerability of all prey, its effect on mortality should be almost apparent for classes of individuals usually preferred by the predator, an vanished progressively as the preference decrease. We tested these hypotheses on a special case considering wild boar as prey and hunters with their dogs as predators, at different distance from a protected area. In this system, predators (i.e. hunters) show a large preference for large individuals. Predator habitat selection was estimated using trajectories of hunting dogs equipped with GPS-transmitters, while wild boar habitat selection was assessed from radio-tracked individuals. We observed, as expected, a transition from matching to mismatching patterns between wild boar and dogs long-term habitat selections as the proximity of the reserve increases. The survival of adults (larger individuals) increased as expected with this mismatch, but survival of younger age classes followed an unexpected decrease at the same time. The spatial avoidance of individual classes usually preferred by the predator therefore lead to an increase mortality of other categories, presumably through changes in predator selectivity. Our results confirmed that refuges strongly modified the predator-prey spatial interaction, leading potentially to deep changes in their numerical dynamic. Key words: wild boar, Sus scrofa, hunting, spatial interaction, refuge, survival. 131 INTRODUCTION Whereas numerical aspects of predator-prey interactions have been studied for several decades (Lotka 1925, Volterra 1926, Leslie 1945), spatial interactions have been investigated only recently (e.g. Hugie and Dill 1994, Mitchell and Lima 2002, Sih 2005). Indeed, most studies investigated long-term spatial distribution of preys or predators independently, and considered either preys and unresponsive predators (e.g. Abrahams and Dill 1989) or predators and unresponsive preys (e.g. Sutherland 1996). In these conditions, preys were expected to avoid risk until reaching an optimal energy-risk tradeoff (Brown 1988, Abrahams and Dill 1989, Houston et al. 1993) while predators were supposed to maximise their energy gain by searching for high prey density (Fauchald et al. 2000). However, the spatial distribution of both preys and predators is highly interactive (Lima 2002), since they both selected their habitat according to the distribution of each other and have opposed interests. These behavioural interactions should lead to a two-way spatial game where preys tend to dissociate their habitat use from predators, whereas predators tend to match the habitat use of their preys (Fauchald et al. 2000, Mitchell and Lima 2002, Sih 2005, Fauchald 2009). The long-term spatial relationships between prey and predator might therefore be more elusive than expected and lead to positive (match), negative (mismatch), or even unstable associations (Fauchald 2009). Considering this spatial relationship might be essential to understand of the numerical response of the system. Indeed the matching intensity between both distributions should be positively related to the encounter probability between prey and predator, and influence ultimately prey mortality. To predict the outcome of this spatial interaction a particular attention must be given on the respective spatial constraints of predators and preys movements (Sih 2005, Fauchald 2009). For instance, long-term mismatches could occur when prey can use specific habitats where predators have difficulty to move around. Such habitats are refuges for preys that avoid mortality due to predators (e.g. trees of cliffs for baboons, Cowlishaw 1997, solid ground for sea otters, Estes et al. 1998, steep slopes or high elevation for ungulates, Caro 2005, rocks for lizards, Martin et al. 2009a). Without any spatial 132 constraints on predator movement prey may not have opportunities to outdistance durably predators, leading to a matching pattern between both space use (Fauchald 2009) and a decrease of prey survival. The availability, the size and the spatial structure of refuges should therefore determine the long-term matching intensity between prey and predator habitat use, and ultimately, the numerical response of the system via prey mortality. This behavioural-mediated interaction could not be fully understood without considering the potential selectivity of predators for particular preys. Indeed, individuals of a given species often suffer from different predation rates (e.g. effect of sex, Fitzgibbon 1990, age and fat reserve, Huggard 1993, age and life cycle, Molinari-Jobin et al. 2004, Owen-Smith 2008) because predators generally focused on the more profitable prey categories . Predators can also switch between prey categories in response to changes in their abundance, accessibility and/or vulnerability (Oaten and Murdoch 1975, e.g. Owen-Smith 2008). Hence, if the long-term predator-prey mismatch decreases uniformly the predator accessibility for all prey categories, the resulting increase of survival might be mostly noticeable for the most preferred prey, and progressively fadeout as the preference decrease amongst categories (i.e. benefits of refuges are almost apparent for highly predated animals). Whereas natural predator-prey systems are the main examples of these behaviourmediated interactions (Hugie and Dill 1994, Lima 2002, Mitchell and Lima 2002, Sih 2005), human-wildlife systems could provide a helpful case study to better understand the mechanisms and the consequences of these interactions. Indeed, while hunters can search actively for their target species in their preferred habitats (Lande et al. 2009), game species tend to avoid hunters by moving in available refuges (Fox and Madsen 1997, Madsen 1998, Tolon et al. 2009). Both players are consequently engaged in a spatial game similar to a natural predator-prey spatial interaction. In such systems, protected areas play the role of refuges and should lead to a strong mismatch between prey (game species) and predators (hunters) long-term space use, which in turn may lead to an increased survival of prey inside the protected area (e.g. Revilla et al. 2001, Duriez et al. 2005). 133 We investigated how the artificial constraints created by a protected area, determines the long-term level of mismatch between a prey and a predator habitat selection, and influences the mortality of the preys, in a system where (1) only a part of the prey population has access to refuges; (2) predators have a strong preference for one category of prey (large). We used hunters and wild boars (Sus scrofa) as case studies, in a study area encompassing a small protected area. We used locations of GPS-collared hunting dog (Canis lupus familiaris) as a measure of the predator habitat selection (e.g. Broseth and Pedersen 2000, Stedman et al. 2004, Lande et al. 2009). At the same time, we studied prey habitat selection from radio-marked individuals. A former study of wild boar in this area (Tolon et al. 2009) has shown that only animals located near the reserve (threshold distance is around 2km) were attracted by the protected area during the hunting season. Therefore, we should observe an increasing mismatch between habitat selection of wild boar and hunting dogs closer to the reserve than far away (prediction 1). Then, if the use of the reserve is efficient as an anti-predator strategy, we expect that the level of mismatch to be negatively correlated with mortality of the preferred prey category (larger animals, prediction 2). Finally, we predict that the level of mismatch between habitat selection of predators and prey will have lower consequences on mortality of less preferred prey (smaller animals, prediction 3). MATERIALS AND METHODS Study area The Geneva basin is located on the France-Switzerland border and is surrounded by mountain chains and the lake of Geneva. The study site is a 16×11 km area around a nature reserve of 318 ha (see Tolon et al. 2009). Natural large predators of wild boar are absent in the Geneva basin and the main source of mortality is mostly due to human encounters (hunting or culling: 81.6%, collision: 15.3%, other: 3.1%; n=196 dead 134 recoveries). Hunting occurs from the second sunday of september to the last sunday of january, between 8:00 a.m. and 5:00 or 7:00 p.m. The study site supported 7 hunting teams, each using a delimited hunting territory available on the study site. During the hunting season, hunters perform between one and six drive hunts per week within forests of hunting territories, except in the nature reserve. Road and railway traffic are mostly diurnal and are especially intense around 7:00-10:00 a.m. and 5:00-8:00 p.m. due to the proximity of Geneva (17 km). Wild boar tracking Wild boar were trapped, weighted and ear-tagged from 2003 to 2007. Forty one individuals heavier than 30 kg were fitted with Very High Frequency radio-transmitters (♀: N=29, ♂: N=12). Six additional animals were fitted with GPS devices (LOTEK®, ♀: N=4, ♂: N=2). Wild boar were mostly nocturnal in hunted areas and remain on their resting site at day during the disturbance phase (Boitani et al. 1994, Keuling et al. 2008b). Animals were therefore located once a day between 7:00 am to 7:00 pm, during their resting period and tracked between one and six times a month at night (interval of 30-60 minutes between successive locations, 15 minutes with GPS devices) from 7:00 pm to 7:00 am. Only 11 individuals were observed two years and a single female was surveyed three years. We then obtained 60 “individual-years” (♀: 8 juveniles, 11 yearlings, 25 adults; ♂:5 juveniles, 3 yearlings, 8 adults; survey duration during hunting = 84 +/- 47 days, mean+/-SD) considered as the statistical unit in the following analyses. We considered the age class combined with the sex to be a general proxy of body weight categories (juveniles: 10-30 kg, yearlings: ♀ 20-60 kg, ♂ 30-70 kg, adults: ♀ 50-90 kg, ♂ 60-110 kg, e.g. Moretti 1995, Markina et al. 2004). Juveniles and yearlings were regrouped in a unique age class (young) in order to equilibrate the number of individuals in the different classes. 135 Hunting dogs tracking Dogs were fitted with GEOPOINTER® GPS collars and located every minute during hunting actions performed within seasons 2006 and 2007, between 07:00 am and 06:00 pm. After 105 observed hunting actions of 133 +/-79 minutes, dogs were shown to move on 5.85 +/- 3.31 km in ranging areas of 70 +/- 66 ha (MCP 100%). Approach of wild boar was done with dogs on leashes. They were just released close to boars resting place and move free after being released. Space use of dogs therefore revealed a mix between hunters and dogs behaviors. Because the central protected area was expected to drive the potential mismatch between wild boar and hunting dogs, and because the reserve perimeter was stable for all year of the study, we considered the two years of dogtracking (2006-2007) as representative of the large scale spatial variations of hunting intensity for all other years (from 2003 to 2005). In order to describe the large scale space use of hunting, we estimate the relative probability density of hunting dogs with a 2D kernel smoothing. The probability density function was estimated separately for each hunting territory by a 2D kernel smoothing and then combined on the entire study site (see Hebblewhite and Merrill 2007). Habitat measures We extracted several habitat variables on a 50×50 m resolution grid using a Geographical Information System. We selected a set of variables potentially linked with mortality or risk levels for wild boar, ungulates or game species (Table 6.1). We focused in the following analysis on habitat selection performed within home ranges of wild boar (third scale of Johnson 1980) during the hunting season, because 90% of mortality (n=196 dead recoveries) occurred during this period. We computed for each individual-year a 100% Minimum Convex Polygon with all individual locations performed during the year (with both day and night locations) to estimate the maximal annual home range. Within these home ranges, we estimated the habitat utilization within the wild boar home range for 136 both wild boar (using diurnal location) and local hunting dogs (hunting dogs GPS locations). Table 6.1: Description of habitat variables potentially linked with risk levels for hunted wild boar. Names Description Link with risk References d-Road Distance to paved road Generate collisions. Bruinderink & Hazebroek (1996): unglulates. Distance to trails d-Trail Enhance hunters’ accessibility. Slope derived from a Digital Slope Elevation Model° Altitude Landscape d-Reserve Reduces (hunters) Caro (2005): short review in accessibility Elevation Model° success. forests, hunted hare. predators Altitude derived from a Digital Hygrophilous Oak- Broseth & Pedersen (2000): and/or Determines ungulates with natural predators. Boitani et al.(1994): wild boar. beech woods, urban areas, open landscape openness field. and resource (mast) Log (Distance to the reserve Provides boundary)□ areas. low risk Tolon et al. (2009): wild boar. Madsen (1998): waterfowl. *source IGN °source © 2004. CGIAR - Consortium for Spatial Information (CGIAR-CSI). Habitat selection analysis We used K-select analysis (Calenge et al. 2005, see Hansen et al. 2009 for a biological application) to analyze habitat selection by wild boars. The K-select analysis is a multivariate method based on the concept of marginality (i.e., the difference between the vector of average available habitat conditions and the vector of average used conditions). For each individual-year, the “availability” corresponds to average habitat conditions for the total home range and the “use” is estimated as the habitat measured for the observed locations (i.e. we studied habitat selection within the home range) . The K- 137 select analysis is a particular case of principal component analysis that returns a linear combination of habitat variables for which the average marginality is maximized. Hence, it is a synthesis of variables which contributes the most to the habitat selection. As we only used wild boar data, the results of the K-select analysis focus only on the habitat selection by the target species (i.e. the prey). Data concerning predators were considered a posteriori by projecting the marginality vectors for hunting dogs on the outputs of the previous K-select analysis. We therefore obtained a summary of the habitat selection by the preys (wild boar within their home ranges) using K-select analysis,and a visualisation of the predator response (hunting dogs habitat selection within each wild boar home range) by the projection of marginality vectors. Mismatch between wild boar and hunting dogs For each individual home range, we used the outputs of the K-select analysis to measure the mismatch between local hunting dogs and wild boar habitat selections. This mismatch was computed as the Euclidian distance between both marginality vectors of wild boar and local hunting dogs measured on the first retained axes of the K-select analysis (i.e., the distance between black and grey arrows extremities of a given individual-year on the Fig. 6.1.c). Null distance described perfect matching patterns of habitat selections (both marginality vectors are identical) whereas increased distance progressively described a mismatch between habitat selections of wild boar and hunting dog. Survival analysis We used a Cox Proportional Hazard model to model the survival of animal along the hunting season. We assigned 1 to individual dying before the end of the hunting season, 0 to survivors and we measured the time elapsed from the beginning of the hunting season and the final event (death, end of hunting season, collar loss). We modelled the survival probability depending on the sex, the age class and the mismatch 138 intensity between hunting dogs and wild boar habitat selection within home ranges (names mis in following analyses). We performed model selection procedure (Burnham and Anderson 2002) based on the Akaike Information Criterion corrected for small sample size (AICc), the difference of AICc with the best models (∆i) and the Akaike weight (wi). We included all additive terms (age, sex, mis) and all interactions in the full model. Indeed, the selectivity of hunters for large animals could lead old males to die preferentially (age×sex) and could report the effect of the mismatch intensity only on old animals (age×mis), on males (age×mis) or on old male (age×sex×mis). We then built 17 models based on age, sex, age+sex and age*sex by adding the term mis and its interactions. Following Burnham and Anderson (2002), we considered that models with a difference of AICc (∆i) lower than 2 with the best model are all plausible and could be interpreted biologically. In order to asses the biological interpretation of the best model we returned its coefficients with their standard error and tested if they significantly differed from 0. All analyses were performed using the R software (R Development Core Team 2008) with the ‘adehabitat’ package (Calenge 2006) for analyses of habitat selections and ‘survival’ for the survival analysis. RESULTS Habitat selection The K-select analysis showed a strong pattern of habitat selection within home ranges. The first and the second axes explained respectively 70.12 % and 11.51 % of the marginality (Fig.6.1a). The first axis opposed hygrophilous forests in the reserve core to the more mountainous periphery characterized by a diversified landscape covers (right and left respectively, Fig 6.1a). The second axis showed that the residual habitat selection opposed mountainous oak and beech woods, poorly accessible to humans, against flat open fields covered by humans (top and bottom respectively, Fig. 6.1a). 139 The plot of marginality vectors (Fig. 6.1b) showed that wild boar expressed differential habitat selection according to home-ranges composition. When animals had home ranges settled near the reserve they tended to strongly select hygrophilous forests within the protected perimeter (arrows on the right pointing on the right, Fig. 6.1b). Others animals that settled far from the reserve, in mountainous environment, tended to intensify the selection for oak and beech woods in steep slopes, far from trails and road (arrows on the left pointing upward, Fig. 6.1b). Habitat selection of hunting dogs within wild boar home-ranges was also dependent of their composition. Dogs chasing wild boar settled near the reserve clearly counter-selected the protected area, in opposition to the local behaviour of wild boar (black arrow on the right, pointing on the left, Fig. 6.1c). However, dogs chasing wild boar in mountainous environment selected oak and beech woods in steep slopes, far from trails and road, in accordance with the local wild boar habitat selection (black arrow on the left, pointing upward, Fig 6.1c). The relative dog-probability density is related to the distance from the reserve and is maximal in the close periphery of the protected area (Fig. 6.2). The density increased from the reserve core (-447m) until its close periphery (around 250-750m), then it decreased and become relatively stable beyond 1500m. 140 a) d = 0.5 oak-beech-forest slope d-trail hygrophilous-forest altitude d-reserve d-road open field urban area b) c) d=1 d=1 Figure 6.1: K-select analysis of habitat selection within home ranges of 60 wild boar. a) Arrow-plot showing the contribution of the six habitat variables on the first two axes of the K-select analysis (see Table 6.1 for details on habitat variables). The length of an arrow described the general strength of the selection amongst individuals on the corresponding habitat variable. Embedded: histogram of eigenvalues showing the proportion of marginality (selection) explained by each axis. b) Plot of marginality vectors for the 60 individuals (wild boar) on the first factorial plan. Each individual corresponds to an arrow that links the average available habitat conditions (black points: home range) to the average used conditions (arrow: relocations in the home range). c) Plot of marginality vectors for hunting-dogs projected on the K-select analysis. Each black arrow links the average available habitat conditions (black points: wild boar home range) to the average used conditions (arrows: locations of hunting dogs in the wild boar home range). Grey arrows correspond to wild boar habitat selection (same vectors than in the previous plot). 141 8e-04 6e-04 4e-04 2e-04 0e+00 Relative probability density of hunting dogs dog-density -500 0 500 1250 2000 2750 3500 4250 5000 5750 6500 7250 8000 Distance from the reserve boundary Distance from the reserve boundary Figure 6.2: Boxplot of relative probability density of hunting dogs according to the distance from the reserve boundary (distance classes = 250m). Survival analysis The mean survival probability to hunting-season for all animals was equal to 0.667 [0.532;0.780] (N=60; mean [CI95%]). Survival only slightly varied amongst age classes (juveniles: S = 0.615 [0.323;0.849], N = 13; yearling: S = 0.786 [0.488;0.943], N=14; adults: S = 0.636 [0.451;0.790], N=33; mean [CI95%]) and between sexes (females S = 0.682 [0.523;0.809], N=44; males S = 0.625 [0.359;0.837], N=16; mean [CI95%]). The best Cox Proportional-Hazard regression selected by AICc included the effect of the age (age) and the mismatch intensity (mis) with their interaction (age×mis, Table 6.2, Likelihood ratio test = 8.42, df = 3, p = 0.0382). The following model (∆i <2) contained in addition the term sex (Table 6.2) whereas others models had a ∆i higher than 2. The best model showed that the effect of the matching intensity on survival was dependent on the age class of individual (Table 6.3). Adults wild boar (N=33) clearly 142 benefited, in term of risk avoidance, from the mismatch of their habitat selection regarding to local hunting dogs habitat selection (Fig. 6.3a & 6.4). Younger animals (yearling and juveniles, N=27) however died more when hunting dogs mismatched their habitat selection than in a matching situation (Fig. 6.3b & 6.4). Table 6.2: Model selection of Cox Proportional-Hazard regression predicting the risk of death of wild boar during hunting season (N=60) with age class (age: adult, young), sex (sex: female, male) and the mismatch intensity (mis: numeric). The mismatch intensity correspond to the Euclidian distance between both marginality vectors of wild boar and hunting dogs in a given wild boar home ranges (see the text for more details). The model selection was performed according to the Corrected Akaike Information Criteria (AICc). k = number of parameter, log-lik = Log-likelihood, ∆i = difference of AICc with the best model, wi = AIC weights relative to all models. Models K log-lik AICc ∆AICc wi age*mis 3 -74.100 154.927 0.000 0.417 age*mis+sex 4 -73.797 156.704 1.777 0.171 age 1 -78.129 158.468 3.540 0.071 sex 1 -78.224 158.658 3.731 0.064 age*mis+sex+ sex:mis 5 -73.537 158.659 3.731 0.064 age*sex+mis+age:mis 5 -73.678 158.940 4.013 0.056 age+mis 2 -77.967 160.362 5.435 0.028 age+sex 2 -78.004 160.436 5.509 0.027 age*sex+mis+sex+sex:mis+age:mis 6 -73.214 160.583 5.655 0.025 sex+mis 2 -78.118 160.665 5.738 0.024 age*sex 3 -77.459 161.645 6.718 0.014 sex*mis 3 -77.817 162.362 7.435 0.010 age+sex+mis 3 -77.876 162.480 7.552 0.010 age*sex*mis 7 -72.869 162.562 7.635 0.009 age*sex+mis 4 -77.318 163.748 8.821 0.005 143 Table 6.3: Coefficient estimates for the best Cox Proportional-Hazard regression predicting the risk of death of wild boar during hunting season (N=60) with age class (age: adult, young) and the mismatch intensity (mis). Coefficient Estimate Std Error Z P(>Z) age 3.009 1.219 2.47 0.0140 mis 0.987 0.536 1.84 0.0660 age: mis -1.746 0.653 -2.67 0.0075 DISCUSSION Our results highlighted a strong spatial interaction between predators (hunting dogs) and prey (wild boar) with important impact on prey survival. The K-select analysis identified two main strategies of habitat selection of preys: selection the protected perimeter when it was available, or forested steep slopes far from human access when it was not (Fig. 6.1a & 6.1b). When movements of dogs were not spatially constrained (far from the reserve), their habitat selection matched with local prey habitat selection, leading both preys and predators to select forested steep slopes (Fig. 6.1). The experiences of hunters (Lande et al. 2009) combined with hunting dog perception capacities (Caley and Ottley 1995, Fernandez-Llario et al. 2003) may easily explain such patterns. The use of steep slopes as refuges were often reported in ungulates (short review in Caro 2005), but such response was not here sufficient for prey to outdistance the risk source. In this case, the prey may have no other choice to increase vigilance levels (Quenette and Gerard 1992) or movement between consecutive resting sites (Maillard and Fournier 1995, Sodeikat and Pohlmeyer 2003) to decrease the risk of being killed. On the other hand, the spatial constraint created by the protected area on the predator space use, allowed prey to spatially avoid the risk source by shifting their activity centres into the refuge (Fig. 6.1, Tolon et al. 2009). In response, hunting dogs and hunters intensified their effort near the reserve boundary in order to benefit from an eventual outcome of animals (Fig. 6.2, analogous to the "fishing-the-line" strategy around marine reserves, 144 Gell and Roberts 2003, Roberts et al. 2005, Kellner et al. 2007). Seasonal habitat selection patterns of both preys and predators showed therefore a clear transition from matching to mismatching patterns as the proximity of the reserve increased (Fig. 6.1c, P1 supported). These results supported the hypothesis that spatial constraints on predator movements determine the long-term matching intensity between habitat use of prey and predators. As expected (P2), the observed mismatch between both habitat selections patterns induced a decrease of mortality of adult wild boar, the prey usually preferred by the predator (Fig. 6.3a & 6.4). However, as the mismatch intensity increased, we observed against expectation that the mortality risk was reported on younger age class (P3 rejected, Table 6.2 & 6.3, Fig. 6.3b & 6.4). b) Others (yearlings + juveniles) 0.6 0.4 Cumulative survival Mismatch 0.0 0.2 0.8 0.6 0.2 0.4 Match 0.8 Match Mismatch 0.0 Cumulative survival 1.0 1.0 a) Adults 0 50 100 150 0 Time (day) b) Others (yearlings + juveniles) 50 100 150 Time (day) Figure 6.3: Cumulative survival of 60 wild boar (individuals-year) during a hunting season, according the age class and the match or mismatch of their habitat use with local hunting dogs. a) Adults, b) Yearling + Juveniles. Match or Mismatch status of individuals showed if their mismatch intensity previously computed (mis, see text for details) was below or above the general median. Therefore, hunters seemed to report their pressure on smaller prey to compensate the behavioural avoidance of larger prey. Juveniles wild boar were reported to progressively leave their mother’s home range at the age of 3-4 months (Cousse et al. 1994) and then disperse at the yearling stage (Truvé and Lemel 2003). By leaving their 145 mother settled in the protected area, inexperienced juveniles and yearlings may sometime relocate at boundaries of the protected area, and therefore die, even if their long term space use mismatched local hunting dogs. This effect my be reinforced by hunters harassment around the reserve combined with the small size of the protected area (318 ha) regarding to wild boar average home range size (♀: 760 +/-290 ha, ♂: 1010 +/-250 5 4 rs he Ot 2 3 ad ul ts 0 1 Risk Change Ratio 6 7 ha, mean+/-sd, Tolon et al. 2009). 0.5 1.0 1.5 2.0 2.5 3.0 3.5 Mismatch Intensity Figure 6.4: Relative change in risk of dying during a hunting season for 60 wild boar (individuals-year) according to the mismatch intensity with local hunting dogs and the age class. Solid lines: predicts (bold) +/- se for adults. Dashed lines: predicts (bold) +/- se for others age classes (yearling and juveniles). The risk change ratio corresponds to the multiplicative factor of the instantaneous risk of dying corresponding to a given value of mismatch (the risk increase if the ratio is positive and decrease if negative). In our system, the effect of the long-term spatial mismatch intensity on prey mortality was therefore highly dependent on the predator selectivity. From an ecological perspective, the predator preference for particular prey categories ensures an indirect protection for others animals, because preferred prey pulled on themselves most of the predation risk. However, predators may switch from one prey category to another one according to their relative abundance, accessibility and/or vulnerability (Oaten and 146 Murdoch 1975, e.g. Owen-Smith 2008). For instance, the behavioural response of preferred prey to actual predation risk (e.g. selection of refuges) can induce switches in predator selectivity. In our system, large preys were usually subject to higher predation rates when predators were free to move in prey preferred habitat. However, the presence of a refuge and the resulting mismatch of predator and prey habitat selection leads to a switch of selectivity and thus, to a higher mortality of smaller prey. Our results highlight how the movement constraints of the predator relative to its prey play a determinant role in the long-term outcome of their spatial interactions. The participant with lowest spatial constraints (e.g. hunting dogs far from the reserve or wild boar near the reserve) can take the advantage of the other (matching for predators or mismatching for prey) and finally succeed in the interaction (feeding for predator or surviving for prey, Sih 2005, Fauchald 2009). Prey refuge were shown to modify stability and equilibrium of predator-prey systems (Berryman and Hawkins 2006) by providing survival opportunities for prey. However, if predators respond to the mismatch of their preferred prey (e.g. large individuals) by switching on secondary prey (e.g. small individuals), the response of the prey population should depend on the relative impact of both prey categories on the population dynamic. In our system the population elasticity to the adult age class were quite similar to juveniles and yearling elasticity (Bieber and Ruf 2005). Therefore the impact of the protected areas on local population dynamic remains unclear because yearling and juvenile mortality replaced adult mortality near the reserve. Our approach however confirmed that simple movements’ constraints of predator relative to its prey may have deep impact on their spatial interactions, finally leading to strong changes in the numerical dynamic of the system. ACKNOWLEDGEMENTS This study was commissioned and funded by the Office National de la Chasse et de la Faune Sauvage (France), the Fédération Départementale des Chasseurs de l’Ain et de la Haute Savoie (France), the Chambre d’Agriculture de la Haute Savoie (France), the 147 Direction générale de la Nature et du Paysage of Canton of Geneva (Switzerland), the Federal Office for the Envionnement (Switzerland), ECOTEC Environnement SA (Switzerland) and the Service des Forêts, de la Faune et de la Nature of Canton of Vaud (Switzerland). The authors thank the Conseil General de l’Ain (France) for additional funding. Many thanks to all the field personnel involved, especially to J.O. Chappuis, M. Comte, F. Corcelle and J. Vasse and to M. Valeix its insightful comments on an earlier version of the manuscript. REFERENCES Abrahams, M. V. and Dill, L. M. 1989. A determination of the energetic equivalence of the risk of predation. — Ecology 70: 999-1007. Berryman, A. A. and Hawkins, B. A. 2006. The refuge as an integrating concept in ecology and evolution. — Oikos 115: 192-196. Bieber, C. and Ruf, T. 2005. Population dynamics in wild boar Sus scrofa: ecology, elasticity of growth rate and implications for the management of pulsed resource consumers. — Journal of Applied Ecology 42: 1203-1213. Boitani, L. et al. 1994. Spatial and activity patterns of wild boars in Tuscany, Italy. — Journal of Mammalogy 75: 600-612. Broseth, H. and Pedersen, H. C. 2000. Hunting effort and game vulnerability studies on a small scale: a new technique combining radio-telemetry, GPS and GIS. — Journal of Applied Ecology 37: 182-190. Brown, J. S. 1988. Patch use as an indicator of habitat preference, predation risk, and competition. — Behavioral Ecology and Sociobiology 22: 37-47. Bruinderink, G. W. T. A. G. and Hazebroek, E. 1996. Ungulate traffic collisions in Europe. — Conservation Ecology 10: 1059-1067. Burnham, K. P. and Anderson, D. R. 2002. Model selection and multimodel inference. A practical information-theoretic approach. — Springer-Verlag New York, Inc. 148 Calenge, C. 2006. The package adehabitat for the R software: a tool for the analysis of space and habitat use by animals. — Ecological Modelling 197: 516-519. Calenge, C. et al. 2005. K-select analysis: a new method to analyse habitat selection in radio-tracking studies. — Ecological Modelling 186: 143-153. Caley, P. and Ottley, B. 1995. The Effectiveness of Hunting Dogs for Removing Feral Pigs (Sus scrofa). — Wildlife Research 22: 147-154. Caro, T. M. 2005. Antipredator defenses in birds and mammals. — University of Chicago Press. Cousse, S. et al. 1994. Use of space by juveniles in relation to their postnatal range, mother, and siblings: an example in the wild boar Sus scrofa L. — Canadian Jounal of Zoology 72: 1691–1694. Cowlishaw, G. 1997. Refuge use and predation risk in a desert baboon population. — Animal Behaviour 54: 241-253. Duriez, O. et al. 2005. Factors affecting population dynamics of Eurasian woodcocks wintering in France: assessing the efficiency of a hunting-free reserve. — Biological Conservation 122: 89-97. Estes, J. A. et al. 1998. Killer whale predation on sea otters linking oceanic and nearshore ecosystems. — Science 282: 473-476. Fauchald, P. 2009. Spatial interaction between seabirds and prey: review and synthesis. — Marine Ecology - Progress Series 391: 139-151. Fauchald, P. et al. 2000. Scale-dependent predator-prey interactions: The hierarchical spatial distribution of seabirds and prey. — Ecology 81: 773-783. Fernandez-Llario, P. et al. 2003. Habitat effects and shooting techniques on two wild boar (Sus scrofa) populations in Spain and Portugal. — Zeitschrift für Jagdwissenschaft 49: 120-129 Fitzgibbon, C. D. 1990. Why do hunting cheetahs prefer male gazelles? . — Animal Behaviour 40: 837-845. Fox, A. D. and Madsen, J. 1997. Behavioural and distributional effects of hunting disturbance on waterbirds in Europe: Implications for refuge design. — Journal of Applied Ecology 34: 1-13. 149 Gell, F. R. and Roberts, C. M. 2003. Benefits beyond boundaries: the fishery effects of marine reserves. — Trends in Ecology and Evolution 18: 448-455. Hansen, B. B. et al. 2009. Functional response in habitat selection and the tradeoffs between foraging niche components in a large herbivore. — Oikos 118: 859-872. Hebblewhite, M. and Merrill, E. H. 2007. Multiscale wolf predation risk for elk: does migration reduce risk? — Oecologia 152: 377-387. Houston, A. I. et al. 1993. General results concerning the trade-off between gaining energy and avoiding predation. — Philosophical Transactions: Biological Sciences 341: 375-397. Huggard, D. J. 1993. Prey selectivity of wolves in Banff National Park. II. Age, sex, and condition of elk. — Canadian Jounal of Zoology 71: 140-147. Hugie, D. M. and Dill, L. M. 1994. Fish and game: a game theoretic approach to habitat selection by predators and prey. — Journal of Fish Biology 45: 151-169. Johnson, D. H. 1980. The comparison of usage and availability measurements for evaluating resource preference. — Ecology 61: 65-71. Kellner, J. B. et al. 2007. Fishing the line near marine reserves in single and multispecies fisheries. — Ecological Applications 17: 1039-1054. Keuling, O. et al. 2008. How does hunting influence activity and spatial usage in wild boar Sus scrofa L.? — European Journal of Wildlife Research 54: 729-737. Lande, U. S. et al. 2009. Use of hunters in wildlife surveys: does hunter and forest grouse habitat selection coincide? . — European Journal of Wildlife Research DOI: 10.1007/s10344-009-0291-2: Leslie, P. H. 1945. Some further notes on the use of matrices in population mathematics. — Biometrika 35: 213-245. Lima, S. L. 2002. Putting predators back into behavioral predator-prey interactions. — Trends in Ecology and Evolution 17: 70-75. Lotka, A. J. 1925. Elements of physical biology. — Williams & Wilkins company. Madsen, J. 1998. Experimental refuges for migratory waterfowl in Danish wetlands. II. Tests of hunting disturbance effects. — Journal of Applied Ecology 35: 398-417. 150 Maillard, D. and Fournier, P. 1995. Effect of shooting with hounds on size of resting range of Wild Boar (Sus scrofa L.) groups in Mediterranean habitat. — Ibex J.M.E. 3: 102-107. Markina, F. A. et al. 2004. Physical development of wild boar in the Cantabric Mountains, Álava, Nothern Spain. — Galemys 16: 25-34. Martin, J. et al. 2009. Temporal patterns of predation risk affect antipredator behaviour allocation by Iberian rock lizards. — Animal Behaviour 77: 1261-1266. Mitchell, W. A. and Lima, S. L. 2002. Predator-prey shell games: large-scale movement and its implications for decision-making by prey. — Oikos 99: 249-259. Molinari-Jobin, A. et al. 2004. Life cycle period and activity of prey influence their susceptibility to predators. — Ecography 27: 323-329. Moretti, M. 1995. Biometric data and growth rates of a mountain population of wild boar (Sus scrofa L.), Ticino, Switzerland. — Ibex J.M.E. 3: 56-59. Oaten, A. and Murdoch, W. W. 1975. Switching, functional response, and stability in predator-prey systems. — American Naturalist 109: 299-318. Owen-Smith, N. 2008. Changing vulnerability to predation related to season and sex in an African ungulate assemblage. — Oikos 117: 602-610. Quenette, P. Y. and Gerard, J. F. 1992. From individual to collective vigilance in wild boar (Sus scrofa). — Canadian Jounal of Zoology 70: 1632-1635 R Development Core Team 2008. R: A Language and Environment for Statistical Computing. — R Foundation for Statistical Computing, Vienna, Austria, <http://www.R-project.org/>. Revilla, E. et al. 2001. Edge-core effects and the effectiveness of traditional reserves in conservation: Eurasian badgers in Donana National Park. — Conservation Biology 15: 148-158. Roberts, C. M. et al. 2005. The role of marine reserves in achieving sustainable fisheries. — Philosophical Transactions: Biological Sciences 360: 123-132. Sih, A. 2005. Predator–prey space use as an emergent outcome of a behavioral response race. — In: Barbosa, P. and Castellanos, I. (eds), Ecology of predator–prey interactions. Oxford University Press, pp. 240-255. 151 Sodeikat, G. and Pohlmeyer, K. 2003. Escape movement of family groups of wild boar Sus scrofa influenced by drive hunts in Lower Saxony, Germany. — Wildlife Biology 9: 43-49. Stedman, R. et al. 2004. Integrating wildlife and human-dimensions research methods to study hunters. — Journal of Wildlife Management 68: 762-773. Sutherland, W. J. 1996. From individual behaviour to population ecology. — Oxford University Press. Tolon, V. et al. 2009. Responding to spatial and temporal variations in predation risk: space use of a game species in a changing landscape of fear. — Canadian Jounal of Zoology 87: 1129-1137. Truvé, J. and Lemel, J. 2003. Timing and distance of natal dispersal for wild boar Sus scrofa in Sweden. — Wildlife Biology 9: 51-57. Volterra, V. 1926. Fluctuations in the abundance of a species considered mathematically. — Nature 118: 558-560. 152 153 154 7. PARTIE 4 : LE MOUVEMENT DANS LE PAYSAGE DU RISQUE 155 MOVING IN A LANDSCAPE OF RISK: THE ROLE OF DIRECTED AND NON-DIRECTED HABITAT SELECTION PROCESSES. Vincent Tolon1,2, Bram van Moorter3, Stéphane Dray2, Anne Loison1 & Eric Baubet4. 1 Laboratoire d’Ecologie Alpine CNRS UMR5553, Université de Savoie, Bâtiment Belledonne, F-73376 Le Bourget-du-Lac, France. 2 Laboratoire de Biométrie et Biologie Evolutive CNRS UMR5558, Université de Lyon, Bâtiment Gregor Mendel, F-69622 Villeurbanne, France. 3 Centre for Conservation Biology, Department of Biology, Norwegian University of Science and Technology, NO-7491 Trondheim, Norway. 4 Office National de la Chasse et de la Faune Sauvage, Centre National d’Etude et de Recherche Appliquée - Cervidés et sanglier, F-55 000 Bar-le-Duc, France. In prep. 156 ABSTRACT To select their habitat, animals can simply modify intrinsic movements’ properties according to proximate environmental conditions (non-directed movements) but can also adjust their direction and velocity according to distant habitats (directed movements). This last process however involves a large scale perception of habitats which is not assured especially for prey moving in an elusive landscape of risk. We studied inter-day movements of wild boar Sus scrofa L. resting sites, in response to the landscape openness, the distance to road, a seasonal probability density of hunting dogs’ presence and the hunting day status (open, close). We used a Step Selection Function to study if animals only modify the distance moved according to immediate risk levels, or selected directly lower risk habitats. As expected (directed response), wild boar directly selected forests and areas of low dog probability density far from roads to relocate their new resting site. Against expectation, the distance moved was not affected by the dog probability density on the initial resting site. The non-directed response was therefore not effective in our case. Moreover, we did not observed any differences in the strength of the responses to hunting according to the day hunting status, i.e. the true presence or absence of the risk source at a given date. These results suggested that wild boar use indirect sources of information about risk levels (e.g. olfactory cue, spatial memory, social acquired information) to perceive the large scale spatial variations of the landscape of risk and perform truly informed choices of habitats, even if the risk source was temporary absent. Our approach showed how movement patterns of prey in contrasted landscape of risk can help understanding their mechanisms of risk perception. Key words: wild boar, Sus scrofa, movement, predation risk, refuge, Step Selection Function. 157 INTRODUCTION Habitat selection – the non-random use of the available habitat (Johnson 1980) – is an important behavioural process affecting fitness (McLoughlin et al. 2006, McLoughlin et al. 2007) as it alters the resources and risk conditions met by individuals during their lifetime. Habitat selection patterns have been studied in situ for decades, however our understanding of processes underlying observed patterns in natural environments is more recent (reviewed in Martin et al. 2009b). Within their home ranges (third scale of Johnson 1980), animal habitat selection corresponds to differential time allocation among habitats and results from individual movements in response to environmental variations (Morales et al. 2004, Frair et al. 2005). Therefore, understanding how, where and why animals move is essential to explain individual space use and habitat selection patterns (Nathan et al. 2008). Two general types of movement are traditionally distinguished based upon their patterns, underlying processes and efficiency: directed versus non-directed movements (Mueller and Fagan 2008). Non-directed movements (or kinesis) involve only changes in intrinsic movements properties (i.e. animals change how they move) due to proximate environmental conditions of the individual location (Benhamou and Bovet 1989). Animals can therefore spend more time in a food-rich patch (Knegt et al. 2007) or a lowrisk area (Lewis et al. 2001) by reducing their velocity or increasing their turn frequency when moving in these habitats. In contrast, directed movements are characterised by adjustments of the movement direction and velocity according to conditions distant from the individual (animals choose where to move). Animals can therefore move in the direction or at higher speeds towards perceived distant environmental structures (e.g. vision and olfaction cue in domestic pig, Croney et al. 2003) or memorised targets places (cognitive maps, O'Keefe and Nadel 1978, e.g. path integration, Benhamou 1997, roadfollowing processes, Lipp et al. 2004). Non-directed and directed movements can coexist for a given taxon (e.g. large ungulates, non-directed: Bergman et al. 2000, Morales et al. 2004, directed: Fortin et al. 2005, Dalziel et al. 2008, Wolf et al. 2009). Directed movements, which require more 158 information, are usually more efficient than non-directed movements, which are partially random (Armsworth and Roughgarden 2005, Brooks and Harris 2008). The strength of directed responses (vs. non-directed) may be accentuated by the increased perceptual capacities of individuals, the reduced scale of the environment spatial heterogeneity (in fine grained environments individuals can easily move toward the preferred habitat types) and the spatiotemporal predictability of habitats conditions (Zollner and Lima 1999, Mueller and Fagan 2008). For a prey, the spatial and temporal variations of the predation risk often result from partially imperceptible and unpredictable predators. Their responses to the landscape of risk could therefore be subjected to both types of movements. Indeed one of the simplest anti-predator behaviour of prey is first to leave their current location, at different spatial scale, in response to local predator presence (e.g. flight, Ydenberg and Dill 1986, patch giving-up, Brown 1999, seasonal migration, Hebblewhite and Merrill 2007, Stankowich 2008). This process alone could theoretically induces habitat selection because preys will simply remain longer in low risk areas due to a decreased movement frequency following the relative absence of predators (Lewis et al. 2001). However, if preys have the opportunities to perceive a marked spatial heterogeneity of their landscape of risk, they would strongly benefit from movements directed toward low risk areas to minimise the time exposed to predators (e.g. Cowlishaw 1997). We investigated the directed and non-directed responses of a prey performing “flea-jumps” between daily resting sites in a structured landscape of risk. We used the wild boar Sus scrofa L., a nocturnal game species (Boitani et al. 1994, Keuling et al. 2008b) that faces extreme hunting pressures (e.g. 40% of chance of being harvested annually, up to 70% for adult males, Toïgo et al. 2008). These animals were shown to be highly sensitive to the structure of the landscape of risk, and select less hunted areas for their diurnal resting sites (Maillard and Fournier 1995, Sodeikat and Pohlmeyer 2003, Tolon et al. 2009). Wild boar usually return to their previous diurnal resting site after their nocturnal feeding phase (e.g. Janeau et al. 1995) except if they are disturbed on it by hunting (Maillard and Fournier 1995, Sodeikat and Pohlmeyer 2003). Non-directed responses could therefore partly or wholly explain the avoidance of high risk areas. When hunters are on the field, wild boars can however rely on direct cues about their locations 159 and could then relocate their next resting site directly in lower risk areas. We studied from an individual-centred point of view, inter-day movements of wild boar resting sites in a landscape of risk structured on the long-term by the presence of a protected area (Tolon et al. 2009). Controlling for the general preference of wild boar for forest (Boitani et al. 1994, D'Andrea et al. 1995) we first predicted that animals relocated directly their resting sites in surrounding lower risk areas (directed response), regarding to both the spatial variations of hunting activities (the landscape of risk, P1) and roads proximity (the second sources of mortality on our study site, P2). Moreover, we expected animals to increase their distance moved in response to the hunting disturbance level of their previous resting site (non-directed response, P3). Finally, we predicted that the strength of directed and non-directed responses to the hunting were higher during open hunting days (P3) because wild boar are not directly confronted with this disturbance source during others days. MATERIALS AND METHODS Study area The Geneva basin is located on the border between France and Switzerland and is surrounded by mountain chains and the Lake of Geneva. The study site is a 16×11 km area around a nature reserve of 318 ha (see a map in Tolon et al. 2009). Natural large predators of wild boar are absent in the Geneva basin and the main source of mortality is mostly anthropogenic (hunting or culling: 81.6%, collision: 15.3%, other: 3.1%; n=196 dead recoveries, unpublished data). Hunting occurs from the second Sunday of September to the last Sunday of January, between 8:00 a.m. and 5:00 or 7:00 p.m, exclusively on Thursday, Saturday and Sunday. The study site supported 7 hunting teams, each using a restricted hunting territory, ranging from 2.9 to 23 km² (median = 12.2 km²). Five hunting territories included protected areas, forbidden to hunters but potentially covered by their hunting dogs. During the hunting season, hunters perform 160 between one and six drive hunts per week within different areas of their hunting territories. Wild boar movements Wild boar were trapped, weighted and ear-tagged from 2003 to 2007. Around the focus protected area, 26 individuals heavier than 30 kg were fitted with Very High Frequency radio-transmitters (♀: N=19, ♂: N=4) or with GPS devices (LOTEK®, ♀: N=3, ♂: N=1). Animals were located once a day between 7:00 am to 7:00 pm, during their resting period. Resting sites were mostly unique throughout the day (confirmed with fine-scale GPS tracking, unpublished data). During a single night, animals generally travelled around 5 or 6 km for females and males respectively (Tolon et al. 2009) to feed on local resources, before returning close to their previous resting site. These movements occurred in ranges of 7.6 or 10.1 km² (MCP 95%), for females and males respectively (Tolon et al. 2009). Therefore, movements observed between consecutive resting sites are not the consequence of fine-scale nocturnal movements but results from the decision of the animal to return on the previous resting site or to change its location after a disturbance phase (day) and a feeding phase (night). In the following analyses, we focused exclusively on inter-day movements performed between consecutive resting sites during the open hunting season. In addition, we limited our dataset to animals possessing at least 10 daily locations associated with locations of the day before and after, i.e. 10 observed steps (a statistical unit required three successive locations, see below and Fortin et al. 2005). We therefore obtained 976 observed steps amongst the 26 individuals (range = [11, 91]). Landscape of risk and others habitat variables The seasonal landscape of risk was quantified with GPS tracking of hunting dogs performed during two hunting season (2006 and 2007). A previous study of hunters habits showed that around a thousand of drive hunts have been performed during both 161 years on the seven hunting territories (around 70 drive hunts per season and hunting territories × 2 years of study × 7 hunting territories = 980 drive hunts, unpublished data). For 105 of these drive hunts, one or two dogs were fitted with GEOPOINTER® GPS collars and located every minute from the beginning of the hunt to their recovery by their master (duration: 133 +/-79 minutes, mean +/-sd). At the beginning of drive hunts, dogs were on leashes and became free to move after having been released close to expected wild boar resting places. Space use of dogs therefore revealed a mix between hunters and dogs behaviors. We use these data to compute a spatially explicit measure of seasonal hunting activity on a 50×50 meter grid map with a kernel utilization distribution (UD, Worton 1989). The number of drive hunts recorded per hunting territory varied from 5 to 24 for the two years considered. To overcome this sampling bias we computed independently kernel UDs (bivariate method) for each hunting territory. We then combined these UDs on the entire study site (see Hebblewhite and Merrill 2007), giving the number of hunting days per season as weights of each UD (varying from 30 to 45 days). We obtained the two-dimensional probability density of hunting dogs (named dog in the following), considered as an estimate of the seasonal landscape of risk (map in Fig. 7.1). The central protected area was designed several decades ago so that the large-scale structure of the landscape of risk around the reserve was expected to be similar between years. Moreover, spatial variations of the seasonal dog probability density appeared to be well correlated between 2006 and 2007 (r = 0.702, Pearson correlation coefficient). In our analysis, we included two other variables potentially related to disturbance levels for wildlife: the landscape cover (forest: 1, open field: 0) and the distance to the closest road (e.g. Coulon et al. 2008). Roads were also the second source of mortality of wild boar on our study site (see above). We removed the positive skew of the dog-density and the distance to road by a log transformation. Movement analysis We used a Step Selection Function (SSF, Fortin et al. 2005) to study the successive choices of resting sites within wild boar home ranges according to the 162 environmental conditions, the intensity of risk, the distance to the previous resting site and the hunting status of each day (open, close). Home ranges were measured as the 100% Minimum Convex Polygon of all day and night locations recorded during the open hunting season (i.e. seasonal home ranges). We associated each observed step (a movement between two consecutives resting sites) of a given individual with 199 control steps (as in Fortin et al. 2005), simulated within its home ranges. We generated step lengths from an exponential distribution with rate parameter (2µi)-1, where µi is the mean length of all observed step for the individual i (following the recommendation of Forester et al. 2009). For a given date, the starting point of the simulated control steps was the starting point of the observed step. The direction of the simulated step was drawn from a uniform distribution because we have no biological reason to expect directional persistence for this time interval (one day) within the home range (Forester et al. 2009). We measured on endpoint of each (observed and control) step the corresponding landscape cover (landscape), the distance to road (droad), the dog probability density (dog) and the distance from the starting point (i.e. step length: dist). For following analyses we also measured the dog probability density on the starting point (dogt-1) and the previous distance moved (distt-1). Note that for an observed step and its associated control ones, all the values measured at t-1 were equal. As seasonal home range sizes greatly vary among individuals, we standardized both dist and distt-1 variables by dividing the observed values by the individual dist standard deviations. Finally, as hunting occurred only three day per week (see below), we returned the hunting status hunt (open = 1, close = 0) associated to each step. Models parameters We used a Conditional Logistic Regression (CLR) to model the step selection probability (Fortin et al. 2005). To test previous predictions, we included, in the CLR, terms expected to describe non-directed and directed response to the dog probability density, in relation to the hunting day status (Fig. 7.1, Table 7.1). First we included the term dist in order to reduce potential bias on habitat coefficients estimates (following the 163 recommendation of Forester et al. 2009). Additive terms like landscape, droad and dog indicate that steps arrival points generally fall within particular habitat surrounding the starting point. We therefore considered these coefficients as measures of the directed response to habitats (see: Fig. 7.1a & 1b). Interaction between the distance and the dog probability density of the starting point, i.e. dist:dogt-1, showed that arrival points were more or less distant according to habitat conditions of the previous resting site. We therefore considered this interaction as a measure of the non-directed response to the hunting risk (the non-directed response is observable on the distance moved, Fig. 7.1c & 7.1d). Table 7.1: Coefficients used in the Step Selection Function of wild boar moving between resting sites, with their expected effect, attended by their biological justification. Variables used were the moving distance (dist, and its value at the previous step distt-1) the landscape cover (forest, dichotomous), the log-distance to the nearest road (droad), the log of the dog probability density (dog, and its values on the starting point dogt-1) and the day hunting status (hunt, dichotomous). Process Coefficient Effect Distance effect dist 0 landscape + droad + dog - Directed response (see Fig 7.1a & 7.1b) Non-directed response (see Fig Temporal variations 164 . The step simulation process removes the effect of the distance. . Forests provide shelter and were highly perceptible. . Roads create mortality and were used every day by humans. . Wild boar perceive hunters and dogs’ presence during hunting days. . Wild boar relocated their resting site dist:dogt-1 + 7.1c & 7.1c) Inertia Justification farther in response to the previous disturbance. dog:huntt - dist: huntt:dogt-1 + dist:distt-1 + . Stronger response to hunters and dogs when they were truly on the field. . Similarity of consecutive distances moved Terms dog:hunt and dist:hunt:dogt-1 verified that the strength of the responses to the dog probability density depended on the day hunting status, i.e. the true presence or absence of the risk source at a given date. Finally, we took into account of the potential inertia in movements due to the internal state of animals (e.g. energetic reserves, basal stress, social or reproductive status) or the environment (e.g. climate along the hunting season). In case of a strong inertia, the actual selected distance should be similar to the precedent (e.g. alternation of slow and fast movement phases, acceleration or deceleration). We therefore included the interaction between the previous distance moved and the actual moving distance, i.e dist:distt-1. In order to detect eventual confounding effects amongst explanatory variables, we explored the distribution of Pearson coefficients of determination (r²) between the distance and habitat variables observed amongst steps of a given date (simulated and observed), to conclude on the interpretability of coefficients. We applied the procedure described by Fortin et al. (2005, Appendix 1) to detect serial correlation in the residuals of step selection. 165 a) βr = -3, βd = -0.003, βd:r’ = 0 b) βr = -3, βd = -0.003, βd:r’ = 0 c) βr = 0, βd = -0.006, βd:r’ = 150 d) βr = 0, βd = -0.006, βd:r’ = 150 Figure 7.1: Risk map and predicted redistribution probability for the next resting site of an individual (i.e. presence probability of arrival points) according to the distance from the previous resting site (d), the surrounding risk level (r), the initial risk level (r’) and the movement type. The score of each pixel was computed as exp(rβr+dβd+dr’βd:r’) (see, Boyce and McDonald 1999, Forester et al. 2009). White points: location of the starting resting site. Black contours: isoclines of the redistribution density (8 levels regularly spaced between the maximal and the minimal density for each plot). The purely directed response (P1 & P2, plots a & b) involve the relocation of arrival points in lower risk areas surrounding the starting point (βr = -3), given the fixed distance effect (βd = -0.003, βd:r’ = 0). The purely non-directed response (P3, plots c & d) involve an increased distancing of arrival points according to the risk at the starting points (βd = -0.006, βd:r’ = 150) without any direct selection for low risk areas (βr = 0). βd and βd:r’ therefore measured the strength of directed and non-directed responses respectively. Note that, at the difference of these examples, spatial units in the analysis were not pixels but random steps simulated from each starting. 166 RESULTS The median distance moved between two consecutives resting sites was equal to 387 m (quantile 95% = 1591 m). Slight differences of step length appeared between sexes (♂: median = 427 m, quantile 95% = 1780 m; ♀: median = 379 m, quantile 95% = 1579, W=78208, P = 0.054, Wilcoxon rank sum test). Distributions of Pearson coefficients of determination showed neither strong confusion between the distance and habitat variables (dist vs. dog: r² = 0.148 [0.000336; 0.709]; dist vs. landscape: r² = 0.0263 [7.67×10-5; 0.204]; dist vs. droad: r² = 0.0778 [0.000193; 0.573], median [quantile 2.5%; 97.5%]) nor between habitat variables amongst steps of a given date (dog vs. droad: r² = 0.0329 [4.89×10-5; 0.336]; landscape vs. dog: r² = 0.0449 [0.000125; 0.283]; landscape vs. droad: r² = 0.0316 [8.54×10-5; 0.248], median [quatile 2.5%; 97.5%])). The autocorrelation function did not showed any significant temporal autocorrelation in the step choices (note that it was not the case when the term dist:distt-1 was removed). Standard errors of coefficients can then be considered as unbiased. The Conditional Logistic Regression did not showed any effect of the distance on the step selection probability (dist, Table 7.2). All habitat variables on step endpoints had strongly significant effects (Table 7.2): wild boar directly selected for increased landscape cover (landscape, Fig. 7.2a), whereas they avoided high dog probability density (dog, Fig. 7.2b) and road proximity (droad, Fig. 7.2a). We could not observe any difference in the distance moved according to the dog probability density on the starting point (dist:dogt-1, Table 7.2). Effects of the hunting day status on the directed and the non-directed response to the dog probability density were not significant (dog:hunt & dist:hunt:dogt-1, Table 7.2, Fig. 7.2b). The interaction of the distance moved with its previous value (dist:distt-1, Table 7.2) was highly significant. A strong inertia in the inter-day resting site movement therefore leads animal to select the distance moved similarly to the precedent (long distance moved at t-1 involved a preference for long distance at t, Fig. 7.2c). 167 Table 7.2: Coefficients of the Step Selection Function of wild boar moving between resting sites. Variables used were the moving distance (dist, and its value at the previous step distt-1) the landscape cover (forest, dichotomous), the log-distance to the nearest road (droad), the log of the dog probability density (dog, and its values on the starting point dogt-1) and the day hunting status (hunt, dichotomous). See Table 7.1 for justifications of the interactions used. Robust Standard Variable Coefficient Error Z P(>|z|) dist 0.10421 0.63390 0.1644 0.87 landscape 1.12627 0.08217 13.7063 <10-5 droad 0.38139 0.06903 5.5254 <10-5 dog -0.42492 0.06039 -7.0368 <10-5 dist:dogt-1 0.00301 0.05063 0.0594 0.95 dog:hunt 0.04761 0.12829 0.3711 0.71 dist:hunt:dogt-1 0.00718 0.00648 1.1071 0.27 dist:distt-1 0.12576 0.02450 5.1331 <10-5 DISCUSSION Our results showed strong direct selection patterns for surrounding refuges habitats when wild boar moved from a resting site to the following. Indeed, when wild boar relocated their new resting site, they moved directly toward forested plots and low dog probability density areas (P1 supported) far from roads (P2 supported, Table 7.2, Fig. 7.2a & 7.2b). Against expectation we did not observed any effect of the dog density of the initial resting on the next distance moved (P3 not supported). Wild boar were reported to increase movements between resting sites when the hunting season starts (temporal effect) in response to the increased disturbance (Maillard and Fournier 1995, Calenge et al. 2002, Sodeikat and Pohlmeyer 2003). Our results showed that this process was not clearly apparent within the hunting seasons, when animals meet different levels of disturbance in space. 168 0.007 0.008 0.005 0.006 hunt = open hunt = closed 0.003 0.002 0.003 Step Selection Probability 0.004 landscape = forest landscape = open 0.004 0.005 b) 0.001 Step Selection Probability a) 5.0 5.5 6.0 -13.5 log(droad) -13.0 -12.5 -12.0 -11.5 -11.0 log(dog) 0.004 0.008 dist(t-1) = 0.03 dist(t-1) = 2.08 dist(t-1) = 4.13 0.000 Step Selection Probability 0.012 c) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 dist(t) Figure 7.2: Step Selection Probability (Conditional Logistic Regression) from one resting site to the following for hunted wild boar according to habitats and distances variables. a) Additives effects (directed response) of the log-distance to roads (droad) and the landscape cover (landscape) of steps’ endpoints. b) Additive effect (directed response) of the log dog probability density (dog) of endpoints, according to the day hunting status (hunt). c) Interaction between the distance of steps arrival points (dist) and the previous distance moved (dist(t-1)). In our case, the strength of the directed response to the landscape of risk (dog in Table 7.2) may explain this absence of non-directed response. Indeed the landscape of risk was structured every year by a protected area, such as it spatial predictability was very high within a season, even between years. In this context directed responses may be strongly favoured (Mueller and Fagan 2008) and lead to a marked use of low risk areas 169 without enough possibilities to observe animals starting their movement from high risk areas (a necessary conditions to observe the non-directed response). The intra-home range habitat selection of wild boar in such a predictable landscape of risk was therefore mostly driven by directed rather than non-directed responses. If the directed component of movements (vs. non-directed) is determined by the perceptibility of habitats (Zollner and Lima 1999), the observed directed responses may involve mechanisms of perception at large scale (within the home range) for the three habitat variables considered. Movements directed toward forests were not surprising considering the easy perceptibility of this habitat for a large mammal like the wild boar (Mech and Zollner 2002). Because the diurnal road traffic was strong on our study site and persisted residually at night, wild boar could also easily perceive disturbance cues while choosing their resting at dawn and select a site far from roads (Fig. 7.2a). To select low dog probability density wild boar could rely on the direct perception of hunters and dog presence during drive hunt if they relocated their resting site at day after disturbance, instead of at dawn. However, against expectation, we could not observe any differences in the strength of the responses to the landscape of risk according to the hunting day status, i.e. the true presence or absence of hunters and hunting dogs on the field at a given date (P4 not supported, Table 7.2, Fig. 7.2c). This suggested that wild boar used indirect sources of information about risk level in space, sure enough to respond with the same intensity than during hunted days. When choosing their new resting site at dawn, they could for instance use olfactory cues (e.g. domestic pig, Croney et al. 2003) to perceive indirectly the long term presence of hunters and dogs, and chose a resting site with a lower (or the lowest) smell of danger. Wild boar could also easily associate risk levels to particular sites or particular habitats (ex: the marsh of the protected area) after previous disturbance experiments (e.g. domestic pig, Held et al. 2000, Held et al. 2005, Jansen et al. 2009). In a stable landscape of risk (over years), animals could also learn the location of best refuges from the maternal behaviour (juveniles wild boar followed their mother during their first year, Cousse et al. 1994, Truvé and Lemel 2003) and adopt a site fidelity strategy. Individually acquired and socially acquired information could then act 170 conjointly (Dall et al. 2005) to produce an adaptive habitat selection in the landscape of risk, even if no hunters nor dog where present during this choice. From all environmental components, predation risk is one of the most important determinant of prey fitness (Kjellander et al. 2004). Prey may therefore have a strong interest to use the most reliable sources of information (sensu Koops 2004) about risk levels, with the lowest possible costs, in order to perform truly informed adaptive choices of habitats (Dall et al. 2005, i.e. being omniscient). Under the postulate that animals tend to maximise their survival (e.g. non-limiting food, critical phase of life cycle, high predation rate) the study of their movements in a given landscape of risk could therefore give relevant information about their actual risk perception in space (i.e. directed reponses: large scale perception, non-directed response: proximate perception). At the same time, the potential changes of movement characteristics, in response to an unpredictable presence or absence of the risk source in the environment, could reveal the importance of direct and indirect sources of information to assess risk levels (i.e. no changes: indirect source, disappearance of the response: direct sources). This approach requires considering both directed and non-direct processes of habitat selection to get an overview of the spacing response of animals to their environment. Our study showed Step Selection Functions could provide such opportunities for free ranging animals by considering conjointly direct effects of habitats (where animals move) and theirs interactions with intrinsic movements’ properties (how animals move). This can be done by using the distance, or even the turning angle, as predictors of the step selection (e.g. distance in Forester et al. 2009) rather than simple constraints on the step simulation (e.g. Fortin et al. 2005, Coulon et al. 2008). We believe that such approach may strongly help understanding habitat selection processes within home ranges of free ranging animals. ACKNOWLEDGEMENTS This study was commissioned and funded by the Office National de la Chasse et de la Faune Sauvage (France), the Fédération Départementale des Chasseurs de l’Ain et de la 171 Haute Savoie (France), the Chambre d’Agriculture de la Haute Savoie (France), the Direction générale de la Nature et du Paysage of Canton of Geneva (Switzerland), the Federal Office for the Envionnement (Switzerland), ECOTEC Environnement SA (Switzerland) and the Service des Forêts, de la Faune et de la Nature of Canton of Vaud (Switzerland). The authors thank the Conseil General de l’Ain (France) for additional funding. Many thanks to all the field personnel involved, especially to J.O. Chappuis, M. Comte, F. Corcelle and J. Vasse. REFERENCES Armsworth, P. R. and Roughgarden, J. E. 2005. The impact of directed versus random movement on population dynamics and biodiversity patterns. — American Naturalist 165: 449-465. Benhamou, S. 1997. Path integration by swimming rats. — Animal Behaviour 54: 321327. Benhamou, S. and Bovet, P. 1989. How animals use their environment: a new look at kinesis. — Animal Behaviour 38: 375-383. Bergman, C. M. et al. 2000. Caribou movement as a correlated random walk. — Oecologia 123: 364-374. Boitani, L. et al. 1994. Spatial and activity patterns of wild boars in Tuscany, Italy. — Journal of Mammalogy 75: 600-612. Boyce, M. S. and McDonald, L. L. 1999. Relating populations to habitats using resource selection functions. — Trends in Ecology and Evolution 14: 268-272. Brooks, C. J. and Harris, S. 2008. Directed movement and orientation across a large natural landscape by zebras, Equus burchelli antiquorum. — Animal Behaviour 76: 277-285. Brown, J. S. 1999. Vigilance, patch use and habitat selection: Foraging under predation risk. — Evolutionary Ecology Research 1: 49-71. 172 Calenge, C. et al. 2002. Summer and hunting season home ranges of wild boar (Sus scrofa) in two habitats in France. — Game and Wildlife Science 19: 281-301. Coulon, A. et al. 2008. Inferring the effects of landscape structure on roe deer (Capreolus capreolus) movements using a step selection function. — Landscape Ecology 23: 603-614. Cousse, S. et al. 1994. Use of space by juveniles in relation to their postnatal range, mother, and siblings: an example in the wild boar Sus scrofa L. — Canadian Jounal of Zoology 72: 1691–1694. Cowlishaw, G. 1997. Refuge use and predation risk in a desert baboon population. — Animal Behaviour 54: 241-253. Croney, C. C. et al. 2003. A note on visual, olfactory and spatial cue use in foraging behavior of pigs: indirectly assessing cognitive abilities. — Applied Animal Behaviour Science 83: 303-308. D'Andrea, L. et al. 1995. Preliminary data of the Wild Boar (Sus scrofa) Space use in mountain environment. — Ibex J.M.E. 3: 117-121. Dall, S. R. X. et al. 2005. Information and its use by animals in evolutionary ecology. — Trends in Ecology and Evolution 20: 187-193. Dalziel, B. D. et al. 2008. Fitting probability distributions to animal movement trajectories: Using artificial neural networks to link distance, resources, and memory. — American Naturalist 172: 248-258. Forester, J. D. et al. 2009. Accounting for animal movement in estimation of resource selection functions: sampling and data analysis. — Ecology 90: 3554-3565. Fortin, D. et al. 2005. Wolves influence elk movements: Behavior shapes a trophic cascade in Yellowstone National Park. — Ecology 86: 1320-1330. Frair, J. L. et al. 2005. Scales of movement by elk (Cervus elaphus) in response to heterogeneity in forage resources and predation risk. — Landscape Ecology 20: 273-287. Hebblewhite, M. and Merrill, E. H. 2007. Multiscale wolf predation risk for elk: does migration reduce risk? — Oecologia 152: 377-387. 173 Held, S. et al. 2005. Foraging behaviour in domestic pigs (Sus scrofa): remembering and prioritizing food sites of different value. — Animal Cognition 8: 114-121. Held, S. et al. 2000. Social tactics of pigs in a competitive foraging task: the 'informed forager' paradigm. — Animal Behaviour 59: 569-576. Janeau, G. et al. 1995. Daily movement patterns variations in Wild Boar (Sus scrofa L.). — Ibex J.M.E. 3: 98-101. Jansen, J. et al. 2009. Spatial learning in pigs: effects of environmental enrichment and individual characteristics on behaviour and performance. — Animal Cognition 12: 303-315. Johnson, D. H. 1980. The comparison of usage and availability measurements for evaluating resource preference. — Ecology 61: 65-71. Keuling, O. et al. 2008. How does hunting influence activity and spatial usage in wild boar Sus scrofa L.? — European Journal of Wildlife Research 54: 729-737. Kjellander, P. et al. 2004. Predation risk and longevity influence variation in fitness of female roe deer (Capreolus capreolus L.). — Procceding of the Royal Society B 271: 338-340. Knegt, H. J. et al. 2007. Patch density determines movement patterns and foraging efficiency of large herbivores. — Behavioral Ecology 18: 1065-1072. Koops, M. A. 2004. Reliability and the value of information —Animal Behaviour 67: 103-111. Lewis, S. et al. 2001. Evidence of intra-specific competition for food in a pelagic seabird. — Nature 412: 816-819. Lipp, H. P. et al. 2004. Pigeon homing along highways and exits. — Current Biology 14: 1239-1249. Maillard, D. and Fournier, P. 1995. Effect of shooting with hounds on size of resting range of Wild Boar (Sus scrofa L.) groups in Mediterranean habitat. — Ibex J.M.E. 3: 102-107. Martin, J. et al. 2009. On the use of telemetry in habitat selection studies. — In: Barculo, D. and Daniels, J. (eds), Telemetry: research, technology and applications Nova Science Publishers Inc. 174 McLoughlin, P. D. et al. 2006. Lifetime reproductive success and density-dependent, multi-variable resource selection. — Procceding of the Royal Society B 273: 1449-1454. McLoughlin, P. D. et al. 2007. Lifetime reproductive success and composition of the home range in a large herbivore. — Ecology 88: 3192-3201. Mech, S. G. and Zollner, P. A. 2002. Using body size to predict perceptual range. — Oikos 98: 47-52. Morales, J. M. et al. 2004. Extracting more out of relocation data: building movement models as mixtures of random walks. — Ecology 85: 2436-2445. Mueller, T. and Fagan, W. F. 2008. Search and navigation in dynamic environments from individual behaviors to population distributions. — Oikos 117: 654-664. Nathan, R. et al. 2008. A movement ecology paradigm for unifying organismal movement research. — Proccedings of the National Academy of Sciences 105: 19052-19059. O'Keefe, J. and Nadel, L. 1978. The hippocampus as a cognitive map. — Oxford University Press. Sodeikat, G. and Pohlmeyer, K. 2003. Escape movement of family groups of wild boar Sus scrofa influenced by drive hunts in Lower Saxony, Germany. — Wildlife Biology 9: 43-49. Stankowich, T. 2008. Ungulate flight responses to human disturbance: A review and meta-analysis. — Biological Conservation 141: 2159-2173. Toïgo, C. et al. 2008. Disentangling natural from hunting mortality in an intensively hunted wild boar population. — Journal of Wildlife Management 72: 1532–1539. Tolon, V. et al. 2009. Responding to spatial and temporal variations in predation risk: space use of a game species in a changing landscape of fear. — Canadian Jounal of Zoology 87: 1129-1137. Truvé, J. and Lemel, J. 2003. Timing and distance of natal dispersal for wild boar Sus scrofa in Sweden. — Wildlife Biology 9: 51-57. Wolf, M. et al. 2009. The attraction of the known: the importance of spatial familiarity in habitat selection in wapiti Cervus elaphus. — Ecography 32: 401-410. 175 Worton, B. J. 1989. Kernel methods for estimating the utilization distribution in homerange studies. — Ecology 70: 164-168. Ydenberg, R. C. and Dill, L. M. 1986. The economics of fleeing from predators. — Advances in the study of behavior 16: 229-249. Zollner, P. A. and Lima, S. L. 1999. Search strategies for landscape-level interpatch movements. — Ecology 80: 1019-1030. 176 177 178 8. PARTIE 5 : EXPLOITATION DES ESPECES ET EFFET REFUGE 179 L’EXPLOITATION DES ESPECES PRES D’UN REFUGE CONTEXTE ET OBJECTIF De nombreux modèles de dynamique de populations spatialisés ont été développés afin de déterminer le prélèvement optimal à réaliser à proximité des refuges (e.g. Mangel 1998, voir revue dans Gerber et al. 2003, Neubert 2003). Néanmoins, peu de ces approches considèrent la capacité des individus à répondre véritablement au risque d’un point de vue comportemental par une redistribution dans les zones refuges (e.g. Lundberg and Jonzen 1999). L’objectif de ce travail est d’explorer, à l’aide de simples simulations, différents scénarios théoriques de dynamique d’une population dont les individus expriment cette réponse comportementale. De façon à lier les patrons de distribution avec la dynamique de la population, nous utiliserons la théorie des isodars développé par Morris (1987, 2003, 2006) basée sur la Distribution Libre Idéale (Ideal Free Distribution, IFD, Fretwell 1972). Ce travail consiste dans un premier temps à présenter la dynamique de la population et l’évolution des prélèvements réalisés lorsque les animaux expriment une réponse optimale. Dans un second temps, nous étudierons l’influence de l’altération de la perception du risque par les individus sur le fonctionnement du système et nous verrons comment elle pourrait théoriquement permettre de résoudre certaines problématiques de gestion autour des zones protégées. 180 REGLES DE DISTRIBUTION AUTOUR DU REFUGE Distribution Libre Idéale et Isodars Considérons deux habitats dans lesquels les dynamiques de populations suivent intrinsèquement un modèle de croissance logistique, ou modèles de Verhulst (1845), mais dont un seul subit une pression d’exploitation. Nous utiliserons dans ce cas le modèle de pêche de Schaefer (1954) qui consiste simplement à inclure dans les modèles de croissance le prélèvement d’une partie de la population. En temps continu, les dynamiques de population de l’habitat protégé (p) et de l’habitat exploité (e) suivent donc respectivement les modèles de Verhulst et Schaeffer soit: Np N dN e et = re N e 1 − e − EN e = r p N p 1 − K dt dt p Ke dN p avec Ne et Np le nombre d’individus dans les habitats exploité (e) et protégé (p), re et rp les taux de croissance lorsque la densité est nulle, Ke et Kp les capacités d’accueil des deux habitats et E le taux de prélèvement (la proportion d’individus prélevés dans l’habitat exploité). Rappelons que ces deux modèles tendent vers une situation d’équilibre lorsque : dN p dt = 0 soit − r p N p2 Kp + rp N p = 0 et dN e r N2 = 0 soit − e e + (re − E )N e = 0 . dt Ke Les populations deviennent donc stable lorsque : N e* = 0 N *p = 0 et * re − E * N p = K p N e = K e r e Le premier équilibre (N*=0) correspond à une population éteinte. Les autres équilibres correspondent respectivement à une saturation du refuge au niveau de sa capacité d’accueil, et un équilibre entre prélèvement et croissance de la population dans la zone exploitée. 181 Nous considérons ensuite des individus ayant une perception parfaite de leur environnement, libres de se distribuer parmi les deux habitats et maximisant leur fitness. Dans ce cas l’IFD prédit que la distribution optimale des individus parmi les habitats est celle aboutissant à une égalisation des fitness individuelles (isofitness). A mesure qu’une partie de la population croît dans un refuge dont la capacité d’accueil est limitée, certains individus peuvent avoir intérêt à s’établir progressivement hors du refuge afin de bénéficier, malgré le risque, de la faible densité (et donc d’une compétition intraspécifique moindre), jusqu’à l’égalisation des fitness parmi les habitats (Abrahams and Dill 1989, Moody et al. 1996). La fitness des individus peut dans notre cas être définie comme le taux de croissance instantané de la population rapporté à un individu (Morris 2003). Nous obtenons ainsi : 1 dN e 1 dN p = N e dt N p dt En développant cette équation il est possible de calculer la relation entre Np et Ne qui devrait être respectée quelque soit la densité totale. Nous obtenons ainsi l’isodar : (1) re − E − rp N e = K e re K e rp + Np K p re Les isodars décrivent donc les relations théoriques entre les effectifs dans plusieurs habitats lorsque l’isofitness (l’IFD) est respectée parmi les individus (Morris 1987). Dans notre cas, le nombre d’individus dans la zone exploitée est une fonction linéaire du nombre d’individus dans le refuge. En fixant Ke = Kp = 1000, re = rp = 0.5 nous pouvons décrire l’effet de E sur les isodars (Fig. 8.1). 182 1000 800 600 400 E=0.2 200 Ne = Nombre d'individus dans l'habitat exploité E=0 0 E=0.4 0 200 400 600 800 1000 Np = Nombre d'individus dans l'habitat protégé Figure 8.1: Isodars représentant les lignes d’effectifs entre l’habitat protégé (Np) et l’habitat exploité (Ne) pour lesquelles l’isofitness est respectée. Les trois droites correspondent à des taux d’exploitations (proportion des individus prélevés à chaque instant) différents (E = 0, 0.2 et 0.4). Notons que pour des effectifs faibles l’isodar n’est plus exclusivement défini dans le positif (ex : E=0.2 et Np=200) ce qui se traduit dans la réalité par l’occupation d’un seul habitat (ici l’habitat protégé). Sans aucune pression d’exploitation (E=0) la stratégie de distribution optimale consiste à équilibrer les effectifs entre les deux habitats. Lorsqu’une pression d’exploitation existe (E>0) la stratégie optimale pour les individus consiste à occuper d’abord l’habitat protégé si l’effectif total est faible. L’occupation de l’habitat exploité ne devient optimale que lorsque le nombre d’individus dans l’habitat protégé dépasse un certain seuil (ex : 400 individus lorsque E=0.2). Au delà de ce seuil, l’effectif optimale dans l’habitat exploité croît linéairement avec la densité dans l’habitat protégé mais reste toujours inférieure à celle-ci, et ce d’autant plus que la pression d’exploitation est grande. 183 Nous pouvons également exprimer le nombre d’individus dans la zone exploitée en fonction de la densité totale N (Fig. 8.2) en posant N p = N − N e dans (1). Nous obtenons : re − E − rp Ne = K p Ke K r +K r e p p e 300 400 500 600 K e rp + N K r +K r p e e p 200 N=1000 100 N=500 N=100 0 Ne = Nombre d'individus dans l'habitat exploité (2) 0.0 0.1 0.2 0.3 0.4 0.5 0.6 E = Pression d'exploitation Figure 8.2: Droites représentant le nombre d’individus dans la zone exploitée Ne en fonction du taux d’exploitation E pour des densités totales N fixes (N = 100, 500 et 1000 individus). Ainsi pour une densité totale fixe, la densité optimale dans la zone exploitée décroît bien avec l’augmentation de la pression d’exploitation (effet refuge). 184 Notons que les équilibres calculés précédemment (en considérant les deux habitats indépendamment) respectent tout de même l’isodar (l’IFD) puisque si Np = Kp (saturation du refuge) dans (1) alors : re − E − rp N e = K e re (3) K e rp r −E + K p = Ke e = N e* re K p re Evolution des prélèvements réalisés Milieu non saturé Nous définissons d’abord le prélèvement instantané (le nombre d’individus prélevés à un temps t) comme le produit du taux de prélèvement et du nombre d’individus dans la zone exploitée, soit ENe (Lundberg and Jonzen 1999). A partir de (2) nous obtenons ainsi: (4) EN e = − K p Ke K p re + K e rp E 2 + Ke K p (re − rp ) + r p N K p re + K e r p E Le prélèvement instantané est donc une fonction quadratique du taux de prélèvement et varie avec la densité ainsi qu’avec les propriétés des deux habitats. Le terme quadratique étant négatif, il existe un taux de prélèvement instantané optimal correspondant au taux Eopt pour lequel la dérivée de la parabole s’annule (Fig. 8.3), et pour lequel le prélèvement instantané est donc maximal. A partir de (4) nous obtenons : re − rp rp K p Ke K p (re − r p ) + rp N dEN e + N = 0 ⇔ −2 E opt + K e = 0 ⇔ Eopt = 2 2K p dE K p re + K e rp K p re + K e rp En présence d’un effet refuge, un E trop important implique que tous les individus s’établissent dans la zone protégée ce qui annule le prélèvement. Le taux Eopt n’est donc pas le taux maximal qu’il est possible d’atteindre mais un taux intermédiaire dépendant entre autre de la densité et des propriétés du refuge. 185 70 50 40 30 20 ENe = prélèvement instantanné 60 N=1000 10 N=500 0 N=100 0.0 0.1 0.2 0.3 0.4 0.5 E = Pression d'exploitation Figure 8.3: Courbes représentant le prélèvement instantané (ENe) en fonction du taux de prélèvement E et pour des densités totales d’individus fixes (N = 100, 500 et 1000 individus). ENe = taux de prélèvement × nombre d’individus dans l’habitat exploité. Milieu saturé Nous savons que, malgré la contrainte d’IFD, les populations devraient se stabiliser sur des effectifs identiques à la situation d’indépendance entre les habitats (3). Lorsque le milieu est à saturation, le prélèvement vaut ainsi : EN e* = − r K r Ke 2 * * = e et vaut alors E opt N e* = e e . E + K e E . Il est maximisé lorsque E opt 2 4 re Il n’est évidemment plus dépendant des propriétés du refuge et correspond au rendement équilibré maximal classique du modèle de Schaefer (1954). 186 INFLUENCE D’UNE ERREUR DE PERCEPTION Introduction d’un biais de perception du risque sur l’isodar Dans cette seconde partie, les animaux se distribuent toujours suivant une IFD, mais sur la base de ce qu’ils perçoivent et non du risque de prélèvement réel. Ainsi nous remplaçons dans l’isodar le paramètre E par (1-b)E. Le paramètre b représente le biais de perception qu’ont les individus du prélèvement réalisé, c'est-à-dire le taux de sousestimation du risque (voir Shochat et al. 2005). A partir de (1) et (2) nous obtenons respectivement: (5) re − (1 − b )E − rp N e = K e re (6) re − (1 − b )E − rp Ne = K p Ke K r +K r p e e p K e rp + N p et K p re K e rp + N K r +K r p e e p Le biais de perception n’intervient ici que sur l’ordonnée à l’origine de l’isodar qui croît lorsque le biais augmente (Fig. 8.4). Lorsque b vaut 0 les individus perçoivent le risque réel et devrait se comporter idéalement (cf. Fig 8.1). Lorsque 0<b<1, les animaux sous-estiment le prélèvement réalisé et fréquentent la zone exploitée plus que dans l’idéal. Le paramètre b traduit bien le fait que, pour un prélèvement constant E, l’effet refuge sera d’autant plus faible que le biais de perception est grand. 187 1000 800 600 400 b=0.5 200 Ne = Nombre d'individus dans l'habitat exploité b=1 0 b=0 0 200 400 600 800 1000 Np = Nombre d'individus dans l'habitat protégé Figure 8.4: Isodars biaisés représentant les lignes de densités entre l’habitat protégé (Np) et l’habitat exploité (Ne) pour lesquelles l’isofitness apparente (incluant un biais de perception b) est respectée. Dans ces trois cas E=0.4. Les trois droites correspondent à des biais de perception du risque (taux de sousestimation) décroissants (b = 1, 0.5 et 0). Simulations Procédure Nous utiliserons dans cette partie des simulations en temps discret pour estimer les propriétés des populations lorsque le milieu est saturé en fonction du prélèvement réalisé E et du biais de perception b. Nous fixons les mêmes taux de croissance 188 intrinsèque et les mêmes capacités d’accueil que précédemment (re = rp = 0.5 et Ke = Kp = 1000). Les populations se distribueront néanmoins à chaque pas de temps de façon à respecter l’isodar biaisé présenté ci-dessus. Nous considérons d’abord 100 valeurs de E reparties régulièrement entre 0 et re+ rp, ainsi que 100 valeurs de b reparties régulièrement entre 0 (perception exacte du risque) et 1 (aucune perception du risque). Pour chacune des 100×100 combinaisons nous réalisons une simulation sur 1000 pas de temps (nous considérons qu’au-delà de cette durée les populations se sont stabilisées, ce qui à été confirmé lors d’exploration graphique préalable). Nous considérons une population de départ de Ne = Np = 500. A chaque pas de temps, nous calculons d’abord indépendamment les nouveaux effectifs dans les habitats, puis l’effectif total de la population: N t +1 p N tp = N + r p N 1 − Kp t p t p Nt N et +1 = N et + re N et 1 − e − EN et Ke N t +1 = N tp+1 + N et +1 Nous calculons ensuite les nouveaux effectifs au sein des deux habitats permettant, pour un effectif total Nt+1 constant, de respecter l’isodar biaisé. D’après (6), nous obtenons : re − (1 − b )E − rp N et +1' = K p K e K p re + K e rp K e rp + N t +1 K p re + K e rp et N tp+1' = N t +1 − N et +1' Ces nouveaux effectifs sont différents des effectifs initiaux mais l’effectif total reste le même. Cela implique donc un transfert d’individus entre les habitats. Lorsque le transfert attendu est supérieur à l’effectif présent dans l’habitat de provenance des individus, nous considérons simplement que cet habitat se vide des individus restants. Dans ce cas l’isodar biaisé n’est plus respecté et tous les individus sont dans l’habitat 189 perçu comme étant le meilleur. A l’issu des 1000 itérations, nous relevons les paramètres suivants : - les effectifs Ne et Np vers lesquels la simulation converge dans les habitats exploités et protégés respectivement. - le flux d’individus nécessaire pour que la population se maintienne à terme sur l’isodar biaisé : N tp+1 − N tp+1' - le produit du taux de prélèvement et de l’effectif final de la population exploité, soit ENe (prélèvement réalisé). Nous étudions ensuite comment évoluent ces grandeurs de façon descriptive en fonction du taux de prélèvement E et du biais de perception b (Fig. 8.5 & 8.6). Résultats Le biais de perception modifie le nombre d’individus à l’équilibre dans le refuge. Plus il approche de 1 (pas de perception du risque) plus la densité à l’équilibre décroît dans l’habitat protégé (Np) surtout lorsque le prélèvement est important. Il est ainsi possible de maintenir la population du refuge à environ 700 individus, contre 1000 dans la situation idéale (b=0), si le biais de perception vaut 0.8 (les animaux perçoivent un risque cinq fois inférieur au risque réel) et pour un taux de prélèvement de 0.4. Cette réduction d’effectif à l’équilibre dans le refuge s’accompagne d’une augmentation d’effectif à l’équilibre dans l’habitat exploité (Ne). Ainsi pour les mêmes biais de perception (b=0.8) et taux de prélèvement (E=0.4), la population de l’habitat exploité s’équilibre à environ 550 individus, contre 200 dans la situation d’IFD réelle (b=0). 190 Figure 8.5: Propriétés des populations à l’équilibre (Kp = Ke = 1000, rp = re = 0.5, à l’issue de 1000 pas de temps) en fonction du taux d’exploitation E (proportion d’individus prélevés dans l’habitat exploité) et du biais de perception du risque b (pourcentage de sous-estimation). Np : nombre d’individus à l’équilibre dans l’habitat protégé ; Ne : nombre d’individus à l’équilibre dans l’habitat exploité ; Flux : nombre d’individus transitant de l’habitat protégé vers l’habitat exploité de façon à respecter l’isodar biaisé à l’équilibre; ENe : prélèvement réalisé à l’équilibre. Les valeurs des paramètres sont codées en niveau de gris (0=blanc, gris foncé=valeur maximale) et les lignes noir représentent les isoclines. 191 Ces changements d’effectifs à l’équilibre sont provoqués par des flux d’individus provenant du refuge et dirigés vers la zone exploitée. Pour les mêmes valeurs que précédemment, un flux d’environ 100 individus se maintient entre les habitats. Nous observons au final que le prélèvement réalisé augmente avec le biais de perception. Ainsi dans notre exemple le prélèvement vaut environ 225 individus contre 125 dans la situation d’IFD. Il apparaît donc bien que le biais de perception amène les individus du refuge à venir combler les niches « apparemment » laissées vacantes par les individus prélevés dans la zone exploitée. Nous pouvons noter que, dans notre exemple, le refuge et la zone exploitée ont exactement les mêmes propriétés intrinsèques (re = rp et Ke = Kp). De façon exploratoire, nous réitérons les mêmes simulations que précédemment mais en fixant cette fois le taux de croissance intrinsèque du refuge à rp = 0.1 (Fig. 8.6). Nous observons alors qu’il est possible de « vider » le refuge de sa population et même de mener la population entière à l’extinction en maintenant un biais de perception réaliste (0.4 < b < 1) pour un prélèvement modéré (E=0.4). Comme l’habitat exploité est intrinsèquement de meilleure qualité que le refuge, les individus peuvent être amenés à coloniser d’abord la zone exploitée, surtout s’ils ont une perception biaisée du risque. Ainsi le refuge peut rester vide à l’équilibre et la population entière peut même disparaître si le prélèvement devient trop important. 192 Figure 8.6: Propriétés des populations à l’équilibre (Kp = Ke = 1000, rp = 0.1, re = 0.5, à l’issu de 1000 pas de temps) en fonction du taux d’exploitation E (proportion d’individus prélevés dans l’habitat exploité) et du biais de perception du risque b (pourcentage de sous-estimation). Np : nombre d’individus à l’équilibre dans l’habitat protégé ; Ne : nombre d’individus à l’équilibre dans l’habitat exploité ; Flux : nombre d’individus transitant de l’habitat protégé vers l’habitat exploité de façon à respecter l’isodar biaisé à l’équilibre; ENe : prélèvement réalisé à l’équilibre. Les valeurs des paramètres sont codées en niveau de gris (0=blanc, gris foncé=valeur maximale) et les lignes noir représentent les isoclines. 193 DISCUSSION Dans ce travail, l’IFD fait office de modèle théorique de la réponse comportementale des proies au risque de prélèvement (effet refuge). Cet effet pourrait néanmoins provenir de mécanismes différents, comme l’utilisation optimale des patchs à l’échelle individuelle (Giving Up Density, Brown 1988, Kotler et al. 1991, Brown 1999, Hochman and Kotler 2007, van der Merwe and Brown 2008) ou encore l’allocation optimale du temps dans le refuge (Risk Allocation Hypothesis, Sih 1992, Cooper 1998 , Lima and Bednekoff 1999, Martin and Lopez 1999b, Cooper 2009, Martin et al. 2009a). Suivant le système biologique (espèce, densité, environnement, échelle de la réserve), d’autres mécanismes que l’IFD peuvent donc être à même de modéliser l’effet refuge. Néanmoins, le compromis énergie-risque (Houston et al. 1993) devrait toujours impliquer une fréquentation moindre des zones exploitées lorsque le risque augmente, affectant ainsi les prélèvements dans le cas où ce phénomène est important. Sous l’hypothèse d’une réelle IFD et lorsque le milieu n’est pas saturé, nous observons en effet que l’augmentation du taux de prélèvement accroît l’utilisation du refuge et réduit à terme le prélèvement (Fig 8.2 & 8.3). Dans ce cas, seul un taux de prélèvement raisonné devrait permettre de rétablir le prélèvement en laissant sortir suffisamment d’individus du refuge (Lundberg and Jonzen 1999). Les possibilités de contrôle de la population de la zone exploitée et du refuge deviennent en revanche bien plus importantes lorsqu’un biais de perception du risque existe chez les individus. Pour une proie, la perception du risque peut parfois être inexacte, voire inexistante (Bouskila and Blumstein 1992, Sih 1992, Koops 2004), surtout si le risque n’est pas manifeste au moment du choix de l’habitat. Les proies pourraient donc être amenées à fréquenter les zones risquées plus que dans l’idéal si elles sous-estiment le risque. Dans ce cas, le refuge devrait générer un système puits-source (Pulliam 1988) puisque le refuge devrait alimenter en individus les zones risquées adjacentes, dans lesquelles ces proies tendent à disparaître (Remes 2000, Gilroy and Sutherland 2007). Nous observons dans les 194 simulations qu’une sous-estimation du risque de la part des individus crée effectivement un système source-puits à l’équilibre entre la réserve et sa périphérie (Fig. 8.5 & 6). Un biais de perception du risque aboutit au fait que l’IFD réelle est perçu comme un déséquilibre apparent que les individus tendent à rétablir. Le biais de perception provoque donc une déviation de l’isofitness, aboutissant à un système puits-source stable (Pulliam 1988, Remes 2000, Gilroy and Sutherland 2007). Réduire la perception du risque chez l’espèce exploitée, devrait donc permettre de contrôler la densité à l’équilibre dans le refuge (voire de supprimer la population du refuge dans certaines conditions) et d’augmenter les prélèvements par rapport à la situation idéale (Fig. 8.5 & 8.6). L’information peut donc être un concept clef pour la gestion des espèces animales (Whittingham et al. 2006, Blackwell and Seamans 2009) surtout lorsqu’elles disposent de moyen d’échapper à l’exploitation. En dissociant le dérangement du risque, les gestionnaires pourront « aplanir » le paysage de la peur des individus, de façon à rééquilibrer leur distribution, tout en maintenant un paysage du risque structuré du fait de la présence de zones protégées. RÉFÉRENCES Abrahams, M. V. and Dill, L. M. 1989. A determination of the energetic equivalence of the risk of predation. — Ecology 70: 999-1007. Blackwell, B. F. and Seamans, T. W. 2009. Enhancing the Perceived Threat of Vehicle Approach to Deer. — Journal of Wildlife Management 73: 128-135. Bouskila, A. and Blumstein, D. T. 1992. Rules of thumb for predation hazard assessment - predictions from a dynamic-model. — American Naturalist 139: 161-176. Brown, J. S. 1988. Patch use as an indicator of habitat preference, predation risk, and competition. — Behavioral Ecology and Sociobiology 22: 37-47. 195 Brown, J. S. 1999. Vigilance, patch use and habitat selection: Foraging under predation risk. — Evolutionary Ecology Research 1: 49-71. Cooper, W. E. 1998 Risk factors and emergence from refuge in the lizard Eumeces laticeps —Behaviour 135: 1065-1076. Cooper, W. E. 2009. Theory successfully predicts hiding time: new data for the lizard Sceloporus virgatus and a review. — Behavioral Ecology 20: 585-592. Fretwell, S. D. 1972. Populations in a seasonal environment. — Princeton University Press. Gerber, L. R. et al. 2003. Population models for marine reserve design: A retrospective and prospective synthesis. — Ecological Applications 13: 47-64. Gilroy, J. J. and Sutherland, W. J. 2007. Beyond ecological traps: perceptual errors and undervalued resources. — Trends in Ecology and Evolution 22: 351-356. Hochman, V. and Kotler, B. P. 2007. Patch use, apprehension, and vigilance behavior of Nubian Ibex under perceived risk of predation. — Behavioral Ecology 18: 368374. Houston, A. I. et al. 1993. General results concerning the trade-off between gaining energy and avoiding predation. — Philosophical Transactions: Biological Sciences 341: 375-397. Koops, M. A. 2004. Reliability and the value of information —Animal Behaviour 67: 103-111. Kotler, B. P. et al. 1991. Factors affecting gerbil foraging behavior and rates of owl predation. — Ecology 72: 2249-2260. Lima, S. L. and Bednekoff, P. A. 1999. Temporal Variation in Danger Drives Antipredator Behavior: The Predation Risk Allocation Hypothesis. — The American Naturalist 153: 649-659. Lundberg, P. and Jonzen, N. 1999. Spatial population dynamics and the design of marine reserves. — Ecology Letters 2 129-134. Mangel, M. 1998. No-take areas for sustainability of harvested species and a conservation invariant for marine reserves. — Ecology Letters 1: 87-90. 196 Martin, J. and Lopez, P. 1999. When to come out from a refuge: risk-sensitive and statedependent decisions in an alpine lizard. — Behavioral Ecology 10: 487-492. Martin, J. et al. 2009. Temporal patterns of predation risk affect antipredator behaviour allocation by Iberian rock lizards. — Animal Behaviour 77: 1261-1266. Moody, A. L. et al. 1996. Ideal free distributions under predation risk. — Behavioral Ecology and Sociobiology 38: 131-143. Morris, D. W. 1987. Spatial scale and the cost of density-dependent habitat selection. — Evolutionary Ecology 1: 379-388. Morris, D. W. 2003. Toward an ecological synthesis: a case for habitat selection. — Oecologia 136: 1-13. Morris, D. W. 2006. Moving to the ideal free home. — Nature 443: 645-646. Neubert, M. G. 2003. Marine reserves and optimal harvesting. — Ecology Letters 6: 843–849. Pulliam, H. R. 1988. Sources, Sinks, and Population Regulation. — The American Naturalist 132: 652-661. Remes, V. 2000. How can maladaptive habitat choice generate source-sink population dynamics? — Oikos 91: 579-582. Schaefer, M. B. 1954. Some aspects of the dynamics of populations important to the management of marine fisheries. — Bulletin of Inter-American Tropical Tune Commission 1: 27-56. Shochat, E. et al. 2005. Ecological traps in isodars: effects of tallgrass prairie management on bird nest success. — Oikos 111: 159-169. Sih, A. 1992. Prey uncertainty and the balancing of antipredator and feeding needs. — American Naturalist 139: 1052-1069. van der Merwe, M. and Brown, J. S. 2008. Mapping the Landscape of Fear of the Cape Ground Squirrel (Xerus inauris). — Journal of Mammalogy 89: 1162-1169. Verhulst, P. F. 1845. Recherches mathématiques sur la loi d'accroissement de la population. — Nouveaux Mémoires de l'Académie Royale des Sciences et BellesLettres de Bruxelles 18: 1-42. 197 Whittingham, M. J. et al. 2006. Altering perceived predation risk and food availability: management prescriptions to benefit farmland birds on stubble fields. — Journal of Applied Ecology 43: 640-650. 198 199 200 9. SYNTHESE 201 9.1. RESUMES ETENDUS PARTIE 1 : SUR L’UTILISATION DE LA TELEMETRIE DANS LES ETUDES DE SELECTION D’HABITAT Martin J., V. Tolon, B. Van Moorter, M. Basille & C. Calenge Comprendre les relations entre un organisme et son environnement est une question centrale de l’écologie. La sélection d’habitat par les individus représente un aspect fondamental de ces relations, puisqu’elle détermine directement les conditions environnementales rencontrées par les individus. Lorsqu’elle est étudiée in situ, la mesure de la sélection d’habitat correspond souvent à la description statique des patrons résultant du processus de sélection par les individus. Néanmoins, le processus de sélection d’habitat est par nature dynamique, puisqu’il résulte du mouvement des individus et comporte ainsi une dynamique temporelle intrinsèque. La considération du mouvement apparaît donc comme incontournable lorsque l’on s’intéresse à la thématique de la sélection d’habitat. Plus précisément la prise en compte du mouvement permet de lier les aspects proximaux aux aspects ultimes de la sélection d‘habitat. Dans cette optique, l’apparition du Global Positioning Système dans le domaine de l’écologie au début des années 1990 a facilité l’acquisition de données permettant la localisation très fréquente d’individus sur de longues périodes de temps. D’un point de vue statistique, ce changement de type de données entraîne une augmentation de l’autocorrélation temporelle dans la position des localisations successives. Les méthodes classiques de sélection d’habitat reposent néanmoins sur l’hypothèse d’une indépendance entre les localisations. L’autocorrélation a donc souvent été ignorée, supprimée ou contournée par différentes procédures. Néanmoins, de nouvelles méthodes ont récemment permis d’intégrer et d’utiliser cette autocorrélation pour développer une approche plus dynamique de la sélection d’habitat. Dans un premier temps, le processus de mouvement a été intégré en considérant les variations de la disponibilité des habitats 202 en fonction de la position initiale des individus, ou en utilisant des trajets de marche aléatoire corrélée pour estimer l’habitat disponible. Plus récemment, de nouvelles méthodes ont permis d’intégrer la structure temporelle du mouvement résultant des variations comportementales des individus (ex : alternance de phases de recherche locale intensive et de déplacements longs) dans la mesure de la sélection d’habitat. L’intégration progressive de l’autocorrélation dans les méthodes d’analyse de la sélection d’habitat s’accompagne donc d’une transition conceptuelle en passant d’une vision statique à une vision plus dynamique de ce processus. L’utilisation de la technologie du GPS dans une optique de gestion ou de conservation de la faune sauvage reste néanmoins plus souvent motivée par ses avantages technologiques, qui permettent de produire d’importantes quantités de données, que par des questions biologiques sous-jacentes. La compréhension des mécanismes proximaux de la sélection d’habitat nécessite souvent l’étude détaillée du mouvement par un suivi intensif des individus, justifiant ainsi l’utilisation de la technologie GPS. Toutefois, les questions relatives aux causes plus ultimes de la sélection d’habitat (ex : effet sur la fitness des individus) ne nécessitent pas forcement un échantillonnage aussi fréquent, mais requièrent un revanche le suivi d’un grand nombre d’individus pendant toute la durée de leur vie. Dans ce cas, l’utilisation de technologies plus anciennes, moins précises mais aussi moins chères pour l’utilisateur (ex : Very High Frequency), peut s’avérer plus pertinente que celle du GPS. Nous souhaitons à travers ce chapitre rappeler que c’est avant tout la question biologique initiale qui doit déterminer la méthode d’échantillonnage et donc la technologie employée, et non l’inverse. Nous encourageons ainsi les utilisateurs de ces technologies à effectuer une réflexion biologique approfondie à propos des caractéristiques écologiques de l’espèce étudiée et de l’échelle spatiale et temporelle à laquelle il sera possible de répondre à la question initiale, avant de choisir l’outil à employer pour suivre les animaux dans leur environnement. 203 PARTIE 2 : UTILISATION DE L’ESPACE D’UNE ESPECE CHASSEE DANS UN PAYSAGE DE LA PEUR FLUCTUANT. Tolon V., S. Dray, A. Loison, A. Zeileis, C. Fischer & E. Baubet. Introduction : Les prédateurs génèrent un « paysage de la peur » où les proies peuvent minimiser le risque de prédation en sélectionnant les zones de faible risque. Suivant la structure spatiale de ce « paysage », i.e. à large ou fine échelle, les proies pourraient répondre à un risque accru respectivement par des déplacements de domaines vitaux ou des redistributions à fine échelle au sein de ces domaines vitaux. Pour les espèces exploitées par l’homme, la présence de zones protégées devrait induire de larges structures spatiales du risque. Nous avons étudié comment des sangliers (Sus scrofa Linnaeus, 1758) répondaient aux variations temporelles du risque dans des zones de chasses (habitat risqué) environnant une réserve naturelle (habitat refuge). Les individus pour qui la présence de la réserve crée de larges contrastes de risque au sein de leur habitat disponible devraient exprimer les plus grands déplacements saisonniers de leurs centres d’activité (P1). A proximité de la réserve, les variations journalières du risque induites par les activités humaines devraient créer une dynamique circadienne de l’utilisation de la réserve (P2). La forte augmentation du risque diurne créé lors de la saison de chasse devrait pousser les animaux encore plus profondément dans la réserve le jour (P3). Néanmoins, ces déplacements saisonniers ne devraient pas être observables sur l’occupation de l’espace nocturne puisque le risque nocturne n’est théoriquement soumis à aucune variation saisonnière (P4). Méthodes: Dans la partie française du bassin genevois, la chasse s’effectue entre septembre et janvier, uniquement le jour. Entre 2003 et 2007, 40 sangliers ont été équipés de colliers émetteurs VHF et ont été localisés une fois par jour lors de leur phase de repos, et suivis de façon continue trois fois par mois la nuit pendant leur phase d’activité. La position relative à la réserve (PRR, distance à la plus proche bordure) de leur centre d’activité a été mesurée pour les deux périodes (jour/nuit) au sein des deux saisons de 204 chasse (close/ouverte). L’attractivité de la réserve a été étudiée avec un modèle linéaire prédisant la PRRouverte en fonction de la PRRclose. Nous avons ensuite appliqué une procédure d’estimation de changement de structure de modèle linéaire en fonction de la PRRclose afin de déterminer le seuil au-delà duquel la réserve n’attire plus les individus. Dans un deuxième temps, nous avons étudié les variations de la taille des domaines vitaux, du recouvrement moyen, du recouvrement des forêts et de l’utilisation de la réserve en fonction de la saison, de la période et du groupe (réserve/autre, d’après le seuil estimé précédemment). Résultats: Des changements significatifs de structure sont détectés dans les modèles linéaires de la PRRouverte en fonction de la PRRclose. Ces modèles indiquent une quasiabsence d’effet de la réserve au-delà de certains seuils (seuil diurne: 1207-2123 m, seuil nocturne: 975-1087 m) et une attraction en deçà. Sur toute la zone d’étude les domaines vitaux sont plus grands la nuit que le jour et plus petits en saison de chasse qu’en saison close. Loin de la zone protégée, les recouvrements des domaines vitaux ne varient pas quelle que soit la période ou la saison. En revanche, l’utilisation des forêts diminue la nuit par rapport au jour et augmente pendant la saison de chasse. A proximité de la réserve, l’utilisation des forêts, le recouvrement et l’utilisation de la réserve diminuent la nuit par rapport au jour. En saison de chasse, le recouvrement et l’utilisation de la réserve augmentent de façon générale alors que l’utilisation des forêts diminue. Discussion: Nos résultats suggèrent que les animaux dont le domaine vital est “en contact” avec la zone protégée en saison de chasse ouverte sont les seuls à exprimer des déplacements saisonniers vers la réserve (P1 supportée). Cette réponse différentielle pourrait résulter de la limitation spatiale du champ de perception du risque, probablement restreint au domaine vital des individus. A proximité de la réserve, les sangliers répondent à l’augmentation du risque diurne et du risque saisonnier par un déplacement de leur centre d’activité dans la zone protégée (P2 et P3 supportées). Malgré l’absence de variation saisonnière du risque nocturne, l’effet du déplacement saisonnier diurne semble se répercuter sur l’utilisation de l’espace nocturne (P4 rejetée). Ce déplacement 205 saisonnier réduit l’utilisation des forêts (surtout présentes à l’extérieur de la réserve) alors que ces habitats offrent protection et nourriture (fruits forestiers) pendant cette saison et sont préférentiellement utilisés par les animaux vivant loin de la réserve. Ceci confirme la prédominance du paysage de la peur pour expliquer l’utilisation de l’espace par des espèces soumises à une forte pression de prédation. Nos résultats suggèrent que la réponse des individus aux variations spatio-temporelles du risque en termes d’utilisation de l’espace dépend de l’échelle et varie suivant la structure sous-jacente du paysage de la peur. 206 PARTIE 3 : IMPACT DES REFUGES SUR L’INTERACTION SPATIALE PREDATEUR-PROIE ET LA MORTALITE DES PROIES : L’EXPLOITATION DE LA FAUNE SAUVAGE COMME CAS D’ETUDE. Tolon V., J. Martin, S. Dray, A. Loison, C. Fischer & E. Baubet. Introduction: Afin de réduire le risque de prédation, les proies tendent à dissocier leur utilisation des habitats de celle du prédateur. En revanche, celui-ci a souvent intérêt à ajuster sa distribution sur celle des proies. Sur le long terme, cette interaction peut entraîner une association ou une dissociation spatiale des deux protagonistes, ce qui peut avoir d’importantes conséquences en termes de mortalité pour la proie. Dans ce contexte, lorsque les prédateurs ne peuvent accéder à un habitat, celui-ci devient pour les proies un refuge permettant la dissociation spatiale et l’augmentation de leur survie. L’influence de cette dissociation sur la survie des proies devrait néanmoins dépendre de la préférence du prédateur pour certains types d’individus (ex : grands ou petits, mâles ou femelles, jeunes ou adultes). Si les refuges réduisent la vulnérabilité de tous les individus d’une population, l’effet de la dissociation spatiale sur la mortalité devrait surtout apparaître pour le type de proies préféré par les prédateurs. Nous testons cette hypothèse dans un système chassé, en considérant le sanglier Sus scrofa L. comme une proie, et les chasseurs ainsi que leurs chiens de chasse comme des prédateurs, ceci à différentes distances d’une zone protégée (refuge). Dans le système étudié, les chasseurs expriment une forte sélectivité pour les gros individus. Nous prédisons d’abord une augmentation de la dissociation spatiale des deux protagonistes à mesure que la proximité de la zone protégée augmente (P1). Nous prédisions ensuite que cette dissociation spatiale provoque une baisse de mortalité pour les classes d’individus habituellement préférées par les chasseurs (P2), mais que cet effet devrait être moins visible sur les classes moins nettement préférées. Matériel et méthode: Dans la partie française du basin genevois, 47 sangliers ont été régulièrement localisés le jour pendant plusieurs saisons de chasse (N = 41 VHF, 6 GPS 207 suivis entre 2003 et 2007). Pendant les saisons 2006 et 2007, des chiens de chasse ont été suivis par GPS pendant 105 battues, nous permettant de quantifier l’effort de chasse dans les habitats. Nous avons réalisé une analyse K-select afin d’étudier les patrons de sélection d’habitat des sangliers au sein de leurs domaines vitaux (variables : pente, altitude, paysage, distance aux sentiers, distance aux routes, distance à la zone protégée). Nous avons ensuite exploré a posteriori les patrons de sélection d’habitat des chiens de chasse au sein de chaque domaine vital des sangliers. Nous avons finalement modélisé la mortalité des sangliers (modèle des hasards proportionnels de Cox) en fonction de l’âge, du sexe et d’une mesure de la dissociation des patrons de sélection d’habitat des chiens de chasse et des sangliers. Résultats: Lorsque les sangliers sont établis près de la zone protégée, ils sélectionnent préférentiellement les forêts hygrophiles localisées dans cette zone, contrairement aux chiens de chasse locaux qui restent en périphérie de la réserve. Lorsque les sangliers sont établis dans des zones montagneuses non protégées, ils intensifient leur sélection pour les forêts de chênes et de hêtres situées dans de fortes pentes et loin de toute voie d’accès pour l’homme. En revanche, les chiens de chasse locaux montrent des patrons de sélection très similaires a ceux des sangliers. Alors que cette dissociation de l’utilisation des habitats aboutit bien à une diminution de la mortalité pour les adultes, nous avons observé parallèlement qu’elle aboutit aussi à une augmentation de la mortalité pour les classes plus jeunes. Discussion: Nos résultats supportent le fait qu’une forte interaction spatiale existe ente les « prédateurs » (chiens de chasse) et leurs « proies » (sangliers). A mesure que les prédateurs rencontrent une contrainte sur leurs mouvements (proximité de la réserve), nous observons effectivement une dissociation progressive de leur patron de sélection d’habitat vis-à-vis des proies (P1 supportée). En réduisant vraisemblablement la probabilité de rencontre avec le prédateur, cette dissociation permet aux individus habituellement sélectionnés par les prédateurs (les adultes) d’accroître leur survie (P2 supportée). En revanche, cette dissociation aboutit à un report de la mortalité adulte sur 208 les classes plus jeunes. Ainsi, le comportement d’évitement des individus les plus sujets à la prédation provoque de façon indirecte une augmentation de la mortalité chez les individus habituellement protégés du fait de la faible préférence qu’a pour eux le prédateur. Ces résultats mettent en lumière comment de simples contraintes sur les mouvements des prédateurs par rapport ceux de leurs proies peuvent déterminer le résultat à long terme de leur interaction spatiale et entraîner une profonde modification de leur dynamique numérique. 209 PARTIE 4 : SE DEPLACER DANS LE PAYSAGE DU RISQUE: LE ROLE DES REPONSES DIRIGE ET NON-DIRIGEES DANS LA SELECTION D’HABITAT. Tolon V., B. Van Moorter, S. Dray, A. Loison, C. Fischer &. E. Baubet. Introduction: Pour sélectionner leur habitat, les animaux peuvent simplement modifier les propriétés intrinsèques de leurs mouvements en réponse aux conditions environnementales immédiates (mouvements non-dirigés) mais peuvent également se déplacer en direction des habitats visés (mouvements dirigés). Les mouvements dirigés impliquent néanmoins de pouvoir disposer d’une perception suffisante des habitats environnants. En réponse aux variations spatiales du risque de prédation, les proies pourraient donc simplement augmenter leur déplacement en réponse à la présence immédiate du prédateur (réponse non-dirigées), et/ou se déplacer directement dans les habitat de moindre risque si elle dispose d’une perception suffisante de ses variations spatiales (réponse dirigée). Nous étudions, par une approche individu-centré, les mouvements interjournaliers du sanglier Sus scrofa L. entre ses sites de repos en fonction du paysage du risque créé par les activités de chasse. Nous prédisons que les animaux devraient sélectionner directement (réponse dirigée) les zones les moins chassées (P1) et les plus éloignés des routes (P2, chasse et collisions sont les deux sources principales de mortalité). Nous prédisons également que les individus devraient augmenter leur distance de déplacement en fonction du niveau de dérangement par la chasse rencontré sur le site de repos initial (P3, réponse non dirigée). Nous prédisons finalement que l’intensité des réponses au risque de chasse sera plus manifeste les jours de chasse puisque les animaux ne sont pas confronté directement à la source du dérangement les autres jours (P4). Matériel et méthodes: Dans la partie française du basin genevois, les sites de repos diurnes de 26 sangliers vivant autour d’une zone protégée ont été régulièrement localisés pendant plusieurs saisons de chasses (N = 22 VHF, 4 GPS suivis entre 2003 et 2007). Pendant les saisons 2006 et 2007, des chiens de chasses ont été suivis par GPS pendant 105 battues, nous permettant de quantifier le paysage du risque autour de la réserve. Nous 210 avons ensuite utilisé une fonction de sélection de pas (Step Selection Function, SSF) pour étudier comment et où les sangliers déplaçaient leur site de repos d’un jour à l’autre, ceci en fonction de la présence de foret, de la distance aux routes et du paysage du risque. Pour chaque déplacement journalier nous avons contrasté les propriétés du pas observé (habitats, distance parcourues) de celles de 199 pas simulés à partir du même point de départ. La SSF nous permet alors de quantifier de façon simultanée l’influence de l’habitat initial sur la distance parcourue (réponse non-dirigée) et celle l’habitat environnant sur la probabilité d’observer le point d’arrivé (réponse dirigée). Résultats: La probabilité de sélection d’un pas est en générale plus forte lorsque le point d’arrivé est situé en forêt, en zone faiblement chassée et loin des routes. La distance parcourue n’est en revanche pas significativement influencée par l’intensité de la chasse sur le site de repos initial. Nous n’observons également aucune différence significative dans l’intensité de ces réponses entre les jours chassés et ceux non-chassés, c.-à-d. la présence ou l’absence réelle de la source du risque. Discussion: Nos résultats montrent clairement que les sangliers rétablissent leur sites de repos directement dans les forêts et dans les zones éloignées des routes (P2 supportée) et faiblement chassées (P1 supportée). Il n’apparaît une en revanche aucun effet du risque de chasse mesuré sur le site de repos initial, sur la distance à laquelle les animaux établissent leur prochain site de repos (P3 non supportée). La réponse au paysage du risque est donc essentiellement dirigée, ce qui pourrait s’expliquer par la forte stabilité du paysage du risque au sein d’une saison de chasse (du fait de la présence d’une zone protégée). La sélection directe de zones faiblement chassées d’un site de repos à l’autre, pourrait résulter de la perception directe par les sangliers des chasseurs et de leur chien pendant les jours de chasse. Nous n’observons néanmoins aucune différence d’intensité de cette réponse lorsque la source du risque est absente de l’environnement (P4 non supportée). Cela suggère que les sangliers utilisent des sources d’informations indirectes (ex : odeur, mémorisation, processus sociaux) de façon à estimer le niveau de risque dans l’espace. 211 212 9.2. IMPLICATIONS POUR LA GESTION 9.2.1. Synthèses des résultats La mise en parallèle des travaux effectués dans le cadre de cette thèse montre bien comment la considération simultanée des mécanismes proximaux et des causes ultimes des comportements (cf. Fig. 1.6) permet de développer de nouvelles approches du comportement de sélection d’habitat dans le paysage du risque. Ceci est d’autant plus important que ce travail s’inscrit dans une optique appliquée destinée à améliorer la gestion du sanglier lorsqu’il dispose de refuge telles que les zones protégées. Nous observons effectivement que les animaux sont extrêmement sensibles aux variations spatiales du risque créées par la réserve. Dès qu’elle apparaît dans leurs domaines vitaux, les animaux la rejoignent lorsque le risque augmente, quitte à abandonner des zones offrant d’abondantes ressources (Partie 2). A l’échelle inter-journalière, les animaux ont également la capacité de repositionner, d’un jour à l’autre, leur site de repos directement dans les zones les moins chassées (Partie 4). Ce comportement est probablement rendu possible par la grande stabilité de la structure spatiale du risque crée par la réserve, facilitant sa perception par les animaux. Comme attendu, les individus les plus exposés au risque (les adultes dans notre cas) bénéficient d’une forte augmentation de leur survie lorsqu’ils utilisent la réserve (Partie 3). En revanche, ce même comportement aboutit à un report de mortalité sur les plus jeunes, probablement à cause de leur plus forte sensibilité à la ceinture de risque existante autour de la réserve. Ceci jette un doute sur le caractère réellement adaptatif de ce comportement (si les adultes survivent mieux mais que leur 213 jeunes meurent à leur place la fitness peut ne pas augmenter). Il est probable que la réserve considérée dans ce travail soit finalement trop petite, et sa périphérie trop chassée, pour qu’une augmentation générale de la survie puisse apparaître dans toutes les classes d’âge. Les animaux pourraient donc adopter la meilleure stratégie locale (la réserve ou les fortes pentes) suivant la disponibilité des refuges, sans qu’une stratégie ne soit clairement plus avantageuse. Notre système correspond donc à une situation où la réponse comportementale correspond aux prédictions de l’approche adaptative (les animaux évitent la source du risque en rejoignant un refuge), mais où les bénéfices réels de ce comportement sont remis en question. C’est précisément ce que nous avons exploré de façon théorique dans la Partie 5, où nous montrons que des erreurs de perception du risque pourrait réduire l’effet positif du refuge sur la population qu’il abrite. Dans la partie suivante nous rappelons d’abord en quoi l’information peut être un concept clef pour la gestion des espèces et nous proposons ensuite un ensemble de mesures de gestion destinées à réduire l’effet réserve chez le sanglier. 9.2.2. Réduire l’effet réserve : l’importance de l’information 9.2.2.a. Erreur de perception et système source-puits La réduction de l’effet réserve tel qu’il a été décrit dans ce travail pourrait passer par la création d’un système source-puits entre la réserve (source) et sa périphérie (puits). Beaucoup de populations sont organisées spatialement en systèmes source-puits (Pulliam 1988) au sein desquels des individus transitent depuis des habitats productifs (source : taux de croissance supérieur à 1) vers des habitats improductifs (puits : taux de croissance inférieur à 1). Ainsi, des habitats de mauvaise qualité peuvent être occupés suite aux flux d’individus provenant d’une source voisine. De façon théorique, la Distribution Libre Idéale (IFD, Fretwell 1972) prédit que lorsque des individus sélectionnent leur habitat de façon à maximiser leur fitness individuelle, la distribution résultante devrait être celle correspondant à la situation d’égalité des fitness (isofitness, Morris 1987, 2006). Par 214 nature, les systèmes source-puits ne respectent pas l’IFD puisque certains individus quittent des habitats productifs pour aller occuper des habitats ne pouvant assurer la persistance d’une population (réduisant ainsi considérablement leur fitness par rapport aux individus de la source). L’émergence et le maintient des systèmes source-puits implique donc l’existence de certaines contraintes sur les mouvements des individus, capables de les faire dévier du comportement idéal (Remes 2000). Plusieurs auteurs considèrent que les erreurs d’estimation de la qualité d’habitat par les individus sont les principales causes de l’émergence et du maintien à long terme des systèmes source-puits (Kristan 2000, Remes 2000, Delibes et al. 2001a, Delibes et al. 2001b). Les individus sont en effet supposés se distribuer seulement d’après les informations dont ils disposent, et l’erreur faite sur la perception de la qualité des habitats n’est jamais nulle même si elle tend à être réduite (Dall et al. 2005). Si la distribution des individus tend vers une IFD, une erreur de perception les fera dévier de la distribution idéale, aboutissant à un déséquilibre des fitness. Or ce déséquilibre est justement une propriété des systèmes source-puits, qui se stabilisent à terme par un transfert d’individus des habitats à forte fitness vers les habitats à faible fitness. La création d’un biais de perception pourrait donc être une des clefs permettant de répondre aux problématiques de gestion créées par l’effet réserve. Cela reviendrait pour les gestionnaires à dissocier le risque réel des stimuli de dérangement. 9.2.2.b. Risque et dérangement Au delà du risque réel de mortalité due à l’exploitation par l’homme, le dérangement est un important facteur déterminant l’utilisation des habitats et le comportement des individus (ex: Blumstein et al. 2003, Blumstein et al. 2005). Le dérangement peut être défini comme l’ensemble des stimuli émis par l’humain et provoquant une modification du comportement des animaux. Ces stimuli peuvent être associés à un risque réel (ex : chasse) ou non (ex : tourisme). Certains auteurs avancent l’hypothèse selon laquelle les animaux réagiront de la même manière face au dérangement créé par l’humain que face à un risque de prédation même si le risque réel 215 n’est pas véritablement présent (risk-disturbance hypothesis, Frid and Dill 2002, Gavin and Komers 2006, Jayakody et al. 2008). Ceci provient du fait que les animaux tendent à réagir à des stimuli généraux comme des bruits soudains ou une approche rapide pour éviter la prédation. L’hypothèse de l’équivalence entre risque réel et dérangement n’est en revanche pas optimale puisqu’elle peut impliquer des réponses anti-prédatrices inutiles si le risque réel n’existe pas. Peu d’espèces pourraient néanmoins être capable de dissocier de façon fiable les stimuli de dérangement seul de ceux impliquant un risque réel (ex: primates, Bshary 2001, Nevin and Gilbert 2005). La création de dérangement sans risque, ou de risque sans dérangement pourrait donc permettre de modifier le comportement des individus de façon à atteindre l’objectif de gestion sans qu’ils puissent le contourner. 9.2.2.c.Gestion par l’information Il peut être intéressant pour les gestionnaires de pouvoir moduler l’association entre le paysage du risque et le paysage de la peur en jouant sur la perception du risque par les individus. La gestion de l’ouverture de la végétation peut par exemple, influencer la distribution, et à terme l’impact écologique, des individus simplement parce qu’elle détermine le niveau de risque perçu par les animaux (Schmitz 2005, Whittingham et al. 2006). Dans certains cas il devient même nécessaire d’ajouter de l’information sur le risque de façon à rétablir sa perception et éviter les problèmes de gestion (ex: collisions routières, Blackwell and Seamans 2009). La prise en compte de la perception du risque chez les animaux a finalement toujours existé dès lors que l’homme a eu un intérêt à gérer ou à exploiter la faune sauvage. Le camouflage, l’affût, l’appât, le leurre, l’effarouchement, la dissuasion sont autant de moyens de biaiser la perception du risque, afin d’éviter ou d’accentuer la fréquentation d’un site par les individus et d’atteindre l’objectif de gestion (fuite ou maintien dans une zone, exploitation et/ou conservation). L’information peut donc constituer un important moyen de gestion des espèces. Ceci pourrait permettre de réduire la concentration d’individus dans les zones protégées lorsque le risque augmente à l’extérieur, et donc potentiellement d’augmenter le 216 rendement de l’exploitation ainsi que réduire les effets écologiques non souhaités. Ainsi, via la création d’un système source-puits, il deviendrait possible de gérer la population dans et en dehors du refuge. Les gestionnaires devraient alors retrouver un effet réserve, n’aboutissant pas cette fois à la réduction des prélèvements via le comportement d’évitement du risque par les proies, mais à une augmentation des prélèvements et des possibilités de gestion de la population dans le refuge grâce au rôle de source tenu par la réserve 9.2.3. Implications pour la gestion du sanglier 9.2.3.a. Mesures de gestion potentielles Nos résultats montrent que le sanglier a la possibilité de s'extraire de façon efficace de la pression de chasse dès que celle-ci présente une forte structuration spatiale. Nous observons que les sangliers favorisent pendant la saison de chasse, l'utilisation des zones de moindre risque plutôt que l'utilisation des forêts pourtant riches en fruits forestiers (Partie 2). Ainsi, le paysage du risque apparaît comme prépondérant pour expliquer la distribution du sanglier. Cela signifie qu'il ne sera probablement pas possible de rééquilibrer les distributions entre les zones refuges et les zones chassées uniquement par le nourrissage dissuasif. Les gestionnaires de la faune sauvage ont donc tout intérêt à modifier la structure spatiale du risque et/ou jouer sur la perceptibilité du risque chez les animaux concernés, de façon à regagner des possibilités de gestion efficace des populations. Dans cette optique nous pouvons dorénavant lister quelques propositions de gestion ayant toutes pour but d’accroître l’exposition des individus au risque de prélèvement. Il est important de rappeler que les mesures de gestion présentées ci dessous sont issues d’une approche purement scientifique destinée à réduire l’effet réserve, mais que leur mise en œuvre nécessiterait néanmoins la prise en compte du contexte écologique, social et politique inhérent à toute problématique de gestion de la faune sauvage (voir paragraphe suivant). 217 Réduction de l’effet réserve chez le sanglier. - 1. Aplanir le paysage du risque. o 1.1 Lorsque aucune zone protégée n’existe, favoriser une répartition uniforme de l’effort de chasse dans l’espace de façon à éviter que des zones ne deviennent refuges. o 1.2 Si la chasse dans les refuges est impossible pour des raisons pratiques (zones trop escarpées) ou légales (zones protégées, proximité des habitations) maintenir une pression de chasse à proximité du refuge. Cependant cette pression doit être adaptée façon à éviter que tous les individus ne rejoignent le refuge en réponse à un différentiel de risque trop important entre la réserve et sa périphérie. - 2. Remodeler le paysage du risque. o 2.1 Si une proportion fixe du territoire de chasse doit être protégée et qu’il est possible de choisir les périmètres concernés, favoriser la fragmentation des réserves de façon à accroître "l'effet bordure" et l’exposition des individus au risque périphérique. o 2.2 Créer de l’irrégularité temporelle dans la structuration spatiale du risque en ne chassant pas toujours aux mêmes endroits, quitte à laisser certaines zones "au repos" pendant quelques jours ou quelques semaines. Le but de cette méthode est de réduire la prévisibilité de la structure spatiale du risque dans les zones périphériques chez les animaux. o 2.3 Lorsque cela est possible, favoriser les changements interannuels de localisation des réserves de chasse de façon à créer de la nouveauté dans la structure spatiale du risque. - 3. Biaiser la perception de la qualité des habitats. o 3.1 Favoriser les méthodes de régulation ne créant pas trop de dérangement mais gardant tout de même une bonne efficacité en terme de gestion 218 des populations (ex: tirs de régulation nocturnes sur le canton de Genève) de façon à réduire la perception du risque dans les habitats ou il se produit. o 3.2 Créer des stimuli de dérangement dans les zones refuges lorsque la chasse est interdite de façon à accroître le risque perçu dans cet habitat (ex : effarouchement). o 3.3 Modofier les habitats de façon à réduire le niveau de sécurité perçu par les animaux dans les refuges (ex: ouvrir le paysage). o 3.4 Inversement, modifier les habitats de façon à augmenter le niveau de sécurité perçu dans les zones chassées (ex: laisser des broussailles). o 3.5 Parfois, lorsque cela est possible, changer de méthode de régulation pour réduire les possibilités d’apparition d’une contre réponse chez les individus via l’apprentissage ou la sélection naturelle (Ex : éviter que les animaux ne quittent le périmètre de battues dès qu’ils perçoivent l’arrivé des chasseurs). 9.2.3.b. Importance des contextes écologique et humain Ces mesures de gestion doivent ensuite, pour être appliquées, être considérées dans un cadre plus général dépendant du contexte biologique (ex : taille de la réserve, habitat, densité), des enjeux de conservation (présence d’espèces ou d’habitats protégés dans la réserve) et du contexte humain (pratiques de chasse, politiques de gestion, contexte social). L’ensemble de ces mesures de gestion sont destinées à accroître les chances que les animaux s’établissent (par erreur) dans la zone périphérique de la réserve et subisse la pression de chasse. Si les réserves sont de taille inférieure à la taille des domaines vitaux des sangliers (<10-30 km2), il est raisonnable de penser que les animaux puissent se déplacer entre la réserve et sa périphérie. Mais au-delà de cette taille (ex : parcs naturels régionaux), les sangliers établis au centre de la zone protégée ont peu de chance de fréquenter les zones périphériques et leur régulation ne peut dans ce cas qu’être effectuée à l’intérieur du périmètre protégé (ex: battues administratives, piégeages, tirs de nuit, autorisation de chasse exceptionnelle, Maillard 1998). 219 La chasse se fait souvent dans le même environnement que la conservation des espèces protégées. Bien souvent les réserves sont avant tout destinées à la protection des habitats et ne peuvent donc être déplacées (les mesures 1.1, 2.1, 2.3 ne sont plus applicables) ou ne peuvent subir du dérangement ou des modification d’habitat (les mesures 3.2 et 3.3 ne sont plus applicables). Néanmoins, la présence en trop grand nombre de sangliers dans une réserve naturelle peut avoir d’importantes conséquences écologiques sur l’écosystème protégé, nécessitant une intervention directement dans le périmètre en question (ex : battues d’effarouchement). La décision d’agir ou non directement dans la réserve doit donc être un compromis entre l’absence totale de dérangement, risquant d’entraîner une trop forte concentration de sangliers dans la zone protégée (et potentiellement des effets écologiques non-souhaités) et un trop fort dérangement affectant directement les espèces protégées locales. Suivant le contexte humain, variant bien souvent fortement d’un endroit à un autre, certaines mesures de gestion peuvent également devenir inadaptées ou tout du moins difficiles à mettre en œuvre. La régulation du sanglier peut par exemple être effectuée par des chasseurs bénévoles sur des territoires communaux, par des professionnels de l’état ou des fédérations de chasseurs ou encore par des particuliers sur des domaines de chassées privées. Suivant ce contexte, la maîtrise et les modifications des pratiques de chasse peuvent être plus ou moins aisées. De façon plus globale, le contexte légal (ex : réglementation de la chasse), politique (ex : objectif de gestion), culturel (ex : mode de chasse) ou social (place de la chasse dans la communauté) peut fortement conditionner l’applicabilité des méthodes de gestion. Nous pouvons maintenant appliquer ce raisonnement au site d’étude concerné dans ce travail : la réserve de l’Etournel dans le bassin genevois. Cette zone est une réserve naturelle dont il n’est possible de modifier ni le périmètre (exclusion de 1.1, 2.1 et 2.3) ni l’habitat (exclusion de 3.3). La régulation du sanglier est pratiquée sur la zone par des chasseurs bénévoles habitués à chasser en battue (exclusion de 3.1 et 3.5). Au sein de la zone périphérique concernée par l’effet réserve (environ 2km, c.f. Partie 2) nous pouvons donc recommander de maintenir globalement une pression de chasse un peu plus retenue surtout en bordure du refuge (1.2), de créer de l’irrégularité temporelle dans 220 l’application de la chasse autour de la réserve (2.2), de créer du dérangement dans la zone protégée (3.2, méthode déjà mise en place) et de laisser des habitats en périphérie de la réserve offrant de bonnes conditions pour l’établissement de la bauge du sanglier (3.4, ex : végétation très fermée). Ces mesures de gestion ne sauraient néanmoins montrer leur efficacité sans la prise en compte des règles de tirs liés à la structure de la population (ex : classe de taille) pouvant influencer dans un sens ou dans un autre l’accroissement d’une population (Servanty 2007). Nous pouvons souligner l’exemple de la gestion du sanglier dans l’état de Genève, réalisés entièrement par tirs nocturnes effectués par des gardes faunes de l’Etat. S’il est irréaliste pour des raisons légales, politiques et culturelles d’imaginer transposer cette pratique sur la partie française à court terme, nous pouvons souligner que cette méthode pourrait être, d’un point de vue strictement biologique, une bonne alternative pour réduire un éventuel effet réserve chez le sanglier. En effet, lorsque les animaux se nourrissent la nuit, ceux-ci sont souvent contraints de se répartir dans les environs des zones protégées (cf. Partie 2). De plus, nous pouvons faire l’hypothèse que cette méthode crée moins de dérangement qu’une battue et provoquera moins d’effet refuge (cf. Partie 5). Ainsi, si cette méthode permet d’appliquer une pression de chasse suffisante pour réguler les populations de sanglier (il semble que ce soit le cas, Hebeisen 2007) nous pouvons penser qu’elle peut aider à mieux contrôler les populations établies dans les réserves lorsque le contexte humain permet son application. 221 9.3. PERSPECTIVES 9.3.1. La nécessité d’approches expérimentales L’étude et la compréhension des processus de distribution, de sélection d’habitat et des mouvements de la faune sauvage reposent souvent sur l’observation in situ des patrons de distribution. Or, en écologie il devient souvent difficile de comprendre les processus simplement par l’étude des patrons de distribution (Turner 1989) du fait de l’existence de nombreuses variables non-contrôlées, confondues ou même inconnues. Dans ce contexte, la possibilité de développer des approches expérimentales ou semiexpérimentales devient un précieux atout pour mettre à jour ou tester les mécanismes conduisant aux patrons observés. Cepandant, il est bien souvent impossible d’utiliser une approche véritablement expérimentale sur la faune sauvage lorsque celle-ci vit et se déplace sur d’importantes surfaces. Les systèmes chassés offrent néanmoins de précieuses opportunités pour comprendre les mécanismes de distribution, de sélection d’habitat et de mouvements des individus vis-à-vis du risque. Ceux-ci présentent premièrement d’importantes et brusques variations spatiales et temporelles du risque (réserve, période de chasse) qui ne peuvent en général être observées en milieu naturel en présence de prédateurs souvent très mobiles et constamment présents. De plus, lorsqu’il devient possible, en partenariat avec les gestionnaires, d’introduire des variations contrôlées dans le système, de nouvelles voies de compréhension émergent. Le cas de la gestion du sanglier est particulièrement 222 propice à ce type d’expérimentation. Nous exposons ci-dessous un exemple d’approche semi-expérimentale qui pourrait permettre de comprendre les mécanismes de perception du risque chez cette espèce. L’agrainage dissuasif est une pratique de gestion très souvent utilisée pour éviter les dégâts de sanglier sur les terres cultivées (Vassant and Breton 1986, Vassant 1994, Calenge et al. 2004, Geisser and Reyer 2004). D’un point de vue théorique cette pratique consiste à offrir aux animaux des patches de nourriture que ceux-ci exploiterons jusqu’au seuil où ils estiment que le patch n’est plus rentable (ex: Giving Up Density, GUD, Charnov 1976, Brown 1988). Or, c’est précisément par cette méthode que certains auteurs ont mesuré le paysage de la peur, sous l’hypothèse que les animaux laissaient plus de nourriture dans le patch lorsque la peur y était forte (Shrader et al. 2008, van der Merwe and Brown 2008). Les paysages de la peur pourraient donc être facilement estimés chez le sanglier dans des zones relativement homogènes. Mais l’intérêt de cette approche résiderait surtout dans le fait de pouvoir ajouter de façon expérimentale différents stimuli potentiellement liés au risque (ex: Shrader et al. 2008) et d’observer leurs effets sur la GUD. Ces stimuli pourraient par exemple être modulés par le nombre de personnes venant sur les points d’agrainage pendant la journée, par la présence éventuelle de chiens, par l’ouverture du paysage local, par l’utilisation de sang de sangliers chassés récemment ou même par de légers dérangements nocturnes. Il serait également intéressant de mesurer la résilience de la réponse des animaux de façon à comprendre l’impact d’une perturbation sur le comportement ultérieur des individus (ex : persistance du stimulus, mémorisation). Il serait ainsi possible de mettre à jour certains mécanismes de perception du risque et de traduire leurs effets en terme écologique via la GUD et l’utilisation de l’espace. Cette approche montre comment les systèmes chassées ou exploités peuvent fournir un cadre propice au développement des approches semiexpérimentales in situ et comprendre les processus aboutissant aux patrons de distribution observés chez la faune sauvage. 223 9.3.2. Etude des systèmes sans réserve Sur notre site d’étude, la réserve existe depuis 1964 créant ainsi une structure spatiale du risque très stable, donc très prévisible pour les individus. Nos résultats montrent que les mouvements des sites de repos sont fortement dirigés vers les zones les moins chassées (Partie 4). Il serait intéressant d’utiliser cette approche dans un système différent où la structuration spatiale du risque est moins évidente et moins prévisible. Par exemple, sur le territoire d’étude de l’ONCFS de Chateauvillain les sangliers ne disposent pas de zones clairement protégées et utilisent comme refuge des parcelles de fourrés denses tout au long de l’année. Contre toute attente, les animaux réduisent partiellement l’utilisation de ces zones pendant la saison de chasse (Saïd et al. in prep). Ce comportement pourrait être expliqué par deux processus différents mais non-exclusifs. Premièrement, la chasse pourrait intervenir comme un facteur de perturbation aboutissant à l’augmentation des déplacements entre les sites de repos. Naturellement, comme l’essentiel des individus se trouve initialement dans les zones de fourrés, l’augmentation des mouvements aboutit à une réduction de l’utilisation de cet habitat (réponse nondirigée). Deuxièmement, les chasseurs pourraient intensifier la pression de chasse dans les habitats préférés par le sanglier (cf. Partie 3) et la sécurité des fourrés pourrait devenir moins importante qu’attendue. Dans ce cas, les individus devraient diminuer leur sélectivité directe pour cet habitat (réponse dirigée). Ainsi, l’étude des mouvements interjournaliers du site de bauge dans cet environnement, telle qu’elle a été présenté dans ce travail (Partie 4) pourrait aider à comprendre le mécanisme aboutissant à la réduction de l’utilisation de ces zones de fourrés dense alors que le risque augmente pendant la chasse. Ce travail pourrait permettre d’adopter une approche comparative intéressante avec la situation de la réserve étudiée ici dans le bassin genevois. 224 9.3.3. Utilisation des refuges et réponse fonctionnelle Les résultats présentés dans ce travail soulignent l’importance de la position du domaine vital dans le paysage du risque sur les stratégies de sélection d’habitat et la survie des proies. Le domaine vital définit l’espace dans lequel les proies peuvent ajuster leur réponse aux variations spatiales et temporelles du risque. Si les individus ne peuvent modifier l’emplacement de leur domaine vitaux (ex : migration), celui-ci représentera alors une forte contrainte sur les stratégies de sélection d’habitat de la proie puisqu’elle ne pourra qu’apporter une réponse aux conditions locales du risque alors que d’autres opportunités existent potentiellement au delà de son espace perçu et/ou accessible. Nos résultats montrent que la position du domaine vital dans le paysage du risque joue sur l’échelle spatiale de réponse aux variations du risque (Partie 2), sur la possibilité ou non de fréquenter un habitat habituellement favorable (la forêt en période de chasse, Article 2), sur la possibilité réelle de distancer le prédateur lorsque celui-ci n’a pas accès à certains habitats et finalement sur la survie des proies (Partie 3). Pour intégrer ces résultats dans un cadre théorique plus général, nous présentons ci-dessous un raisonnement pour l’instant simplifié, mais qui pourrait être développé plus en profondeur à l’avenir. Nous faisons d’abord le postulat que les proies ne sont pas fortement limitées par les ressources (ex : habitat riche, pas de compétition alimentaire, prédation forte, phases critiques du cycle de vie) et que la mortalité n’est pas densité dépendante. Les proies devraient alors simplement sélectionner sur leurs domaines vitaux, l’habitat leur permettant de maximiser leur survie. Considérons qu’une proie dispose, dans un espace à une dimension, de deux types de refuges, a et b, et du reste de son environnement c, chacun défini par leur probabilité de survie journalière sa et sb et sc. Dans cet exemple, la quantité totale de refuge disponible est constante (soit Da + Db = cte). Nous étudions ici le cas où la proie choisit toujours un seul type de refuge (stratégie unique) pendant une durée T (ex : une saison). En sélectionnant le refuge, les proies font néanmoins des erreurs de localisation qui peuvent aboutir, en bordure de celui-ci, à la fréquentation de 225 l’habitat c (risqué). Considérons ε comme le rayon d’erreur maximal que les individus peuvent réaliser autour du refuge (zone d’erreur fixe). En considérant des refuges d’un seul tenant et des individus ne chevauchant qu’une seule bordure et se distribuant uniformément dans l’espace utilisé (Fig. 9.1), la proportion de temps passé à l’intérieur du refuge (ici a) peut donc être estimé par : (1) Ta = T Da . ε + Da Nous reconnaissons ici l’équation d’une réponse fonctionnelle de type II (Holling 1959). Bien que les individus sélectionnent toujours le refuge, le temps passé à l’intérieur de celui-ci décroît lorsque sa disponibilité décroît puisque la zone d’erreur en bordure du refuge représente une partie de plus en plus importante de l’habitat utilisé (Fig 9.1 & 9.2a). La survie à l’issue de la période T peut ensuite être calculée en multipliant les survies journalières des habitats par elles-mêmes, ce autant de fois que le temps passé dans chaque habitat, soit pour la stratégie a : (2) T S a = sa a sc T −Ta Nous observons logiquement que la survie décroît lorsque la disponibilité du refuge utilisé diminue (Fig 9.2b). L’importance relative de l’erreur par rapport à la quantité de refuge disponible s’accroît lorsque la disponibilité du refuge diminue (Fig. 9.1), ce qui provoque une fréquentation accrue de l’habitat risqué (Fig. 9.2a), aboutissant à une baisse de survie (Fig. 9.2b). Alors que la survie journalière est, dans l’absolu, plus forte dans le refuge a, nous observons que la stratégie b devient plus intéressante en dessous d’un certain seuil de disponibilité de a (ou au dessus d’un certain seuil de b, Fig. 9.2b). Si les individus expriment des stratégies uniques (un seul type de refuge sélectionné), un switch de refuge devrait se produire lorsque la disponibilité du meilleur refuge passe en dessous d’un certain seuil. 226 Da Individu ε Db ε Sa>Sb Environnement a c Da ε Individu b ε Db Sa<Sb Environnement a c b Figure 9.1 : Mécanisme de la réponse fonctionnelle des proies disposant de deux refuges a et b, offrant des probabilités de survie journalière différentes, mais supérieures à l’habitat c (sa>sb>> sc) et dont la disponibilité Da et Db varie. La taille du refuge est fixe de telle sorte que Da + Db = cte. Les erreurs de localisation des individus aboutissent en bordure des refuges à une utilisation de l’habitat risqué c sur un espace fixe défini par ε. Les individus ne choisissent qu’un seul type de refuge pour une durée T. A mesure que la disponibilité d’un refuge diminue, l’effet relatif de l’erreur sur l’habitat utilisé augmente (équation 1), ce qui réduit la survie à l’issue de la période T (équation 2). Ainsi, alors que sa>sb, la stratégie b devient plus intéressante (Sa<Sb) lorsque Da<<Db. Sans aucune erreur de localisation, la meilleure stratégie est évidement de toujours choisir le refuge a (équation 1). Ce modèle simplifié laisse donc entrevoir l’impact que pourrait avoir la qualité de la perception des habitats, les contraintes de mouvements ou d’utilisation de l’espace (ex : taille du domaine vital), ou encore la fragmentation des refuges (les trois pouvant faire varier l’importance de ε) sur la sélection des refuges par les proies et leur survie lorsqu’elles disposent de différents types de refuges. Cette approche pourrait gagner en réalisme en étant développée dans deux dimensions (dans ce cas l’importance de l’erreur dépendrait du ratio périmètre/aire) et en permettant l’existence de stratégies mixtes chez les individus (ex : fréquentation simultanée de a et b). Nos résultats suggèrent qu’une telle réponse fonctionnelle peut exister puisque les animaux basculent de la réserve vers les fortes pentes lorsque la disponibilité de la réserve devient trop faible (les fortes pentes sont considérées comme des refuges secondaires car elles ne permettent pas de distancer le prédateur). Malgré cette réponse, le changement de condition de risque à large échelle s’accompagne d’une baisse de la 227 survie pour les proies les plus exposées au risque de prédation (comme décrit dans la Stratégie b 20 40 60 80 Stratégie a 0 Temps dans le refuge a) 100 figure 9.2b). 0.0 0.2 0.4 0.6 0.8 1.0 0.8 b) a 0.4 Stratégie b 0.0 0.2 Survie 0.6 tégie Stra 0.0 0.2 0.4 0.6 0.8 1.0 Da Figure 9.2 : Effet des variations de la disponibilité relative des refuges a et b (mesuré ici par Da) sur a) le temps passé dans ces refuges et b) la survie à l’issue d’une période T, lorsque les proies adoptent des stratégies uniques. Dans cet exemple Da + Db = 1, T=100, sa=0.999, sb=0.995, sc=0.95 et ε = 0.05. Le point noir indique le seuil de disponibilité du refuge a en dessous duquel les individus ont intérêt à basculer sur le refuge b. Il pourrait donc être intéressant d’aborder l’utilisation des refuges sous l’angle de la réponse fonctionnelle (Holling 1959) en les considérant comme des ressources au sens large du terme (Berryman and Hawkins 2006). Puisque l’erreur (ou l’effet bordure des 228 refuges) est le paramètre faisant ici émerger une telle réponse, ce raisonnement montre en quoi l’intégration des contraintes proximales de la sélection d’habitat (mesurées ici par ε) pourrait avoir un fort impact sur les modalités de réponse et la survie attendue des individus lorsqu’ils disposent de plusieurs types de refuges. 9.4. CONCLUSION De façon à conclure ce travail de thèse, il peut être intéressant de rappeler brièvement son historique, de son émergence à son aboutissement. Ce travail est avant tout né de la confrontation des gestionnaires de la faune sauvage du bassin genevois à une problématique appliquée bien précise (l’effet réserve), nécessitant des mesures de gestion rapidement applicables. Pour ces raisons, une grande variété de données a été récoltée autour de cette problématique dans le but d’apporter ultérieurement des éléments de réponses par une approche scientifique. Cette façon de procéder est différente de la démarche scientifique classique nécessitant de poser une question biologique clairement définie et résultant d’un raisonnement théorique élaboré et éprouvé, avant de récolter les données. Il existe en réalité un important décalage entre l’échelle de temps à laquelle se déroule le processus de production scientifique et l’échelle de temps à laquelle les solutions doivent être appliquées dans le cadre de la gestion de l’environnement (une fois soumis, un article scientifique met environ 18 mois pour être publié dans le domaine de l’écologie, Kareiva et al. 2002). Pour s’adapter à l’urgence des problématiques de gestion de l’environnement, il est donc nécessaire de pouvoir rapidement rétablir le sens habituel de la démarche 229 scientifique tout en considérant les données à disposition. Cet exercice particulièrement difficile peut néanmoins faire émerger des approches intéressantes qui n’auraient pas forcement vu le jour dans un contexte purement fondamental. Ainsi le principal intérêt de cette thèse est probablement son approche multi facettes, motivée par la nécessité d’aborder différents aspects du comportement pour répondre à la problématique de gestion. Il a ainsi été ainsi nécessaire d’étudier différents types de données (habitat, mouvement, survie, utilisation de l’espace) dans différents cadres théoriques (causes ultimes et proximales du comportement) et d’utiliser des approches scientifiques variées (étude de cas, simulations, développements théoriques). Ce type d’approche intégrative permet d’acquérir une vision globale des processus comportementaux et des systèmes écologiques complexes, ce qui a pour intérêt principal d’aider à fournir des réponses adaptées aux problématiques de gestion. D’un point de vue théorique, ce type de démarche intégrative permet également de développer de nouvelles approches en intégrant les contraintes et mécanismes du système écologique dans son évolution (cf. paragraphe 9.3.3). Ces approches sont très prometteuses puisqu’elles offrent l’opportunité de lier la puissance euristique de l’écologie évolutive avec le réalisme des approches mécanistiques (Sih et al. 2004, Dall et al. 2005, Nathan et al. 2008). Il sera encore très probablement nécessaire de poursuivre à l’avenir le développement de ce type de démarche afin d’amélioration de la compréhension des systèmes écologiques complexes dans lesquels intervient le comportement des individus. 230 231 232 Remerciements Je remercie, en tout premier lieu, l’ensemble des membres de mon jury de thèse, notamment Messieurs Stefano Focardi et Mark Hewison, rapporteurs de cette thèse, pour avoir eu la gentillesse de bien vouloir relire et commenter mon travail, ainsi que Mr Claude Miaud pour avoir accepté de présider mon jury de thèse. J’adresse ma plus profonde gratitude à l’ensemble des Fédérations Départementales des Chasseurs de l’Ain et de la Haute Savoie pour leur soutien financier et leur contribution majeure à la réalisation de l’Etude des sangliers du Bassin genevois. En particulier, je remercie Messieurs les directeurs Daniel Rousset (Ain) et Philippe Arpin (Haute Savoie), ainsi que Michel Dunand, Patrick Gaulard (Ain), Jean Louis Ducruet, et Jean Jacques Pasquier (Haute Savoie) pour la confiance qu’ils m’ont accordée tout au long de l’étude. Un merci tout particulier à François Corcelle, Gabriel Lagriffoul, Pascal Fol et Brigitte Combaz pour leur enthousiasme communicatif à participer à cette étude. J’adresse également tous mes remerciements à l’ensemble des chasseurs des fédérations pour leur participation à l‘organisation des captures et du suivi des sangliers. Je remercie grandement, pour avoir permis la réalisation d’une étude transfrontalière, tous nos partenaires Suisses de l’étude, notamment Gottlieb Dandliker, Marc Obermann, Jean-Louis Delabays et tout le personnel du Domaine Nature et Paysage du Département du Territoire de l’Etat de Genève, Sébastien Sachot, Patrick Deleury et l’ensemble du Service des Forêts, de la Faune et de la Nature du Canton de Vaud ainsi que Patrick Durand d’ECOTEC Environnement. J’adresse des remerciements particuliers au Conseil Général de l’Ain pour son soutien financier additionnel ainsi qu’aux Chambres d’Agriculture de l’Ain et de la Haute Savoie, en particulier à Corinne Lavorel pour son active participation au projet. 233 Parce que derrière chaque localisation utilisée dans cette étude se cache la capture d’un animal et des heures entières passées sur le terrain dans des conditions difficiles mais toujours intenses, je me dois d’adresser d’immenses remerciements à tous les pisteurs ayant participé aux captures et aux suivis des sangliers de l’étude, en particulier Jacques Olivier Chappuis, Mathieu Comte et Julien Vasse avec qui j’ai partagé bien plus que des heures terrain. J’adresse mes plus sincères remerciements au personnel du CNERA Cervidés et Sanglier de l’Office National de la Chasse et de la Faune Sauvage, notamment à François Klein, pour m’avoir offert la possibilité de réaliser un thèse aussi enrichissante tant elle m’a permis de toucher des aspect théoriques et appliqués de l’écologie. Je ne saurais résumer en quelques lignes tout ce que je souhaite adresser à mon co-encadrant, Eric Baubet de l’ONCFS, qui m’a accompagné d’une manière des plus actives, enrichissantes et humaines depuis le début de ce projet en tant que stagiaire, jusqu’à ma soutenance de thèse. Je le remercie pour d’avoir été le réfèrent idéal permettant à un jeune étudiant en écologie baigné de concepts théoriques de travailler avec des chasseurs vivant les pieds sur terre. A mon co-directeur, Stéphane Dray du LBBE, je voudrais adresser mes plus profonds remerciements, pas seulement pour m’avoir orienté sur les méthodes les plus pertinentes dans ce travail, mais aussi pour m’avoir aider à force de réflexions et de discussions, à éclaircir mon raisonnement et prendre le recul nécessaire à la production d’un travail scientifique que j’espère être de qualité. Son soutien scientifique et moral sans faille a été une des clefs permettant à ce travail d’aboutir et je l’en remercie encore. Parce que reprendre la direction d’une thèse déjà en cours doit être un exercice particulièrement difficile, je remercie tout particulièrement ma directrice de thèse, Anne Loison du LECA pour m’avoir permis de poursuivre mon travail dans conditions d’encadrement plus sereines. Je lui suis infiniment reconnaissant d’avoir su percevoir et développer une solide dimension biologique au sein de ce travail. 234 Si nous n’avons eu que peu de temps pour interagir, je souhaite tout de même adresser mes pensées à Mr Jean François Dobremez, à qui je dois sans conteste l’opportunité d’avoir pu réaliser cette thèse au sein de l’étude du bassin genevois. J’adresse également tous mes remerciements à Christiane Gallet pour m’avoir accueilli dans le Laboratoire d’Ecologie Alpine de l’Université de Savoie et Dominique Mouchiroud pour son accueil au sein du Laboratoire de Biométrie et Biologie Evolutive de l’UCB Lyon 1. J’adresse des remerciements particuliers à Dominique Allainé pour m’avoir ouvert la porte de son bureau lorsque je cherchais un stage « à tout hasard » au sein du laboratoire, et à Jean Michel Gaillard pour avoir toujours su, en peu de temps, m’orienter sur les pistes de recherche les plus pertinentes. Un grand merci à toutes les autres personnes avec qui j’ai eu la chance de collaborer lors de la rédaction des publications, notamment Claude Fischer, Achim Zeileis, Sonia Saïd, Céline Richomme, Eve Afonso, Emanuelle Gilot-Fromont, David Christianson, Herver Fritz et Marion Valeix. J’adresse un salut nostalgique à la cohorte de doctorants « habitat » apparue en 2005-2006 et disparaissant avec moi aujourd’hui : Mathieu Basille, Bram Van Moorter et Jodie Martin, tous filles et fils spirituels de notre maître Clément Calenge et avec qui j’ai eu un immense plaisir à construire petit à petit mon esprit scientifique (à la base plutôt tortueux mais qui apprend, je l’espère, à filer plus droit quand il le faut). Il me faut désormais rendre hommage à l’ensemble du groupe de recherche international sur le sanglier pour la pertinence de leur travaux (il fallait le dire) et surtout, pour ces jours à chanter, boire et danser aux quatre coins du monde (quand on a vu Chypre, on a vu le monde). Merci à Ulf, Carlos, Sabrina, Giovanna, Cornelia, Carme, Henrik, ainsi que Sophie, Sandra, Laura, Oliver et Céline et les autres pour ces moments inoubliables. 235 A mes amis du bureau de la salle 60, duquel personne ne ressort indemne, et de ses environs plus ou moins lointains, je souhaite une très belle vie dans quelques milieux professionnels que ce soit. Salut à toi Pierrick, Stephanie, Patrice, Jodie, Benjamin, Sophie, Mathieu, Bram, Dominique, Karine, Emilie, Eve, Gaëlle, Alexis, Maud, Simon, Emilien, Hélène, Karine, Delphine, David et Céline. Etienne, Micheline et Arnaud savent à quel point leur présence a été importante pour moi. Merci également à mes amis « hors science » et ma famille, Clément et Elodie, Nicolas, Pauline et Lutz, Rémi et Cindy, Damien et Béatrice, Dominique, Monique, Marion et Nicolas, May et Pascal, Natacha et Julien, Pierrette et Marie et toutes les familles Tolon Coulet, Atten et Rey pour leur présence pendant toutes ces années. Merci à Chantal, Bruno et Bernadette d’avoir été là. A ma mère et mon père, pour ce qu’ils m’ont donné, j’adresse plus encore que ne pourrait contenir ce simple mot : merci. A toi Agnès, je ne peux adresser de réels remerciements tellement ce mot ne saurait décrire ce tout ce que je te dois. Que notre aventure soit belle et nous pousse encore à devenir nous même. Merci à mon fils d’avoir attendu le lendemain de la date du dépôt de la thèse avant d’arriver au monde. Je lui revaudrai ça. 236 REFERENCES 237 Abrahams, M. V. and Dill, L. M. 1989. A determination of the energetic equivalence of the risk of predation. — Ecology 70: 999-1007. Abramsky, Z. et al. 1997. Gerbils under threat of owl predation: Isoclines and isodars. — Oikos 78: 81-90. Aebischer, N. J. et al. 1993. Compositional analysis of habitat use from animal radiotracking data. — Ecology 74: 1313-1325. Allee, W. C. et al. 1949. Principles of animal ecology. — W.B. Saunders. Anderson, D. P. et al. 2005. Scale-dependent summer resource selection by reintroduced elk in Wisconsin, USA. — Journal of Wildlife Management 69: 298-310. Andrews, D. W. K. 1993. Tests for parameter instability and structural-change with unknown change point. — Econometrica 61: 821-856. Armsworth, P. R. and Roughgarden, J. E. 2005. The impact of directed versus random movement on population dynamics and biodiversity patterns. — American Naturalist 165: 449-465. Banks, S. C. et al. 2007. Sex and sociality in a disconnected world: a review of the impacts of habitat fragmentation on animal social interactions. — Canadian Jounal of Zoology 85: 1065-1079. Baubet, E. L. 1998. Biologie du sanglier en montagne : biodémographie, occupation de l'espace et régime alimentaire. Department of Evolutionary Ecology. University of Lyon 1, p 299. Baxter, D. J. M. et al. 2006. Behavioural response of bats to perceived predation risk while foraging. — Ethology 112: 977-983. Benhamou, S. 1997. Path integration by swimming rats. — Animal Behaviour 54: 321327. Benhamou, S. and Bovet, P. 1989. How animals use their environment: a new look at kinesis. — Animal Behaviour 38: 375-383. Bergman, C. M. et al. 2000. Caribou movement as a correlated random walk. — Oecologia 123: 364-374. 238 Bernstein, C. et al. 1991. Individual decisions and the distribution of predators in a patchy environment .2. The influence of the travel costs and structure of the environment. — Journal of Animal Ecology 60: 205-225. Berryman, A. A. and Hawkins, B. A. 2006. The refuge as an integrating concept in ecology and evolution. — Oikos 115: 192-196. Bieber, C. and Ruf, T. 2005. Population dynamics in wild boar Sus scrofa: ecology, elasticity of growth rate and implications for the management of pulsed resource consumers. — Journal of Applied Ecology 42: 1203-1213. Blackwell, B. F. and Seamans, T. W. 2009. Enhancing the Perceived Threat of Vehicle Approach to Deer. — Journal of Wildlife Management 73: 128-135. Blumstein, D. T. et al. 2003. Testing a key assumption of wildlife buffer zones: is flight initiation distance a species-specific trait? — Biological Conservation 110: 97100. Blumstein, D. T. et al. 2005. Inter-specific variation in avian responses to human disturbance. — Journal of Applied Ecology 42: 943-953. Boisaubert, B. and Klein, F. 1984. Contribution à l'étude de l'occupation de l'espace chez le sanglier (Sus scrofa) par capture et recapture. In: Publ., I. (ed), Symposium International sur le Sanglier. Les Colloques de l'INRA, pp. 135-150. Boitani, L. et al. 1994. Spatial and activity patterns of wild boars in Tuscany, Italy. — Journal of Mammalogy 75: 600-612. Bonenfant, C. et al. 2009. Empirical evidence of density dependence in populations of large herbivores. — Advances in Ecological research 41: 314-345. Bonenfant, M. and Kramer, D. L. 1996. The influence of distance to burrow on flight initiation distance in the woodchuck, Marmota monax. — Behavioral Ecology 7: 299-303. Bouskila, A. and Blumstein, D. T. 1992. Rules of thumb for predation hazard assessment - predictions from a dynamic-model. — American Naturalist 139: 161-176. Boyce, M. S. and McDonald, L. L. 1999. Relating populations to habitats using resource selection functions. — Trends in Ecology and Evolution 14: 268-272. 239 Brandt, S. et al. 2005. La chasse en battue modifie l'utilisation de l'espace par le Sanglier: quelles conséquences pour la gestion ? — Faune Sauvage 266: 12-17. Brandt, S. et al. 1998. Domaine vital diurne des sangliers en forêt de Châteauvillain Arc-en-Barrois. — Bulletin Mensuel O.N.C. 234: 4-11. Bratton, S. P. 1975. The effect of the european wild boar, Sus scrofa, on gray beech forest in the Great Smoky Mountains. — Ecology 56: 1356-1366. Brooks, C. J. and Harris, S. 2008. Directed movement and orientation across a large natural landscape by zebras, Equus burchelli antiquorum. — Animal Behaviour 76: 277-285. Broseth, H. and Pedersen, H. C. 2000. Hunting effort and game vulnerability studies on a small scale: a new technique combining radio-telemetry, GPS and GIS. — Journal of Applied Ecology 37: 182-190. Brown, J. S. 1988. Patch use as an indicator of habitat preference, predation risk, and competition. — Behavioral Ecology and Sociobiology 22: 37-47. Brown, J. S. 1999. Vigilance, patch use and habitat selection: Foraging under predation risk. — Evolutionary Ecology Research 1: 49-71. Brown, J. S. et al. 1999. The ecology of fear: Optimal foraging, game theory, and trophic interactions. — Journal of Mammalogy 80: 385-399. Bruinderink, G. W. T. A. G. and Hazebroek, E. 1996. Ungulate traffic collisions in Europe. — Conservation Ecology 10: 1059-1067. Bshary, R. 2001. Diana monkeys, Cercopithecus diana, adjust their anti-predator response behaviour to human hunting strategies. — Behavioral Ecology and Sociobiology 50: 251-256. Burnham, K. P. and Anderson, D. R. 2002. Model selection and multimodel inference. A practical information-theoretic approach. — Springer-Verlag New York, Inc. Byers, J. A. 1997. American pronghorn: social adaptations and the ghosts of predators past. — University of Chicago Press. Cahill, S. et al. 2003. Spacing and nocturnal activity of wild boar Sus scrofa in a Mediterranean metropolitan park. — Wildlife Biology 9: 3-13. 240 Calenge, C. 2006. The package adehabitat for the R software: a tool for the analysis of space and habitat use by animals. — Ecological Modelling 197: 516-519. Calenge, C. and Dufour, A. B. 2006. Eigenanalysis of selection ratios from animal radiotracking data. — Ecology 87: 2349-2355. Calenge, C. et al. 2005. K-select analysis: a new method to analyse habitat selection in radio-tracking studies. — Ecological Modelling 186: 143-153. Calenge, C. et al. 2004. Efficiency of spreading maize in the garrigue to reduce Wild Boar (Sus scrofa) damage to Mediterranean vineyards. — European Journal of Wildlife Research 50: 112-120. Calenge, C. et al. 2002. Summer and hunting season home ranges of wild boar (Sus scrofa) in two habitats in France. — Game and Wildlife Science 19: 281-301. Caley, P. 1997. Movements, Activity Patterns and Habitat Use of Feral Pigs (Sus scrofa) in a Tropical Habitat. — Wildlife Research 24: 77-87. Caley, P. and Ottley, B. 1995. The Effectiveness of Hunting Dogs for Removing Feral Pigs (Sus scrofa). — Wildlife Research 22: 147-154. Caro, T. M. 2005. Antipredator defenses in birds and mammals. — University of Chicago Press. Carter, J. and Leonard, B. P. 2002. A review of the literature on the worldwide distribution., spread of, and efforts to eradicate the coypu (Myocastor coypus). — Wildlife Society Bulletin 30: 162-175. Charnov, E. L. 1976. Optimal foraging, the marginal value theorem. — Theoretical Population Biology 9: 129-136. Choquenot, D. and Ruscoe, W. A. 2003. Landscape complementation and food limitation of large herbivores: habitat-related constraints on the foraging efficiency of wild pigs. — Journal of Animal Ecology 72: 14-26. Clarke, M. F. et al. 1993. Site familiarity affects escape behaviour of the eastern chipmunk, Tamias striatus. — Oikos 66: 533-537. Coltman, D. W. et al. 2003. Undesirable evolutionary consequences of trophy hunting. — Nature 426: 655-658. 241 Cooper, W. E. 1998 Risk factors and emergence from refuge in the lizard Eumeces laticeps —Behaviour 135: 1065-1076. Cooper, W. E. 2008. Visual monitoring of predators: occurrence, cost and benefit for escape. — Animal Behaviour 76: 1365-1372. Cooper, W. E. 2009. Theory successfully predicts hiding time: new data for the lizard Sceloporus virgatus and a review. — Behavioral Ecology 20: 585-592. Coulon, A. et al. 2008. Inferring the effects of landscape structure on roe deer (Capreolus capreolus) movements using a step selection function. — Landscape Ecology 23: 603-614. Cousse, S. et al. 1994. Use of space by juveniles in relation to their postnatal range, mother, and siblings: an example in the wild boar Sus scrofa L. — Canadian Jounal of Zoology 72: 1691–1694. Cowlishaw, G. 1997. Refuge use and predation risk in a desert baboon population. — Animal Behaviour 54: 241-253. Creel, S. and Christianson, D. 2008. Relationships between direct predation and risk effects. — Trends in Ecology and Evolution 23: 194-201. Croney, C. C. et al. 2003. A note on visual, olfactory and spatial cue use in foraging behavior of pigs: indirectly assessing cognitive abilities. — Applied Animal Behaviour Science 83: 303-308. D'Andrea, L. et al. 1995. Preliminary data of the Wild Boar (Sus scrofa) Space use in mountain environment. — Ibex J.M.E. 3: 117-121. Dall, S. R. X. et al. 2005. Information and its use by animals in evolutionary ecology. — Trends in Ecology and Evolution 20: 187-193. Dalziel, B. D. et al. 2008. Fitting probability distributions to animal movement trajectories: Using artificial neural networks to link distance, resources, and memory. — American Naturalist 172: 248-258. Danchin, E. et al. 2004. Public information: From nosy neighbors to cultural evolution. — Science 305: 487-491. 242 Dardaillion, M. and Beugnon, G. 1987. The influence of some environmental characteristics on the movements of wild boar Sus scrofa. — Biology of Behaviour 12: 82-92. Darwin, C. 1859. On the origin of species by means of natural selection, or the preservation of favoured races in the struggle for life. — John Murray. Delibes, M. et al. 2001a. Attractive sinks, or how individual behavioural decisions determine source-sink dynamics. — Ecology Letters 4: 401-403. Delibes, M. et al. 2001b. Effects of an attractive sink leading into maladaptive habitat selection. — The American Naturalist 158: 277-285. Dexter, N. 1996. The Effect of an Intensive Shooting Exercise from a Helicopter on the Behaviour of Surviving Feral Pigs. — Wildlife Research 23: 435-441. Diamond, J. M. 1975. The island dilemma: Lessons of modern biogeographic studies for the design of natural reserves. — Biological Conservation 7: 129-146. Diamond, J. M. 1997. Aaaaaaaaaaaaargh, no! — Nature 385: 295-296. Dobzhansky, T. 1973. Nothing in biology makes sense except in the light of evolution. — The American Biology Teacher 35: 125-129. Dolédec, S. and Chessel, D. 1987. Rythmes saisonniers et composantes stationnelles en milieu aquatique I- Description d'un plan d'observations complet par projection de variables. — Acta Oecologica, Oecologia Generalis 8: 403-426. Duriez, O. et al. 2005. Factors affecting population dynamics of Eurasian woodcocks wintering in France: assessing the efficiency of a hunting-free reserve. — Biological Conservation 122: 89-97. Dussault, C. et al. 2005. Linking moose habitat selection to limiting factors. — Ecography 28: 619-628. Endler, J. A. 1991. Interaction between predators and prey. — In: Krebs, J. R. and Davies, N. B. (eds), Behavioral ecology: an evolutionary approach. Blackwell Scientific Publications, pp. 169-196. Escoufier, Y. 1987. The duality diagram: a means of better practical applications. — In: Legendre, P. and Legendre, L. (eds), Development in numerical ecology. Springer Verlag. 243 Estes, J. A. et al. 1998. Killer whale predation on sea otters linking oceanic and nearshore ecosystems. — Science 282: 473-476. Fauchald, P. 2009. Spatial interaction between seabirds and prey: review and synthesis. — Marine Ecology - Progress Series 391: 139-151. Fauchald, P. et al. 2000. Scale-dependent predator-prey interactions: The hierarchical spatial distribution of seabirds and prey. — Ecology 81: 773-783. Fernandez-Llario, P. et al. 2003. Habitat effects and shooting techniques on two wild boar (Sus scrofa) populations in Spain and Portugal. — Zeitschrift für Jagdwissenschaft 49: 120-129 Fieberg, J. and Kochanny, C. O. 2005. Quantifying home-range overlap: The importance of the utilization distribution. — Journal of Wildlife Management 69: 1346-1359. Fitzgibbon, C. D. 1990. Why do hunting cheetahs prefer male gazelles? . — Animal Behaviour 40: 837-845. Forester, J. D. et al. 2009. Accounting for animal movement in estimation of resource selection functions: sampling and data analysis. — Ecology 90: 3554-3565. Fortin, D. et al. 2005. Wolves influence elk movements: Behavior shapes a trophic cascade in Yellowstone National Park. — Ecology 86: 1320-1330. Fowler, C. W. 1987. A review of density dependence in populations of large mammals. — Current Mammalogy 1: 401-441. Fox, A. D. and Madsen, J. 1997. Behavioural and distributional effects of hunting disturbance on waterbirds in Europe: Implications for refuge design. — Journal of Applied Ecology 34: 1-13. Frair, J. L. et al. 2005. Scales of movement by elk (Cervus elaphus) in response to heterogeneity in forage resources and predation risk. — Landscape Ecology 20: 273-287. Fraker, M. 2009. The effect of prior experience on a prey's current perceived risk. — Oecologia 158: 765-774. Fretwell, S. D. 1972. Populations in a seasonal environment. — Princeton University Press. 244 Fretwell, S. D. and Lucas, H. L. J. 1969. On territorial behavior and other factors influencing habitat distributions in birds. I. Theoretical development. — Acta Biotheoretica 19: 16-36. Fretwell, S. D. and Lucas, H. L. J. 1970. On territorial behavior and other factors influencing habitat distributions in birds. — Acta Biotheoretica 19: 16-36. Frid, A. and Dill, L. M. 2002. Human-caused disturbance stimuli as a form of predation risk. Ecology and Society. p 11. Gavin, S. D. and Komers, P. E. 2006. Do pronghorn (Antilocapra americana) perceive roads as a predation risk? — Canadian Jounal of Zoology 84: 1775-1780. Geisser, H. and Reyer, H. U. 2004. Efficacy of hunting, feeding, and fencing to reduce crop damage by Wild Boars. — Journal of Wildlife Management 68: 939-946. Gell, F. R. and Roberts, C. M. 2003. Benefits beyond boundaries: the fishery effects of marine reserves. — Trends in Ecology and Evolution 18: 448-455. Gerber, L. R. et al. 2003. Population models for marine reserve design: A retrospective and prospective synthesis. — Ecological Applications 13: 47-64. Gilroy, J. J. and Sutherland, W. J. 2007. Beyond ecological traps: perceptual errors and undervalued resources. — Trends in Ecology and Evolution 22: 351-356. Gimenez-Anaya, A. et al. 2008. Food habits of wild boars (Sus scrofa) in a Mediterranean coastal wetland. — Wetlands 28: 197-203. Gould, S. J. and Lewontin, R. C. 1979. The spandrels of San Marco and the Panglossian paradigm: a critique of the adaptationist program. — Procceding of the Royal Society B 205 581-598. Grand, T. C. and Dill, L. M. 1997. The energetic equivalence of cover to juvenile coho salmon (Oncorhynchus kisutch): ideal free distribution theory applied. — Behavioral Ecology 8: 437-447. Grand, T. C. and Dill, L. M. 1999a. The effect of group size on the foraging behaviour of juvenile coho salmon: reduction of predation risk or increased competition? — Animal Behaviour 58: 443-451. Grand, T. C. and Dill, L. M. 1999b. Predation risk, unequal competitors and the ideal free distribution. — Evolutionary Ecology Research 1: 389-409. 245 Gurung, B. et al. 2008. Factors associated with human-killing tigers in Chitwan National Park, Nepal. — Biological Conservation 141: 3069-3078. Hall, L. S. et al. 1997. The habitat concept and a plea for standard terminology. — Wildlife Society Bulletin 25: 173-182. Hamilton, W. D. 1971. Geometry for the selfish herd. — Journal of Theoretical Biology 31: 295-311. Hansen, B. B. et al. 2009. Functional response in habitat selection and the tradeoffs between foraging niche components in a large herbivore. — Oikos 118: 859-872. Haugen, T. O. et al. 2006. The ideal free pike: 50 years of fitness-maximizing dispersal in Windermere. — Proceeding of the Royal Society B 273: 2917–2924. Hebblewhite, M. and Merrill, E. H. 2007. Multiscale wolf predation risk for elk: does migration reduce risk? — Oecologia 152: 377-387. Hebeisen, C. 2007. Population size, density and dynamics, and social organization of wild boar (Sus scrofa) in the Basin of Geneva. Department of Behavioural Ecology. University of Neuchâtel, p 81. Heinken, T. et al. 2006. Soil seed banks near rubbing trees indicate dispersal of plant species into forests by wild boar. — Basics and Applied Ecology 7: 31-34. Held, S. et al. 2005. Foraging behaviour in domestic pigs (Sus scrofa): remembering and prioritizing food sites of different value. — Animal Cognition 8: 114-121. Held, S. et al. 2000. Social tactics of pigs in a competitive foraging task: the 'informed forager' paradigm. — Animal Behaviour 59: 569-576. Hochman, V. and Kotler, B. P. 2007. Patch use, apprehension, and vigilance behavior of Nubian Ibex under perceived risk of predation. — Behavioral Ecology 18: 368374. Holden, C. 2006. Inching toward movement ecology. — Science 313: 779-+. Holling, C. S. 1959. Some characteristics of simple types of predation and parasitism. — Canadian Entomologist 91: 385-398. Holyoak, M. et al. 2008. Trends and missing parts in the study of movement ecology. — Proccedings of the National Academy of Sciences 105: 19060-19065. 246 Hone, J. 1995. Spatial and temporal aspects of vertebrate damage with emphasis on Feral Pigs. — Journal of Applied Ecology 32: 311-319. Houston, A. I. et al. 1993. General results concerning the trade-off between gaining energy and avoiding predation. — Philosophical Transactions: Biological Sciences 341: 375-397. Huggard, D. J. 1993. Prey selectivity of wolves in Banff National Park. II. Age, sex, and condition of elk. — Canadian Jounal of Zoology 71: 140-147. Hugie, D. M. and Dill, L. M. 1994. Fish and game: a game theoretic approach to habitat selection by predators and prey. — Journal of Fish Biology 45: 151-169. Jackson, J. B. C. et al. 2001. Historical overfishing and the recent collapse of coastal ecosystems. — Science 293: 629-638. Janeau, G. et al. 1995. Daily movement patterns variations in Wild Boar (Sus scrofa L.). — Ibex J.M.E. 3: 98-101. Janeau, G. and Spitz, F. 1984. Budget espace temps des sangliers (Sus scrofa L.) en forêt de Grésigne. — Les Colloques de l'INRA 22: 123-134. Janeau, G. and Spitz, F. 1990. Spatial strategies: an attempt to classify daily movements of wild boar. — Acta Theriologica 35: 129-149. Jansen, J. et al. 2009. Spatial learning in pigs: effects of environmental enrichment and individual characteristics on behaviour and performance. — Animal Cognition 12: 303-315. Jayakody, S. et al. 2008. Red deer Cervus elephus vigilance behaviour differs with habitat and type of human disturbance. — Wildlife Biology 14: 81-91. Johnson, C. J. et al. 2002. Movement parameters of ungulates and scale-specific responses to the environment. — Journal of Animal Ecology 71: 225-235. Johnson, D. H. 1980. The comparison of usage and availability measurements for evaluating resource preference. — Ecology 61: 65-71. Kaminski, G. et al. 2005. Life-history patterns in female wild boars (Sus scrofa): motherdaughter postweaning associations. — Canadian Jounal of Zoology 83: 474-480. Kareiva, P. et al. 2002. Slow-moving journals hinder conservation efforts. — Nature 420: 15-15. 247 Kellner, J. B. et al. 2007. Fishing the line near marine reserves in single and multispecies fisheries. — Ecological Applications 17: 1039-1054. Kennedy, M. and Gray, R. D. 1993. Can ecological theory predict the distribution of foraging animals? A critical analysis of experiments on the ideal free distribution. — Oikos 68: 158-166. Keuling, O. et al. 2008a. Annual and seasonal space use of different age classes of female wild boar Sus scrofa L. — European Journal of Wildlife Research 54: 403-412. Keuling, O. et al. 2008b. How does hunting influence activity and spatial usage in wild boar Sus scrofa L.? — European Journal of Wildlife Research 54: 729-737. Kittle, A. M. et al. 2008. The scale-dependent impact of wolf predation risk on resource selection by three sympatric ungulates. — Oecologia 157: 163-175. Kjellander, P. et al. 2004. Predation risk and longevity influence variation in fitness of female roe deer (Capreolus capreolus L.). — Procceding of the Royal Society B 271: 338-340. Knegt, H. J. et al. 2007. Patch density determines movement patterns and foraging efficiency of large herbivores. — Behavioral Ecology 18: 1065-1072. Koops, M. A. 2004. Reliability and the value of information —Animal Behaviour 67: 103-111. Kotler, B. P. et al. 1991. Factors affecting gerbil foraging behavior and rates of owl predation. — Ecology 72: 2249-2260. Kristan, W. B. 2000. The role of habitat selection behavior in population dynamics: source-sink systems and ecological traps. — Oikos 103: 457-468 Kruz, J. C. and Marchinton, R. L. 1972. Radiotelemetry studies of Feral Hogs in South Carolina. — Journal of Wildlife Management 36: 1240-1248. Lande, U. S. et al. 2009. Use of hunters in wildlife surveys: does hunter and forest grouse habitat selection coincide? . — European Journal of Wildlife Research DOI: 10.1007/s10344-009-0291-2: Laundré, J. W. et al. 2001. Wolves, elk, and bison: reestablishing the “landscape of fear” in Yellowstone National Park, U.S.A. — Canadian Journal of Zoology 79: 14011409. 248 Lemel, J. et al. 2003. Variation in ranging and activity behaviour of European wild boar Sus scrofa in Sweden. — Wildlife Biology 9: 29-36. Lenihan, H. S. et al. 2001. Cascading of habitat degradation: Oyster reefs invaded by refugee fishes escaping stress. — Ecological Applications 11: 764-782. Leslie, P. H. 1945. Some further notes on the use of matrices in population mathematics. — Biometrika 35: 213-245. Lewis, S. et al. 2001. Evidence of intra-specific competition for food in a pelagic seabird. — Nature 412: 816-819. Lima, S. L. 2002. Putting predators back into behavioral predator-prey interactions. — Trends in Ecology and Evolution 17: 70-75. Lima, S. L. and Bednekoff, P. A. 1999. Temporal Variation in Danger Drives Antipredator Behavior: The Predation Risk Allocation Hypothesis. — The American Naturalist 153: 649-659. Lima, S. L. and Zollner, P. A. 1996. Towards a behavioral ecology of ecological landscapes. — Trends in Ecology and Evolution 11: 131-135. Lipp, H. P. et al. 2004. Pigeon homing along highways and exits. — Current Biology 14: 1239-1249. Lotka, A. J. 1925. Elements of physical biology. — Williams & Wilkins company. Lundberg, P. and Jonzen, N. 1999. Spatial population dynamics and the design of marine reserves. — Ecology Letters 2 129-134. MacArthur, R. H. and Pianka, E. E. 1966. On the optimal use of a patchy environment. — American Naturalist 100: 603-609. Madsen, J. 1998. Experimental refuges for migratory waterfowl in Danish wetlands. II. Tests of hunting disturbance effects. — Journal of Applied Ecology 35: 398-417. Maillard, D. 1996. Occupation et utilisation de la garrigue et du vignoble mediterraneens par le sanglier (Sus scrofa L.). Université d'Aix-Marseille 3, p 275. Maillard, D. 1998. Approche du fonctionnement de la population de sangliers (Sus scrofa L.) de la Réserve Naturelle de Roque-Haute à partir des résultats scientifiques obtenus sur l'espèce en milieu méditerranéen. — Ecologia Mediterranea 24: 223234. 249 Maillard, D. and Fournier, P. 1995. Effect of shooting with hounds on size of resting range of Wild Boar (Sus scrofa L.) groups in Mediterranean habitat. — Ibex J.M.E. 3: 102-107. Mangel, M. 1998. No-take areas for sustainability of harvested species and a conservation invariant for marine reserves. — Ecology Letters 1: 87-90. Manly, B. F. J. et al. 2002. Resource selection by animals. Statistical design and analysis for field studies. — Kluwer Academic Publisher. Markina, F. A. et al. 2004. Physical development of wild boar in the Cantabric Mountains, Álava, Nothern Spain. — Galemys 16: 25-34. Martin, J. 2009. Habitat selection and movement by brown bears in multiple-use landscapes. University of Claude Bernard Lyon 1. Martin, J. and Lopez, P. 1999a. An experimental test of the costs of antipredatory refuge use in the wall lizard, Podarcis muralis. — Oikos 84: 499-505. Martin, J. and Lopez, P. 1999b. When to come out from a refuge: risk-sensitive and statedependent decisions in an alpine lizard. — Behavioral Ecology 10: 487-492. Martin, J. et al. 2003. Loss of mating opportunities influences refuge use in the Iberian rock lizard, Lacerta monticola. — Behavioral Ecology and Sociobiology 54: 505510. Martin, J. et al. 2009a. Temporal patterns of predation risk affect antipredator behaviour allocation by Iberian rock lizards. — Animal Behaviour 77: 1261-1266. Martin, J. et al. 2009b. On the use of telemetry in habitat selection studies. — In: Barculo, D. and Daniels, J. (eds), Telemetry: research, technology and applications Nova Science Publishers Inc. Martinez, J. A. et al. 2003. Predictive models of habitat preferences for the Eurasian eagle owl Bubo bubo: a multiscale approach. — Ecography 26: 21-28. Massei, G. et al. 1997. Factors influencing home range and activity of wild boar (Sus scrofa) in a Mediterranean coastal area. — Journal of Zoology 242: 411-423. McIlroy, J. C. and Saillard, R. J. 1989. The effect of hunting with dogs on the numbers and movements of feral pigs, Sus scrofa, and the subsequent success of poisoning 250 exercises in Namadgi National Park, A.C.T. — Australian Wildlife Research 16: 353-363. McLoughlin, P. D. et al. 2006. Lifetime reproductive success and density-dependent, multi-variable resource selection. — Procceding of the Royal Society B 273: 1449-1454. McLoughlin, P. D. et al. 2005. Relating predation mortality to broad-scale habitat selection. — Journal of Animal Ecology 74: 701-707. McLoughlin, P. D. et al. 2007. Lifetime reproductive success and composition of the home range in a large herbivore. — Ecology 88: 3192-3201. McNamara, J. M. et al. 2006. Bayes' theorem and its applications in animal behaviour. — Oikos 112: 243-251. McNamara, J. M. and Houston, A. I. 1980. The application of statistical decision-theory to animal behavior. — Journal of Theoretical Biology 85: 673-690. Mech, S. G. and Zollner, P. A. 2002. Using body size to predict perceptual range. — Oikos 98: 47-52. Meriggi, A. and Sacchi, O. 2001 Habitat requirements of wild boars in the northern Apennines (N Italy): a multi-level approach. — Italian Journal of Zoology 68: 4755. Mitchell, W. A. and Lima, S. L. 2002. Predator-prey shell games: large-scale movement and its implications for decision-making by prey. — Oikos 99: 249-259. Molinari-Jobin, A. et al. 2004. Life cycle period and activity of prey influence their susceptibility to predators. — Ecography 27: 323-329. Moody, A. L. et al. 1996. Ideal free distributions under predation risk. — Behavioral Ecology and Sociobiology 38: 131-143. Morales, J. M. et al. 2004. Extracting more out of relocation data: building movement models as mixtures of random walks. — Ecology 85: 2436-2445. Moretti, M. 1995. Biometric data and growth rates of a mountain population of wild boar (Sus scrofa L.), Ticino, Switzerland. — Ibex J.M.E. 3: 56-59. Morini, P. et al. 1995. Space use by pen-raised Wild Boars (Sus scrofa) released in Tuscany (Central Italy) -II: Home Range. — Ibex J.M.E. 3: 112-116. 251 Morris, D. W. 1987. Spatial scale and the cost of density-dependent habitat selection. — Evolutionary Ecology 1: 379-388. Morris, D. W. 1989. Density-dependent habitat selection: testing the theory with fitness data. — Evolutionary Ecology 3: 80-94. Morris, D. W. 2003. Toward an ecological synthesis: a case for habitat selection. — Oecologia 136: 1-13. Morris, D. W. 2006. Moving to the ideal free home. — Nature 443: 645-646. Mueller, T. and Fagan, W. F. 2008. Search and navigation in dynamic environments from individual behaviors to population distributions. — Oikos 117: 654-664. Nathan, R. 2008. An emerging movement ecology paradigm. — Proccedings of the National Academy of Sciences 105: 19050-19051. Nathan, R. et al. 2008. A movement ecology paradigm for unifying organismal movement research. — Proccedings of the National Academy of Sciences 105: 19052-19059. National Forestry Office 2008. REseau National de suivi à long terme des ECOsystèmes FORestiers. Neubert, M. G. 2003. Marine reserves and optimal harvesting. — Ecology Letters 6: 843–849. Nevin, O. T. and Gilbert, B. K. 2005. Measuring the cost of risk avoidance in brown bears: Further evidence of positive impacts of ecotourism. — Biological Conservation 123: 453-460. Novaro, J. A. et al. 2000. Effect of hunting in source-sink systems in the neotropics. — Conservation Biology 14: 713-721. O'Keefe, J. and Nadel, L. 1978. The hippocampus as a cognitive map. — Oxford University Press. Oaten, A. and Murdoch, W. W. 1975. Switching, functional response, and stability in predator-prey systems. — American Naturalist 109: 299-318. Owen-Smith, N. 2008. Changing vulnerability to predation related to season and sex in an African ungulate assemblage. — Oikos 117: 602-610. 252 Parker, G. A. and Maynard Smith, J. 1990. Optimality theory in evolutionary biology. — Nature 348: 27-33. Pellerin, M. et al. 2008. Roe deer Capreolus capreolus home-range sizes estimated from VHF and GPS data. — Wildlife Biology 14: 101-110. Pierce, G. J. and Ollason, J. G. 1987. Eight reasons why optimal foraging theory is a complete waste of time. — Oikos 49: 111-118. Pinheiro, J. C. and Bates, D. M. 2000. Mixed-effects models in S and S-PLUS. — Springer-Verlag. Pulliam, H. R. 1988. Sources, Sinks, and Population Regulation. — The American Naturalist 132: 652-661. Pyšek, P. et al. 2002. Patterns of invasion in temperate nature reserves. — Biological Conservation 104: 13-24. Quenette, P. Y. and Gerard, J. F. 1992. From individual to collective vigilance in wild boar (Sus scrofa). — Canadian Jounal of Zoology 70: 1632-1635 R Development Core Team 2008. R: A Language and Environment for Statistical Computing. — R Foundation for Statistical Computing, Vienna, Austria, <http://www.R-project.org/>. Ranta, E. et al. 1999. Resource matching with limited knowledge. — Oikos 86: 383-385. Remes, V. 2000. How can maladaptive habitat choice generate source-sink population dynamics? — Oikos 91: 579-582. Rettie, W. J. and Messier, F. 2000. Hierarchical habitat selection by woodland caribou: its relationship to limiting factors. — Ecography 23: 466-478. Revilla, E. et al. 2001. Edge-core effects and the effectiveness of traditional reserves in conservation: Eurasian badgers in Donana National Park. — Conservation Biology 15: 148-158. Ripple, W. J. and Beschta, R. L. 2004. Wolves and the ecology of fear: Can predation risk structure ecosystems? — Bioscience 54: 755-766. Roberts, A. M. and Liebgold, E. B. 2008. The effects of perceived mortality risk on habitat selection in a terrestrial salamander. — Behavioral Ecology 19: 621-626. 253 Roberts, C. M. et al. 2005. The role of marine reserves in achieving sustainable fisheries. — Philosophical Transactions: Biological Sciences 360: 123-132. Roberts, G. 1996. Why individual vigilance declines as group size increases. — Animal Behaviour 51: 1077-1086. Romesburg, H. C. 1985. Exploring, confirming and randomization tests. — Computers and Geosciences 11: 19-37. Rosell, C. et al. 2004. Activity patterns and social organization of Wild Boar (Sus scrofa, L.) in a wetland environment: preliminary data on the effect of shooting individuals. — Galemys 16: 185-193. Roth, T. C. et al. 2008. Can foraging birds assess predation risk by scent? — Animal Behaviour 76: 2021-2027. Russo, L. et al. 1997. Daily home range and activity of wild boar in a Mediterranean area free from hunting. — Ethology Ecology & Evolution 9: 287-294. Saez-Royuela, C. and Telleria, J. L. 1986. The increased population of the Wild Boar (Sus scrofa L.) in Europe. — Mammal Review 16: 97-101. Saïd, S. et al. in prep. Sex effect on habitat selection in response to hunting disturbance: the study of Wild boar. Santos, P. et al. 2004. Habitat selection by Wild Boar Sus scrofa L. in Alentejo, Portugal. — Galemys 16: 167-184. Saunders, G. and Kay, B. 1991. Movements of Feral Pigs (Sus scrofa) at Sunny Corner, New South Wales. — Wildlife Research 18: 49-61. Schaefer, M. B. 1954. Some aspects of the dynamics of populations important to the management of marine fisheries. — Bulletin of Inter-American Tropical Tune Commission 1: 27-56. Schley, L. and Roper, T. J. 2003. Diet of wild boar Sus scrofa in Western Europe, with particular reference to consumption of agricultural crops. — Mammal Review 33: 43-56. Schmidt, K. A. et al. 2008. Eastern chipmunks increase their perception of predation risk in response to titmouse alarm calls. — Behavioral Ecology 19: 759-763. 254 Schmitz, O. J. 2005. Scaling from plot experiments to landscapes: studying grasshoppers to inform forest ecosystem management. — Oecologia 145: 225-234. Servanty, S. 2007. Dynamique d’une population chassée de sangliers (Sus scrofa scrofa) en milieu forestier. . Département Ecologie Evolutive. Université Claude Bernard Lyon 1, p 240. Sherman, P. W. 1977. Nepotism and evolution of alarm calls. — Science 197: 12461253. Shochat, E. et al. 2002. Density-dependent habitat selection in migratory passerines during stopover: what causes the deviation from IFD? — Evolutionary Ecology 16: 469-488. Shochat, E. et al. 2005. Ecological traps in isodars: effects of tallgrass prairie management on bird nest success. — Oikos 111: 159-169. Shrader, A. M. et al. 2008. Do free-ranging domestic goats show 'landscapes of fear'? Patch use in response to habitat features and predator cues. — Journal of Arid Environments 72: 1811-1819. Sih, A. 1992. Prey uncertainty and the balancing of antipredator and feeding needs. — American Naturalist 139: 1052-1069. Sih, A. 2005. Predator–prey space use as an emergent outcome of a behavioral response race. — In: Barbosa, P. and Castellanos, I. (eds), Ecology of predator–prey interactions. Oxford University Press, pp. 240-255. Sih, A. et al. 2004. Behavioral syndromes: an ecological and evolutionary overview. — Trends in Ecology and Evolution 19: 372-378. Sih, A. et al. 2000. New insights on how temporal variation in predation risk shapes prey behavior. — Trends in Ecology and Evolution 15: 3-4. Simberloff, D. and Abele, L. G. 1982. Refuge Design and Island Biogeographic Theory: Effects of Fragmentation. — The American Naturalist 120: 41-50. Singer, F. J. et al. 1981. Home ranges, movements, and habitat use of European Wild Boar in Tennessee. — Journal of Wildlife Management 45: 343-353. 255 Sodeikat, G. and Pohlmeyer, K. 2003. Escape movement of family groups of wild boar Sus scrofa influenced by drive hunts in Lower Saxony, Germany. — Wildlife Biology 9: 43-49. Spitz, F. 1989. Mortalité et dispersion chez le sanglier (Sus scrofa) de Camargue. — Gibier Faune Sauvage 6: 27-42. Spitz, F. and Janeau, G. 1995. Daily selection of habitat in wild boar (Sus scrofa). — The Zoological Society of London 237: 423-434. Stankowich, T. 2008. Ungulate flight responses to human disturbance: A review and meta-analysis. — Biological Conservation 141: 2159-2173. Stedman, R. et al. 2004. Integrating wildlife and human-dimensions research methods to study hunters. — Journal of Wildlife Management 68: 762-773. Stephen, D. W. and Krebs, J. R. 1986. Foraging theory. — Princeton University Press. Stephens, D. W. 1989 Variance and the value of information. — American Naturalist 134: 128-140. Stewart, K. M. et al. 2005. Density-dependent effects on physical condition and reproduction in North American elk: an experimental test. — Oecologia 143: 8593. Surovell, T. et al. 2005. Global archaeological evidence for proboscidean overkill. — Procceding of the National Academy of Sciences 102: 6231-6236. Sutherland, W. J. 1996. From individual behaviour to population ecology. — Oxford University Press. Tinbergen, N. 1963. On aims and methods of Ethology. — Zeitschrift fur Tierpsychologie 20: 410-433. Toïgo, C. et al. 2008. Disentangling natural from hunting mortality in an intensively hunted wild boar population. — Journal of Wildlife Management 72: 1532–1539. Tolon, V. et al. 2009. Responding to spatial and temporal variations in predation risk: space use of a game species in a changing landscape of fear. — Canadian Jounal of Zoology 87: 1129-1137. Truvé, J. and Lemel, J. 2003. Timing and distance of natal dispersal for wild boar Sus scrofa in Sweden. — Wildlife Biology 9: 51-57. 256 Turner, M. G. 1989. Landscape ecology: the effect of pattern on process. — Annual Review of Ecology and Systematics 20: 171-197. Usher, M. B. 1988. Biological invasions of nature reserves: a search for generalizations. — Biological Conservation 44: 119-135. Valeix, M. et al. 2009a. Does the risk of encountering lions influence African herbivore behaviour at waterholes? — Behavioral Ecology and Sociobiology 63: 14831494. Valeix, M. et al. 2009b. Behavioral adjustments of African herbivores to predation risk by lions: Spatiotemporal variations influence habitat use. — Ecology 90: 23-30. van der Merwe, M. and Brown, J. S. 2008. Mapping the Landscape of Fear of the Cape Ground Squirrel (Xerus inauris). — Journal of Mammalogy 89: 1162-1169. Vassant, J. 1994. L'agrainage dissuasif: résultats d'expériences. — Bulletin Mensuel Office National de la Chasse 191: 101-105. Vassant, J. and Breton, D. 1986. Essai de réduction des dégâts de sangliers (Sus scrofa) sur blé (Triticum sativum) au stade laiteux par distribution de maïs (Zea mais) en forêt. — Gibier Faune Sauvage 3: 83-95. Verhulst, P. F. 1845. Recherches mathématiques sur la loi d'accroissement de la population. — Nouveaux Mémoires de l'Académie Royale des Sciences et BellesLettres de Bruxelles 18: 1-42. Volterra, V. 1926. Fluctuations in the abundance of a species considered mathematically. — Nature 118: 558-560. Webster, M. M. and Laland, K. N. 2008. Social learning strategies and predation risk: minnows copy only when using private information would be costly. — Procceding of the Royal Society B 275: 2869-2876. Welton, N. J. et al. 2003. Assessing predation risk: optimal behaviour and rules of thumb. — Theoretical Population Biology 64: 417-430. Werner, E. E. and Anholt, B. R. 1993. Ecological consequences of the trade-off between growth and mortality-rates mediated by foraging activity. — The American Naturalist 142: 242-272. 257 White, G. C. and Garrott, R. A. 1990. Analysis of wildlife radio-tracking data. — Academic Press. Whittaker, R. H. et al. 1973. Niche, habitat, and ecotope. — American Naturalist 107: 321-338. Whittingham, M. J. et al. 2006. Altering perceived predation risk and food availability: management prescriptions to benefit farmland birds on stubble fields. — Journal of Applied Ecology 43: 640-650. Willems, E. P. and Hill, R. A. 2009. Predator-specific landscapes of fear and resource distribution: effects on spatial range use. — Ecology 90: 546-555. Wilson, C. J. 2004. Rooting damage to farmland in Dorset, southern England, caused by feral Wild Boar Sus scrofa. — Mammal Review 34: 331-335. Wolf, M. et al. 2009. The attraction of the known: the importance of spatial familiarity in habitat selection in wapiti Cervus elaphus. — Ecography 32: 401-410. Wong, B. B. M. et al. 2005. Response to perceived predation threat in fiddler crabs: trust thy neighbor as thyself? — Behavioral Ecology and Sociobiology 58: 345-350. Worton, B. J. 1989. Kernel methods for estimating the utilization distribution in homerange studies. — Ecology 70: 164-168. Wright, J. P. and Jones, C. G. 2004. Predicting effects of ecosystem engineers on patchscale species richness from primary productivity. — Ecology 85: 2071-2081. Wright, S. 1932. The roles of mutation, inbreeding, crossbreeding, and selection in evolution. Proccedings of the 6th International Congress of Genetics. pp. 355-366. Ydenberg, R. C. and Dill, L. M. 1986. The economics of fleeing from predators. — Advances in the study of behavior 16: 229-249. Zeileis, A. 2005. A unified approach to structural change tests based on ML scores, F statistics, and OLS residuals. — Econometric Reviews 24: 445-466. Zeileis, A. et al. 2003. Testing and dating of structural changes in practice. — Computational Statistics & Data Analysis 44: 109-123. Zollner, P. A. and Lima, S. L. 1999. Search strategies for landscape-level interpatch movements. — Ecology 80: 1019-1030. 258 259 260 ANNEXE 1 261 SEX EFFECT ON HABITAT SELECTION IN RESPONSE TO HUNTING DISTURBANCE: THE STUDY OF WILD BOAR Sonia Saïd1, Vincent Tolon1,2, 3, Serge Brandt1, Eric Baubet1* 1 Office National de la Chasse et de la Faune Sauvage, Centre National d’Etudes et de Recherches Appliquées Cervidés-Sanglier, « Montfort », 01330 Birieux,, France 2 Laboratoire d’Ecologie Alpine CNRS UMR5553, Université de Savoie, Bâtiment Belledonne, F-73376 Le Bourget-du-Lac, France 3 Unité Mixte de Recherche N°5558 “ Biométrie et Biologie Evolutive ”, Bâtiment 711, Université Claude Bernard Lyon 1, 43 Boulevard du 11 novembre 1918, 69622 Villeurbanne Cedex, France In prep. 262 ABSTRACT In Europe, ungulate game population dynamic is mainly affected by human, than by natural predators, through hunting activities. Spatial variation of the “predation risk” due to the predator activities or distribution may reinforce the sexual difference of habitat selection as females with offspring are usually more risk adverse than males. From the intensive and long-term monitoring of wild boar, we study the distribution of female and male wild boar before, during and after hunting with highlight of habitat selection for different vegetative covers. Wild boar selected as expected dense vegetative cover during their resting periods, but, against expectations, they release this selection in response to hunting disturbance. As expected, females responded more to the hunting disturbance than males leading to more pronounced sexual difference during the riskier season. The use of dense bushes plots while resting is an established feature of wild boar behaviour as it probably reduces detectability or accessibility for human or other predators. The unexpected decrease of this selectivity may be explained eitheir by the increased hunting effort of hunters in bushes plots, or by the fact that the disturbance increased movements between resting, leading to a more random habitat selection pattern of animals. The observed sexual difference could results from higher flight response of females with offspring leading to an increased frequentation of secondary habitats whereas males can support more risks and remain hidden in bush plots. Our results showed how hunting disturbance can lead in the game species to an unexpected decrease of the prefered refuge, and a more pronounced differentiation of sexes amongst habitats. Key words: wild boar, Sus scrofa. hunting disturbance, habitat selection, sex effect. 263 INTRODUCTION Recently an increasing number of studies have revealed the impact of human activities on evolutionary changes observed in wildlife populations (Palumbi 2001). In Europe, ungulate game population dynamic is mainly affected by human, than by natural predators, through hunting activities (Jedrzjewski et al. 2002; Nores et al. 2008; Theuerkauf and Rouys 2008). These games species tend to avoid “predation risk”, often, by modifying their behaviour. For instance, animals can reallocate time budgets and, switch activity patterns (Lima and Bednekoff 1999), modulate vigilance levels (Roberts 1996) or group size (Hamilton 1971) in response to an increased predation risk. Because both sexes may have different sensitivity to predation risk. Many authors (Main et al. 1996; Ruckstuhl and Neuhaus 2002; Bonenfant et al. 2004 for review) defined the “predation-risk hypothesis” which states that anti-predator behaviour may also shape the sexual segregation. Indeed, several studies have reported that sexual segregation is common in ungulates outside the rutting season (Miquelle et al. 1992). Spatial variation of the “predation risk” due to the predator activities or distribution may reinforce the sexual segregation of females, especially those with offspring, which choose safe habitats even at the expense of resource abundance. In wild boar populations, the difference between female and male habitat use in response to human hunting have not been clearly studied (but see (Spitz and Janeau 1995). Many studies have stressed the influence of hunting activity on wild boar home range size (Brandt et al. 1998; Calenge et al. 2002) often showing an increase of the home range due to hunting activities (Maillard and Fournier 1995, Sodeikat and Pohlmeyer 2007), but such pattern is not a rule (Sodeikat et Pohlmeyer 1999; McIlroy and Saillard 1989). Knowleged on habitat use by wild boar have been investigated by indirect methods (Dardaillon 1986, Merrigi and Sacchi 1991; Abaigar et al. 1994) and shown that this species is very sensitive to the vegetation type (the denser the vegetation the more likely the wild boar will inhabit the area; Meriggi and Sacchi 2001) or direct methods with radiotracking and shown human disturbance (regarding hunting activities and reserve effect; Tolon et al. 2009). 264 Futhermore, through radio-telemetry studies the avoidance of open vegetation type have been confirmed and was adopted as a rule, even variations due to the seasonal or groups effect might exist (Barber and Coblentz 1986; Boitani et al. 1994; Caley 1997; Santos et al. 2004). Indeed wild boar live in matrilineal social system, adult male living solitary outside of the rutting period (Kaminski et al. 2005). Familial groups (females and their offspring) are then expected to select denser cover for the safetiness than solitary male (Spitz and Janeau 1995). From the intensive and long-term monitoring of wild boar over a period of 9 years (ca. 34 female-years and 12 male-years) in the well-studied population of Arc en Barrois (e. g., Servanty et al. 2007, 2009; Toigo et al. 2008), we study the distribution of female and male wild boar before, during and after hunting with highlight of habitat selection between forest plots for both sexes. We tested following predictions: (i) wild boar select the densest cover for their resting site settlement, (ii) they enhance this dense cover selection during hunting in order to avoid risk of being killed (iii) familial groups (females with offspring) express a stronger response to hunting disturbance than solitary males during hunting season. MATERIAL AND METHODS Study site Data were collected in the Châteauvillain-Arc-en-Barrois forest area (11 000 ha), situated in east France (48°02N, 4°56E). The study area lies essentially on calcareous soil (Figure 1). The main species of woody ticket are Carpinus betulus, Fagus sylvatica, Quercus petreae. The bush is characterised by shrubs such as Prunus spinosa, Crataegus laevigata, Crataegus monogyna). This national forest are divided in 735 parcels, the limits between them being materialized by forest’s trails. The forest is managed at the finer sub-parcel’s scale. For each sub-parcel, the total cover (in %) of species were determined by Brandt, S. for 735 sub-parcels for 5 years : 1995, 1997, 1999, 2001 and 265 2002. Habitats were considered to be stable for two years and value for 1996, 1998, 2000 and 2003 were set from respective preceding years. Habitat types were split in four categories: bush for bush and brush, heavy thickets and dense copsewoods, coppice for coppice and coppice-with-standards, more open shoulder-high vegetative cover like clear copsewood or mature timbers, mix for bush mixed with coppice and open for meadows and pathways. These four habitat types present in Arc-en-Barrois were drawn and incorporated into a Geographic Information System (GIS; ArcView® 3.2, ESRI, Redlands, California, USA) from a 1/10000 IGN background (Saïd and Gegout 2000). Proportions of each habitat are quite stable among year and range within 69-74% for coppice, 16-23% for bush, 5-11% for mix and 2-3 % for open. All the three species of large ungulates (roe deer, red deer and wild boar) present on the Châteauvillain-Arc-enBarrois forest are hunted. Capture methods Wild boar was abundant with densities up to 9 per 100 ha (Servanty et al., 2007) in Châteauvillain. The use of space by wild board were studied by radiotracking. During nine years study (from 1995 to 2003), 34 different females and 12 different males wild boar have been trapped, weighed sexed and aged before being fitted with ear-tags and Televilt TXH-3 radio collars (Brandt et al., 2004). Single diurnal locations were performed on animals as long as possible during the year round. The number of locations was varying a lot according to the date of hunting (October to January) and to the date of end of monitoring. To avoid problems generated by unequal sample size among periods, we randomly drew between 17 and 61 locations within each period (White and Garrott 1990). We selected individuals with quite balanced numbers of locations between periods. To appreciate the hunting influence on habitat selection, we defined three periods (Janis and Clark 2002): before hunting from July to about mid-October (the date of hunting start changes annually), during hunting from mid-October to February and after hunting from March to June. Because some individuals survived several years we 266 finaly obtained for females 67 “individual-year” before and during hunting period and 49 after hunting period, and for males 14 “individual-year” before and during hunting period and 8 after hunting period. The 7541 daily locations were considered to be independent for each individual in the statistical analyses (see Powell 1996; 2000). We estimated kernel home ranges for each period using the 95% fixed kernel estimator (Worton 1989) with h fixed at 80 corresponding to the mean of h-ref values of all animals and periods. For each period, female wild boar had a different mean home range size (Kernel 95 : before hunting: 247.51 ± 30.13 ha; during hunting: 719.67 ± 92.92 ha, and after hunting: 412.75 ±53.45 ha), and male wild boar had a mean home range size (Kernel 95: before huntings: 830.95 ±198.45 ha; during hunting: 976.47 ±175.85 ha, and after hunting: 800.69 ±198.94 ha) also. Pellerin et al. (2008) showed that fixing h at the same value for all home ranges provides a reliable way to standardize the estimate of home-range size and thereby provides a better way to compare home ranges of different size and number of locations. Describing habitat selection Studies of habitat selection are generally performed by comparing use and availability (Manly 1993; Thomas and Taylor 1990). We studied in our case the third order of ressource selection which is defined as the habitat selection of individuals within their respective home ranges (Johnson 1980). We considered habitat included in a 95% kernel computed indiviually with all annual location (the three periods) as available for a given individual-year (the availability can change between individuals and years). The habitat utilization (use) corresponded to the vegetation type found directly at individual locations. To describe habitat selections we used an Eigenanalysis of Selection Ratios (ESR; (Calenge and Dufour 2006) because we used categorical variables of habitat. This analysis generalizes the White and Garrott statistic (White and Garrott 1990) and the Manly’s selection ratios (Manly et al. 2002) within the framework of eigenanalyses (Escoufier 1987). This method could be viewed as a generalisation of the compositional analysis (Aebischer et al. 1993) when animals do not express the same selection pattern. 267 The Eigenanalysis of Selection Ratios is therefore optimal in habitat selection studies as recommanded by Calenge and Dufour (2006). We used ESR to describe the selection pattern between individuals because this method takes into account individual heterogeneity in contrast to the compositional analysis (Calenge and Dufour. 2006). We applied the ESR within each period to describe seasonal habitat selections patterns. Testing habitat selection We first tested within each period and each sex the hypothesis of identical habitat use for all animals. We computed a standard chi-square on the contingency table containing numbers of locations in each habitat for each individual (Manly et al. 2002). Habitat selection within period were then tested for each sex by summing the White and Garrott chi-square statistic for all individuals (see (Calenge and Dufour 2006). To test significance of habitat changes over seasons we first selected a pool of animals observed during two periods (before-during or during-after). We then summed standard chisquares computed on individual contingency tables containing numbers of locations in each habitat for each period for a given individual (a sum of contigency chi-square, d.f.=Nindividual × (Nhabitats – 1) × (Nperiods - 1)). As some individuals never used some scarce habitats we merged habitats together such as 0 values were avoided. To explore intra-season differences of habitat selection between sexes we performed a non-centred non-normed between-class PCA (Dolédec and Chessel 1987), Chessel et al. 2004). Statistical significance was evaluated by using a Monte-Carlo randomisation test on the between-groups inertia percentage (Romesburg 1985); Chessel et al. 2004) with 9999 iterations. We interpreted the P-value as the strength of the evidence against H0 (no sex difference), conditionally to the data at hand (Edgington, 1987). The whole analysis was carried out using the “ade4” package (Chessel et al. 2004) and the "adehabitat" package (Calenge 2006) for the R (R Development Core Team. 2008. R: A Language and Environment for Statistical Computing. R Foundation for 268 Statistical Computing, Vienna, Austria.) distributed under the GNU General Public License. RESULTS Among the 46 wild boar equipped in the nine years, none disappeared from the study zone. First axis of the ESR explained the quasi totality of the selection ratio variance (97.7% before hunting, 94.0% during hunting, 97.4% after hunting, for first axis). Within each period, habitat selection patterns appeared to be significantly different among individuals (p < 0.001 for all periods). Heterogeneities of selection patterns seemed to come from a differential intensity of habitat selection rather than a selection for differents habitat (Figure 2b: vectors have heterogenous length but were oriented in the same direction). Within seasons and within sexes, diurnal locations were not randomly placed with respect to habitats: we found a significant difference (p<0.0001) in the distribution of habitat use regarding to available (Table 1). Table 1 Results of chi-square test of differential habitat utilisation among individuals and habitat selection regarding to habitat available for each individuals Females Males Before During After Differential Chi2=925.86 Chi2=598.10 Chi2=550.18 utilisation p<0.0001 p<0.0001 p<0.0001 Habitat Chi2=4632.67 Chi2=1917.08 Chi2=2834.81 selection p<0.0001 p<0.0001 p<0.0001 Differential Chi2=102.73 Chi2=120.58 Chi2=38.10 utilisation p<0.0001 p<0.0001 p=0.0014 Habitat Chi2=1258.54 Chi2=674.81 Chi2=625.20 selection p<0.0001 p<0.0001 p<0.0001 269 Plots showed that whatever sexes and periods wild boars selected intensively bush and avoid coppice, mix and open fields (Figures 1, 2). During hunting a general reduction of bush selection seemed to occur (Figures 2a, b: vectors appeared to be shorter during hunting than before and after hunting). Before hunting, 80 % and 82 % of cumulated locations, for females and males respectively, were found in bush. During hunting the proportion of locations found in bush dropped to 54 % and 63 % for females and males respectively. Fig. 1: Study area of ChateauVillain-Arc-en-Barrois and resting sites locations (black points) of wild boars (females and males) before, during and after hunting season in 2000. From white to grey: open, coppice, mix and bush. Before During # ## # ## #### # # ## # ## ### ### ## # ## # ## ## ## ## # ## # # ## ## ## # # # # # # ## ## ## # # # ### # ### # ## # # # ### ## # #### # # ### ##### ### # # # # ## # # # # # # ## ## # # # # # # #### # ## # ## # # # # # # ### # ## # ## # # ## # ######## ## ### # ## ## ### # # # ## #### ### # #### # # # # # # ## # # ## ### # # ### # ### # ## # ## ## # ## # ## # ## # ## # ## ## ### # # ### # ## # # ## # ## ### ## # ## ### ### # # ## # ## ## # # # ## ### ### # # # ## ## # # ## # ## #### ## ## # ## # #### ### # ### # # # ## ## # ### ### #### ## # ## # # ### # ## ### ## # ## ## ## # ##### ##### # ## ## ## ### ## # # # ### # ### # ## # #### #### ### ## ####### ### # ## ## # # # # # ##### # # # # # # # # # # ## ####### ### ## # ## # ##### # ### # ## # ## # ## ### #### # # ## # # ## ## ## # ## #### ##### ## ## ## ## # # ## ## #### ##### ## # #### ## #### # ## # # # ## ## ## #### ## # # #### # ## # ### #### # ## ## # ## ##### ## ## # # # # # # # # # # # # # # # # # # # #### ## ###### ### # # ### ## ## ### # ##### # ## # ### #### # # # ## #### # # # # #### ##### # ### ### # # ## ## # #### # ##### # #### # # ## # ## ## ## ## ### # # ## # # ## ## # # # # # ## # # # # #### ### ### ## #### # # # # # # ## #### ## ## ### # ### # # # ## ## # ## # # # ### # # # # ## ### ## ### ## ### ## # ## ## ## #### ## ####### # ## ### ## ### # ### ## ### # ### ## # # # ## ### # ##### # # ### # # ## ## # # # # ## ### ## ##### # # ### ###### ## ### ### ## # # # ## ### # ## #### # # ## ### # # # ## ## ## ## ## # # # # # ## # # # # # # # ## # ## ### ###### # # ## # ## # # # # ### # ##### ##### # # # ## # # # # ### # # # ## ## ## # # # # ## # ## # # # # # # # ### ## # ## # # ## ### ### # ## # # #### ## ## ## # ### # ## # # # # # # # # # # # # # # # # # ## ## # # # ## # # ## ## # # ## ####### ## # ## # #### # # # ## # ## #### # ## # ### ## ## # # # # # # # ## # # # ### # # # ### ## ## ### # #### #### ## ## # # # # ## ### ### ## ## # # # # ## # ### ### ## # # # # # ## ### ###### # # # # ## ###### # # ### ### ## ## # # # # #### ## # # ##### ## ##### ## ## # # # ## # # ## # # ## ## # ## # ##### # # ##### # # # ### # # # # ## # # # # # ## # ## # # # # ## # # # # # # # # ## # # ### # # # # # ## ## # # # # # # # # ## # ## # # # ## ### ## # #### ### # # # ## # ## ## ## # # ## ## # ## ## ## ## # # # # # ## # ## # # # # # ## # ###### ### ## # ## # # ### # # ## # # ## # ## # # # ## ## ### # # # # # #### # ##### ## ## # # # ## # ##### ## # # ## # # # # ## # # # ## # ## # ## # # # # ## # # # ### # ##### ###### # # # # #### # # # # # # # # # # # # # # ## # # # ## ## ## # # ### # # ## # ### # ### # ## ## # # # ### ## # # ## # # # # # # #### # ## ## # # # ## # # ## # ## # # # # ## # ## # # ## ### ### # ### # # # ## # # # # ## # ## ## # # ## ## ### # # ## ## # ## # # # # # ## # # ## # # ## ## # ## # # #### ## # # # ## # ## # # ## # # ## # # ## # # ### #### ## # # # # # # # # # # # # # # # # # ## # # # # # ## ## # # ## ## ## ## # ## ## ## # # # ## ## # # ## # ## # # # # # # # ## # # # ## # # ## ## ## ## # ## # # # ## ## # # ## ## # # # # # # # # # # # # ## # # # # # # # # # # # # ## # # # ## # # ## # ## # # ## ### # # # # ## # # # # # # # # # # ## # ## # ## # ## #### ## # ### # # ## # # #### # ## # # ## # # ## # # # # # # ### ## # # # # # # ## ##### # # # ### ## # # ## # ## # # # # # # # # ## # # ## # # # # # ## ## # # ## ## ### # ### # ## ##### ### ### #### # ## ### # # ## #### ## # # # # # # ### ## ### ##### ## ## # # # # # # ### # ## # ## ### # # #### # #### ### ## # # # # #### ### ## # ## ## #### # # ### # ## ## # ##### ### # # ## ## # ## ### # ## ##### ## ######## ## ## #### # # # # # ## # # # ### # #### ####### # # # # # # # # # # # # # # # # # ## ### ### # ##### ###### # ######### ##### ### ### # ## ##### #### ### ## ### # # # ### # ### ## # # ## ## # ### # ### ## #### ## # # # # # # ## # ## ### # ## ## ## # # # # # # # # # # # # # # # #### ## # # ## ### # # ## ## ## ####### ## # ###### ## ## ### ##### ### # ## ## #### ## # # ## ## ## ## # # ## ## # ## ### ## ## # ### #### #### ####### # # # ## # ## ### # ### ## ### ### # ## ###### # ## ### ## ## ##### ## ## # ## ## ## # ### # # ### # # # # ## ## # ## # ## ## ### # # ### ##### #### ## # ## # # # ## ## # ## # # # # # # # ## ### # # # # # # # # # # # # # # # ####### # # # ### # ## ### ## ## ## ##### ## # ## ## # ### # ## ##### # ### # ## ## ## # ## # ## ## # ##### # ### ### ## ## # # ## ## ## #### # ## ## ## #### ## ### ### ## ## ## # # # # #### ## ## ### ## #### ## #### # # ## # ## ## # ###### ## # # ### # # # # ### ## # ## # # # # # # ## ## ## # # # # ### # ## # # # # # ## # #### ## ## ### ## ##### # ## ## ## # # # ## #### ### # ## ### # ## ### ## ## ### # # # # # #### ## # # ## # # # # # ###### ### #### # ### # ## ## # # # ## #### # # # # # # # # # ## ## # ## # # # # # # # # # #### # # ## # ## # # ## # # # ## ## ## ## # # ## ## # # # ## # ## # # #### # # ## ## ### ### ## ### # # # # ## ### # ## ## ## # # ## ## ## # ## #### # # # # # # ### ## ## ##### # ## ## ## #### # ## ### ## # # ## ## ### # # #### # ## # # ## # # ## ## ## # # # ## ### ### # # ## ### # # # # # # ## # # # # # # ## # ## # ## # # ## # # ## # ## # # ## ## # ## # ## # # # # # # # # #### # # # # # ## ### # ## ## # # # # ## ## #### # # # # # # # # # #### # # # # # ## # # # # # # # # # # # # ## # ### # # # # # # # # # ## #### # # # # #### # # ## ## # # # # # # # ## ### # # # ## ## After # # # # # ## # ### ## # # # #### # # # ## ## ### # # # ## ## # # # ## # ## # # ## # # ### # # ## # ## # # ## # # # # # # # # # # # # # # # # # # # # # # ## # # # # # #### # # # # # ## # # ## # ## # ## # ## ## # ### # ## ## ## # ## # ## ##### ## ## # # ## # # # # ## # # ## # # # # # # # #### # ### # ## ####### # # # #### # ## # ### # # # # # # # # # # # # # # #### ## # #### # # # ## ### # # # ## # # ### ### ## # # # ## # # ## # # # ######### # # # # #### # ### # ## # ### # # # # ## # # ### ## # # # ## ##### ## #### # ### ### # ## ## ### # # ## #### ### # #### # ## # ### ## # # ## ## ## # #### ### ## # #### #### ### ## # ## # ## # ### # ## ## ## # # # # # ## # ###### ### # # # #### ### # # # # ## # ### # ## ### ## ### ### ### ## ## ## ## ##### #### # # # ### # ### ### # ## # # ### # # #### ## ## ### ### ##### ### # ## ### # ## # ## ### ## ### ## ## ### ### # # # # # ### #### # ## # # ## ## ## # # # # # # # ## ### # # ## # # ## ## ### ## # ## # ## # # ## # # # #### # # # # # # # # # # # # # # # # # ## # ## # # # ## #### ## ## #### ### # ### # # ### ## ## ## # ## # # ## ## #### ### ## # ## ## # ## # # ## ## #### ## ### # #### ## ## ### ##### # # # ## ## ## # # ###### # # # # # # # # # ### # # ## # # ## ## # ## # # # ## # ### # ## # ## ##### # # # # ## ## # # ## # ## # ## # # ### #### #### ## # ## # ### # ## # # # # # # # ###### # ## # # ## # ### # # # # # # # # # # # ## # ##### # # # # # ## ### # #### # ## ## ## ## ## # # # # # # # ## # ### ## ## # ## # # ## ##### # # # # # ## ## # # # ## # # # # ## # # # # # ### # # # # # # # ## ### # # ## # # # # # # ## ##### ### # # # # # # # # # # ## # # # # ### # # # ### # ##### # # # #### # ## ## ## # ## # ### ### ## # # # # # ### # ## ### # ## # # # ### ## # ## # ## # # # ### ### # # # ### ## # # # # ## # ### # ### # # # # # # # # # # # # # # ## # # ## # ### ## # ## # # # # ## # # ## # # # ## ## # # # # # # ## # # ## # # # # ## # ### ## # # # ### # ### # # ## ### # ## ### # # # # # # # #### ### # #### ### # # # # # # # # # # # # # ## # ## ## ## # # # ### # ##### ## # # ## # ## ## # # ## ## ### # ## # #### # ### #### #### # # # ## ## # ### # ## ## # # ### # # # # # # ## ### ## ##### # # # ## # ### # # # ## # # ## ### ### ### # # ## # #### ## # # # ## # # # ## # ## ## # ## # # ## # # ## # # ## # # # ## # ### # # # # ## ## ## # ## # ### # # # # ## # # # # # # ### ## # # # # ## # # # # # # ## # # ### # ## # # # # # # # # ## # ## # # ### ### ## ### # # # # # # # # # # ## ## ## # # # # # ## # # # # # ## ## # # # # # ## ## ### ## # # # # # # # # ## # # # ## # # # # # # ## ## 270 # # # # # # # # Plots of the ESR even showed than few individuals tended to select coppice during hunting (Fig. 2b). After hunting this percentage return to higher values equal to 69 and 72 %. Respective availablitity of bush for female and males range within 11-58% and 12-32% which showed that despite variations among period this habitat was always positively selected (Figures 1, 2). The use of three other habitats followed an opposite patterns especially clear for coppice utilisation: cumulated percentages of locations found in coppice are 16 % and 14 % before hunting, 40% and 31% during hunting and 26% and 26% after hunting for females and males respectively. However coppice availabilities range within 37-78 % and 62-86 % for females and males respectively which showed that despite variations among season this habitat was mainly avoided (Figure 2). Such changes in habitat selection during hunting regarding to before hunting were significant for both sexes (habitat grouping: bush vs. coppice-mix-open; females: chi-square = 506.085, ddl = 54, p < 0.0001; males: chi-square =55.78, ddl = 12, p< 0.0001). After hunting bush selection return to higher value than during hunting for both sexes (Figures 2a, b). Such changes were significant for females and quasi significant for males (habitat grouping: bush vs. coppice-mix-open; females: chi-square = 148.43, ddl = 41, p < 0.0001; males: chi-square = 13.24, ddl = 7, p = 0.07). Within all periods selections for bush versus others habitats seemed higher for males than for females (Figure 2c). Monte-Carlo randomisation tests on between-groups inertia percentage performed on the selection ratios table show that habitat selection was significantly different between females and males only during hunting period (before: p=0.14, during, p=0.015, after, p=0.076). 271 (a) Habitat types coppice bush open mixt (c) sexes After hunting During hunting Before hunting (b) individuals ++ + + + + Fig. 2: Plots of Eigenanalysis of Selection Ratios before, during and after hunting. (a) habitat types on the first two factorial axis. (b) individual scores on the first factorial plane. (c) convex hull (50 and 100%) of individual score depending on sex (full lines: females; dashed lines: males). 272 DISCUSSION We aimed to assess the hunting disturbance effect on habitat selection of wild boar. Our results supported two out of the three hypotheses we tested. Wild boar select as expected dense vegetative cover during its resting periods (prediction (i) supported), but, against expectations, they do not increase this selection in response to hunting disturbance. On the contrary we found a decrease of this selection during the riskier season (prediction (ii) rejected). As expected , females responded more to the hunting disturbance than males leading to more prononced sexual difference during the riskier season (prediction (iii) supported). Some studies have shown that core areas of wild boar home range might be related to scarcely hunted areas (Baubet 1998; Calenge et al. 2003). Some other case, wild boar shift their home range core toward protected area (Maillard, 1998; Tolon et al., 2009). Nevertheless, such changes did not integrate the fine scale habitat choice regarding to the density of the vegetative cover (more or less crowded). We observed in our study, whatever sexes and periods, that wild boar strongly select dense vegetation for their resting sites, whereas other forest type were available. The use of bushes plots while resting is an established feature of wild boar behaviour (Barber and Coblentz 1986; Boitani et al. 1994; Spitz and Janeau 1995; Caley 1997; Merrigi and Sacchi 2000; Welander 2001) as it probably reduces detectability or accessibility for human or other predators. It can be easily explain the high selection for bush all year long the effect of hunting disturbance on the use of these potential refuges was clearly against our expectation. Indeed wild boar decreased their selection for bushes during the riskier seasons, without however totally shifting toward an other habitat. Hunters had probably some expectations on habitat preferences of local wild boar and could intensify their effort near or into bush plots. (Lande et al. 2009). Hunting disturbance could then be higher than expected in bush and the decreasing selection of this habitat during hunting could reflect a spatial avoidance of such disturbance. Such changes could be viewed as an adaptive behavioural response from hunted animals to hunting stress as it have been 273 proved in some other ungulates (Kufeld et al. 1988, for mountain mule deer (Maillard 1998); Baubet 1998; Rosell et al. 2004 for wild boar) and other game species (Fox and Madsen. 1997 ; Madsen, J. 1998). However, our unexpected results could be also interpreted in a more simplistic way: hunting could act as a simple disturbance factor on a system at equilibrium. Indeed animal can express a constant tendency to select dense vegetative cover for resting, but hunting force them to change their resting sites more or less randomly, increasing their probability to relocate in a different habitat. The spacing behaviour of wild boar was shown to change rapidly as hunting season goes along, animals reacting through increasing mobility between resting sites (Maillard and Fournier 1995). This can explain why animal simply decreased their selection for bush without clearly shifting toward others habitats. In this case, the behavioral response to hunting disturbance in term of habitat selection resulted more from a random diffusion process (e.g. Westerberg et al. 2005) due to an increase of movements than an adaptive behavior. Our results indicated that during the hunting period, familial groups and solitary males selected differently their habitats. They adjust themselves by slight changes in their used of the different forested type plots encompassed in their usual home range. Even males showed a partial switch in their habitat selection, groups led by sows indicated a stronger reaction, leaving more clearly bush plots location during hunting. A previous study (Spitz et Janeau 1995) has shown that sows appeared to be more reactive than male in their utilization of different habitats, especially regarding the vegetation coverage This variation of habitat selection between sexes during the high risk season is then not totally surprising and it could be interpreted as a possible difference in regard to anti-predation tactics. In a recent review, Stankowich (2008) states that flight responses to human disturbance in ungulate were highly dependent on group size, sex and age class. Indeed large groups, especially females with offspring exhibit the greatest flight responses to human disturbance. Males wild boar live solitary whereas females lived in matriarcal groups composed by relatives females (sister or daughters) as well as young of the year (Mauget 1980 ; Kaminski et al. 2005; Poteaux et al., 2009). The observed sexual difference might then first results from a differential tactic of both sexes in response to predation. Many authors have stressed that, in ungulates, females with newborns selected 274 secured habitat in order to ensure their survival, whereas males choice is turn to habitat maximizing food availability (e.g. Corti et al. 2002 for Alpine Ibex). Therefore, females should exhibit strong flight responses leading to an increased frequentation (random or not) of secondary habitats whereas males can support more risks and remain hidden in bush plots. However sex and group size are confounded in wild boar and the observed sexual difference might also results from higher flight responses of large groups (Stankowich 2008). Indeed group cohesion can be a high constraint for females especially because of the presence of juveniles. Therefore, all members of a familial groups can have more difficulty to remain hidden at the same place whereas solitary males can easily hide themselves in a single shrub and survive to hunting actions without moving. Moreover, it might be suggested that the synchronization of a group escape movement will be much easier in open habitat than in bush (Jarman, 1974). We could probably argue that cohesion of group could be maintain easily in an open habitat, individuals being more aggregate when flying away, than in bushed formation. However, such shift of habitat used toward coppice plots looked unusual for females wild boar which naturally selected the bushed areas of their ranges contrary to male that more often used open plots (Spitz and Janeau 1995) might be related to the differential vulnerability between sexes toward hunting practise (Toigo and al. 2008). Our study therefore demonstrated how hunting disturbance can lead in a game species to an unexpected decrease of the prefered refuge, and a more pronounced differentiation of sexes amongts habitats. ACKNOWLEDGEMENTS The study was conducted by the Office National de la Chasse et de la Faune Sauvage. I would like to thank Jean-Michel Gaillard, Clément Calenge and François Klein for improving earlier drafts of this paper and their help with the English. 275 REFERENCES Abaigar T., Del Barrio G., Vericad J.R. 1994. Habitat preference of wild boar (Sus scrofa L. 1758) in a mediterranean environment. Indirect evaluation by signs. Mammalia 58(2): 201-210. Aebischer, N. J., Robertson, P. A., Kenward, R. E. 1993. Compositional analysis of habitat use from animal radiotracking data. Ecology, 74, 1313-1325. Apollonio, M; Ciuti, S; Luccarini, S. 2005. Long-term influence of human presence on spatial sexual segregation in fallow deer (Dama dama). Journal of Mammalogy, 86: 937946 Barber D.W. and Coblentz B. E. 1986. Density, home range, habita use and reproduction in feral pigs on santa catalina island. Journal of mammalogy 67(3) : 512-525. Baubet E. 1998. Biologie du sanglier en montagne : biodémographie, occupation de l’espace et régime alimentaire. Thèse de doctorat – UCBL I – 281 p. Baubet E., Brandt S. et Touzeau C. 1998. Effet de la chasse sur les stratégies d'occupation de l'espace des sangliers (Sus scrofa). Analyses préliminaires. Gibier Faune Sauvage, Game Wildl. 15 (hors série tome 2) : 655-658. Baubet, E. L. 1998. Biologie du sanglier en montagne : biodémographie, occupation de l'espace et régime alimentaire. pp. 299. Villeurbanne: University of Lyon 1. Bieber, C., T. Ruf, 2005. Population dynamics in wild boar Sus scrofa: ecology, elasticity of growth rate and implications for the management of pulsed resource consumers. Journal of Applied Ecology, 42:6 1203-1213. Boitani L., Mattei L., Nonis D. and Corsi F. 1994. Spatial and activity patterns of wild boars in Tuscany, Italy. Journal of Mammalogy. 75(3) : 600-612. Brandt S., Vassant J., Baubet E. 2004. Adaptation d’un collier émetteur extensible pour sanglier. Faune Sauvage 263 : 14-19 Calenge, C. & Dufour, A. B. 2006. Eigenanalysis of selection ratios from animal radiotracking data. Ecology, 87, 2349-2355. 276 Calenge, C. 2006. The package adehabitat for the R software: A tool for the analysis of space and habitat use by animals. Ecological Modelling 197:516-519 Caley P. 1997. Movements, activity patterns and habitat use of feral pigs (Sus scrofa) in a tropical habitat. Wildlife Research. 24 : 77-87. Cargnelutti B., Gerard J-F.,Spitz F. Valet G. and Sardin T. 1990. occupation par le sanglier (Sus scrofa) des habitats d’un agro-écosystème modifié lors de la mécanisation de l’agriculture. Gibier Faune Sauvage 7 : 53-66. Chessel, D., Dufour, A.B. & Thioulouse, J. 2004: The ade4 package - I: One-table methods. - R News 4: 5- 10. Chessel, D., Dufour, A.B. & Thioulouse, J. 2004: The ade4 package - I: One-table methods. - R News 4: 5- 10. Conner M.M., White G.C. and Freddy D.J. 2001. Elk movement in response to earlyseason hunting in northwest colorado. Journal of Wildlife Management. 65(4) :926-940. Cruz F., Donlan C.J., Campbell K. and Carrion V. 2005. Conservation action in the Galapagos: feral pigs (Sus scrofa) eradication from Santiago Island. Biological conservation 121 : 473-478 D’Andrea L., Durio P., Perrone A., Pirone S. 1995. Preliminary data of the wild boar (Sus scrofa) space use in mountain environment. IBEX Journal of Mountain Ecology 3 : 117-121. Dardaillon, M. 1986. Seasonal-variations in habitat selection and spatial-distribution of wild boar (sus-scrofa) in the camargue, southern France. Behavioural processes, 13: 251268. Dolédec, S. & Chessel, D. 1987. Rythmes saisonniers et composantes stationnelles en milieu aquatique I- Description d'un plan d'observations complet par projection de variables. Acta Oecologica, Oecologia Generalis, 8, 403-426. Edgington, E.S. 1987: Randomization tests. – Marcel Dekker, New York, NY, 341 pp. Escamilla A., Sanvicente M., Sosa M. and Galindo-Leal C. 2000. Habitat mosaic, wildlife availability, and hunting in the tropical forest of Calakmul, Mexico. Conservation Biology. 14(6) :1592-1601. 277 Escoufier, Y. 1987. The duality diagram: a means of better practical applications. In: Development in numerical ecology (Ed. by Legendre, P. & Legendre, L.). Berlin, Germany: Springer Verlag. Fox A. D., and Madsen J. 1997. Behavioural and distributional effects of hunting disturbance on waterbirds in Europe: Implications for refuge design. Journal of Applied Ecology 34:1-13. Fretwell, S. D. & Lucas, H. L. J. 1970. On territorial behavior and other factors influencing habitat distributions in birds. Acta Biotheoretica, 19, 16-36. Gaillard JM, Vassant J., Klein F., 1987. Quelques caractéristiques de la dynamique des populations de sangliers (Sus scrofa) en milieu chassé. Gibier Faune Sauvage 4: 31-47. Gell, F. R. & Roberts, C. M. 2002. The Fishery Effects of Marine Reserves and Fishery Closures. WWF-US, 1250 24th Street, NW, Washington, DC 20037, USA. Houston, A. I., McNamara, J. M. & Hutchinson, J. M. C. 1993. General results concerning the trade-off between gaining energy and avoiding predation. Philosophical Transactions: Biological Sciences, 341, 375-397. Ihaka, R., et R. Gentleman. 1996. R: A Language for Data Analysis and Graphics. Journal of Computational and Graphical Statistics 5:299-314. Janis M.W. and Clark J.D. 2002. Response of forida panthers to recreational deer and hog hunting. Journal of Wildlife Management. 66(3) :839-848. Jarman. P.J. 1974. The social organisation of Antelope in relation to their ecology. Behaviour, 48(3/4) : 215-267. Johnson, D. H. 1980. The comparison of usage and availability measurements for evaluating resource preference. Ecology, 61, 65-71. Kaminski, G., Brandt, S., Baubet, E. and Baudoin, C. 2005. Life-history patterns in female wild boars (Sus scrofa): mother-daughter postweaning associations. Canadian Jounal of Zoology, 83, 474-480. Kilgo J.C. Labisky R.F. and Fritzen D.E. 1998. Influences of hunting on the behavior of white-tailed deer : implications for conservation of the Florida panther. Conservation Biology 12(6) :1359-1364. 278 Kufeld R.C., Bowden D.C and Schrupp D.L. 1988. Influence of hunting on movements of female mule deer. Journal of Range Management. 41(1) :70-72. Lande U. S., Herfindal I. , Finne M. H. and Kastdalen L.. 2009. Use of hunters in wildlife surveys: does hunter and forest grouse habitat selection coincide ? . European Journal of Wildlife Research Lopes M.A. and Ferrari S.F., 2000. Effects of human colonization on the abundance and diversity of mammals in eastern Brazilian Amazonia. Conservation Biology. 14(6) :16581665. Madsen, J. 1998. Experimental refuges for migratory waterfowl in Danish wetlands. II. Tests of hunting disturbance effects. Journal of Applied Ecology 35:398-417. Maillard D. 1998. Approche du fonctionnement de la population de sangliers (Sus scrofa L.) de la Réserve Naturelle de Roque-Haute à partir des résultats scientifiques obtenus sur l'espèce en milieu méditérranéen. Ecologia mediterranea. 24(2), 223-234. In French with English abstract. Maillard, D and Fournier P. 1995. Effects of shooting with hounds on size of resting range of wild boar (Sus scrofa L.) groups in mediterranean habitat. IBEX Journal of Mountain Ecology 3: 102-107. Maillard, D. 1998. Approche du fonctionnement de la population de sangliers (Sus scrofa L.) de la Réserve Naturelle de Roque-Haute à partir des résultats scientifiques obtenus sur l'espèce en milieu méditerranéen. Ecologia Mediterranea, 24, 223-234. Manly, B. F. J., McDonald, L. L., Thomas, D. L., MacDonald, T. L. & Erickson, W. P. 2002. Resource selection by animals. Statistical design and analysis for field studies. London, UK: Kluwer Academic Publisher. McIlroy, J. C. and Saillard, R. J. 1989. The effect of hunting with dogs on the numbers and movements of feral pigs, Sus scrofa, and the subsequent success of poisoning exercises in Namadgi National Park, A.C.T. Australian Wildlife Research, 16, 353-363. Millspaugh, J. J., Brundige, G. C., Gitzen, R. A. & Raedeke, K. J. 2000. Elk and hunter space-use sharing in South Dakota. Journal of Wildlife Management, 64, 994-1003. Pellerin, M. Saïd S. and J-M. Gaillard 2008: Roe deer Capreolus capreolus home-range sizes estimated from VHF and GPS data. - Wildl. Biol. 14: 101-110. 279 Poteaux, C. Baubet, E. Kaminski, G. Brandt, S. Dobson, F.S. Baudoin C. 2009. Sociogenetic structure and mating system of a wild boar population. Journal of Zoology 278 :116-125. R Development Core Team. 2008. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Romesburg, H. C. 1985. Exploring, confirming and randomization tests. Computers and Geosciences, 11, 19-37. Rosell C., Navas F., Romero S. and De Dalmases I. 2004. Activity patterns and social organization of wild boar (Sus scrofa, L.) in a wetland environment : preliminary data on the effect of shooting individuals. In: C. Fonseca, J. Herrero, A. L. and A.M.V.M. Soares (eds.). Wild boar research 2002. A slection and edited papers from the “4th international Wild boar symposium” Galemys 16(ne) : 157-166. Russo L., Massei G. and Genov P.V. 1997. Daily home range and activity of wild boar in a Mediterranean area free from hunting. Ethology, Ecology and Evolution. 9:287-294. Saïd S. and Gegout J.C. 2000. Using the age of the oldest woody specimen for studying post-pasture successions in Corsica (Mediterranean island), Acta Oecologica 21: 193−201. Santos P. Mexia de Almeida L. and Pretucci-Fonseca F. 2004. Habitat selection by wild boar Sus scrofa L. in Alentejo, Portugal In: C. Fonseca, J. Herrero, A. L. and A.M.V.M. Soares (eds.). Wild boar research 2002. A slection and edited papers from the “4th international Wild baor symposium” Galemys 16(ne) : 167-184. Santos, P., L. Mexia-de-Almeida & F. Petrucci-Fonseca, 2004. Habitat selection by Wild Boar Sus scrofa L. in Alentejo, Portugal. Galemys, 16: 167-184. Sodeikat G. and Pohlmeyer K. 2003. Escape movments of family groups of wild boar Sus scrofa influenced by drive hunts in Lower Saxony, Germany. Wildlife Bilogy. 9 (1) : 43- 49. Sodeikat G., K. Pohlmeyer. 2007. Impact of drive hunts on daytime resting site areas of wild boar family groups (Sus scrofa l.) Wildl. Biol. Pract., July 3(1): 28-38. Spitz, F. & Janeau, G. 1995. Daily selection of habitat in wild boar (Sus scrofa). The Zoological Society of London, 237, 423-434. 280 Stankowich, T. 2008. Ungulate flight responses to human disturbance: A review and meta-analysis. Biological Conservation 141: 2159-2173. Toigo C. Servanty S. Gaillard JM. Brandt S. and Baubet E. 2008. Disentangling natural from hunting mortality in an intensively hunted wild boar population. Journal of Wildlife Management 72,7 : 1532-1539. Tustin K.G and Challies C.N. 1978. The effect of hunting on the numbers and group sizes of Himalayan thar (Hemitragus jemlahicus) in Carneys creeks, rangitata catchment. New Zealand Journal of Ecology 1:153-157. Westerberg, L; Ostman, O; Wennergren, U (2005) Movement effects on equilibrium distributions of habitat generalists in heterogeneous landscapes. Ecological modelling 188: 432-447. White, G. C. and Garrott, R. A. 1990. Analysis of wildlife radio-tracking data. London, UK: Academic Press. 281