Understanding the link between environmental factors such as disturbance events, land cover, and soil productivity to spatial variation in animal distributions and vital rates is fundamental to population ecology and wildlife management. The Longleaf pine (Pinus palustris; hereafter, LLP) ecosystem is an archetypal fire-mediated ecosystem, which has seen drastic reductions in land area due to fire suppression. Current restoration utilizes prescribed fire and hardwood removal, but more research is needed to understand the influence of these restoration efforts on the wildlife that exist in that ecosystem. As such, we conducted field surveys on Camp Blanding Joint Training Center and Wildlife Management Area to understand how fire influences relative abundances of mammalian predators, occupancy and species richness of avian species, guilds and communities, and vital rates of white-tailed deer (Odocoileus viginianus) fawns. Our results indicated that mammalian predator space use, and avian species richness were influenced by fire and land cover. Mammalian predator space use was altered by fire conditions and land cover. This mechanism may support predator management strategies that utilizes commonly management techniques for the restoration and conservation of the LLP ecosystem to indirectly alter predator distributions, which has the potential to positively affect the management of important species within this ecosystem. Some mammalian mesocarnivores historically common throughout the southeastern United States were rarely detected, suggesting more research is needed to identify the cause of the potential decline in mesocarnivores in the Southeastern United States. Avian species richness at the community level was positively influenced by the heterogeneity of post fire conditions, or pyrodiversity. Avian species richness of the cavity nesting guild was negatively influenced by increasing time-since-fire. Our results suggest that managers can promote avian community diversity by reducing the size of burn units to create areas with multiple adjacent burn units, with unique fire histories and a mosaic of post-fire conditions. Lastly, fawn recruitment was greater on the higher productivity site than the low productivity site on CB. However, within sites soil productivity did not have a demonstrable effect. In fact, we observed differences between sites, but did not observe any effects of covariates on spatial variation in density or survival of fawns within sites. Although we did not explicitly test the factors influencing our parameters between sites, we hypothesize that the variation in coyote activity rates as well as soil productivity and its subsequent effects (i.e. forage availability, concealment cover, and land cover type) likely drove the differences we saw between sites. These results are relevant to local managers and provide support for unit-specific, deer management on CB. In conclusion, understanding the influence of fire in a frequently burned landscape allows us to better inform management of predators and avian communities using prescribed burns, and the differences in deer populations between areas allowed us to better in inform managers on harvest quotas so that the magnitude of the effect of harvest can better match the population vital rates of each area. / Master of Science / Understanding the link between environmental factors such as fire, land cover and soil productivity is essential for wildlife managers to maintain healthy wildlife populations. The Longleaf pine (Pinus palustris) ecosystem requires frequent fire and has seen drastic reductions in land area due to fire suppression. Current restoration utilizes prescribed fire, controlled burning of an area, and hardwood removal, logging hardwood trees such as oaks, but more research is needed to understand how this restoration influences the wildlife in the longleaf pine ecosystem. As such, we collected data collected from Camp Blanding Joint Training Center and Wildlife Management Area to understand how fire influences the relative numbers of mammalian predators, the distribution and species richness (i.e. number of unique species) of avian species, guilds and communities, and vital rates (i.e. births, survival to a certain age) of white-tailed deer fawns. Our results indicated that mammalian predator distributions, and avian species richness were influenced by fire and land cover. Mammalian predator space use was altered by fire conditions and landcover, which supports a predator management strategy that utilizes prescribed burning and hardwood removal used in restoration and conservation of the LLP ecosystem to indirectly alter predator distributions. Some mammalian mesocarnivores (i.e. foxes, skunks, weasels, etc.) historically common throughout the southeastern United States were rarely detected, suggesting more research is needed to identify the cause of the potential decline of cryptic mesocarnivores in the Southeastern United States. Avian species richness, number of unique species, at the community level was positively influenced by pyrodiversity, the number of unique burn years in an area. This supports and further extends the 'pyrodiversity begets biodiversity' hypothesis for avian species, which states that greater pyrodiversity increases the diversity of bird species in that area. Avian species richness of cavity nesting birds decreased with increasing time since fire. Our results suggest that managers can promote avian community diversity by reducing the size of burn units to create areas with multiple adjacent burn units, with unique fire histories and a mosaic of post-fire conditions. Lastly, fawn recruitment was greater on the higher productivity site than the low productivity site, however, within sites soil productivity did not seem to influence birth and recruitment. Although we did not statistically test the factors influencing our parameters between sites, we hypothesize that the variation in coyote activity rates as well as soil productivity and its subsequent effects (i.e. forage availability, concealment cover, and land cover type) likely drove the differences we saw between sites. These results are relevant to local managers and provide support for managing deer differently across both sites. In conclusion, understanding the influence of fire in a frequently burned landscape allows us to better inform management of predators and avian communities using prescribed burns, and the differences in deer populations between areas allowed us to better in inform managers on harvest quotas so that the magnitude of the effect of harvest can better match the population vital rates of each area.

En Europe, la présence de grands carnivores dans des paysages anthropisés entraîne une forte compétition avec l'homme et alimente d'importantes polémiques concernant leur protection légale. La perception antagoniste de ces espèces à la fois emblématiques pour certains et sources de conflits pour d'autres, rend la gestion de leurs populations très délicate. Depuis la recolonisation spontanée du loup (Canis lupus) dans les Alpes françaises au début des années 1990, la population s'est accrue numériquement et spatialement. Parallèlement, les dégâts occasionnés par le loup sur la filière élevage ont suivi la même tendance. L'Etat met en place aujourd'hui un contrôle raisonné de la population, sous réserve que les objectifs de conservation, exigés par la Directive Habitat, soient respectés. En s'inspirant du cas d'étude du loup en France, nous proposons dans cette thèse un cadre de prise de décision structurée adapté pour la gestion et la conservation d'espèces rares et difficilement observables, protégées par des accords législatifs mais qui, dans un contexte social conflictuel, peuvent être régulées. La modélisation séquentielle du processus décisionnel s'est déroulée dans un contexte de forte incertitude selon plusieurs étapes : 1) appréhender les objectifs de conservation et/ou contrôle du loup en France pour les formaliser sous forme mathématique via une fonction d'utilité, 2) suivre la population par une méthode non-invasive pour définir des indicateurs de gestion fiables et évaluer le statut de conservation de la population, 3) coupler les mesures létales adoptées actuellement à un modèle démographique décrivant la dynamique du loup et intégrant sa structure sociale, 4) et déterminer la décision. Cette dernière étape est réalisée à l'aide d'une méthode d'optimisation qui calcule la stratégie optimale de gestion en fonction de la structure sociale de la population et des différentes sources d'incertitude accumulées à chaque étape du processus décisionnel. Nous avons choisi comme indicateur de gestion le taux de croissance, à partir duquel nous avons défini l'utilité. Cet indicateur était robuste à l'incertitude d'échantillonnage émergeant de la détection partielle et hétérogène des individus. Des analyses de sensibilité de la décision ont montré une forte influence de la fonction d'utilité sur la stratégie optimale, soulignant ainsi l'importance de définir correctement les objectifs. Nous avons également montré que la stratégie optimale était sensible aux variations des paramètres démographiques, montrant ainsi l'intérêt des méthodes de capture-marquage-recapture pour les estimer correctement. Nous discutons enfin de l'extension de notre approche à un cadre décisionnel de gestion adaptative pour traiter des problèmes de conservation dans un contexte conflictuel. / Large carnivore management in Europe is controversial because of conflictive objectives arising from the legal protection of threatened species vs. the possible necessity of culling individuals to prevent severe damages on human activities. Since the wolf recovery in the French Alps in the early 90's, the population has been numerically and spatially increasing. In parallel, livestock depredations have been following the same trend. As an EU member state, France is bound to the European Habitat Directive, which provides full protection of wolf populations and their habitat. Nevertheless, derogatory killings are allowed for individuals causing problems on livestock and some lethal control is now incorporated into the national management plan, as long as the population growth and its distribution range are not being threatened. Illustrating with the case study of the wolf in France, my dissertation proposes a structured decision making framework for the management and the conservation of elusive species that are legally protected but, in a conflictive context, are subject to population control. The sequential modeling of our decision process occurred in the following steps: 1) define the multiple objectives and formulate them in terms of a utility function, 2) monitor the population through a non-invasive approach in order to define the population conservation status, 3) build a demographic model to predict the consequences of harvesting on population dynamics and social structure, 4) obtain optimal state-dependent decisions. The last step is done with stochastic dynamic programming (SDP), acknowledged to be one of the most useful optimization methods in decision making. We provide an optimal solution for wolf management that gives the highest chance of meeting objectives, defined on population growth rate. This demographic indicator was found to be robust to sampling uncertainty arising from partial and heterogeneous detection of individuals. We ran decision sensibility analyses and found a strong effect of the utility function on the optimal strategy, highlighting the importance of defining explicit objectives. We also found that the optimal strategy was sensitive to demographic parameters, which demonstrate the general need of using solid statistical approaches to estimate them properly. This structured decision making framework can further be extended to adaptive management, acknowledged as being a convenient framework for wildlife management.

La région méditerranéenne, hot-spot de biodiversité avec un fort taux d'endémisme, est classée parmi les zones les plus touchées par le changement climatique. La conservation des espèces nécessite de comprendre finement leur démographie face à ces changements, mais aussi leur capacité à s’adapter à ces nouvelles conditions. Au cours de cette thèse, deux espèces végétales rares méditerranéennes ont été étudiées, Centaurea corymbosa et Brassica insularis, en utilisant des modèles mathématiques récents pour analyser des suivis démographiques de long-terme (22 ans et 18 ans). Dans la première partie de cette thèse, des modèles de projection matriciels ont été construits afin d’analyser la variation spatio-temporelle des taux d’accroissement des populations de C. corymbosa. Cela a permis d'identifier les facteurs climatiques clefs qui impactent les taux d’accroissement des populations, et plus finement, les paramètres de survie, de floraison et de fécondité. Une analyse de viabilité des populations a été réalisée sous différents scénarios climatiques. Dans un second temps, une analyse de la capacité des populations de C. corymbosa à répondre à ces changements climatiques a été effectuée avec un modèle de projection intégral. Les changements des traits d’histoire de vie au cours du temps ont montré que la stratégie de floraison observée répond aux variations climatiques en suivant la même direction que la stratégie optimale. Ceci montre que les populations sont capables de s’adapter au changement climatique. Enfin, des modèles de capture-recapture (CR) ont été construits afin de déterminer l’influence de la faible détectabilité des individus sur l’estimation des paramètres démographiques chez B. insularis. Les probabilités de survie estimées par l’approche classique sont inférieures à celles estimées par les modèles CR. Les paramètres estimés par les deux approches sont en général très corrélés, à l'exception d'une population montrant des problèmes d'identification des individus. L'ensemble de cette thèse est replacé dans le cadre de la biologie de la conservation des espèces végétales, notamment en suggérant de prendre en compte leur capacité de réponse au changement climatique. / Mediterranean region is a biodiversity hot-spot with a high endemism rate and is classified among the areas most sensitive to climate change. Deep understanding of demography and evolution following these changes is a necessity for species conservation. During this thesis, two rare Mediterranean plant species have been studied, Centaurea corymbosa and Brassica insularis, using recent mathematical models to analyse long term demographic datasets (22 and 18 years). In the first part of this thesis, matrix projection models have been used to analyse spatio-temporal variation in population growth rate in C. corymbosa. Major climatic parameters that impact population growth rate and survival, flowering and fecundity have been identified. A population viability analysis has been performed under various climatic scenarios. In a second part, an analysis of the population ability of C. corymbosa to response to these climatic changes has been performed using integral projection models. Temporal variation of life-history traits showed that the realized flowering strategy varied with climatic conditions following the same direction than the optimal flowering strategy. It can thus be concluded that the studied populations can adapt to climatic changes. Finally, capture-recapture models have been applied on the Brassica insularis dataset to analyse the effect of plant detectability on demographic parameter estimations. Survival probabilities obtained with the classical approach are lower than the ones estimated using the CR models. However, the parameters estimated using both methods are highly correlated, except for one population where individual identification is problematic. Results of this thesis are linked to some considerations in conservation biology of plants, particularly in link to their ability to respond to climatic changes.

Made available in DSpace on 2016-06-02T20:06:10Z (GMT). No. of bitstreams: 1 6083.pdf: 1151427 bytes, checksum: 24c39bb02ef8c214a3e10c3cc5bae9ef (MD5) Previous issue date: 2014-03-14 / Financiadora de Estudos e Projetos / In this work, we consider the estimation of the number of errors in a software from a closed population. The process of estimating the population size is based on the capture-recapture method which consists of examining the software, in parallel, by a number of reviewers. The probabilistic model adopted accommodates situations in which reviewers are independent and homogeneous (equally efficient), and each error is an element that is part of a disjoint partition in relation to its detection probability. We propose an iterative process to obtain maximum likelihood estimates in which the EM algorithm is used to the nuisance parameters estimation. The estimates of population parameters were also obtained under the Bayesian approach, in which Monte Carlo on Markov Chains (MCMC) simulations through Gibbs sampling algorithm with insertion of latent variables were used on the conditional posterior distributions. The two approaches were applied to simulated data and in two real data sets from the literature. / Neste trabalho, consideramos a estimação do número de erros em um software provenientes de uma população fechada. O processo de estimação do tamanho populacional é baseado no método de captura-recaptura, que consiste em examinar o software, em paralelo, por certo número de revisores. O modelo probabilístico adotado acomoda situações em que os revisores são independentes e homogêneos (igualmente eficientes) e que cada erro é um elemento que faz parte de uma partição disjunta quanto à sua probabilidade de detecção. Propomos um processo iterativo para obtenção das estimativas de máxima verossimilhança em que utilizamos o algoritmo EM na estimação dos parâmetros perturbadores. As estimativas dos parâmetros populacionais também foram obtidas sob o enfoque Bayesiano, onde utilizamos simulações de Monte Carlo em Cadeias de Markov (MCMC) através do algoritmo Gibbs sampling com a inserção de variáveis latentes nas distribuições condicionais a posteriori. As duas abordagens foram aplicadas em dados simulados e em dois conjuntos de dados reais da literatura.