| Forest-dwelling woodland caribou are considered threatened in Canada. The main hypothesis to the decrease of populations is the intensification of predation provoked by anthropogenic perturbations in the landscape. To deal with this situation, it is necessary to study and to understand the impact of the environment on the predator-prey interactions between the caribou and the wolf, and with the moose, its main alternative prey. In this perspective, this thesis presents the design of an individual-based model of displacements of these three species with respect to their environment, from which the predator-prey interactions will result. To allow for the application of this model for long periods of time, i.e. for changing environments, a methodology has been developed, which is based on two key points. First, the notion of levels of emergence is introduced, allowing to order the different observable comportments of the system according to their inter-dependencies, to help with choosing a trait of the model corresponding to the intended domain of applicability. Ordering the comportments according to their levels of emergence also permits to identify redundancy between patterns, which can lead to over-fitting when they are used for calibration. Second, a new methodology for calibration and validation of the trait(s) chosen by means of the levels of emergence framework is also presented, named Individual Based System Replication (IBSRtion). This protocol emphasizes forward modelling, contrary to the main existing methodology, Pattern Oriented Modelling (POM), and allows to use an empirical approach by artificially generating data that are unavailable or that cannot be obtained by means of field studies. IBSRtion can also be integrated into POM, to contribute to the establishment of a universal methodology for the design of individual based models. The design process of this model allowed for a synthesis of existing knowledge and to point out some gaps. More specifically, a study conducted to deal with the lack of sufficient information on the short-term response of prey to predation risk allowed to observe that their behavioural responses to prior presence of predators are a combination of chronic and ephemeral behaviours, and that the mechanisms that produce them are complex and non-linear. The outcome of this work is a complex model, using many sub-models, and calibrated in an empirical fashion, that can be applied to a wide variety of environments. This model allowed to test the impact of the encroachment of deciduous trees on predator-prey relations. Simulations have been run for different quantities of encroachment, according to two different spatial configurations. Simulation results suggest that management plans taking into account the moose's habitat might benefit woodland caribou, because they could increase spatial segregation between the two species, and thus between caribou and wolf. Coupling this model with a module of birth and natural death, along with a model of landscape transformation, would allow the assessment of the impact of different forest management plans on the viability of woodland caribou populations. |
