Human activities are resulting in many land-use changes, particularly due to urbanisation and intensification of agricultural practices. Because of these changes, in addition to climate change, many species are facing habitat degradation. In order to avoid extinction under these conditions, they can either move to more favourable areas or adapt to their new environment. To develop appropriate conservation measures, it is essential to identify vulnerable populations that may not be able to disperse or adapt. In this context, modelling tools can be used to predict the potential impact of environmental changes on species and populations. However, to date, few approaches take into account the degree of exposure, the possibility of dispersal and the potential for adaptation. In this thesis, we present modelling approaches based on geo-environmental data to integrate these three elements. First, we use ecological niche models to estimate the distribution of suitable habitats for a given species as a function of environmental conditions. We propose an improvement of commonly used models by developing an approach to integrate spatio-temporal variability of environmental predictors. In addition, we develop a nested model to predict the distribution of vector-borne pathogens. This model can be used to identify populations that may be threatened by an increasing presence of pathogens in their habitat. We use it to model the evolution of the distribution of Ixodes ricinus ticks and their Chlamydiales bacterial pathogen. We then use landscape graphs to analyse the connectivity between habitats and estimate the possibilities for threatened species to move to more favourable areas. Connectivity is also essential for maintaining gene flow and genetic diversity, which is necessary for greater adaptive capacity. We propose here an approach combining landscape graphs, simulations and empirical genetic data to identify the impact of reduced connectivity on population persistence and genetic diversity. We use it to identify butterfly populations that are threatened by increasing fragmentation in an urban landscape. Finally, we develop the concept of "Spatial Areas of Genotypes Probability" (SPAG) to better analyse the adaptive potential of populations. SPAGs make it possible to model the probability of finding locally adapted genetic variants in a given territory, as well as to identify vulnerable populations lacking in genetic variants that would favour adaptation to future climate conditions. We use it to highlight populations of Moroccan and European goats that are poorly adapted to the climatic conditions predicted under a climate change scenario for 2070 (strong variations in precipitation or increased drought). To conclude, we show how the three modelling approaches presented can be combined and integrated into a more general conservation framework to identify vulnerable populations facing high exposure to environmental changes, low dispersal possibilities and reduced adaptive capacity.