The combined use of remote sensing and spatial modelling for animal movement - Application to the study of wildlife/livestock contacts and the risk of pathogen transmission in Southern Africa

Interface areas in Southern Africa that consist of communal lands located at the edge of protected areas face a growing number of human/wildlife coexistence related issues and among them, the risk of pathogen transmission between wild and domesticated species.In this context, the present thesis aims to 1) Characterize the environmental variables, at a landscape scale at three different interfaces located in Southern Africa (hwange National park, Gonarezhou National Park, North Kruger National Park), that potentially influence the movements of two focal species, one wild ungulates species (the buffalo – Syncerus caffer caffer) and one domesticated ungulates species (the cattle – Bos taurus & Bos indicus) 2) Simulate the movements of the focal species, at the individual and herd scales, in relation with their respective environments, 3) Determine the nature, frequency and localization of the contacts between the focal species to better apprehend the risks of pathogen transmission.A temporal series of Sentinel-2 satellite images have been classified to produce monthly surface water and landcover maps at 10 meters spatial resolution. These environmental variables have then been integrated into a spatialized mechanistic movement model based on a collective motion of self-propelled individuals to simulate buffalo and cattle movements and contacts in response to the surface water seasonality and the type of landcover. To spatialized the movements and contacts models, the domain specific language Ocelet has been used. Telemetry data collected in previous studies have been used as reference data to design, calibrate and validate the movement and contact models.Results highlighted strong space and time variabilities of water availability in the three study areas. Landcover classified maps accurately reproduced the specificities in landscape compositions of the three study areas. By only taking surface water into account, the mechanistic movement models showed a positive and significant correlation between observations/simulations movements and space-use of buffalo's and cattle herds despite overestimating the presence of buffalo individuals at proximity of the surface water. The addition of the landcover increased the overall accuracy of the movement models. Contacts patterns and their accuracies have been reproduced but their respective accuracy differed according to the study area.Overall, combining remote sensing and spatial modeling offers possibilities to develop simple models to simulate animal movements and contacts in direct relation with the environment. This methodology has the advantage of being scalable and reproducible. It subsequently offers the possibility of integrating an epidemiological model to estimate the risk of pathogen transmission in direct link with the movements and contacts of animal species.