Spatial mechanistic modelling to simulate movements and contacts between wildlife and livestock in Southern Africa

The open interfaces between protected areas and rural communal lands in southern Africa are characterized by semi-arid savannas where wildlife-livestock interactions vary in frequency and intensity. In a context of increasing anthropization of land and trans-frontier conservation, the multiplication of these interactions may facilitate human-wildlife coexistence such as competition for natural resources, livestock predation, crop destruction by wildlife, and/or the risk of pathogen transmission between wild and domestic species. To better understand potential contacts between domestic and wild animals at these wildlife/livestock interfaces, we developed a method combining remote sensing and spatial modelling to simulate the movements of African buffalo (Syncerus caffer) and domestic cattle (Bos taurus, Bos indicus). Satellite-derived maps of surface water and vegetation, the primary determinants of movement for these ungulate species, were integrated into a mechanistic and stochastic model of collective movements of individuals interacting according to group cohesion and alignment. This model allowed simulations of herd movements and the location of contact areas with their seasonal dynamics in space and time at the periphery of three national parks in Zimbabwe and South Africa. Model outputs were compared to Global Positioning Systems collar location data of 32 individuals (14 buffalo and 18 cattle). The modelled results show a high spatial and seasonal variability of contacts between buffalo and cattle in the three study sites, and a landscape scale correspondence between spatial extensions of the modelled and observed contact areas. These results illustrate the potential of spatial modelling combined with remote sensing to generically simulate animal movements and contacts at landscape scale while providing opportunities to explore the management of these wildlife/livestock interfaces through, for example, a further coupling with epidemiological modelling.