Using Gaussian Bayesian Networks to disentangle direct and indirect associations between landscape physiography, environmental variables and species distribution

•We model the associations between topography, local climate and species distributions.•Temperature, soil moisture and solar radiation (local climate) are associated with topography.•Species distributions are associated with temperature, soil moisture and solar radiation.•Topography is indirectly associated with species distribution through local climate.•Separating direct versus indirect associations improves our understanding on species distribution. Landscape physiography affects temperature, soil moisture and solar radiation. In turn, these variables are thought to determine how species are distributed across landscapes. Systems involving direct and indirect associations between variables can be described using path models. However, studies applying these to species distribution modelling are rare. Bayesian Networks are path models designed to represent associations across observed variables. Here, we demonstrate the use of Bayesian Networks to disentangle the direct and indirect associations between landscape physiography, soil moisture, solar radiation, temperature and the distribution patterns of four plants at their northern range limit in Sweden. Fine scale variations in maximum temperatures were associated with variations in elevation, distance to coast and solar radiation. In contrast, fine scale variations in minimum temperature were associated with distance to coast, cold air drainage and soil moisture. These associations between landscape physiography and minimum and maximum temperature were predicted, furthermore, to be associated with growing season length, growing degree day and ultimately species distributions. All species were indirectly associated with aspect through their responses to either solar radiation or temperature. The models demonstrated strong indirect associations between landscape physiography and species distributions. The models suggested that local variation in light can be as important as temperature for species distributions. Disentangling the direct and indirect associations between landscape physiography, environmental variables and species distribution can provide new and important insights into how landscape components are linked to species distributions.