Optimizing scale search in species distribution models

The influence of spatial scale on ecological processes and pattern formation, such as species distribution is a major research topic since decades. It has become even more relevant in the context of global change. In many studies the influence of a predictor on a response derived over multiple and discrete spatial scales is evaluated. Due to inherent issue of multiple testing, this approach can be problematic. In this study on bonobos distribution, we suggest a 3-steps procedure that overcomes this problem. It takes into account the decay of a predictor by using a weighting function of distance to the observation of interest. We use variables to account for human pressure, food availability, patchy structure of the forest and nesting sites re-use. In a first step, we run a model with a fixed scale for every predictor based on expert opinion. For significant variables, we derive then in a second step the distance weighted influence over a range of scales. This helps to narrow down the search for the final model parameter estimates. Findings indicate that bonobo distribution is driven on the intermediate scale by forest patches structure. Food availability explains their abundance only at smaller scales. Those results demonstrate the sadly well-known influence of habitat fragmentation on animals’ density and distribution but also highlight the importance to understand influences of scale, the animal perceptions of their environment, by using appropriate statistical procedures. Our method can be particularly useful to formulate specific management hypotheses for conservation. Furthermore, its principles can be of use to other types of studies, such as behavior.