Predicting species distribution combining multi-scale drivers

Abstract

Species Distribution Models (SDMs) are often used to predict the potential range of invasive species. Unfortunately, most studies do not evaluate variables relevance before selecting them to fit their models. Moreover, multiple variables such as climate and land use may drive species distribution at different spatial scales but most studies either use a single type of drivers, or combine multiple types without respecting their operating scale. We propose a three steps framework to overcome this limitation. First, use SDMs to select the most relevant climatic variables to predict a given species distribution, at continental scale. Then, characterize the species-habitat relationships, at a local scale, to produce species and area specific habitat filters. Finally, combine both information, each obtained at a relevant scale, to refine climatic predictions according to habitat suitability. We illustrate this framework with 14,794 Asian hornet (Vespa velutina nigrithorax) records. We show that integrating multiple drivers, while still respecting their scale of effect, produced a potential range 55.9% smaller than that predicted using the climatic model alone, suggesting a systematic overestimation in many published predictions. This general framework illustrated by a well-documented invasive species is applicable to other taxa and scenarios of future climate and land-cover changes.

Keywords

• Climatic variables
• Spatial scale
• Environmental filtering
• Habitat
• SDM
• Yellow-legged hornet