When modelling landscape connectivity, one of the main challenges is to have an a priori knowledge of resistance values for each landscape feature. Originally, these values were often assigned based on expert opinion, which was strongly criticized due to the subjective nature of such decisions. To avoid this bias, most current studies infer these values from ecological niche models. With the rise of naturalist databases from citizen science programs for more than a decade, this approach has met with considerable enthusiasm but has not been fully evaluated yet, despite being debatable on certain theoretical aspects. More recently, another approach, implemented in the resistanceGA package in R, has been proposed to optimize resistance surfaces from landscape variables without a priori assumption about the effect of landscape elements on the focal species. Using population genetic differentiation metrics as a proxy for gene flow, this genetic algorithm (GA) implements several methods to infer connectivity, including circuit theory models, which are expected to perform well for species with diffuse dispersal like amphibians.

We implemented both approaches to assess landscape connectivity in the Yellow-bellied toad (Bombina variegata), a European amphibian species particularly vulnerable to global change. More specifically, we first used a species distribution modelling approach to obtain a relative habitat and climatic quality index, based on opportunistic presence data from citizen science databases. These suitability maps were then used to obtain a pairwise landscape connectivity matrix using circuit theory modelling. On the other hand, we optimized the raw landscape variables with resistanceGA to obtain a second connectivity matrix. We then compared the adjustment of both approaches (i.e., based on suitability index or raw landscape variables) according to spatial genetic differentiation, taking advantage of a dense genetic sampling from the Rhône-Alpes region in France (RADseq, 73,825 SNP).