Inferring landscape effects on gene flow: the endangered brown bear in the Cantabrian range (NW Spain).

The ability of populations to be connected across broad landscapes is critical to long term viability for many species. This is particularly the case of endangered brown bear in northwest Spain which presents a fragmented distribution that has led to a highly limited gene flow and its subsequent deleterious effects for the conservation of this emblematic species (Pérez et al. 2009, 2010). Despite the clear importance of genetic connectivity for population persistence, the best ways to mitigate population isolation remain an object of intense debate (Crooks and Sanjayan 2006; Awade 2012) and specific factors mediating connectivity are largely unknown for most species. In the last years a myriad of studies have produced maps of corridors and linkages relying on expert opinion due to lack of detailed information on animal movement, which is not desirable (Seoane et al. 2005). To overcome this limitation, some authors proposed that resistance to movement could be estimated as a inverse function of a habitat occurrence model (e.g Ferreras 2001, Chetkiewicz et al. 2006, O´Brien et al. 2006, Beier et al. 2008). In this context, the field of landscape genetics has great potential to provide more rigorous methods to infer the influences of landscape on population connectivity (Spear 2005, Storfer 2007, Cushman 2006, Balkenhol et al. 2009). Individual-based analysis comparing pairwise genetic distances to pairwise effective distances under multiple landscape resistance hypothesis are a powerful tool for conservation efforts. Moreover, recent work has refined and extended landscape genetics approaches on optimizing and improving model selection frameworks (Shirk 2010, Wasserman 2010, Cushman et al. 2013). Our goal in study is to use rigorous connectivity models to provide detailed and quantitative guidance to conservation planning efforts aimed at improving landscape permeability for brown bear population in Spain. We used (1) a model selection framework based on univariate and multivariate optimization (Shirk et al. 2010) of the potential landscape variables considered as the most important resistors for brown bear movements and therefore to gene flow and (2) the calculation of the relative support of the causal modelling tests, which has been shown as an improvement of causal modelling performance (Cushman et al. 2013). We compared this selection framework with the use of habitat suitability as a surrogate of landscape resistance by utilizing a number of multiscale suitability models (based on the maximum entropy method) developed under different ecological hypotheses developed for brown bear in the study area (Mateo-Sánchez et al. 2013). Our findings showed that the best model for habitat suitability failed to be a strong indicator of gene flow for brown bear in the study area. However, other habitat models that showed a weaker ability to predict bear ocurrence were strongly related to brown bear genetic isolation. Overall gene flow appears highly related to unfragmented forested areas. This study attempts to demonstrate the utility of this approach to determine the influence of landscape on population genetic structure and to account for potential implications of the use of species-habitat relationships as surrogates of resistance surfaces.