Ecological connectivity assessment in a strongly structured fire salamander (Salamandra salamandra) population.

Small populations are more prone to extinction if the dispersal among them is not adequately maintained by ecological connections. The degree of isolation between populations could be evaluated measuring their genetic distance, which depends on the respective geographic (isolation by distance, IBD) and/or ecological (isolation by resistance, IBR) distances. The aim of this study was to assess the ecological connectivity of fire salamander Salamandra salamandra populations by means of a landscape genetic approach. The species lives in broad-leaved forest ecosystems and is particularly affected by fragmentation due to its habitat selectivity and low dispersal capability. We analyzed 477 biological samples collected in 47 sampling locations (SLs) in the mainly continuous populations of the Prealpine and Eastern foothill lowland (PEF) and 10 SLs in the fragmented populations of the Western foothill (WF) lowland of Lombardy (northern Italy). Pairwise genetic distances (Chord distance, DC) were estimated from allele frequencies of 16 microsatellites loci. Ecological distances were calculated using one of the most promising methodology in landscape genetics studies, the circuit theory, applied to habitat suitability maps. We realized two habitat suitability models: one without barriers (EcoD) and a second one accounting for the possible barrier effect of main roads (EcoDb). Mantel tests between distance matrices highlighted how the Log-DC in PEF populations was related to log-transformed geographic distance (confirming a prevalence of IBD), while it was explained by the Log-EcoD, and particularly by the Log-EcoDb, in WF populations, even when accounting for the confounding effect of geographic distance (highlighting a prevalence of IBR). Moreover, we also demonstrated how considering the overall population, the effect of Euclidean or ecological distances on genetic distances acting at the level of a single group (PEF or WF populations) could not be detected, when population are strongly structured.