Assessing Brazilian turtles’ vulnerability BY USING species distribution models AND dispersal constraints

Most assessments of the conservation status of Brazilian turtles use the IUCN geographic range criteria performed by the Minimum Convex Polygon (MCP). This technique often leads to over- or under-estimating the geographic distribution of rare, vulnerable, or endangered species. We aimed to demonstrate that using Species Distribution Models (SDM) on the geographic range assessment of turtles could be more accurate than using the minimum polygon convex. We reduced overestimation of species’ extent of occurrence by adding dispersal constraints, which avoids under- or over-estimating the impact of threatening events. The extent of occurrence derived from MCP was 31% higher than SDM on average, ranging from 4 to 311% higher. Using remaining habitat variables, we found that habitat loss within the predicted extent of occurrence increased by 79% from 1985 to 2019, and inferred population fragmentation increased by 161%. The distribution of turtles Acanthochelys radiolata, Acanthochelys spixii, Hydromedusa maximiliani, Hydromedusa tectifera, Mesoclemmys vanderhaegei, Phrynops williamsi, and Ranacephala hogei is severely fragmented, with most of their extent of occurrence being split into patches that are unavailable to the species persistence. Our findings highlight the importance of using SDM combined with dispersal constraints, which may further benefit from future information about the dispersal capacity of turtles. Furthermore, adding environmental layers to this combination makes it possible to discuss processes affected by habitat fragmentation, such as the fragmentation of species populations, an aspect essential to evaluate population viability and local extinctions.