Development and evaluation of species distribution models for five endangered elasmobranchs in southwestern Atlantic

Abstract

Species distribution models (SDMs) are tools to obtain habitat suitability maps based on historical species occurrences and environmental variables. Those maps can be used to restrict fishing grounds or to assist in planning and reserve selection. This is especially important for species at risk of extinction. We developed SDMs for five endangered elasmobranch species, namely Squatina guggenheim, S. occulta, Rhinobatos horkelii, Galeorhinus galeus, and Mustelus schmitti, using Boosted Regression Trees. Data from 1,704 bottom trawls carried out between 1972 and 2005 as part of research surveys on the southern Brazilian shelf between 28°36′S and 33°45′S, combined with satellite imagery and environmental atlases, were used in the models. Based on 10-fold cross-validation statistics, all models had a reasonable performance, though S. guggenheim models had an excellent discrimination (AUC > 0.9) and R. horkelii models had just a fair discriminatory power (AUC 0.7–0.8). Except for R. horkelii, all models showed good association between observed and predicted occurrences (PBC > 0.5). Squatina guggenheim models provided the greatest explained deviance (49–54%), whereas R. horkelii models the smallest (14–17%). Models’ predictions were consistent with the current knowledge of all species. Moreover, those models made reasonable predictions using the great spatial and temporal coverage of satellite data.