Comparative assessment of different methods for using land-cover variables for distribution modelling of Salamandra salamandra longirotris

Abstract

Predictive models are frequently used to define the most suitable areas for species protection or reintroduction. Land-cover variables can be used in different ways for distribution modelling. The surface area of a set of land-cover classes is often used, each land-cover presence/absence or the distance to them from any point of the study area can be preferred; multiple types of landcover variables may be combined to produce a single model. This paper assesses whether different approaches to using land-cover variables may lead to different ecological conclusions when interpreted for conservation by focusing on the distribution of the salamader Salamandra salamandra longirostris, an endangered amphibian subspecies in the south of the Iberian Peninsula. Twenty-eight land-cover classes and another 42 environmental variables were used to construct four different models. Three models used a unique type of land-cover variable: either the presence of each class, the surface area of each class or the distance to each class, with all three variable types jointly entered in a fourth model. All models attained acceptable scores according to some criteria (discrimination, descriptive and predictive capacities, classification accuracy and parsimony); however most of the assessment parameters computed indicated a better performance of the models using either the surface area of land classes or the distance to them from every sampled square, compared to the model using class presences. The best scores were obtained with the fourthmodel, which combined different types of land-cover variables. Thismodel suggested that oak forest fragmentation in favour of herbaceous crops and pastures may have negative effects on the distribution of S. s. longirostris. This was only partially suggested by the first threemodels, which considered a single type of land-cover variable, demonstrating the importance of considering a multi-variable analysis for conservation planning.