Modelling correlates of microhabitat use of two sympatric lizards: a model selection approach

Studies of the ecological characteristics of sympatric species are important for developing and testing ecological theory, but may be of interest also for conservation biology research when the study species are threatened, endemic and with narrow distribution. Here, we studied a particular aspect of the ecology (i.e. microhabitat use) of two sympatric lizards (Archaeolacerta bedriagae, Podarcis tiliguerta) endemic to Sardinia and Corsica (Tyrrhenian islands). We studied this issue by modelling procedures, using field data collected at six study areas in both Sardinia and Corsica. We recorded 18 microhabitat variables for each lizard spot. The variables were entered as independent variables in logistic regression analysis with the presence/absence data for the lizards as the dependent variable, and Akaike Information Criterion was applied to select the best models describing the ecological equation of each study species. In total, we based our modelling approach on 296 individuals of A. bedriagae and 182 of P tiliguerta. The general logistic regression models revealed that five distinct variables were significantly correlated to the presence/absence of A. bedriagae, and six to that of P tiliguerta. We found that three variables were important for only P tiliguerta, two for only A. bedriagae, and three for both species and with an identical sign. We also found some similarities in microhabitat choice between species. Indeed, some variables were always present in the best models of both A. bedriagae and P tiliguerta. In general, A. bedriagae was more related to spots with large stones and low vegetation than P tiliguerta, which, on the contrary, choose spots relatively closer to vegetation. The various reasons explaining the observed similarities and differences between species were examined. It is suggested that our modelling procedure may be widely used for studies of lizard community ecology, because it easy to use and allows a more-in-depth analysis than normal 'count approaches'. (C) Koninklijke Brill NV, Leiden, 2009.