Null models, co-occurrence patterns, and ecological modelling of a Mediterranean community of snakes

Abstract

A Mediterranean community of snakes, inhabiting a hilly area in central Italy (Monterano, Tolfa Mountains) was statistically modelled to investigate whether: (i) it shows non-random structure as predicted by competition theories, (ii) there are proximate elements of the landscape that may predict the presence/absence of the various snake species at the study area, and (iii) the snakes select 'resting sites' which are relatively simple or relatively complicated in landscape structure. To investigate issue (i) we used co-occurrence pattern analyses and Monte Carlo simulations. To investigate issue (ii) we used logistic regression analysis on presence/absence data of eleven 'proximate landscape' variables around sites of presence for the snakes and around random (= absence) sites at the same study area. To investigate issue (iii), we used expected-versus-observed x2 to test whether snake presences were associated more often than random with a combination of variables, i.e. with at least four different variables for each point of presence. The snake community was constituted by eight species, but only the five commonest species (i.e. Coluber viridiflavus, Elaphe longissima, Elaphe quatuorlineata, Natrix natrix, and Vipera aspis) were included in our models. Our modeling and simulation analyses suggested that: (i) the community structure was not competitively structured both in terms of habitat resource and in terms of 'proximate landscape' variables influencing snake presences; (ii) all five snake species were positively influenced by at least two of the eleven variables considered in the logistic regression model; (iii) four out of five species clearly selected sites with complicated structure, the only exception being Coluber viridiflavus. © Koninklijke Brill NV 2006.