To assess the influence of variation in environmental conditions and habitat structure on recent mammal communities, we developed two matrices of distances among the UTM 100x100 grid squares: *The first one, denv, accounts for environmental variation. It was built using the information from six climatic variables (minimum, maximum and mean annual temperature, mean annual and summer precipitation, and percentage of sunny hours per year), and three bedrock geologic (calcareous rock, acid rock and clay area). Climate data extracted from the 50x50 km resolution data provided by W. Cramer (CLIMATE database version 2; http://www.pik-potsdam.de/~cramer/climate .html). To obtain bedrock data, we used a previously digitised three-category map (IGN 1995). All these variables were successfully used in other analyses on the distribution of species richness and composition of several groups of
Iberian plants and insects (e.g., Lobo et al. 2001, Lobo and Martin-Piera 2002, or Hortal et al. 2004). These analyses usually include geomorphologic and structural variables, such as altitudes, slopes, aspects, etc. However, we use such information to develop friction surfaces and connectivity analyses (see below). Thus, we have excluded them from the calculation of the environmental matrix, in order to avoid self-replication and to separate more clearly both sources of faunistic variability. *The second one, diuic, accounts for the differences in the land use/land cover of each area, that is, for differences in habitat structure among areas. To develop this matrix, the original CORINE Land Use/Land Cover map (EEA 2000) were reclassified into 13 categories (urban areas, dry cultures, irrigated cultures, patchy cultures, olive, vineyard and fruit plantations, anthropogenic pastures, natural pastures, broadleaf forests, conifer forests, mixed forests, shrubland, bare rock and land masses). The percentage of land coverage devoted to each category was extracted using GIS techniques. Similar variables were also successfully used in several of the formerly related studies.
Following the recommendations of Legendre and Legendre (1998), after the standardization of the truly continuous variables (those related to climate), we used the Gower Similarity Index (Gower 1971) to measure the environmental and habitat proximities among grid squares. In the case of the development of the denv matrix, different weights were assigned to bedrock and climate variables, to account for the differential importance of each group of environmental variables in the distribution of the species (see discussion at Hortal and Lobo 2005). Although bedrock geology is not a direct determinant of variations in mammal communities, it does for plants (see, e.g., Lobo et al. 2001 for the Iberian Peninsula), a well known determinant of mammal diversity (see, e.g., Olff et al. 2002 or Cristoffer and Peres 2003). Thus, substrate composition may have an indirect effect over mammal communities. Although no analysis is available on the different weights of geology and climate in the configuration of mammal communities (see Hortal and Lobo 2005 for an example of such analysis on Iberian dung beetles), we have assumed the latter having greater importance. In absence of any hypotheses of the magnitude of such difference, and following the empirical evidence obtained by Hortal and Lobo (2005), climate was assumed to be three times more important than geology. Thus, every climatic variable had a weight of 1/2 (3/6 variables), giving a total weight of 3 to climate (0.5 x 6 climatic variables). In the same way, every geologic category had a weight of 1/3 (1/3 variables), and geology a total weight of 1 (0.33 x 3 variables). The resulting similarity matrices (calculations made with Clustan Graphics software; Clustan Ltd.
2002) were transformed into distance matrices by calculating the inverse of the index value (that is, d = 1 - Gower value).
Was this article helpful?