Environmental constraints regulating the distribution and abundance of small mammals.

Abstract

I computed 15 landscape variables at four different scales (40 variables total). Cluster analysis of these weighted data produced three multispecies clusters based on associations of species distributions and abundances to landscape factors.

The landscape predictive models constructed using discriminant function analysis determined which landscape variable or combination of variables were most efficient in classifying species into the appropriate cluster and allowed small-mammal distributions across the landscape to be predicted. Cluster classification accuracy was 59%. When local-level variables were combined with the landscape data classification accuracy was 58%.

The PCA of local variables showed that four species (C. hispidus, N. floridana, P. attwateri, and P. leucopus) occupied barren or rocky areas with a woody canopy, while six species (C. parva, M. ochrogaster, P. maniculatus, R. fulvescens, R. montanus, and S. hispidus) preferred open grassy areas. Weighted discriminant analysis of the local variables produced better predictive accuracy (75% correctly classified) than the unweighted data (63% correctly classified). Discriminant analysis using only the two largest clusters produced classification accuracy of 72% (unweighted) and 83% (weighted). Total number of broadleaf trees and rocky around cover were the most important factors in discriminating among groups.

I assessed the influence of 19 local-level, 40 landscape-level, and 59 combined variables on the distribution and abundance of small mammals at 60 plots across Fort Sill Military Reservation in Comanche County, Oklahoma. I collected 15 small-mammal species and used 10 of these (n > 10) in my analyses. Variables for each mammal species were evaluated as unweighted measures based on the presence/absence of each mammal species at a plot and as weighted measures based on the abundance of each mammal species at each plot. These data were subjected to cluster analysis, principal-components analysis, and discriminant-function analysis. General trends of the local, landscape, and combined affinities of species in these clusters were summarized on principal components.