ABSTRACT
By completely censusing a 1 ha forest dynamics plot it was possible to identify the variables (spider mass, size, sex and tree species, size, and bark roughness) that influenced the spatial distribution of adult Drapetisca alteranda Chamberlin 1909 (Araneae: Linyphiidae), a sheet web spider that specializes in lower tree trunks in North American forests. To account for spatial autocorrelation, a conditional autoregressive random effect was included in the zero-inflated Poisson generalized linear mixed model. Parameters estimated were produced by Bayesian inference with vague prior probability distributions and the best of 16 models were selected using Watanabe-Akaike Information Criterion. The best model showed that larger diameter trees located at higher plot elevations were more likely to have D. alteranda present. Smooth bark tree species such as paper birch and American basswood tended to have the most spiders while rough bark species had the least. The relationship between tree diameter and D. alteranda abundance also varied by tree species. Paper birch and quaking aspen tend to produce a greater slope compared to the other species, indicating that as these trees get larger, the abundance of D. alteranda increases at a higher rate than on other tree species. Spider sex and size were not associated with height on the trunk or tree species selection, nor were they associated with microhabitats such as bark furrow depth. Landscape-level factors largely predict D. alteranda abundance and distribution, suggesting that spatial autocorrelation should be considered when modeling the abundance of even small organisms, such as spiders.
