Testing the Generality of Patch and Landscape-Level Predictors of Tree Squirrel Occurrence at a Regional Scale

Abstract

Few investigators have considered whether predictors of species occurrence from 1 locality are useful predictors in others, or how regional factors may constrain occurrence. We used landowner questionnaires over 6 regions in Indiana to test the performance of predictors from prior studies of more limited spatial extent. Logistic regression and classification tree models were constructed using data on 577 forest patches supplied by 362 landowners. Consistent with predictions, probability of occurrence of eastern gray squirrels (Sciurus carolinensis) was related positively to forest patch size and percent of forest in the surrounding landscape. Also, probability of occurrence of North American red squirrels (Tamiasciurus hudsonicus) was related positively to the presence of black walnut trees (Juglans nigra) and conifers in forest patches, and negatively to the presence of gray squirrels. Occurrence rates of S. carolinensis increased from north to south, whereas occurrence rates of T. hudsonicus increased from south to north and from west to east. Based on landowner perceptions, abundance of T. hudsonicus was more likely to increase in north and central regions, and to decline in patches with walnuts and at sites with longer periods of observation by landowners. Although less detailed than studies restricted to a particular locality, questionnaire data permit ecologists to test the generality of local findings and to further formulate and refine hypotheses about processes underlying large-scale patterns.