Indiana bat maternity roost habitat preference within Midwestern United States upland Oak-Hickory (Quercus-Carya) forests

•Locations and characteristics of Indiana bats' maternity roost trees were studied.•Feature selection was performed for correlated variables in fragmented forests.•Tree height, width, bark, and distance to forest edge and water were key factors.•Live trees were chosen only when tree and location features were highly favorable.•Shagbark hickory was important, but other species can contribute to bat survival.

The Indiana bat (Myotis sodalis) is a federally listed endangered species negatively impacted by human disturbance, habitat change, and disease. Habitat protection and management of summer roosting habitat and cave hibernacula are recommended for the recovery of this species. We studied roost tree and landscape characteristics of 19 known summer maternity colony habitats in Illinois and Iowa upland oak-hickory (Quercus-Carya) forests. Landscape variables can be highly correlated in fragmented forest habitats and not all the roost tree and landscape variables are relevant to roost tree selection. We employed an algorithm to approximate the data set by using singular value decomposition (SVD) to identify the primary factors governing the selection of maternity roosts. The proposed method (formally referred to as a feature selection algorithm) approximates the data by discarding highly correlated features and features that can be removed without incurring much loss of information. Results indicated that summer maternity roosts were trees closer to forest edge, larger in diameter and typically trees with crowns in the upper canopy of the forest. In addition, partial bark cover and locations within 1 km of water bodies positively influenced the tree selection. Although live or dead shagbark hickory (Carya ovata) were preferred as maternity roosts, snags of other trees species common to Midwest upland oak-hickory forests were also used. We incorporated a joint probability density estimation and the Wilcoxon rank sum test to investigate the relationship between tree species and dead/live tree status for roost tree selection. There was sufficient evidence to infer that the chosen live trees were taller and in more favorable locations compared to the chosen snags. We further observe that the joint distribution for quantitative attributes among dead or declining shagbark hickory was not significantly different compared to the other chosen tree species, indicating that a lack of shagbark hickory may not limit the Indiana bat population. Knowledge of these complex relationships regarding maternity roost habitat preferences is useful for future management of the Indiana bat throughout Midwest oak-hickory forests.