Deriving Habitat Models for Northern Long-Eared Bats from Historical Detection Data: A Case Study Using the Fernow Experimental Forest

Abstract

Abstract The listing of the northern long-eared bat (Myotis septentrionalis) as federally threatened under the Endangered Species Act following severe population declines from white-nose syndrome presents considerable challenges to natural resource managers. Because the northern long-eared bat is a forest habitat generalist, development of effective conservation measures will depend on appropriate understanding of its habitat relationships at individual locations. However, severely reduced population sizes make gathering data for such models difficult. As a result, historical data may be essential in development of habitat models. To date, there has been little evaluation of how effective historical bat presence data, such as data derived from mist-net captures, acoustic detection, and day-roost locations, may be in developing habitat models, nor is it clear how models created using different data sources may differ. We explored this issue by creating presence probability models for the northern long-eared bat on the Fernow Experimental Forest in the central Appalachian Mountains of West Virginia using a historical, presence-only data set. Each presence data type produced outputs that were dissimilar but that still corresponded with known traits of the northern long-eared bat or are easily explained in the context of the particular data collection protocol. However, our results also highlight potential limitations of individual data types. For example, models from mist-net capture data only showed high probability of presence along the dendritic network of riparian areas, an obvious artifact of sampling methodology. Development of ecological niche and presence models for northern long-eared bat populations could be highly valuable for resource managers going forward with this species. We caution, however, that efforts to create such models should consider the substantial limitations of models derived from historical data, and address model assumptions.