Second Guessing the Maximum Likelihood Estimator Values for Bat Surveys

Abstract

The U.S. Fish and Wildlife Service allows acoustical surveys and automated identification software to determine the presence of the endangered northern long-eared bat (Myotis septentrionalis) and Indiana bat (Myotis sodalis). Analytical software is required to assess presence probability on a site-night basis using a maximum likelihood estimator (MLE) that accounts for interspecific bat misclassification rates. The current standard for occupancy is a returned MLE P-value < 0.05 at the nightly level irrespective of the number of files identified as either northern long-eared bats or Indiana bats. These MLE P-values can vary based on presence of other bat species with similar calls and the relative proportions of all species recorded. Accordingly, there is concern that with few nightly northern long-eared bat or Indiana bat recordings or the presence of large numbers of high-frequency bats, false-negative findings from a swamping effect could result. Using data collected in 2020–2021 by the U.S. Fish and Wildlife Service to set nationwide acoustic monitoring guidelines, we examined the relationship of returned software MLE P-values from 4873 site-nights of acoustic detector data relative to nightly counts of northern long-eared bats and Indiana bats, overall counts of other high-frequency bats, and habitat cover type. For both northern long-eared bats and Indiana bats, nights with one or more echolocation pass files identified as either species but above the MLE P-value threshold largely occurred where nightly counts of the target species was <15 and their proportion to the count of high-frequency bat species was low. We followed this analysis with a simulation using a known call library and observed similar patterns. Accordingly, with few nightly cholocation passes, post-hoc visual assessment following automated software identification easily could be undertaken. Evidence of swamping by other high-frequency species causing positive file identification creating false-negative or false-positives of northern long-eared bats and Indiana bats was not apparent at nightly counts of either species > 10.