Estimating Probabilities of Detection for Bat Echolocation Calls: An Application of the Double-Observer Method

Abstract

Differential detectability is an issue of great practical importance in bat species (Order Chiroptera) surveys. Bat echolocation recorders increasingly are being used in survey efforts, but recorder-based surveys have not measured differences in detection probability quantitatively. A recently developed avian survey technique uses 2 observers at a site to estimate the probability of detection by comparing birds recorded independently by the 2 observers. We applied this method to bat surveys using remotely operated Anabat echolocation recorders. We placed 2 detectors at each sample point, which were randomly selected from forested habitat within 2 watersheds in Indiana and 2 in Missouri. The first detector was oriented toward the most open area of forest and the second was set 5 m away and oriented toward the most open area that allowed sampling of an area distinct from the first detector. We estimated detection probabilities using the Huggins closed-capture model in program MARK. We compared models of detection probability using the corrected quasi-likelihood Akaike's Information Criterion adjusted for overdispersion (QAICc). The best models accounted for detector placement and region but not species. Overall probability of detection for bats using 2 detectors was 0.62 in Indiana and 0.42 in Missouri. Results confirmed that a second detector increased the probability of detecting different species of bats at a site. Future regional comparisons should account for differences in detection probability to avoid biases associated with surveys relying solely on recorded calls.