Site-scale variability of calling activity of insectivorous bats:
implications for passive acoustic monitoring
Abstract
Detectors for the passive acoustic monitoring (PAM) of bats have become
invaluable research tools, especially for surveys, monitoring programs
and environmental impact assessments. However, little is known about the
small-scale (within-site) variability of PAM recordings and especially
about the influence of detector identity and distance, and of microphone
orientation on the statistical confidence of activity estimates and
species detection probabilities. We recorded vocalizations in a
homogeneous meadow with no trees, bushes or tall ground vegetation.
Eight detector pairs were arranged in an octagon, the two detectors of a
pair facing in opposite directions. The call sequences of eight species
were analyzed. The deviations of individual detectors from the overall
mean were generally small, but large outliers occurred both at the file
(temporal resolution: five seconds) and the night (resolution: one
night) scale. All devices detected the main temporal patterns of calling
activity in the study period, but three devices deviated systematically
from the others and the sensitivity of two devices deteriorated over
time. Detector orientation and distance were significant, yet small,
sources of variability. The probability of detecting the presence of
species correlated with species’ activity and ranged on average from 100
% for bats in total to only 18.8 % for the least active Myotis myotis.
The sample sizes necessary to achieve 90 % statistical confidence of
activity estimates ranged from 7 to 16 detectors and from 5 to 12
nights, depending on taxon. Increasing the number of nights resulted in
much higher confidence than increasing the number of detectors. We
recommend PAM studies of bats to frequently calibrate detector
sensitivity; deploying at least three detectors per study site; sampling
longer periods instead of deploying more detectors; randomly assigning
and swapping detectors among sites, treatments, strata, etc.; and
statistically scrutinizing the sample data, especially for outliers.