Just Accepted

This article has been peer reviewed and accepted for publication. It is in production and has not been edited, so may differ from the final published form.

Species-specific spatial and temporal variability in anuran call detection: implications for deploying autonomous recording units

Andrew Hall , Amelia Walcott, Ali Borrell, Dale Nimmo , Skye Wassens

Abstract

Context Ecosystem assessment using acoustic monitoring technologies can be an efficient method for determining species community composition and breeding activity, but many factors affect the quality of acoustics-data and subsequent level of confidence in derived inferences. Aims We aimed to assess variability in detection probabilities of five frog species using autonomous recording units (ARUs) deployed across a single 1 km2 wetland, comprising a lagoon and surrounding area, and subsequently determine the required number of ARUs for 95% confidence in derived presence-absence data. Methods Ten ARUs were deployed in two rings around the lagoon’s centroid close to the water’s edge. Occupancy models were used to derive detection probabilities of species calling in the lagoon from data describing the temporal pattern of calling at each site, which were derived using call recognition software. Key results Only two of the five target species were detected by all 10 ARUs. All target species’ non-zero ARU detection probabilities varied by a factor of 14, and the coefficients of variation in individual ARU detection probability for each species varied by a factor of 7. Simulations revealed 7 or 8 ARUs are required for 95% confidence in detection on any one day of the two species with the highest detection probabilities that were known to be calling. Even with ten deployed ARUs, the probability of successful detection of the other three species known to be calling on any day was less than 40% . Conclusions Effective detection was not achieved for all targeted species by several ARUs during a period when hydrology and season suited recruitment activity. Despite all ARUs being deployed at locations favourable for detecting targeted species, stochastic factors drove spatial variability in detection resulting in markedly different data for each ARU and each species. Implications Data describing species presence derived from automated recording units may not be representative due to spatiotemporal variability in detection that varies by species. To improve ARU deployment strategies, a priori knowledge of typical detection probabilities and species spatial variability can be used to determine the required number of call recorders for a set level of confidence.

WR24036  Accepted 20 December 2024

© CSIRO 2024