A spatial capture–recapture model to estimate call rate and population density from passive acoustic surveys

Abstract

Spatial capture–recapture (SCR) models are commonly used to estimate animal population density from detections and subsequent redetections of individuals across space. In particular, acoustic SCR models deal with detections of animal vocalisations across an array of acoustic detectors. Previously published acoustic SCR methods either estimate call density (calls per unit space per unit time) rather than animal density itself, require an independently estimated call rate to estimate animal density, or discard data from all but one detected call from each individual. In this manuscript, we develop a new spatial capture–recapture model that estimates both call rate and animal density from the acoustic survey alone, without requiring an independently estimated call rate. Our approach therefore alleviates the need for the additional fieldwork of physically locating and monitoring individual animals. We illustrate our method and compare it to an existing approach using a simulation study and an application to data collected on an acoustic survey of the visually cryptic Cape peninsula moss frog Arthroleptella lightfooti. In the context of our acoustic survey, our calling animal density estimator has low bias, good precision and confidence intervals with appropriate coverage, yielding results that are consistent with previous studies of the same species. Our method can obtain accurate and precise estimates of animal density while eliminating the fieldwork burden associated with separately estimating call rate. We discuss how the development of our model's likelihood reveals a clear path to further extensions, which may incorporate features such as animal movement processes and uncertain individual identification.

The development of a convolutional neural network for the automatic detection of Northern Bobwhite Colinus virginianus covey calls

“…ARUs are being increasingly used to obtain estimates of key population parameters such as occupancy, abundance, call rate and density (Marques et al., 2013; Pérez‐Granados & Traba, 2021; Sebastián‐González et al., 2018; Stevenson et al., 2021). Nevertheless, our simple occupancy case study demonstrated how impactful the choice of method to analyze acoustic data can be on those estimates.…”

A review of spatial capture–recapture: Ecological insights, limitations, and prospects

“…SCR has shown to integrate biotic and abiotic observations at large spatio‐temporal scales to investigate complex population‐level processes. Methodological developments, such as data integration (Chandler & Clark, 2014), continuous time detection (Borchers et al, 2014), noncircular home ranges (Sutherland et al, 2015), density‐dependent home ranges (Efford et al, 2016), and passive acoustic SCR (Stevenson et al, 2021) extend our ability to incorporate complex data structures and hierarchical relationships scaled from the individual to population and species level. Availability of software packages and comprehensive support by the community of users and developers has helped researchers formulate SCR models according to their specific ecological state and observation processes of interest.…”

Accommodating false positives within acoustic spatial capture-recapture, with variable source levels, noisy bearings and an inhomogeneous spatial density

“…Standard SCR relies on animals moving between detection locations and hence typically requires multiple capture occasions, while in ASCR the sound travels almost instantaneously from its source and hence can be detected on multiple sensors in a single occasion. Estimation is of cue spatial density; to convert to animal density an estimate of average cue production rate is required (Marques et al, 2013;Stevenson et al, 2021). As well as the location of detection, additional information is often available about the location of the vocalisation, e.g., the bearing, received sound level or time of arrival on multiple sensors.…”