Accurate prediction of olive-sided flycatcher breeding status using song rate measured with autonomous recording units

Abstract

Autonomous recording units (ARUs) are recognized for their use in detecting vocalizing bird species to assess presence, occupancy, and density, but their potential to monitor reproductive status of individuals and reproductive rates is not well known. We investigated whether song rates derived from ARU data, when combined with the known date, can be used to predict the proportion of male songbirds in 3 breeding status classes (single, paired, and feeding young). We monitored breeding status with weekly field visits and collected daily ARU recordings at 46 olive-sided flycatcher (Contopus cooperi) breeding territories in northwestern Canada in 2016–2017. We tested 4 variations of a hierarchical multinomial regression model that used time of day, day of year, and song rate derived from 2-minute recordings to predict breeding status, and evaluated models using a novel, likelihood-based approach. We found the top model correctly estimated 79% of the observed proportions of birds in each breeding status across the length of the breeding season. Although date was the primary predictor of breeding status, singing rate reduced some of the uncertainty and provided more accurate estimates for a given time. A major challenge to prediction accuracy and data interpretation was accounting for bird movement and the associated impact on detection, which we partly addressed by limiting our study to individuals who were detected on at least 30% of ARU sampling days. We demonstrate that ARUs can be used to assess breeding status in a cryptic, low-density species at risk such as the olive-sided flycatcher, suggesting this method could be applied to a wider range of species to better understand demographics and population dynamics, and inform management decisions, for bird species of concern.