Predicting Breeding Status of a Forest Songbird from Singing Rate

For male breeding songbirds, song rate varies throughout the breeding season and tends to be correlated with breeding-cycle stages. Although these patterns have been well documented, to our knowledge, this relationship has not been used to predict a bird’s breeding status through acoustic monitoring. The first objective of this study was to determine if variation in song rate can be used to predict the breeding status of the Olive-sided Flycatcher (Contopus cooperi; OSFL), a Species at Risk in Canada. In 2016, song rates from 27 male OSFLs in Alberta and the Northwest Territories were collected from human observers (n = 454 5-min counts), and breeding status (i.e. single, paired, and feeding young) was monitored throughout the breeding season. I evaluated the predictive ability of three modeling approaches (i.e. regression, hierarchical, and machine learning) using model sensitivity and specificity. The hierarchical model was the best at predicting all three breeding statuses, with 69%, 50% and 87% sensitivities and 80%, 82% and 78% specificities for predicting single, paired, and feeding young, respectively. This resulted in a mean sensitivity of 69%, compared with 54% and 50% from the regression and machine learning models, respectively. A second objective was to use the hierarchical modelling framework to predict breeding status from song rates collected by Autonomous Recording Units (ARUs) processed using automatic recognition software. For 24 of these OSFLs, I collected 4,302 5-min song counts and used daily song rate to compare the relationship of rates and breeding status as determined by ARUs versus human-observers. We then tested four hierarchical models accounting for imperfect detection. Song rates derived from ARU data followed a similar pattern to that of human-observer song rates, where single males had the higher rates, paired males had lower rates, and those feeding young had lowest rates, but the absolute values for rates were much lower with ARUs. All ARU data predictive models performed poorly at predicting single (sensitivity range 0 – 7%) and well at predicting paired (sensitivity range 77 – 84%). The ARU models had mixed success at predicting feeding young (sensitivity range of 25 – 68%) but adjusting for imperfect detection did not improve model sensitivity to predict any breeding statuses. Low predictive ability was likely due to the low detectability of ARUs (e.g. bird movement out of detection range of ARU) and the automatic recognition software we used. Considering the high predictive ability of models using human-observer data and that the challenges currently associated with our acoustic processing methods can be addressed, I recommend that the breeding status of forest birds should be monitored using acoustic data. I provided a hierarchical modelling framework than can be applied to other species and improved to account for bird movement or number of conspecifics. This novel approach could provide a cost-effective tool to infer much needed demographic information over large spatial extents, and inform species status assessments, recovery strategies, and management plans for many species of conservation interest.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.