Canadian Forest Service Publications

Comparison of semiautomated bird song recognition with manual detection of recorded bird song samples. 2017. Venier, L.A.; Mazerolle, M. J.; Rodgers, A.; McIlwrick, K.A.; Holmes, S.;Thompson, D. Avian Conservation and Ecology 12(2): article 2.

Year: 2017

Issued by: Great Lakes Forestry Centre

Catalog ID: 37411

Language: English

Availability: PDF (request by e-mail)

Available from the Journal's Web site.
DOI: 10.5751/ACE-01029-120202

† This site may require a fee

Mark record

Plain Language Summary

Automated recording units are increasingly being used to sample wildlife populations. These devices can produce large amounts of data that are difficult to process manually. However, the information in the recordings can be summarized with automated sound recognition software. Our objective was to assess the utility of the automated bird song recognizers for conservation and sustainable forest management applications. Specifically, we compared detection data generated from expert-interpreted recordings of bird songs collected with automated recording units and data derived from a semi-automated recognition process. We recorded bird songs at 109 sites in boreal forest in 2013 and 2014 using automated recording units. Our study consisted of two components. First, we selected ten minute recordings from each site for interpretation. Each of the 436 recordings was interpreted by a bird song expert who noted all birds heard. We developed bird-song recognizers for ten species using Song Scope software (Wildlife Acoustics) and each recognizer was used to scan the same set of recordings. We used occupancy models to estimate the detection probability associated with each method. In the second part of our study, we estimated the detection probability of song recognizer software from four one-week recordings (350 - 630 min per week) at each of the 109 sites. Results show that the detection probability of most species from 10-min recordings is substantially higher using expert-interpreted bird song recordings than using the song recognizer software. However, our results also indicate that using automatic recording units and song recognizer software can be valuable tools to estimate detection probability and occupancy of boreal forest birds, when sampling for sufficiently long periods.