Comparing manual and automated methods for identifying individual great gray owls via territorial calls

Abstract Unique identification is essential for understanding population demography and often relies on capturing, tagging, and resighting or tracking known individuals. Recent advances in acoustic analysis show that individual birds can be identified using spectral analysis of vocalizations. However, spectral analysis can be a time‐consuming process, often requiring measurement of multiple sound attributes by hand. The use of mel‐frequency cepstral coefficients (MFCC), which analyze the entire spectrum of a vocalization, is an alternative method to identify individuals within a species. Our goal was to compare the effectiveness of 2 techniques, 1) manual spectral analysis and 2) automated MFCC analysis, for identifying individuals across space and time, using vocalizations of the great gray owl ( Strix nebulosa ). We combined GPS tracking data of known individuals with audio data from autonomous recording units (ARUs) deployed in great gray owl territories during the breeding season to compare spectral analysis methods to automated MFCC methods for identifying individuals. Our analysis utilized territorial calls from 26 ARUs across 4 years (2019–2022) and 14 territories in the Greater Yellowstone Ecosystem in Wyoming, USA. We found that the average classification accuracy of the spectral analysis was 77.2%, whereas the accuracy of the MFCC method was 97.6% based on discriminant analysis. Call analysis using MFCCs was also successful in identifying the same unique individual owl across multiple ARU locations, both within and across years and territories. Our results demonstrated that automated MFCC methods are effective and efficient non‐invasive tools to identify individual great gray owls using territorial calls. The automated MFCC audio analysis method can be applied to other species for which individual identification is possible based on vocalizations. The use of ARUs and MFCC analysis of vocalizations to distinguish individuals can allow for non‐invasive monitoring of individuals and contribute to improved understanding of population dynamics, for example, by providing information on individual fitness and movement behavior.