Vocal individuality measures reveal spatial and temporal variation in roding behaviour in Woodcock (Scolopax rusticola)

Abstract

Use of species-specific field methods may be required for taxa that are inherently difficult to survey, for example species with cryptic camouflage or secretive behaviour. However, these methods often require more manual effort and therefore cost. Passive acoustic monitoring (PAM) is now an established tool to reduce manual effort to monitor species, and analysis of spectrograms provides the means to discriminate individuals by call characteristics. At night, male Eurasian Woodcock Scolopax rusticola make distinct, audible 'roding' displays above the tree canopy to advertise to females. Studying this behaviour at an individual level while using PAM presents opportunities to improve monitoring methods of this cryptic, Red-Listed species. This study evaluates the potential use of vocal individuality measurements in distinguishing Woodcock individuals and interpreting spatial and temporal patterns in their roding displays across woodland sites. Woodcock roding calls were recorded from woodland fragments across two regions comprising 20 sites. A random forest classifier was applied to reduce the time needed to find and manually verify calls. Principal component analysis (PCA) and hierarchical clustering algorithms were used on call measurements, describing duration and frequency differences in calls. When clusters formed, they were used to qualitatively assess supposed individual spatial and temporal variation in roding behaviour. The variance of dimensionally reduced measurements was used to interpret local Woodcock abundance and changes over time. Supposed individuals used many sites within a region, and many sites were used by multiple birds. However, sites showed clusters of calls from supposed individuals in different proportions. It was difficult to discriminate individuals using PCA with more than six birds because the degree of call overlap increased. Though the call measurement variance is associated with number of call events, it may provide a suitable method for representing population size without call count bias.