A case study on Bornean gibbons highlights the challenges for incorporating individual identity into passive acoustic monitoring surveys

Passive acoustic monitoring (PAM) has the potential to revolutionize how we study vocal animals, given recent advances in battery life, data storage capabilities and design of recording devices. A major obstacle to the wide-spread implementation of terrestrial PAM programs—particularly with respect to density estimation—is the difficulty in effectively discriminating between calling individuals. Using PAM, identity can be inferred based on caller location or on individually distinct call features. Here, we report the results of a two-month intensive acoustic survey of two gibbon groups in Sabah, Malaysia. We aimed to (1) test a novel acoustic localization method and (2) test how relative distance of recorders to the vocalizing animal, along with recording day and time, influenced our estimates of the call features important for individual identification. We were able to localize calling individuals from within our array with relatively high accuracy, but localization was less accurate at the edges of the array. We found that features estimated from the spectrogram remained relatively stable across recording distances and conditions, whereas Mel-Frequency Cepstral Coefficients were more variable. Our results have important implications for scalability of PAM programs, and based on our results we provide recommendations for incorporating individual identity into PAM programs.