Occurrence of delphinids off Egmont Atoll, Chagos Archipelago: Detection, characterisation, and temporal variability of vocalisations

Passive acoustic monitoring techniques are useful for studying vocally active marine species, particularly in remote and difficult to access areas. In this study, three months of acoustic recordings were collected off Egmont Atoll, Chagos Archipelago in the central Indian Ocean to detect, characterise, and investigate temporal variability of delphinid whistles. A subsample (3.6% of the total dataset) was used to manually annotate whistles and test automated whistle detectors. Higher frequency whistles (6-25 kHz, n = 126) were manually classified into six categories: upsweep (56%), convex (19%), downsweep (13%), concave (5%), sine (4%), and constant frequency (3%). An automated whistle detector was evaluated under five detection thresholds (measured in dB): 5.5, 6.0, 6.5, 7.0, and 7.5 above background noise. The 6.5 dB threshold demonstrated the best balance between precision (0.9) and recall (0.5). The whistles were detected on 78% of days, with a bimodal diel pattern, where whistles peaked after sunrise (0800-1000 h) and before sunset (1600-1800 h), with fewer detections in the middle of the day and at night. This study highlights the value of passive acoustic monitoring techniques to better understand the delphinid occurrence in remote and understudied areas.