Environmental data:
Sunrise, sunset and civil twilight times were extracted from (www.timeanddate.com/sun) for Sherkin Island, and were used to calculate the diel cycle phases (morning, day, evening, and night) (Carlström, 2005; Todd et al. , 2009). Tide data were extracted from tide tables for Roaringwater Bay (www.tides4fishing.com/ie/munster/roaringwater-bay). Time difference to nearest high tide was calculated, as well as phases of the tidal cycle (ebb/flow/high/low water). Tidal range was also calculated as an indicator of spring and neap tides. Hourly water temperature data were obtained from the PODs, taken as the average value recorded by both PODS to avoid any recording bias between PODs. While the PODs cannot provide a direct measure of environmental noise levels, the click detection algorithms record unfiltered short click-like events within the 20kHz – 160kHz bandwidth (Nall ), and is recorded for each sampling minute. A positive correlation has been found betweenNall and environmental noise levels using a full-bandwidth recorder and is used as a proxy for general noise levels (Clausen et al., 2018; Nuuttila et al., 2018). We included this in models to investigate the effect of varying environmental noise levels on POD detection performance. For the purpose of interpretation of the results, seasons were defined as Spring (March to May), Summer (June to August), Autumn (September to November) and Winter (December to February).