Modeling sperm whale seasonal dynamics using probabilistic machine learning of passive acoustic data

Seasonal variability in sperm whale acoustic activity in the northern Gulf of Mexico remains poorly understood. Using two years of passive acoustic data, hourly click counts were analyzed to characterize seasonal patterns in whale presence and behavior. A generalized linear modeling framework was used to accommodate temporally-correlated, over-dispersed, and zero-inflated click-count data. Baseline activity differed significantly among seasons, with the highest levels occurring in winter and fall and the lowest in spring. These results establish a quantitative seasonal baseline of regional sperm whale acoustic activity important for future studies linking habitat-use behavior to environmental and anthropogenic drivers in the Gulf.