Predictive model of sperm whale prey capture attempts from time-depth data

Area of Science:

• Marine Mammal Ecology
• Bioacoustics
• Behavioral Ecology

Background:

• High-resolution tags provide detailed foraging data but are costly.
• Time-Depth Recorders (TDRs) are affordable but offer limited, bi-dimensional data.
• Quantifying foraging effort from TDR data alone is challenging.

Purpose of the Study:

• Develop a predictive model to estimate sperm whale foraging effort using TDR data.
• Identify prey capture attempts (PCAs) from time-depth profiles.
• Assess the feasibility of using low-cost TDRs for fine-scale foraging behavior analysis.

Main Methods:

• Downsampled high-resolution tag data to match TDR resolution (1 Hz).
• Developed generalized linear mixed models using dive metrics (depth, velocity variance) to predict buzzes (PCAs).
• Evaluated model performance using different dive segment durations (30, 60, 180, 300 s).

Main Results:

• Average depth, depth variance, and vertical velocity variance were key predictors of buzzes.
• Models using 180-second dive segments showed optimal predictive performance (AUC=0.78 ± 0.05).
• Predicted buzzes closely matched observed numbers, with a median difference of 30%.

Conclusions:

• Accurate indices of sperm whale prey capture attempts can be derived from TDR data.
• This approach enhances the study of foraging ecology in sperm whales and other echolocating cetaceans.
• Utilizing TDR data democratizes research, enabling long-term and historical analyses of cetacean foraging.