An automated lunge detector for baleen whale tag data: Application and results for fin whales off Southern California

Publications >> An automated lunge detector for baleen whale tag data: Application and results for fin whales off Southern California

Citation

Allen, A, J Goldbogen, A Friedlaender, J Calambokidis, and B Southall. 2015. An automated lunge detector for baleen whale tag data: Application and results for fin whales off Southern California. Abstract (Proceedings) 21st Biennial Conference on the Biology of Marine Mammals, San Francisco, California, December 14-18, 2015.

Abstract

Multi-sensor archival tags have become a relatively common tool for studying the underwater behavior of diving animals, including whales. Rorqual whales (Balaenopteridae) feed via an energetically costly, complex behavior called lunge feeding, an intermittent ram filtration mechanism. This process includes kinematic maneuvers at depth and near the surface that have signatures evident in a number of tag sensors. The extreme size of rorqual species requires high energetic demands and consequently high feeding rates. Developing an automated method to detect lunges from tag records would significantly decrease data analysis time and could be used to augment studies across a range of species and tag types. Here we identify lunges in 19 fin whales tagged with digital acoustic recording tags (DTAG). We demonstrated that surface lunge feeding in fin whales is identifiable from tag data. While there are similarities in kinematic behavior to deep lunges, surface lunge feeding is less stereotypic. Most surface lunging whales rolled >20º to the right, and have a large peak and then drop in both jerk (a metric of body acceleration) and flow noise. These features likely correspond to acceleration towards schooling prey, and then an abrupt deceleration due to high drag as the mouth opens and ventral pouch expands. Using accelerometer and hydrophone data, an automated lunge detector was designed that correctly identified an average of 92.8% of all lunges, with a false positive rate of only 9.5%. The very strong performance of our lunge detector has significant promise for applications in studies of foraging ecology, baseline behavior, energetics, and behavioral response to controlled sound exposure of fin (and other baleen) whales. These studies require standardized methods for determining feeding rates. Our methods can be applied to other cetacean species and could be incorporated into in situ lunge counters in next generation, long-term tag designs.