ABSTRACT: After declining by over 80% in the last 30 yr, Steller sea lion Eumetopias jubatus (SSL) populations in Alaska remain low with no sign of recovery. To better examine the potential effects of nutritional stress, foraging studies are needed. Currently available biotelemetry techniques are either incapable of precisely quantifying prey captures or have not been validated for SSLs. Our objectives were to (1) determine whether SSL fish-capture attempts (FCAs) produce a head surge acceleration signal; (2) determine whether acceleration of the head relative to the body (differential) produces a more distinct signal than head acceleration alone (dynamic); and (3) develop a method for predicting FCAs using accelerometry. We attached accelerometers to the head and torso of 2 captive SSLs at the Alaska SeaLife Center and allowed them to feed on live fish. For the first SSL, we used video to classify FCAs during foraging trials and modeled the association between FCAs and parameters created using both dynamic and differential acceleration. Model coefficients were used to predict FCAs for a second SSL, and the results were compared with video. Head acceleration spikes were found in all FCAs. The dynamic model performed slightly better than the differential model at correctly predicting FCAs. Although the model did not perform exceptionally at identifying individual FCAs, it did predict at least 1 FCA in 96% of fish-chasing events and made only 1 false FCA prediction during 392 min of non-foraging. This study demonstrates that head accelerometry is useful for estimating SSL foraging effort.
KEY WORDS: Foraging behavior · Remote monitoring · Eumetopias jubatus · Acceleration data logger · Signal detection