Assessment of goose-beaked whale responses to mid-frequency active sonar using a hierarchical hidden Markov model.

backgroundAcoustic disturbance is increasingly recognized as an ecological concern, particularly at sea. In light of marine mammal stranding events associated with use of naval mid-frequency active sonar (MFAS), which often involve goose-beaked whales (Ziphius cavirostris), research has sought to detect and characterize Ziphius behavioral responses to MFAS.

methodsWe deployed Sound and Motion Recording and Telemetry (SMRT) tags on 13 Ziphius in the Southern California Bight. The study area includes the United States Navy’s Southern California Anti-submarine Warfare Range (SOAR), and deployments overlapped temporally with training exercises including MFAS use. We used hierarchical hidden Markov models (HHMMs), which provided a framework to model whale behavior and putative responses to MFAS at two relevant time scales: foraging dive cycles, which may last hours, and five-minute intervals, where finer-scale changes in movement and behavior are evident.

resultsWe analyzed 70.7 days of tag data (1697.6 h) representing 361 Ziphius foraging dive cycles, 52 with MFAS detections. Cumulative MFAS sound energy level (cSEL) per dive cycle was 69.9–160.3 dB re 1µPa2s (median 121.3, inter-quartile range (IQR) 21.7). We identified two dive-cycle states: “Typical” and “Variant.” In Variant state, durations of dive cycles, foraging dives, non-foraging dives, and echolocation periods were shorter on average but more variable, while time spent near the surface and net distance traveled were longer and more variable. According to Akaike’s information criterion (AIC), MFAS cSEL (rather than per-sound maximum received level) increased the probability of switching from Typical to Variant dive-cycle state, with MFAS cSEL also modulating transition rates between four fine-scale behavior states during Variant dive cycles only. Whales transitioned to Variant state in 12 of 52 MFAS-exposed dive cycles (23.1%), with observed Variant state dwell times ranging from 1 to 72.4 h (median 9.0, IQR 9.2).

conclusionsThis work has relevance for conservation and management, enhancing our understanding of Ziphius behavioral responses to operational MFAS in an area where naval training activities occur regularly. These analysis tools may prove useful in assessing responses to other acoustic disturbance and understanding how shorter-term behavior changes translate to longer-term consequences.