Dose response severity functions for acoustic disturbance in cetaceans using recurrent event survival analysis

• Published in: Ecosphere
• Main Authors: Harris, C. M., Sadykova, D., DeRuiter, S. L., Tyack, P. L., Miller, P. J. O., Kvadsheim, P. H., Lam, F. P. A., Thomas, L.
• Format: Article in Journal/Newspaper
• Language: English
• Published: Wiley 2015
• Subjects: Killer Whale; Orca; Orcinus orca; Physeter macrocephalus; Sperm whale; Killer whale
• Online Access: http://dx.doi.org/10.1890/es15-00242.1; https://api.wiley.com/onlinelibrary/tdm/v1/articles/10.1890%2FES15-00242.1; https://esajournals.onlinelibrary.wiley.com/doi/pdf/10.1890/ES15-00242.1

Behavioral response studies (BRSs) aim to enhance our understanding of the behavior changes made by animals in response to specific exposure levels of different stimuli, often presented in an increasing dosage. Here, we focus on BRSs that aim to understand behavioral responses of free‐ranging whales and dolphins to manmade acoustic signals (although the methods are applicable more generally). One desired outcome of these studies is dose‐response functions relevant to different species, signals and contexts. We adapted and applied recurrent event survival analysis (Cox proportional hazard models) to data from the 3S BRS project, where multiple behavioral responses of different severities had been observed per experimental exposure and per individual based upon expert scoring. We included species, signal type, exposure number and behavioral state prior to exposure as potential covariates. The best model included all main effect terms, with the exception of exposure number, as well as two interaction terms. The interactions between signal and behavioral state, and between species and behavioral state highlighted that the sensitivity of animals to different signal types (a 6–7 kHz upsweep sonar signal [MFAS] or a 1–2 kHz upsweep sonar signal [LFAS]) depended on their behavioral state (feeding or non‐feeding), and this differed across species. Of the three species included in this analysis (sperm whale [ Physeter macrocephalus ], killer whale [ Orcinus orca ] and long‐finned pilot whale [ Globicephala melas ]), killer whales were consistently the most likely to exhibit behavioral responses to naval sonar exposure. We conclude that recurrent event survival analysis provides an effective framework for fitting dose‐response severity functions to data from behavioral response studies. It can provide outputs that can help government and industry to evaluate the potential impacts of anthropogenic sound production in the ocean.