Modelling post-release mortality of loggerhead sea turtles exposed to the Hawaii-based pelagic longline fishery

Loggerhead sea turtles Caretta caretta are an endangered species exposed to anthropogenic hazards such as pelagic longline fisheries. Many loggerheads caught in these fisheries are alive when released from the gear, but many probably die soon after because of hook injuries or line entanglement. Robust estimates of post-release mortality are essential for stock assessment and evaluating the benefit of releasing turtles caught alive in the gear, yet none are available for any sea turtle species. Here, the post-release mortality of 40 loggerheads caught in the Hawaii-based pelagic longline fishery was investigated using satellite telemetry deployed by a National Marine Fisheries Service (NMFS) observer program. We modelled time-to-failure of all transmitters using nonparametric statistical modelling (Kaplan-Meier-Turnbull, local regression) to derive survival and hazard functions for light and deep hooked loggerheads. There was a significant difference between the survival functions for light and deep hooked loggerheads within 90 d of release, but no difference between survival functions after this time. But satellite transmitters fail for many reasons (defects, battery failure, transmitter detachment, turtle death), which results in a hazard function that confounds these competing risks. Hence we propose that it might not be possible to infer true post-release mortality based on satellite telemetry unless the cause of each transmitter failure is known, which is rarely the case. We discuss other survey design and statistical modelling challenges involved in the evaluation of post-release mortality based on satellite telemetry.