Modeling the potential for vessel collision with southern resident killer whales

Abstract

Agent-based models can be used to simulate the behavior of animals within a stimulus field. For example, such models are used to estimate the sound exposure of simulated animals ("animats") moving within a computed sound field. This approach has not yet been used to estimate the probability of vessel collision between marine mammals and ships. One reason is the difficulty of defining aversive behavior in response to disturbance. While many animals display aversion, models often use a simplified approach due to insufficient data. Building on the JASCO animal simulation model including noise exposure (JASMINE), a vessel collision framework was developed for southern resident killer whales (SRKWs) using modeled vessel sound fields and AIS data. Animats were programmed to increasingly avoid louder, closer, and additional vessels by changing their heading, speed, and behavioral state based on published data from SRKWs. Animats that remained within a calculated "encounter radius" despite aversive behavior were considered struck by the vessel. Parameters governing aversive behaviors were calibrated by comparing the proportion of animats struck in the simulation with the collision probability for real-world SRKWs. This model provides a starting point to model the risk of vessel collision for SRKW and other species.