Forecasting trajectories of Southern Resident killer whales with stochastic movement models incorporating direction modification

Animal movement forecasting is a novel area in ecological modelling that can significantly aid marine coastal management and facilitate timely conservation actions, particularly for endangered species. Animal movement modelling research has focused largely on statistical inference for long-term spatial distributions of animal movement and changes in behavioural states, with relatively little attention being given to short-term animal movement forecasting. We propose a straightforward forecasting framework that employs a continuous-time Ornstein–Uhlenbeck (O–U) velocity process as the foundation for a movement forecast system. Specifically, we incorporate a direction modification method to ensure directional persistence and to guide the movement towards preferred historical locations and pathways. We demonstrate our forecasting methods using movement data from 11 years of Southern Resident killer whale (SRKW) movement data. We evaluate its forecasting performance on a historical trajectory of the SRKW. The resultant ensemble forecast outcome defines a 90% probability region indicating the most likely region where animals may be found for a specific forecast horizon. Our stochastic dynamic framework successfully predicts an SRKW trajectory up to three hours ahead with incoming observations covered by our 90% probability regions. This shows the approach is suitable for our conservation objectives of using short-term SRKW forecasts to aid in dynamic management of marine traffic and to reduce whale-vessel interactions. Importantly, our forecasting framework is versatile and can be readily applied to a wide range of animal species, provided there is a historical trajectory database available. It can be initiated with observed locations and conducted in real time for ecological management plans, and it can also be integrated into data-assimilative forecasting.