A real-time data assimilative forecasting system for animal tracking.

Area of Science:

• Movement Ecology
• Ecological Forecasting
• Computational Biology

Background:

• Real-time animal tracking data is increasingly available.
• Predictive models for animal movement are crucial for ecological research and conservation.
• State-space models are effective for retrospective analysis but less explored for forecasting.

Purpose of the Study:

• To develop and implement a data assimilative framework for probabilistic animal movement forecasting.
• To integrate real-time location data with stochastic movement models.
• To demonstrate the framework's utility in predicting animal trajectories and estimating behavioral parameters.

Main Methods:

• Utilized a state-space model within a data assimilative framework.
• Employed ensemble-based sequential Monte Carlo methods (particle filter).
• Applied the framework to an idealized continuous-time random walk model and to southern resident killer whales (SRKW).

Main Results:

• Successfully forecasted animal locations and trajectories using real-time data assimilation.
• Demonstrated online estimation of behavioral parameters and the impact of habitat preference functions.
• Achieved short-term forecasts for SRKW in the Salish Sea with moderate prediction error (<5 km) up to 2.5 hours in advance.

Conclusions:

• The developed forecasting framework effectively synthesizes diverse data types for improved animal movement modeling.
• This approach enhances behavioral understanding and has the potential to advance movement ecology.
• Forecasting animal movements can provide critical lead time for mitigating human-wildlife conflicts, such as vessel-whale interactions.