Equation-free mechanistic ecosystem forecasting using empirical dynamic modeling.

Area of Science:

• Ecology
• Fisheries Science
• Time Series Analysis

Background:

• Traditional equilibrium-based models inadequately predict natural and fisheries systems with nonlinear dynamics.
• These models often fail due to time-varying parameters, poor out-of-sample prediction, and misidentified drivers in nonlinear systems.
• Static equilibrium models persist despite these limitations.

Purpose of the Study:

• To evaluate Empirical Dynamic Modeling (EDM) as an alternative to conventional fisheries models.
• To compare the forecasting accuracy of EDM with traditional models using real-world fisheries data.
• To assess the utility of EDM in incorporating environmental variables for improved recruitment forecasting.

Main Methods:

• Utilized time series data from nine sockeye salmon (Oncorhynchus nerka) stocks in the Fraser River.
• Performed a direct comparison between contemporary fisheries models and equivalent EDM formulations.
• Incorporated spawning stock and environmental variables within EDM to forecast recruitment.

Main Results:

• EDM models demonstrated superior accuracy and precision in forecasting sockeye salmon recruitment.
• EDM showed significant improvements when environmental factors were included, outperforming extensions of the Ricker spawner-recruit equation.
• The study provides the first real-data comparison of EDM and contemporary fisheries models for recruitment forecasting.

Conclusions:

• EDM is strategically valuable for integrating environmental influences into fisheries forecasts.
• EDM provides insights into the operational mechanisms of environmental factors in forecasting models.
• EDM facilitates equation-free mechanistic forecasting applicable to fisheries management contexts.