Deep learning predicts ocean plastic drift more accurately than models
Drift Field Net: Learning Ocean Lagrangian advection fields from in-situ and satellite observations
Machine Learning
Summary
Plastic trash in the North Pacific Ocean builds up because of ocean currents that move floating debris. The authors develop a new deep learning tool called Drift Field Net that uses satellite data to better predict how ocean currents will move floating particles. They train the model first on simulations and then improve it by making sure its predictions follow physical laws about particle movement. This method predicts drifting paths with smaller errors than current operational ocean models, helping future cleanup efforts.
What this means in practice
- •For environmental monitoring teams: Improve forecasts of plastic debris movement in ocean gyres to guide cleanup operations more precisely.
- •For marine traffic management: Use more accurate ocean surface flow predictions to improve routing and safety advisories for ships in affected areas.
Authors
Théo Archambault, Pierre Garcia, Mattia Romero, Anastase Charantonis, Dominique Béréziat
Abstract
The North Pacific Subtropical Gyre (NPSG) is a major accumulation zone for floating plastic debris, resulting from basin-scale convergent ocean circulation. Effective cleanup strategies in this region rely on accurate forecasts of Lagrangian particle drift. Here, we introduce Drift Field Net (DFN), a deep neural network that predicts ocean surface flow fields from operational satellite observations. DFN is trained using a novel two-stage strategy that combines pretraining on simulated data with Lagrangian fine-tuning based on an advection-consistent loss function. This physics-informed optimization directly improves the accuracy of particle trajectory predictions. We evaluate DFN against an operational physics-based forecasting system and demonstrate the potential of deep learning for ocean surface flow prediction. On in situ drifter trajectories, DFN reduces the mean positioning error by 20 km after a 7-day forecast compared with the operational model. Furthermore, Lagrangian fine-tuning with the proposed advection loss further reduces the positioning error by 10 km, highlighting the benefits of incorporating Lagrangian constraints into the training process.