Typhoon track tracking and forecasting algorithm based on multi-source information.

The article focuses on a novel typhoon trajectory inversion model that integrates regional ocean environment forecasting with deep learning techniques to improve the accuracy of typhoon path predictions. Utilizing multi-source environmental data—including wind components at 10 meters and significant wave height—and employing a U-Net convolutional neural network framework, the model predicts typhoon center positions iteratively without relying on extensive historical track data. Tested on 57 North Atlantic typhoons from 2017 to 2020, the model demonstrated a 6.3% to 33.3% reduction in forecast errors for 24- to 48-hour lead times compared to traditional neural network methods, with strong correlations maintained up to 72 hours. Statistical and interpretability analyses revealed that the model performs best for typical typhoons with pronounced meteorological features, while challenges remain in forecasting atypical or rapidly moving storms, suggesting directions for future refinement.