Breakthrough in Typhoon Forecasting: Shanghai Typhoon Institute Develops Hybrid Machine Learning Model

Researchers from the Shanghai Typhoon Institute have made significant strides in typhoon forecasting by developing a hybrid machine learning model that leverages both traditional physics-based regional models and large-scale constraints from machine learning weather prediction models. This study, titled "Evaluating the Shanghai Typhoon Model Against State-of-the-art Machine-learning Weather Prediction Models: a Case Study for Typhoon Danas (2025)", showcases the potential of machine learning in improving forecast accuracy and highlights the technical roadmap for transitioning to a fully data-driven model.

Key Takeaways:

• The Shanghai Typhoon Model (SHTM) has been upgraded to leverage large-scale constraints from machine learning weather prediction models, resulting in an ML-physics hybrid framework.
• During Typhoon Danas (2025), the hybrid SHTM achieved substantially lower track errors than both the advanced ECMWF Integrated Forecasting System (IFS) and leading MLWP models such as PanGu and FuXi.
• The hybrid SHTM consistently maintained mean track errors below 200 km up to a forecast lead time of 108 hours, representing a significant advancement in forecast accuracy.
• This study highlights the importance of advances in physical modeling framework for improving the performance of future data-driven ML typhoon models.
• The research emphasizes the technical roadmap for transitioning from a physics-based typhoon model to a fully data-driven ML typhoon forecast system.
• The project was funded by the National Youth Science Foundation of China and the Research and Development of Key Technologies for Artificial Intelligence Regional Typhoon Forecasting Model project.

Statistics:

• The hybrid SHTM achieved track errors of 25% lower than the advanced ECMWF IFS model during Typhoon Danas (2025).
• The hybrid SHTM consistently maintained mean track errors below 200 km up to a forecast lead time of 108 hours.
• The panel of experts and reviewers acknowledged the research as a critical step towards developing data-driven ML typhoon models.
• The research was peer-reviewed and published in the journal Advances in Atmospheric Sciences.

Sources:

• VerticalNews. New Research on Machine Learning (Shanghai Typhoon Model) Detailed. October 14, 2025.
• Evaluating the Shanghai Typhoon Model Against State-of-the-art Machine-learning Weather Prediction Models: a Case Study for Typhoon Danas (2025). Advances in Atmospheric Sciences, 2025.
• NewsRx. Studies from Shanghai Typhoon Institute Provide New Data on Machine Learning [Evaluating the Shanghai Typhoon Model Against State-of-the-art Machine-learning Weather Prediction Models: a Case Study for Typhoon Danas (2025)]. Information Technology Newsweekly. October 14, 2025; p 1134.