Chinese researchers have developed a new AI-powered ensemble forecasting system that can predict typhoon tracks more accurately while also explaining why a storm's path may change. The system was created by scientists from the Institute of Atmospheric Physics of the Chinese Academy of Sciences and Fudan University, combining China's FuXi AI weather model with a physics-based forecasting method known as Orthogonal Conditional Nonlinear Optimal Perturbations (O-CNOPs). Unlike many AI weather models that focus primarily on speed, the new approach aims to improve both forecast accuracy and interpretability.
A key innovation is the system's ability to generate multiple realistic forecast scenarios rather than a single predicted path. Because tiny differences in atmospheric conditions can significantly alter a typhoon's trajectory, the AI evaluates how these uncertainties influence future movement. In tests covering 62 typhoon cases and 91 comparative forecasting experiments, the system matched leading global forecasting models for 24-hour predictions and outperformed them on forecasts ranging from 24 to 120 hours, reducing maximum track errors by as much as 32.33%. It also improved the accuracy of uncertainty estimates by up to 29.2%, helping forecasters better understand the range of possible storm tracks.
Researchers say the system represents an important step beyond purely data-driven AI models because it integrates physical atmospheric principles with machine learning. Rather than simply predicting where a typhoon will travel, it identifies the atmospheric conditions that could cause the storm to deviate from its expected path. This makes forecasts more transparent and useful for meteorologists, emergency planners, and disaster response agencies. Another advantage is that the framework does not require training additional massive AI models, making it more practical for operational weather forecasting.
The research team plans to further refine the system to improve forecasts of typhoon intensity in addition to storm tracks and to extend the framework to other extreme weather events. As climate change contributes to more frequent and severe weather disasters, AI-augmented forecasting systems like this could provide earlier and more reliable warnings, giving governments and communities more time to prepare, reduce damage, and protect lives.
