China launches AI models to automate typhoon and storm analysis

From observation to prediction

China has officially released its first batch of artificial intelligence models designed for extreme-weather analysis. These models were unveiled on October 29 at the 2026 Fengyun Satellite User Conference. The primary goal of this technology is to shift how satellite data is used from simply describing current weather to predicting what will happen next.

The models are built specifically on data from the Fengyun meteorological satellite series. By applying AI to this data, the systems can interpret complex imagery that humans previously had to review manually. This transition represents a move toward more automated and proactive weather monitoring systems.

As reported by pandaily.com, the release marks a significant step in integrating AI into national meteorological infrastructure. The initiative aims to provide a faster and more consistent starting point for weather professionals. This allows for a more streamlined workflow when handling rapidly changing weather events.

Automating typhoon analysis

The first specific model released is designed to interpret typhoon cloud imagery. It works by combining multichannel Fengyun cloud images with more than 20 related satellite products. The AI automatically analyzes a typhoon's position, its physical structure, and its current intensity to determine its likely evolution.

The system is capable of identifying seven different categories of typhoon features. Once it processes the data, it writes up its findings as a text-based interpretation. This process is highly efficient, as the model can run an interpretation every 15 minutes and requires only 27 seconds to generate the final text.

Developers tested the model using the 26th typhoon of the year, known as Surigae. The AI successfully identified the storm and produced a combined analysis of its wind radius and rainfall characteristics. This automated first pass helps support monitoring and forecasting efforts by providing immediate summaries of complex data.

Predicting severe thunderstorms

The second model focuses on providing earlier warnings for severe convective weather, such as thunderstorms. It utilizes continuous observations from the Fengyun-4 satellite over a two-hour period. The AI then applies these observations to forecast how specific cloud formations will develop in the near future.

The core logic of this model involves splitting cloud evolution into two distinct parts. It separates the overall movement of large weather systems from the local growth and decay of smaller ones. This distinction allows the system to track both broad atmospheric trends and localized, high-intensity events.

By using this dual-track approach, the technology extends the warning lead time to four hours. This provides more time for authorities to prepare for sudden, severe weather. It represents a shift toward using AI to understand the mechanics of cloud development rather than just tracking their current location.

New research and infrastructure

To support these new tools, China has introduced an artificial intelligence satellite meteorology technology and application innovation team. This group is based at the Xu Jianmin Meteorological Satellite Innovation Center. The team is composed of 49 researchers who represent 12 different institutions.

The researchers will work together on the multimodal representation of satellite weather data. Multimodal representation refers to the way different types of data, such as various satellite products and imagery, are combined for the AI to process. This work is essential for creating more comprehensive and accurate weather models.

The team's primary focus will be on the monitoring and early warning of extreme weather events. Their goal is to refine how AI interacts with satellite data to improve public safety. This collaborative effort ensures that the technology continues to evolve as more data becomes available from the Fengyun fleet.

Integration into existing tools

These new satellite models join a growing set of AI weather tools currently being used in China. Previously, many of these tools focused on medium-range forecasting models that ran on reanalysis data. Reanalysis data is a type of data used to study past weather patterns to improve future models.

The new models are different because they work directly from satellite observations. This is the specific type of data that forecasters rely on most when a storm is already forming. By processing live satellite feeds, the AI can provide more immediate utility during active weather events.

The integration of these models into the national framework suggests a move toward real-time AI assistance. While human forecasters still play a vital role, these tools provide a high-speed analysis of complex imagery. This allows for a more rapid response to the development of typhoons and severe thunderstorms across the region.