China has deployed a new machine-learning model capable of forecasting typhoon rapid intensification (RI) within 12 and 24-hour windows, according to researchers at the Shenzhen Institutes of Advanced Technology (SIAT). The system, which utilizes ensemble algorithms to predict sudden wind speed spikes, is now operational at the National Meteorological Center to help mitigate risks from destructive storms like Yagi and Hato.
How does the new typhoon model improve forecasting?
The model replaces traditional statistical-dynamical methods that often struggle with the nonlinear, fast-moving nature of storm intensification. According to SIAT lead researcher Li Qinglan, the system identifies physical links between a typhoon’s inner-core structure and its potential for sudden growth. By calculating a “symmetric ratio,” the model detects when a storm develops a ring-like structure, which serves as a primary indicator that rapid intensification is imminent. The system issues an alert once a majority of its four integrated sub-models predict an intensification event.
In meteorology, “rapid intensification” is defined as a maximum sustained wind speed increase of more than 15 meters per second within 24 hours, or 10 meters per second within 12 hours.
Why is rapid intensification a persistent challenge?
Forecasting intensity remains difficult because it depends on a volatile mix of environmental backgrounds, land-sea surface interactions, and inner-core convection. Conventional models frequently fail to capture these complex interactions, leading to delays in emergency warnings. As noted by the China Association for Science and Technology, identifying these patterns was ranked as one of the top 10 frontier scientific problems for 2025. By introducing a “sea-land ratio” index, the SIAT team has created a quantitative way to track how land distribution affects a storm’s behavior, providing a sharper analytical tool than previous methods.
How does this model compare to global systems?
During testing, the SIAT team compared their ensemble model against the operational forecast system used by the US National Hurricane Center. Using data from North Atlantic tropical cyclones between 2016 and 2020, researchers found that their model achieved a higher probability of detection while maintaining a lower false alarm rate. Lyu Xinyan, a senior engineer at the National Meteorological Center, stated that this technology now serves as a critical reference point for China’s national typhoon intensity forecasts.
Future trends in meteorological machine learning
The shift toward ensemble machine learning suggests a move away from reliance on single-physics equations. Future meteorological research is expected to lean heavily on high-resolution spatial data to better understand “inner-core symmetry.” As these models integrate more real-time sensor data, the window for disaster preparedness—currently set at 24 hours—could potentially expand, giving coastal communities more time to evacuate or reinforce infrastructure.
Frequently Asked Questions
What is the main advantage of the new SIAT model?
It uses an ensemble of four machine-learning algorithms to predict rapid wind speed increases, resulting in fewer false alarms compared to traditional statistical models.
How does a typhoon’s “inner core” predict its strength?
According to Li Qinglan, a highly symmetric, ring-like inner core structure is a reliable precursor to rapid intensification.
Is this model currently in use?
Yes, the model has completed operational deployment and is currently being utilized by China’s National Meteorological Center.
Stay updated on the latest breakthroughs in climate science and disaster mitigation. Subscribe to our newsletter for weekly deep dives into emerging technology.
