Before rapid intensification, a typhoon's inner core becomes a perfect ring
Along the coastlines of East Asia, where typhoons have long arrived with sudden and devastating force, a new forecasting system now watches for the quiet structural signatures that precede catastrophe. Researchers at the Shenzhen Institutes of Advanced Technology have deployed a machine-learning model capable of predicting rapid intensification — the abrupt surge in storm strength that has repeatedly outpaced human preparation — up to 24 hours before it occurs. The system, born from the recognition that a typhoon's inner core grows eerily symmetric before it accelerates, represents science's answer to one of meteorology's most persistent and costly blind spots. In the space between a forecast and a landfall, lives are measured.
- Typhoons Rammasun, Hato, and Yagi each intensified suddenly before striking land, leaving casualties and destruction in their wake — proof that the gap between forecast and reality carries a human price.
- Rapid intensification, where wind speeds surge 15 meters per second or more in a single day, has consistently overwhelmed conventional forecasting models built on equations that cannot capture nature's nonlinear leaps.
- A team led by Li Qinglan discovered that a storm's inner core becomes strikingly symmetric just before dangerous acceleration — a physical fingerprint that machine learning could be trained to recognize.
- Their ensemble model, combining four algorithms into a single forecasting system, outperformed the U.S. National Hurricane Center's operational tools in historical testing, detecting more events with fewer false alarms.
- The system has cleared operational testing at China's National Meteorological Center and is now active, offering both 24-hour and 12-hour forecast windows to emergency managers and coastal communities.
A typhoon's most dangerous moment often arrives without warning. In the hours before landfall, a storm can abruptly surge in strength — wind speeds climbing 15 meters per second or more in a single day — transforming from a serious threat into a catastrophic one. Typhoon Rammasun in 2014, Hato in 2017, and Yagi in 2024 all underwent this sudden acceleration before striking land, each leaving behind significant casualties and economic devastation. The problem has been understood for years. A reliable solution has remained out of reach.
Now researchers at the Shenzhen Institutes of Advanced Technology, part of the Chinese Academy of Sciences, have deployed a forecasting system designed to predict these dangerous accelerations up to 24 hours in advance. Led by Li Qinglan, the team built an ensemble of four machine-learning algorithms that issues a forecast when a majority of its sub-models agree rapid intensification is coming. The effort directly addresses a challenge China's scientific community named in 2025 as one of the nation's top ten frontier problems.
The key insight was physical: before a typhoon rapidly intensifies, its inner core develops a highly symmetric, ring-like structure. The more geometrically regular that core becomes, the greater the danger. The researchers translated this observation into two quantitative indices — one tracking the storm's interaction with coastlines, another measuring the symmetry of its inner convective structure — giving the machine-learning models a grounded physical foundation to work from.
Tested against North Atlantic historical data from 2016 to 2020 and benchmarked against the U.S. National Hurricane Center's operational system, the new model detected rapid intensification more reliably and produced fewer false alarms — a balance that matters enormously, since missed warnings cost lives while excessive false alarms erode the public trust that makes preparedness possible.
The system has now entered real-world use at China's National Meteorological Center, providing both 24-hour and 12-hour forecast windows to emergency managers and coastal communities. It will not stop storms from intensifying. But it may give people the warning they need to get out of the way.
A typhoon's most dangerous moment often comes without warning. In the hours before landfall, a storm can abruptly gather strength—wind speeds climbing by 15 meters per second or more in a single day—transforming from a serious threat into a catastrophic one. This phenomenon, called rapid intensification, has repeatedly caught forecasters off guard. Typhoon Rammasun in 2014, Hato in 2017, and Yagi in 2024 all underwent this sudden acceleration before striking land, each leaving behind significant casualties and economic devastation. The problem has been clear for years. The solution has been elusive.
Now researchers at the Shenzhen Institutes of Advanced Technology, part of the Chinese Academy of Sciences, have deployed a new forecasting system designed to predict these dangerous accelerations up to 24 hours in advance. The model, developed by a team led by Li Qinglan, combines four machine-learning algorithms into a single ensemble system. When more than half of these sub-models agree that rapid intensification will occur, the system issues a forecast. The work represents a direct response to a challenge that China's scientific community identified in 2025 as one of the nation's top ten frontier problems.
The core insight came from studying the physics of what happens inside a typhoon before it intensifies. Li and his team discovered that rapid intensification is preceded by a distinctive structural change: the storm's inner core develops a highly symmetric, ring-like formation. The more symmetric that inner core becomes, the higher the likelihood of dangerous acceleration. To operationalize this observation, the researchers created two quantitative measures—a sea-land ratio that tracks how the storm's path interacts with coastlines and land masses, and a symmetric ratio that measures the geometric regularity of the inner core's convective structure. These indices provided the physical foundation for the machine-learning models to work from.
The challenge of forecasting typhoon intensity has always been that multiple factors interact in ways that resist simple prediction. A storm's evolution depends on its internal structure, the environmental conditions surrounding it, and the complex interactions between ocean and land surfaces. Conventional statistical and dynamical methods struggle to capture the nonlinear character of these changes—the way small shifts in conditions can trigger sudden, outsized responses. Machine learning offered a way to recognize patterns that human-designed equations might miss.
When the team tested their new model against historical data from the North Atlantic between 2016 and 2020, comparing it directly to the operational forecast system used by the U.S. National Hurricane Center, the results showed measurable improvement. The new model detected rapid intensification events more reliably and generated fewer false alarms—a critical balance in forecasting, where missed warnings cost lives but excessive false alarms erode public trust and preparedness.
The system has now completed operational testing at China's National Meteorological Center and entered real-world use. It provides both a 24-hour forecast window and a shorter 12-hour service, giving emergency managers and coastal communities more time to prepare. According to Lyu Xinyan, a senior engineer at the National Meteorological Center, this technology now serves as an important reference point for China's typhoon intensity forecasting operations.
What makes this development significant is not just the technical achievement but the practical consequence. Typhoons that undergo rapid intensification before landfall are among the most destructive natural disasters in the Western Pacific. Better forecasting means better evacuation planning, more time for communities to secure infrastructure, and potentially fewer lives lost. The model won't prevent storms from intensifying. But it may give people the warning they need to get out of the way.
Notable Quotes
Prior to rapid intensification, a typhoon's inner core typically develops a highly symmetric ring-like structure. A more symmetric inner core indicates a higher likelihood of rapid intensification occurrence.— Li Qinglan, lead researcher, Shenzhen Institutes of Advanced Technology
24-hour rapid intensification forecast technology now provides an important reference for China's typhoon intensity forecasting.— Lyu Xinyan, senior engineer, National Meteorological Center
The Hearth Conversation Another angle on the story
Why does a typhoon suddenly intensify? What's actually happening inside the storm?
The inner core develops a symmetric structure—almost like a perfect ring. When that symmetry increases, the storm's organization improves, and it can rapidly strengthen. It's a physical signal that the conditions are right for acceleration.
So you're saying the shape of the storm predicts its behavior?
Exactly. Before rapid intensification, there's a structural transformation. The researchers quantified this—they measured the symmetry and tracked how the storm interacts with land and sea. Those measurements became the foundation for the machine-learning models.
Why couldn't the old forecasting methods catch this?
They relied on equations designed by humans to capture linear relationships. But typhoon intensification is nonlinear—small changes can trigger sudden, large effects. Machine learning can recognize patterns in data that traditional equations miss.
How much better is this new system?
In testing against U.S. data, it detected rapid intensification events more often and with fewer false alarms. That's the hard part—you need both sensitivity and specificity. Too many false alarms and people stop listening.
What happens now that it's deployed?
It's in real-world use at China's National Meteorological Center. It gives forecasters and emergency managers a 24-hour window to prepare—time to evacuate, to secure infrastructure, to save lives.