Hong Kong scientists develop AI model that predicts severe thunderstorms four hours in advance

Hong Kong Scientists Achieve Breakthrough in Weather Forecasting

Researchers from The Hong Kong University of Science and Technology have developed a groundbreaking AI model that can predict extreme weather events including severe thunderstorms, Black Rainstorms, and heavy rainfall up to four hours in advance1

1

. The system, created in collaboration with the China Meteorological Administration and national meteorological institutions, represents a critical advance for climate resilience as communities face increasingly frequent weather extremes linked to climate change2

2

.

Led by Su Hui, Chair Professor in the Department of Civil and Environmental Engineering at The Hong Kong University of Science and Technology, the research team published their findings in the Proceedings of the National Academy of Sciences under the title "Four-hour thunderstorm nowcasting using a deep diffusion model for satellite data"1

1

. "We hope to use AI and satellite data to improve prediction of extreme weather so we can be better prepared," Su Hui told a press conference describing the work2

2

.

Deep Diffusion Model Transforms Forecast Accuracy

The new AI model, known as the Deep Diffusion Model based on Satellite Data (DDMS), applies state-of-the-art generative AI techniques to weather forecasting1

1

. During training, noise is added to satellite data, enabling the system to learn the reverse process of generating high-quality forecasts. This approach allows the model to predict extreme weather events with improving forecast accuracy of more than 15% at the 48-kilometer spatial scale compared to existing systems1

1

.

The research team trained the Deep Diffusion Model using infrared brightness temperature data collected by China's FengYun-4A satellite from 2018 to 2021, incorporating professional meteorological domain expertise to accurately capture the evolution of convective cloud structures1

1

. Model performance was validated using samples from the spring and summer seasons of 2022 and 20232

2

.

Addressing Critical Gaps in Early Warning Systems

Conventional weather forecasting relies on numerical weather prediction models that simulate future atmospheric conditions by solving complex fluid-dynamical equations. However, these systems require intensive computation and are highly sensitive to atmospheric chaos and limitations in observational data1

1

. For rapidly evolving convective systems including severe thunderstorms and rainstorms, accurate forecasts are often limited to just 20 minutes to two hours in advance, leaving governments, emergency services, and the public with critically limited time to prepare1

1

.

Dr. Dai Kuai, Postdoctoral Fellow and first author of the paper, explained the advantage of satellite data over traditional methods: "Conventional weather forecasting models rely mainly on ground-based radar, but radar signals are easily affected by terrain and precipitation composition and often detect changes only after convective clouds have already formed. This results in delays in forecast lead time"1

1

. By leveraging satellite data that monitor cloud evolution from space, the AI model can detect signs of convective development much earlier1

1

.

Operational Capabilities and Regional Coverage

The AI model delivers high-resolution, high-frequency forecasts updated at approximately 15-minute intervals, covering a region of about 20 million square kilometers including China, Korea, Southeast Asia, and surrounding areas1

1

. The system demonstrates stable performance across multiple spatial scales ranging from 4 kilometers to 48 kilometers and throughout different seasons1

1

.

Particularly notable is the model's accuracy in the critical 2-4 hour forecast window, where it excels precisely where conventional models fail most. Within this lead time, the system delivers reliable forecasts with accuracy improvements ranging from 3% to 16%, averaging 8.26%1

1

. Both the China Meteorological Administration and Hong Kong's Observatory are working to incorporate the model into their operational forecasts2

2

.

Responding to Escalating Climate Risks

The timing of this breakthrough is critical as extreme weather events have become increasingly frequent in recent years. Hong Kong issued four Black Rainstorm Warnings within just eight days last summer, while the city issued its highest rainstorm warning five times in 2025 and the second highest 16 times, setting new records according to its observatory1

1

2

2

. Regions such as Bali in Indonesia, southern Thailand, and other areas have also experienced severe flooding resulting in significant casualties and economic losses1

1

.

This research aligns with the core objectives of the State Key Laboratory of Climate Resilience for Coastal Cities, established with approval from the Ministry of Science and Technology of China. The laboratory operates under the directorship of Prof. Charles Ng Wang-Wai, Vice President for Institutional Advancement at HKUST1

1

. The system's ability to predict extreme weather events four hours in advance promises to strengthen disaster preparedness and transform early warning systems for vulnerable communities across Asia, providing governments and emergency services the critical time needed to respond effectively to increasingly frequent weather extremes1

1

2

2

.