Hong Kong Scientists Launch DDMS AI Model: New Breakthrough in Predicting Extreme Storms Earlier

Professor from the HKUST Department of Civil and Environmental Engineering, Su Hui (photo: x @theAGU)

JAKARTA - A team of scientists from the Hong Kong University of Science and Technology (HKUST) officially introduced the latest artificial intelligence (AI) system designed specifically to predict lightning storms and heavy rain with a much longer duration.

In a press conference held on Wednesday, January 28, the research team revealed that this system is capable of providing early warnings up to four hours before extreme weather occurs. This is a big leap compared to current technology which only has a prediction range of between 20 minutes to two hours.

The project leader and Professor from the Department of Civil and Environmental Engineering at HKUST, Su Hui, emphasized that the presence of this technology is very vital in the midst of increasing frequency of extreme weather due to climate change.

He explained that the use of AI combined with satellite data aims to give the government and emergency services more time to prepare in the face of flash flood and large storms that are increasingly difficult to predict.

The system, called Deep Diffusion Model based on Satellite Data (DDMS), works in a unique way compared to conventional forecasting models. DDMS implements a generative AI technique that inserts noise into the training data, where the system then learns to reverse the process to generate much more precise forecasts.

This technology was developed through collaboration with Chinese weather authorities and uses infrared brightness temperature data collected by the Fengyun-4 satellite between 2018 and 2021.

The sharpness of the DDMS model is proven by the increase in accuracy which reaches more than 15 percent with data update frequency every 15 minutes. The emphasis on the use of satellites is the main key because space sensors are able to detect the formation of convective clouds much earlier than land radar systems.

This advantage is considered a solution to the weather anomalies that have hit Hong Kong and southern China throughout 2025, where the number of typhoons and rainfall recorded far exceeded the seasonal norm to break the record for the highest storm warning five times in one year.

Currently, the China Meteorological Administration and the Hong Kong Observatory are working to integrate this DDMS AI model into their official weather forecasting protocol. With the data validation that has been carried out on the spring and summer samples of 2022 and 2023, this model is expected to become a new standard in the global disaster mitigation system.

This step is seen as a concrete effort to minimize material losses and human casualties through the use of artificial intelligence technology in the forefront of public protection.