AI Predicts Severe Storms 4 Hours Early: Hong Kong Breakthrough

The researchers at Hong Kong have recently made some innovations by providing a disruptive model of artificial intelligence that can identify severe thunderstorms as early as four hours before they occur. 

The dependence on the 20 minute lead time will be done away with, because this contribution made by Hong Kong University of Science and Technology has the potential to save lives in the wake of rising climatic disruptions.

Said Su Hui, chair professor of the university's civil and environmental engineering department, who led the project.

We hope to use AI and satellite data to improve prediction of extreme weather so we can be better prepared.

The Storm Detection Technological Progress

The team, headed by Chair Professor Su Hui, developed the Deep Diffusion Model (DDMS) based on Satellite Data. 

The Deep Diffusion Model based on Satellite Data (DDMS), the team's new AI framework, was trained using infrared brightness temperature data gathered by China's Fengyun-4 satellite between 2018 and 2021.

It was created in cooperation with China's weather authorities and, according to the team, has increased forecast accuracy by more than 15%  and forecasts are updated every 15 minutes.

Unlike radar systems, satellites are able to capture cloud formation sooner, hence minimizing prediction errors and issuing early warnings.

Why Timing Matters: Climate Pressure and impact?

The previous year reached unprecedented heights of rainfall in Hong Kong with authorities issuing rainstorm red warnings 5 times and amber warnings 16 times as typhoons and heavy rains caused havoc to southern China. 

Extreme weather events, which were caused by the change in climate, created pressure on the legacy forecasting structures.  The four-hour lead time provides emergency responders with actionable early-warning capability to the emergency responders which can possibly reduce the damage associated with floods.

Implementation Outlook

It is projected to be placed into practical use during late 2026, and when combined with physical weather models, it will create a higher level of predictability with the aid of AI. With the approaching typhoon season, this will be a great move forward in the field of AI-based meteorology, being met with protection of millions of people against the threat of climate challenges in Asia. 

Chen Zhenlin, head of the China Meteorological Administration said,

The country will pilot a new imminent warning system, apply artificial intelligence to refine typhoon and heavy rainfall forecasts, and develop new prediction products for extreme weather.

If this system is launched on a worldwide scale in order to include the satellite networks across the globe then this system has the potential of redefining disaster preparedness worldwide.