Hong Kong Observatory Boosts AI for Extreme Weather Forecasting

Hong Kong’s Meteorological Authority Explores AI for Extreme Weather Prediction

The Hong Kong Observatory is exploring the potential of artificial intelligence (AI) to enhance its ability to forecast extreme weather events, according to a meteorologist. This initiative comes after AI models successfully predicted a heavy rain event that occurred on Tuesday, with forecasts issued as far as a month in advance.

He Yuheng, acting senior scientific officer at the Observatory, revealed that AI models had accurately anticipated the movement of a westward rainband that affected the Guangdong coast just days before it struck. The AI system also provided early warnings about the rainfall distribution, even though the predictions may not be 100% precise. He emphasized that these tools could still offer valuable insights to help the Observatory make more timely and informed decisions.

This development follows a record-breaking rainfall event in Hong Kong. On Tuesday, the city recorded 358.8 millimetres of rain at its Tsim Sha Tsui headquarters by 5pm, marking the highest daily rainfall in August since records began in 1884. The downpour led to the fourth black rainstorm signal being raised this year, which is the city's highest-level rain warning.

The black rainstorm signal was activated twice within six hours earlier in the week. First, it was raised at 11:45 pm on Monday, then lowered to a red rainstorm warning at 2:10 am on Tuesday. A second black rainstorm signal was issued more than three hours later at 5:50 am, and it lasted over 11 hours, causing significant flooding across the city.

The use of AI models by the Hong Kong Observatory has been in trial since mid-2023. According to He, these models have proven to be as effective as, or even better than, traditional methods when it comes to predicting large-scale weather patterns such as anticyclones or subtropical ridges.

Although the integration of AI is still in its early stages, He noted that the models have reached or even slightly exceeded the accuracy of conventional forecasting techniques. One of the key advantages of AI is its reduced margin of error when predicting tropical cyclone paths. For five-day forecasts, AI models show an error range of 300 to 500 kilometres, compared to 500 to 800 kilometres for traditional models.

However, He pointed out that predicting the intensity of typhoons remains a challenge for AI systems. This area requires further refinement and development to match the precision of traditional methods.

In addition to the weather-related challenges, infrastructure damage caused by the recent downpours has also drawn attention. A major road in the New Territories was damaged by a landslide triggered by Tuesday’s heavy rainfall. The Highways Department announced on Thursday evening that one lane of Bride’s Pool Road had been reopened for two-way traffic, signaling progress in restoring normal conditions.

As the Hong Kong Observatory continues to refine its AI-based forecasting systems, the goal is to improve the accuracy and timeliness of weather alerts, ultimately enhancing public safety and preparedness for future extreme weather events.