糖心视频

For decades, medium-range weather forecasting鈥攑redicting conditions 1 to 5 days ahead鈥攈as relied heavily on traditional numerical models. However, this approach often struggles when applied to specific regions with limited historical data.

Researchers at Northwestern Polytechnical University in China have now proposed a novel deep learning鈥揵ased framework that dramatically improves the accuracy of forecasts, even when data are limited.

To address key challenges in regional forecasting, the team introduced a new method that integrates three major innovations: the use of semantic segmentation models originally designed for medical image analysis; a learnable Gaussian noise mechanism that improves the model's robustness; and a cascade prediction strategy that breaks the forecasting task into manageable stages. The study is published in recently.

"Our goal was to make regional forecasting smarter, faster, and more reliable, even in data-limited scenarios." says Associate Professor Congqi Cao, corresponding author of the study. "This is especially valuable for areas where a dense network of meteorological observations is not available."

The method was tested on the East China Regional AI Medium Range Weather Forecasting Competition dataset, which includes 10 years of reanalysis data from ERA5. The task involved using past atmospheric variables to predict five key surface weather indicators鈥攊ncluding temperature, wind, and precipitation鈥攅very 6 hours for the next 5 days.

The results speak for themselves: the model achieved significant improvements in prediction performance, outperforming many mainstream global AI forecasting models. Specifically, the method reduced temperature forecast errors by 9.3%, improved the precipitation F1-score by 6.8%, and lowered wind speed errors by 12.5%.

"This is the first time semantic segmentation and learnable noise mechanisms have been used together for regional weather forecasting," explains Prof. Cao. "It opens up new possibilities for accurate forecasting in other data-scarce regions."

Looking ahead, the team plans to extend their method to real-time systems and apply it to more regions across China. They hope their work will eventually serve public safety, agriculture, and disaster prevention needs鈥攄elivering smarter, faster, and more local forecasts when they matter most.