Artificial intelligence makes weather forecasting more advanced and accurate

Recently, Chinese and American scientists reported on the potential of artificial intelligence (AI) to assist in weather forecasting in two separate studies published in Nature. One method predicts global weather patterns up to a week in advance, while the other predicts short-term weather, such as extreme precipitation events. These AI-assisted weather forecasting methods are as accurate as existing methods and may be able to predict weather phenomena that were previously difficult to predict. However, the researchers say further assessment and involvement in the traditional weather forecasting field is needed before considering how and whether these new methods complement or replace existing forecasting systems.

Weather forecasting plays a key role in helping to save lives and minimize property damage, especially as climate change leads to more frequent extreme weather events. By far the most accurate forecasting system is numerical weather forecasting, which relies heavily on physical equations but is demanding and often slow, taking hours on a single simulation. In recent years, some AI-based methods have the potential to significantly speed up weather forecasting, but often less accurate than numerical weather forecasting.

Tian Qi, chief scientist of HUAWEI CLOUD AI field, and collaborators reported that Pangu-Weather’s AI-based weather forecasting system can predict global weather up to a week in advance. The model was trained using 39 years of global reanalysis weather data. The prediction accuracy of Pangea meteorology is equivalent to the comprehensive forecasting system used by the European Medium-Range Weather Forecasting Center, the world’s best numerical weather prediction system, and is more than 10,000 times faster at the same spatial resolution. Pangea can also use a three-dimensional model to predict various altitude levels, providing more complete and detailed predictions than its predecessor AI system.

In another study, Michael Jordan and colleagues at the University of California, Berkeley reported that the NowcastNet model predicts precipitation in real time by combining physical laws and deep learning. Forecasting refers to very short-term weather forecasts, up to 6 hours in advance, and therefore provide detailed information about the immediate weather. Nowcasting is important for risk prevention and crisis management of extreme precipitation events. Based on radar observation data from the United States and China, NowcastNet can make high-resolution precipitation predictions for an area of 2048 km × 2048 km up to 3 hours in advance. 62 meteorologists assessed the predictive power and value of extreme precipitation from different models; NowcastNet outperformed other leading methods in about 70% of its predictions, ranking first. The findings suggest a predictive advantage in rainfall rates, especially extreme precipitation events that were previously considered difficult to predict.

Imme Ebert-Uphoff and Kyle Hilburn of Colorado State University noted in a News & Opinion article published contemporaneously that AI has “great potential” for weather forecasting tasks. But they also note that “the risks require meteorologists to learn to design, evaluate and interpret such systems.” (Source: Feng Weiwei, China Science News)

Image courtesy of Pixabay

Related paper information:

Source link

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button