Google-related company DeepMind greatly improves the accuracy of weather forecasts up to 90 minutes ahead, accurately approved by weather forecasters

With deep learning, with the same writing as Google as the parent companyStructural analysis of proteinsOrThe world’s strongest Go AIKnown for the development of an artificial intelligence companyDeep MindHowever, an in-depth generating model that can predict the probability of precipitation 90 minutes in advance with high accuracy.DGMRAnnounced growth. Predicting climate change within two hours is considered to be the most difficult problem in weather forecasting, and the introduction of this model is expected to significantly improve the accuracy of weather forecasts.

Efficient rainfall landscape using deep generating models of radar
https://www.nature.com/articles/s41586-021-03854-z

Now airing the next hour of rain | Deep Mind
https://deepmind.com/blog/article/nowcasting

In modern weather forecasts, the movement of fluids in the atmosphere is calculated to predict future weather.Weather number forecastHas been used. The Mathematical Weather Forecast is good for forecasting weather from about 6 hours ahead to about 2 weeks later, but it is said to be less accurate when forecasting weather within 2 hours.

Deep Mind Co-Chairman Suman Rowrie and colleagues have developed an in-depth creation model called “DGMR” to improve the accuracy of so-called “short-term forecasts” within two hours. Image creation, etc.Generative Enemy Network (GAN)The same method as above is designed to learn the movement of rain clouds captured by weather radar and to predict and create the movement of rain clouds 5 to 90 minutes ahead.

To verify the accuracy of the forecast results generated by DGMR, Rowrie et al. Two existing rainfall probability forecast models were prepared and asked to hide the names of 56 weather forecasters and assess the accuracy. As a result, DGMR was rated “highly accurate and effective” in 89% of cases. In the following, the actual cloud motion seen on the top left, and the DGMR on the top right is. Unlike PySTEPS (bottom left), rainfall intensity is often high, and DGMR balances the range of rainfall probabilities when deep learning blurs simulation results (bottom right). The actual tracking is close to the record.

The development team said, “We consider this to be an exciting piece of research, and our research will form the basis of new research, helping to integrate machine learning and environmental science, and make better decisions in weather forecasting. I hope it will be supported.”