Predicting long-term climate trends is possible with the help of artificial intelligence

Researchers in the US and UK claim to have trained an atmospheric model, NeuralGCM-2.8, for a mid-latitude wind balance within five years and tested it for over 21 unseen years. It was more accurate in depicting the vertical structure of the winds, compared with GraphCast data. However, ECMWF-HRES still shows a closer alignment to ERA5 data than NeuralGCM does.