– Two papers published in Nature present advancements in AI technology for weather prediction.
– Huawei's AI model, Pangu-Weather, can quickly predict weekly weather patterns worldwide with comparable accuracy to traditional methods.
– A deep-learning algorithm outperformed other methods in predicting extreme rainfall, ranking first in around 70% of tests.
– These AI models can complement conventional forecasting methods, aiding authorities in preparing for adverse weather conditions.
– Pangu-Weather utilizes a deep neural network trained on 39 years of reanalysis data, analyzing weather variables simultaneously in seconds.
– The model demonstrated the ability to track tropical cyclones accurately, even without specific training data on cyclones.
– Other AI models, such as NowcastNet, can predict extreme rain up to three hours in advance, surpassing existing methods.
– NowcastNet incorporates data from various weather radars and is rooted in the principles of atmospheric physics, resulting in more comprehensive predictions.
– While AI can help determine the path of tropical cyclones, it may underestimate the intensity of extreme weather events.
– The performance of these AI systems in practical applications is yet to be fully assessed, considering the changing climate and potential unknowns.
– AI-based weather forecasting shows promise in improving predictions and providing timely information for preparations against extreme rain and other weather events.