"Improving Weather Forecast Accuracy" Google DeepMind Develops AI-Based Model

Google has developed a weather simulator model that improves accuracy using artificial intelligence (AI).

Google's AI organization, Google DeepMind, published a paper on the 22nd (local time) in the scientific journal Nature introducing the weather simulator 'Neural GCM'. Developed in collaboration with the European Centre for Medium-Range Weather Forecasts (ECMWF), this system combines traditional weather forecasting methods with AI machine learning (ML).

Traditional models are used for calculations involving large atmospheric changes necessary for weather prediction, while AI is applied to fine-scale climate phenomena, such as fog in specific regions where the traditional model does not perform well.

In particular, AI enables highly advanced weather forecasting faster and with less computational power. The paper explained that the weather forecasts of the Neural GCM model are as accurate as ECMWF's forecasts, which extend up to 15 days. The research team is currently developing a function to predict hurricanes one year in advance, and if this function proves useful, it could help prepare for storms in advance and build infrastructure accordingly.