Google, as one of the leaders in AI-based weather forecasting models, has introduced the third version of its WeatherNext model. This update introduces the most significant change by incorporating raw satellite data into the model, thereby reducing the latency between current weather conditions and the generation of new forecasts.
AI models can deliver forecasting performance comparable to that of traditional models while requiring significantly less computational power, enabling them to be run repeatedly. In this version, WeatherNext leverages the "reanalysis" process, in which various atmospheric data sources are combined to create a coherent global picture of the atmosphere, ensuring that forecasts remain reliable even in areas lacking direct measurements.