WeatherNext 3 directly learns from satellite observation data, bringing a route to weather forecasting that differs from traditional simulation.
WeatherNext 3 directly learns from satellite observation data, bringing a route to weather forecasting that differs from traditional simulation.
WeatherNext 3's new route is quite interesting: it directly learns from real-time satellite data to make weather forecasts, reducing dependence on traditional physical simulation pipelines. Google says the new version can provide finer hourly forecasts and hopes to improve forecast coverage in parts of Africa, Latin America, and the Asia-Pacific.
This type of model has potential value for agriculture, logistics, energy scheduling, and disaster warning, but weather services cannot be judged by just a pretty forecast map. When putting it into practice, it should continuously be checked against local observation station data, observing errors during extreme weather, missing data, and regional transfer, while retaining the judgment of meteorological professionals.
#天气AI #GoogleDeepMind #卫星数据 #气候科技 #预测模型 #人工智能