谷歌DeepMind和谷歌研究团队推出了WeatherNext 3，这是一款天气预报AI模型，称其为迄今最先进准确的全球天气预测系统1。该模型在分辨率和预报精度上实现了重大突破，相比前代产品性能显著提升。
WeatherNext 3采用5至10公里分辨率生成每小时更新的天气预报，较前代25公里分辨率、6小时更新周期的模式提升约5倍1。在降水预报准确度方面，相比IMERG基线改进60%，相比MRMS改进30%，在早期引导时间内雨量计方面改进10%1。该模型还在1天以上提前期的中期全球预报中，降水预报准确度提升50%1。该模型的独特之处在于直接学习实时卫星观测数据而非依赖传统数值天气预报模型，从而消除了6小时数据滞后1。
WeatherNext 3不仅能预报传统气象要素，还具备100米高度风速预报能力，并可提供高分辨率云覆盖率和太阳辐射水平的预测1。该模型已集成至谷歌搜索、Gemini应用、谷歌地图和谷歌地球引擎等多个产品中1。
Google DeepMind and Google Research have introduced WeatherNext 3, a significantly more advanced weather forecasting AI model designed to deliver hourly global weather predictions at substantially higher resolution than its predecessors 1. The system learns directly from real-time satellite observations and generates forecasts at a 5- to 10-kilometer resolution, representing approximately a fivefold improvement over the previous generation's 25-kilometer resolution and 6-hour update frequency 1. By processing satellite data directly rather than relying on traditional numerical weather prediction models, WeatherNext 3 eliminates the six-hour data lag that has historically constrained weather forecasting accuracy 1.
The model demonstrates marked improvements across multiple meteorological parameters. Precipitation forecasting shows a 60% accuracy gain compared to the IMERG baseline, a 30% improvement over MRMS, and a 10% enhancement relative to early lead-time rain gauge measurements 1. For extended global forecasts with lead times exceeding one day, precipitation accuracy has improved by 50% 1. Beyond precipitation, WeatherNext 3 can predict wind speeds at 100-meter altitude—the typical hub height for wind turbines—as well as high-resolution cloud cover and solar radiation levels 1.
Google has already begun deploying WeatherNext 3 across its product ecosystem, including Google Search, Gemini, Google Maps, and Google Earth Engine 1. These integrations demonstrate the model's operational readiness and its potential to enhance weather information accessibility for users worldwide.
评论
还没有评论，欢迎留下第一条。