The system, FengWu, is based on deep learning architecture of multi-model and multi-task, capable of generating high-precision global meteorological forecast results for the next 10 days in 30 seconds.
Global medium-range weather forecasts are significant in meteorological and climate forecasting to predict the atmospheric system status in the next 14 days.
Compared with the traditional physical models, FengWu reduces the error by 19.4 percent and extends the forecast range to 10.75 days based on reanalysis data, according to the Shanghai Artificial Intelligence Laboratory, one of its developers.
In the future, the system can complement the traditional physical models in weather forecast digitalization and support industries such as agriculture and forestry.