The Chongqing Meteorological Bureau released the "Talent·12h" artificial intelligence weather forecast model, marking Chongqing's significant progress in the field of weather forecasting. This model focuses on the accurate forecast of disaster-causing precipitation in the Chengdu-Chongqing region. It is based on Huawei's Pangu global meteorological model and integrates local data and advanced technology to achieve high spatial and temporal resolution forecasts and significantly improve forecast accuracy. This move can not only effectively serve urban governance and low-altitude economic development, but also provide important support for disaster prevention and reduction, reflecting the important role of technological innovation in meteorological services.
At the 2024 Western Science and Technology Innovation Ecological Development Conference held recently, the Chongqing Meteorological Bureau officially released the "Talent·12h" artificial intelligence (AI) weather forecast model, which focuses on accurate forecasts of disaster-causing precipitation in the Chengdu-Chongqing region, marking Chongqing has taken an important step in the field of weather forecasting.
The "Talent·12h" large model is based on Huawei's Pangu global meteorological large model, which integrates ground meteorological conditions and high-precision terrain data in the Chengdu and Chongqing regions. It adopts a global and regional large model fusion architecture and combines the three-dimensional earth-specific transformer (3DEST) Models, multi-source heterogeneous data fusion technology and heavy precipitation small sample processing technology provide a solid foundation for improving forecast accuracy.
This large model can predict surface precipitation, temperature, wind and relative humidity in the Chengdu-Chongqing region every 6 hours for the next 12 hours, with a spatial and temporal resolution of 3 hours and 3 kilometers; at the same time, it can also predict the geopotential of 13 high-altitude pressure layers Altitude, specific humidity, temperature and wind with a spatiotemporal resolution of 3 hours and 25 km.
Gao Yudong, director of the Key Technology Research Laboratory for Disaster-Causing Precipitation Numerical Models and Artificial Intelligence Integrated Forecasting, said that the main forecast timeliness of this large model is between 2 hours and 12 hours in the future. The period within 2 hours mainly relies on meteorological observation data analysis, because this period is very close to meteorological disasters and most weather changes can be observed. Forecasts beyond 12 hours still need to rely on traditional numerical weather models, and existing AI large model technology requires further research and iteration.
Zhang Yan, director of the Chongqing Meteorological Observatory, said that this large model can be used in many fields such as megacity governance, low-altitude economy, and dangerous rock disaster early warning. Taking low-altitude economy as an example, meteorological conditions are an important factor affecting low-altitude flight safety. Since the entire low-altitude flight is in an atmospheric environment with strong convection and drastic changes, it is necessary to pay close attention to meteorological conditions from takeoff to landing, and the large AI weather forecast model can provide guarantee for the smooth operation of low-altitude aircraft such as drones.
In terms of forecast accuracy, the rainfall forecast accuracy for the first 6 hours of the "Talent·12h" artificial intelligence weather forecast model is up to 36% higher than that of traditional forecasts.
The successful release of the "Talent·12h" large model heralds the huge potential of artificial intelligence technology in the field of weather forecasting, provides a new direction for more accurate and timely weather forecasting in the future, and also provides a new direction for the economic and social development of Chengdu and Chongqing regions. Strong technological support.