基于人工智能大模型的中期全球气象预报新方法
Accurate Medium-Range Global Weather Forecasting Based on Large AI Models
田奇 1毕恺峰 1谢凌曦1
作者信息
- 1. 华为云计算技术有限公司,深圳 518129
- 折叠
摘要
天气预报对科学和社会具有重要意义.传统数值气象预报(NWP)方法需要大量算力,且在预报精度方面遭遇瓶颈.基于三维神经网络,构建盘古气象大模型,用于准确预报中期全球气象,并且采取地球特定先验和层次化时域聚合策略,处理天气数据中的复杂模式,同时降低中期全球气象预报中的累积误差.利用1979-2017年全球气象数据训练之后,发现盘古气象大模型对所有测试变量的确定性气象预报精度都超越了欧洲中期天气预报中心的综合预报系统(IFS),且在极端气象预报和集成气象预报中也表现良好.当使用再分析数据做初始化时,其跟踪热带气旋的准确性也超过了欧洲中期天气预报中心高分辨率预报系统的结果.
Abstract
Weather forecasting is important for science and society.The conventional numerical weather prediction(NWP)method is computationally expensive and it is increasingly difficult to improve the forecast accuracy.This paper establishes Pangu-Weather,an AI-based weather forecasting system based on 3D neural networks,and equips it with Earth-specific priors and a hierarchical temporal aggregation strategy to deal with complex patterns in weather data and reduce accumulation errors in medium-range forecasting.Trained on 1979-2017 global weather data,Pangu-Weather obtains stronger deterministic forecast results on reanalysis data in all tested variables when compared with the operational integrated forecasting system(IFS)of the European Centre for Medium-Range Weather Forecasts(ECMWF).Pangu-Weather also works well with extreme weather forecasts and ensemble forecasts.When initialized with reanalysis data,the accuracy of tracking tropical cyclones is also higher than that of ECMWF-HRES.
关键词
数值气象预报/人工智能/深度神经网络/中期气象预报/盘古气象大模型Key words
numerical weather prediction/artificial intelligence/deep neural networks/medium-range weather forecasting/Pangu-Weather引用本文复制引用
出版年
2024
NETL
NSTL
万方数据