{"title":"Accurate initial field estimation for weather forecasting with a variational constrained neural network","authors":"Wuxin Wang, Jinrong Zhang, Qingguo Su, Xingyu Chai, Jingze Lu, Weicheng Ni, Boheng Duan, Kaijun Ren","doi":"10.1038/s41612-024-00776-1","DOIUrl":null,"url":null,"abstract":"Weather forecasting is crucial for scientific research and society. Recently, deep learning (DL) methods have achieved significant advancements in medium-range weather forecasting. However, they generally depend on the initial fields generated by the computationally expensive four-dimensional variational (4DVar) data assimilation (DA) technique, which limits their real-time applicability in multivariate three-dimensional (3D) weather forecasting. Here we propose 4DVarFormer by exploring the potential of integrating the 4DVar constraint into an attention-based neural network. 4DVarFormer eliminates the need for background error covariance statistics and the complex adjoint model development. It can generate multivariate 3D weather states within 0.37 s. Moreover, 4DVarFormer can capture inter-variable relationships, allowing the assimilation of observed variables to correct unobserved variables. Hence, medium-range forecasts initiated by 4DVarFormer outperform those of DL-based DA methods and achieve performance comparable to the forecasts initiated by ERA5 reanalyses. These promising findings contribute to future advancements in integrated end-to-end DL weather forecasting systems.","PeriodicalId":19438,"journal":{"name":"npj Climate and Atmospheric Science","volume":" ","pages":"1-17"},"PeriodicalIF":8.5000,"publicationDate":"2024-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.nature.com/articles/s41612-024-00776-1.pdf","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"npj Climate and Atmospheric Science","FirstCategoryId":"89","ListUrlMain":"https://www.nature.com/articles/s41612-024-00776-1","RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"METEOROLOGY & ATMOSPHERIC SCIENCES","Score":null,"Total":0}
引用次数: 0
Abstract
Weather forecasting is crucial for scientific research and society. Recently, deep learning (DL) methods have achieved significant advancements in medium-range weather forecasting. However, they generally depend on the initial fields generated by the computationally expensive four-dimensional variational (4DVar) data assimilation (DA) technique, which limits their real-time applicability in multivariate three-dimensional (3D) weather forecasting. Here we propose 4DVarFormer by exploring the potential of integrating the 4DVar constraint into an attention-based neural network. 4DVarFormer eliminates the need for background error covariance statistics and the complex adjoint model development. It can generate multivariate 3D weather states within 0.37 s. Moreover, 4DVarFormer can capture inter-variable relationships, allowing the assimilation of observed variables to correct unobserved variables. Hence, medium-range forecasts initiated by 4DVarFormer outperform those of DL-based DA methods and achieve performance comparable to the forecasts initiated by ERA5 reanalyses. These promising findings contribute to future advancements in integrated end-to-end DL weather forecasting systems.
期刊介绍:
npj Climate and Atmospheric Science is an open-access journal encompassing the relevant physical, chemical, and biological aspects of atmospheric and climate science. The journal places particular emphasis on regional studies that unveil new insights into specific localities, including examinations of local atmospheric composition, such as aerosols.
The range of topics covered by the journal includes climate dynamics, climate variability, weather and climate prediction, climate change, ocean dynamics, weather extremes, air pollution, atmospheric chemistry (including aerosols), the hydrological cycle, and atmosphere–ocean and atmosphere–land interactions. The journal welcomes studies employing a diverse array of methods, including numerical and statistical modeling, the development and application of in situ observational techniques, remote sensing, and the development or evaluation of new reanalyses.