Article(id=1329420060856267557, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1329420000110161984, articleNumber=null, orderNo=null, doi=10.7544/issn1000-1239.202660002, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=null, receivedDate=1767110400000, receivedDateStr=2025-12-31, revisedDate=1777996800000, revisedDateStr=2026-05-06, acceptedDate=null, acceptedDateStr=null, onlineDate=1794888041937, onlineDateStr=2026-11-17, pubDate=1782835200000, pubDateStr=2026-07-01, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1794888041937, onlineIssueDateStr=2026-11-17, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1794888041937, creator=13701087609, updateTime=1794888041937, updator=13701087609, issue=Issue{id=1329420000110161984, tenantId=1146029695717560320, journalId=1324759914146529478, year='2026', volume='63', issue='7', pageStart='1639', pageEnd='1903', issueExtLink='null', onlineDate='null', pubDate='1782835200000', pubDateStr='2026-07-01', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1794888027455, creator='13701087609', updateTime=1794892162905, updator='13701087609', preIssue=null, nextIssue=null, articleTotal=null, ext={EN=IssueExt(id=1329437345918595829, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1329420000110161984, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1329437345918595830, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1329420000110161984, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null, downloadFileDto=null}, startPage=1690, endPage=1711, ext={EN=ArticleExt(id=1329420061120508710, articleId=1329420060856267557, tenantId=1146029695717560320, journalId=1324759914146529478, language=EN, title=Survey of Large Meteorological Models in Agricultural Applications, columnId=1329420000701558849, journalTitle=Journal of Computer Research and Development, columnName=Special Issue on “AI4Science: A New Paradigm in Scientific Research”, runingTitle=null, highlight=null, articleAbstract=

In recent years, deep learning has achieved remarkable progress in meteorological forecasting and climate modeling, exerting a growing influence on agricultural applications. While numerous review articles have systematically examined related methodologies, most focus on model architectures, task types, or application domains, with limited attention given to the agricultural perspective. This paper provides a comprehensive review of the development of large meteorological models (LMMs), structured around model architecture. We trace the evolution from traditional numerical methods to deep learning approaches and categorize existing models into six major classes: classical deep neural networks, graph neural networks, Transformer-based architectures, state space models, generative models, and physics-informed/multimodal fusion models. Furthermore, we summarize representative applications of LMMs in agriculture, including meteorological forecasting services, extreme weather early warning and disaster mitigation, and decision support systems for precision agriculture. We also examine commonly used datasets and evaluation metrics in the field. This is followed by a discussion of current challenges, particularly in areas of generalization capability, uncertainty quantification, and physical consistency. Finally, we outline key research directions for future exploration in this rapidly evolving field. This work aims to provide a systematic reference for advancing both the research and application of LMMs in agricultural contexts.

, authors=Ruxue Xing, Jianjun Zhu, Ziting Nie, Qinghao Guo, Shan Li, Yinxue Shi, Yaojun Wang*, authorsList=Ruxue Xing, Jianjun Zhu, Ziting Nie, Qinghao Guo, Shan Li, Yinxue Shi, Yaojun Wang, authorCompany=null, correspAuthors=Yaojun Wang, authorNote=

born in 1998. Master candidate. Her main research interests include artificial intelligence and meteorological forecasting

born in 1997. PhD candidate. His main research interests include deep learning and time series prediction

born in 2003. Master candidate. Her main research interest includes deep learning

born in 2003. Master candidate. His main research interest includes agricultural production warning

born in 1995. PhD, lecturer. Her main research interests include statistical pattern recognition and agricultural large model

born in 1974. PhD, associate professor. Her main research interests include computer vision, smart agriculture, and smart healthcare

born in 1985. PhD, associate professor. His main research interests include knowledge graphs, artificial intelligence, and large models

, correspAuthorsNote=null, copyrightStatement=Copyright © 2026 Journal of Computer Research and Development. All rights reserved., copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, fund=null), CN=ArticleExt(id=1329420063792280387, articleId=1329420060856267557, tenantId=1146029695717560320, journalId=1324759914146529478, language=CN, title=气象大模型在农业应用中的综述, columnId=1329420000856748099, journalTitle=计算机研究与发展, columnName=AI4Science:科学研究的新范式专题, runingTitle=null, highlight=null, articleAbstract=

近年来,深度学习在气象预测与气候建模中的应用取得了显著进展,并在农业气象服务中展现出重要价值。尽管已有综述文章对相关方法进行了系统梳理,但多聚焦于模型架构、任务类型或应用场景,尚缺乏从农业视角出发的系统总结。聚焦气象大模型(large meteorological models,LMMs)的技术演进与应用实践,围绕模型结构建立分类体系,系统回顾传统数值方法与深度学习架构的发展脉络。根据建模范式的不同,将主流方法归纳为六大类:经典深度神经网络、图神经网络、Transformer 架构、状态空间模型、生成式模型以及物理引导与多模态融合模型,并进一步梳理了典型模型在农业场景中的应用,包括农业气象预报、极端天气预警、防灾减灾及精准农业决策支持等关键任务。此外,还汇总了常用数据集与评估指标,分析当前气象大模型在泛化能力、不确定性建模与物理一致性方面的挑战,并展望了未来研究的关键方向,旨在为气象大模型在农业场景中的深入研究与应用提供系统性参考。

, authors=邢茹雪, 朱建军, 聂紫婷, 郭庆昊, 李珊, 史银雪, 王耀君*, authorsList=邢茹雪, 朱建军, 聂紫婷, 郭庆昊, 李珊, 史银雪, 王耀君, authorCompany=null, correspAuthors=王耀君, authorNote=

作者贡献声明

邢茹雪完成论文框架设计、文献调研与全文撰写;朱建军参与算法分类体系的梳理与章节结构的设计;聂紫婷、郭庆昊参与文献整理与图表制作;李珊、史银雪参与论文修改;王耀君提出研究选题与论文思路,并对论文进行总体指导与修改完善。邢茹雪与朱建军是共同第一作者。

1998年生。硕士研究生。主要研究方向为人工智能、气象预测

1997年生。博士研究生。主要研究方向为深度学习、时序预测

2003年生。硕士研究生。主要研究方向为深度学习

2003年生。硕士研究生。主要研究方向为农业生产预警

1995年生。博士,讲师。主要研究方向为统计模式识别、农业大模型

1974年生。博士,副教授。主要研究方向为计算机视觉、智慧农业、智能医疗

1985年生。博士,副教授。主要研究方向为知识图谱、人工智能、大模型

, correspAuthorsNote=
王耀君()
, copyrightStatement=版权所有 © 2026 计算机研究与发展编辑部, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=/gdKKrr8ODlBscjQkJHe3A==, magXml=aktTwynwmdqulDBXG+GbjA==, pdfUrl=null, pdf=1W+IJrPWNQxM4KMT3iXhgw==, pdfFileSize=4734993, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=KBe2V/xmHlBMYFkI9CNzww==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=nowJivQe1LXD6pO9NkwN7Q==, mapNumber=null, fund=null)}, authors=[Author(id=1329420064127824712, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=xingruxue@cau.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420064194933578, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064127824712, language=EN, stringName=Ruxue Xing, firstName=Ruxue, middleName=null, lastName=Xing, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, bio={"content":"

born in 1998. Master candidate. Her main research interests include artificial intelligence and meteorological forecasting

"}, bioImg=null, bioContent=

born in 1998. Master candidate. Her main research interests include artificial intelligence and meteorological forecasting

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420064253653835, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064127824712, language=CN, stringName=邢茹雪, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=中国农业大学信息与电气工程学院 北京 100083, bio={"img":"E5wLCWSlGIC6yLm2TXG8KQ==","content":"

1998年生。硕士研究生。主要研究方向为人工智能、气象预测

"}, bioImg=E5wLCWSlGIC6yLm2TXG8KQ==, bioContent=

1998年生。硕士研究生。主要研究方向为人工智能、气象预测

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420064018772804, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, xref=null, ext=[AuthorCompanyExt(id=1329420064027161413, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083), AuthorCompanyExt(id=1329420064035550022, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国农业大学信息与电气工程学院 北京 100083)])]), Author(id=1329420064312374093, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420064392065871, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064312374093, language=EN, stringName=Jianjun Zhu, firstName=Jianjun, middleName=null, lastName=Zhu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, bio={"content":"

born in 1997. PhD candidate. His main research interests include deep learning and time series prediction

"}, bioImg=null, bioContent=

born in 1997. PhD candidate. His main research interests include deep learning and time series prediction

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420064459174736, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064312374093, language=CN, stringName=朱建军, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=中国农业大学信息与电气工程学院 北京 100083, bio={"img":"IPJmrTFIcR2eZD4JSfaRRw==","content":"

1997年生。博士研究生。主要研究方向为深度学习、时序预测

"}, bioImg=IPJmrTFIcR2eZD4JSfaRRw==, bioContent=

1997年生。博士研究生。主要研究方向为深度学习、时序预测

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420064018772804, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, xref=null, ext=[AuthorCompanyExt(id=1329420064027161413, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083), AuthorCompanyExt(id=1329420064035550022, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国农业大学信息与电气工程学院 北京 100083)])]), Author(id=1329420064517894994, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420064593392468, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064517894994, language=EN, stringName=Ziting Nie, firstName=Ziting, middleName=null, lastName=Nie, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, bio={"content":"

born in 2003. Master candidate. Her main research interest includes deep learning

"}, bioImg=null, bioContent=

born in 2003. Master candidate. Her main research interest includes deep learning

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420064656307029, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064517894994, language=CN, stringName=聂紫婷, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=中国农业大学信息与电气工程学院 北京 100083, bio={"img":"kaWFNaJLNpW1IiPxECoDUw==","content":"

2003年生。硕士研究生。主要研究方向为深度学习

"}, bioImg=kaWFNaJLNpW1IiPxECoDUw==, bioContent=

2003年生。硕士研究生。主要研究方向为深度学习

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420064018772804, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, xref=null, ext=[AuthorCompanyExt(id=1329420064027161413, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083), AuthorCompanyExt(id=1329420064035550022, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国农业大学信息与电气工程学院 北京 100083)])]), Author(id=1329420064715027287, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, orderNo=3, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420064782136153, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064715027287, language=EN, stringName=Qinghao Guo, firstName=Qinghao, middleName=null, lastName=Guo, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, bio={"content":"

born in 2003. Master candidate. His main research interest includes agricultural production warning

"}, bioImg=null, bioContent=

born in 2003. Master candidate. His main research interest includes agricultural production warning

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420064866022234, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064715027287, language=CN, stringName=郭庆昊, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=中国农业大学信息与电气工程学院 北京 100083, bio={"img":"pKXq7Fgf5dfO22y0hRqFGA==","content":"

2003年生。硕士研究生。主要研究方向为农业生产预警

"}, bioImg=pKXq7Fgf5dfO22y0hRqFGA==, bioContent=

2003年生。硕士研究生。主要研究方向为农业生产预警

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420064018772804, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, xref=null, ext=[AuthorCompanyExt(id=1329420064027161413, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083), AuthorCompanyExt(id=1329420064035550022, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国农业大学信息与电气工程学院 北京 100083)])]), Author(id=1329420064937325404, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, orderNo=4, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420065000239966, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064937325404, language=EN, stringName=Shan Li, firstName=Shan, middleName=null, lastName=Li, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, bio={"content":"

born in 1995. PhD, lecturer. Her main research interests include statistical pattern recognition and agricultural large model

"}, bioImg=null, bioContent=

born in 1995. PhD, lecturer. Her main research interests include statistical pattern recognition and agricultural large model

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420065054765919, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064937325404, language=CN, stringName=李珊, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=中国农业大学信息与电气工程学院 北京 100083, bio={"img":"BjCOJrYaP8EU0fZ0o9AjsA==","content":"

1995年生。博士,讲师。主要研究方向为统计模式识别、农业大模型

"}, bioImg=BjCOJrYaP8EU0fZ0o9AjsA==, bioContent=

1995年生。博士,讲师。主要研究方向为统计模式识别、农业大模型

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420064018772804, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, xref=null, ext=[AuthorCompanyExt(id=1329420064027161413, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083), AuthorCompanyExt(id=1329420064035550022, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国农业大学信息与电气工程学院 北京 100083)])]), Author(id=1329420065134457697, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, orderNo=5, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420065205760867, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420065134457697, language=EN, stringName=Yinxue Shi, firstName=Yinxue, middleName=null, lastName=Shi, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, bio={"content":"

born in 1974. PhD, associate professor. Her main research interests include computer vision, smart agriculture, and smart healthcare

"}, bioImg=null, bioContent=

born in 1974. PhD, associate professor. Her main research interests include computer vision, smart agriculture, and smart healthcare

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420065264481124, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420065134457697, language=CN, stringName=史银雪, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=中国农业大学信息与电气工程学院 北京 100083, bio={"img":"AcS7IUc3PKvkCUdvWzEfgg==","content":"

1974年生。博士,副教授。主要研究方向为计算机视觉、智慧农业、智能医疗

"}, bioImg=AcS7IUc3PKvkCUdvWzEfgg==, bioContent=

1974年生。博士,副教授。主要研究方向为计算机视觉、智慧农业、智能医疗

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420064018772804, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, xref=null, ext=[AuthorCompanyExt(id=1329420064027161413, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083), AuthorCompanyExt(id=1329420064035550022, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国农业大学信息与电气工程学院 北京 100083)])]), Author(id=1329420065360950118, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, orderNo=6, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=wangyaojun@cau.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1329420065428058984, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420065360950118, language=EN, stringName=Yaojun Wang, firstName=Yaojun, middleName=null, lastName=Wang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, bio={"content":"

born in 1985. PhD, associate professor. His main research interests include knowledge graphs, artificial intelligence, and large models

"}, bioImg=null, bioContent=

born in 1985. PhD, associate professor. His main research interests include knowledge graphs, artificial intelligence, and large models

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420065503556457, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420065360950118, language=CN, stringName=王耀君, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=中国农业大学信息与电气工程学院 北京 100083, bio={"img":"cDSG5wGKQETezA+E2GuB2w==","content":"

1985年生。博士,副教授。主要研究方向为知识图谱、人工智能、大模型

"}, bioImg=cDSG5wGKQETezA+E2GuB2w==, bioContent=

1985年生。博士,副教授。主要研究方向为知识图谱、人工智能、大模型

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420064018772804, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, xref=null, ext=[AuthorCompanyExt(id=1329420064027161413, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083), AuthorCompanyExt(id=1329420064035550022, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国农业大学信息与电气工程学院 北京 100083)])])], keywords=[Keyword(id=1329420065620996970, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=EN, orderNo=1, keyword=large meteorological models (LMMs)), Keyword(id=1329420065734243179, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=EN, orderNo=2, keyword=deep learning), Keyword(id=1329420065792963436, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=EN, orderNo=3, keyword=time series modeling), Keyword(id=1329420065864266605, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=EN, orderNo=4, keyword=weather forecasting), Keyword(id=1329420065931375470, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=EN, orderNo=5, keyword=agricultural applications), Keyword(id=1329420066002678639, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=CN, orderNo=1, keyword=气象大模型), Keyword(id=1329420066115924848, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=CN, orderNo=2, keyword=深度学习), Keyword(id=1329420066187228017, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=CN, orderNo=3, keyword=时间序列建模), Keyword(id=1329420066250142578, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=CN, orderNo=4, keyword=天气预报), Keyword(id=1329420066325640051, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=CN, orderNo=5, keyword=农业应用)], refs=[Reference(id=1329420069588808592, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=1, rfOrder=0, authorNames=null, journalName=null, refType=null, unstructuredReference=Hultgren A, Carleton T, Delgado M, et al. Impacts of climate change on global agriculture accounting for adaptation[J]. Nature, 2025, 642(8068): 644−652, articleTitle=null, refAbstract=null), Reference(id=1329420069647528849, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=2, rfOrder=1, authorNames=null, journalName=null, refType=null, unstructuredReference=Lipper L, Thornton P, Campbell B M, et al. Climate-smart agriculture for food security[J]. Nature Climate Change, 2014, 4(12): 1068−1072, articleTitle=null, refAbstract=null), Reference(id=1329420069723026322, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=3, rfOrder=2, authorNames=null, journalName=null, refType=null, unstructuredReference=Yang Yi, Tilman D, Jin Zhenong, et al. Climate change exacerbates the environmental impacts of agriculture[J]. Science, 2024, 385(6713): eadn3747, articleTitle=null, refAbstract=null), Reference(id=1329420069823689619, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=4, rfOrder=3, authorNames=null, journalName=null, refType=null, unstructuredReference=Parolini G. Weather, climate, and agriculture: Historical contributions and perspectives from agricultural meteorology[J]. WIREs Climate Change, 2022, 13(3): e766, articleTitle=null, refAbstract=null), Reference(id=1329420069920158612, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=5, rfOrder=4, authorNames=null, journalName=null, refType=null, unstructuredReference=Schepen A, Everingham Y, Wang Q J. An improved workflow for calibration and downscaling of GCM climate forecasts for agricultural applications: A case study on prediction of sugarcane yield in Australia[J]. Agricultural and Forest Meteorology, 2020, 291: 107991, articleTitle=null, refAbstract=null), Reference(id=1329420069983073173, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=6, rfOrder=5, authorNames=null, journalName=null, refType=null, unstructuredReference=Jha P K, Athanasiadis P, Gualdi S, et al. Using daily data from seasonal forecasts in dynamic crop models for yield prediction: A case study for rice in Nepal’s Terai[J]. Agricultural and Forest Meteorology, 2019, 265: 349−358, articleTitle=null, refAbstract=null), Reference(id=1329420070054376342, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=7, rfOrder=6, authorNames=null, journalName=null, refType=null, unstructuredReference=Materia S, García L P, Van Straaten C, et al. Artificial intelligence for climate prediction of extremes: State of the art, challenges, and future perspectives[J]. WIREs Climate Change, 2024, 15(6): e914, articleTitle=null, refAbstract=null), Reference(id=1329420070129873815, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=8, rfOrder=7, authorNames=null, journalName=null, refType=null, unstructuredReference=Yu Siwei, Ma Jianwei. Deep learning for geophysics: Current and future trends[J]. Reviews of Geophysics, 2021, 59(3): e2021RG000742, articleTitle=null, refAbstract=null), Reference(id=1329420070217954200, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=9, rfOrder=8, authorNames=null, journalName=null, refType=null, unstructuredReference=Bi Kaifeng, Xie Lingxi, Zhang Hengheng, et al. Accurate medium-range global weather forecasting with 3D neural networks[J]. Nature, 2023, 619(7970): 533−538, articleTitle=null, refAbstract=null), Reference(id=1329420070310228889, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=10, rfOrder=9, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen Lei, Zhong Xiaohui, Zhang Feng, et al. FuXi: A cascade machine learning forecasting system for 15-day global weather forecast[J]. npj Climate and Atmospheric Science, 2023, 6(1): 190, articleTitle=null, refAbstract=null), Reference(id=1329420070377337754, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=11, rfOrder=10, authorNames=null, journalName=null, refType=null, unstructuredReference=Bodnar C, Bruinsma W P, Lucic A, et al. A foundation model for the Earth system[J]. Nature, 2025, 641(8065): 1180−1187, articleTitle=null, refAbstract=null), Reference(id=1329420070440252315, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=12, rfOrder=11, authorNames=null, journalName=null, refType=null, unstructuredReference=Kurth T, Subramanian S, Harrington P, et al. FourCastNet: Accelerating global high-resolution weather forecasting using adaptive Fourier neural operators[C]//Proc of the Platform for Advanced Scientific Computing Conf. New York: ACM, 2023: 1−11, articleTitle=null, refAbstract=null), Reference(id=1329420070515749788, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=13, rfOrder=12, authorNames=null, journalName=null, refType=null, unstructuredReference=Price I, Sanchez-Gonzalez A, Alet F, et al. Probabilistic weather forecasting with machine learning[J]. Nature, 2025, 637(8044): 84−90, articleTitle=null, refAbstract=null), Reference(id=1329420070578664349, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=14, rfOrder=13, authorNames=null, journalName=null, refType=null, unstructuredReference=Niu Dan, Niu Daben, Li Yinghao, et al. CastDiffuser: Cascaded latent diffusion framework for high-resolution precipitation nowcasting via multi source fusion[J]. Neurocomputing, 2026, 670: 132435, articleTitle=null, refAbstract=null), Reference(id=1329420070633190302, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=15, rfOrder=14, authorNames=null, journalName=null, refType=null, unstructuredReference=Sun Xiuyu, Zhong Xiaohui, Xu Xiaoze, et al. A data-to-forecast machine learning system for global weather[J]. Nature Communications, 2025, 16(1): 6658, articleTitle=null, refAbstract=null), Reference(id=1329420070691910559, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=16, rfOrder=15, authorNames=null, journalName=null, refType=null, unstructuredReference=Boukabara S A, Krasnopolsky V, Stewart J Q, et al. Realizing the benefits of AI across the numerical weather prediction value chain[J]. Bulletin of the American Meteorological Society, 2020, 101(1): 29−33, articleTitle=null, refAbstract=null), Reference(id=1329420070759019424, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=17, rfOrder=16, authorNames=null, journalName=null, refType=null, unstructuredReference=Mass C. The uncoordinated giant II: Why US operational numerical weather prediction is still lagging and how to fix it[J]. Bulletin of the American Meteorological Society, 2023, 104(4): E851−E871, articleTitle=null, refAbstract=null), Reference(id=1329420070826128289, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=18, rfOrder=17, authorNames=null, journalName=null, refType=null, unstructuredReference=Steppeler J, Gan P, Fang F, et al. Numerical weather prediction in the time of Exaflop computers[J]. Bulletin of the American Meteorological Society, 2025, 106(6): E1087−E1097, articleTitle=null, refAbstract=null), Reference(id=1329420070910014370, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=19, rfOrder=18, authorNames=null, journalName=null, refType=null, unstructuredReference=Baran Á, Baran S. A two-step machine-learning approach to statistical post-processing of weather forecasts for power generation[J]. Quarterly Journal of the Royal Meteorological Society, 2024, 150(759): 1029−1047, articleTitle=null, refAbstract=null), Reference(id=1329420071014871971, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=20, rfOrder=19, authorNames=null, journalName=null, refType=null, unstructuredReference=Sacco M A, Ruiz J J, Pulido M, et al. Evaluation of machine learning techniques for forecast uncertainty quantification[J]. Quarterly Journal of the Royal Meteorological Society, 2022, 148(749): 3470−3490, articleTitle=null, refAbstract=null), Reference(id=1329420071081980836, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=21, rfOrder=20, authorNames=null, journalName=null, refType=null, unstructuredReference=An-Vo D A, Radanielson A M, Mushtaq S, et al. A framework for assessing the value of seasonal climate forecasting in key agricultural decisions[J]. Climate Services, 2021, 22: 100234, articleTitle=null, refAbstract=null), Reference(id=1329420071140701093, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=22, rfOrder=21, authorNames=null, journalName=null, refType=null, unstructuredReference=Wanthanaporn U, Supit I, Chaowiwat W, et al. Skill of rice yields forecasting over Mainland Southeast Asia using the ECMWF SEAS5 ensemble prediction system and the WOFOST crop model[J]. Agricultural and Forest Meteorology, 2024, 351: 110001, articleTitle=null, refAbstract=null), Reference(id=1329420071199421350, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=23, rfOrder=22, authorNames=null, journalName=null, refType=null, unstructuredReference=Naveau P, Hannart A, Ribes A. Statistical methods for extreme event attribution in climate science[J]. Annual Review of Statistics and Its Application, 2020, 7(1): 89−110, articleTitle=null, refAbstract=null), Reference(id=1329420071283307431, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=24, rfOrder=23, authorNames=null, journalName=null, refType=null, unstructuredReference=Hersbach H, Bell B, Berrisford P, et al. The ERA5 global reanalysis[J]. Quarterly Journal of the Royal Meteorological Society, 2020, 146(730): 1999−2049, articleTitle=null, refAbstract=null), Reference(id=1329420071350416296, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=25, rfOrder=24, authorNames=null, journalName=null, refType=null, unstructuredReference=Lam R, Sanchez-Gonzalez A, Willson M, et al. Learning skillful medium-range global weather forecasting[J]. Science, 2023, 382(6677): 1416−1421, articleTitle=null, refAbstract=null), Reference(id=1329420071509799849, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=26, rfOrder=25, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen Kang, Han Tao, Ling Fenghua, et al. The operational medium-range deterministic weather forecasting can be extended beyond a 10-day lead time[J]. Communications Earth & Environment, 2025, 6(1): 518, articleTitle=null, refAbstract=null), Reference(id=1329420071593685930, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=27, rfOrder=26, authorNames=null, journalName=null, refType=null, unstructuredReference=Allen A, Markou S, Tebbutt W, et al. End-to-end data-driven weather prediction[J]. Nature, 2025, 641(8065): 1172−1179, articleTitle=null, refAbstract=null), Reference(id=1329420071669183403, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=28, rfOrder=27, authorNames=null, journalName=null, refType=null, unstructuredReference=LeCun Y, Bottou L, Bengio Y, et al. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 1998, 86(11): 2278−2324, articleTitle=null, refAbstract=null), Reference(id=1329420071732097964, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=29, rfOrder=28, authorNames=null, journalName=null, refType=null, unstructuredReference=Shi Xingjian, Chen Zhourong, Wang Hao, et al. Convolutional LSTM network: A machine learning approach for precipitation nowcasting[C]//Proc of the 28th Int Conf on Neural Information Processing Systems. Red Hook, NY: Curran Associates Inc. , 2015: 802−810, articleTitle=null, refAbstract=null), Reference(id=1329420071807595437, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=30, rfOrder=29, authorNames=null, journalName=null, refType=null, unstructuredReference=周立, 吴洋洋, 苗晓烨. 基于深度学习的不完整时序数据补全方法综述[J]. 计算机研究与发展, 2025, 62(9): 2233−2258, articleTitle=null, refAbstract=null), Reference(id=1329420071887287214, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=30, rfOrder=30, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhou Li, Wu Yangyang, Miao Xiaoye. Survey of incomplete time series imputation methods based on deep learning[J]. Journal of Computer Research and Development, 2025, 62(9): 2233−2258 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329420071946007471, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=31, rfOrder=31, authorNames=null, journalName=null, refType=null, unstructuredReference=Scarselli F, Gori M, Tsoi A C, et al. The graph neural network model[J]. IEEE Transactions on Neural Networks, 2009, 20(1): 61−80, articleTitle=null, refAbstract=null), Reference(id=1329420072008922032, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=32, rfOrder=32, authorNames=null, journalName=null, refType=null, unstructuredReference=Alexe M, Boucher E, Lean P, et al. GraphDOP: Towards skilful data-driven medium-range weather forecasts learnt and initialised directly from observations[J]. arXiv preprint, arXiv: 2412.15687, 2024, articleTitle=null, refAbstract=null), Reference(id=1329420072067642289, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=33, rfOrder=33, authorNames=null, journalName=null, refType=null, unstructuredReference=Lang S, Alexe M, Chantry M, et al. AIFS - ECMWF’s data-driven forecasting system[J]. arXiv preprint, arXiv: 2406.01465, 2024, articleTitle=null, refAbstract=null), Reference(id=1329420072130556850, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=34, rfOrder=34, authorNames=null, journalName=null, refType=null, unstructuredReference=Nguyen T, Brandstetter J, Kapoor A, et al. ClimaX: A foundation model for weather and climate[C]//Proc of the 40th Int Conf on Machine Learning. San Diego, CA: PMLR, 2023: 25904−25938, articleTitle=null, refAbstract=null), Reference(id=1329420072214442931, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=35, rfOrder=35, authorNames=null, journalName=null, refType=null, unstructuredReference=Hu Yuan, Chen Lei, Wang Zhibin, et al. SwinVRNN: A data-driven ensemble forecasting model via learned distribution perturbation[J]. Journal of Advances in Modeling Earth Systems, 2023, 15(2): e2022MS003211, articleTitle=null, refAbstract=null), Reference(id=1329420072273163188, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=36, rfOrder=36, authorNames=null, journalName=null, refType=null, unstructuredReference=Andrychowicz M, Espeholt L, Li D, et al. Deep learning for day forecasts from sparse observations[J]. arXiv preprint, arXiv: 2306.06079, 2023, articleTitle=null, refAbstract=null), Reference(id=1329420072361243573, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=37, rfOrder=37, authorNames=null, journalName=null, refType=null, unstructuredReference=Xu Siyi, Zhang Yize, Chen Junping, et al. Short-to medium-term weather forecast skill of the AI-based Pangu-Weather model using automatic weather stations in China[J]. Remote Sensing, 2025, 17(2): 191, articleTitle=null, refAbstract=null), Reference(id=1329420072432546742, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=38, rfOrder=38, authorNames=null, journalName=null, refType=null, unstructuredReference=Yang S, Liu Z, Shi Z, et al. WSSM: Geographic enhanced hierarchical state-space model for global station weather forecast[J]. arXiv preprint, arXiv: 2501.11238, 2025, articleTitle=null, refAbstract=null), Reference(id=1329420072524821431, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=39, rfOrder=39, authorNames=null, journalName=null, refType=null, unstructuredReference=Qin H, Chen Y, Jiang Q, et al. MetMamba: Regional weather forecasting with spatial-temporal Mamba model[J]. arXiv preprint, arXiv: 2408.06400, 2024, articleTitle=null, refAbstract=null), Reference(id=1329420072583541688, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=40, rfOrder=40, authorNames=null, journalName=null, refType=null, unstructuredReference=Verma Y, Heinonen M, Garg V. ClimODE: Climate and weather forecasting with physics-informed neural ODEs[C/OL]//Proc of the 12th Int Conf on Learning Representations. 2024[2026-06-18]. https://openreview.net/forum?id=xuY33XhEGR, articleTitle=null, refAbstract=null), Reference(id=1329420072642261945, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=41, rfOrder=41, authorNames=null, journalName=null, refType=null, unstructuredReference=Liu P, Zhou T, Sun L, et al. Mitigating time discretization challenges with WeatherODE: A sandwich physics-driven neural ODE for weather forecasting[J]. arXiv preprint, arXiv: 2410.06560, 2024, articleTitle=null, refAbstract=null), Reference(id=1329420072751313850, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=42, rfOrder=42, authorNames=null, journalName=null, refType=null, unstructuredReference=Gu A, Dao T. Mamba: Linear-time sequence modeling with selective state spaces[C/OL]//Proc of the 1st Conf on Language Modeling. 2024[2026-04-30]. https://openreview.net/forum?id=tEYskw1VY2, articleTitle=null, refAbstract=null), Reference(id=1329420072818422715, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=43, rfOrder=43, authorNames=null, journalName=null, refType=null, unstructuredReference=Ravuri S, Lenc K, Willson M, et al. Skilful precipitation nowcasting using deep generative models of radar[J]. Nature, 2021, 597(7878): 672−677, articleTitle=null, refAbstract=null), Reference(id=1329420072902308796, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=44, rfOrder=44, authorNames=null, journalName=null, refType=null, unstructuredReference=Sha Yingkai, Sobash R A, Gagne D J. Generative ensemble deep learning severe weather prediction from a deterministic convection-allowing model[J]. Artificial Intelligence for the Earth Systems, 2024, 3(2): e230094, articleTitle=null, refAbstract=null), Reference(id=1329420072973611965, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=45, rfOrder=45, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Yuchen, Long Mingsheng, Chen Kaiyuan, et al. Skilful nowcasting of extreme precipitation with NowcastNet[J]. Nature, 2023, 619(7970): 526−532, articleTitle=null, refAbstract=null), Reference(id=1329420073040720830, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=46, rfOrder=46, authorNames=null, journalName=null, refType=null, unstructuredReference=Gong Junchao, Bai Lei, Ye Peng, et al. CasCast: Skillful high-resolution precipitation nowcasting via cascaded modelling[C]//Proc of the 41st Int Conf on Machine Learning. San Diego, CA: PMLR, 2024: 15809−15822, articleTitle=null, refAbstract=null), Reference(id=1329420073103635391, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=47, rfOrder=47, authorNames=null, journalName=null, refType=null, unstructuredReference=Kochkov D, Yuval J, Langmore I, et al. Neural general circulation models for weather and climate[J]. Nature, 2024, 632(8027): 1060−1066, articleTitle=null, refAbstract=null), Reference(id=1329420073162355648, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=48, rfOrder=48, authorNames=null, journalName=null, refType=null, unstructuredReference=Das P, Posch A, Barber N, et al. Hybrid physics-AI outperforms numerical weather prediction for extreme precipitation nowcasting[J]. npj Climate and Atmospheric Science, 2024, 7(1): 282, articleTitle=null, refAbstract=null), Reference(id=1329420073225270209, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=49, rfOrder=49, authorNames=null, journalName=null, refType=null, unstructuredReference=Boussioux L, Zeng C, Guénais T, et al. Hurricane forecasting: A novel multimodal machine learning framework[J]. Weather and Forecasting, 2022, 37(6): 817−831, articleTitle=null, refAbstract=null), Reference(id=1329420073288184770, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=50, rfOrder=50, authorNames=null, journalName=null, refType=null, unstructuredReference=Ding Junsheng, Chen Wu, Chen Junping, et al. Real-time high-resolution global PWV retrieval based on weather forecast foundation models and cross-validation with radiosonde, GNSS, and ERA5[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2025, 18: 15972−15981, articleTitle=null, refAbstract=null), Reference(id=1329420073351099331, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=51, rfOrder=51, authorNames=null, journalName=null, refType=null, unstructuredReference=Park H, Park S, Kang D, et al. A super-resolution framework for downscaling machine learning weather prediction toward 1-km air temperature[J]. npj Climate and Atmospheric Science, 2026, 9(1): 56, articleTitle=null, refAbstract=null), Reference(id=1329420073418208196, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=52, rfOrder=52, authorNames=null, journalName=null, refType=null, unstructuredReference=Tarraf B, Brun F, Raynaud L, et al. Assessing the impact of weather forecast uncertainties in crop water stress model predictions[J]. Agricultural and Forest Meteorology, 2024, 349: 109934, articleTitle=null, refAbstract=null), Reference(id=1329420073481122757, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=53, rfOrder=53, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen Lei, Zhong Xiaohui, Li Hao, et al. A machine learning model that outperforms conventional global subseasonal forecast models[J]. Nature Communications, 2024, 15(1): 6425, articleTitle=null, refAbstract=null), Reference(id=1329420073552425926, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=54, rfOrder=54, authorNames=null, journalName=null, refType=null, unstructuredReference=Garcia D, Rolim J, Cameira M R, et al. Assessing seasonal forecast performance to predict crop irrigation requirements to support water management decision-making in the Mediterranean region[J]. Agricultural Water Management, 2025, 313: 109467, articleTitle=null, refAbstract=null), Reference(id=1329420073615340487, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=55, rfOrder=55, authorNames=null, journalName=null, refType=null, unstructuredReference=Luo Qunying, Wen Li, Cowan T, et al. Seasonal climate forecast - an important tool in managing the risk of extreme weather events in Australia’s wheat industry[J]. Agricultural and Forest Meteorology, 2024, 351: 110005, articleTitle=null, refAbstract=null), Reference(id=1329420073686643656, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=56, rfOrder=56, authorNames=null, journalName=null, refType=null, unstructuredReference=Garza A, Mergenthaler-Canseco M. TimeGPT−1[J]. arXiv preprint, arXiv: 2310.03589, 2023, articleTitle=null, refAbstract=null), Reference(id=1329420073753752522, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=57, rfOrder=57, authorNames=null, journalName=null, refType=null, unstructuredReference=Shin J Y, Min B, Kim K R. High-resolution wind speed forecast system coupling numerical weather prediction and machine learning for agricultural studies—a case study from South Korea[J]. International Journal of Biometeorology, 2022, 66(7): 1429−1443, articleTitle=null, refAbstract=null), Reference(id=1329420073858610123, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=58, rfOrder=58, authorNames=null, journalName=null, refType=null, unstructuredReference=Su Qiong, Ale S, Himanshu S K, et al. Calibration and bias correction of seasonal weather forecasts from the North American multi-model ensemble: Potential applications for regional crop modelling and irrigation management[J]. The Journal of Agricultural Science, 2025, 163(2): 145−159, articleTitle=null, refAbstract=null), Reference(id=1329420074009605068, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=59, rfOrder=59, authorNames=null, journalName=null, refType=null, unstructuredReference=Garcia D, Silva N, Rolim J, et al. Prediction of crops cycle with seasonal forecasts to support decision-making[J]. Agronomy, 2025, 15(6): 1291, articleTitle=null, refAbstract=null), Reference(id=1329420074101879757, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=60, rfOrder=60, authorNames=null, journalName=null, refType=null, unstructuredReference=Skok G, Lledó L. Spatial verification of global precipitation forecasts[J]. Quarterly Journal of the Royal Meteorological Society, 2025, 151(771): e5006, articleTitle=null, refAbstract=null), Reference(id=1329420074173182926, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=61, rfOrder=61, authorNames=null, journalName=null, refType=null, unstructuredReference=Rasp S, Dueben P D, Scher S, et al. WeatherBench: A benchmark data set for data-driven weather forecasting[J]. Journal of Advances in Modeling Earth Systems, 2020, 12(11): e2020MS002203, articleTitle=null, refAbstract=null), Reference(id=1329420074252874703, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=62, rfOrder=62, authorNames=null, journalName=null, refType=null, unstructuredReference=Rasp S, Hoyer S, Merose A, et al. WeatherBench 2: A benchmark for the next generation of data-driven global weather models[J]. Journal of Advances in Modeling Earth Systems, 2024, 16(6): e2023MS004019, articleTitle=null, refAbstract=null), Reference(id=1329420074336760784, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=63, rfOrder=63, authorNames=null, journalName=null, refType=null, unstructuredReference=Nathaniel J, Qu Yongquan, Nguyen T, et al. ChaosBench: A multi-channel, physics-based benchmark for subseasonal-to-seasonal climate prediction[C]//Proc of the 38th Int Conf on Neural Information Processing Systems. Red Hook, NY: Curran Associates Inc., 2024: 43715−43729, articleTitle=null, refAbstract=null), Reference(id=1329420074408063953, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=64, rfOrder=64, authorNames=null, journalName=null, refType=null, unstructuredReference=Dosovitskiy A, Beyer L, Kolesnikov A, et al. An image is worth 16x16 words: Transformers for image recognition at scale[C/OL]//Proc of the 9th Int Conf on Learning Representations. 2021[2026-04-30]. https://openreview.net/forum?id=YicbFdNTTy, articleTitle=null, refAbstract=null), Reference(id=1329420074479367122, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=65, rfOrder=65, authorNames=null, journalName=null, refType=null, unstructuredReference=史宏志, 赵健, 赵雅倩, 等. 大模型时代的混合专家系统优化综述[J]. 计算机研究与发展, 2025, 62(5): 1164−1189, articleTitle=null, refAbstract=null), Reference(id=1329420074538087379, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=65, rfOrder=66, authorNames=null, journalName=null, refType=null, unstructuredReference=Shi Hongzhi, Zhao Jian, Zhao Yaqian, et al. Survey on system optimization for mixture of experts in the era of large models[J]. Journal of Computer Research and Development, 2025, 62(5): 1164−1189 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329420074605196244, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=66, rfOrder=67, authorNames=null, journalName=null, refType=null, unstructuredReference=李文斌, 熊亚锟, 范祇辰, 等. 持续学习的研究进展与趋势[J]. 计算机研究与发展, 2024, 61(6): 1476−1496, articleTitle=null, refAbstract=null), Reference(id=1329420074663916501, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=66, rfOrder=68, authorNames=null, journalName=null, refType=null, unstructuredReference=Li Wenbin, Xiong Yakun, Fan Zhichen, et al. Advances and trends of continual learning[J]. Journal of Computer Research and Development, 2024, 61(6): 1476−1496 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329420074731025366, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=67, rfOrder=69, authorNames=null, journalName=null, refType=null, unstructuredReference=王睿, 张留洋, 高志涌, 等. 面向边缘智能的大模型研究进展[J]. 计算机研究与发展, 2025, 62(10): 2565−2582, articleTitle=null, refAbstract=null), Reference(id=1329420074798134231, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=67, rfOrder=70, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Rui, Zhang Liuyang, Gao Zhiyong, et al. Research progress on large models for edge intelligence[J]. Journal of Computer Research and Development, 2025, 62(10): 2565−2582 (in Chinese), articleTitle=null, refAbstract=null)], funds=null, companyList=[AuthorCompany(id=1329420064018772804, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, xref=null, ext=[AuthorCompanyExt(id=1329420064027161413, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083), AuthorCompanyExt(id=1329420064035550022, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国农业大学信息与电气工程学院 北京 100083)])], figs=[ArticleFig(id=1329420066464052084, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=EN, label=Figure 1, caption=Four core challenges of large meteorological models in agricultural applications, figureFileSmall=ImQpC9FO33odjDADib2C7Q==, figureFileBig=KBe2V/xmHlBMYFkI9CNzww==, tableContent=null), ArticleFig(id=1329420066543743861, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=CN, label=图1, caption=气象大模型在农业应用中的4类核心挑战, figureFileSmall=ImQpC9FO33odjDADib2C7Q==, figureFileBig=KBe2V/xmHlBMYFkI9CNzww==, tableContent=null), ArticleFig(id=1329420066828956534, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=EN, label=Figure 2, caption=Formal architecture and universal evaluation framework for large meteorological modeling tasks, figureFileSmall=+A2P62voMrXbjFgSEd1rYg==, figureFileBig=pMBHhW29T6OC+4RWe8SpBw==, tableContent=null), ArticleFig(id=1329420067034477431, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=CN, label=图2, caption=气象大模型建模任务的形式化架构与通用评估框架, figureFileSmall=+A2P62voMrXbjFgSEd1rYg==, figureFileBig=pMBHhW29T6OC+4RWe8SpBw==, tableContent=null), ArticleFig(id=1329420067244192632, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=EN, label=Figure 3, caption=Global distribution of leading research institutions for large meteorological models, figureFileSmall=YzRyzlgF0K0/wzY/LzW2MA==, figureFileBig=nnoIR/JWq+v4TF9tSgOdcQ==, tableContent=null), ArticleFig(id=1329420067319690105, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=CN, label=图3, caption=主流气象大模型全球研究机构分布, figureFileSmall=YzRyzlgF0K0/wzY/LzW2MA==, figureFileBig=nnoIR/JWq+v4TF9tSgOdcQ==, tableContent=null), ArticleFig(id=1329420067508433786, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=EN, label=Figure 4, caption=Technical evolution of large meteorological models, figureFileSmall=Gnb/5FPk8ud5nQvzdsNMxA==, figureFileBig=0Bye1rmAHwpPwQ4pejMBcA==, tableContent=null), ArticleFig(id=1329420067621679995, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=CN, label=图4, caption=气象大模型技术演进脉络, figureFileSmall=Gnb/5FPk8ud5nQvzdsNMxA==, figureFileBig=0Bye1rmAHwpPwQ4pejMBcA==, tableContent=null), ArticleFig(id=1329420067747509116, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=EN, label=Figure 5, caption=Value chain and quantitative benefits of meteorological forecasting accuracy improvement for agricultural KPI conversion, figureFileSmall=PQj89rDB0mkhGs81UxmKmA==, figureFileBig=JqbcL0XV6teWQvro1BWW9A==, tableContent=null), ArticleFig(id=1329420067852366717, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=CN, label=图5, caption=气象预报精度改进向农业 KPI 转化的价值链条与定量效益, figureFileSmall=PQj89rDB0mkhGs81UxmKmA==, figureFileBig=JqbcL0XV6teWQvro1BWW9A==, tableContent=null), ArticleFig(id=1329420067915281278, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=EN, label=Figure 6, caption=A unified evaluation protocol framework for agricultural meteorological applications, figureFileSmall=5+taoXiUoGp9aQ013vljjA==, figureFileBig=gJeNqpW0xnU2NdyrBbrpqg==, tableContent=null), ArticleFig(id=1329420068011750271, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=CN, label=图6, caption=面向农业气象应用的统一评估协议框架, figureFileSmall=5+taoXiUoGp9aQ013vljjA==, figureFileBig=gJeNqpW0xnU2NdyrBbrpqg==, tableContent=null), ArticleFig(id=1329420068133385088, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=EN, label=Figure 7, caption=Future research directions and timeline of large meteorological models for agricultural applications, figureFileSmall=ljzuEQg4dCQHBdRCKd6+jg==, figureFileBig=vNYuUbs7uAuucgrQ+Wf7Uw==, tableContent=null), ArticleFig(id=1329420068234048385, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=CN, label=图7, caption=气象大模型面向农业应用的未来研究方向与时间线, figureFileSmall=ljzuEQg4dCQHBdRCKd6+jg==, figureFileBig=vNYuUbs7uAuucgrQ+Wf7Uw==, tableContent=null), ArticleFig(id=1329420068305351554, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=EN, label=Table 1, caption=

Evolution of Large Meteorological Models (2015–2025)

, figureFileSmall=null, figureFileBig=null, tableContent=
年份代表模型核心创新对应架构族农业适配点遗留局限
2015ConvLSTM卷积+LSTM 融合CNN/RNN县域短临遥相关弱
2022Earthformer立方体注意力时空建模Transformer站点集合显存瓶颈
2022FourCastNetAFNO 神经算子算子/谱域全球田块长程发散
2022—2023Pangu-Weather3D 地球特定 TransformerTransformer全季长程推理成本高
2023GraphCast40k mesh 节点 GNNGNN不规则站点下采样敏感
2023—2024FuXi/FengWu级联多时效延15天Transformer季节决策业务待验证
2025GenCast扩散集合概率化生成式极端事件推理开销大
2025Aurora多源地球物理预训练Foundation多任务迁移需域适应
2025Aardvark端到端观测→预报Transformer+观测稀疏区可用观测不均
), ArticleFig(id=1329420068389237635, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=CN, label=表1, caption=

气象大模型演进脉络(2015—2025年)

, figureFileSmall=null, figureFileBig=null, tableContent=
年份代表模型核心创新对应架构族农业适配点遗留局限
2015ConvLSTM卷积+LSTM 融合CNN/RNN县域短临遥相关弱
2022Earthformer立方体注意力时空建模Transformer站点集合显存瓶颈
2022FourCastNetAFNO 神经算子算子/谱域全球田块长程发散
2022—2023Pangu-Weather3D 地球特定 TransformerTransformer全季长程推理成本高
2023GraphCast40k mesh 节点 GNNGNN不规则站点下采样敏感
2023—2024FuXi/FengWu级联多时效延15天Transformer季节决策业务待验证
2025GenCast扩散集合概率化生成式极端事件推理开销大
2025Aurora多源地球物理预训练Foundation多任务迁移需域适应
2025Aardvark端到端观测→预报Transformer+观测稀疏区可用观测不均
), ArticleFig(id=1329420068456346500, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=EN, label=Table 2, caption=

Weather Prediction Models Based on GNN Architecture

, figureFileSmall=null, figureFileBig=null, tableContent=
模型位置编码原生分辨率(°)时效跨度概率输出球面感知参数规模参考文献
GraphCast节点图消息传递0.251~10 d否是约 37×106文献[25]
GenCastGNN+扩散时间步嵌入0.250~15 d是是约 41×106文献[13]
GraphDOP观测空间图编码11~5 d否是未披露文献[32]
AIFS*GNN 编码器 + 滑窗 Transformer0.251~15 d混合是未披露文献[33]
), ArticleFig(id=1329420068540232581, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=CN, label=表2, caption=

基于GNN架构的气象预测模型

, figureFileSmall=null, figureFileBig=null, tableContent=
模型位置编码原生分辨率(°)时效跨度概率输出球面感知参数规模参考文献
GraphCast节点图消息传递0.251~10 d否是约 37×106文献[25]
GenCastGNN+扩散时间步嵌入0.250~15 d是是约 41×106文献[13]
GraphDOP观测空间图编码11~5 d否是未披露文献[32]
AIFS*GNN 编码器 + 滑窗 Transformer0.251~15 d混合是未披露文献[33]
), ArticleFig(id=1329420068687033222, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=EN, label=Table 3, caption=

Weather Prediction Models Based on Transformer Architecture

, figureFileSmall=null, figureFileBig=null, tableContent=
模型位置编码原生分辨率时效跨度概率输出球面感知参数规模参考文献
Pangu-Weather3D Earth-Specific Positional Bias0.25°1~7 d否是约 256×106文献[9]
FuXiSwin V2 相对位置偏置0.25°0~15 d是否约1.5×109×3级联文献[10]
AuroraPerceiver编解码+3D Swin(相对位置偏置)0.25°1~10 d否否约1.3×109文献[11]
ClimaXViT 绝对位置编码0.25°~5.625°1~30 d否否约 107×106文献[34]
FengWuSwin Transformer 相对位置0.25°1~11 d混合否约 75×106文献[26]
SwinVRNNSwin + 变分 RNN5.625°1~5 d是否约 25×106文献[35]
MetNet-3轴向注意力 + 大感受野卷积1~2 km0~24 h否否约 225×106文献[36]
AardvarkPerceiver 跨注意力0.25°1~14 d是是未披露文献[27]
), ArticleFig(id=1329420068766724999, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=CN, label=表3, caption=

基于Transformer架构的气象预测模型

, figureFileSmall=null, figureFileBig=null, tableContent=
模型位置编码原生分辨率时效跨度概率输出球面感知参数规模参考文献
Pangu-Weather3D Earth-Specific Positional Bias0.25°1~7 d否是约 256×106文献[9]
FuXiSwin V2 相对位置偏置0.25°0~15 d是否约1.5×109×3级联文献[10]
AuroraPerceiver编解码+3D Swin(相对位置偏置)0.25°1~10 d否否约1.3×109文献[11]
ClimaXViT 绝对位置编码0.25°~5.625°1~30 d否否约 107×106文献[34]
FengWuSwin Transformer 相对位置0.25°1~11 d混合否约 75×106文献[26]
SwinVRNNSwin + 变分 RNN5.625°1~5 d是否约 25×106文献[35]
MetNet-3轴向注意力 + 大感受野卷积1~2 km0~24 h否否约 225×106文献[36]
AardvarkPerceiver 跨注意力0.25°1~14 d是是未披露文献[27]
), ArticleFig(id=1329420068854805384, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=EN, label=Table 4, caption=

Weather Prediction Models Based on Generative Models and Neural Operators

, figureFileSmall=null, figureFileBig=null, tableContent=
模型核心生成机制原生分辨率时效跨度概率输出物理约束参数规模参考文献
FourCastNetAFNO 谱域算子0.25°1~10 d否无约 74×106文献[12]
NowcastNet条件 GAN + 神经演化算子1~2 km0~3 h是强约 95×106文献[45]
CasCast级联确定性 + 潜在扩散1 km0~2 h是无约 140×106文献[46]
), ArticleFig(id=1329420068930302857, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=CN, label=表4, caption=

基于生成式模型与神经算子的气象预测模型

, figureFileSmall=null, figureFileBig=null, tableContent=
模型核心生成机制原生分辨率时效跨度概率输出物理约束参数规模参考文献
FourCastNetAFNO 谱域算子0.25°1~10 d否无约 74×106文献[12]
NowcastNet条件 GAN + 神经演化算子1~2 km0~3 h是强约 95×106文献[45]
CasCast级联确定性 + 潜在扩散1 km0~2 h是无约 140×106文献[46]
), ArticleFig(id=1329420069018383242, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=EN, label=Table 5, caption=

Cross-Architecture Comparison of Mainstream LMMs

, figureFileSmall=null, figureFileBig=null, tableContent=
架构计算
复杂度
长程依
赖能力
概率/生
成能力
物理
一致性
农业
适配度
代表模型提出年份训练数据
量需求
业务就绪度
CNN/RNN/LSTM低中低低—中高ConvLSTM/PredRNN2015低研究—试点
GNN(GraphCast 等)中—高中—高中中高GraphCast2023高业务试点
Transformer
(Pangu/FuXi/Aurora)
高高中中中—高Pangu/FuXi/Aurora2023高业务试点—业务化
SSM/Mamba中极高中中中—高Mamba/MetMamba2023中研究版
Diffusion/GAN高中—高极高中中—高GenCast/CastDiffuser/DGMR2023高研究版—业务试点
物理引导+多模态中—高中—高中高高WeatherGFT/NeuralGCM2023中—高原型—研究版
), ArticleFig(id=1329420069177766795, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=CN, label=表5, caption=

主流气象大模型架构横向对比

, figureFileSmall=null, figureFileBig=null, tableContent=
架构计算
复杂度
长程依
赖能力
概率/生
成能力
物理
一致性
农业
适配度
代表模型提出年份训练数据
量需求
业务就绪度
CNN/RNN/LSTM低中低低—中高ConvLSTM/PredRNN2015低研究—试点
GNN(GraphCast 等)中—高中—高中中高GraphCast2023高业务试点
Transformer
(Pangu/FuXi/Aurora)
高高中中中—高Pangu/FuXi/Aurora2023高业务试点—业务化
SSM/Mamba中极高中中中—高Mamba/MetMamba2023中研究版
Diffusion/GAN高中—高极高中中—高GenCast/CastDiffuser/DGMR2023高研究版—业务试点
物理引导+多模态中—高中—高中高高WeatherGFT/NeuralGCM2023中—高原型—研究版
), ArticleFig(id=1329420069253264268, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=EN, label=Table 6, caption=

Compatibility Matrix of Architectures vs Agricultural Scenario Constraints

, figureFileSmall=null, figureFileBig=null, tableContent=
架构族分辨率
失配
(田块级)
变量不匹
配(农艺
次生)
极端事
件命
中率
决策链
漏损
弥合
综合推荐场景
CNN/RNN/LSTM弱中弱强县域短临、田块物联
GNN中—强强中中不规则站点、灌区图
Transformer中强中—强中全生育期长程、
区域业务
SSM/Mamba中—强中中中—强长时序作物监测
生成式
(扩散/GAN)
中—强中强中极端事件预警、
集合投放
多模态物理引导强强中—强弱田块小气候、
作物模型耦合
), ArticleFig(id=1329420069320373133, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=CN, label=表6, caption=

架构与农业场景约束兼容性矩阵

, figureFileSmall=null, figureFileBig=null, tableContent=
架构族分辨率
失配
(田块级)
变量不匹
配(农艺
次生)
极端事
件命
中率
决策链
漏损
弥合
综合推荐场景
CNN/RNN/LSTM弱中弱强县域短临、田块物联
GNN中—强强中中不规则站点、灌区图
Transformer中强中—强中全生育期长程、
区域业务
SSM/Mamba中—强中中中—强长时序作物监测
生成式
(扩散/GAN)
中—强中强中极端事件预警、
集合投放
多模态物理引导强强中—强弱田块小气候、
作物模型耦合
), ArticleFig(id=1329420069395870606, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=EN, label=Table 7, caption=

Summary of Key Cases of LMM Applications in Agriculture

, figureFileSmall=null, figureFileBig=null, tableContent=
应用场景主要气象大模型方法气象指标农业 KPI/经济价值引用文献空间尺度时间尺度定量数据
灌溉调度FuXi-S2S/ClimaX/GenCast参考蒸散, 累积降水
(集合)
节水 5%~15%;
净收益增加85 美元/公顷
[21, 53−54]田块—灌区
(1~10 km)
15~90 dSWF:ETo MAE
17.0~170.3 mm
霜冻/热浪预警GenCast/CastDiffuser/HybridPOD/FAR/Brier
(条件概率)
提前期增加3~9 d;
产量增益增加13%
[25, 48, 55]县域—田块
(1~10 km)
0~10 dGraphCast:
飓风 Lee 提前 9 d
病虫害风险FuXi/FengWu + LR/GBT冠层温度、叶面湿度提前预警 1~2 周;
预测稳定增加20%~30%
[52]县域(1~5 km)3~15 dCWSM 误差
下降20%~30%
产量预测ACCESS-S2+APSIM/
SEAS5+WOFOST/Aurora
T/P/Rs 全季节序列R2 0.4~0.88;
产量增加281 kg/ha
(增加13%)
[11, 22, 55]县域—田块
(1~10 km)
60~180 d澳洲小麦
68% 案例增产
精准作业调度NowcastNet/MetNet-3/
TimeGPT
小时级风速、降水喷药漂移下降5%~10%;
机械效率增加
[57]田块
(100~500 m)
0~24 h风速预报
RMSE 0.8 m/s
), ArticleFig(id=1329420069483950991, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, language=CN, label=表7, caption=

气象大模型在农业应用中的核心案例汇总

, figureFileSmall=null, figureFileBig=null, tableContent=
应用场景主要气象大模型方法气象指标农业 KPI/经济价值引用文献空间尺度时间尺度定量数据
灌溉调度FuXi-S2S/ClimaX/GenCast参考蒸散, 累积降水
(集合)
节水 5%~15%;
净收益增加85 美元/公顷
[21, 53−54]田块—灌区
(1~10 km)
15~90 dSWF:ETo MAE
17.0~170.3 mm
霜冻/热浪预警GenCast/CastDiffuser/HybridPOD/FAR/Brier
(条件概率)
提前期增加3~9 d;
产量增益增加13%
[25, 48, 55]县域—田块
(1~10 km)
0~10 dGraphCast:
飓风 Lee 提前 9 d
病虫害风险FuXi/FengWu + LR/GBT冠层温度、叶面湿度提前预警 1~2 周;
预测稳定增加20%~30%
[52]县域(1~5 km)3~15 dCWSM 误差
下降20%~30%
产量预测ACCESS-S2+APSIM/
SEAS5+WOFOST/Aurora
T/P/Rs 全季节序列R2 0.4~0.88;
产量增加281 kg/ha
(增加13%)
[11, 22, 55]县域—田块
(1~10 km)
60~180 d澳洲小麦
68% 案例增产
精准作业调度NowcastNet/MetNet-3/
TimeGPT
小时级风速、降水喷药漂移下降5%~10%;
机械效率增加
[57]田块
(100~500 m)
0~24 h风速预报
RMSE 0.8 m/s
)], attaches=null, journal=Journal(id=1324759606947315908, delFlag=0, nameCn=计算机研究与发展, nameEn=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, issn=1000-1239, eissn=null, cn=11-1777/TP, coden=null, periodic=0, language=CN, oaType=null, ccby=null, superviseOffice=null, ownerOffice=null, pubOffice=null, editorOffice=null, officeType=null, aims=null, clcCode=null, officeProv=null, officeCity=null, officeAddr=null, officeZip=null, officeEmail=null, officePhone=null, editDirector=null, officeDirector=null, officeDirectorPhone=null, officeStaffNum=null, officeEmpNum=null, coverPicUrl=k2l8zgqwstuxxevFX2qxVA==, journalPrice=null, startedYear=null, abbrevIsoEn=Journal of Computer Research and Development, journalRemark=null, publicationField=null, createdTime=1793776903141, updatedTime=1793850698612, createdBy=18614031015, updatedBy=13701087609, firstLetterCn=J, firstLetterEn=J, subjectCode=Electronic Information, subjectName=null, subjectCodeEn=Electronic Information, subjectNameEn=null, picCn=k2l8zgqwstuxxevFX2qxVA==, picEn=r6k0Ry+RD+c/VXj5QarmDw==, jcr=null, cjcr=null, exts=[JournalExt(id=1325069127710286688, language=CN, name=计算机研究与发展, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698640, updatedTime=1793850698640, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1325069127777395553, language=EN, name=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698656, updatedTime=1793850698656, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1324759914146529478, websiteList=[Website(id=1325070300139892771, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN, language=CN, createTime=1793850978169, createBy=18614031015, updateTime=1793851040259, updateBy=18614031015, name=计算机研究与发展-中文, tplId=1146099689490845704, title=计算机研究与发展, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070893751349360, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=articleTextType, value=kx, createTime=1793851119697, updateTime=1793851119697, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893730377837, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=banner, value=null, createTime=1793851119692, updateTime=1793851119692, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893772320883, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=grayFlag, value=0, createTime=1793851119702, updateTime=1793851119702, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893721989228, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851119690, updateTime=1793851119690, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893784903797, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=minRunFlag, value=0, createTime=1793851119705, updateTime=1793851119705, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893742960751, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic, createTime=1793851119695, updateTime=1793851119695, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893780709492, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=silenceFlag, value=0, createTime=1793851119704, updateTime=1793851119704, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893738766446, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1793851119694, updateTime=1793851119694, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893755543665, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeColor, value=null, createTime=1793851119698, updateTime=1793851119698, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893763932274, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeStyle, value=null, createTime=1793851119700, updateTime=1793851119700, creator=18614031015, updator=18614031015)]), Website(id=1325070300580294698, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN, language=EN, createTime=1793850978274, createBy=18614031015, updateTime=1793851036638, updateBy=18614031015, name=计算机研究与发展-英文, tplId=1146101810881728533, title=Journal of Computer Research and Development, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070925288321148, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=articleTextType, value=kx, createTime=1793851127216, updateTime=1793851127216, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925263155321, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=banner, value=null, createTime=1793851127210, updateTime=1793851127210, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925309292671, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=grayFlag, value=0, createTime=1793851127221, updateTime=1793851127221, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925254766712, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851127208, updateTime=1793851127208, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925326069889, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=minRunFlag, value=0, createTime=1793851127225, updateTime=1793851127225, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925279932539, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic, createTime=1793851127214, updateTime=1793851127214, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925317681280, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=silenceFlag, value=0, createTime=1793851127223, updateTime=1793851127223, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925271543930, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_en_623/, createTime=1793851127212, updateTime=1793851127212, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925292515453, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeColor, value=null, createTime=1793851127217, updateTime=1793851127217, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925300904062, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeStyle, value=null, createTime=1793851127219, updateTime=1793851127219, creator=18614031015, updator=18614031015)])], journalTitle=计算机研究与发展, weixinUrl=null, journalUrl=https://crad.ict.ac.cn/, iacademicId=null, status=1, seqNo=null, journalTitleEn=Journal of Computer Research and Development, journalPhotoCn=k2l8zgqwstuxxevFX2qxVA==, journalPhotoEn=r6k0Ry+RD+c/VXj5QarmDw==, journalFirstLetter=J, journalRecommend=null, journalNew=null, journalCollection=null, jcrJf=null, cjcrJf=null, jcrJfStr=null, cjcrJfStr=null, submissionFirstDecision=null, sciSubjectClassification=null, casSubjectClassification=null, citeScore=null, totalCitationFrequency=null, icpCode=null, psCode=null, advertisingLicenseCode=null, copyrightInformation=null, country=null, option=, provinceCode=null, provinceName=null, collectFlag=false, interPubPlatform=, interPubPlatformUrl=null), detailUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202660002, detailUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/10.7544/issn1000-1239.202660002, pdfUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/PDF/10.7544/issn1000-1239.202660002, pdfUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/PDF/10.7544/issn1000-1239.202660002, aliStartDate=0, aliEndDate=0, collectionFlag=false, citedCount=null, citedUrl=null, previewStatus=0, delFlag=0, hasFullText=1, orderTime=1782835200000, fullTextJson=null, articleText=null, reference=null)
收藏切换
气象大模型在农业应用中的综述
邢茹雪 , 朱建军 , 聂紫婷 , 郭庆昊 , 李珊 , 史银雪 , 王耀君 *
计算机研究与发展 | AI4Science:科学研究的新范式专题 2026,63(7): 1690-1711
收起
收藏切换
计算机研究与发展 |AI4Science:科学研究的新范式专题 2026 , 63 (7) : 1690 -1711
气象大模型在农业应用中的综述
全屏
[Author(id=1329420064127824712, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=xingruxue@cau.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420064194933578, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064127824712, language=EN, stringName=Ruxue Xing, firstName=Ruxue, middleName=null, lastName=Xing, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, bio={"content":"

born in 1998. Master candidate. Her main research interests include artificial intelligence and meteorological forecasting

"}, bioImg=null, bioContent=

born in 1998. Master candidate. Her main research interests include artificial intelligence and meteorological forecasting

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420064253653835, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064127824712, language=CN, stringName=邢茹雪, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=中国农业大学信息与电气工程学院 北京 100083, bio={"img":"E5wLCWSlGIC6yLm2TXG8KQ==","content":"

1998年生。硕士研究生。主要研究方向为人工智能、气象预测

"}, bioImg=E5wLCWSlGIC6yLm2TXG8KQ==, bioContent=

1998年生。硕士研究生。主要研究方向为人工智能、气象预测

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420064018772804, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, xref=null, ext=[AuthorCompanyExt(id=1329420064027161413, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083), AuthorCompanyExt(id=1329420064035550022, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国农业大学信息与电气工程学院 北京 100083)])]), Author(id=1329420064312374093, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420064392065871, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064312374093, language=EN, stringName=Jianjun Zhu, firstName=Jianjun, middleName=null, lastName=Zhu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, bio={"content":"

born in 1997. PhD candidate. His main research interests include deep learning and time series prediction

"}, bioImg=null, bioContent=

born in 1997. PhD candidate. His main research interests include deep learning and time series prediction

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420064459174736, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064312374093, language=CN, stringName=朱建军, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=中国农业大学信息与电气工程学院 北京 100083, bio={"img":"IPJmrTFIcR2eZD4JSfaRRw==","content":"

1997年生。博士研究生。主要研究方向为深度学习、时序预测

"}, bioImg=IPJmrTFIcR2eZD4JSfaRRw==, bioContent=

1997年生。博士研究生。主要研究方向为深度学习、时序预测

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420064018772804, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, xref=null, ext=[AuthorCompanyExt(id=1329420064027161413, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083), AuthorCompanyExt(id=1329420064035550022, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国农业大学信息与电气工程学院 北京 100083)])]), Author(id=1329420064517894994, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420064593392468, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064517894994, language=EN, stringName=Ziting Nie, firstName=Ziting, middleName=null, lastName=Nie, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, bio={"content":"

born in 2003. Master candidate. Her main research interest includes deep learning

"}, bioImg=null, bioContent=

born in 2003. Master candidate. Her main research interest includes deep learning

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420064656307029, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064517894994, language=CN, stringName=聂紫婷, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=中国农业大学信息与电气工程学院 北京 100083, bio={"img":"kaWFNaJLNpW1IiPxECoDUw==","content":"

2003年生。硕士研究生。主要研究方向为深度学习

"}, bioImg=kaWFNaJLNpW1IiPxECoDUw==, bioContent=

2003年生。硕士研究生。主要研究方向为深度学习

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420064018772804, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, xref=null, ext=[AuthorCompanyExt(id=1329420064027161413, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083), AuthorCompanyExt(id=1329420064035550022, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国农业大学信息与电气工程学院 北京 100083)])]), Author(id=1329420064715027287, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, orderNo=3, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420064782136153, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064715027287, language=EN, stringName=Qinghao Guo, firstName=Qinghao, middleName=null, lastName=Guo, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, bio={"content":"

born in 2003. Master candidate. His main research interest includes agricultural production warning

"}, bioImg=null, bioContent=

born in 2003. Master candidate. His main research interest includes agricultural production warning

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420064866022234, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064715027287, language=CN, stringName=郭庆昊, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=中国农业大学信息与电气工程学院 北京 100083, bio={"img":"pKXq7Fgf5dfO22y0hRqFGA==","content":"

2003年生。硕士研究生。主要研究方向为农业生产预警

"}, bioImg=pKXq7Fgf5dfO22y0hRqFGA==, bioContent=

2003年生。硕士研究生。主要研究方向为农业生产预警

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420064018772804, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, xref=null, ext=[AuthorCompanyExt(id=1329420064027161413, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083), AuthorCompanyExt(id=1329420064035550022, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国农业大学信息与电气工程学院 北京 100083)])]), Author(id=1329420064937325404, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, orderNo=4, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420065000239966, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064937325404, language=EN, stringName=Shan Li, firstName=Shan, middleName=null, lastName=Li, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, bio={"content":"

born in 1995. PhD, lecturer. Her main research interests include statistical pattern recognition and agricultural large model

"}, bioImg=null, bioContent=

born in 1995. PhD, lecturer. Her main research interests include statistical pattern recognition and agricultural large model

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420065054765919, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420064937325404, language=CN, stringName=李珊, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=中国农业大学信息与电气工程学院 北京 100083, bio={"img":"BjCOJrYaP8EU0fZ0o9AjsA==","content":"

1995年生。博士,讲师。主要研究方向为统计模式识别、农业大模型

"}, bioImg=BjCOJrYaP8EU0fZ0o9AjsA==, bioContent=

1995年生。博士,讲师。主要研究方向为统计模式识别、农业大模型

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420064018772804, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, xref=null, ext=[AuthorCompanyExt(id=1329420064027161413, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083), AuthorCompanyExt(id=1329420064035550022, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国农业大学信息与电气工程学院 北京 100083)])]), Author(id=1329420065134457697, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, orderNo=5, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420065205760867, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420065134457697, language=EN, stringName=Yinxue Shi, firstName=Yinxue, middleName=null, lastName=Shi, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, bio={"content":"

born in 1974. PhD, associate professor. Her main research interests include computer vision, smart agriculture, and smart healthcare

"}, bioImg=null, bioContent=

born in 1974. PhD, associate professor. Her main research interests include computer vision, smart agriculture, and smart healthcare

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420065264481124, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420065134457697, language=CN, stringName=史银雪, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=中国农业大学信息与电气工程学院 北京 100083, bio={"img":"AcS7IUc3PKvkCUdvWzEfgg==","content":"

1974年生。博士,副教授。主要研究方向为计算机视觉、智慧农业、智能医疗

"}, bioImg=AcS7IUc3PKvkCUdvWzEfgg==, bioContent=

1974年生。博士,副教授。主要研究方向为计算机视觉、智慧农业、智能医疗

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420064018772804, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, xref=null, ext=[AuthorCompanyExt(id=1329420064027161413, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083), AuthorCompanyExt(id=1329420064035550022, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国农业大学信息与电气工程学院 北京 100083)])]), Author(id=1329420065360950118, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, orderNo=6, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=wangyaojun@cau.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1329420065428058984, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420065360950118, language=EN, stringName=Yaojun Wang, firstName=Yaojun, middleName=null, lastName=Wang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, bio={"content":"

born in 1985. PhD, associate professor. His main research interests include knowledge graphs, artificial intelligence, and large models

"}, bioImg=null, bioContent=

born in 1985. PhD, associate professor. His main research interests include knowledge graphs, artificial intelligence, and large models

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420065503556457, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, authorId=1329420065360950118, language=CN, stringName=王耀君, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=中国农业大学信息与电气工程学院 北京 100083, bio={"img":"cDSG5wGKQETezA+E2GuB2w==","content":"

1985年生。博士,副教授。主要研究方向为知识图谱、人工智能、大模型

"}, bioImg=cDSG5wGKQETezA+E2GuB2w==, bioContent=

1985年生。博士,副教授。主要研究方向为知识图谱、人工智能、大模型

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420064018772804, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, xref=null, ext=[AuthorCompanyExt(id=1329420064027161413, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=College of Information and Electrical Engineering, China Agricultural University, Beijing 100083), AuthorCompanyExt(id=1329420064035550022, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420060856267557, companyId=1329420064018772804, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国农业大学信息与电气工程学院 北京 100083)])])]
邢茹雪 , 朱建军, 聂紫婷, 郭庆昊, 李珊, 史银雪, 王耀君*
作者信息
  • 中国农业大学信息与电气工程学院 北京 100083
通讯作者:
王耀君()
作者简介:

作者贡献声明

邢茹雪完成论文框架设计、文献调研与全文撰写;朱建军参与算法分类体系的梳理与章节结构的设计;聂紫婷、郭庆昊参与文献整理与图表制作;李珊、史银雪参与论文修改;王耀君提出研究选题与论文思路,并对论文进行总体指导与修改完善。邢茹雪与朱建军是共同第一作者。

1998年生。硕士研究生。主要研究方向为人工智能、气象预测

1997年生。博士研究生。主要研究方向为深度学习、时序预测

2003年生。硕士研究生。主要研究方向为深度学习

2003年生。硕士研究生。主要研究方向为农业生产预警

1995年生。博士,讲师。主要研究方向为统计模式识别、农业大模型

1974年生。博士,副教授。主要研究方向为计算机视觉、智慧农业、智能医疗

1985年生。博士,副教授。主要研究方向为知识图谱、人工智能、大模型

Survey of Large Meteorological Models in Agricultural Applications
Ruxue Xing , Jianjun Zhu, Ziting Nie, Qinghao Guo, Shan Li, Yinxue Shi, Yaojun Wang*
Affiliations
  • College of Information and Electrical Engineering, China Agricultural University, Beijing 100083
  • born in 1998. Master candidate. Her main research interests include artificial intelligence and meteorological forecasting

    born in 1997. PhD candidate. His main research interests include deep learning and time series prediction

    born in 2003. Master candidate. Her main research interest includes deep learning

    born in 2003. Master candidate. His main research interest includes agricultural production warning

    born in 1995. PhD, lecturer. Her main research interests include statistical pattern recognition and agricultural large model

    born in 1974. PhD, associate professor. Her main research interests include computer vision, smart agriculture, and smart healthcare

    born in 1985. PhD, associate professor. His main research interests include knowledge graphs, artificial intelligence, and large models

出版时间: 2026-07-01 doi: 10.7544/issn1000-1239.202660002
文章导航
收藏切换

近年来,深度学习在气象预测与气候建模中的应用取得了显著进展,并在农业气象服务中展现出重要价值。尽管已有综述文章对相关方法进行了系统梳理,但多聚焦于模型架构、任务类型或应用场景,尚缺乏从农业视角出发的系统总结。聚焦气象大模型(large meteorological models,LMMs)的技术演进与应用实践,围绕模型结构建立分类体系,系统回顾传统数值方法与深度学习架构的发展脉络。根据建模范式的不同,将主流方法归纳为六大类:经典深度神经网络、图神经网络、Transformer 架构、状态空间模型、生成式模型以及物理引导与多模态融合模型,并进一步梳理了典型模型在农业场景中的应用,包括农业气象预报、极端天气预警、防灾减灾及精准农业决策支持等关键任务。此外,还汇总了常用数据集与评估指标,分析当前气象大模型在泛化能力、不确定性建模与物理一致性方面的挑战,并展望了未来研究的关键方向,旨在为气象大模型在农业场景中的深入研究与应用提供系统性参考。

气象大模型  /  深度学习  /  时间序列建模  /  天气预报  /  农业应用

In recent years, deep learning has achieved remarkable progress in meteorological forecasting and climate modeling, exerting a growing influence on agricultural applications. While numerous review articles have systematically examined related methodologies, most focus on model architectures, task types, or application domains, with limited attention given to the agricultural perspective. This paper provides a comprehensive review of the development of large meteorological models (LMMs), structured around model architecture. We trace the evolution from traditional numerical methods to deep learning approaches and categorize existing models into six major classes: classical deep neural networks, graph neural networks, Transformer-based architectures, state space models, generative models, and physics-informed/multimodal fusion models. Furthermore, we summarize representative applications of LMMs in agriculture, including meteorological forecasting services, extreme weather early warning and disaster mitigation, and decision support systems for precision agriculture. We also examine commonly used datasets and evaluation metrics in the field. This is followed by a discussion of current challenges, particularly in areas of generalization capability, uncertainty quantification, and physical consistency. Finally, we outline key research directions for future exploration in this rapidly evolving field. This work aims to provide a systematic reference for advancing both the research and application of LMMs in agricultural contexts.

large meteorological models (LMMs)  /  deep learning  /  time series modeling  /  weather forecasting  /  agricultural applications
邢茹雪, 朱建军, 聂紫婷, 郭庆昊, 李珊, 史银雪, 王耀君. 气象大模型在农业应用中的综述. 计算机研究与发展, 2026 , 63 (7) : 1690 -1711 . DOI: 10.7544/issn1000-1239.202660002
Ruxue Xing, Jianjun Zhu, Ziting Nie, Qinghao Guo, Shan Li, Yinxue Shi, Yaojun Wang. Survey of Large Meteorological Models in Agricultural Applications[J]. Journal of Computer Research and Development, 2026 , 63 (7) : 1690 -1711 . DOI: 10.7544/issn1000-1239.202660002
全球气候变化背景下,粮食安全面临多重压力。联合国粮食及农业组织(FAO)长期跟踪的全球粮食产量与气象波动数据显示,极端高温、季节性干旱、突发性强降水与热带气旋等灾害性天气事件的发生频率与强度近二十年持续上升[1-3],其影响已远超过单一作物或单一区域的边界。主产带的产量异常会沿全球贸易链向价格、储备与营养结构传导,进而对人口众多的发展中国家造成连锁性的食物可获得性冲击。历史性的农业气象研究指出,从近代农业起步至今,气象条件一直是制约作物生育期推进、产量稳定性与农事决策窗口的首要外生变量;20 世纪初的农业气象学曾以提升农场盈利为目标,21 世纪以来其学科使命逐步转向“保障粮食安全与可持续生产”双重目标[4]。气象服务在粮食系统韧性建设中的作用日益重要。
与交通、能源等其他气象敏感行业相比,农业对气象信息具有三重特异性依赖。其一是时间关键性:播种、施药、灌溉、抢收等关键作业受限于数小时至数日的“决策窗”,错过即可能造成不可逆损失。其二是尺度敏感性:田块管理单元普遍处于 30~500 m 量级,其上的小气候差异(如冠层温度、叶面湿润时间、土壤含水率)足以决定病害爆发与产量分级;而全球 0.25°(约 25 km)业务网格无法分辨这些差异[5-6]。其三是变量多样性:农艺决策不仅依赖大气主流变量(气温、降水、风速),更需要冠层温度、有效辐射、有效积温、近地面湿度等次生变量与极端事件的窗口化概率。这些维度往往超出通用气象指标体系的优化目标[4,7]。上述特异性意味着,气象大模型在农业领域的价值评估必须以“农业 KPI 转化率”为最终标尺,而非仅以通用气象指标的改进幅度为依据。
近年来,深度学习方法在气象建模领域取得了里程碑式进展。基于全球再分析资料训练的气象大模型(large meteorological models,LMMs)不仅在 500 hPa 位势高度、海平面气压等传统指标上比肩甚至超越主流业务模式,其推理时延较传统数值天气预报降低3~4个数量级,为面向多变量、多场景、多用户的高频次农业气象服务提供了前所未有的供给能力[7-8]。然而,气象大模型从“通用基础模型”跨越到“农业可用工具”,仍需克服分辨率、变量丰度、概率表达与决策链漏损等多重障碍。本文即以“农业问题—模型选型—效益转化”为主线,系统综述这一新兴技术体系,辨析其相对于传统数值预报的边际价值,并指出未来面向田块级、决策级、价值级的研究方向。
气象大模型在农业应用中面临 4 类核心挑战:1)空间分辨率失配——主流全球模型仍以 0.25°(约 25 km)为原生分辨率,与田块 30~500 m 的管理单元存在2~3个数量级差距;2)农艺次生变量缺失——对冠层温度、叶面湿润时长、有效积温等农艺次生变量的覆盖有限,多依赖后处理二次推演;3)极端事件命中率不足——数据驱动架构在能量守恒与极端事件物理机制上缺乏显式约束,对气候态尾部事件的命中率有限;4) 决策链逐级漏损——预报技巧在“预报—解读—发布—接收—调整”多级中介中逐环衰减,影响业务部门与一线农户的决策信任。本文将按此坐标系展开综述。
在气象大模型出现之前,自回归整合移动平均(ARIMA)等传统统计模型曾长期承担短期气象预测的辅助角色,但其线性结构难以刻画大气系统中显著的非线性与多尺度耦合行为。与之形成鲜明对比的是,深度神经网络可以从海量历史观测与再分析数据中自动学习高维时空依赖,实现端到端、低延迟的天气预测。研究表明,新一代气象大模型在中期预报、极端事件检索与概率化输出3个方面展现出显著潜力:Pangu-Weather 通过三维地球特定 Transformer 在 0.25° 网格上 1~7 天预报中超越了 ECMWF IFS[9],FuXi 通过级联U-Transformer 将稳定技巧延伸至15天中期尺度[10],而微软研究院牵头的Aurora模型进一步将“基础模型”理念引入地球系统[11]。
尽管早期气象大模型(如 FourCastNet[12])在精度上曾不及最优数值预报系统,但伴随结构创新与训练策略的迭代,近年来已有多个气象大模型进入业务测试阶段。与此同时,扩散模型与生成对抗网络的引入使得高分辨率短临降水预报与集合化概率预报获得新的动力,GenCast、CastDiffuser 等代表性工作正不断刷新极端天气预报的可靠性边界[13-14]。
以基础模型范式为标志,气象大模型正从单点替代模型逐步演进为支持多下游任务的通用基础架构。通过在全球历史气象数据上完成大规模预训练,同一模型可经由轻量微调适配到台风路径、极端降水、海温异常乃至生态遥相关等多种下游任务[11,15]。微软研究院的Aurora整合大气-海洋-陆面-生态系统多源数据,其预训练数据时长超过 100 万小时,涵盖ERA5、海洋、空气质量等多源地球系统数据,在空气质量、海浪、台风路径、0.1°高分辨率天气等下游任务上展现出可观的迁移能力[11]。基础模型与垂直微调相结合的范式,为气象大模型走向农业等高度专业化场景提供了切实可行的技术路径。
现有针对深度学习与气象建模的综述已为该领域奠定基础[8,16],但在分析视角与分类框架上各有侧重,尚缺乏从农业问题出发、贯穿“模型—指标—决策—效益”全链条的系统总结。基于此,本文以中国农业生产与全球粮食安全的现实需求为出发点,梳理气象大模型的技术演进脉络,重点分析其在农业场景下的适用性与边界。
数值天气预报(numerical weather prediction,NWP)[17]通过在三维网格上数值求解描述大气流体动力学与热力学过程的偏微分方程组,来推进未来大气状态的时空演化,是现代气象业务的物理基石。代表性的全球NWP系统包括欧洲中期天气预报中心的IFS、美国国家环境预测中心的GFS与英国气象局的MOGREPS等;区域尺度模式则以WRF、MM5及国内的GRAPES等为典型代表[18]。在长期演进过程中,NWP确立了“观测—资料同化—动力积分—产品后处理”的标准工作流,其物理可解释性与守恒律保障使其在业务化预报中长期占据主导地位。
在统计建模方向,自回归整合移动平均(ARIMA)等线性模型曾广泛用于短期气温与降水预测,但难以刻画大气系统的非线性动力学行为。此后,隐马尔可夫模型、支持向量机、随机森林、梯度提升树等非线性方法被尝试用于降水概率预报与气候指数预测;近十年里,机器学习还以“NWP偏差订正”和“集合后处理”两类边缘任务嵌入到既有业务链中,显著提升了概率预报的可靠性[19-20]。然而,受限于早期数据规模与网络深度,端到端的数据驱动天气建模长期未能形成实用系统。近年来再分析数据规模的扩大、计算硬件的迭代与深度神经网络架构的成熟,共同推动了气象大模型的兴起,并在多项业务指标上接近或超越传统NWP[7-8]。
将气象预报真正转化为田间决策时,传统NWP暴露出若干在通用气象评估中并不突出,却在农业链条中放大的结构性短板。审视农业用户的需求侧不难发现:作物生产对气象信息的依赖具有强烈的“时间窗”特征,即播种、施药、抢收等关键作业往往集中在数小时至数日的窗口内[21];同时,田块管理单元的空间尺度普遍在数十米至公里量级,远小于全球NWP 0.25°(约25 km)或区域模式数公里的有效分辨率[5-6]。正是这种“决策窗短、空间窗细、变量窗专”的不对称需求,使得通用气象指标上的微小偏差在农业端被显著放大,具体表现有4点。
其一,空间分辨率与农业管理单元失配。当前业务化全球模式的水平分辨率多在25 km量级,区域模式可下沉至3~5 km,但精准农业所需的田块尺度往往在30~500 m,二者差距在1~3个数量级。已有研究将ECMWF SEAS5(0.5°×0.5°,约50 km)作为东南亚水稻生产的预报输入,发现其在主季首播阶段尚可提供1~3个月的可用技巧,但在延后播种与旱季阶段技巧迅速衰减[22];类似地,CFSv2在尼泊尔水稻产量预测中受到约100 km×100 km网格分辨率的根本性制约,难以刻画山地—平原交错地带的降水分异[6]。简单地统计降尺度(如分位数映射)虽可缩小网格,但不可避免地破坏雨日的时间自相关结构,使得后续作物模型对干湿交替过程的模拟出现系统性偏差[5]。
其二,预报变量与农艺需求不匹配。传统NWP以500 hPa位势高度、海平面气压、地面气温、降水等大气主流变量的预报技巧为优化目标,但农业决策真正关心的是冠层温度、叶面湿润时间、土壤体积含水率、有效辐射、累积积温与极端窗口等“次生变量”。这些变量需要在NWP后处理阶段借助经验关系或作物模型再二次推演,每一次转换都引入新的不确定性。例如,Jha等人[6]的研究指出,SEAS5对水稻产量预测技巧的限制因素并非降水,而是温度变量上的偏差。这一温度偏差在通用气象指标评估中不易察觉。
其三,极端事件命中率与农业灾损分布不匹配。霜冻、高温热害、短时强降水等农业关键性极端事件,往往位于气候态分布的尾部,且具有局地化与突发性的双重特征。传统NWP的均方根误差类指标对均值附近的偏差敏感,对尾部命中率的反映则十分有限[23];与此同时,受次网格物理过程参数化与初始扰动设计的影响,集合预报对低概率高影响事件的可靠性评估仍是公开难题[17]。在农业语境下,一次错过的霜冻预警或一次错失的强降水提示所带来的损失可达数百美元每公顷,远超过常规气象指标改进所能体现的“价值差”[21]。
其四,预报价值在农业决策链条上的逐级漏损。即使NWP产品本身具备技巧,其转化为可执行决策仍要经过“预报—解读—信息发布—农户接收—管理调整”的多级中介,任何一环的延迟、误读或语义损耗都将削弱最终的农业效用[21]。已有的经济学评估显示,在湄公河三角洲水稻系统中,将季节性降水预报技巧从当前约70%提升到理想的100%水平,可使最优播期决策的农户净收益由 135 美元/公顷提升至 220 美元/公顷(净增 85 美元/公顷)——这是衡量“气象精度→农业效益”转化效率的少数定量证据之一[21]。正是上述4方面短板,构成了传统NWP在农业服务深度上的边界,也为下一节将要讨论的气象大模型在分辨率、变量丰度、概率表达与端到端学习等维度上提供了改进空间。
上述4类短板并非孤立,每一类都对 AI 时代的方法设计提出了具体约束:对分辨率失配,深度学习超分辨率与3维 Transformer 类架构(见第 2.3 节与第 2.6 节)通过球面自注意力与 1 km 级降尺度网络弥合空间尺度鸿沟;对预报变量与农艺需求不匹配,多模态融合与物理引导神经网络(见第 2.6 节)通过将卫星、雷达、地面站与作物模型统一嵌入,补齐冠层温度、叶面湿润时间等次生变量;对极端事件命中率瓶颈,生成式与扩散类架构(见第 2.5 节)利用条件概率采样显著改善尾部命中率;对决策链漏损,则有赖第 3 节的问题驱动应用框架与第 4 节的农业导向评估指标共同收口。下一节即按此坐标系展开架构梳理。
图1直观汇总了上述4类核心挑战,为后续架构选型与农业应用分析提供参照。
为便于后续对深度学习架构进行统一比较,本节给出气象大模型所涉及的预报任务与评估指标的形式化描述。如图2综合展示了气象大模型的端到端建模架构,将输入输出的形式化定义、核心算子演进、典型损失函数与多时效评估协议整合为统一的理论框架。
1)预报任务定义。给定过去$ k $个时刻的多变量大气状态序列 $ {\boldsymbol X}_{t-k:t}\in {\mathbb R}^{\left(k+1\right)\times V\times H\times W\times L} $($ V $为变量数, $ H\times W $为水平网格,$ L $为垂直层数),气象大模型 $ {f}_{\theta } $ 学习的端到端映射形式为
$ {\hat{\boldsymbol X}}_{t+1: t+h}={f}_{\theta }\left({\boldsymbol X}_{t-k: t}\right)\;, $
其中$ h $为预报步数,$ \theta $为模型参数。对于确定性模型,输出为单一预报场;对于概率化模型(如 GenCast),输出为条件分布的样本集合,其概率密度函数如式(2)所示。
$ {\hat{\boldsymbol X}}_{t+1: t+h}\sim {p}_{\theta }\left(\cdot |{\boldsymbol X}_{t-k: t}\right)\;。 $
2)主流损失函数。训练目标以均方误差(MSE)为基础,主流模型多采用纬度加权以校正球面面积差异,其形式定义如式(3)所示。
$ {{MSE}}=\frac{1}{N}\displaystyle\sum_{i=1}^Nw\left({\varphi }_{i}\right){\left({\hat{y}}_{i}-{y}_{i}\right)}^{2}\;。 $
概率化模型则广泛采用连续排序概率得分(CRPS),其形式如式(4)所示,其中$ F $ 为预报累积分布、$ y $ 为观测,CRPS 同时度量预报分布的位置与离散度。
$ {CRPS}\left(F,y\right)=\int_{-\infty }^{+\infty }{\left[F\left(x\right)-1\left\{x\geq y\right\}\right]}^{2}{\rm{d}}x\;。 $
对极端事件预报,常附加如式(5)所示的分位数损失 $ {L}_{\tau } $ 以压低尾部偏差,$ \tau $为目标分位数。
$ {L}_{\tau }\left(y,\hat{y}\right)=\max \left(\tau \left(y-\hat{y}\right),\left(\tau -1\right)\left(y-\hat{y}\right)\right) \;。$
3)评估指标与协议。中期预报技巧的两大基线指标为均方根误差 (RMSE) 与异常相关系数 (ACC),分别如式(6)与式(7)所示,其中$ {\hat y}' = \hat y - w$,$ y' = y -w $,w为气候态。
$ {RMSE=}\sqrt{\frac{1}{N}\displaystyle\sum_{i=1}^N{\left({\hat{y}}_{i}-{y}_{i}\right)}^{2}}\;, $
$ {ACC}=\frac{\displaystyle\sum ({\hat y}'\cdot y' )}{\sqrt{\displaystyle\sum{{\hat y}'^{2}}\cdot \sum {y'}^{2}}} \;。$
概率化输出补充 CRPS、Brier Score 与可靠性图;事件类预报(降水、雷暴)采用 POD、FAR、ETS 等。Lead Time 约定为预报起报时刻至目标时刻的小时数或天数,实验中按 24 h、72 h、120 h、168 h、240 h 多档分报以刻画随时效的技巧衰减。
在数据基线方面,ERA5[24]、AgERA5、AgMERRA 等再分析与专项数据集是当前气象大模型训练与评估的事实底座,本文第 4.3节、第4.4 节将结合通用与农业2类视角对其作进一步展开。
本节按深度学习架构族展开综述,依次讨论早期 CNN/RNN 基线、图神经网络、Transformer、状态空间模型、生成式模型与物理引导多模态方法六大分支,每一子节末结合第1节列出的农业场景约束进行分析。之所以选择架构族作为分类维度,主要考虑3点:其一,架构族直接对应模型对气象场时空结构的归纳偏置(GNN 对应不规则站点拓扑,Transformer 对应全局长程依赖,扩散模型对应概率化表达),便于与第 1节列出的 4 类农业约束建立一一映射;其二,架构族演进是该领域近十年最显著的技术主线,按此组织可同时反映方法论变迁与时间脉络;其三,同一架构族内的模型在计算复杂度、数据需求与部署成本上更具可比性,便于后续按农业问题做架构选型。架构族之外,空间离散、时间推进、输出形式等亦可作为分类维度,本节以架构族为主线,其余维度的对照见第 2.7节。
在第1节对传统数值天气预报与统计建模方法回顾的基础上,本节按架构类别系统综述近十年代表性气象大模型的进展,并在每个子节末尾结合第1.2节列出的农业场景约束进行分析,为后续第3节在选型层面建立“农业问题→模型架构”的对应关系奠定基础。
从演进脉络看,气象大模型的架构选择并非偶然,而是历代典型模型在解决前代局限中逐步成长起来的。2015 年 Shi 等提出的 ConvLSTM(详见第 2.1节) 将卷积与长短期记忆融合,奠定了雷达回波外推与短临降水预测的端到端基线,但卷积核的局部感受野难以刻画跨大陆的遥相关,时间维度亦受梯度消失制约。2022 年提出的 FourCastNet将自适应傅里叶神经算子(AFNO)引入全球尺度谱域建模,证明AI模型可在 500 hPa 位势等核心指标上逼近 IFS,但能量守恒与长预报期发散仍难解决[12]。2022年末至 2023 年,华为云 Pangu-Weather 以3维地球特定 Transformer(3D earth-specific Transformer)补强球面几何与垂直层耦合,在 1~7 天时效全面对标 IFS,但对不规则台站与农业次生变量仍无直接接口[9];DeepMind GraphCast 则以 40 962 个mesh节点(六级二十面体精化网格)上的 GNN 消息传递解决拓扑不规则性问题[25]。2023—2024 年,国内 FuXi 与 FengWu 以级联3维 Transformer将稳定技巧延展至10~15天时效[10,26],但概率化与极端尾部表达仍偏弱。2025 年 GenCast 把条件扩散机制引入中期集合预报,概率技能超越 ECMWF ENS,显著提升了极端事件的尾部命中率[13]。2025 年,微软 Aurora 将基础模型范式扩展至大气-海洋-陆面-生态多圈层,展示了向“通用地球系统模型”演化的路径[11];剑桥-ECMWF 联合提出的 Aardvark 则摒弃逐时滚动推理,以单步端到端生成完整时序进一步压缩业务化推理成本[27]。“2015—2025年”这条演进主线提示:后续研究应从追求全球平均技巧转向田块级、作物级的精细化建模。
表1 所列9个代表模型揭示出3条耦合的驱动线:数据维度持续扩展,从 ConvLSTM 的单变量雷达场,到 Pangu-Weather 的 13 变量3维大气立体场,再到 Aurora 的大气-海洋-陆面-生态多圈层耦合,输入变量与时空维度跃迁推动了模型族从 CNN/RNN → Transformer → Foundation 的逐代演化。任务尺度持续下沉,全球模型从 0.25° 网格的中尺度替代模型逐步向 1 km 超分辨率的田块级业务基础架构演进,与第 3 节问题驱动的农业场景需求正面相连;不确定性显式化成为新主线。GenCast 的扩散集合显著提升了极端事件的概率尾部命中率,与第 2.5 节生成式架构对农业极端事件预警的分析直接对应。表1将与第 2.7 节的横向对比表形成“时间维”与“特征维”双视角,为后续选型建立完整坐标。
图3主要展示了全球气象大模型研究的三大核心阵营——北美、欧洲与中国,这些机构通过各具侧重的技术路径,共同构成了推动该领域技术演进的主要力量。
图4展示了2015—2025年气象大模型的演进主线,覆盖CNN/RNN、GNN、Transformer、生成式与基础模型五大范式,3条耦合驱动线(数据维度扩展、任务尺度下沉、不确定性显式化)贯穿其中,为后续各架构族的分节综述提供时序坐标。
深度学习在气象建模中的早期应用主要依赖2类基础架构:CNN[28]用于刻画空间局部相关性,RNN及其变体(LSTM、GRU)负责对时间动态进行建模。早期工作已将 CNN 用于 500 hPa 高度场预测,在粗分辨率下验证了端到端学习的可行性;Shi等人于 2015 年提出的 ConvLSTM[29] 进一步将卷积嵌入 LSTM 单元,实现对时空序列的联合建模,至今仍是雷达回波外推与短临降水预测的经典基线。
此后,PredRNN、MM-RNN、DeepESN 等结构在递归连接、记忆机制与多模态输入方面持续探索,尝试缓解长程依赖衰减与梯度消失等通病。尽管这些早期架构在理论复杂度与全球泛化能力上落后于后来的 Transformer/GNN,其轻量、可训练成本低、可与卫星/雷达观测自然耦合等特点,使其在面向农业气象站、灌区雷达网等“小样本+局地化”场景中仍具有不可替代的工程价值。
CNN/RNN 系架构计算门槛低、数据需求适中,已在县域/灌区一级的雷达回波短临预报、田块尺度温度/湿度时序预测以及与作物生长模型耦合的“轻量在线推理”链条中取得稳定可用的业务表现,尤其在小麦、水稻主产区的灌区分钟级短临任务上具备可直接落地的成熟度。针对农业气象站观测高缺失率的场景,基于深度学习的时序补全方法[30]为 CNN/RNN 类模型的训练数据预处理提供了可直接集成的通用工具。然而,CNN 的局部卷积核难以刻画跨大陆遥相关,RNN/LSTM 在多日跨度下误差累积明显,致使此类模型难以单独承担次季节及以上时效,以及跨气候带的迁移预报任务。此外,在“农业 Foundation Model 的预训练-微调范式”(见第 5.2 节)的视角下,CNN/RNN 主干可作为轻量下游适配器与田块 IoT、作物机理模型的融合层,与大规模 Transformer 基础模型形成互补的县域级业务部署方案。
GNN将气象要素的空间分布建模为图结构[31],其中节点表示空间网格点或观测站点,边表示空间相邻或物理耦合关系。相对于规整网格上的 CNN,GNN 天然适合不规则站点布局与多分辨率混合,这一性质对全球地形高度异质、观测密度极不均衡的气象观测网具有特殊意义[8]。
GraphCast 的处理器(processor)由 16 层未共享参数的消息传递 GNN 构成,作用于 6 级二十面体精化得到的 40962 个 mesh 节点(节点间平均间距约 110 km,覆盖 6 级 mesh 的 multi-mesh 边集)。每个 grid 节点(共 1038240 个)的输入特征为 474 维,由2(输入时刻)×(5(地表变量)+6(高空变量)×37(气压层))+20(静态)强迫场(土地海洋掩码、地形等)构成。第 $ l $ 层节点 $ i $ 的隐状态更新公式如式(8)所示,其中 $ {\boldsymbol e}_{ij} $为编码节点间相对位置向量。
$ h_{i}^{\left(l+1\right)}={{MLP}}_{v}\left(h_{i}^{\left(l\right)}\| \displaystyle\sum_{j\in \mathcal{N}(i)}{{MLP}}_{\boldsymbol e}\left(h_{i}^{\left(l\right)}\| h_{j}^{\left(l\right)}\| {\boldsymbol e}_{ij}\right)\right)\;。 $
模型每 6 小时推进一次,以 4 次递归完成 24 h 预报;训练采用课程学习,从 1 步到 12 步逐步扩展自回归地平线以缓解误差累积。基于GNN架构的代表性气象预测模型及其特征对比如表2所示。
GNN 在农业场景中的最大优势体现为“对不规则空间结构的天然亲和性”:GraphCast 与 GraphDOP 等工作已能把作物试验站、农田小气候站、商业雷达与卫星反演节点纳入统一图结构,在山地农田、丘陵茶园、戈壁绿洲等非规整空间中完成跨单元的消息传递[25,31]。然而,现有 GNN 气象模型对“作物—田块—县域—灌区”多层级嵌套拓扑的显式表达仍不充分,边的物理权重主要由距离或邻接经验给出,尚未引入大气动力学或水文传输过程的因果约束;且 GNN 推理对节点数敏感,当站点规模扩展到省级以上时存储与通信开销显著上升。此外,面向“气象大模型与作物模型的耦合”(见第 5.2 节)的下一步工作,可将 GNN 的多尺度图范式进一步扩展到“气象节点×作物群落节点”的异构图结构,为群落—田块—县域三级预报提供统一的形式化工具。
Transformer 凭借自注意力机制对全局长程依赖的高效建模能力,已成为气象大模型的主流架构,如表3所示。华为云团队提出的 Pangu-Weather 是这一方向的代表性工作[9]。其引入3维地球特定 Transformer,在 0.25°全球网格上实现了对位势高度、风速、气温、湿度等核心变量的端到端预报,在 1 h至 1 周时效内多项指标超越 ECMWF IFS。近期 Remote Sensing 上的独立第三方评估进一步证实了 Pangu-Weather 在站点级 2 m 气温与风速预报中的可用性[37]。Pangu-Weather 的 3D地球特定Transformer 采用分层 Swin 自注意力:输入为 13 个垂直层×0.25°网格的 4 个地表变量(2 m 气温、10 m 风速 u/v 分量、海平面气压)与 5 个高空变量(位势高度、比湿、温度、风速 u/v 分量),经 3D Patch Embedding(patch size = 2×4×4)后映射为 $ C $ 维 token 序列,随后通过“8 层 Encoder + 8 层 Decoder”架构(共 16 个 Earth-Specific Block)处理。其核心创新 Earth-Specific Positional Bias 为每个经度-纬度-层级组合学习独立的位置偏置 $ B $,以显式编码球面几何与垂直分层的非均匀性;这一设计使同一模型在全球任意子域上的精度保持稳定。训练使用 ERA5 数据(1979—2017年)按多种 Lead Time 划分,损失函数为纬度加权 MSE,总参数约 2.56×108(256M)。推理时以 1 h、3 h、6 h、24 h四个模型级联生成 1~168 h 预报,单 GPU 单步推理约 1.4 s,整体吞吐较 IFS 快约 104 倍。
国内研究在该方向亦取得快速进展。由复旦大学人工智能创新与产业研究院推出的FuXi系列模型采用多层级3维Transformer架构,通过级联推理将稳定预报时效延伸至15天中期尺度[10];FuXi-Weather 进一步整合资料同化—预报全链条,实现“数据到预报”的端到端机器学习系统[15]。FuXi 系列的关键设计是 U-Transformer 级联架构:训练3个独立模型 FuXi-Short(0~5 d)、FuXi-Medium(5~10 d)、FuXi-Long(10~15 d),分别在对应时效上优化;其骨干由 48 个 Swin Transformer V2 块构成 U-Transformer,时间步长为6 h,处理 70 个气象变量(5(地表)+5(高空)×13(气压层))。推理时按目标 Lead Time 选取对应模型,有效缓解单模型在不同时效上的精度—稳定性冲突。FuXi-Weather 在此基础上引入可微分 4D-Var 同化模块,实现观测到预报的端到端训练,使整体系统在无需外部资料同化的条件下仍能达到 ECMWF 业务水平。上海人工智能实验室研发的 FengWu 系列在多变量集合预报与极端事件检索方面亦表现突出。
Aurora 的基础模型设计包含3个要点。1) Perceiver 编/解码器:对不同空间分辨率(0.1°~1°)与采样频率(小时~日)的异构输入,由Perceiver编码器统一压缩到固定潜在空间,对应的 Perceiver 解码器再将潜在状态还原到目标变量与分辨率。2) 3D Swin Transformer 处理器:在统一潜在空间上用一个 3D Swin Transformer 主干推进时间演化(输入为 $ t-1 $与 $ t $时刻表示,输出 $ t+1 $时刻),结合 LoRA 实现高效微调。3) 大规模多源预训练:在超过 100 万小时的 ERA5、CMIP6、CAMS、GLORYS 等多源地球系统数据上联合预训练。多任务联合损失形式如式(9)所示,其中 $ {L}_{k} $ 为第$ k $个预训练子任务的损失、$ {\lambda }_{k} $ 为对应权重。
$ L=\displaystyle\sum_{k}{\lambda }_{k}\cdot {L}_{k}\;。 $
Aurora 训练目标为 6 h Lead-Time 的 MAE,在 32 张 A100 GPU 上完成约 2.5 周共150000步。在下游空气质量、海浪、台风路径、0.1° 高分辨率天气等微调任务中,Aurora 仅需 LoRA 模块即可达到或超越专用业务系统的精度。
Transformer 的全局自注意力机制与农业对“作物全生育期长程依赖”的天然需求高度契合:Pangu-Weather、FuXi、Aardvark、Aurora 等模型在 1~15 天乃至次季节时效上已展现出对累积温度、累积辐射等关键农艺时序的可用刻画能力[9-11,27]。然而,训练与微调成本对县域与合作社一级机构构成显著门槛,且现有模型对冠层温度、叶面湿润时间等次生变量尚缺乏原生输出接口,仍依赖后处理或作物模型二次推演,进而引入额外不确定性。此外,在“农业 Foundation Model 的预训练-微调范式”(见第 5.2 节)指向下,有必要基于现有全球气象 Transformer 主干开展面向主粮作物关键生育期的轻量微调、参数高效化与多任务联合训练,以在保持通用气象技巧的同时将模型边际收益有效转化为农业决策价值。
随着气象大模型在分辨率、时序跨度与推理效率上的需求不断提升,传统 Transformer 在长时间序列任务中面临 $ O ({n}^{2} ) $计算复杂度的瓶颈。状态空间模型(state space model,SSM)及其代表性实现 Mamba,通过隐式状态更新机制以接近线性的复杂度完成长距离时序建模,在效率与建模长度上提供了新的可能。
Yang 等人[38]针对全球气象站点预测任务提出的 Mamba-based框架展现了 SSM 在站点级时序中的潜力;MetMamba[39]在区域气象建模中通过区域—全球协同训练初步实现了多尺度耦合;ClimODE[40]、WeatherODE[41]等基于神经常微分方程(Neural ODE)的工作,将 SSM 的连续时间形式与气候模拟相结合,为长时序与稀疏观测下的气象建模提供了新工具。尽管目前 SSM/Mamba 系列工作多处于实验阶段,但其轻量性与稳定性使其有望在未来与基础模型范式互补。
SSM/Mamba 类架构在农业场景中的核心价值体现为长时序建模与低成本2方面:已有工作在全球站点级长时序预测、区域-全球协同建模以及基于 Neural ODE 的连续时间气象建模中初步展示出以接近线性复杂度处理季节乃至年度跨度序列的潜力[42]。然而,当前 SSM 气象模型的外推稳定性、对极端分布尾部的刻画能力以及与高分辨率空间表达(如球面 Transformer)的融合策略仍有待系统研究,尤其在冬季长期低温累积、夏季高温热浪连发等“长时序+极端”复合场景中样本稀缺问题突出。此外,“田块尺度极端事件临近预警”(见第 5.2 节)的落地路径,SSM 可作为低成本、可在县域硬件上长期运行的时序基底,与扩散类生成式模型在极端事件分支上联合使用,形成“长时序主干+事件概率分支”的轻量业务架构。
生成式模型(generative models)聚焦于不确定性表达、高分辨率降水建模与极端事件模拟。相比传统神经网络的单点预报输出,生成式架构通过对预报场的概率分布建模,能够提供与物理一致的“集合样本”,为下游决策提供概率化输入。
早期工作多基于生成对抗网络(GAN)与变分自编码器(VAE),例如DeepMind 的 DGMR 雷达生成模型[43]是临近降水预测的经典工作,AIES 2024 上发表的 CGAN 集合生成式严重天气预报[44]进一步将概率化扩展到高影响事件。近年来,扩散模型(diffusion models)逐步成为主流:GenCast 将条件扩散机制引入中期集合预报,在 ERA5 数据集上多项核心变量的概率技能超越 ECMWF ENS。
GenCast 以图神经网络为去噪主干,将条件分布 $ p_\theta (\boldsymbol X_{(t+1)} |\boldsymbol X_t) $建模为对 $ T=20 $步扩散隐变量 $ {z}_{1:T} $的边际化积分,其反向扩散步采样过程如式(10)所示。$ s $ 为扩散步索引、$ \boldsymbol \varepsilon \sim N\left(0,I\right) $、$ {\sigma }_{s} $ 为该步噪声尺度。
$ \boldsymbol X_{t+1}^{\left(s-1\right)}={\boldsymbol \mu }_{\theta }\left(\boldsymbol X_{t+1}^{\left(s\right)},{\boldsymbol X}_{t},s\right)+{\sigma }_{s}\cdot \boldsymbol \varepsilon\;。 $
训练阶段,随机采样扩散步$ s $并优化噪声预测损失,目标如式(11)所示。
$ {{L}}_{\text{diff}}={{\mathbb E}}_{s,\boldsymbol \varepsilon ,\left({\boldsymbol X}_{t},{\boldsymbol X}_{t+1}\right)}{\|{\boldsymbol \varepsilon }_{\theta }\left(\boldsymbol X_{t+1}^{\left(s\right)},{\boldsymbol X}_{t},s\right)-\boldsymbol \varepsilon \|}^{2}\;。 $
模型由此学到从高噪声到清洁预报的去噪轨迹;单次采样生成1个集合成员,50次独立采样构建50个成员集合。相比传统 EPS(ensemble prediction system)的随机扰动法,GenCast 直接学习后验分布的形状,在台风路径与极端降水命中率上显著优于 ECMWF ENS;Neurocomputing 2026年发表的CastDiffuser 通过级联潜在扩散在短临降水任务上实现高分辨率与高稳定性的平衡[14]。表4 汇总了主要生成式与算子类模型。
生成式模型在农业场景中的独特价值在于“对极端天气集合预报的概率化能力”:GenCast、CastDiffuser、CGAN 集合等工作已能为“未来 7 天内≥$ X $ mm 单点降水概率”“花期连续 3 日≥35 ℃概率”等条件事件提供与物理一致的集合样本,直接服务于农业风险定价、保险产品设计与灌溉/施药决策[13,26,46]。然而,扩散与 GAN 类模型推理成本仍偏高,面向乡镇与合作社一级的实时部署需要进一步的轻量化与蒸馏研究;且现有概率输出主要针对通用气象变量,对“小麦越冬期日最低气温≤−15 ℃ 持续 24 h”等作物-气候耦合阈值事件的概率校准仍显不足。此外,在“田块尺度极端事件临近预警”(见第 5.2 节)方向上,生成式模型可与 1 km 超分辨率框架、物理引导子模块联合构建集合化、物理一致的临近预警系统,是该子方向最具落地可能的候选架构。
随着深度学习在气象建模中的广泛应用,研究者越来越重视模型的物理一致性与跨模态泛化能力。物理引导神经网络(physics-informed neural network,PINN)与多模态融合(multi-modal fusion)逐步成为支撑下一代气象大模型的两大方法论支柱。
物理引导方面的核心目标是将守恒律、动力方程等显式物理知识嵌入神经网络结构或损失函数,增强模型对自然规律的刻画能力。WeatherGFT、NeuralGCM[47]等工作通过将连续动力方程作为软约束嵌入端到端训练,在中期预报与气候模拟之间架起桥梁;NPJ Climate 2024 报道的混合物理-AI短临降水模型证明了“物理子模块 + AI 校正”对极端降水任务的有效性[48]。
在多模态融合方面,随着卫星、雷达、地面站、海洋浮标等多源数据持续积累,统一融合架构成为提升模型精度与鲁棒性的关键。代表性工作包括用于台风的 Hurricast[49]、PWV 检索的气象基础模型[50],以及面向 1 km 超分辨率的 npj Climate 2026 通用框架[51],后者为后续农业田块尺度精细化预报提供了重要的方法论输入。
物理引导与多模态融合是“基础模型+下游农业任务”范式落地的两大关键支点:物理引导(WeatherGFT、NeuralGCM、物理+AI 短临降水模型)确保气象大模型在农业气象站与田块小气候等数据稀疏场景中保持能量与水量守恒,多模态融合(Hurricast、PWV 基础模型、1 km 超分辨率通用框架)则让模型同时消化卫星遥感、地面观测、农气试验等异质数据[47-51]。然而,当前物理约束多以软约束的损失项引入,与动力方程的严格闭合仍有差距;多模态融合亦以拼接与晚融合为主,对冠层-土壤-大气耦合界面的机理性刻画不足。此外,面向农业 Foundation Model 的预训练-微调范式(见第 5.2 节)与气象大模型与作物模型的耦合(见第 5.2 节)两个方向,2类方法的进展有望显著降低“通用气象大模型→农业专用模型”的微调门槛,并为可微作物模型的联合训练提供可导的物理-数据桥梁。
为帮助读者在面向具体农业问题进行模型选型时具备直观的全局视图,本节对前述6类主流架构进行横向对比,维度涵盖计算复杂度、数据量需求、长程依赖能力、生成式(概率化)能力、物理一致性以及在农业场景中的适配度,详见表5。
从表5 可以看出,主流架构在通用气象指标上各有优势,但在农业场景的适配度并非简单地与计算成本或参数规模成正比。具体而言:CNN/RNN/LSTM 类架构以“轻量、可解释、可与作物模型耦合”取胜,更适合县域与田块级的业务化部署;GNN 在“不规则站点—灌区—行政单元混合空间”上具有结构性优势,是中国大农业中“多尺度合作社”业务的天然候选;Transformer 系基础模型在全季气象演变与多变量耦合方面展现出强大潜力,但训练与推理成本对县域级机构构成显著门槛;SSM/Mamba 在“长时序+低成本”双重维度上具有未来空间;Diffusion 与 GAN 的概率化能力对极端事件—保险—金融链条不可替代;物理引导与多模态融合则提供了“基础模型→农业专用模型”的桥梁。在面向具体农业问题进行选型时,决策者应基于“问题—架构—预算—部署”四维进行权衡,而非简单追求“最强模型”。
表5给出了主流架构在通用维度上的定量画像,但农业应用对模型的期望往往是多约束耦合的——既要田块级精度,又要长时序跨生育期,还要在多源观测稀疏的局地环境中稳定推理。本节将第 1.2 节列出的 4 类农业特异性短板(分辨率失配/变量不匹配/极端命中率/决策链漏损)与第 2.1~2.6 节的 6 类架构两两交叉,得到一张兼容性矩阵(表6),为第 3节的问题驱动选型提供逆向参考。
表6 可得到5个观察结论:1)没有一种架构在 4 列上全部取得“强”,印证农业应用需要“多架构混合”;2)Transformer 在长程依赖列表现最均衡但代价是显存与训练成本;3)生成式架构在极端事件列明显优于其他,是霜冻、热浪、冰雹等关键农业气象事件预警的第一候选;4)物理引导在变量匹配上强但复杂度阻碍大规模部署,更适合田块小气候与作物模型耦合的深度任务;5)从“综合推荐场景”列看,问题选型的路径应是:先按第1.2节的 4 类短板定位,再在表6 对应列找“强/中—强”候选,最后用表5 横向对比做最终权衡——3张表即构成“短板诊断→架构候选→定量权衡”的完整选型闭环。
在第2节按架构梳理了主流气象大模型之后,本节转入“农业应用”维度,采用“问题驱动”的组织方式:从农业生产中的若干关键决策需求出发,逆向讨论适用的气象大模型与方法选型,并尽可能将气象精度的提升量化转化为农业 KPI(节水率、产量增益、灾损减少率、净收益等)。这一框架与第2节末尾的“架构横向对比表”形成呼应。
本节在3.1~3.5节按5类农业问题展开,每节首先界定农业需求与决策时效,其次讨论与该需求最匹配的气象大模型与气象指标,最后以已发表案例的定量数字呈现“气象精度→农业效益”的转化效率。
图5系统展示了气象精度改进通过转化链条最终反映在节水、增产、减损等农业 KPI 上的逻辑路径,并为文中讨论的五大核心应用场景提供了定量效益标定基线。
灌溉决策面临“何时灌、灌多少”2个基本问题。其有效性取决于未来 7~30 天的累积降水、参考蒸散(ETo)与土壤含水率三者的联合预测,决策时效从天级(临时调度)跨到月级(整季水量调配)。相应地,所需气象产品需覆盖中期(7~15天)与次季节—季节(15~120天)两个时效,空间分辨率需达到田块或灌区级别(1~10 km),并以概率化输出为佳,以适配灌溉决策对极端干旱与连续降雨概率的双向需求[52]。
在中期到次季节时效,基于 Transformer 的 FuXi-S2S[53]、ClimaX 与基于扩散的 GenCast 等模型已展现出超越传统数值预报的概率化能力,其集合输出可直接接入灌溉调度决策模型。在田块尺度,“AgERA5 + LMM 输出”的混合方案进一步配合 1 km 尺度超分辨率框架(Park等人[51]提出的 DL 超分辨率通用方法),将灌溉建议从灌区平均细化到地块级别。
定量证据方面,地中海地区针对硬质小麦与玉米的研究表明,原始季节性气象预报(SWF)在 ETo 上的 MAE 为 17.0~170.3 mm,在降水上的 MAE 为 94.8~438.6 mm,其中 ETo 通常被高估约 20%、累积降水被低估可达 90%[54]。在偏差订正与气象大模型联合后处理之后,该研究估算可使灌溉用水量预测的方向性正确率从 50% 提升到 70% 以上,为节水 5%~15% 提供了上限空间。湄公河三角洲水稻系统的经济学评估则进一步量化:将季节降水预报技巧从 70% 提升至 100%,可使最优播期决策的农户净收益由 135 美元/公顷上升至 220 美元/公顷(净增 85 美元/公顷)[21]。这些数字为“气象精度→灌溉节水/收益”链条提供了具体的标定基线。将上述定量结果映射到主产区业务链条:地中海硬质小麦与玉米片区中,ETo 预报 MAE 从 170 mm 量级压缩至偏差订正后的可接受水平,可使得灌溉决策前置窗口延长数日,经折算可实现灌区平均节水 5%~15%(对应万亩级灌区年均节水量约数十万立方米、电耗节约数万千瓦时量级)[51];湄公河三角洲水稻系统中,按线性外推可估计预报技巧每提升 10 个百分点对应农户净收益约增加 30 美元/公顷(即上述 70%→100% 全档共增加85 美元/公顷的逐档边际折算),按主产县规模放大后具备可观的保险与金融产品空间[21]。
极端天气事件对农业的破坏往往在数小时内发生,却可能造成整个生长季的颗粒无收。霜冻、春寒、热浪、骤旱、强对流降水与冰雹是中国农业最常见的“高影响、低概率”事件,其预警的核心挑战在于:事件位于气候态分布的尾部、具有强烈的局地化与窗口化特征。
从模型选型看,生成式与扩散类模型(GenCast、CastDiffuser、CGAN 集合)是当前在概率化极端事件预测上最有竞争力的架构[14,44],其输出可天然形成“集合样本”以驱动农业决策的概率风险评估。GraphCast 在 2023 年飓风 Lee 的预测中提前 9 天准确预测了 Lee 在加拿大新斯科舍省的登陆位置,为沿海农业灾前转移与设施加固赢得了实质性提前期[25]。在区域强对流尺度,“物理引导+AI”混合模型(如 NPJ Climate 2024 的 Hybrid 短临降水)结合了物理子模块与生成式扩散,已在多次极端降水个例中表现出对落区与强度结构的可靠预报能力[48]。
定量评估方面,澳大利亚小麦产区基于 ACCESS-S2 季节预报与 APSIM-Wheat 模型联合应用研究表明,使用季节预报支持播期决策可使 68% 案例获得产量增益,6 个东部产区平均产量增益达 281 kg/ha(约增加13%);在干旱年份,采用季节预报的策略表现尤为突出,产量增益高于湿润年份与中性年份[55]。这一结果直接量化了“极端事件预测+作物模型”链条的农业 KPI 转化效率。进一步将该结果解读到业务链条:6 个东部产区以 11 年(2010—2020年)×多县站样本规模计,281 kg/ha 的平均增益经折算对应区域年度总产约多出数万至数十万吨,且在干旱年份优势更显著,直接支撑了粮食战略储备与播期保险2类业务决策;配合 GenCast 等概率化预警,霜冻/热浪窗口的预防性作业调度(果园熏烟、设施防风)可将单次事件的灾损避免率提升到“集合置信度可追踪”水平[13,55]。
作物病虫害风险与气象条件高度耦合:小麦条锈病、稻瘟病、棉铃虫、玉米螟、草地贪夜蛾等典型病虫害的发生与流行,依赖于湿度、气温、叶面湿润时间、风速等多个气象变量在特定窗口内的组合。相应的预测任务对未来 3~15 天的高分辨率多变量预报有强需求,尤其是叶片小气候(冠层温度、叶面湿度)的精细化输出。
从方法上,此类风险预测通常采用“气象输入→统计/机器学习风险模型→空间风险地图”两阶段范式。气象大模型主要承担第1阶段:提供多变量、高时空分辨率的预报输入。FuXi、FengWu、AIFS 等中期预报模型经下游降尺度后,可为县域至灌区一级的病虫害风险模型提供 1~7 天的稳定输入;风险模型本身则多为逻辑回归、梯度提升树、随机森林等可解释机器学习方法,便于植保部门按“高、中、低”三级生成预警地图。
在定量转化层面,已有研究将气象预报与作物水分胁迫模型(CWSM)联合,量化了集合预报与单值预报在水分胁迫预测上的差异;基于9个葡萄园的对比表明,采用“集合气象预报+后处理”可使水分胁迫预测误差较确定性预报降低 20%~30%,由此提高病害风险评估的稳定性[52]。在中国实践层面,基于气象大模型与历史病虫害发生记录构建县域级风险地图,有望在小麦赤霉病、稻飞虱与草地贪夜蛾等防控中提供提前预警支持。从业务链条视角解读:9 个葡萄园对比实验中,采用“集合气象预报+后处理”使水分胁迫预测误差较确定性预报降低 20%~30%,对应病害风险等级的误判率测算下降约 1/4,在县域植保机调度层面可减少无效施药次数并降低农药残留风险[52];与此同时,基于气象大模型与历史病虫害记录构建的县域风险地图,已为主产区的分级预警与防控资源调度提供了可直接接入现有植保业务流程的输入[48]。
产量预测是从国家粮食安全到地方种植结构调整的核心决策依据。其时效跨度大(从生育早期到收获前)、空间尺度多样(从全国到田块),对应的气象输入需求也最为综合:既要全季节多变量序列(温度、降水、辐射、蒸散),又要关键期窗口的极端事件概率,还需将气象输入与卫星遥感、土壤性质、管理记录等多模态信息融合。
从模型选型看,产量预测本质上是“多模态融合+时序建模”任务。近年来的代表性工作可分为3类:1)“气象+作物机理”模型:“ACCESS-S2 + APSIM-Wheat”“ECMWF SEAS5 + WOFOST”等系统耦合气象大模型提供的气象输入与作物模型生育-产量过程,在区域产量预测上具有较强物理可解释性[22,55];2)“气象+遥感+ML”融合模型:将气象大模型输出与 NDVI、EVI、叶面积指数等遥感变量融合,通过 CNN-LSTM、Transformer 或图神经网络完成产量回归;已有针对小麦的多源融合模型 $ {R}^{2} $ 达到 0.4~0.88,显著优于单一数据源方案;3) 基础模型微调:Aurora、TimeGPT[56]等地球系统/时序基础模型的引入,为下游产量预测提供了轻量微调路径[11]。上述不同模型家族的产量预测 $ {R}^{2} $ 差异(0.4~0.88)映射到业务链条,对应“县域产量月更新发布误差”从 15% 以上压缩到 5%~8% 的量级;配合气象-作物—遥感多模态融合,现有研究已在小麦、水稻主产区实现了生育早期 30~60天的产量趋势方向性判断,直接支撑国家级粮食安全决策与省级种植结构调整[22,55]。
在尺度问题上,田块级产量预测的难点在于全球气象大模型的 0.25°(约 25 km)分辨率与农场管理单元(30~500 m)之间的2~3个数量级差距。解决路径主要有3个:高分辨率超分辨率模型(Park 等人[51]提出的DL超分辨率通用框架可下沉至1 km)、区域 WRF/AROME 与气象大模型嵌套(如 Pangu-Weather 驱动的 WRF 区域系统已在台风路径案例中显著优于经典 GFS-WRF),以及面向特定作物-区域的微调基础模型。需要特别说明的是,FourCastNet 本身在 0.25° 网格上完成全球预报,可作为区域降尺度的高质量背景场输入,但其本身并非超分辨率模型[12]。
精准农业的核心理念是“按需、按时、按地”实施差异化作业管理。喷药、施肥、抢收抢种、机械作业等环节均对气象条件高度敏感:风速过大会导致药液漂移与肥料损失,突发降雨会冲刷叶面药效,抢收过程对降水间歇期的把握直接决定了产量变现。这一类任务对气象输入的需求是“短时(0~24 h)、高分辨率(1 km 以下)、多变量(风速、降水、湿度、温度)、高频次(小时级或更细) ”。
在模型选型上,NowcastNet、DGMR、CastDiffuser、MetNet-3 等短临生成式与高分辨率模型是核心候选[14,44],其分钟级、公里级输出可直接对接田间作业窗口决策。TimeGPT 等时间序列基础模型在土壤水分等慢变量的零样本预测中也展现出独特价值,为低频信息融合提供了通用工具。
在定量评估层面,韩国针对农业用途的高分辨率风速预报系统在 100 m×100 m 空间分辨率下实现 3 m 高度风速预报,基于随机森林后处理的 48 小时,其RMSE 约 0.8 m/s、MAE 约 0.5 m/s,可有效支持田间机械作业与喷药决策[57]。在中国实践中,将气象大模型短临输出与县域植保机调度结合,可对喷药窗口给出精细化提示,有望减少药液漂移损失。将该类定量结果解读到业务链条:100 m×100 m 分辨率、48 小时 RMSE 约 0.8 m/s 的风速预报,足以把田间喷药窗口的作业决策提前至班次级调度,按万亩规模测算可减少无效作业与药液漂移损失各数个百分点;短临生成式模型与县域植保机调度联动后,抢收抢种窗口的把握进一步获得“分钟级更新、公里级落区”的新输入,在试点区已观察到减少因误判窗口造成的农机空驶与抢收延误[46,57]。
上述5类应用对气象输入的空间尺度要求差异显著:灌溉与产量预测可在 1~10 km 量级运行,极端事件预警与精准作业调度则需要数百米至 1 km 的分辨率;而主流全球气象大模型(GraphCast、Pangu-Weather、FuXi)的原生分辨率多在 0.25°(约 25 km),弥合这2~3个数量级的“尺度鸿沟”是气象大模型走向农业可用工具的核心工程问题。综合3.1~3.5节的讨论,主要工程路径包括深度学习超分辨率(可下沉至 1 km)[49]、全球 LMM 驱动区域 NWP(WRF/AROME)嵌套,以及面向作物-区域的基础模型轻量微调。这3条路径并非互斥而是互补,面向不同农业问题的最优组合取决于具体的时效、精度与成本约束[58-59]。
为便于读者横向比较前述5类应用中的典型方案及其农业 KPI 转化数据,表7给出了气象大模型在农业应用中的核心案例汇总。
现代智慧农业的发展建立在对海量、多源数据的有效利用之上。高质量的气象数据集与科学的评估指标体系,在模型训练、诊断、长势监测、产量预测和自动化管理等任务中发挥关键作用。本节首先梳理气象大模型领域常用的通用评估指标与数据资源;随后引入面向农业的导向性评估视角,补充关键农事期预报技巧、极端农业气象事件命中率、与产量相关性等农业 KPI 指标,以及面向农业的专项数据集,为后续第5节对未来研究方向的讨论奠定方法论基础。
图6提出面向农业的统一评估协议,涵盖三层数据协同(three-tier data integration)、通用与农业双轨指标(dual-track metrics)、四时效分层评估(multi-horizon evaluation)及严格时间锁划分(time-locked train/test split)四个维度,为跨机构、跨模型的横向比较提供标准化框架。
评估指标是衡量预测模型性能和比较不同方法的核心工具。气象预测领域的常用定量指标可大致分为5类。
1) 误差类指标。RMSE衡量预报场与真实场之差的均方根,反映整体误差大小,RMSE 越小越好。数值天气预报中,500 hPa 高度场 5 天 RMSE 是评估全球中期预报技巧的常用指标。MAE取绝对值平均,可减小大误差的过大影响,对均值偏差较为稳健。
2) 相关性指标。相关系数(CORR)衡量预报与实况之间的线性相关性;ACC用于评价高度场形态相似度,与 RMSE 并列作为中期预报技巧分数(skill score)的两大核心指标。
3) 概率性指标。针对集合预报与概率化输出,常用连续排序概率得分(CRPS)、Brier Score、可靠性图(reliability diagram)与展度-技能比(spread-skill)等;CRPS 综合反映了预测分布与观测之间的距离,是当前气象大模型概率化能力评估的事实标准[19-20]。
4) 分类与分位指标。对降水、雷暴、强对流等事件类预报,常用命中率(POD)、虚警率(FAR)、公平技巧评分(ETS)以及HSS(Heidke skill score)进行评估。在极端事件评估场景下,基于广义极值分布(GEV)的超阈值统计与重现期分析提供了更稳健的尾部度量[23]。
5) 空间结构指标。FSS(fractions skill score)评估降水落区匹配度,结构-振幅-位置(SAL)度量预报降水场的整体结构误差,近期 QJRMS 2025 提出的 PAD 空间检验进一步丰富了“分布近似”维度的评估工具[60]。这类指标特别适合于面向降水临近预报与高分辨率场预报的评估。此外,对深度学习模型还需关注推理时延与计算成本,例如 FourCastNet 相对于 IFS 的提速可达 4.5×104倍,这类“效率指标”对面向农业的高频次业务化部署尤为关键。
指标的选择需针对具体任务进行。全球中期预报通常以 500 hPa 高度场 RMSE/ACC 为核心指标;降水定量预报强调 POD/FAR/ETS;台风路径预报采用平均距离误差与登陆点偏差;集合预报则需要 CRPS、可靠性图与展度-技能比联合评估。在引用文献时常出现“模型 $ A $ 在 500 hPa 上 RMSE 比模型 $ B $ 低 X%”的表述,其严格意义需结合具体年份、区域与基线参考再做横向对比。
4.1节中通用气象指标对气象预报系统的整体技巧给出了可比较的量化标尺,却未能直接反映“对农业决策的可用性”。本节在通用指标体系基础上,提出4类面向农业应用的导向性评估指标,与第3节“问题驱动”的应用框架形成对照。
1) 关键农事期预报技巧。不同作物在生育周期内存在若干“关键窗口”——小麦的抽穗扬花期、水稻的孕穗期、玉米的乳熟期等阶段对气象极端事件高度敏感。建议在通用 RMSE/ACC 基础上,对关键农事期内的相关变量(气温、降水、风速、湿度)单独计算技巧得分,形成“窗口化技巧”评估;例如“花期连续3日≥35 ℃ 概率”的预报技巧,可同时通过 Brier Score 与可靠性图评估[48]。
2) 极端农业气象事件命中率。霜冻、春寒、热浪、连阴雨、骤旱等是直接造成农业灾损的关键事件。建议以“事件-窗口-强度”3维定义典型事件:事件维度采用作物-气候耦合阈值(如冬小麦越冬期日最低气温≤−15 ℃ 持续 24 h);窗口维度对应作物物候关键期;强度维度采用事件强度分级。基于此,使用 POD/FAR/ETS及条件偏差(conditional bias)对预报的尾部命中率进行评估[48],可有效弥补 RMSE 类指标对极端事件的“麻木”。
3) 与产量(及品质)相关性。将气象预报或再分析驱动的作物模型输出与实测产量(或品质)进行回归分析,可计算 $ {R}^{2} $、Nash-Sutcliffe 效率(NSE)及偏差-方差分解,由此把“气象预报技巧”转化为“农业产出可解释方差”。该指标在县域、灌区、田块多个尺度上的差异性,为模型尺度适配提供了直接证据。
4) 预报价值的经济转化。使用决策-响应模型(如 CRP、灌溉调度模型)将气象预报作为输入,量化最优决策与气候态决策之间的农户净收益差异,即“气象预报的经济价值”(value of information)。湄公河三角洲水稻系统的实证研究表明,将季节降水预报技巧从 70% 提升到 100%,可使最优播期决策的净收益由 135 美元/公顷上升至 220 美元/公顷(净增 85 美元/公顷)[21];该类研究为“通用气象指标→农业 KPI”的换算提供了少有的标定数据。
深度学习在气象建模中的快速发展,依赖于高质量、多源、多尺度的气象数据资源。当前常用数据可分为再分析与数值模式数据、观测类数据、标准基准数据集及多模态/专项任务数据四大类。再分析数据如 ECMWF 的 ERA5 提供了 1940 年以来全球小时级、0.25°分辨率、近 240 个变量的高质量大气—海洋—陆面再分析序列,是训练全球气象大模型(Pangu-Weather、GraphCast、FuXi 等)最常用的“通用底座”;CMIP6 多模式集合则为长期气候演变与情景模拟提供基础。
在观测数据方面,雷达与卫星资料为短时降水、对流系统与极端天气建模提供了关键支持。SEVIR 数据集整合了美国本土的雷达、卫星与地面观测图像,广泛用于发展 Nowcasting 模型与多模态学习,以及TempestExtremes和TropycalDB 等专项数据集则面向台风、气旋等高影响事件的检索与评估。
标准基准数据集方面,WeatherBench[61] 与 WeatherBench 2[62]为深度学习气象预测提供了统一评估平台;ChaosBench 在 S2S 时效的多变量、物理一致性维度上提出了基于物理的基准[63];PAD 空间检验框架进一步丰富了空间结构维度的评估工具[60]。总体而言,上述数据资源为大样本、可重复、可对比的气象大模型研究奠定了基础。
通用气象数据集虽然为模型训练与基准评估提供了基础,但其变量、分辨率与采样频率往往未直接对接农业决策需求。在面向农业应用的研究中,建议补充4类专项数据资源。
1) AgERA5 与 AgMERRA。AgERA5 由 ECMWF 在 ERA5 基础上为农业应用专门重新插值与降尺度,提供 0.1°分辨率、日尺度的近地面气温、降水、辐射、相对湿度、风速及参考蒸散等农气直接相关变量;AgMERRA 是 NASA 在 MERRA 再分析基础上面向农业研究构建的姊妹数据集。二者已成为全球作物模型(APSIM、WOFOST、DSSAT)区域评估的事实输入标准[6,22]。
2) 作物试验站与作物气候网观测。中国、美国、欧盟与澳大利亚均维护了规模不一的作物气候联网观测系统(如中国农业气象观测站网、AgMIP 全球作物模型对比联盟提供的多站点数据)。这类数据涵盖物候期、产量、生育阶段田间小气候等“农业 KPI”维度,是评估“气象预报→农业效益”转化率的关键真值。
3) 田块尺度无人机/物联网数据。近年来兴起的农业物联网与无人机遥感为田块级气象与生长状态提供了高密度观测。NDVI、冠层温度、土壤含水率、作物高度等指标可在 30~500 m 尺度上与气象大模型的高分辨率输出做联合验证[51]。此类数据多以省级或试点为单位发布,系统性整合仍处于起步阶段。
4) 农业经济与决策数据。在“预报价值经济转化”维度上,需配套农作物产量统计、灾损保险数据与农户决策行为数据,才能完成“气象指标→农业 KPI→经济价值”的完整链条评估。已有研究将气象预报与灌溉决策模型联合,量化出每提升10%季节降水预报技巧可减少 1~5% 灌溉用水的关系[21,54],这类工作需要的多源决策数据仍是当前农业气象服务体系的薄弱环节。
为推动气象大模型在农业领域的横向对比与可复现研究,建议后续工作采用4点统一评估协议。1)数据集三层协同:以 AgERA5 或 AgMERRA 作为区域再分析底座,与中国农业气象观测站网、AgMIP 等作物气候试验站观测以及田块级 IoT/无人机观测数据共同构成“再分析—试验站—田块”三层数据协同,避免评估退化为“比谁的再分析更好”[6,22,51]。2) 指标两类联合:在同一套实验中同时报告通用气象指标(RMSE、ACC、CRPS、FSS)与农业导向指标(关键农事期窗口技巧、极端事件 ETS/POD/FAR、县域产量 $ {R}^{2} $、预报经济价值),把“气象指标→农业 KPI→经济价值”的完整链条纳入评测[21,23]。3) 任务分层对标:按临近(0~6 h)、短期(1~7 d)、次季节(7~90 d)、季节(3~9个月)四个时效分层评估,每个时效对接各自适用的农业决策(作业窗口、病虫害、灌溉、战略规划)。4) 开源与时间锁要求:复现代码、模型权重与数据分配方式公开,训练集与测试集按严格的时间锁划分(如2020年前为训练集,2022年后为测试集),以避免数据泄露并为不同机构方法的横向对比提供共同语言。
在前述4节的基础上,本节总结气象大模型在农业应用中面临的关键挑战,并就未来的研究方向给出展望。与前期通用气象综述不同,本节特别强调“农业导向”,即从作物—田块—决策—效益完整链条出发,识别气象大模型尚未跨越的几个工程性与方法论瓶颈,并在 5.2 节给出面向农业的具体研究方向,替代以往综述中常见的航空、能源等通用领域举例。如图7主要从时间维度梳理了气象大模型农业应用的5条未来研究方向,为后续各节的讨论提供整体框架。
本节从数据与计算、不确定性量化、极端事件、物理一致性、业务部署与标准评测6个维度综合识别当前气象大模型在农业应用中尚未跨越的关键瓶颈。
训练气象大模型通常需要处理几十年全球数据(数百 TB 规模)与大量计算资源。在向农业应用迁移时,此瓶颈被进一步放大:面向特定作物-特定区域的微调需要田块级真值,而此类真值的获取成本高、空间稀疏、时间不连续。近年来 ClimaX(一种基于视觉 Transformer(ViT)[64]的气候-气象基础模型)等工作的“在气候长期数据上预训练、再微调特定任务”路径,为缓解农业数据稀缺提供了通用框架[11,34]。与此同时,面向农业的轻量化(SSM/Mamba 类)、联邦学习、自监督预训练等方向值得深入探索。此外,针对观测稀疏带来的训练数据质量瓶颈,深度学习时序补全方法[30]可在预处理阶段缓解观测缺失的影响;在模型侧,大模型混合专家系统的优化综述[65]为降低“Transformer+MoE”架构的推理成本提供了系统化路径。
尽管生成式与扩散类模型已能提供概率化预报,如何全面、直观地把不确定性传达给农业决策者仍是难题。与航空、新能源等行业不同,农业决策具有窗口窄和回报曲线非线性2个特征:例如作物关键生育期的热累积预报,少量极端日数即可显著改变产量结果;又如霜冻窗口期的极端概率预测,即使 20% 的概率提示也可能触发整片果园的防护行动。面向这类应用,应发展将“窗口化条件概率”作为标准输出的概率表达体系,并把面向农业的风险指标(如“花期≥35 ℃ 持续 3 日”的 Brier Score)直接融入模型训练损失,而不是事后再做“翻译”[21,23]。
当前气象大模型在常规天气场上有很好表现,但对极端事件(如特大暴雨、强台风、骤旱)的预测仍不稳定。这些事件样本稀少,模型容易低估其强度。可能的解决方案包括:结合重采样或以极端为重点的损失函数来强化模型对尾部事件的学习;构建专门的极端事件训练子集,或将物理引导(PINN)与生成式扩散结合,在保证物理一致性的前提下增强对小概率高影响事件的概率化能力[60,63]。
气象大模型容易出现物理不一致(如能量不守恒、不符合动力约束)的问题,限制了气象专家与农业用户对其信任。提高可解释性的路径主要有4点:1)把质量、能量、正定性等约束以软约束形式嵌入损失函数;2)采用混合架构(NeuralGCM、WeatherGFT、“物理+AI”短临降水模型)将动力子模块与 AI 校正结合[48];3)事后校正与物理一致检查(如收支闭合、能量守恒检验);4)加强可解释 AI 与因果发现工具,帮助专家追踪模型对关键变量的依赖结构。
即使 AI 模型准确度接近或超过数值模式,业务化应用仍需考虑稳定性、鲁棒性以及与现有预报流程的兼容。AI 模型在训练数据分布内表现良好,但遇到未知情况(如前所未有的极端事件或气候新态)时是否可靠仍是开放问题。需要通过持续学习/在线适应[66]、面向分布外的稳健性评估,以及与现有 NWP 流水线的并行业务化机制来逐步建立信任。在面向中国县域、合作社一级的农业气象服务时,“基础模型 + 局地自适应”的轻量部署架构是关键的工程方向,而边缘智能大模型的最新研究进展[67]为这一部署范式提供了重要技术参考。
随着越来越多机构投入气象大模型研究,建立统一的评测基准与开放平台变得迫切。WeatherBench 与 WeatherBench 2 在通用气象指标上提供了基线,ChaosBench 在 S2S 时效与物理一致性维度补充了基准[63];业务化检验的统一方法论仍在构建中。在农业方向,尚缺乏统一的“农业 KPI 转化基准”——建议社区共建涵盖关键农事期预报技巧、极端事件命中率、产量相关性的“农业气象大模型 benchmark”,为不同方法的横向对比提供共同语言。
基于前述4节的综述与 5.1 节所列挑战,面向中国乃至全球农业应用,气象大模型未来的研究空间可凝练为5个方向。
1) 作物关键生育期精细化预报。当前以小麦抽穗扬花期、水稻孕穗期、玉米乳熟期为代表的“窗口化、事件化、条件化”预报已能借助 FuXi-S2S、ClimaX、GenCast 等基础模型与扩散模型提供中期至次季节时效的通用气象输入,并在关键窗口内实现基本可用的概率化输出[10-11,13]。然而,通用气象大模型尚未把作物物候模型的关键期定义原生嵌入训练目标,导致“花期 3 日≥35 ℃ 累积概率”等作物-气候耦合事件的概率校准仍以事后后处理为主,窗口内的极端事件样本稀缺问题突出。建议在 Transformer 与生成式扩散模型(见第 2.3 节、第 2.5 节)的联合框架下,把作物物候模型的关键期定义作为损失项直接嵌入训练,并配套关键农事期窗口技巧评估(见第 4.5 节),形成窗口化目标函数与窗口化评测的闭环;在跨季节、跨气候带部署中引入持续学习机制[66]亦可缓解作物-气候联合分布漂移带来的模型技巧衰减。
2) 田块尺度极端事件临近预警。面向霜冻、热浪、强对流降水、冰雹等典型极端事件,NowcastNet、CastDiffuser、MetNet-3、1 km 超分辨率框架等工作已在分钟至小时、公里至百米尺度上提供了可用的高分辨率输出,证明了“全球 LMM+超分辨率+生成式集合”技术路线的初步可行性[46,51]。然而,1 km 及以下的算法稳定性、极端事件样本扩充与重采样策略,以及面向区域差异的微调机制仍存显著缺口,且县域硬件对持续运行集合化预警的支撑能力有限。建议在生成式扩散模型与物理引导多模态方法的交叉框架下,构建物理子模块、扩散集合与区域微调的统一管线,并借助轻量化与联邦学习降低部署成本,使之适用于主产区县域与合作社一级的实时预警。
3) 气象大模型与作物模型的耦合。当前 APSIM、WOFOST、DSSAT 等作物模型已与 ACCESS-S2、ECMWF SEAS5 及若干气象大模型形成“气象输入→作物模型”的串行管道,在区域产量预测与战略规划中具备较强物理可解释性[22,55]。然而,串行耦合无法把作物模型的产量梯度反向传播回气象大模型,致使整体技巧的提升依赖于各自独立迭代,且气象与作物模型在时空网格、变量定义、不确定性表达上不一致导致的信息损耗难以系统评估。建议沿物理引导神经网络(见第 2.6 节)与可解释性约束探索基于神经常微分方程或差分形式的可微作物模型,与气象大模型联合训练并共享梯度,使“气象精度→农业效益”链条的整体端到端优化成为可能。
4) 农业 Foundation Model 的预训练-微调范式。在 Aurora 等地球系统基础模型的启发下,面向农业的“专用基础模型”已在 ERA5、AgERA5、CMIP6 与多源遥感数据的联合预训练上出现早期工作,PWV 检索、土壤水分预测等下游任务验证了“大规模预训练+轻量微调”在农业数据稀疏条件下的可行性[11,50]。然而,现有尝试仍以单任务或小规模多任务为主,缺少覆盖作物物候、极端事件、产量回归等异构任务的统一预训练目标,且对中国主粮作物-主产区的特定先验融入不足。建议以多任务自监督与物理一致性损失构建农业 Foundation Model 的预训练目标,并在第 4.5 节统一评估协议下对不同微调方案进行可复现比较。
5) 农业特定 benchmark 与开放评测。WeatherBench、ChaosBench、PAD 等通用与物理一致性基准已为深度学习气象模型的横向对比提供了基础设施,其覆盖的时效、变量与空间结构维度亦在快速扩展[60,63]。然而,农业视角下的“关键农事期窗口预报、极端事件命中率、县域产量回归、经济价值”4类评估任务目前仍散落在各自的独立研究中,缺乏统一、开源、可复现的开放评测基准。建议在第 4.5 节统一评估协议基础上,由学术机构与气象、农业业务部门共同牵头,构建包含确定性与概率化双轨输出、覆盖中国七大主产区、跨三类时效、配套时间锁训练测试划分的开放评测基准,以显著加速气象大模型在农业领域的迭代速度与可重复性。
综上,5 个方向中,第2方向与第5方向属短期(1~2 年)可集中突破者,前者具备 1 km 超分辨率与扩散生成式的现成技术组件,后者则可以现有WeatherBench/ChaosBench为基础快速扩展;第1方向属中期(2~3 年)积累方向,有赖于物候模型与气象模型的联合目标函数设计;第3方向耦合与第4方向属中长期(3~5 年)方向,前者受限于可微作物模型的工程成熟度,后者依赖大规模多任务预训练与评测生态的共同成长。
随着观测资料的不断累积与计算力的持续提升,气象大模型在中期—次季节—季节多尺度上的技巧仍有显著上升空间。物理驱动的数值模式与数据驱动的气象大模型有望融合为“物理-AI混合”架构,在通用气象指标与农业 KPI 上实现协同优化。在天气预报之外,气象大模型在气候变化预测、农业气象灾害风险评估、粮食安全战略规划等场景中亦将持续扩展应用边界。本文从“农业问题—模型选型—效益转化”主线出发,系统综述了气象大模型在农业应用中的研究进展、现有挑战与未来方向,希望为后续的工程实践与基础研究提供具有针对性的参考[4,7]。

参考文献 引证文献
排序方式:
1
Hultgren A, Carleton T, Delgado M, et al. Impacts of climate change on global agriculture accounting for adaptation[J]. Nature, 2025, 642(8068): 644−652
2
Lipper L, Thornton P, Campbell B M, et al. Climate-smart agriculture for food security[J]. Nature Climate Change, 2014, 4(12): 1068−1072
3
Yang Yi, Tilman D, Jin Zhenong, et al. Climate change exacerbates the environmental impacts of agriculture[J]. Science, 2024, 385(6713): eadn3747
4
Parolini G. Weather, climate, and agriculture: Historical contributions and perspectives from agricultural meteorology[J]. WIREs Climate Change, 2022, 13(3): e766
5
Schepen A, Everingham Y, Wang Q J. An improved workflow for calibration and downscaling of GCM climate forecasts for agricultural applications: A case study on prediction of sugarcane yield in Australia[J]. Agricultural and Forest Meteorology, 2020, 291: 107991
6
Jha P K, Athanasiadis P, Gualdi S, et al. Using daily data from seasonal forecasts in dynamic crop models for yield prediction: A case study for rice in Nepal’s Terai[J]. Agricultural and Forest Meteorology, 2019, 265: 349−358
7
Materia S, García L P, Van Straaten C, et al. Artificial intelligence for climate prediction of extremes: State of the art, challenges, and future perspectives[J]. WIREs Climate Change, 2024, 15(6): e914
8
Yu Siwei, Ma Jianwei. Deep learning for geophysics: Current and future trends[J]. Reviews of Geophysics, 2021, 59(3): e2021RG000742
9
Bi Kaifeng, Xie Lingxi, Zhang Hengheng, et al. Accurate medium-range global weather forecasting with 3D neural networks[J]. Nature, 2023, 619(7970): 533−538
10
Chen Lei, Zhong Xiaohui, Zhang Feng, et al. FuXi: A cascade machine learning forecasting system for 15-day global weather forecast[J]. npj Climate and Atmospheric Science, 2023, 6(1): 190
11
Bodnar C, Bruinsma W P, Lucic A, et al. A foundation model for the Earth system[J]. Nature, 2025, 641(8065): 1180−1187
12
Kurth T, Subramanian S, Harrington P, et al. FourCastNet: Accelerating global high-resolution weather forecasting using adaptive Fourier neural operators[C]//Proc of the Platform for Advanced Scientific Computing Conf. New York: ACM, 2023: 1−11
13
Price I, Sanchez-Gonzalez A, Alet F, et al. Probabilistic weather forecasting with machine learning[J]. Nature, 2025, 637(8044): 84−90
14
Niu Dan, Niu Daben, Li Yinghao, et al. CastDiffuser: Cascaded latent diffusion framework for high-resolution precipitation nowcasting via multi source fusion[J]. Neurocomputing, 2026, 670: 132435
15
Sun Xiuyu, Zhong Xiaohui, Xu Xiaoze, et al. A data-to-forecast machine learning system for global weather[J]. Nature Communications, 2025, 16(1): 6658
16
Boukabara S A, Krasnopolsky V, Stewart J Q, et al. Realizing the benefits of AI across the numerical weather prediction value chain[J]. Bulletin of the American Meteorological Society, 2020, 101(1): 29−33
17
Mass C. The uncoordinated giant II: Why US operational numerical weather prediction is still lagging and how to fix it[J]. Bulletin of the American Meteorological Society, 2023, 104(4): E851−E871
18
Steppeler J, Gan P, Fang F, et al. Numerical weather prediction in the time of Exaflop computers[J]. Bulletin of the American Meteorological Society, 2025, 106(6): E1087−E1097
19
Baran Á, Baran S. A two-step machine-learning approach to statistical post-processing of weather forecasts for power generation[J]. Quarterly Journal of the Royal Meteorological Society, 2024, 150(759): 1029−1047
20
Sacco M A, Ruiz J J, Pulido M, et al. Evaluation of machine learning techniques for forecast uncertainty quantification[J]. Quarterly Journal of the Royal Meteorological Society, 2022, 148(749): 3470−3490
21
An-Vo D A, Radanielson A M, Mushtaq S, et al. A framework for assessing the value of seasonal climate forecasting in key agricultural decisions[J]. Climate Services, 2021, 22: 100234
22
Wanthanaporn U, Supit I, Chaowiwat W, et al. Skill of rice yields forecasting over Mainland Southeast Asia using the ECMWF SEAS5 ensemble prediction system and the WOFOST crop model[J]. Agricultural and Forest Meteorology, 2024, 351: 110001
23
Naveau P, Hannart A, Ribes A. Statistical methods for extreme event attribution in climate science[J]. Annual Review of Statistics and Its Application, 2020, 7(1): 89−110
24
Hersbach H, Bell B, Berrisford P, et al. The ERA5 global reanalysis[J]. Quarterly Journal of the Royal Meteorological Society, 2020, 146(730): 1999−2049
25
Lam R, Sanchez-Gonzalez A, Willson M, et al. Learning skillful medium-range global weather forecasting[J]. Science, 2023, 382(6677): 1416−1421
26
Chen Kang, Han Tao, Ling Fenghua, et al. The operational medium-range deterministic weather forecasting can be extended beyond a 10-day lead time[J]. Communications Earth & Environment, 2025, 6(1): 518
27
Allen A, Markou S, Tebbutt W, et al. End-to-end data-driven weather prediction[J]. Nature, 2025, 641(8065): 1172−1179
28
LeCun Y, Bottou L, Bengio Y, et al. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 1998, 86(11): 2278−2324
29
Shi Xingjian, Chen Zhourong, Wang Hao, et al. Convolutional LSTM network: A machine learning approach for precipitation nowcasting[C]//Proc of the 28th Int Conf on Neural Information Processing Systems. Red Hook, NY: Curran Associates Inc. , 2015: 802−810
30
周立, 吴洋洋, 苗晓烨. 基于深度学习的不完整时序数据补全方法综述[J]. 计算机研究与发展, 2025, 62(9): 2233−2258
Zhou Li, Wu Yangyang, Miao Xiaoye. Survey of incomplete time series imputation methods based on deep learning[J]. Journal of Computer Research and Development, 2025, 62(9): 2233−2258 (in Chinese)
31
Scarselli F, Gori M, Tsoi A C, et al. The graph neural network model[J]. IEEE Transactions on Neural Networks, 2009, 20(1): 61−80
32
Alexe M, Boucher E, Lean P, et al. GraphDOP: Towards skilful data-driven medium-range weather forecasts learnt and initialised directly from observations[J]. arXiv preprint, arXiv: 2412.15687, 2024
33
Lang S, Alexe M, Chantry M, et al. AIFS - ECMWF’s data-driven forecasting system[J]. arXiv preprint, arXiv: 2406.01465, 2024
34
Nguyen T, Brandstetter J, Kapoor A, et al. ClimaX: A foundation model for weather and climate[C]//Proc of the 40th Int Conf on Machine Learning. San Diego, CA: PMLR, 2023: 25904−25938
35
Hu Yuan, Chen Lei, Wang Zhibin, et al. SwinVRNN: A data-driven ensemble forecasting model via learned distribution perturbation[J]. Journal of Advances in Modeling Earth Systems, 2023, 15(2): e2022MS003211
36
Andrychowicz M, Espeholt L, Li D, et al. Deep learning for day forecasts from sparse observations[J]. arXiv preprint, arXiv: 2306.06079, 2023
37
Xu Siyi, Zhang Yize, Chen Junping, et al. Short-to medium-term weather forecast skill of the AI-based Pangu-Weather model using automatic weather stations in China[J]. Remote Sensing, 2025, 17(2): 191
38
Yang S, Liu Z, Shi Z, et al. WSSM: Geographic enhanced hierarchical state-space model for global station weather forecast[J]. arXiv preprint, arXiv: 2501.11238, 2025
39
Qin H, Chen Y, Jiang Q, et al. MetMamba: Regional weather forecasting with spatial-temporal Mamba model[J]. arXiv preprint, arXiv: 2408.06400, 2024
40
Verma Y, Heinonen M, Garg V. ClimODE: Climate and weather forecasting with physics-informed neural ODEs[C/OL]//Proc of the 12th Int Conf on Learning Representations. 2024[2026-06-18]. https://openreview.net/forum?id=xuY33XhEGR
41
Liu P, Zhou T, Sun L, et al. Mitigating time discretization challenges with WeatherODE: A sandwich physics-driven neural ODE for weather forecasting[J]. arXiv preprint, arXiv: 2410.06560, 2024
42
Gu A, Dao T. Mamba: Linear-time sequence modeling with selective state spaces[C/OL]//Proc of the 1st Conf on Language Modeling. 2024[2026-04-30]. https://openreview.net/forum?id=tEYskw1VY2
43
Ravuri S, Lenc K, Willson M, et al. Skilful precipitation nowcasting using deep generative models of radar[J]. Nature, 2021, 597(7878): 672−677
44
Sha Yingkai, Sobash R A, Gagne D J. Generative ensemble deep learning severe weather prediction from a deterministic convection-allowing model[J]. Artificial Intelligence for the Earth Systems, 2024, 3(2): e230094
45
Zhang Yuchen, Long Mingsheng, Chen Kaiyuan, et al. Skilful nowcasting of extreme precipitation with NowcastNet[J]. Nature, 2023, 619(7970): 526−532
46
Gong Junchao, Bai Lei, Ye Peng, et al. CasCast: Skillful high-resolution precipitation nowcasting via cascaded modelling[C]//Proc of the 41st Int Conf on Machine Learning. San Diego, CA: PMLR, 2024: 15809−15822
47
Kochkov D, Yuval J, Langmore I, et al. Neural general circulation models for weather and climate[J]. Nature, 2024, 632(8027): 1060−1066
48
Das P, Posch A, Barber N, et al. Hybrid physics-AI outperforms numerical weather prediction for extreme precipitation nowcasting[J]. npj Climate and Atmospheric Science, 2024, 7(1): 282
49
Boussioux L, Zeng C, Guénais T, et al. Hurricane forecasting: A novel multimodal machine learning framework[J]. Weather and Forecasting, 2022, 37(6): 817−831
50
Ding Junsheng, Chen Wu, Chen Junping, et al. Real-time high-resolution global PWV retrieval based on weather forecast foundation models and cross-validation with radiosonde, GNSS, and ERA5[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2025, 18: 15972−15981
51
Park H, Park S, Kang D, et al. A super-resolution framework for downscaling machine learning weather prediction toward 1-km air temperature[J]. npj Climate and Atmospheric Science, 2026, 9(1): 56
52
Tarraf B, Brun F, Raynaud L, et al. Assessing the impact of weather forecast uncertainties in crop water stress model predictions[J]. Agricultural and Forest Meteorology, 2024, 349: 109934
53
Chen Lei, Zhong Xiaohui, Li Hao, et al. A machine learning model that outperforms conventional global subseasonal forecast models[J]. Nature Communications, 2024, 15(1): 6425
54
Garcia D, Rolim J, Cameira M R, et al. Assessing seasonal forecast performance to predict crop irrigation requirements to support water management decision-making in the Mediterranean region[J]. Agricultural Water Management, 2025, 313: 109467
55
Luo Qunying, Wen Li, Cowan T, et al. Seasonal climate forecast - an important tool in managing the risk of extreme weather events in Australia’s wheat industry[J]. Agricultural and Forest Meteorology, 2024, 351: 110005
56
Garza A, Mergenthaler-Canseco M. TimeGPT−1[J]. arXiv preprint, arXiv: 2310.03589, 2023
57
Shin J Y, Min B, Kim K R. High-resolution wind speed forecast system coupling numerical weather prediction and machine learning for agricultural studies—a case study from South Korea[J]. International Journal of Biometeorology, 2022, 66(7): 1429−1443
58
Su Qiong, Ale S, Himanshu S K, et al. Calibration and bias correction of seasonal weather forecasts from the North American multi-model ensemble: Potential applications for regional crop modelling and irrigation management[J]. The Journal of Agricultural Science, 2025, 163(2): 145−159
59
Garcia D, Silva N, Rolim J, et al. Prediction of crops cycle with seasonal forecasts to support decision-making[J]. Agronomy, 2025, 15(6): 1291
60
Skok G, Lledó L. Spatial verification of global precipitation forecasts[J]. Quarterly Journal of the Royal Meteorological Society, 2025, 151(771): e5006
61
Rasp S, Dueben P D, Scher S, et al. WeatherBench: A benchmark data set for data-driven weather forecasting[J]. Journal of Advances in Modeling Earth Systems, 2020, 12(11): e2020MS002203
62
Rasp S, Hoyer S, Merose A, et al. WeatherBench 2: A benchmark for the next generation of data-driven global weather models[J]. Journal of Advances in Modeling Earth Systems, 2024, 16(6): e2023MS004019
63
Nathaniel J, Qu Yongquan, Nguyen T, et al. ChaosBench: A multi-channel, physics-based benchmark for subseasonal-to-seasonal climate prediction[C]//Proc of the 38th Int Conf on Neural Information Processing Systems. Red Hook, NY: Curran Associates Inc., 2024: 43715−43729
64
Dosovitskiy A, Beyer L, Kolesnikov A, et al. An image is worth 16x16 words: Transformers for image recognition at scale[C/OL]//Proc of the 9th Int Conf on Learning Representations. 2021[2026-04-30]. https://openreview.net/forum?id=YicbFdNTTy
65
史宏志, 赵健, 赵雅倩, 等. 大模型时代的混合专家系统优化综述[J]. 计算机研究与发展, 2025, 62(5): 1164−1189
Shi Hongzhi, Zhao Jian, Zhao Yaqian, et al. Survey on system optimization for mixture of experts in the era of large models[J]. Journal of Computer Research and Development, 2025, 62(5): 1164−1189 (in Chinese)
66
李文斌, 熊亚锟, 范祇辰, 等. 持续学习的研究进展与趋势[J]. 计算机研究与发展, 2024, 61(6): 1476−1496
Li Wenbin, Xiong Yakun, Fan Zhichen, et al. Advances and trends of continual learning[J]. Journal of Computer Research and Development, 2024, 61(6): 1476−1496 (in Chinese)
67
王睿, 张留洋, 高志涌, 等. 面向边缘智能的大模型研究进展[J]. 计算机研究与发展, 2025, 62(10): 2565−2582
Wang Rui, Zhang Liuyang, Gao Zhiyong, et al. Research progress on large models for edge intelligence[J]. Journal of Computer Research and Development, 2025, 62(10): 2565−2582 (in Chinese)
2026年第63卷第7期
73
32
0
引用本文
BibTeX
文章信息
doi: 10.7544/issn1000-1239.202660002
  • 接收时间:2025-12-31
  • 首发时间:2026-11-17
  • 出版时间:2026-07-01
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2025-12-31
  • 修回日期:2026-05-06
基金
作者信息
    中国农业大学信息与电气工程学院 北京 100083

通讯作者:

王耀君()
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202660002
分享至
全文二维码

扫描看全文

引用本文
BibTeX
本文的引用情况
2种不同金属材料的力学参数
科
Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)
属
Genus
种数
Number of
species
占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏