Article(id=1261262690375353074, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1261262687258985194, articleNumber=null, orderNo=null, doi=10.12404/j.issn.1671-1815.2406929, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=research-article, receivedDate=1726243200000, receivedDateStr=2024-09-14, revisedDate=1745683200000, revisedDateStr=2025-04-27, acceptedDate=null, acceptedDateStr=null, onlineDate=1778638058513, onlineDateStr=2026-05-13, pubDate=1752768000000, pubDateStr=2025-07-18, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1778638058513, onlineIssueDateStr=2026-05-13, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1778638058513, creator=13701087609, updateTime=1778638058513, updator=13701087609, issue=Issue{id=1261262687258985194, tenantId=1146029695717560320, journalId=1146123166801305609, year='2025', volume='25', issue='20', pageStart='8317', pageEnd='8759', issueExtLink='null', onlineDate='null', pubDate='null', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1778638057769, creator=13701087609, updateTime=1778753106634, updator=13701087609, preIssue=null, nextIssue=null, ext={EN=IssueExt(id=1261745237240722095, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1261262687258985194, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1261745237240722096, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1261262687258985194, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null}, startPage=8379, endPage=8391, ext={EN=ArticleExt(id=1261262691096773367, articleId=1261262690375353074, tenantId=1146029695717560320, journalId=1146123166801305609, language=EN, title=Daily Precipitation Forecasting Using Global Weather Model and Regional Pre-training Optimization: A Case Study in Shaanxi Province, columnId=1156262729351549255, journalTitle=Science Technology and Engineering, columnName=Papers·Astronomy and Geosciences, runingTitle=null, highlight=null, articleAbstract=

To improve the accuracy of precipitation forecasts and address the limitations of traditional numerical weather prediction models in forecast precision and computational efficiency, a meteorological large model was combined with a deep learning post-processing approach was combined. A case study was conducted for precipitation forecasts over Shaanxi Province during 2008—2018. Based on meteorological variable fields output by the FourCastNet model, a pre-trained model mapping meteorological fields to regional precipitation was constructed using Bayesian-optimized convolutional neural networks (CNN)/long short-term memory (LSTM) networks. The results indicate that this method outperforms traditional numerical weather prediction models in terms of spatial resolution and forecast accuracy. The regionally fine-tuned forecasts more accurately capture the spatiotemporal distribution of precipitation. Furthermore, the Bayesian-optimized deep learning post-processing algorithm effectively mitigates the impact of initial field biases on forecast results. These findings demonstrate the significant potential of integrating meteorological large models with deep learning post-processing algorithms for accurate precipitation forecasting, providing scientific support for disaster prevention, agricultural production, and water resource management.

, correspAuthors=Ling HAN, authorNote=null, correspAuthorsNote=null, copyrightStatement=null, 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, authorCompany=null, fund=null, authors=null, authorsList=Hao-yu WANG, Ling HAN, Liang-zhi LI), CN=ArticleExt(id=1261262700206801730, articleId=1261262690375353074, tenantId=1146029695717560320, journalId=1146123166801305609, language=CN, title=利用气象大模型和区域预训练调优进行日降水预报:以陕西省为例, columnId=1156262730077163858, journalTitle=科学技术与工程, columnName=论文·天文学、地球科学, runingTitle=null, highlight=null, articleAbstract=

为提高降水预报的精度并解决传统数值模式在预报精度和运算效率上的瓶颈问题,通过结合气象大模型和深度学习后处理方法,对陕西省2008—2018年的降水预报进行实例研究。以FourCastNet气象大模型输出的气象要素场为基础,利用贝叶斯优化的卷积神经网络(convolutional neural network,CNN)/长短期记忆网络(long short-term memory, LSTM)算法构建气象要素场-区域降水信息的预训练模型,生成高分辨率的日降水预报。结果表明:该方法在空间分辨率和预报精度上优于传统数值天气预报模式;区域预训练调优预报结果更准确地反映了区域降水的时空分布;基于贝叶斯优化的深度学习后处理算法能够有效缓解初始场偏差对预报的影响。可见,气象大模型结合深度学习后处理算法在降水精准预报中具有显著的应用潜力,为防灾减灾、农业生产及水资源管理提供了科学支持。

, correspAuthors=韩玲, authorNote=null, correspAuthorsNote=
* 韩玲(1964—),女,汉族,辽宁沈阳人,博士,教授,博士研究生导师。研究方向:遥感信息提取、资源环境遥感和遥感影像智能识别。E-mail:
, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=6H9rw/uRbw4Z1mAUUF4lsg==, magXml=YqV+jMysgHKJPu0dDZKzbg==, pdfUrl=null, pdf=1rKDkjNqm7hwkRu3RJhtFA==, pdfFileSize=12601827, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=Ys7UlP9Vs5wWiqm1/U5cUw==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=TQsglO1cGyqmjg7MHmr7eA==, mapNumber=null, authorCompany=null, fund=null, authors=

王浩宇(2000—),男,汉族,山东滕州人,硕士研究生。研究方向:遥感信息提取。E-mail:

, authorsList=王浩宇, 韩玲, 李良志)}, authors=[Author(id=1261377014376874250, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=1019404761@qq.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1261377016008458511, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, authorId=1261377014376874250, language=EN, stringName=Hao-yu WANG, firstName=Hao-yu, middleName=null, lastName=WANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1 School of Geological Engineering and Geomatics, Chang’an University, Xi’an 710054, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1261377016394334482, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, authorId=1261377014376874250, language=CN, stringName=王浩宇, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1 长安大学地质工程与测绘学院, 西安 710054, bio={"content":"

王浩宇(2000—),男,汉族,山东滕州人,硕士研究生。研究方向:遥感信息提取。E-mail:

"}, bioImg=null, bioContent=

王浩宇(2000—),男,汉族,山东滕州人,硕士研究生。研究方向:遥感信息提取。E-mail:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1261377012200030450, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, xref=1, ext=[AuthorCompanyExt(id=1261377012241973492, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, companyId=1261377012200030450, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 School of Geological Engineering and Geomatics, Chang’an University, Xi’an 710054, China), AuthorCompanyExt(id=1261377012267139316, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, companyId=1261377012200030450, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 长安大学地质工程与测绘学院, 西安 710054)])]), Author(id=1261377017182863639, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=hanling@chd.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1261377018088833309, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, authorId=1261377017182863639, language=EN, stringName=Ling HAN, firstName=Ling, middleName=null, lastName=HAN, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, 3, *, address=2 School of Land Engineering, Chang’an University, Xi’an 710054, China
3 Key Laboratory of Land Consolidation, Shaanxi Province, Xi’an 710054, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1261377018676035872, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, authorId=1261377017182863639, language=CN, stringName=韩玲, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, 3, *, address=2 长安大学土地工程学院, 西安 710054
3 陕西省土地整治重点实验室, 西安 710054, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1261377012791427319, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, xref=2, ext=[AuthorCompanyExt(id=1261377012824981752, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, companyId=1261377012791427319, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 School of Land Engineering, Chang’an University, Xi’an 710054, China), AuthorCompanyExt(id=1261377012988559610, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, companyId=1261377012791427319, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 长安大学土地工程学院, 西安 710054)]), AuthorCompany(id=1261377013554790657, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, xref=3, ext=[AuthorCompanyExt(id=1261377013571567875, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, companyId=1261377013554790657, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3 Key Laboratory of Land Consolidation, Shaanxi Province, Xi’an 710054, China), AuthorCompanyExt(id=1261377013579956484, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, companyId=1261377013554790657, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3 陕西省土地整治重点实验室, 西安 710054)])]), Author(id=1261377020584444201, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, 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=1261377021050011959, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, authorId=1261377020584444201, language=EN, stringName=Liang-zhi LI, firstName=Liang-zhi, middleName=null, lastName=LI, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2 School of Land Engineering, Chang’an University, Xi’an 710054, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1261377021456859453, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, authorId=1261377020584444201, language=CN, stringName=李良志, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2 长安大学土地工程学院, 西安 710054, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1261377012791427319, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, xref=2, ext=[AuthorCompanyExt(id=1261377012824981752, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, companyId=1261377012791427319, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 School of Land Engineering, Chang’an University, Xi’an 710054, China), AuthorCompanyExt(id=1261377012988559610, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, companyId=1261377012791427319, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 长安大学土地工程学院, 西安 710054)])])], keywords=[Keyword(id=1261377023470125392, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=EN, orderNo=1, keyword=precipitation forecasting), Keyword(id=1261377024992657748, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=EN, orderNo=2, keyword=deep learning), Keyword(id=1261377025252704605, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=EN, orderNo=3, keyword=large meteorological models), Keyword(id=1261377025693106531, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=EN, orderNo=4, keyword=pre-training), Keyword(id=1261377026074788199, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=CN, orderNo=1, keyword=降水预报), Keyword(id=1261377026531967341, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=CN, orderNo=2, keyword=深度学习), Keyword(id=1261377027416965492, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=CN, orderNo=3, keyword=气象大模型), Keyword(id=1261377029430231419, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=CN, orderNo=4, keyword=预训练)], refs=[Reference(id=1261377051706180213, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2021, volume=28, issue=31, pageStart=41576, pageEnd=41595, url=null, language=null, rfNumber=[1], rfOrder=0, authorNames=Srivastav A L, Dhyani R, Ranjan M, journalName=Environmental Science and Pollution Research, refType=null, unstructuredReference=Srivastav A L, Dhyani R, Ranjan M, et al. Climate-resilient strategies for sustainable management of water resources and agriculture[J]. Environmental Science and Pollution Research, 2021, 28(31): 41576-41595., articleTitle=Climate-resilient strategies for sustainable management of water resources and agriculture, refAbstract=null), Reference(id=1261377052096250492, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2024, volume=35, issue=1, pageStart=156, pageEnd=166, url=null, language=null, rfNumber=[2], rfOrder=1, authorNames=赵铜铁钢, 张弛, 田雨, journalName=水科学进展, refType=null, unstructuredReference=赵铜铁钢, 张弛, 田雨, . 全球气象预报驱动流域水文预报研究进展与展望[J]. 水科学进展, 2024, 35(1): 156-166., articleTitle=全球气象预报驱动流域水文预报研究进展与展望, refAbstract=null), Reference(id=1261377053027385990, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2024, volume=35, issue=1, pageStart=156, pageEnd=166, url=null, language=null, rfNumber=[2], rfOrder=2, authorNames=Zhao Tongtiegang, Zhang Chi, Tian Yu, journalName=Advances in Water Science, refType=null, unstructuredReference=Zhao Tongtiegang, Zhang Chi, Tian Yu, et al. Progress and prospects ofglobal meteorological forecasting to drive basin hydrological forecasting[J]. Advances in Water Science, 2024, 35(1): 156-166., articleTitle=Progress and prospects ofglobal meteorological forecasting to drive basin hydrological forecasting, refAbstract=null), Reference(id=1261377053782360715, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2012, volume=12, issue=11, pageStart=2602, pageEnd=2606, url=null, language=null, rfNumber=[3], rfOrder=3, authorNames=周宏博, 孙树林, 尚文涛, journalName=科学技术与工程, refType=null, unstructuredReference=周宏博, 孙树林, 尚文涛, . 降雨雨型和强度对土边坡稳定性影响分析[J]. 科学技术与工程, 2012, 12(11): 2602-2606., articleTitle=降雨雨型和强度对土边坡稳定性影响分析, refAbstract=null), Reference(id=1261377054956765845, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2012, volume=12, issue=11, pageStart=2602, pageEnd=2606, url=null, language=null, rfNumber=[3], rfOrder=4, authorNames=Zhou Hongbo, Sun Shulin, Shang Wentao, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Zhou Hongbo, Sun Shulin, Shang Wentao, et al. Analysis of the impact of rainfall patterns and intensity on soil slope stability[J]. Science Technology and Engineering, 2012, 12(11): 2602-2606., articleTitle=Analysis of the impact of rainfall patterns and intensity on soil slope stability, refAbstract=null), Reference(id=1261377056273777309, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=9, pageStart=3543, pageEnd=3553, url=null, language=null, rfNumber=[4], rfOrder=5, authorNames=许浩然, 杨兆前, 陈中举, journalName=科学技术与工程, refType=null, unstructuredReference=许浩然, 杨兆前, 陈中举, . 湖南省极端气候时空特征分析及风险评估[J]. 科学技术与工程, 2024, 24(9): 3543-3553., articleTitle=湖南省极端气候时空特征分析及风险评估, refAbstract=null), Reference(id=1261377057037140648, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=9, pageStart=3543, pageEnd=3553, url=null, language=null, rfNumber=[4], rfOrder=6, authorNames=Xu Haoran, Yang Zhaoqian, Chen Zhongju, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Xu Haoran, Yang Zhaoqian, Chen Zhongju, et al. Analysis of spatiotemporal characteristics and risk assessment of extreme climate in Hunan Province[J]. Science Technology and Engineering, 2024, 24(9): 3543-3553., articleTitle=Analysis of spatiotemporal characteristics and risk assessment of extreme climate in Hunan Province, refAbstract=null), Reference(id=1261377057586594477, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2019, volume=45, issue=4, pageStart=445, pageEnd=457, url=null, language=null, rfNumber=[5], rfOrder=7, authorNames=金荣花, 代刊, 赵瑞霞, journalName=气象, refType=null, unstructuredReference=金荣花, 代刊, 赵瑞霞, . 我国无缝隙精细化网格天气预报技术进展与挑战[J]. 气象, 2019, 45(4): 445-457., articleTitle=我国无缝隙精细化网格天气预报技术进展与挑战, refAbstract=null), Reference(id=1261377057959887541, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2019, volume=45, issue=4, pageStart=445, pageEnd=457, url=null, language=null, rfNumber=[5], rfOrder=8, authorNames=Jin Ronghua, Dai Kan, Zhao Ruixia, journalName=Meteorological Monthly, refType=null, unstructuredReference=Jin Ronghua, Dai Kan, Zhao Ruixia, et al. Progress and challenges of seamless fine-grid weather forecasting technology in China[J]. Meteorological Monthly, 2019, 45(4): 445-457., articleTitle=Progress and challenges of seamless fine-grid weather forecasting technology in China, refAbstract=null), Reference(id=1261377059763438275, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2024, volume=54, issue=5, pageStart=1680, pageEnd=1701, url=null, language=null, rfNumber=[6], rfOrder=9, authorNames=钟琦, 张智察, 姚秀萍, journalName=中国科学: 地球科学, refType=null, unstructuredReference=钟琦, 张智察, 姚秀萍, . 应用物理结合的人工智能方法提升“21·7”河南极端强降水预报[J]. 中国科学: 地球科学, 2024, 54(5): 1680-1701., articleTitle=应用物理结合的人工智能方法提升“21·7”河南极端强降水预报, refAbstract=null), Reference(id=1261377060933649096, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2024, volume=54, issue=5, pageStart=1680, pageEnd=1701, url=null, language=null, rfNumber=[6], rfOrder=10, authorNames=Zhong Qi, Zhang Zhicha, Yao Xiuping, journalName=Science China: Earth Sciences, refType=null, unstructuredReference=Zhong Qi, Zhang Zhicha, Yao Xiuping, et al. Improving the forecast of the “21.7” Henan extreme heavy rainfall using AI methods combined with applied physics[J]. Science China: Earth Sciences, 2024, 54(5): 1680-1701., articleTitle=Improving the forecast of the “21.7” Henan extreme heavy rainfall using AI methods combined with applied physics, refAbstract=null), Reference(id=1261377061722178252, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2021, volume=21, issue=17, pageStart=7008, pageEnd=7016, url=null, language=null, rfNumber=[7], rfOrder=11, authorNames=凌敏华, 韩洪宝, journalName=科学技术与工程, refType=null, unstructuredReference=凌敏华, 韩洪宝. 1960—2018年河南省降水时空变化特征及重心[J]. 科学技术与工程, 2021, 21(17): 7008-7016., articleTitle=1960—2018年河南省降水时空变化特征及重心, refAbstract=null), Reference(id=1261377062565233365, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2021, volume=21, issue=17, pageStart=7008, pageEnd=7016, url=null, language=null, rfNumber=[7], rfOrder=12, authorNames=Ling Minhua, Han Hongbao, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Ling Minhua, Han Hongbao. Spatiotemporal variation characteristics and center of precipitation in Henan Province from 1960 to 2018[J]. Science Technology and Engineering, 2021, 21(17): 7008-7016., articleTitle=Spatiotemporal variation characteristics and center of precipitation in Henan Province from 1960 to 2018, refAbstract=null), Reference(id=1261377065090204386, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2021, volume=21, issue=6, pageStart=2186, pageEnd=2193, url=null, language=null, rfNumber=[8], rfOrder=13, authorNames=王宗敏, 王治中, 杨瑶, journalName=科学技术与工程, refType=null, unstructuredReference=王宗敏, 王治中, 杨瑶, . 多时间尺度下遥感降水产品与再分析降水产品在海河流域适用性对比分析[J]. 科学技术与工程, 2021, 21(6): 2186-2193., articleTitle=多时间尺度下遥感降水产品与再分析降水产品在海河流域适用性对比分析, refAbstract=null), Reference(id=1261377065538994921, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2021, volume=21, issue=6, pageStart=2186, pageEnd=2193, url=null, language=null, rfNumber=[8], rfOrder=14, authorNames=Wang Zongmin, Wang Zhizhong, Yang Yao, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Wang Zongmin, Wang Zhizhong, Yang Yao, et al. Comparative analysis of the applicability of remote sensing precipitation products and reanalysis precipitation products in the Haihe River Basin at multiple time scales[J]. Science Technology and Engineering, 2021, 21(6): 2186-2193., articleTitle=Comparative analysis of the applicability of remote sensing precipitation products and reanalysis precipitation products in the Haihe River Basin at multiple time scales, refAbstract=null), Reference(id=1261377065958425333, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2016, volume=27, issue=5, pageStart=534, pageEnd=549, url=null, language=null, rfNumber=[9], rfOrder=15, authorNames=毕宝贵, 代刊, 王毅, journalName=应用气象学报, refType=null, unstructuredReference=毕宝贵, 代刊, 王毅, . 定量降水预报技术进展[J]. 应用气象学报, 2016, 27(5): 534-549., articleTitle=定量降水预报技术进展, refAbstract=null), Reference(id=1261377066759537406, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2016, volume=27, issue=5, pageStart=534, pageEnd=549, url=null, language=null, rfNumber=[9], rfOrder=16, authorNames=Bi Baogui, Dai Kan, Wang Yi, journalName=Journal of Applied Meteorology, refType=null, unstructuredReference=Bi Baogui, Dai Kan, Wang Yi, et al. Progress in quantitative precipitation forecasting technology[J]. Journal of Applied Meteorology, 2016, 27(5): 534-549., articleTitle=Progress in quantitative precipitation forecasting technology, refAbstract=null), Reference(id=1261377067178967813, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2022, volume=80, issue=5, pageStart=649, pageEnd=667, url=null, language=null, rfNumber=[10], rfOrder=17, authorNames=杨绚, 代刊, 朱跃建, journalName=气象学报, refType=null, unstructuredReference=杨绚, 代刊, 朱跃建. 深度学习技术在智能网格天气预报中的应用进展与挑战[J]. 气象学报, 2022, 80(5): 649-667., articleTitle=深度学习技术在智能网格天气预报中的应用进展与挑战, refAbstract=null), Reference(id=1261377067367711500, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2022, volume=80, issue=5, pageStart=649, pageEnd=667, url=null, language=null, rfNumber=[10], rfOrder=18, authorNames=Yang Xun, Dai Kan, Zhu Yuejian, journalName=Acta Meteorologica Sinica, refType=null, unstructuredReference=Yang Xun, Dai Kan, Zhu Yuejian. Progress and challenges of deep learning technology in intelligent grid weather forecasting[J]. Acta Meteorologica Sinica, 2022, 80(5): 649-667., articleTitle=Progress and challenges of deep learning technology in intelligent grid weather forecasting, refAbstract=null), Reference(id=1261377068026217243, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2016, volume=16, issue=6, pageStart=9, pageEnd=16, url=null, language=null, rfNumber=[11], rfOrder=19, authorNames=张同, 鲍艳松, 陆其峰, journalName=科学技术与工程, refType=null, unstructuredReference=张同, 鲍艳松, 陆其峰. IASI卫星资料同化对江淮暴雨预报的试验研究[J]. 科学技术与工程, 2016, 16(6): 9-16, 36., articleTitle=IASI卫星资料同化对江淮暴雨预报的试验研究, refAbstract=null), Reference(id=1261377069540360991, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2016, volume=16, issue=6, pageStart=9, pageEnd=16, url=null, language=null, rfNumber=[11], rfOrder=20, authorNames=Zhang Tong, Bao Yansong, Lu Qifeng, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Zhang Tong, Bao Yansong, Lu Qifeng. Experimental study on the assimilation of IASI satellite data forthe forecasting of torrential rains in the Yangtze-Huaihe River Basin[J]. Science Technology and Engineering, 2016, 16(6): 9-16, 36., articleTitle=Experimental study on the assimilation of IASI satellite data forthe forecasting of torrential rains in the Yangtze-Huaihe River Basin, refAbstract=null), Reference(id=1261377069817185063, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2019, volume=566, issue=7743, pageStart=195, pageEnd=204, url=null, language=null, rfNumber=[12], rfOrder=21, authorNames=Reichstein M, Camps-Valls G, Stevens B, journalName=Nature, refType=null, unstructuredReference=Reichstein M, Camps-Valls G, Stevens B, et al. Deep learning and process understanding for data-driven Earth system science[J]. Nature, 2019, 566(7743): 195-204., articleTitle=Deep learning and process understanding for data-driven Earth system science, refAbstract=null), Reference(id=1261377070098203434, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2021, volume=23, issue=null, pageStart=100178, pageEnd=null, url=null, language=null, rfNumber=[13], rfOrder=22, authorNames=Ren X L, Li X Y, Ren K J, journalName=Big Data Research, refType=null, unstructuredReference=Ren X L, Li X Y, Ren K J, et al. Deep learning-based weather prediction: a survey[J]. Big Data Research, 2021, 23: 100178., articleTitle=Deep learning-based weather prediction: a survey, refAbstract=null), Reference(id=1261377070333084463, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2024, volume=47, issue=1, pageStart=46, pageEnd=54, url=null, language=null, rfNumber=[14], rfOrder=23, authorNames=黄小猛, 林岩銮, 熊巍, journalName=大气科学学报, refType=null, unstructuredReference=黄小猛, 林岩銮, 熊巍, . 数值预报AI气象大模型国际发展动态研究[J]. 大气科学学报, 2024, 47(1): 46-54., articleTitle=数值预报AI气象大模型国际发展动态研究, refAbstract=null), Reference(id=1261377070802846518, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2024, volume=47, issue=1, pageStart=46, pageEnd=54, url=null, language=null, rfNumber=[14], rfOrder=24, authorNames=Huang Xiaomeng, Lin Yanluan, Xiong Wei, journalName=Journal of Atmospheric Sciences, refType=null, unstructuredReference=Huang Xiaomeng, Lin Yanluan, Xiong Wei, et al. International development trends of numerical weather prediction AI large models[J]. Journal of Atmospheric Sciences, 2024, 47(1): 46-54., articleTitle=International development trends of numerical weather prediction AI large models, refAbstract=null), Reference(id=1261377071243248441, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2024, volume=47, issue=1, pageStart=55, pageEnd=64, url=null, language=null, rfNumber=[15], rfOrder=25, authorNames=张弛, 陈国兴, 杨洪涛, journalName=大气科学学报, refType=null, unstructuredReference=张弛, 陈国兴, 杨洪涛. 利用深度学习预报美国东北部日降水分布[J]. 大气科学学报, 2024, 47(1): 55-64., articleTitle=利用深度学习预报美国东北部日降水分布, refAbstract=null), Reference(id=1261377071780119365, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2024, volume=47, issue=1, pageStart=55, pageEnd=64, url=null, language=null, rfNumber=[15], rfOrder=26, authorNames=Zhang Chi, Chen Guoxing, Yang Hongtao, journalName=Journal of Atmospheric Sciences, refType=null, unstructuredReference=Zhang Chi, Chen Guoxing, Yang Hongtao. Forecasting daily precipitation distribution in the northeastern United States using deep learning[J]. Journal of Atmospheric Sciences, 2024, 47(1): 55-64., articleTitle=Forecasting daily precipitation distribution in the northeastern United States using deep learning, refAbstract=null), Reference(id=1261377072019194697, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2020, volume=20, issue=28, pageStart=11447, pageEnd=11454, url=null, language=null, rfNumber=[16], rfOrder=27, authorNames=王坤, 蒋宁, 李敏, journalName=科学技术与工程, refType=null, unstructuredReference=王坤, 蒋宁, 李敏, . 基于SMOTE算法和逻辑回归模型算法的江苏短时强降水潜势预报[J]. 科学技术与工程, 2020, 20(28): 11447-11454., articleTitle=基于SMOTE算法和逻辑回归模型算法的江苏短时强降水潜势预报, refAbstract=null), Reference(id=1261377072296018769, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2020, volume=20, issue=28, pageStart=11447, pageEnd=11454, url=null, language=null, rfNumber=[16], rfOrder=28, authorNames=Wang Kun, Jiang Ning, Li Min, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Wang Kun, Jiang Ning, Li Min, et al. Short-term heavy precipitation potential forecasting in Jiangsu based on the SMOTE algorithm and logistic regression model[J]. Science Technology and Engineering, 2020, 20(28): 11447-11454., articleTitle=Short-term heavy precipitation potential forecasting in Jiangsu based on the SMOTE algorithm and logistic regression model, refAbstract=null), Reference(id=1261377074091180887, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2023, volume=619, issue=7970, pageStart=533, pageEnd=538, url=null, language=null, rfNumber=[17], rfOrder=29, authorNames=Bi K F, Xie L X, Zhang H H, journalName=Nature, refType=null, unstructuredReference=Bi K F, Xie L X, Zhang H H, et al. Accurate medium-range global weather forecasting with 3D neural networks[J]. Nature, 2023, 619(7970): 533-538., articleTitle=Accurate medium-range global weather forecasting with 3D neural networks, refAbstract=null), Reference(id=1261377074414142303, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2024, volume=26, issue=1, pageStart=7, pageEnd=13, url=null, language=null, rfNumber=[18], rfOrder=30, authorNames=田奇, 毕恺峰, 谢凌曦, journalName=中国基础科学, refType=null, unstructuredReference=田奇, 毕恺峰, 谢凌曦. 基于人工智能大模型的中期全球气象预报新方法[J]. 中国基础科学, 2024, 26(1): 7-13, 21., articleTitle=基于人工智能大模型的中期全球气象预报新方法, refAbstract=null), Reference(id=1261377074699354983, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2024, volume=26, issue=1, pageStart=7, pageEnd=13, url=null, language=null, rfNumber=[18], rfOrder=31, authorNames=Tian Qi, Bi Kaifeng, Xie Lingxi, journalName=Chinese Journal of Basic Science, refType=null, unstructuredReference=Tian Qi, Bi Kaifeng, Xie Lingxi. A new method for medium-range global meteorological forecasting based on artificial intelligence large models[J]. Chinese Journal of Basic Science, 2024, 26(1): 7-13, 21., articleTitle=A new method for medium-range global meteorological forecasting based on artificial intelligence large models, refAbstract=null), Reference(id=1261377074938430318, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[19], rfOrder=32, authorNames=Kurth T, Subramanian S, Harrington P, journalName=Proceedings of the Platform for Advanced Scientific Computing Conference, refType=null, unstructuredReference=Kurth T, Subramanian S, Harrington P, et al. FourCastNet: accelerating global high-resolution weather forecasting using adaptive fourier neural operators[C]// Proceedings of the Platform for Advanced Scientific Computing Conference. New York: IEEE, 2023: DOI: 10.1145/3592979.3593412., articleTitle=FourCastNet: accelerating global high-resolution weather forecasting using adaptive fourier neural operators, refAbstract=null), Reference(id=1261377075139756918, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2021, volume=426, issue=null, pageStart=285, pageEnd=298, url=null, language=null, rfNumber=[20], rfOrder=33, authorNames=Castro R, Souto Y M, Ogasawara E, journalName=Neurocomputing, refType=null, unstructuredReference=Castro R, Souto Y M, Ogasawara E, et al. STConvS2S: spatiotemporal convolutional sequence to sequence network for weather forecasting[J]. Neurocomputing, 2021, 426: 285-298., articleTitle=STConvS2S: spatiotemporal convolutional sequence to sequence network for weather forecasting, refAbstract=null), Reference(id=1261377075294946172, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2023, volume=61, issue=null, pageStart=715, pageEnd=726, url=null, language=null, rfNumber=[21], rfOrder=34, authorNames=Pirone D, Cimorelli L, Del G G, journalName=Journal of Hydrology, refType=null, unstructuredReference=Pirone D, Cimorelli L, Del G G, et al. Short-term rainfall forecasting using cumulative precipitation fields from station data: a probabilistic machine learning approach[J]. Journal of Hydrology, 2023, 61: 715-726., articleTitle=Short-term rainfall forecasting using cumulative precipitation fields from station data: a probabilistic machine learning approach, refAbstract=null), Reference(id=1261377075395609474, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2004, volume=null, issue=5, pageStart=623, pageEnd=33, url=null, language=null, rfNumber=[22], rfOrder=35, authorNames=陈德辉, 薛纪善, journalName=气象学报, refType=null, unstructuredReference=陈德辉, 薛纪善. 数值天气预报业务模式现状与展望[J]. 气象学报, 2004(5): 623-33., articleTitle=数值天气预报业务模式现状与展望, refAbstract=null), Reference(id=1261377075634684813, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2004, volume=null, issue=5, pageStart=623, pageEnd=633, url=null, language=null, rfNumber=[22], rfOrder=36, authorNames=Chen Dehui, Xue Jishan, journalName=Acta Meteorologica Sinica, refType=null, unstructuredReference=Chen Dehui, Xue Jishan. Current status and prospects of numerical weather prediction operational models[J]. Acta Meteorologica Sinica, 2004(5): 623-633., articleTitle=Current status and prospects of numerical weather prediction operational models, refAbstract=null), Reference(id=1261377075831817104, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2020, volume=78, issue=3, pageStart=451, pageEnd=476, url=null, language=null, rfNumber=[23], rfOrder=37, authorNames=沈学顺, 王建捷, 李泽椿, journalName=气象学报, refType=null, unstructuredReference=沈学顺, 王建捷, 李泽椿, . 中国数值天气预报的自主创新发展[J]. 气象学报, 2020, 78(3): 451-476., articleTitle=中国数值天气预报的自主创新发展, refAbstract=null), Reference(id=1261377076196721564, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2020, volume=78, issue=3, pageStart=451, pageEnd=476, url=null, language=null, rfNumber=[23], rfOrder=38, authorNames=Shen Xueshun, Wang Jianjie, Li Zechun, journalName=Journal of Meteorology, refType=null, unstructuredReference=Shen Xueshun, Wang Jianjie, Li Zechun, et al. Independent innovation and development of numerical weather prediction in China[J]. Journal of Meteorology, 2020, 78(3): 451-476., articleTitle=Independent innovation and development of numerical weather prediction in China, refAbstract=null), Reference(id=1261377076435796896, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2023, volume=43, issue=6, pageStart=1958, pageEnd=1968, url=null, language=null, rfNumber=[24], rfOrder=39, authorNames=董润婷, 吴利, 王晓英, journalName=计算机应用, refType=null, unstructuredReference=董润婷, 吴利, 王晓英, . 深度学习在天气预报领域的应用分析及研究进展综述[J]. 计算机应用, 2023, 43(6): 1958-1968., articleTitle=深度学习在天气预报领域的应用分析及研究进展综述, refAbstract=null), Reference(id=1261377076595180456, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2023, volume=43, issue=6, pageStart=1958, pageEnd=1968, url=null, language=null, rfNumber=[24], rfOrder=40, authorNames=Dong Runting, Wu Li, Wang Xiaoying, journalName=Computer Applications, refType=null, unstructuredReference=Dong Runting, Wu Li, Wang Xiaoying, et al. Application analysis and research progress of deep learning in the field of weather forecasting: a review[J]. Computer Applications, 2023, 43(6): 1958-1968., articleTitle=Application analysis and research progress of deep learning in the field of weather forecasting: a review, refAbstract=null), Reference(id=1261377076737786799, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2019, volume=45, issue=3, pageStart=426, pageEnd=436, url=null, language=null, rfNumber=[25], rfOrder=41, authorNames=孙全德, 焦瑞莉, 夏江江, journalName=气象, refType=null, unstructuredReference=孙全德, 焦瑞莉, 夏江江, . 基于机器学习的数值天气预报风速订正研究[J]. 气象, 2019, 45(3): 426-436., articleTitle=基于机器学习的数值天气预报风速订正研究, refAbstract=null), Reference(id=1261377078470034356, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2019, volume=45, issue=3, pageStart=426, pageEnd=436, url=null, language=null, rfNumber=[25], rfOrder=42, authorNames=Sun Quande, Jiao Ruili, Xia Jiangjiang, journalName=Meteorology, refType=null, unstructuredReference=Sun Quande, Jiao Ruili, Xia Jiangjiang, et al. Research on wind speed correction of numerical weather prediction based on machine learning[J]. Meteorology, 2019, 45(3): 426-436., articleTitle=Research on wind speed correction of numerical weather prediction based on machine learning, refAbstract=null), Reference(id=1261377078742664122, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2020, volume=31, issue=4, pageStart=494, pageEnd=503, url=null, language=null, rfNumber=[26], rfOrder=43, authorNames=陈昱文, 黄小猛, 李熠, journalName=应用气象学报, refType=null, unstructuredReference=陈昱文, 黄小猛, 李熠, . 基于ECMWF产品的站点气温预报集成学习误差订正[J]. 应用气象学报, 2020, 31(4): 494-503., articleTitle=基于ECMWF产品的站点气温预报集成学习误差订正, refAbstract=null), Reference(id=1261377078994322366, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2020, volume=31, issue=4, pageStart=494, pageEnd=503, url=null, language=null, rfNumber=[26], rfOrder=44, authorNames=Chen Yuwen, Huang Xiaomeng, Li Yi, journalName=Journal of Applied Meteorology, refType=null, unstructuredReference=Chen Yuwen, Huang Xiaomeng, Li Yi, et al. Error correction of site temperature forecasts using ensemble learning based on ECMWF products[J]. Journal of Applied Meteorology, 2020, 31(4): 494-503., articleTitle=Error correction of site temperature forecasts using ensemble learning based on ECMWF products, refAbstract=null), Reference(id=1261377079128540098, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, doi=null, pmid=null, pmcid=null, year=2016, volume=97, issue=1, pageStart=49, pageEnd=67, url=null, language=null, rfNumber=[27], rfOrder=45, authorNames=Swinbank R, Kyouda M, Buchanan P, journalName=Bulletin of the American Meteorological Society, refType=null, unstructuredReference=Swinbank R, Kyouda M, Buchanan P, et al. The TIGGE project and its achievements[J]. Bulletin of the American Meteorological Society, 2016, 97(1): 49-67., articleTitle=The TIGGE project and its achievements, refAbstract=null)], funds=[Fund(id=1261377048350736993, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, awardId=42171348, language=CN, fundingSource=国家自然科学基金(42171348), fundOrder=null, country=null), Fund(id=1261377048812110439, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, awardId=D040405, language=CN, fundingSource=国防科工局重点项目(D040405), fundOrder=null, country=null)], companyList=[AuthorCompany(id=1261377012200030450, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, xref=1, ext=[AuthorCompanyExt(id=1261377012241973492, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, companyId=1261377012200030450, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 School of Geological Engineering and Geomatics, Chang’an University, Xi’an 710054, China), AuthorCompanyExt(id=1261377012267139316, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, companyId=1261377012200030450, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 长安大学地质工程与测绘学院, 西安 710054)]), AuthorCompany(id=1261377012791427319, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, xref=2, ext=[AuthorCompanyExt(id=1261377012824981752, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, companyId=1261377012791427319, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 School of Land Engineering, Chang’an University, Xi’an 710054, China), AuthorCompanyExt(id=1261377012988559610, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, companyId=1261377012791427319, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 长安大学土地工程学院, 西安 710054)]), AuthorCompany(id=1261377013554790657, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, xref=3, ext=[AuthorCompanyExt(id=1261377013571567875, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, companyId=1261377013554790657, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3 Key Laboratory of Land Consolidation, Shaanxi Province, Xi’an 710054, China), AuthorCompanyExt(id=1261377013579956484, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, companyId=1261377013554790657, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3 陕西省土地整治重点实验室, 西安 710054)])], figs=[ArticleFig(id=1261377030738854291, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=EN, label=Fig.1, caption=Map of the study area and distribution of stations, figureFileSmall=sc5dqZ5crry89QQGSH7XFg==, figureFileBig=Swr9aCEVDKKjdg1+oqN1Uw==, tableContent=null), ArticleFig(id=1261377031271530903, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=CN, label=图1, caption=研究区域位置及站点分布图, figureFileSmall=sc5dqZ5crry89QQGSH7XFg==, figureFileBig=Swr9aCEVDKKjdg1+oqN1Uw==, tableContent=null), ArticleFig(id=1261377031967785384, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=EN, label=Fig.2, caption=Flow chart of pre-training method for regional precipitation information based on meteorological macromodel with Bayesian optimization, figureFileSmall=7kzgawQruulY5QycwSwCiQ==, figureFileBig=TBIPwjGLZjN4m4ungjPskw==, tableContent=null), ArticleFig(id=1261377032752120243, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=CN, label=图2, caption=基于气象大模型与贝叶斯优化的区域降水信息预训练方法流程图

EAR5为高精度气象再分析数据;FourCastNet为气象大模型;AFNO为一种基于傅里叶空间操作的神经网络骨干;CNN为利用贝叶斯优化的卷积神经网络;LSTM为长短期记忆网络;loss为损失函数;longitudes为经度;latitudes为纬度;backbone为主干网络 ;whl分别为像素宽度、像素高度和变量维度

, figureFileSmall=7kzgawQruulY5QycwSwCiQ==, figureFileBig=TBIPwjGLZjN4m4ungjPskw==, tableContent=null), ArticleFig(id=1261377034077520312, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=EN, label=Fig.3, caption=Schematic diagram of the network architecture adopted by the Institute, figureFileSmall=35YOo2LJIOQzSZPHKBmHVA==, figureFileBig=VTQYDtGopXq/Ql6wvTt8kQ==, tableContent=null), ArticleFig(id=1261377034543088070, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=CN, label=图3, caption=研究所采用网络架构示意图, figureFileSmall=35YOo2LJIOQzSZPHKBmHVA==, figureFileBig=VTQYDtGopXq/Ql6wvTt8kQ==, tableContent=null), ArticleFig(id=1261377035230953929, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=EN, label=Fig.4, caption=Comparison of continuity metrics between the two methods and the traditional numerical model, figureFileSmall=RwyUkvynLy4AG6IHekm1tw==, figureFileBig=plEkzljuAYWRN3uI76rDHQ==, tableContent=null), ArticleFig(id=1261377035700715987, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=CN, label=图4, caption=两种方法与传统数值模式的连续性指标对比, figureFileSmall=RwyUkvynLy4AG6IHekm1tw==, figureFileBig=plEkzljuAYWRN3uI76rDHQ==, tableContent=null), ArticleFig(id=1261377036497633754, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=EN, label=Fig.5, caption=Statistical classification metric evaluation results for different models by station, figureFileSmall=ZnTkTA5/kHtuFo6nWMlXVA==, figureFileBig=I3xACE53Wqcly6JkeJmgqw==, tableContent=null), ArticleFig(id=1261377038296990179, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=CN, label=图5, caption=不同模式分站点统计分类指标检验结果

Count为各指标值出现的次数;TS为风险评分;FAR为误报率;POD为探测率

, figureFileSmall=ZnTkTA5/kHtuFo6nWMlXVA==, figureFileBig=I3xACE53Wqcly6JkeJmgqw==, tableContent=null), ArticleFig(id=1261377038754169326, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=EN, label=Fig.6, caption=Comparison of RMSE and TS for daily precipitation forecast results of different models across different seasons, figureFileSmall=+F8BLCfnRaLJaZ2cDlKaFw==, figureFileBig=CTypj/4ee51VoI+xt3h9WQ==, tableContent=null), ArticleFig(id=1261377039429452282, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=CN, label=图6, caption=不同模式对不同季节的日降水量预报结果RMSE和TS对比, figureFileSmall=+F8BLCfnRaLJaZ2cDlKaFw==, figureFileBig=CTypj/4ee51VoI+xt3h9WQ==, tableContent=null), ArticleFig(id=1261377040071180803, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=EN, label=Fig.7, caption=Comparison of BIAS, RMSE and TS scores of different models for different precipitation intensity thresholds, figureFileSmall=5L5+Cm3DWUInuUJnwpti/Q==, figureFileBig=bfc85CpeLmtZAalzzg7rJA==, tableContent=null), ArticleFig(id=1261377040809378319, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=CN, label=图7, caption=不同模式对于不同降水强度阈值下的BIAS、RMSE和TS评分对比, figureFileSmall=5L5+Cm3DWUInuUJnwpti/Q==, figureFileBig=bfc85CpeLmtZAalzzg7rJA==, tableContent=null), ArticleFig(id=1261377042742952479, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=EN, label=Table 1, caption=

Overview of traditional numerical models

, figureFileSmall=null, figureFileBig=null, tableContent=
预报模式 起报时间
(UTC)
集合成
员数
初始扰
动方法
预报时间
长度/d
CMA 00/12 30 BVs 0~10
ECMWF 00/12 51 EDA-SVINI 0~16
KMA 00/12 26 ETKF 0~12
NCEP 00/06/12/18 31 BV-ETR 0~16
UKMO 00/12 18 ETKF 0~15
), ArticleFig(id=1261377042914918951, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=CN, label=表1, caption=

传统数值模式简介

, figureFileSmall=null, figureFileBig=null, tableContent=
预报模式 起报时间
(UTC)
集合成
员数
初始扰
动方法
预报时间
长度/d
CMA 00/12 30 BVs 0~10
ECMWF 00/12 51 EDA-SVINI 0~16
KMA 00/12 26 ETKF 0~12
NCEP 00/06/12/18 31 BV-ETR 0~16
UKMO 00/12 18 ETKF 0~15
), ArticleFig(id=1261377043653116468, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=EN, label=Table 2, caption=

Continuity evaluation metrics of the two forecasting methods and the traditional numerical model

, figureFileSmall=null, figureFileBig=null, tableContent=
模式 MAE RMSE BIAS CC
CMA 2.62 7.15 0.36 0.508 132
FourCastNet_CNN 1.61 4.10 0.14 0.806 948
ECMWF 2.35 6.23 0.19 0.555 629
KMA 2.34 6.45 0.08 0.491 785
FourCastNet_LSTM 1.41 4.16 -0.06 0.799 116
NCEP 2.19 6.01 0.03 0.539 585
UKMO 2.37 6.41 0.08 0.491 640
), ArticleFig(id=1261377044471005760, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=CN, label=表2, caption=

两种预报方法与传统数值模式连续性检验指标

, figureFileSmall=null, figureFileBig=null, tableContent=
模式 MAE RMSE BIAS CC
CMA 2.62 7.15 0.36 0.508 132
FourCastNet_CNN 1.61 4.10 0.14 0.806 948
ECMWF 2.35 6.23 0.19 0.555 629
KMA 2.34 6.45 0.08 0.491 785
FourCastNet_LSTM 1.41 4.16 -0.06 0.799 116
NCEP 2.19 6.01 0.03 0.539 585
UKMO 2.37 6.41 0.08 0.491 640
), ArticleFig(id=1261377044991099466, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=EN, label=Table 3, caption=

Classification metric evaluation results of the two forecasting methods and the traditional numerical model

, figureFileSmall=null, figureFileBig=null, tableContent=
模式 TS FAR POD
CMA 0.42 0.54 0.83
FourCastNet_CNN 0.33 0.67 0.97
ECMWF 0.41 0.57 0.88
KMA 0.42 0.55 0.86
FourCastNet_LSTM 0.48 0.50 0.90
NCEP 0.42 0.55 0.87
UKMO 0.42 0.55 0.85
), ArticleFig(id=1261377045699936847, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261262690375353074, language=CN, label=表3, caption=

两种预报方法与传统数值模式分类指标检验结果

, figureFileSmall=null, figureFileBig=null, tableContent=
模式 TS FAR POD
CMA 0.42 0.54 0.83
FourCastNet_CNN 0.33 0.67 0.97
ECMWF 0.41 0.57 0.88
KMA 0.42 0.55 0.86
FourCastNet_LSTM 0.48 0.50 0.90
NCEP 0.42 0.55 0.87
UKMO 0.42 0.55 0.85
)], attaches=null, journal=Journal(id=1146119176004939786, delFlag=0, nameCn=科学技术与工程, nameEn=Science Technology and Engineering, nameHistory1=null, nameHistory2=null, issn=1671-1815, eissn=, cn=11-4688/T, coden=null, periodic=4, language=CN, oaType=是, 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=UKU/O7GSka5polgCTkbIIw==, journalPrice=null, startedYear=null, abbrevIsoEn=Sci Technol Eng, journalRemark=null, publicationField=null, createdTime=null, updatedTime=1754445529766, createdBy=null, updatedBy=13701087609, firstLetterCn=S, firstLetterEn=S, subjectCode=Natural Sciences, subjectName=自然科学, subjectCodeEn=Natural Sciences, subjectNameEn=null, picCn=UKU/O7GSka5polgCTkbIIw==, picEn=5hwlULoNwcbj3xUmVi9MAQ==, jcr=null, cjcr=null, exts=[JournalExt(id=1159791870395564357, language=CN, name=科学技术与工程, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=null, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=http://www.stae.com.cn/jsygc/home, createdTime=1754445529793, updatedTime=1754445529793, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=http://www.stae.com.cn/jsygc/site/menus/20090429150146001, submissionAuthorUrl=http://www.stae.com.cn/jsygc/author/login, submissionEditorUrl=http://www.stae.com.cn/jsygc/editor/login, submissionReviewUrl=http://www.stae.com.cn/jsygc/reviewer/login, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1159791870441701702, language=EN, name=Science Technology and Engineering, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=null, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=http://www.stae.com.cn/jsygc/home, createdTime=1754445529804, updatedTime=1754445529804, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=http://www.stae.com.cn/jsygc/author/login, submissionEditorUrl=http://www.stae.com.cn/jsygc/editor/login, submissionReviewUrl=http://www.stae.com.cn/jsygc/reviewer/login, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1146123166801305609, websiteList=[Website(id=1148243202391400884, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1146123166801305609, 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/kxjsygc/CN, language=CN, createTime=1751692112777, createBy=18614031015, updateTime=1753520965431, updateBy=18614031015, name=科学技术与工程-中文站点, tplId=1146099689490845704, title=科学技术与工程, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1148622798802673703, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=articleTextType, value=kx, createTime=1751782615614, updateTime=1751782615614, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798781702180, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=banner, value=null, createTime=1751782615609, updateTime=1751782615609, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798769119267, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=logo, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic?fileId=j86gbwi+p0Idkyl5SzIlmQ==, createTime=1751782615606, updateTime=1751782615606, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798794285094, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic, createTime=1751782615612, updateTime=1751782615612, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798790090789, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1751782615611, updateTime=1751782615611, creator=18614031015, updator=18614031015)]), Website(id=1155914124811976731, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1146123166801305609, 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/kxjsygc/EN, language=EN, createTime=1753521003206, createBy=18614031015, updateTime=1753521003206, updateBy=18614031015, name=科学技术与工程-英文站点, tplId=1146101810881728533, title=Science Technology and Engineering, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1155914371227308235, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=articleTextType, value=kx, createTime=1753521061952, updateTime=1753521061952, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371210531016, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=banner, value=null, createTime=1753521061947, updateTime=1753521061947, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371202142407, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=logo, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic?fileId=j86gbwi+p0Idkyl5SzIlmQ==, createTime=1753521061945, updateTime=1753521061945, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371223113930, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic, createTime=1753521061950, updateTime=1753521061950, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371218919625, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1753521061949, updateTime=1753521061949, creator=18614031015, updator=18614031015)])], journalTitle=科学技术与工程, weixinUrl=null, journalUrl=null, iacademicId=null, status=0, seqNo=null, journalTitleEn=Science Technology and Engineering, journalPhotoCn=UKU/O7GSka5polgCTkbIIw==, journalPhotoEn=5hwlULoNwcbj3xUmVi9MAQ==, journalFirstLetter=S, 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=null, provinceCode=null, provinceName=null, collectFlag=false), detailUrlCn=https://castjournals.cast.org.cn/joweb/kxjsygc/CN/10.12404/j.issn.1671-1815.2406929, detailUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/10.12404/j.issn.1671-1815.2406929, pdfUrlCn=https://castjournals.cast.org.cn/joweb/kxjsygc/CN/PDF/10.12404/j.issn.1671-1815.2406929, pdfUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/PDF/10.12404/j.issn.1671-1815.2406929, aliStartDate=null, aliEndDate=null, collectionFlag=false, citedCount=null, citedUrl=null, reference=null)
收藏切换
利用气象大模型和区域预训练调优进行日降水预报:以陕西省为例
收藏切换
PDF下载
王浩宇 1 , 韩玲 2, 3, * , 李良志 2
科学技术与工程 | 论文·天文学、地球科学 2025,25(20): 8379-8391
收起
收藏切换
科学技术与工程 | 论文·天文学、地球科学 2025, 25(20): 8379-8391
利用气象大模型和区域预训练调优进行日降水预报:以陕西省为例
全屏
王浩宇1 , 韩玲2, 3, * , 李良志2
作者信息
  • 1 长安大学地质工程与测绘学院, 西安 710054
  • 2 长安大学土地工程学院, 西安 710054
  • 3 陕西省土地整治重点实验室, 西安 710054
  • 王浩宇(2000—),男,汉族,山东滕州人,硕士研究生。研究方向:遥感信息提取。E-mail:

通讯作者:

* 韩玲(1964—),女,汉族,辽宁沈阳人,博士,教授,博士研究生导师。研究方向:遥感信息提取、资源环境遥感和遥感影像智能识别。E-mail:
Daily Precipitation Forecasting Using Global Weather Model and Regional Pre-training Optimization: A Case Study in Shaanxi Province
Hao-yu WANG1 , Ling HAN2, 3, * , Liang-zhi LI2
Affiliations
  • 1 School of Geological Engineering and Geomatics, Chang’an University, Xi’an 710054, China
  • 2 School of Land Engineering, Chang’an University, Xi’an 710054, China
  • 3 Key Laboratory of Land Consolidation, Shaanxi Province, Xi’an 710054, China
出版时间: 2025-07-18 doi: 10.12404/j.issn.1671-1815.2406929
文章导航
收藏切换

为提高降水预报的精度并解决传统数值模式在预报精度和运算效率上的瓶颈问题,通过结合气象大模型和深度学习后处理方法,对陕西省2008—2018年的降水预报进行实例研究。以FourCastNet气象大模型输出的气象要素场为基础,利用贝叶斯优化的卷积神经网络(convolutional neural network,CNN)/长短期记忆网络(long short-term memory, LSTM)算法构建气象要素场-区域降水信息的预训练模型,生成高分辨率的日降水预报。结果表明:该方法在空间分辨率和预报精度上优于传统数值天气预报模式;区域预训练调优预报结果更准确地反映了区域降水的时空分布;基于贝叶斯优化的深度学习后处理算法能够有效缓解初始场偏差对预报的影响。可见,气象大模型结合深度学习后处理算法在降水精准预报中具有显著的应用潜力,为防灾减灾、农业生产及水资源管理提供了科学支持。

降水预报  /  深度学习  /  气象大模型  /  预训练

To improve the accuracy of precipitation forecasts and address the limitations of traditional numerical weather prediction models in forecast precision and computational efficiency, a meteorological large model was combined with a deep learning post-processing approach was combined. A case study was conducted for precipitation forecasts over Shaanxi Province during 2008—2018. Based on meteorological variable fields output by the FourCastNet model, a pre-trained model mapping meteorological fields to regional precipitation was constructed using Bayesian-optimized convolutional neural networks (CNN)/long short-term memory (LSTM) networks. The results indicate that this method outperforms traditional numerical weather prediction models in terms of spatial resolution and forecast accuracy. The regionally fine-tuned forecasts more accurately capture the spatiotemporal distribution of precipitation. Furthermore, the Bayesian-optimized deep learning post-processing algorithm effectively mitigates the impact of initial field biases on forecast results. These findings demonstrate the significant potential of integrating meteorological large models with deep learning post-processing algorithms for accurate precipitation forecasting, providing scientific support for disaster prevention, agricultural production, and water resource management.

precipitation forecasting  /  deep learning  /  large meteorological models  /  pre-training
王浩宇, 韩玲, 李良志. 利用气象大模型和区域预训练调优进行日降水预报:以陕西省为例. 科学技术与工程, 2025 , 25 (20) : 8379 -8391 . DOI: 10.12404/j.issn.1671-1815.2406929
Hao-yu WANG, Ling HAN, Liang-zhi LI. Daily Precipitation Forecasting Using Global Weather Model and Regional Pre-training Optimization: A Case Study in Shaanxi Province[J]. Science Technology and Engineering, 2025 , 25 (20) : 8379 -8391 . DOI: 10.12404/j.issn.1671-1815.2406929
高质量高时空分辨率的降水预报信息在水文监测预报、水资源规划管理和洪涝干旱监测等领域发挥重要作用,还可用于监测中小尺度极端暴雨事件及其引发的洪水、地质灾害等防灾减灾工作[1-4]。然而,受大气状态、所处地理位置以及下垫面条件等众多因素共同影响,降水存在显著的时空变异性,因此降水预报数据在上述应用中会带来很大的不确定性[5-8]。如何提高降水预报数据的空间分辨率和准确性是近年来气象水文等领域研究热点和难点。
目前,根据理论基础和数据来源,降水预报方法主要分为:基于数值模式的方法和基于数据驱动的方法[9-11]。基于数值模式的方法正面临着精度和速度方面的挑战,近年来其精度提升缓慢,且运算时间严重影响了预报的时效性。随着深度学习理论的不断进步,基于数据驱动的气象大模型在没有显式定义大气动力学方程的情况下,也能够通过学习大量历史数据,建立当前时刻与未来时刻的气象变量间的复杂关系[12-16]。其中,以华为的盘古气象大模型[17-18]、Nvidia的FourCastNet气象大模型[19]为典型代表,在精度上可与传统数值模式相媲美,且计算速度极快,仅需几秒,即可输出全球未来数天的高分辨率预报结果。
气象大模型通常采用再分析资料或雷达卫星资料进行训练,然而,这些训练数据与区域实际降水之间的映射结果存在一定程度的偏差。如何科学合理利用区域降水观测信息从而增强降水预报的可靠性减小不确定性[12,20 -21],是研究降水过程预报时空变异性与解决观测-预报不确定性的关键问题。
近年来,随着人工智能技术的快速发展,深度学习后处理方法逐渐成为提升数值天气预报(numerical weather prediction,NWP)结果精度的重要手段[22-23]。传统的数值模式虽然能够模拟大尺度天气过程,但在区域尺度的降水预报中往往存在显著偏差,尤其是在复杂地形和局地对流系统影响下,预报结果难以准确反映降水的时空分布特征。为解决这一问题,研究者们尝试通过深度学习模型对数值模式输出进行后处理,从而优化降水预报结果[24]。如基于卷积神经网络(convolutional neural network,CNN)的研究,利用其强大的空间特征提取能力,对预报降水场与观测降水数据建立映射关系,改善了降水预报的空间分辨率。此外,长短期记忆网络(long short-term memory, LSTM)作为一种时间序列建模工具,能够有效捕捉降水预报结果误差随时间变化的动态特征,在降水预报校正展现出良好的效果[25]。然而,现有的后处理方法大多依赖单一数据源如ERA5再分析数据(ECMWF Reanalysis v5)或局部观测数据,且未充分考虑模型超参数和网络层权重对于不同区域降水预报任务的适用性,难以充分结合全球气象大模型的预报能力与区域观测信息,从而导致模型泛化性能受限[26]
基于此,提出了一种基于气象大模型和区域降水信息预训练的降水预报方法,以提升降水预报结果的精度并减小其不确定性。该方法以FourCastNet气象大模型为基础模型,构造了基于贝叶斯优化算法的卷积神经网络和长短期记忆网络的预训练模型。与现有方法相比,新方法优势包括:①利用CNN与LSTM,通过预训练建立了区域降水信息与气象参量的映射关系;②通过贝叶斯优化技术,对神经网络的关键超参数(如隐藏单元数、层数、学习率等)进行自动搜索,提高模型精度与可靠性;③设计预训练模型以充分利用FourCastNet的预报能力,适配无观测资料区域的降水预报需求。以陕西省为研究区域,选取常见的TIGGE[THORPEX(全球观测系统研究与可预报性试验)Interactive Grand Global Ensemble]中心的传统数值模式输出作为对比数据源,使用气象大模型区域预训练调优方法对陕西省区域进行降水预报任务,并将计算结果与传统数值模式预报结果比较,验证新方法的可行性和高效性。
陕西地处中国中部黄河中游地区,经纬度范围为105°29'E~111°15'E,31°42'N~39°35'N(图1),总面积20.6×104 km2。陕西省北部有黄土高原,南方有秦巴山地,中部为关中平原,地形复杂,南北气候差异极其显著。全省降水空间分布不均,年降水量330~1 250 mm,由南至北递减,同时,全省降水年内分配不均,多集中在夏季,多由几次暴雨形成,干旱和洪涝灾害频繁发生,在气候变化背景下,灾害频率逐渐增加。精确且高分辨率的降水预报结果有助于陕西省的防灾减灾和农业生产。
(1)气象数据。气象数据来自中国气象局国家气象信息中心“中国地面气候资料日值数据V3.0”(http://cdc.cma.gov.cn/),主要包括1957—2019年中国2 474个国家级地面站气压、气温、降水量、相对湿度、风向风速和日照时数等日值数据。选取陕西省内2008年1月1日—2018年12月31日期间33个气象站点逐日降水数据,所有数据经过内部一致性检查、极值检查以及空间一致性检查等严格质量控制后,得到每日降水量。
(2)ERA5再分析数据。ERA5数据集(ECMWF Reanalysis 5th Generation)是欧洲中期天气预报中心(European Centre for Medium-range Weather Forecasts,ECMWF)推出的第五代再分析产品,通过先进模式和数据同化系统,整合了丰富的历史观测资料。通过CDS(Climate Data Store)(https://cds.climate.copernicus.eu/)下载全球区域2008年1月—2018年12月逐小时的ERA5数据,空间分辨率为0.25°×0.25°,作为FourCastNet气象大模型推理预报的初始场。
(3)数值预报降水数据。数值预报降水数据收集自TIGGE(交互式全球大集合预报系统)[27](https://apps.ecmwf.int/)。TIGGE数据是由世界气象组织(World Meteorological Organization,WMO)主导的全球集合预报系统,旨在通过多模式、多中心的集合预报数据共享,提高全球数值天气预报的准确性和可靠性。选用预见期24 h,空间分辨率统一处理为0.25°×0.25°的多模式预报产品,包括中国气象局(China Meteorological Administration,CMA)、ECMWF、韩国气象局(Korea Meteorological Administration,KMA)、美国国家环境与预报中心(National Centers for Environmental Prediction,NCEP)和英国气象局(United Kingdom Meteorological Office,UKMO)等模式预报产品(表1)。将其作为研究所采用的对比数据,时间跨度为2008年1月—2018年12月。
(4)DEM数据。采用的DEM数据为SRTM (Shuttle Radar Topography Mission) DEM V4.1数据集,从地理空间数据云(http://www.gscloud.cn/)获取,空间分辨率为90 m。
提出一种多层次、集成化的降水预报框架,旨在弥补现有气象大模型预报结果与区域实际观测之间的偏差,并进一步提升预报的可靠性与精度。具体而言,该方法基于FourCastNet气象大模型与区域降水观测数据,通过结合深度学习算法与贝叶斯优化策略,构建具备预训练与区域调优功能的降水预报系统。框架能够有效整合气象大模型在气象要素场趋势外推优势与区域降水观测数据对局地降水过程的精细化表征,从而提高降水预报的时空分辨率与准确度。方法的理论依据与构建思路如下。
选取FourCastNet气象大模型作为基础模型。FourCastNet 是一种基于自适应傅里叶神经算子(adaptive Fourier neural operator,AFNO)的深度学习模型,已在全球气象要素预报中展现出显著的效率与准确性。其核心优势在于:通过傅里叶变换实现大规模空间特征的高效建模,能够在全球尺度上生成高分辨率的气象要素场。该模型的输入数据源于ERA5(ECMWF第五代再分析数据),具备长期历史数据支撑,可通过端到端的深度学习网络进行气象变量的快速推理。
在此基础上,对2008—2018年陕西省逐6 h的气象要素场进行推理计算,生成历史气象场的次日预报结果。气象要素场的变量包括温度、湿度、风速、地面压力等,这些数据为后续区域降水映射提供输入特征,可表示为
${X}_{\mathrm{F}\mathrm{o}\mathrm{u}\mathrm{r}\mathrm{C}\mathrm{a}\mathrm{s}\mathrm{t}\mathrm{N}\mathrm{e}\mathrm{t}}={f}_{\mathrm{F}\mathrm{o}\mathrm{u}\mathrm{r}\mathrm{C}\mathrm{a}\mathrm{s}\mathrm{t}\mathrm{N}\mathrm{e}\mathrm{t}}\left({X}_{\mathrm{E}\mathrm{R}\mathrm{A}5}\right)$
式(1)中:XERA5为ERA5初始场;fFourCastNet为气象大模型推理函数;XFourCastNet为外推得到的气象要素场。
气象大模型输出结果虽然在全球尺度上表现出良好的拟合效果,但由于其初始场为再分析资料与区域实际降水之间存在固有的偏差,直接依赖其结果进行区域降水预测,难以满足实际需求。因此,必须进一步结合区域降水观测数据,建立区域尺度的降水映射关系。
为弥补气象大模型在区域尺度降水预报中的不足,引入陕西省实测站点降水数据,利用深度学习方法建立气象要素场与区域降水的非线性映射关系。具体而言,通过构建CNN和LSTM的模型,学习气象要素场与降水量之间的时空关联特征。其中CNN能够有效提取输入数据的局部空间特征,适用于气象要素场与降水数据之间的空间映射建模。通过卷积层和池化层的逐级特征提取,CNN能够捕捉气象要素在不同尺度下的空间分布模式。
LSTM是一种适合时间序列建模的神经网络,能够处理长期依赖关系并对气象要素随时间变化的动态特征进行建模。LSTM 网络可进一步学习降水量的时间演变规律,提升模型的时效性与准确性。
利用FourCastNet输出的气象要素场数据与陕西省实测站点降水数据,构建贝叶斯优化的CNN模型和LSTM模型,用于学习气象要素场与区域降水的非线性映射关系,可表示为
${\widehat{y}}_{i}=g({X}_{i};\theta)$
式(2)中:${\widehat{y}}_{i}$为预报降水量;Xi为气象要素特征场;g(·)为深度学习模型;θ为模型超参数。
为确保映射模型的高效性与鲁棒性,采用贝叶斯优化(Bayesian optimization)技术对CNN与LSTM模型的超参数进行自动化搜索与优化。贝叶斯优化通过构建概率模型,对隐藏单元数、网络层数、学习率等关键超参数进行全局最优搜索,并以验证集上的损失函数[如均方误差(mean squared error,MSE)]作为优化目标,最终选择出性能最佳的模型参数组合。该优化策略不仅提高了模型的收敛速度,还能有效增强模型的泛化能力,优化目标函数为验证集上的损失,可表示为
$\tau =\mathrm{M}\mathrm{S}\mathrm{E}(\widehat{y},y)+\lambda \mathrm{R}\mathrm{e}\mathrm{g}\left(\theta \right)$
式(3)中:MSE为均方误差;$\widehat{y}$为预报降雨量;y为实测降雨量;Reg(θ)为正则化项;λ为正则化系数。
将站点划分为70%用于训练(预训练数据集),30%用于验证(检验站点)。通过交叉验证与对比实验,生成高分辨率的日降水预报结果,并与传统数值模式输出(如ECMWF模式)进行比较。
所提方法主要包括气象大模型推理与预训练区域调优的双阶段建模过程。具体如下。
(1)气象大模型推理。利用FourCastNet气象大模型生成高分辨率的气象要素场数据。FourCastNet基于ERA5再分析数据进行长期训练,具备捕捉全球气象要素场的能力。本阶段针对2008—2018年陕西省逐6 h的气象要素场数据进行推理计算,获得包括温度、湿度、风速、地面压力等变量在内的次日气象预报数据,为后续区域建模提供基础输入特征。
(2)预训练及区域调优。将FourCastNet输出的高分辨率气象要素场数据与陕西省实测站点的降水观测数据相结合,构建CNN和LSTM的深度学习模型。具体而言,首先通过预训练阶段初始化CNN/LSTM模型,以捕捉气象变量与区域降水之间的初步映射关系;随后,通过区域调优阶段进一步将实测降水数据纳入模型训练,采用贝叶斯优化技术对关键超参数(如隐藏单元数、层数、学习率等)进行优化调整,从而实现区域尺度下气象要素与降水量映射关系的精确拟合。方法的整体流程如图2所示,包括气象大模型推理、区域预训练调优与对比验证两个主要步骤。采用的网络架构如图3所示。
FourCastNe是一种基于AFNO (adaptive fourier neural operator) 的高分辨率天气预报模型,专为处理高分辨率输入数据而设计。其架构整合了近年来深度学习领域的多项关键进展,包括 FNO(Fourier neural operator) 和ViT(Vision Transformer)。其关键特性如下。
(1)高分辨率支持。AFNO 模型利用快速傅里叶变换(fast Fourier transform,FFT)在傅里叶域中执行空间混合操作,将复杂度从传统自注意力机制的O(N2)降低到 $O\left(N\mathrm{l}\mathrm{o}\mathrm{g}N\right),$使其能够高效处理高分辨率数据,例如当前 0.25°的全球网格(720×1 440分辨率)。
(2)FNO (Fourier neural operator)。FNO 是一种连续算子学习方法,最初用于模拟湍流系统(如 Navier-Stokes 方程),在解决复杂偏微分方程(partial differential equation,PDE)时表现优异。将其应用于气象数据建模,为 FourCastNet 提供了强大的基础算子能力。
(3)ViT (Vision Transformer)。ViT 架构通过多头自注意力机制建模特征间的全局交互,在计算机视觉任务中表现卓越。然而,传统 ViT 在处理高分辨率输入时计算代价较高。AFNO 通过傅里叶域操作解决了这一问题。
(4)模块化架构。AFNO 模型采用模块化设计,包括空间混合(傅里叶域操作)和通道混合[多层感知机(MLP)],并通过残差连接提高训练稳定性。
FourCastNet模型的计算流程如下。
步骤1 输入投影与分块。原始输入是 720×1 440的经纬度网格,首先被分割为h×w 个小块(patches),每个小块为 p×p (如p=8),并嵌入高维空间中形成d维的 tokens,同时加入位置编码。
步骤2 空间混合(Spatial Mixing)。空间混合过程在傅里叶域中执行。
离散傅里叶变换(FFT):将小块 token 转换到傅里叶域,可表示为
${z}_{m,n}=[\mathrm{F}\mathrm{F}\mathrm{T}{\left(X\right)]}_{m,n}$
式(4)中:zm,n为傅里叶域中的 token;m,n为patch的空间位置索引;FFT(X)为二维离散傅里叶变换操作;X为输入m×n的二维矩阵。
权重应用与稀疏化可表示为
$\left\{\begin{array}{l}{\tilde{z}}_{m,n}={S}_{\lambda }\left[\mathrm{M}\mathrm{L}\mathrm{P}\right({z}_{m,n}\left)\right]\\ {S}_{\lambda }\left(x\right)=\mathrm{s}\mathrm{i}\mathrm{g}\mathrm{n}\left(x\right)\mathrm{m}\mathrm{a}\mathrm{x}(\left|x\right|-\lambda,0)\end{array}\right.$
式(5)中:Sλ(x)为稀疏化操作;λ为控制稀疏化强度的阈值参数;MLP(${{{z}_{m}}_{,}}_{n}$)为两层多层感知机,用于学习频域特征;sign(x)为符号函数,用于提取输入x的符号信息;max(·)为取较大值函数,用于阈值化操作或确保结果不小于某个下限值;x为某个权重值,用于在神经网络剪枝或权重稀疏化中进行相应处理。
此过程通过多层感知机(MLP)对 token 赋予权重,并通过软阈值操作控制稀疏性。
然后将稀疏化后的频域$\tilde{Z}$通过逆傅里叶变换(IFFT)返回到补丁域,并添加残差连接,可表示为
${y}_{\mathrm{m},\mathrm{n}}=[\mathrm{I}\mathrm{F}\mathrm{F}\mathrm{T}{\left(\tilde{Z}\right)]}_{m,n}+{X}_{\mathrm{m},\mathrm{n}}$
步骤3 通道混合(Channel Mixing)。每个token的通道维度通过独立的 MLP 进行混合,以进一步提取特征。
步骤4 输出解码。整个过程在多层AFNO模块中重复,每层输出结果作为下一层的输入。最终通过线性解码器(linear decoder)将嵌入特征转换为目标时间步的气象要素场。
贝叶斯优化(Bayesian Optimization)是一种高效的全局优化方法,特别适用于高维、复杂和计算开销大的黑箱函数优化问题。在机器学习模型的超参数优化中,贝叶斯优化能够有效地找到近似最优的超参数组合,显著提升模型性能。贝叶斯优化算法应用于随机森林模型超参数优化的原理和流程如下。
贝叶斯优化的核心思想是将待优化的目标函数视为一个黑箱函数,通过构建该函数的概率模型来指导搜索过程。主要步骤如下。
步骤1 初始化样本点。选择初始超参数点集合$\{{x}_{i}{\}}_{i=1}^{n}$及其对应的目标函数值$\{{y}_{i}{\}}_{i=1}^{n}。$
步骤2 构建和更新高斯过程。使用初始样本点,构建高斯过程模型,计算均值μn(x)和方差函数σn(x)。
步骤3 优化采集函数。
${x}_{n+1}=\mathrm{a}\mathrm{r}\mathrm{g}\underset{x\in X}{\mathrm{m}\mathrm{a}\mathrm{x}}\alpha \left[\left.x\right|{\mu }_{n}\right(x),{\sigma }_{n}(x\left)\right]$
式(7)中:α(·)为选定的采集函数,如期望改进 (expected improvement,EI)或上置信界 (upper confidence bound,UCB))。
步骤4 评估新点。评估新的超参数点${{x}_{n}}_{+1}$,得到目标函数值${{y}_{n}}_{+1}$,并将其添加到数据集中。
步骤5 迭代更新。更新高斯过程模型,重复优化采集函数、评估新点和更新模型的过程,直到满足预定的停止条件。
通过上述步骤,贝叶斯优化能够高效探索超参数空间,找到提升随机森林模型性能的最佳或近似最佳超参数组合。
选择先验分布:初始化一个关于目标函数的先验概率分布,通常使用高斯过程(Gaussian process, GP)。高斯过程是一种非参数化的贝叶斯模型,用于估计目标函数的分布。
CNN是一种针对图像或序列识别建模的前馈神经网络,其由卷积层和全连接层组成,通过卷积和池化学习输入数据的二维结构关系,在计算机视觉领域表现较好。模型相应参数:卷积层(Conv1d)、卷积核大小(kernel_size)、步长(stride)、补零(padding)、激活函数(ReLU)、正则化(Dropout)、线性函数(Linear)、输入节点(in_features)、输出节点(out_features)。
经贝叶斯优化算法确定CNN网络结构如下:①Conv1d[1, 172, kernel_size=(4,), stride=(1,), padding=(2,)],ReLU(), Dropout(p=0.43);②Conv1d[172, 417, kernel_size=(3,), stride=(1,), padding=(1,)], ReLU(), Dropout(p=0.317);③Conv1d[417, 140, kernel_size=(4,), stride=(1,), padding=(2,)], ReLU(), Dropout(p=0.358); ④Conv1d[140, 265, kernel_size=(3,), stride=(1,), padding=(1,)], ReLU(), Dropout(p=0.294); ⑤Conv1d[265, 511, kernel_size=(5,), stride=(1,), padding=(2,)], ReLU(), Dropout(p=0.481);⑥ (fc): Linear(in_features=43 946, out_features=1)。
LSTM是一种针对时间序列建模的递归神经网络(recurrent neural network,RNN),为解决常规递归神经网络先前输入信息经常存在衰减或指数爆炸现象即梯度消失现象,针对难以学习序列间时间依赖关系的难题,其网络结构可以对输入信息进行选择性保留,相较于常规RNN的隐藏层,LSTM模型的基本结构为记忆细胞单元,从而可以提取具有序列长期依赖关系的数据特征信息。经贝叶斯优化算法确定CNN网络结构为LSTMModel((lstm): LSTM(84,227), (fc): Linear(in_features=227, out_features=1)。
通过卷积神经网络和长短期记忆神经网络方法将FourCastNet模型输出的大气状态变量与地面观测降水建立映射关系从而得到高精度高分辨率的降水空间状态预报场,即FourCastNet_CNN和FourCastNet_LSTM模型。
为全面反映降水预报产品在日尺度上的精度特征,采用连续性和分类指标对各模式降水预报数据进行评价。其中连续性指标用于描述降水预报数据对日降水量和过程总体吻合程度和误差特征,包括:相关系数(correlation coefficient,CC)、偏差(BIAS)、平均绝对误差(mean absolute error,MAE)和均方根误差(root mean square error,RMSE);分类指标反映降水预报结果对日降水事件发生与否的估计能力,包括:误报率FAR、探测率POD和风险评分TS。上述指标的计算公式分别为
$MAE=\frac{1}{N}\stackrel{N}{\sum _{i=1}}\left|{P}_{i}-{S}_{i}\right|$
$\mathrm{B}\mathrm{I}\mathrm{A}\mathrm{S}=\frac{S-P}{P}$
$RMSE={\left[\frac{1}{N}\stackrel{N}{\sum _{i=1}}({S}_{i}-{P}_{i}{)}^{2}\right]}^{\frac{1}{2}}$
$CC=\frac{\stackrel{n}{\sum _{i=1}}({P}_{i}-\stackrel{-}{\mathrm{P}})({S}_{i}-\stackrel{-}{\mathrm{S}})}{\sqrt{\stackrel{n}{\sum _{i=1}}({P}_{i}{-\stackrel{-}{P})}^{2}}\sqrt{\stackrel{n}{\sum _{i=1}}({S}_{i}{-\stackrel{-}{S})}^{2}}}$
式中:Pi为地面雨量站点第i时段的观测降水量,mm;Si为降水预报结果的第i时段降水量,mm;N为所拥有的样本总数(数据点的个数);SP分别为降水预报结果和地面雨量站观测的全时段降水量总和$\stackrel{-}{P}$$\stackrel{-}{S}$分别为地面雨量站点观测降水量和降水预报数据降水量的平均值。
$\mathrm{F}\mathrm{A}\mathrm{R}=\frac{F}{H+F}$
$\mathrm{T}\mathrm{S}=\frac{H}{H+M+F}$
$\mathrm{P}\mathrm{O}\mathrm{D}=\frac{H}{H+M}$
式中:H为准确捕捉到的降水事件数,即降水预报结果和地面雨量站点观测降水数据均发生降水事件的频数;M为漏报的降水事件数,即地面雨量站点观测降水数据发生降水事件而降水预报结果未预报的频数;F为误报的降水事件数,即为降水预报结果认为发生降水事件而地面雨量站点观测降水数据未监测到的频数。
需要注意的是根据实测降水数据,研究认为降水量大于0.1 mm为有雨。探测率POD、误报率FAR和风险评分TS 3个指标的取值区间均为0~1,其中POD和TS越大、FAR越小,表明降水预报结果对降水事件的预报能力越高,反之则越低。
为了评估FourCastNet_CNN模型和FourCastNet_LSTM,模型和传统数值模式对日降水量的预报性能,利用33个站点的实测降水进行评估。首先,分别计算FourCastNet_CNN、FourCastNet_LSTM和各模式降水预报结果与地面雨量站点实测降水的连续性指标,如图4所示,4幅图分别表示各模式在各站点的MAE、RMSE、BIAS及CC。可以看出,相较于传统数值模式,FourCastNet_CNN和FourCastNet_LSTM对日降水量的预报性能提升较大,其中FourCastNet_LSTM对总体的BIAS控制最好。在传统数值模式中,ECMWF和NCEP总体表现较好。
表2给出了各模式在全部站点的连续性指标结果,进一步评估其综合性能。在传统数值模式中,NCEP有着更低的MAE、RMSE和BIAS,说明其对降水过程的预报较为可靠,但在CC指标上,ECMWF(0.556)在传统数值模式中领先。与传统数值模式相比,FourCastNet_CNN和FourCastNet_LSTM模型每一项指标提升较大,尤其是在CC指标上,从0.49~0.55提升至约0.8。两者对比,FourCastNet_CNN模型结果有着更高的MAE,却有更低的RMSE,这可能意味着FourCastNet_CNN模型结果对于高降水量事件的预报更为准确,将从不同量级进行展开分析。FourCastNet_LSTM模型结果有着更佳的BIAS指标(-0.06),且与FourCastNet_CNN模型结果呈较大的正偏相反,呈微弱负偏。
表3给出了不同模型在全部站点对日降水事件预报的分类检验指标结果,可以看出,FourCastNet_LSTM模型在误报率FAR、探测率POD和风险评分TS均表现出较高的性能,传统数值模式之间预报性能差异不大。FourCastNet_CNN模型结果有着较低的TS评分和较高的FAR评分,但却有着最高的POD(0.97),说明在学习区域降水信息时过度追求了探测率而引发较高的虚警率和风险。而相对于其他模型,FourCastNet_LSTM模型能够更准确地避免误报降水事件(FAR=0.5)。综合来看,FourCastNet_LSTM模型在分类检验中表现最好,在传统数值模式中NCEP相对表现最好。这些结果表明,与传统数值模式对比,FourCastNet_LSTM模型在各项指标上取得了较好的结果,是可靠的选择,特别是在对降水事件进行准确探测和分类时,而FourCastNet_CNN模型在不同指标上表现迥异,还需要进一步调优平衡其在不同方面的表现。
图5给出了不同模式对日降水量预报性能在不同站点的分类评价指标变化图,竖轴为评分在同一区间的站次。
总体来看,FourCastNet_CNN模型的TS评分在大多数站点低于0.35,而且有着最糟糕的FAR评分表现,大多数站点的FAR大于0.6,有的站点甚至达到0.8。而FourCastNet_LSTM模型在各分类指标上表现更为稳定可靠,相较于传统数值模式仍保持一定的优势。这有可能与CNN和LSTM模型因理论基础和网络结构不同所适用的目标场景和任务也不同有关。
图6给出了不同模式在春季、夏季、秋季和冬季的RMSE和TS指标评估结果。可以看出,不同模式的TS评分年内分布规律为先增大再减小趋势,TS评分均在冬季均达到最低值。比较不同模式的TS评分值可以大致看出利用区域降水信息进行预训练的FourCastNet_LSTM模型TS评分相较于传统数值模式显著增大,这说明在学习到区域降水信息时空分布规律后,准确率得到显著提升,具有较好的可靠性;FourCastNet_CNN模型的TS指标在经过网络学习后不升反降,但是RMSE却明显要低于传统数值模式,这可能与CNN模型核心的卷积操作更关注于全局空间特征有关;而对比5种传统数值模式,各模式的预报性能差异较小,其中CMA模式的夏季RMSE相较其他模式稍高。
为了评估不同模式在不同降水强度阈值下的预报性能表现,将日降水强度分为4种类型:小雨(0.1~10 mm)、中雨(10~25 mm)、大雨(25~50 mm)以及暴雨(>50 mm)。图7给出了各模式在不同降水强度阈值下的BIAS、RMSE和TS评分结果。
图7可知,对于BIAS评分,除小雨外,各模式对小雨以上量级的降水事件均出现了低估现象,且随降水强度增加,BIAS偏差也显著增加。值得注意的是,经预训练调优的FourCastNet_CNN模型在小雨事件上的表现未得到明显改善,显著劣于FourCastNet_LSTM模型,但在较强降水事件上的表现优于FourCastNet_LSTM模型。而对于RMSE评分,可以明显看出,随着降水强度增加,各模式的预报误差均随之增大,相较于FourCastNet_LSTM模型,FourCastNet_CNN模型精度提升较为明显。与前文所提及的FourCastNet_CNN模型TS评分表现不佳相反,对于不同降水强度阈值的降水事件,其表现均优于FourCastNet_LSTM模型,这意味着对于降水事件而言,FourCastNet_CNN模型表现较好,而FourCastNet_LSTM模型则更侧重于晴雨识别,两者各有优劣。
为提高降水预报的准确性和可靠性,以FourCastNet气象大模型为基础,构造了基于贝叶斯优化算法的CNN和LSTM模型,主要包括模型推理和区域降水信息预训练两个阶段。首先利用ERA5初始场和FourCastNet预训练权重进行模型推理,得到全球次日预报气象要素场,然后以区域降水信息为先验信息,基于贝叶斯优化的CNN和LSTM模型对气象要素场预报进行站点降水映射,生成高空间分辨率和高精度降水预报产品。得出如下主要结论。
(1)借助CNN和LSTM模型构建的区域降水信息预训练模型适用性良好,引入的贝叶斯优化技术,用于对关键超参数(如隐藏单元数、学习率、层数等)进行自动搜索和调优,确保了模型的最优性能。不仅提高了模型训练效率,还进一步提升了模型在不同区域和时间尺度上的适用性。突破了传统数值模式对初始场依赖性强且计算开销大的瓶颈,显著提高了区域降水预报的精度和效率。
(2)提出的基于不同理论基础和网络结构的模型对降水空间分布细节特征和准确性具有不同的特点:CNN更适合捕捉降水的空间分布与具体数值特征,而LSTM在降水事件发生概率(如晴雨判别)上表现突出。通过两种模型的协同作用,有效解决了研究区域因传统数值模式分辨率粗糙和精度不足带来的预报不确定性问题,为区域降水预报产品的精细化开发提供了重要参考。
(3)为验证本文方法的有效性,将其实验结果与传统数值模式(如ECMWF模式)进行了系统性对比。结果表明,基于本文方法的降水预报产品在空间分辨率和预报精度上均优于传统模式,尤其在降水空间分布细节特征的捕捉和区域降水准确性上表现更为出色。这进一步证明了结合气象大模型和区域降水信息进行预训练调优的可行性和优越性。
后续研究中,将使用其他高时空分辨率降水相关因素(如土壤湿度、地表温度、风速等)并融合不同传统数值模式结果,进一步提升降水预报产品时空分辨率和精度。
  • 国家自然科学基金(42171348)
  • 国防科工局重点项目(D040405)
参考文献 引证文献
排序方式:
[1]
Srivastav A L, Dhyani R, Ranjan M, et al. Climate-resilient strategies for sustainable management of water resources and agriculture[J]. Environmental Science and Pollution Research, 2021, 28(31): 41576-41595.
[2]
赵铜铁钢, 张弛, 田雨, . 全球气象预报驱动流域水文预报研究进展与展望[J]. 水科学进展, 2024, 35(1): 156-166.
Zhao Tongtiegang, Zhang Chi, Tian Yu, et al. Progress and prospects ofglobal meteorological forecasting to drive basin hydrological forecasting[J]. Advances in Water Science, 2024, 35(1): 156-166.
[3]
周宏博, 孙树林, 尚文涛, . 降雨雨型和强度对土边坡稳定性影响分析[J]. 科学技术与工程, 2012, 12(11): 2602-2606.
Zhou Hongbo, Sun Shulin, Shang Wentao, et al. Analysis of the impact of rainfall patterns and intensity on soil slope stability[J]. Science Technology and Engineering, 2012, 12(11): 2602-2606.
[4]
许浩然, 杨兆前, 陈中举, . 湖南省极端气候时空特征分析及风险评估[J]. 科学技术与工程, 2024, 24(9): 3543-3553.
Xu Haoran, Yang Zhaoqian, Chen Zhongju, et al. Analysis of spatiotemporal characteristics and risk assessment of extreme climate in Hunan Province[J]. Science Technology and Engineering, 2024, 24(9): 3543-3553.
[5]
金荣花, 代刊, 赵瑞霞, . 我国无缝隙精细化网格天气预报技术进展与挑战[J]. 气象, 2019, 45(4): 445-457.
Jin Ronghua, Dai Kan, Zhao Ruixia, et al. Progress and challenges of seamless fine-grid weather forecasting technology in China[J]. Meteorological Monthly, 2019, 45(4): 445-457.
[6]
钟琦, 张智察, 姚秀萍, . 应用物理结合的人工智能方法提升“21·7”河南极端强降水预报[J]. 中国科学: 地球科学, 2024, 54(5): 1680-1701.
Zhong Qi, Zhang Zhicha, Yao Xiuping, et al. Improving the forecast of the “21.7” Henan extreme heavy rainfall using AI methods combined with applied physics[J]. Science China: Earth Sciences, 2024, 54(5): 1680-1701.
[7]
凌敏华, 韩洪宝. 1960—2018年河南省降水时空变化特征及重心[J]. 科学技术与工程, 2021, 21(17): 7008-7016.
Ling Minhua, Han Hongbao. Spatiotemporal variation characteristics and center of precipitation in Henan Province from 1960 to 2018[J]. Science Technology and Engineering, 2021, 21(17): 7008-7016.
[8]
王宗敏, 王治中, 杨瑶, . 多时间尺度下遥感降水产品与再分析降水产品在海河流域适用性对比分析[J]. 科学技术与工程, 2021, 21(6): 2186-2193.
Wang Zongmin, Wang Zhizhong, Yang Yao, et al. Comparative analysis of the applicability of remote sensing precipitation products and reanalysis precipitation products in the Haihe River Basin at multiple time scales[J]. Science Technology and Engineering, 2021, 21(6): 2186-2193.
[9]
毕宝贵, 代刊, 王毅, . 定量降水预报技术进展[J]. 应用气象学报, 2016, 27(5): 534-549.
Bi Baogui, Dai Kan, Wang Yi, et al. Progress in quantitative precipitation forecasting technology[J]. Journal of Applied Meteorology, 2016, 27(5): 534-549.
[10]
杨绚, 代刊, 朱跃建. 深度学习技术在智能网格天气预报中的应用进展与挑战[J]. 气象学报, 2022, 80(5): 649-667.
Yang Xun, Dai Kan, Zhu Yuejian. Progress and challenges of deep learning technology in intelligent grid weather forecasting[J]. Acta Meteorologica Sinica, 2022, 80(5): 649-667.
[11]
张同, 鲍艳松, 陆其峰. IASI卫星资料同化对江淮暴雨预报的试验研究[J]. 科学技术与工程, 2016, 16(6): 9-16, 36.
Zhang Tong, Bao Yansong, Lu Qifeng. Experimental study on the assimilation of IASI satellite data forthe forecasting of torrential rains in the Yangtze-Huaihe River Basin[J]. Science Technology and Engineering, 2016, 16(6): 9-16, 36.
[12]
Reichstein M, Camps-Valls G, Stevens B, et al. Deep learning and process understanding for data-driven Earth system science[J]. Nature, 2019, 566(7743): 195-204.
[13]
Ren X L, Li X Y, Ren K J, et al. Deep learning-based weather prediction: a survey[J]. Big Data Research, 2021, 23: 100178.
[14]
黄小猛, 林岩銮, 熊巍, . 数值预报AI气象大模型国际发展动态研究[J]. 大气科学学报, 2024, 47(1): 46-54.
Huang Xiaomeng, Lin Yanluan, Xiong Wei, et al. International development trends of numerical weather prediction AI large models[J]. Journal of Atmospheric Sciences, 2024, 47(1): 46-54.
[15]
张弛, 陈国兴, 杨洪涛. 利用深度学习预报美国东北部日降水分布[J]. 大气科学学报, 2024, 47(1): 55-64.
Zhang Chi, Chen Guoxing, Yang Hongtao. Forecasting daily precipitation distribution in the northeastern United States using deep learning[J]. Journal of Atmospheric Sciences, 2024, 47(1): 55-64.
[16]
王坤, 蒋宁, 李敏, . 基于SMOTE算法和逻辑回归模型算法的江苏短时强降水潜势预报[J]. 科学技术与工程, 2020, 20(28): 11447-11454.
Wang Kun, Jiang Ning, Li Min, et al. Short-term heavy precipitation potential forecasting in Jiangsu based on the SMOTE algorithm and logistic regression model[J]. Science Technology and Engineering, 2020, 20(28): 11447-11454.
[17]
Bi K F, Xie L X, Zhang H H, et al. Accurate medium-range global weather forecasting with 3D neural networks[J]. Nature, 2023, 619(7970): 533-538.
[18]
田奇, 毕恺峰, 谢凌曦. 基于人工智能大模型的中期全球气象预报新方法[J]. 中国基础科学, 2024, 26(1): 7-13, 21.
Tian Qi, Bi Kaifeng, Xie Lingxi. A new method for medium-range global meteorological forecasting based on artificial intelligence large models[J]. Chinese Journal of Basic Science, 2024, 26(1): 7-13, 21.
[19]
Kurth T, Subramanian S, Harrington P, et al. FourCastNet: accelerating global high-resolution weather forecasting using adaptive fourier neural operators[C]// Proceedings of the Platform for Advanced Scientific Computing Conference. New York: IEEE, 2023: DOI: 10.1145/3592979.3593412.
[20]
Castro R, Souto Y M, Ogasawara E, et al. STConvS2S: spatiotemporal convolutional sequence to sequence network for weather forecasting[J]. Neurocomputing, 2021, 426: 285-298.
[21]
Pirone D, Cimorelli L, Del G G, et al. Short-term rainfall forecasting using cumulative precipitation fields from station data: a probabilistic machine learning approach[J]. Journal of Hydrology, 2023, 61: 715-726.
[22]
陈德辉, 薛纪善. 数值天气预报业务模式现状与展望[J]. 气象学报, 2004(5): 623-33.
Chen Dehui, Xue Jishan. Current status and prospects of numerical weather prediction operational models[J]. Acta Meteorologica Sinica, 2004(5): 623-633.
[23]
沈学顺, 王建捷, 李泽椿, . 中国数值天气预报的自主创新发展[J]. 气象学报, 2020, 78(3): 451-476.
Shen Xueshun, Wang Jianjie, Li Zechun, et al. Independent innovation and development of numerical weather prediction in China[J]. Journal of Meteorology, 2020, 78(3): 451-476.
[24]
董润婷, 吴利, 王晓英, . 深度学习在天气预报领域的应用分析及研究进展综述[J]. 计算机应用, 2023, 43(6): 1958-1968.
Dong Runting, Wu Li, Wang Xiaoying, et al. Application analysis and research progress of deep learning in the field of weather forecasting: a review[J]. Computer Applications, 2023, 43(6): 1958-1968.
[25]
孙全德, 焦瑞莉, 夏江江, . 基于机器学习的数值天气预报风速订正研究[J]. 气象, 2019, 45(3): 426-436.
Sun Quande, Jiao Ruili, Xia Jiangjiang, et al. Research on wind speed correction of numerical weather prediction based on machine learning[J]. Meteorology, 2019, 45(3): 426-436.
[26]
陈昱文, 黄小猛, 李熠, . 基于ECMWF产品的站点气温预报集成学习误差订正[J]. 应用气象学报, 2020, 31(4): 494-503.
Chen Yuwen, Huang Xiaomeng, Li Yi, et al. Error correction of site temperature forecasts using ensemble learning based on ECMWF products[J]. Journal of Applied Meteorology, 2020, 31(4): 494-503.
[27]
Swinbank R, Kyouda M, Buchanan P, et al. The TIGGE project and its achievements[J]. Bulletin of the American Meteorological Society, 2016, 97(1): 49-67.
2025年第25卷第20期
PDF下载
67
32
引用本文
BibTeX
文章信息
doi: 10.12404/j.issn.1671-1815.2406929
  • 接收时间:2024-09-14
  • 首发时间:2026-05-13
  • 出版时间:2025-07-18
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2024-09-14
  • 修回日期:2025-04-27
基金
国家自然科学基金(42171348)
国防科工局重点项目(D040405)
作者信息
    1 长安大学地质工程与测绘学院, 西安 710054
    2 长安大学土地工程学院, 西安 710054
    3 陕西省土地整治重点实验室, 西安 710054

通讯作者:

* 韩玲(1964—),女,汉族,辽宁沈阳人,博士,教授,博士研究生导师。研究方向:遥感信息提取、资源环境遥感和遥感影像智能识别。E-mail:
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/kxjsygc/CN/10.12404/j.issn.1671-1815.2406929
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

Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)

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