杨绚,代刊,朱跃建. 2022. 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, (0):-, doi:[doi]
深度学习技术在智能网格天气预报中的应用进展与挑战
YANG Xuan DAI Kan ZHU Yuejian
投稿时间:2021-11-19  修订日期:2022-04-06
DOI:
中文关键词:  天气预报,智能网格预报,深度学习技术,统计后处理,模式解释应用
英文关键词:Intelligent grid weather forecasting, Deep learning, Statistical post-process technologies, Model-based interpretation and application
基金项目:国家重点研发计划(2021YFC3000905和2017YFC1502004),中国工程院咨询研究项目(FWC2014)
作者单位邮编
杨绚 国家气象中心 100081
代刊 国家气象中心 100081
朱跃建 美国国家环境预报中心/环境模式中心 20740
摘要点击次数: 28
全文下载次数: 17
中文摘要:
      我国智能网格天气预报已初步建立0~30天涵盖基本气象要素的无缝隙气象预报业务体系。近年深度学习技术兴起,给不同领域带来前所未有的变革。[目的]同样,深度学习的非线性映射能力、海量信息提取能力、时空建模能力等优势为进一步提升智能网格预报的准确性和精细化水平提供了新的思路和方法。[结果]越来越多的研究将深度学习技术应用于智能网格预报的各个方面,包括数值预报订正和解释应用、集合天气预报、相似集合、统计降尺度、纯数据驱动的预报模型和极端天气预报等,并展示出良好的应用潜力。然而,目前深度学习技术在天气预报领域的应用仍处于起步阶段,将其引入智能网格预报业务体系还面临诸多挑战,主要包括算法的选择、算法的数据基础、多源数据融合以及模型的可解释性、可信度、可用性和工程化等。通过回顾近年来深度学习技术在智能网格预报中的应用进展和前景,同时对面临的挑战与应对进行探讨,将有利于促进深度学习技术在天气客观预报领域更好、更稳定的发展。
英文摘要:
      Intelligent grid weather forecasting has preliminary established a seamless weather forecasting system covering basic weather elements from 0 to 30 days in China. The rise of deep learning technology in recent years has brought unprecedented changes to every fields. The advantages of deep learning, such as nonlinear mapping capability, massive information extraction capability, and spatio-temporal modeling capability, provide new methods to further improve the accuracy and refinement of intelligent grid forecasting. An increasing number of studies have applied deep learning techniques to various aspects of intelligent grid forecasting, including statistical postprocessing, ensemble weather forecast, analog ensemble, statistical downscaling, pure data-driven forecast models and extreme weather forecast, and have demonstrated exciting application potential. However, application of deep learning techniques in grid weather forecasting is still in its infancy. There are many challenges, mainly including selection algorithms、dataset、blending multi-source data、interpretability、reliability、availability and engineering process of models. By reviewing the progress and prospects of the application of deep learning in seamless intelligent grid forecasting in recent years, and discussing the challenges and measures, it will be beneficial to promote better development of objective weather forecasting techniques applying deep learning.
查看全文   查看/发表评论  下载PDF阅读器
分享按钮