Prediction of ENSO using multivariable deep learning

作     者：Yue Chen Xiaomeng Huang Jing-Jia Luo Yanluan Lin Jonathon S.Wright Youyu Lu Xingrong Chen Hua Jiang Pengfei Lin Yue Chen;Xiaomeng Huang;Jing-Jia Luo;Yanluan Lin;Jonathon S.Wright;Youyu Lu;Xingrong Chen;Hua Jiang;Pengfei Lin

作者机构：Ministry of Education Key Laboratory for Earth System Modeling and Department for Earth System ScienceTsinghua UniversityBeijingChina Institute for Climate and Application Research(ICAR)/CICFEM/KLME/ILCECNanjing University of Information Science and TechnologyNanjingChina Ocean and Ecosystem Sciences DivisionFisheries and Oceans CanadaBedford Institute of OceanographyDartmouthNova ScotiaCanada National Marine Environmental Forecasting CenterState Oceanic AdministrationBeijingChina State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid DynamicsInstitute of Atmospheric PhysicsChinese Academy of SciencesBeijingChina 

出 版 物：《Atmospheric and Oceanic Science Letters》 (大气和海洋科学快报（英文版）)

年 卷 期：2023年第16卷第4期

页      面：51-56页

核心收录：

学科分类：07[理学] 0707[理学-海洋科学] 

基　　金：supported by the National Natural Science Foundation of China[grant Nos.42125503 and 42075137] the National Key Research and Development Program of China[grant Nos.2020YFA0608000 and 2020YFA0607900] 

主　　题：ENSO预报 深度学习 春季预报障碍 多维时空预报因子 

摘      要：本文基于残差神经网络和观测数据构建了一套深度学习多因子预报测模型,以改进厄尔尼诺-南方涛动(ENSO)的预报.该模型基于最大信息系数进行因子时空特征提取,并根据泰勒图的评估标准可自动确定关键预报因子进行预报.该模型在超前8个月以内的预报性能要优于当前传统的业务预报模式.2011–2018年间,该模型的预报性能优于多模式集成预报的结果.在超前6个月预报时效上,模型预报相关性可达0.82,标准化后的均方根误差仅为0.58°C,多模式集成预报的相关性和标准化后的均方根误差分别为0.70和0.73°C.该模型春季预报障碍问题有所缓解,并且自动选取的关键预报因子可用于解释热带和副热带热动力过程对于ENSO变化的影响.