A self-attention-based neural network for three-dimensional multivariate modeling and its skillful ENSO predictions

doi:10.1126/sciadv.adf2827

IOCAS-IR > 海洋环流与波动重点实验室

	A self-attention-based neural network for three-dimensional multivariate modeling and its skillful ENSO predictions
	Zhou, Lu 1,2,3; Zhang, Rong-Hua4,5
	2023-03-10
发表期刊	SCIENCE ADVANCES
ISSN	2375-2548
卷号	9 期号:10 页码:1
通讯作者	Zhang, Rong-Hua(rzhang@nuist.edu.cn)
摘要	Large biases and uncertainties remain in real-time predictions of El Nino-Southern Oscillation (ENSO) using process-based dynamical models; recent advances in data-driven deep learning algorithms provide a promising mean to achieve superior skill in the tropical Pacific sea surface temperature (SST) modeling. Here, a specific selfattention-based neural network model is developed for ENSO predictions based on the much sought-after Transformer model, named 3D-Geoformer, which is used to predict three-dimensional (3D) upper-ocean temperature anomalies and wind stress anomalies. This purely data-driven and time-space attention-enhanced model achieves surprisingly high correlation skills for Nino 3.4 SST anomaly predictions made 18 months in advance and initiated beginning in boreal spring. Further, sensitivity experiments demonstrate that the 3DGeoformer model can depict the evolution of upper-ocean temperature and the coupled ocean-atmosphere dynamics following the Bjerknes feedback mechanism during ENSO cycles. Such successful realizations of the self-attention-based model in ENSO predictions indicate its great potential for multidimensional spatiotemporal modeling in geoscience.
DOI	10.1126/sciadv.adf2827
收录类别	SCI
语种	英语
资助项目	National Natural Science Foundation of China[42030410] ; National Natural Science Foundation of China[LSL202202402] ; Strategic Priority Research Program of the Chinese Academy of Sciences[XDB40000000] ; Startup Foundation for Introducing Talent of NUIST
WOS研究方向	Science & Technology - Other Topics
WOS类目	Multidisciplinary Sciences
WOS记录号	WOS:000960951400021
出版者	AMER ASSOC ADVANCEMENT SCIENCE
WOS关键词	2015-2016 EL-NINO ; MULTIMODEL ENSEMBLE ; TELECONNECTIONS ; PREDICTABILITY ; VARIABILITY ; FORECASTS ; EVOLUTION ; PROGRESS
引用统计	被引频次：21[WOS] [WOS记录] [WOS相关记录]
文献类型	期刊论文
条目标识符	http://ir.qdio.ac.cn/handle/337002/182736
专题	海洋环流与波动重点实验室
通讯作者	Zhang, Rong-Hua
作者单位	1.Chinese Acad Sci, Inst Oceanol, Key Lab Ocean Circulat & Waves, Qingdao 266071, Peoples R China 2.Chinese Acad Sci, Ctr Ocean Mega Sci, Qingdao 266071, Peoples R China 3.Univ Chinese Acad Sci, Beijing 10029, Peoples R China 4.Nanjing Univ Informat Sci & Technol, Sch Marine Sci, Nanjing 210044, Peoples R China 5.Laoshan Lab, Qingdao 266237, Peoples R China
第一作者单位	海洋环流与波动重点实验室
推荐引用方式 GB/T 7714	Zhou, Lu,Zhang, Rong-Hua. A self-attention-based neural network for three-dimensional multivariate modeling and its skillful ENSO predictions[J]. SCIENCE ADVANCES,2023,9(10):1.
APA	Zhou, Lu,&Zhang, Rong-Hua.(2023).A self-attention-based neural network for three-dimensional multivariate modeling and its skillful ENSO predictions.SCIENCE ADVANCES,9(10),1.
MLA	Zhou, Lu,et al."A self-attention-based neural network for three-dimensional multivariate modeling and its skillful ENSO predictions".SCIENCE ADVANCES 9.10(2023):1.