STUNNER: Radar Echo Extrapolation Model Based on Spatiotemporal Fusion Neural Network

被引:12
作者
Fang, Wei [1 ,2 ,3 ]
Pang, Lin [1 ]
Sheng, Victor S. [4 ,5 ]
Wang, Qiguang
机构
[1] Nanjing Univ Informat Sci & Technol, Engn Res Ctr Digital Forens, Sch Comp & Software, Minist Educ, Nanjing 210044, Peoples R China
[2] Nanjing Joint Inst Atmospher Sci, Key Lab Transportat Meteorol China Meteorol Adm, Nanjing 210041, Peoples R China
[3] Jiangsu Collaborat Innovat Ctr Atmospher Environm, Nanjing 210044, Peoples R China
[4] China Meteorol Adm, Meteorol Cadre Training Inst, Beijing 100081, Peoples R China
[5] Texas Tech Univ, Dept Comp Sci, Lubbock, TX 79409 USA
来源
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING | 2023年 / 61卷
基金
中国国家自然科学基金;
关键词
Radar; Radar imaging; Market research; Extrapolation; Convolution; Transient analysis; Spaceborne radar; Dynamic convolution; high-order nonstationarity; precipitation nowcasting; radar echo extrapolation; two-stream spatiotemporal fusion;
D O I
10.1109/TGRS.2023.3268187
中图分类号
P3 [地球物理学]; P59 [地球化学];
学科分类号
0708 ; 070902 ;
摘要
Radar echo extrapolation based on deep learning is an important method for conducting precipitation nowcasting. Radar echo sequence data have spatiotemporal correlations and nonrigid movements of the radar echo. According to the characteristics of radar data, this study proposes a new spatiotemporal fusion neural network called STUNNER. STUNNER implements a two-stream spatiotemporal fusion strategy to extract and fuse spatial and temporal signals. Specifically, it uses a novel cross-network embedding method to achieve efficient spatiotemporal fusion; the fusion integrates a temporal differencing network (TDN) and a spatiotemporal trajectory network (STTN). The TDN models the high-order nonstationarity of the radar sequence to learn the motion trend. The STTN optimizes the convolution operator in a spatiotemporal long short-term memory model to extract the transient variations from the radar images. We compare STUNNER with the other five models on two public datasets. On the radar echo extrapolation task, STUNNER achieves the best performance.
引用
收藏
页数:14
相关论文
共 31 条
[1]   RainNet v1.0: a convolutional neural network for radar-based precipitation nowcasting [J].
Ayzel, Georgy ;
Scheffer, Tobias ;
Heistermann, Maik .
GEOSCIENTIFIC MODEL DEVELOPMENT, 2020, 13 (06) :2631-2644
[2]   Optical flow models as an open benchmark for radar-based precipitation nowcasting (rainymotion v0.1) [J].
Ayzel, Georgy ;
Heistermann, Maik ;
Winterrath, Tanja .
GEOSCIENTIFIC MODEL DEVELOPMENT, 2019, 12 (04) :1387-1402
[3]  
Cramer H., 1961, P 4 BERK S MATH STAT, V2, P57
[4]   Developing Deep Learning Models for Storm Nowcasting [J].
Cuomo, Joaquin ;
Chandrasekar, V. .
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, 2022, 60
[5]   Deformable Convolutional Networks [J].
Dai, Jifeng ;
Qi, Haozhi ;
Xiong, Yuwen ;
Li, Yi ;
Zhang, Guodong ;
Hu, Han ;
Wei, Yichen .
2017 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV), 2017, :764-773
[6]  
De Brabandere B, 2016, ADV NEUR IN, V29
[7]   DeepVideoMVS: Multi-View Stereo on Video with Recurrent Spatio-Temporal Fusion [J].
Duzceker, Arda ;
Galliani, Silvano ;
Vogel, Christoph ;
Speciale, Pablo ;
Dusmanu, Mihai ;
Pollefeys, Marc .
2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021, 2021, :15319-15328
[8]  
Jaderberg M, 2015, ADV NEUR IN, V28
[9]   3D Convolutional Neural Networks for Human Action Recognition [J].
Ji, Shuiwang ;
Xu, Wei ;
Yang, Ming ;
Yu, Kai .
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2013, 35 (01) :221-231
[10]  
Jing JR, 2020, INT CONF ACOUST SPEE, P4142, DOI [10.1109/ICASSP40776.2020.9054232, 10.1109/icassp40776.2020.9054232]