基于双偏振雷达的高时空分辨率临近降水预报

doi:10.3785/j.issn.1008-973X.2026.03.013

浙江大学学报(工学版)

2026, Vol. 60

Issue (3): 574-584 DOI: 10.3785/j.issn.1008-973X.2026.03.013

计算机技术、控制工程

基于双偏振雷达的高时空分辨率临近降水预报

林乐平1,2(

),李量1,欧阳宁1,2,*(

)

1. 桂林电子科技大学认知无线电与信息处理省部共建教育部重点实验室，广西桂林 541004
2. 桂林电子科技大学信息与通信学院，广西桂林 541004

High spatio-temporal resolution precipitation forecast based on dual-polarization radar

Leping LIN1,2(

),Liang LI1,Ning OUYANG1,2,*(

)

1. Ministry of Education Key Laboratory of Cognitive Radio and Information Processing, Guilin University of Electronic Technology, Guilin 541004, China
2. School of Information and Communication, Guilin University of Electronic Technology, Guilin 541004, China

全文: PDF(2664 KB) HTML

摘要：

针对现有算法在复杂快变降水外推中的局限性及Z-R转换带来的误差叠加问题，提出联合双偏振雷达变量的多模态临近降水预报模型DPRF-UNet. 该预报模型利用双偏振雷达变量中的反射率因子、差分反射率和比差分相移，直接对降水量进行预测. 通过纳入降水粒子微物理特性的信息，采用端到端的降水预报框架，该模型避免了因经验性转换带来的误差，实现了降水的直接预测. 研究结果表明，与仅使用雷达反射率因子作为输入的主流模型相比，提出DPRF-UNet模型在所有降水阈值下的临界成功指数、Heidke技巧评分和命中率的平均值分别提高了2.6%、2.7%和1.0%，达到0.64、0.72和0.74.

关键词： 临近降水预报; 双偏振雷达; 深度学习; 时空预测; 时空分辨率

Abstract:

The DPRF-UNet model, a multimodal precipitation forecasting model that integrated dual-polarization radar variables, was proposed aiming at the problems that existing algorithms had limitations in extrapolating complex and rapidly changing precipitation and the Z-R conversion led to the error accumulation. Dual-polarization radar variables such as reflectivity, differential reflectivity, and specific differential phase were used to directly predict precipitation. The error introduced by empirical conversion was avoided by incorporating microphysical characteristics and employing an end-to-end precipitation forecasting framework, enabling direct precipitation prediction. The research results show that the proposed DPRF-UNet model demonstrates superior performance compared with mainstream models that solely utilize radar reflectivity as input. The average critical success index, Heidke skill score and probability of detection of the DPRF-UNet model across all precipitation thresholds were increased by 2.6%, 2.7% and 1.0%, reaching 0.64, 0.72 and 0.74.

Key words: precipitation forecast dual-polarization radar deep learning spatio-temporal prediction spatio-temporal resolution

收稿日期: 2025-01-10 出版日期: 2026-02-04

TP 391

基金资助: 国家自然科学基金资助项目(62001133)；广西科技基地和人才专项资助项目(桂科 AD19110060)；广西自然科学基金资助项目(2017GXNSFBA198212)；广西无线宽带通信与信号处理重点实验室基金资助项目(GXKL06200114).

通讯作者: 欧阳宁 E-mail: linleping@guet.edu.cn;ynou@guet.edu.cn

作者简介: 林乐平（1980—），女，教授，从事机器学习、图像信号处理的研究. orcid.org/0000-0002-9842-8421. E-mail：linleping@guet.edu.cn

	服务
	把本文推荐给朋友
	加入引用管理器
	E-mail Alert
	作者相关文章
	林乐平
	李量
	欧阳宁

引用本文:

林乐平,李量,欧阳宁. 基于双偏振雷达的高时空分辨率临近降水预报[J]. 浙江大学学报(工学版), 2026, 60(3): 574-584.

Leping LIN,Liang LI,Ning OUYANG. High spatio-temporal resolution precipitation forecast based on dual-polarization radar. Journal of ZheJiang University (Engineering Science), 2026, 60(3): 574-584.

链接本文:

https://www.zjujournals.com/eng/CN/10.3785/j.issn.1008-973X.2026.03.013 或 https://www.zjujournals.com/eng/CN/Y2026/V60/I3/574

图 1 DPRF-UNet模型参数的优化过程示意图

图 2 交叉注意力空频融合模块的示意图

图 3 融合降水生成模块的示意图

图 4 多尺度注意力卷积模块的示意图

图 5 跨层融合模块的示意图

表 1 深度学习实验的硬件平台及软件环境

表 2 NJU-CPOL数据集的降水强度分类统计

表 3 不同模型在0.5 h前置时间下的实验结果对比

表 4 不同模型在1 h前置时间下的实验结果对比

表 5 不同模型在1.5 h前置时间下的实验结果对比

图 6 不同模型在连续降水过程中的预测散点图对比

图 7 不同前置时间下的模型预测可视化结果对比

图 8 不同前置时间下的模型预测可视化结果对比

表 6 1 h前置时间下的消融实验结果对比

1	RAVURI S, LENC K, WILLSON M, et al Skilful precipitation nowcasting using deep generative models of radar[J]. Nature, 2021, 597 (7878): 672- 677
2	李扬, 刘玉宝, 许小峰基于深度学习改进数值天气预报模式和预报的研究及挑战[J]. 气象科技进展, 2021, 11 (3): 103- 112 LI Yang, LIU Yubao, XU Xiaofeng Advances and challenges for improving numerical weather prediction models and forecasting using deep learning[J]. Advances in Meteorological Science and Technology, 2021, 11 (3): 103- 112
3	REINOSO-RONDINEL R, REMPEL M, SCHULTZE M, et al Nationwide radar-based precipitation nowcasting: a localization filtering approach and its application for Germany[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2022, 15: 1670- 1691 doi: 10.1109/JSTARS.2022.3144342
4	AYZEL G, HEISTERMANN M, WINTERRATH T Optical flow models as an open benchmark for radar-based precipitation nowcasting (rainymotion v0.1)[J]. Geoscientific Model Development, 2019, 12 (4): 1387- 1402 doi: 10.5194/gmd-12-1387-2019
5	SHI X, CHEN Z, WANG H, et al. Convolutional LSTM network: a machine learning approach for precipitation nowcasting [C]∥Advances in Neural Information Processing Systems 28. Montreal: Curran Associates, Inc, 2015: 802-810.
6	俞小鼎, 周小刚, 王秀明雷暴与强对流临近天气预报技术进展[J]. 气象学报, 2012, 70 (3): 311- 337 YU Xiaoding, ZHOU Xiaogang, WANG Xiuming The advances in the nowcasting techniques on thunderstorms and severe convection[J]. Acta Meteorologica Sinica, 2012, 70 (3): 311- 337
7	黄兴友, 马玉蓉, 胡苏蔓基于深度学习的天气雷达回波序列外推及效果分析[J]. 气象学报, 2021, 79 (5): 817- 827 HUANG Xingyou, MA Yurong, HU Suman Extrapolation and effect analysis of weather radar echo sequence based on deep learning[J]. Acta Meteorologica Sinica, 2021, 79 (5): 817- 827
8	马志峰, 张浩, 刘劼基于深度学习的短临降水预报综述[J]. 计算机工程与科学, 2023, 45 (10): 1731- 1753 MA Zhifeng, ZHANG Hao, LIU Jie A survey of precipitation nowcasting based on deep learning[J]. Computer Engineering and Science, 2023, 45 (10): 1731- 1753
9	SHI X, GAO Z, LAUSEN L, et al. Deep learning for precipitation nowcasting: a benchmark and a new model [C]//Advances in Neural Information Processing Systems 30. Red Hook, NY: Cur Associates, Inc., 2017.
10	WANG Y, LONG M, WANG J, et al. PredRNN: recurrent neural networks for predictive learning using spatiotemporal LSTMs [C]//Proceedings of the Neural Information Processing Systems. Long Beach: Curran Associates, Inc, 2017: 879-888.
11	MA Z, ZHANG H, LIU J DB-RNN: an RNN for precipitation nowcasting deblurring[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024, 17: 5026- 5041
12	AYZEL G, SCHEFFER T, HEISTERMANN M RainNet v1.0: a convolutional neural network for radar-based precipitation nowcasting[J]. Geoscientific Model Development, 2020, 13 (6): 2631- 2644
13	RONNEBERGER O, FISCHER P, BROX T. U-Net: convolutional networks for biomedical image segmentation [C]//Medical Image Computing and Computer-Assisted Intervention. Cham: Springer, 2015: 234–241.
14	TREBING K, STAǸCZYK T, MEHRKANOON S SmaAt-UNet: precipitation nowcasting using a small attention-UNet architecture[J]. Pattern Recognition Letters, 2021, 145: 178- 186
15	YANG Y, MEHRKANOON S. AA-TransUNet: attention augmented TransUNet for nowcasting tasks [C]//Proceedings of the International Joint Conference on Neural Networks. Padua: IEEE, 2022: 1–8.
16	FERNÁNDEZ J G, MEHRKANOON S Broad-UNet: multi-scale feature learning for nowcasting tasks[J]. Neural Networks, 2021, 144: 419- 427
17	REN G, SUN Y, SUN H, et al A case study on two differential reflectivity columns in a convective cell: phased-array radar observation and cloud model simulation[J]. Remote Sensing, 2024, 16 (3): 460
18	PAN X, LU Y, ZHAO K, et al Improving nowcasting of convective development by incorporating polarimetric radar variables into a deep-learning model[J]. Geophysical Research Letters, 2021, 48 (21): e2021GL095302
19	ZHU K, CHEN H, HAN L. MCT U-Net: a deep learning nowcasting method using dual-polarization radar observations [C]//IEEE International Geoscience and Remote Sensing Symposium. Kuala Lumpur: IEEE, 2022: 4665–4668.
20	ROMBEEK N, LEINONEN J, HAMANN U Exploiting radar polarimetry for nowcasting thunderstorm hazards using deep learning[J]. Natural Hazards and Earth System Sciences, 2024, 24 (1): 133- 144
21	RYZHKOV A, ZHANG P, BUKOVČIĆ P, et al Polarimetric radar quantitative precipitation estimation[J]. Remote Sensing, 2022, 14 (7): 1695
22	庄潇然, 郑玉, 王亚强, 等基于深度学习的融合降水临近预报方法及其在中国东部地区的应用研究[J]. 气象学报, 2023, 81 (2): 286- 303 ZHUANG Xiaoran, ZHENG Yu, WANG Yaqiang, et al Research on fusion precipitation prediction method based on deep learning and its application in Eastern China[J]. Acta Meteorologica Sinica, 2023, 81 (2): 286- 303
23	CREMONINI R, VOORMANSIK T, POST P, et al Estimation of extreme precipitation events in Estonia and Italy using dual-polarization weather radar quantitative precipitation estimations[J]. Atmospheric Measurement Techniques, 2023, 16 (11): 2943- 2956
24	HUANG H, ZHAO K, ZHANG G, et al Quantitative precipitation estimation with operational polarimetric radar measurements in Southern China: a differential phase–based variational approach[J]. Journal of Atmospheric and Oceanic Technology, 2018, 35 (6): 1253- 1271
25	张永华. 基于深度学习的华南登陆台风偏振雷达定量降水估测研究 [D]. 南京: 南京信息工程大学, 2022. ZHANG Yonghua. Research on quantitative precipitation estimation of landfalling typhoons in South China based on deep learning and polarimetric radar [D]. Nanjing: Nanjing University of Information Science and Technology, 2022.
26	GU X, WANG L, DENG Z, et al Adaptive spatial and frequency experts fusion network for medical image fusion[J]. Biomedical Signal Processing and Control, 2024, 96: 106478
27	COOLEY J W, TUKEY J W An algorithm for the machine calculation of complex Fourier series[J]. Mathematics of Computation, 1965, 19 (90): 297- 301
28	HOU Q, ZHOU D, FENG J. Coordinate attention for efficient mobile network design [C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville: IEEE, 2021: 13708–13717.
29	KINGMA D P, BA J. Adam: a method for stochastic optimization [EB/OL]. (2014-12-22). https://arxiv.org/abs/1412.6980.
30	SMITH L N, TOPIN N. Super-convergence: very fast training of neural networks using large learning rates [C]//Proceedings of the Artificial Intelligence and Machine Learning for Multi-Domain Operations Applications. Baltimore: SPIE, 2019: 36.

[1]	武晓春,郭宁. 基于HMARU-net的隧道渗漏水轻量化检测方法[J]. 浙江大学学报(工学版), 2026, 60(3): 468-477.
[2]	杨明辉,宋牧原,付大喜,郭炎伟,卢贤锥,张文聪,郑伟龙. 基于多头自注意力-Bi-LSTM模型的盾构掘进引发的土体沉降预测[J]. 浙江大学学报(工学版), 2026, 60(2): 415-424.
[3]	朱志航,闫云凤,齐冬莲. 基于扩散模型多模态提示的电力人员行为图像生成[J]. 浙江大学学报(工学版), 2026, 60(1): 43-51.
[4]	孙月,张兴兰. 基于双重引导的目标对抗攻击方法[J]. 浙江大学学报(工学版), 2026, 60(1): 81-89.
[5]	段继忠,李海源. 基于变分模型和Transformer的多尺度并行磁共振成像重建[J]. 浙江大学学报(工学版), 2025, 59(9): 1826-1837.
[6]	王福建,张泽天,陈喜群,王殿海. 基于多通道图聚合注意力机制的共享单车借还量预测[J]. 浙江大学学报(工学版), 2025, 59(9): 1986-1995.
[7]	张弘,张学成,王国强,顾潘龙,江楠. 基于三维视觉的软体机器人实时定位与控制[J]. 浙江大学学报(工学版), 2025, 59(8): 1574-1582.
[8]	王圣举,张赞. 基于加速扩散模型的缺失值插补算法[J]. 浙江大学学报(工学版), 2025, 59(7): 1471-1480.
[9]	章东平,王大为,何数技,汤斯亮,刘志勇,刘中秋. 基于跨维度特征融合的航空发动机寿命预测[J]. 浙江大学学报(工学版), 2025, 59(7): 1504-1513.
[10]	蔡永青,韩成,权巍,陈兀迪. 基于注意力机制的视觉诱导晕动症评估模型[J]. 浙江大学学报(工学版), 2025, 59(6): 1110-1118.
[11]	王立红,刘新倩,李静,冯志全. 基于联邦学习和时空特征融合的网络入侵检测方法[J]. 浙江大学学报(工学版), 2025, 59(6): 1201-1210.
[12]	徐慧智,王秀青. 基于车辆图像特征的前车距离与速度感知[J]. 浙江大学学报(工学版), 2025, 59(6): 1219-1232.
[13]	陈赞,李冉,冯远静,李永强. 基于时间维超分辨率的视频快照压缩成像重构[J]. 浙江大学学报(工学版), 2025, 59(5): 956-963.
[14]	马莉,王永顺,胡瑶,范磊. 预训练长短时空交错Transformer在交通流预测中的应用[J]. 浙江大学学报(工学版), 2025, 59(4): 669-678.
[15]	陈巧红,郭孟浩,方贤,孙麒. 基于跨模态级联扩散模型的图像描述方法[J]. 浙江大学学报(工学版), 2025, 59(4): 787-794.

Viewed

Full text

Abstract

Cited

Shared

Discussed