A Study on Radar Echo Nowcasting Based on Convolutional Gated Recurrent Unit Neural Network

  • Xunlai CHEN ,
  • Jun LIU ,
  • Qunfeng ZHEN ,
  • Xutao LI ,
  • Jia LIU ,
  • Xiyang JI ,
  • Yuanzhao CHEN ,
  • Yunming YE
Expand
  • <sup>1.</sup>Shenzhen Meteorological Bureau,Shenzhen 518040,Guangdong,China;<sup>2.</sup>Shenzhen Key laboratory of severe weather in south China,Shenzhen 518040,Guangdong,China;<sup>3.</sup>Harbin Institute of Technology (Shenzhen),Shenzhen 518055,Guangdong,China

Received date: 2019-09-20

  Online published: 2021-04-28

Abstract

At present, the extrapolation forecast based on radar echoes is the mainstay of disaster weather 0~2 hours nowcasting.This paper proposes a convolutional gated recurrent unit neural network (ConvGRU) by using radar mosaics at 6 min intervals obtained from the radar images provided by 11 doppler radars in Guangdong Province from 2015 to 2018.Through the automatic learning of massive data, the inherent characteristics of the data and the contained physical laws can be discovered using the proposed network.A multi-loss function weighting and hierarchical weighting strategy are proposed.Based on the ConvGRU framework, a three-layer self-encoding model (Encoder-Decoder) is built for training to establish a radar echo prediction model which predicts radar echoes for 20 consecutive frames in the next 2 hours by 6 minutes.The results are compared with the operationally applied methods including tracking radar echoes by correlation (TREC), optical flow, and particle filter using typical case analysis and long-term verification.All the subjective and objective evaluation results indicate that the proposed ConvGRU method shows better forecasting performance in severe convective weather systems in predicting radar echo position, intensity and shape than other methods.These results indicate that the deep learning method can better grasp the characteristics of the strong echo area, and predict the strong echo accurately to a certain extent by automatic learning of time-series radar echo data.For the long-term evaluation results, the ConvGRU method has higher critical success index (CSI) and probability of detection (POD) scores than those of the traditional TREC, optical flow and particle filtering methods, and has the lowest false alarm rate (FAR) scores among all methods, suggesting it could be widely used in operational applications.However, the deep learning-based method has the limitation of losing spatial detail information in radar echoes due to the up-sampling and down-sampling operators, and the prediction performance of stratiform cloud precipitation is relatively poor.

Key words: Nowcasting; ConvGRU; radar echo

Cite this article

Xunlai CHEN , Jun LIU , Qunfeng ZHEN , Xutao LI , Jia LIU , Xiyang JI , Yuanzhao CHEN , Yunming YE . A Study on Radar Echo Nowcasting Based on Convolutional Gated Recurrent Unit Neural Network[J]. Plateau Meteorology, 2021 , 40(2) : 411 -423 . DOI: 10.7522/j.issn.1000-0534.2020.00023

References

[1]Arulampalam M S, Maskell S, Gordon N, et al, 2002.A tutorial on Particle fillers for online noalineor[J].IEEE Transactions on Signal Processing, 50(2): 174-188
[2]Austin G L, Bellon A, 1982.Very short-range forecast of precipitation by the objective extrapolation of radar and satellite data.[M] In: Browing K., Ed, Nowcasting Academic Press, 177-190.
[3]Bechini R, Chandrasekar V, 2017.An enhanced optical flow technique for radar nowcasting of precipitation and winds[J].Journal of Atmospheric and Ocean Technology, 34(12): 2637-2658.
[4]Carpenter J R, Clifford P, Fearnhead P, 1997.Efficient implementation of particle filters for non-linear systems[C]//The 4<sup>th</sup> Interim Report, DRA contract WSS/U1172.Department of Statistics, Oxford University, Oxford, 639.
[5]Chen Y Z, Lan H P, Chen X L, et al, 2017, A nowcasting technique based on application of the particle filter blending algorithm[J].Journal of Meteorological Research, 31(5), 931-945.
[6]Cho K, Van Merrienboer B, Gulcehre C, et al, 2014.Learning phrase representations using RNN encoder-decoder for statistical machine translation[C]//Conference on Empirical Methods in Natural Language Processing, Doha, 25-29 October 2014.1724-1734.
[7]Gibson J, 1950.The ecological approach to visual perception[M].Boston: Houghton Mifflin.
[8]Golding B W, 1998.Nimrod-A system for generating automated very short range forecast[J].Meteorlogical Applications, 5: 1-16.
[9]Dixon M, Wiener G, 1993.TITAN: Thunderstorm identification, tracking, analysis, and nowcasting-A radar based methodology[J].Journal of Atmospheric and Oceanic Technology, 10: 785-797.
[10]Johnson J T, Mackeen P L, Witt A, et al, 1998, The storm cell identification and tracking algorithm: An enhanced WSR-88D algorithm[J].Weather and Forecasting, 13: 263-276.
[11]Li L, Schmid W, Joss J, 1995.Nowcasting of motion and growth of precipitation with radar over a complex orography[J].Journal of Applied Meteorology, 34: 1286-1300.
[12]Li P W, Lai E S T, 2004.Application of readr-based nowcasting techniques for mesoscale weather forecasting in Hong Kong[J].Meteorlogical Applications, 11: 253-264
[13]Maas A L, Hannun A Y, Ng A Y, 2013.Recti?er nonlinearities improve neural network acoustic models[C]//The 30<sup>th</sup> International Conference on Machine Learning, Atlanta, GA, USA, July2013.
[14]Muller C K, Saxen T, Roberts R, et al, 2003.NCAR auto-nowcast system[J].Weather and Forecasting, 18: 545-561.
[15]Rinehnet R E, Gvarey E T, 1978.Three-dimensional storm motion detection by conventional weather radar[J].Nature, 273: 287-289.
[16]Rosenfeld D, 1987.Object method for analysis and tracking of convective cell as seen by radar[J].Journal of Atmospheric and Oceanic Technology, 4: 422-434.
[17]Shi X J, Chen Z R, Wang H, et al, 2015.Convolutional LSTM Network: A machine learning approach for precipitation nowcasting[C]//Proceedings of the 28th Internationla conference on neural information processing systems, Montreal, Canda: MIT press: 802-810.
[18]Shi X J, Gao Z H, Lausen L, et al, 2017.Deep learning for precipitation nowcasting: A benchmark and a new model.Advances in neural information processing systems [C]//Proceedings of the 31th Internationla conference on neural information processing systems, Long Beach, CA, US: NIPS, 1-11.
[19]Wilson J W, Crook N A, Mueller C K, et al, 1998.Nowcasting thunderstorms: A status report[J].Bulletin of the American Meteorological Society, 79(10): 2079-2099.
[20]Wang Y B, Long M S, Wang J M, et al, 2017.PredRnn: Recurrent neural networks for predictive learning using spatiotemporal LSTMS[C]//Advances in Neural Information Processing Systems, 879-888.
[21]Zhao H, Gallo O, Frosio I, et al, 2017.Loss functions for image restoration with neural networks[J].IEEE Transactions on Computational Imaging, 3(1): 47-57.
[22]曹春燕, 陈元昭, 刘东华, 等, 2015.光流法及其在临近预报中的应用[J].气象学报, 73(3): 471-480.
[23]陈葆德, 王晓峰, 李泓, 等, 2013.快速更新同化预报的关键技术综述[J].气象科技进展, 3(2): 31-37.
[24]陈明轩, 王迎春, 俞小鼎, 2007.交叉相关外推算法的改进及其在对流临近预报中的应用[J].应用气象学报, 18(5): 690-701.
[25]陈元昭, 林良勋, 王蕊, 等, 2019.基于生成对抗网络GAN的人工智能临近预报方法研究[J].大气科学学报, 42(2): 311-320.
[26]韩雷, 王洪庆, 林隐静, 2008.光流法在强对流天气临近预报中的应用[J].北京大学学报(自然科学版), 44(5): 751-755.
[27]韩丰, 魏鸣, 李南, 等, 2013.反射率因子和径向风速共同约束反演多普勒雷达风场化[J].遥感学报, 17(3): 584-589
[28]韩丰, 龙明盛, 李月安, 等, 2019.循环神经网络在雷达临近预报中的应用[J].应用气象学报, 30(1): 61-69.
[29]胡胜, 汪瑛, 陈荣, 等, 2009.“雨燕”中风暴算法在北京奥运天气预报示范项目中的应用及改进[J].高原气象, 28(6): 1434-1442.
[30]郭瀚阳, 陈明轩, 韩雷, 等, 2019.基于深度学习的强对流高分辨率临近预报[J].气象学报, 77(4): 715-727.
[31]李天成, 范红旗, 孙树栋, 2015.粒子滤波理论、 方法及其在多目标跟踪中的应用[J].自动化学报, 41(12): 1981-2002.
[32]马申佳, 陈超辉, 何宏让, 等, 2018.基于BGM的对流尺度集合预报试验及其检验[J].高原气象, 37(2): 495-504.DOI: 10. 7522/j.issn.1000-0534.2017.00073.
[33]孙志远, 鲁成详, 史忠植, 等, 2016.深度学习研究与进展[J].计算机科学, 43(2): 1-8.
[34]王丹, 王改利, 刘黎平, 等, 2014.基于雷达回波外推和中尺度模式预报的短时降水对比分析[J].高原气象, 33(3): 811-822.DOI: 10.7522/j.issn.1000-0534.2013.00038.
[35]王改利, 赵翠光, 刘黎平, 等, 2013.雷达回波外推预报的误差分析[J].高原气象, 32(3): 874-883.DOI: 10.7522/j.issn.1000-0534.2012.00081.
[36]吴剑坤, 陈明轩, 2018.基于雷达回波区域跟踪算法的临近预报技术进展[J].气象科技, 46(5): 899-909.
[37]吴剑坤, 陈明轩, 秦睿, 等, 2019.变分回波跟踪算法及其在对流临近预报中的应用试验[J].气象学报, 77(6): 999-1014.
[38]肖艳娇, 汤达章, 李中华, 等, 1998.风暴的自动识别, 跟踪与预报[J].南京气象学院学报, 21(2): 223-229.
[39]肖艳娇, 刘黎平, 2006.新一代天气雷达网资料的三维格点化及拼图方法研究[J].气象学报, 64(5): 647-657.
[40]徐月飞, 顾松山, 黄兴友, 等, 2011.计算几何法在风暴识别中的应用[J].热带气象学报, 27(2): 173-178.
[41]许小峰, 2018.从物理模型到智能分析——降低天气预报不确定性的新探索[J].气象, 44(3): 341-350.
[42]徐亚钦, 翟国庆, 黄旋旋, 等, 2011.基于TREC法以多重动态区域反演风场[J].浙江大学学报(工学版), 45(10): 1738-1745.
[43]俞小鼎, 姚秀萍, 熊廷南, 等, 2006.多普勒天气雷达原理与业务应用[M].北京: 气象出版社, 63-89.
[44]俞小鼎, 周小刚, 王秀明, 2012.雷暴与强对流临近天气预报技术进展[J].气象学报, 70(3): 311-337.
[45]郑永光, 张小玲, 周庆亮, 等, 2010.强对流天气短时临近预报业务技术进展与挑战[J].气象, 36(7): 33-42.
[46]郑永光, 周康辉, 盛杰, 等, 2015.强对流天气监测预报预警技术进展[J].应用气象学报, 26(6): 641-657.
[47]张兰, 徐道生, 胡东明, 等, 2019.雷达反演资料的Nudging同化对华南暴雨过程短临预报的影响[J].高原气象, 38(6): 1208-1220.DOI: 10.7522/j.issn.1000-0534.2018.00147.
[48]张亚萍, 程明虎, 夏文梅, 等, 2006.天气雷达回波运动场估测及在降水临近预报中的应用[J].气象学报, 64(5): 632-634.
Outlines

/