THE ASSESSMENT OF APPLICATION EFFECTIVENESS OF THREE MACHINE LEARNING METHODS IN AUTOMATIC IDENTIFICATION OF THUNDERSTORM GALE IN GUANGDONG
-
摘要: 基于2012—2019年自动站雷暴大风观测实况和对应雷达回波，利用传统机器学习方法(决策树)和深度学习方法(CNN、YOLO)等三种机器学习方法分别建立雷暴大风自动识别模型。根据广东雷暴大风回波特征，选取50 dBZ高度、反射率因子强度梯度等5个回波参量作为决策树的特征因子；将1~9 km高度的雷达回波分为11层，作为YOLOv3的输入层，使其由原3个特征层扩展到11层，训练优化后的YOLOv3可更合理刻画雷暴大风的空间结构特征。经批量测试和业务试运行试验，检验结果表明：三种模型中基于决策树的模型虚警最高，基于CNN的模型漏报最多，基于YOLO的模型识别效果最好，其POD和CSI均最高。通过对广东2020年汛期5次系统性和5次局地性雷暴大风过程进行分类型自动识别效果评估，并选取任意天气下长达30天连续时段进行不间断识别检验，结果表明该算法对于不同类型的雷暴大风均有较好的识别能力，具备业务化应用前景。Abstract: On the basis of radar data and thunderstorm gale observation in Guangdong from 2012 to 2019, three machine learning algorithms, including a traditional machine learning algorithm (Decision Tree) and two deep learning algorithms (CNN and YOLO), were applied to establish automatic identification models for thunderstorm gale respectively. In line with the radar echo characteristics of thunderstorm gale in Guangdong, six echo parameters, such as 50 dBZ height and reflectivity factor gradient, were selected as the characteristic factors of the decision tree. In addition, the heights of 1 to 9 km were divided into 11 radar echo input layers of YOLOv3, expanding from the original three characteristic layers, which described the spatial structure characteristics of thunderstorm gale more reasonably after optimization. A series of business run-in tests indicated that among the three identification models, the model using the decision tree presents the highest FAR and the one based on CNN misses the most identifications, while the identification model based on YOLO behaves the best, with the highest POD and CSI. By evaluating the automatic identification effects of five systematic and five local thunderstorm gale processes during the rainy season of 2020 in Guangdong, and selecting continuous periods of up to 30 days under any weather condition for ongoing identification tests, the results reveal that the YOLO algorithm possesses good recognition ability for different types of thunderstorm gale with considerable business application prospect.
-
Key words:
- thunderstorm gale /
- automatic identification /
- machine learning /
- radar echo /
- deep learning
-
表 1 三种模型识别效果批量测试对比
模型 POD FAR CSI 决策树 0.934 0.57 0.363 CNN 0.865 0.489 0.421 YOLO 0.994 0.308 0.685 说明：表中YOLO识别结果的置信度=0.7，后文中YOLO识别结果均采用该置信度。 
下载: 导出CSV
表 2 2016年4月13日飑线型雷暴大风天气过程识别评估结果
模型 实况记录 命中数 空报数 漏报数 POD FAR CSI 决策树 200 214 14 0.935 0.517 0.467 CNN 214 189 135 25 0.883 0.417 0.542 YOLO 210 101 4 0.981 0.325 0.667
下载: 导出CSV
表 3 2019年3月2日雷暴大风、强降水混合天气过程识别评估结果
模型 实况记录 命中数 空报数 漏报数 POD FAR CSI 决策树 26 39 2 0.929 0.600 0.388 CNN 28 25 30 3 0.893 0.545 0.431 YOLO 28 28 0 1.000 0.500 0.500
下载: 导出CSV
表 4 基于YOLO的识别模型对于局地性天气过程识别能力的评估
类型 实况记录 POD FAR CSI 局地性过程 71 0.958 0.364 0.618 系统性过程 148 0.986 0.321 0.673
下载: 导出CSV
-
[1] 陈晓欣, 俞小鼎, 王秀明. 中国大范围雷暴大风事件(Derechos)研究: 时空分布, 环境背景和对流系统形态特征[J]. 气象学报, 2022, 80(1): 67-81. [2] 俞小鼎. 强对流天气的多普勒天气雷达探测和预警[J]. 气象科技进展, 2011, 1(3): 31-41. [3] 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. [4] 伍志方, 叶爱芬, 胡胜, 等. 中小尺度天气系统的多普勒统计特征[J]. 热带气象学报, 2004, 20(4): 391-400. [5] 俞小鼎, 张爱民, 郑媛媛, 等. 一次系列下击暴流事件的多普勒天气雷达分析[J]. 应用气象学报, 2006, 17(4): 385-393. [6] 牛淑贞, 鲍向东, 乔春贵, 等. 强对流风暴新一代雷达产品特征分析[J]. 气象, 2008, 34(7): 92-100. [7] 李国翠, 刘黎平, 连志鸾, 等. 利用雷达回波三维拼图资料识别雷暴大风统计研究[J]. 气象学报, 2014, 72(1): 168-181. [8] 周康辉, 郑永光, 王婷波, 等. 基于模糊逻辑的雷暴大风和非雷暴大风区分方法[J]. 气象, 2017, 43(7): 781-791. [9] GAGNE D J. Coupling data science techniques and numerical weather prediction models for high impact weather prediction[D]. Oklahoma: University of Oklahoma, 2016: 185. [10] MCGOVERN A, ELMORE K L, GAGNE D J, et al. Using artificial intelligence to improve real-time decision-masking for high-impact weather[J]. Bull Amer Meteor Soc, 2017, 98(10): 2 073-2 090. [11] 杨璐, 韩丰, 陈明轩, 等. 基于支持向量机的雷暴大风识别方法[J]. 应用气象学报, 2018, 29(6): 680-689. [12] SHI X J, CHEN Z R, WANG H, et al. Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting[C]//Proceedings of the 28th International Conference on Neural Information Processing Systems, Cambridge: MIT, 2015. [13] WANG Y, LONG M, WANG J, et al. Pred RNN: Recurrent neural networks for predictive learning using spatiotemporal LSTMs[C]//Proceedings of the 30th Neural Information Processing Systems. Long Beach, CA, 2017. [14] WANG Y, ZHANG J, ZHU H, et al. Memory in memory: a predictive neural network for learning higher-order non-stationarity from spatiotemporal dynamics[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition, USA, 2019. [15] WANG Y, WU H, ZHANG J, et al. Pred RNN: A recurrent neural network for spatiotemporal predictive learning[C]// Advances in Neural Information Processing Systems 30th, 2017. [16] 陈元昭, 林良勋, 王蕊, 等. 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, 2019, 42(2): 311-320. [17] 顾建峰, 周国兵, 刘伯骏, 等. 人工智能技术在重庆临近预报业务中的初步研究与应用[J]. 气象, 2020, 46(10): 1 286-1 296. [18] 俞小鼎. 多普勒天气雷达原理与业务应用[M]. 北京: 气象出版社, 2006: 11-12. [19] SAFAVIAN S R, LANDGREBE D. A survey of decision tree classifier methodology[J]. IEEE Transactions on Systems, Man, and Cybernetics, 1991, 21(3): 660-674. [20] 王艳兰, 唐桥义. 一次大范围强对流天气卫星云图及雷达回波特征分析[C]//中国气象学会年会, 2012. [21] 龙柯吉, 康岚, 罗辉, 等. 四川盆地雷暴大风雷达回波特征统计分析[J]. 气象, 2020, 46(2): 212-222. [22] 杨晓亮, 柴东红, 杨敏. 河北南部雷暴大风天气分析[C]//中国气象学会年会, 2008. [23] 东高红, 吴涛. VIL在地面灾害性大风中的应用与临近预报[C]//2006年全国重大天气过程总结和预报技术经验交流会, 中国气象学会, 2007. [24] 廖玉芳, 潘志祥, 郭庆. 基于单多普勒天气雷达产品得强对流天气预报预警方法[J]. 气象科学, 2006, 26(5): 565-571. [25] 伍志方, 庞古乾, 贺汉青, 等. 2012年4月广东左移和飑线内超级单体的环境条件和结构对比分析[J]. 气象, 2014, 40(6): 655-667. [26] 吴芳芳, 王慧, 韦莹莹, 等. 一次强雷暴阵风锋和下击暴流的多普勒雷达特征[J]. 气象, 2009, 35(1): 55-64. [27] LAWRENCE S, GILES C L, TSOI A C, et al. Face Recognition: A Convolutional Neural Network Approach[J]. IEEE Transactions on Neural Networks, 1997, 8(1): 98-113. [28] HE S F, RYNSON W H, LIU W X, et al. Super CNN: A superpixelwise convolutional neural network for salient object detection[J]. International Journal of Computer Vision, 2015, 115(3): 330-344. [29] JOSEPH R, SANTOSH K D, ROSS B. Girshick, et al. You only look once: Unified, real-time object detection[C]//2016 IEEE / CVF Conference on Computer Vision and Pattern Recognition, 2016. [30] JING L L, YANG X D, TIAN Y L. Video you only look once: Overall temporal convolutions for action recognition[J]. Journal of Visual Communication and Image Representation, 2018, 52: 58-65 [31] TIAN S, LIU J, JIN Y D, et al. Pedestrian Multi-target Tracking Based on YOLOv3[C]//2020 39th Chinese Control Conference (CCC), 2020. -
点击查看大图
图(6) / 表(6)
计量
- 文章访问数: 214
- HTML全文浏览量: 40
- PDF下载量: 68
- 被引次数: 0
粤公网安备 4401069904700003号