SHORT-TERM PRECIPITATION NOWCASTING BASED ON MULTI-SCALE FEATURE DEEP LEARNING
-
摘要: 雷达回波外推是短临降水预报的一种重要方法。针对雷达回波外推中随着外推时间的增加而出现回波演变信息丢失这一问题，本文提出一种多尺度特征融合的深度学习短临降水预报模型(以下简称为MSF2)。首先，采用多尺度的卷积核对网络的浅层信息进行特征提取，弥补单一特征检测带来的不足。其次，将不同维度的特征信息进行拼接及通道混洗，进一步增强特征图通道之间的信息流通和信息表达能力。最后，将特征图中的多尺度信息进行融合，从而有效保留不同尺度的特征信息。利用华南雷达回波拼图资料数据，在3种不同降水强度（5 mm/h、10 mm/h和25 mm/h）下进行降水预报研究，并与光流法和ConvLSTM两种主流算法进行了对比。结果显示，在3种不同降水强度条件下，MSF2在所有评价指标（命中率POD、临界成功指数CSI、误报率FAR）中表现最优，这表明引入多尺度机制能改善模型的特征提取能力。相比于目前主流的光流法和ConvLSTM，本文提出的模型对于短临降水预报具有较好的适用性和较高的预报精度, 而且实现了业务化运行。Abstract: Radar echo extrapolation is an important method precipitation nowcasting. To address the problem of the loss of characteristic evolution information in echo extrapolation prediction with the increase of echo intensity and prediction time, this paper proposes a deep learning model for precipitation nowcasting based on multi-scale feature fusion (MSF2). Firstly, the multi-scale convolution kernel is used to extract the features of the shallow information of the network to offset the shortcomings caused by the single feature detection. Secondly, the feature information of different dimensions is spliced and the channels are shuffled to further enhance the information circulation and information expression capabilities between the feature map channels. Finally, the multi-scale information in the feature map is fused in order to effectively keep the channel information after the fusion of the feature map. With the South China radar echo data, the fusion experiment was carried out under three different precipitation intensities, and compared with two mainstream algorithms, i. e., ConvLSTM and optical flow. The experimental results show that MSF2 performs best in terms of all evaluation indexes under the conditions of precipitation rates 5 mm/h, 10 mm/h and 25 mm/h. It can be concluded that the introduction of a multi-scale mechanism can improve the feature extraction ability of the nowcasting model. Compared with the current radar echo extrapolation algorithm ConvLSTM and optical flow, the proposed model MSF2 has better potentials in operational applications and higher forecast accuracy for precipitation nowcasting.
-
Key words:
- nowcasting /
- deep learning /
- multi-scale features /
- optical flow /
- convLSTM
-
表 1 降水量大于等于5 mm/h的预报结果
预报方法 CSI FAR POD Flow 0.403 3 0.399 7 0.456 1 ConvLSTM 0.440 3 0.365 7 0.493 8 MSF2 0.479 8 0.323 2 0.5279 
下载: 导出CSV
表 2 降水量大于等于10 mm/h的预报结果
预报方法 CSI FAR POD Flow 0.313 9 0.431 6 0.341 0 ConvLSTM 0.358 4 0.390 0 0.383 4 MSF2 0.402 7 0.326 7 0.444 7
下载: 导出CSV
表 3 降水量大于等于25 mm/h的预报结果
预报方法 CSI FAR POD Flow 0.214 2 0.560 4 0.234 5 ConvLSTM 0.261 6 0.491 9 0.279 7 MSF2 0.310 4 0.412 3 0.338 2
下载: 导出CSV
-
[1] 郭瀚阳, 陈明轩, 韩雷, 等. 基于深度学习的强对流高分辨率临近预报试验[J]. 气象学报, 2019, 77(4): 715-727. [2] JOHNSON J T, MACKEEN P L, WITT A, et al. The storm cell identification and tracking algorithm: An enhanced WSR-88D algorithm[J]. Weather and forecasting, 1998, 13(2): 263-276. [3] RINEHART R E, GARVEY E T. Three-dimensional storm motion detection by conventional weather radar[J]. Nature, 1978, 273(5660): 287-289. [4] 陈训来, 刘军, 郑群峰, 等. 基于卷积门控循环单元神经网络的临近预报方法研究[J]. 高原气象, 2021, 40(2): 411-423. [5] 韩丰, 龙明盛, 李月安, 等. 循环神经网络在雷达临近预报中的应用[J]. 应用气象学报, 2019, 30(1): 61-69. [6] 施恩, 李骞, 顾大权, 等. 基于卷积神经网络的雷达回波外推方法[J]. 计算机应用, 2018, 38(3): 661-665、676. [7] SHI X, CHEN Z, WANG H, et al. Convolutional LSTM network: A machine learning approach for precipitation nowcasting[J]. arXiv preprint arXiv: 1506.04214, 2015. [8] 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): 2 631-2 644. [9] AGRAWAL S, BARRINGTON L, BROMBERG C, et al. Machine learning for precipitation nowcasting from radar images[J]. arXiv preprint arXiv: 1912.12132, 2019. [10] CHOLLET F. Xception: Deep learning with depthwise separable convolutions[C]. Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 1 251-1 258. [11] GAO H, WANG Z, JI S. Channelnets: Compact and efficient convolutional neural networks via channel-wise convolutions[J]. arXiv preprint arXiv: 1809.01330, 2018. [12] ZHANG X, ZHOU X, LIN M, et al. Shufflenet: An extremely efficient convolutional neural network for mobile devices[C]. Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 6 848-6 856. [13] HE K M, ZHANG X G, REN S Q, et al. Deep Residual Learning for Image Recognition[C]. //2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016. [14] 张辉, 朱家明, 唐文杰. 基于聚类和改进型水平集的图像分割算法[J]. 计算机科学, 2017, 44(Z1): 198-201. [15] 高晗, 田育龙, 许封元, 等. 深度学习模型压缩与加速综述[J]. 软件学报, 2021, 32(01): 68-92. [16] BAI L Q, CHEN G X, HUANG L. Image processing of radar mosaics for the climatology of convection initiation in South China[J]. J App Meteor Climat, 2020, 59(1): 65-81. [17] 唐菁, 陈生, 李峙, 等. 基于雷达观测的华南夏季降水类型日变化的时空分布特征[J]. 气象, 2023, 49(3): 307-318. [18] FULTON R A, BREIDENBACH J P, SEO D J, et al. The WSR-88D rainfall algorithm[J]. Weather and forecasting, 1998, 13(2): 377-395. [19] 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 Discussions, 2018: 1-23. -
点击查看大图
图(5) / 表(3)
计量
- 文章访问数: 202
- HTML全文浏览量: 19
- PDF下载量: 66
- 被引次数: 0
粤公网安备 4401069904700003号