摘要
【目的】中尺度对流降水预报是天气预报的重点和难点之一,天气雷达探测的高时空分辨率降水数据是开展0~2 h临近预报的重要依据,对高分辨率中小流域雨洪预报具有重要意义。利用雷达数据开展对流降水临近预报为人们出行、农业生产指导、防灾减灾提供便利,在气象和水文领域具有实际应用价值。【方法】以广州新一代S波段天气雷达体扫资料为基础,采用基于Causal-LSTM记忆模块的条件可逆网络CrevNet开展对流降水临近预报的能力研究,并将模型的预报效果与基于普通ST-LSTM的模型进行对比分析,验证其优越性。为提升预报模型对强回波的记忆能力使用了带权重Huber损失。采用临界成功指数(CSI)、命中率(POD)、虚警率(FAR)评估了测试集在不同预报时效和检验阈值下的预测效果。此外,还采用峰值信噪比(PSNR)、图像结构相似性(SSIM)以及偏差评分(BIAS)检验新生对流回波的预测能力。【结果】结果显示:基于Causal-LSTM记忆单元的CrevNet模型在预报时段内CSI、POD较高,FAR较低。在两次对流个例的预测中,该模型在多个预测时效下具有较高的PSNR、SSIM以及更接近1的BIAS。【结论】研究表明:条件可逆网络CrevNet深度学习模型对于时空序列有较强的时空特征提取能力,搭配不同的卷积循环神经单元,预报效果会有不同;结合Causal-LSTM记忆单元的CrevNet模型能更好地保留对流回波形态,适用于对流降水临近预报。
[Objective]Mesoscale convective precipitation prediction is one of the key and difficult objects of weather forecasting.Precipitation data detected by weather radar has high spatiotemporal resolution,which is the main data source for nowcasting in 0~2 hours.It is of great significance for high-resolution rainfall forecasting in small and medium-sized watersheds.The use of radar data to carry out nowcasting of convective precipitation can facilitate people′s travel,agricultural production guidance,disaster prevention and mitigation,and has practical application value in the field of meteorology and hydrology.[Methods]Based on the volumetric scan data of the new generation S-band Doppler radar in Guangzhou,this study will explore the prediction performance of the conditional reversible network CrevNet based on the Causal-LSTM memory module in convective precipitation nowcasting,and then compare the prediction effect with the model based on ordinary ST-LSTM to verify its superiority.To improve the memory ability of strong echoes,the model was trained with weighted Huber loss function.In this study,the CSI(Critical Success Index),POD(Probability of Detection,or Hit Ratio)and FAR(False Alarm Rate)were used to evaluate the result of the test dataset under different prediction time and test thresholds,and PSNR(Peak signal-to-noise ratio),SSIM(image structure similarity)and BIAS(bias scores)were used to test the predictive ability of the convective event.[Results]The result show that the CrevNet model based on Causal-LSTM memory unit has higher CSI and POD,and lower FAR during the forecast period.In the prediction of two convection cases,the model has a higher PSNR,SSIM and a BIAS closer to 1 under multiple prediction timeliness.[Conclusion]The CrevNet deep learning model of conditional reversible network has a strong ability to extract spatiotemporal features for spatiotemporal sequences,and the prediction effect will be different with different convolutional recurrent neural units.Therefore,the CrevNet model based on the Causal-LSTM memory unit can better preserve the convective echo morphology and is more suitable for convective precipitation nowcasting.
作者
张永轩
黄兴友
王雪婧
于华英
楚志刚
ZHANG Yongxuan;HUANG Xingyou;WANG Xuejing;YU Huaying;CHU Zhigang(School of Atmospheric Physics,Nanjing University of Information Science&Technology,Nanjing 210044,Jiangsu,China)
出处
《水利水电技术(中英文)》
北大核心
2024年第11期1-14,共14页
Water Resources and Hydropower Engineering
基金
国家重点研发计划项目“重大自然灾害监测预警”课题(2018YFC1506102)
山东省自然科学基金双偏振雷达资料在强对流天气中的应用研究(ZR2020MD053)。
关键词
对流临近预报
深度学习
CrevNet
Huber损失
雷达反射率
降水
气候变化
防灾减灾
convection nowcasting
deep learning
CrevNet
Huber loss function
radar reflectivity
precipitation
climate change
disaster prevention and reduction
作者简介
张永轩(2000-),女,硕士研究生,主要从事雷达气象学的研究。E-mail:15651020295@163.com;通信作者:黄兴友(1965-),男,教授,博士,主要从事大气探测和雷达气象学的研究。E-mail:hxyradar@126.com。
