雷达回波预测极端暴雨概率方法构建原理与应用研究

Construction and Application of Extreme Rainstorm Probability Prediction Based on Radar Echo Base Data

基于局地雷达回波基数据强度与雷达体扫面内相应地区强降水量之间的统计关系,并通过LucasKanade局部光流法分析下一小时可能影响该地区的回波区域,将这些区域定义为强降水影响系统的动态回波上游。同时分析雷达各层体扫回波情况,即考虑强降水系统的空间伸展程度。通过分析这些动态回波区域在当前时刻内(每10 min多层体扫信息)的时空变化特征,建立回波强度特征与局地降水的相关,进一步与改善的降水极值概率预测方法相结合,构建对下一小时局地暴雨重现期极值预测预警指标,为应急保障方案和及时应对决策提供技术支持与专业参考信息。整体方案以2006-2011年6-8月盐城多普勒天气雷达回波强度的基数据资料和同时段盐城雷达体扫范围内建湖气象站的小时雨量序列为基础,采用Lucas-Kanade局部光流法确定回波强度反映的降水系统时空动态上游,并利用皮尔逊III型方法和广义帕累托方法建立回波类别与建湖局地白天与夜间降水序列的统计概率关系,计算回波类别对应的局地极端降雨极值以及极值重现概率特征,构建降水系统的雷达回波动态综合特征与下一小时局地强降水极值概率间的统计关系指标,指标的相关性检验达到70%。该方法具有强降水多等级重现期极值预测能力,为下一小时临近极端暴雨预测预警提供了雷达监测动力统计优化方法。

Based on a statistical relation betw een local Radar echo intensity base data and local heavy rainfall,to find the upstream echo areas that will move to the local raining region in next hour by the regional Lucas-Kanade optical flow method. Meanw hile to analyze the echo features at every scan surface,so that to learn the raining system stretching altitude. By analysis of the temporal and spatial characteristics of dynamic echo areas in current hour with information at scan surfaces every 10 minutes,to build the relation betw een Radar echo intensity features and local rainfall,then applied it on a predicting method of improved rainfall extreme probability,thus to construct forecast and warning index of extreme rainfall return period in next hour. The results will provide the technique support and professional reference for emergency schemes and strategy and tactics. The whole scheme is based on the Doppler radar echo intensity base data at scanning surfaces and the hourly precipitation data of Jianhu weather station during June,July,August in 2006-2011. Jianhu station is within the scanning area of the radar located at Yancheng. The echo dynamic upstream of the local heavy rainfall system is determined by the regional Lucas-Kanade optical flow method. The statistical probability correlation is calculated betw een echo intensity classification and precipitation series in day and night by Peirson III and Generalized Pareto Distribution methods. The probability characteristics of return period of graded extreme rainfall and the local extreme rainfall under the echo types are also calculated,then the statistic index is constructed betw een the radar echo dynamic features including the trend of evolution and the probability of local extreme rainfall in next hour,the index correlation and its test all get to 70%. The index can predict the probabilities of different grade extreme rainfall in return period at local place. It is an optimization method of radar dynamic statistics for now casting and warning of the extreme rainstorm in next hour.