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

Abstract

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.