Study On The Echo Feature Extraction And Forecast Of Convective Storm

Doppler radar is a powerful tool to monitor meso-scale and micro-scale weather system. The echo data can be used for severe storm detection and warning, mitigating the loss by weather disaster.This paper describes firstly the radar echo feathers of convective storm and two widely accepted storm identification algorithms, SCIT(Storm Cell Identification and Tracking Algorithm) and TITAN(Thunder storm Identification, Tracking, Analysis and Nowcasting). TITAN can only recognize single storm but not a cluster. SCIT can find the strong center of storm cluster, but ignores low-threshold recognition results. For the shortcomings of two algorithms, a new algorithm is proposed.This article draws the advantages of the two identification methods, uses composite reflectivity for analysis, references the multi-threshold identification method of SCIT, imposes open operation on the recognition results of every level threshold. In the open process, we need the dynamic template to control erosion, and let the storms dilate until them contacts with each other or reaches on the profile of low threshold. This allows the storm recognized result restore information within the body as much as possible, such as storm size, liquid water content, volume, etc. The identified storm components are considered to have been treated about false combined. The centroid and quality of storms can be calculated through mapping the components to each level of PPI in the three-dimensional space. Using the method of TREC to get the beginning guess of the storm position, tracking the path of history storm with storm cell center identification to make forecasts of the future storm location and outline through weighted extrapolation.The process observed by Hefei radar at July 22 and 23,2008 is used to test and analyze. Compared with TITAN and SCIT, it is proved that the identification results proposed in this paper is much better. It could not only separate the storms from cluster, but also restore some of the information discarded by high threshold identification.As the main features of severe weather, to extract the parameters of bow echo, the paper do closed operations on echo image after binarization. The near cells can be connected through dilation operation which is part of closed operation, in order to inhibit the breaking of erosion operation, the erosion pixel of image can be limited before operation. Because of many branches and burrs existing in thining operated image after closed operation, a new pruning algorithm is presented to solve this problem. Single connectivity of a thined curve could be achieved through scaning branch point one by one, tracking the development of branch, and cutting off the branches. After pruning, the line image can be used to analyse storm parameters intuitively.There are some innovations in this paper:(1)Based on the foundation of muti-threshold recognition, joining open operation to the identification results of each level of the threshold, which can not only identify the storm in the cluster through the senior threshold but also keep the storm information by low threshold recognition.(2) The storm thinning method is presented. And a innovative pruning algorithm based on branch point about line of image is first proposed. The method is confirmed good enough for experiment, it is able to correctly reserve the trunk and cut off all of the burrs.(3) After fitting the segment of storm skeleton,and using it to calculate the curvature, the result of fitting arc is consistent with the storm trend.