Research On X-band Radar Networking Strategy And Strong Convective Weather Identification

With the advent of weather radar,disaster prevention and mitigation for severe disaster weather became possible.Today,weather radars play an irreplaceable role in meteorological observations.The main weather radars are S/C band radars.Due to the curvature of the earth,topographical obstacles,and the long scanning time of large radars,problems such as low-altitude blind zones and leakage of rapidly growing small and medium-sized convection fluids have arisen.In order to overcome these shortcomings,in 2006,the US Atmospheric Cooperative Adaptive Remote Sensing Engineering Research Center(CASA)first proposed the use of multiple small X-band radars to form a networked radar to achieve high spatial and temporal resolution observations of target weather objects.In 2013,relying on the special "Multi-radar network adaptive observation technology research and data quality control" project of the meteorological industry,China's Nanjing Enruite Industrial Co.,Ltd.completed the use of 3 X-band weather radars to complete the radar network installation in Nanjing For meteorological observations and experiments.This article relies on the key special project “Monitoring and Precautions of Vertical Comprehensive Meteorological Observation Technology for Megacities” in the National Key R & D Plan of 2018,“Monitoring and Precaution of Major Natural Disasters”,and uses three existing X-band networked radar of Chengdu University of Information Technology in Chengdu,China.The radar test platform aims to obtain a refined vertical structure of the atmospheric low-altitude boundary layer,and is designed to meet the radar network scanning strategy with local characteristics.At the same time,an algorithm for automatic identification and classification of strong convective targets was developed based on disaster prevention and mitigation for severe disaster weather and the scanning strategy in conjunction with networking radar.The main specific contents of this paper are as follows:(1)The research focuses on the specific scanning strategies of three networked radars.In order to meet the project's requirements for meteorological observation of vertical low-altitude high-temporal resolution,two working modes of network radar are designed.Clear sky mode and cooperative mode.Under the clear sky mode,the three radars each perform a 14-layer volume scan mode.If a strong convective weather event occurs,the cooperative mode of the network radar will be triggered.There are two scanning modes in the cooperative mode,namely,the fast body scan mode and the multiple RHI mode.In the fast body scan mode,each radar will perform five-layer 360°full-angle scanning at elevation angles of 0.5°,1.45°,2.4°,3.35°,and 4.3°;In the multi-RHI mode,operations are performed according to the type of strong convection.If the strong convection is a single-cell strong convection,each of the three radars performs a RHI scan on the center of the single monomer.If the strong convection is multi-cell strong convection,each of the three radars performs a RHI scan on the head,middle and tail of the multi-cell.(2)The collaborative control algorithm of network radar is mainly studied.In the emergence of severe weather,many strong convective cells often appear,and the specific radar allocation is particularly important.Targets scanned by radar should be sorted according to their threat level.The specific operation is divided into three steps.The first step is the selection of key areas(strong convection recognition).The second step is priority calculation.The priority is mainly determined by 5 factors,which are the highest reflectance of the area,the average reflectance of the area,the area of the area,and the importance of the area.The third step is to assign radars based on priorities.(3)The research focuses on the identification algorithms of strong convection.Due to the shortcomings of the existing strong convection recognition algorithms,such as the Nanjing test platform and the US CASA simply use thresholds for identification.This paper proposes an algorithm for identifying strong convection using a neural network(BPNN).First,through the feature selection,six horizontal and vertical radar echo characteristics of the weather target are selected,which are: continuous height greater than 40 d BZ,average background reflectivity,cloud height,horizontal gradient of reflectivity,and greater than35 d BZ The body area and vertical integration of liquid water finally divide the precipitation echo into stratiform cloud precipitation and strong convective precipitation.By comparing with several classic algorithms of strong convection recognition Fuzzy logic,SHY95,BL,the F-Score scores of this algorithm are3%,24%,and 30% higher than them.The operation speed of this algorithm is54 times their,2 times,4 times.(4)The research focuses on the specific types of classification algorithms for strong convection regions.The strong convection recognition in(3)is mainly used for the delineation of key areas and the triggering of modes,and there are different scanning methods for specific types of strong convection.This classification algorithm uses the morphological features of strong convection to classify the recognition results based on(3).The main core idea is to use convolutional neural network(CNN)combined with its morphology to separate strong convection into two types: single-cell and multi-cell.Using the neural network trained in(3)to identify the strong convection area in the radar precipitation echo,and then use the density-based clustering algorithm DBSCAN to divide the strong convection area into independent individuals.Finally,independent strong convection individuals are input into the CNN,and they are classified into single and multiple cells.