Research On Lightning Data Simulation And Filtering Methods Of Geostationary Lightning Imager

China will launch the FY-4geostationary satellite into space to perform tasks. As asubsystem of the satellite, the lightning imager on FY-4will monitor and track convectiveweather by real-time and continuous observation of lightning. And it will play an importantrole in the field of lightning protection, early warning of forest fire, aviation weathersecurity and military meteorological. In order to meet the need to generate lightningproduct, my school commitments the welfare research project The Product GenerationAlgorithm and Software of Lightning Imager. The study content of the paper which is basedon the project is on lightning data simulation and filtering method of geostationarylightning imager.The paper adapts a Monte-Carlo model to simulate a random walk in cloud for a largenumber of photons to analyze the imaging characteristics of lightning. Then it presents thedistribution pattern of photons in the surface of the cloud, which shows the aggregation oflightning signal. The data called original data of the lightning imager on stationary orbitconsists of lightning signal, background signal, and the noise generated by device andparticles hitting. In order to filter the noise signal and extract real lightning signal, thisarticle puts forward the processing flow of lightning data and filter method which consistsof L1Event filtering methods and L2Flash filtering methods.The lightning original data is the key of designing and verifying filter algorithms. Dueto the lack of original data, we simulate some different noisy signals of geostationarylightning imager. The paper simulates some common equipment noise such as redundancynoise and ghost noise firstly. And then the L1Event filtering will delete the noiseeffectively. Particle noise is a random event following a normal distribution in time anduniform distribution in space and it is the main noise in the original data. The papersimulates particle noise based in the time-space characteristic of the noise using theBox-Muller method. Then we get the original data for L2Flash filtering by adding thenoise to the lightning signal transformed from the product of LIS product of America.In the single-frame image, the particle noise is the same as lightning signal. However,the particle noise is random in space and time, and the lightning signal is continuous inspace and time. The paper implements the L2Flash filtering method based in the differentcharacteristics of them. The method can be described at three steps. At first, the clusteringalgorithm based in time continuity and space continuity of lightning clusters the Event datato Flash data. Then, the partial filtering algorithm based in the space aggregation oflightning pulses and the clustering properties of noise the false Flash delete the Flashes that are clustered from noise. At last, the putting back algorithm based in space aggregation oflightning is proposed and it decides whether the deleted Flash in the partial filtering true ornot.We have proved through experiments and data that the L2filter method can filtersimulated data safety, which meets the requirements of the data process system ofgeostationary lightning imager. In the experiments, we have189seconds simulated datawhich has71613Events that contain lightning signal and random particle noise. Afterclustering algorithm, we get8486Groups and7221Flashes with6841Flash1,294Flash2,26Flash3, and others are considered real lightning Flashes. The clustering algorithm has agood performance after data analysis. The noise level t makes a great influence on theperformance of particle filtering algorithm. If t is large, a noisy Flash can be considered areal lightning Flash, on the contrary, a real lightning Flash may be a noise if t is small. Wetake on different values to observe the performance of the particle filtering. Afterrepeated experiments, When is5%, the performance is better and the all and43are deleted, and then at last there are337Flashes after particle filtering. Namely,we have deleted95%Flashes which are almost noisy Flashes. In the simulated data,99%ofare noise which tells that the particle filtering works well. At last, we use puttingback algorithm to take380Flashes back, which instructions the algorithm takes back0.6%Flashes of the deleted Flashes that areFlash2.This article designs and verifies the filter methods of lightning data processing systemabout geostationary lightning imager. And then the article gradually realizes the filteringmethods and filters the original data safety. This provides the technical and theoreticalsupport. However, Because of project conditions and the lack of real original data oflightning imager, we adapt simulated data to replace real data in the experiments. And theparameters are all artificially set. So the filter methods are limited. In addition, the filtermethods have not been proven on geostationary orbit. These are the deficiencies of thisarticle. With the progress of the project and the completion of the hardware facilities, weexpect the improvement of the deficiencies.