Research And Application Of Clustering Algorithm For FY-4A Lightning Mapping Imager

As the newest geostationary lightning detection system,FY-4A Lightning Mapping Imager(LMI)provides a large amount of observation data for the characteristics analysis of lightning activities and the forecasting of thunderstorm activities in China.At present,the lightning detection data published by the National Meteorological Satellite Data Center is an Level 2 event and has no physical meaning.Therefore,clustering LMI detection data into lightning is the first step of satellite lightning data application.However,the clustering algorithm currently used by LMI is the clustering parameter of the polar-orbiting satellite lightning imager.It does not take into account the large spatial resolution difference of different CCD pixels of the geostationary satellite lightning mapping imager.Based on the spatial resolution of the pixels in different regions of the LMI observation range,the clustering algorithm suitable for the LMI observation region is determined by the parameterization scheme of the clustering algorithm.The data of the WWLLN(World Wide Lightning Location Network)is used to analyze the detection performance of LMI in different regions.On this basis,the characteristic distribution of lightning matched by LMI and WWLLN is analyzed,which provides a reference for the identification of IC pluse and CG lightning flash.The main research methods and conclusions are as follows:The Geo-location algorithm of the LMI is used to convert the grid coordinates of the CCD array pixels into the latitude and longitude coordinates of the Earth coordinate system,and the spatial resolution of the CCD array pixels of the LMI is calculated.The observation area of LMI is divided into A region with a pixel spatial resolution of 20 km.The region ranges from 50°~ 60°N,60°~ 160°E,and the B region with a spatial resolution of 8 km,The range is 15°~50°N,60°~160°E.For the lightning data of A and B regions,the clustering parameter sensitivity test of event to group and group to lightning is found,and the spatial resolution has no effect of the event to group clustering.The event to group clustering time parameter of the A and B regions is 2 ms,and the spatial parameters are adjacent pixel.In the group to lightning clustering algorithm,in different spatial resolution regions,The time parameter is 330 ms,the spatial parameter of zone A is 25 km,and the spatial parameter of zone B is 16.5 km.Using the synchronization data of WWLLN and TBB to assess the detection performance of LMI,it is found that the density distribution of detected lightning activity of LMI and WWLLN in China region shows a good correction.The number of lightning activity detected by LMI is three times more than the data WWLLN detected.LMI lightning activity is mainly distributed in the cold cloud area where the TBB temperature is lower than-60 °C,which has a good correspondence with the strong convective weather process.When the lightning data detected by WWLLN is matched with the LMI data,the matching time threshold of WWLLN and LMI satellite detection system is determined to be 2 s and the spatial threshold is 0.2° by the time threshold and spatial parameter sensitivity test.Due to the spatial resolution of the pixel,the detection performance in the A region with a spatial resolution of 20 km is poor,and the matching rate with WWLLN is only 3%.The lightning time difference between the two systems is 0.9 s,and the distance is 22 km.The B region with a spatial resolution of 8 km has a matching ratio of LMI of 13%,lightning time difference of 0.6 s,and distance of 11 km.Using the matched and unmatched data of LMI and WWLLN,the intensity and size of the detected group,the difference in intensity,duration and space of the lightning are analyzed.The difference between the radiant energy density value and the spatial size of lightning is found can be used as a reference for the identification of IC pluse and CG lightning flash.