Joint Application Of Infrared Imaging And Microwave Detection In Early Warning Of Local Convection

There is large region in China affected by with high temperature,rainy and frequent convections especially the local severe convection s in summer due to the influence of the East Asian monsoon and the distribution of land and ocean.Therefore,the study on the prediction of convective occurrence can help identify the local severe convective weather as early and accurately as possible,thus is very useful for disaster prevention and mitigation.The geostationary weather satellite-based infrared(IR)imager with high temporal resolution can observe the convection frequently In addition,its high spatial resolution can well observe the characteristics of convection.However,IR imager's observation ability is limited by its poor penetration to cloud and rain.It can only obtain the information from the cloud top,not the information inside the clouds.Therefore,it is necessary to also use the data from microwave(MW)instrument that can penetrate the clouds and provide information in the cloudy skies for better prediction of convective occurrence.Based on the observation data of AHI(Advanced Himawari Imager),an advanced imager onborad the new generation of geostationary meteorological satellite(Himawari-8),and combined with the methods of area overlap(AO),optical flow(OF)and brightness temperature threshold(BTT),this paper first identifies the convective initiation(CI),and then matches the observation data of ATMS,an advanced microwave sounder onboard SNPP(Suomi-NPP),for obtaining the dataset of CI.Using this dataset,the parallax corrections for both polar orbit and geostationary microwave sounder observations are analyzed,with focus on the factors affecting the parallax of observations of geostationary microwave sounder,which is scheduled be to onboard China's FengYun-4 series in future.In addition,the IR observation data set,GFS(global forecast system)data and ATMS observation data of CI over East Asia from 2016 to 2017 are analyzed using random forest(RF)method for deep learning,and the precipitation data of CMORPH is used as label data for classification to establish the early warning model of convective occurrence.The rankings from inputs that affect the accuracy and out-of-bag error show that it is very important to combine both high temporal resolution IR imager and microwave sounder data for early warning of CI,which can be achieved in future when both IR imager and MW sounder are onboard the same geostationary satellite.