Research On Hydrometeor And Cloud Recognition Based On Fuzzy Logic And Deep Learning

Clouds are visible aggregates of hydrometeor in the air that floating in the atmosphere after condensation.It is the external representation of atmospheric thermodynamic and dynamic processes,and also an important carrier of water vapor flow and circulation processes.Different precipitation clouds often accompany different precipitation processes.Accurate identification of precipitation clouds is significant for the prediction of severe precipitation processes.The identification of hydrometeor in clouds can help people to understand the variation of hydrometeor phase in precipitation process.It plays an important role in the research of radar quantitative precipitation estimation,weather modification and aviation meteorology.In this paper,we will explore the identification of precipitation clouds and hydrometeors.Based on the differences of radar reflectivity distribution morphology between stratiform and convective precipitation clouds in three-dimensional space,the conventional method of precipitation cloud identification utilize some parameters calculated from radar reflectivity to identify the types of precipitation clouds.The accuracy of the traditional methods depends on the reliability and quantity of the selected parameter.However,the increment of identification parameters leads to the multiplication of computation,which causes the great reduction of the efficiency of the algorithm.In this paper,a precipitation cloud identification method based on deep learning is proposed,including two parts(i.e.,data preprocessing and precipitation clouds identification).During the data preprocessing,in order to obtain the distribution of radar reflectivity at multiple altitudes,the adaptive Barnes interpolation method is used to perform CAPPI(Constant Altitude P1 an Position Indicator)inversion by radar volume scan data.Subsequently,the Faster-RCNN model applies the reflectivity data of several altitude layers as input to predict the area of convective cloud.Meanwhile,the non-convective cloud area is regarded as the stratiform cloud.Finally,the experimental results show that the identification results of this method based on the Faster-RCNN are basically consistent with those traditional methods.However,this method boasts great advantages in running time and adaptive ability.In recent years,numerous scholars in domestic and foreign conducted a lot of studies about the identification of hydrometeors,and obtained some good results.CINRAD-SA radar,referred as the dual-polarization Doppler weather radar,is produced by the BEIJING METSTAR RADAR CO.LTD in China.There is little literature on the identification of hydrometeors using the data detected by the CINRAD-SA radar.In this paper,a hydrometeor identification method based on fuzzy logic algorithm using the CINRAD-SA radar data is proposed,which mainly includes four steps(i.e.,fuzzification,inference,aggregation,and defuzzification).The experimental results demonstrate that the identification results of this method are basically correct and reasonable,which can be used as a reference for forecasters and researchers.