Research On Hail And Short-term Intense Precipitation Recongition And Hail Intensity Estimation Method

Severe convective weather such as hail and short-term intense precipitation has the characteristics of short life cycle,strong locality and small spatial scale,and it is very difficult to make accurate forecasts.In addition,because of its strong destructive nature,it has a serious impact on people’s production and life.Although many achievements have been made in the study of hail and intense precipitation,most of them are aimed at identifying single weather types such as hail or intense precipitation,and there are few quantitative studies on the prediction of hail intensity in China.Based on Doppler weather radar data,this thesis studies the use of pattern recognition and machine learning to construst a recognition model for hail and short-term strong,and the construction method of hail intensity estimation model is discussed.The main work of this thesis is as follows:(1)Convert the radar base data format to the three-dimensional grid data format.In this thesis,the bilinear interpolation method is used to obtain the radar reflectivity three-dimensional grid point data.Under this data format,six mechanism features and two elevation features of hail in the S-band are reconstructed.(2)The classification and recognition model of hail was constructed.Due to the lack of perfect records of hail in China,this paper first uses web crawler technology to collect hail information on microblog platform,and obtains effective hail information through manual screening to make samples.A hail recognition model based on support vector machine is constructed by using 8-dimensional characteristics of hail.Through experimental tests,the model has a hit rate of hail reaches 93.0%,and critical success index reacheds 80.2%.(3)A classification and recognition model for short-term intense precipitation was constructed.This thesis designs the characteristics of short-term heavy precipitation based on three-dimensional grid data,including density characteristics,intensity characteristics,gradient characteristics and liquid water content characteristics.Taking non-heavy precipitation samples as a negative example,a short-term intense precipitation classification and recognition model based on support vector machine was constructed,and the hit rate of intense precipitation was 81.7%.This model is cascaded with the hail recognition model to form a two-level classification and recognition method of hail and short-term intense precipitation.(4)Based on the characteristics of hail mechanism and early identification of hail,this thesis proposes a hail intensity estimation model based on conformal prediction.This model combines random forest regression model and conformal prediction algorithm to estimate the size of hail,and the prediction error is less than the MEHS(Maximum Exprcted Hail Size)algorithm commonly used in the business system.Finally,an attempt was made to transform the hail size regression problem into a hail grade classification problem,and a hail intensity grade classification model based on random forest is constructed,which provides a research idea for hail intensity prediction.