Study On The Method Of Classification And Recognition Of Small Convective Cells

Intense convective thunderstorms such as hail,thunderstorms,locally heavy rainfall and other weather disasters mainly occurs in in the convection system,on the mesoscale meteorological weather systems.The horizontal scale of these thunderstorm systems is generally less than 200 kilometers,some even only a few kilometers.Obvious features of hail,heavy rainfall does not appear in a small system within 50 km,so severe weather classification model suitable for large-scale cells does not apply to small cells.In order to solve the problem of automatical classification for thunderstorm systems whose horizontal scale are within 50 km classification,small single new features are constructed based on data got from a new generation of weather radar and radar reflectivity graph,and appropriate machine learning methods are selected,to construct the classification model to achieve the classification of hail,heavy rainfall and severe weather triggered by small cells.According to the different characteristics of small-sized hail and heavy rain cells on the radar reflectivity graphs,the system structure density characteristics are constructed.The feature is got by calculating the average distance between the cell contour and cell center contour in 72 directions.Affected by vertically integrated liquid water content,the paper constructs small cell liquid water content.Reflectivity values are equally converted to liquid water content,and integrated between 0 ?layer and-20 ? layer in the cell region.The reason that small cell generates heavy rainfall in one area is also relevant with the length of time the cell remain in the region,by which the paper construct moving rate.The moving rate is the specific value of the displacement relative to the initial time and the initial time span width in the direction of the final displacement.Meanwhile,the moving rate of provides the cumulative weight for the accumulation of liquid water content.These features not only have significant difference between hail cell and heavy rainfall cell,but also can collectively describe complex systems generating both hail and heavy rainfall.In this paper,based on these features,a small single hail \ heavy rainfall \ non-severe weather classification decision tree model is built.In line with the accumulation of liquid water content which is associated with the use of time,iterative mechanism is introduced in the realization of the classification tree.Experiments show that the recognition rate of systems generating short-time heavy precipitation is as high as 89.1% and the false positive rate is 9.5%,besides the recognition rate of systems generating hail is 79.8% and the false positive rate is 3.5%.The average critical success index(CSI)reaches 80.0%.