Abstract: Phased array radar is faster in scanning speed than mechanical scanning radar, but due to the influence of antenna structure and attenuation, phased array radar will produce larger system error and random error. The mainstream fuzzy logic hydrometeor classification method has little difference in the weight coefficients of each parameter, and the calculated composite values are often close, so the hydrometeor classification results are easily affected by data errors. The detection data of the X-band dual polarization phased array radar at Qiuyutan, Shenzhen, from March to September in 2020 are compared with the S-band dual polarization radar at the same location. The points close to the elevation, azimuth and radial distance of the two radars are obtained to establish matching datasets to calculate the errors of X-band dual-polarization phased array radar. Quantitative analysis of the causes for the introduction of errors through certain restriction conditions reveals that the calibration error and random error of the reflectance factor Z_H and the differential reflectance Z_DR are relatively large. The error range of Z_H is -0.5-4.5 dB, and the error of Z_DR is -0.7-0.2 dB. After the preliminary correction of calibration error and random error, it is found that there are still some errors in data, which make the hydrometeor classification result of fuzzy logic method unreliable, so the decision tree hydrometeor classification method with the basic structure of binary tree is established according to the characteristic range of radar parameters of different hydrometeors and the height of the melting layer. In order to verify the practical application effects of the above methods, the error sensitivity of hydrometeor classification results and the rationality of hydrometeor spatial distribution are evaluated respectively. Typical examples are selected to further evaluate the rationality of the decision tree hydrometeor classification method by comparing the parameters and hydrometeor classification results of X-band dual polarization phased array radar and S-band dual polarization radar. The evaluation results show that the stability of the decision tree hydrometeor classification method is higher than that of the fuzzy logic method, and the hydrometeor distribution in the convective cloud is more reasonable, which can give full play to the advantage of X-band dual polarization phased array radar in studying the phase evolution of particles in the cloud.