Abstract: The size of hails directly affects the severity degree of weather disasters. To meet the need of hail size identification, based on statistical data of hails detected by two S-band dual-polarization radars in Jinan and Qingdao, a hail dataset is trained first, and the probability distributions of radar horizontal reflectance factor (Z_H), differential reflectance (Z_DR) and correlation coefficient (CC) for small, large and giant hails are then obtained. Finally, the hail size discrimination model (HSDM) based on Bayesian method is developed and verified by two supercell storm processes. The study yields the following results. (1) The results obtained from HSDM are consistent with the real situation and agree with the analysis of the scattering and polarization parameters characteristics of different size hails and dynamic and microphysical characteristics for supercell hail storm. (2) Horizontal and vertical distribution characteristics of hail size agree with the precipitation particles filtering mechanisms and the hail growth mechanism of supercell storm. Large hails and giant hails are mainly located in strong echo area near the front side of the V-shaped groove. On the lower side of the suspension echo of supercell, small hails are found. Large hails and giant hails are mainly generated in the high reflectance area with large gradient above the weak echo area and then fall along the strong echo wall, but smaller hails fall on the side far away from the updraft. (3) Small hails near the updraft are mainly distributed in the Z_DR column, K_DP column and between them, large hails and giant hails are mainly located on the other side of the K_DP column away from the Z_DR column. The landing concentration of heavy hails can be determined comprehensively by combining the identification results at the height below 2 km.