Abstract: In order to construct the cloudradiation database, 11 precipitating systems were simulated by the MM5 to obtain the rainfall profiles whose upwelling radiative brightness temperatures were computed by the MonteCarlo radiative transfer model. Then as the priori databases, the Db_m1 and Db_m2 were extracted from the cloudradiation database by using check list. The rain rates of three precipitation systems during July 2007 were retrieved by importing Db_m1 and Db_m2 to the GPROF algorithm respectively. Comparing with the results from the TMI and three statistic algorithms, the rain rate from Db_m2 was close to the TMI's and better than that from Db_m1 obviously, even better than those from NESDIS, GSCAT and PCTSI algorithms sometimes. The rainfall intensities from the PR and the Doppler radar on land surface were used to inspect the results calculated by the aforementioned methods. It is found that the Db_m2 and TMI are preferable and the result from the PCTSI algorithm is better than the other statistic algorithms. Among the 4 kinds of hydrometeors retrieved by Db_m2, the precipitable water (pw) displays consistent with that from the TMI or Doppler radar on land surface. The vertical distributions of the other three hydrometeors including cloud water (cw), cloud ice (ci) and precipitable ice (pi) are close to those from the TMI. However the contents of those 3 kinds of hydrometeors are several times as large as those from the TMI respectively. The hydrometeor structures obtained from the Db_m1 are consistent to those from the Db_m2, except that the content of pw is lower, and so does cw sometimes. Thus, the algorithm constructed in this study has a positive effect on retrieving rainfall intensity and the structures of hydrometeors to some extent. It can be regarded as a successful attempt to improve the physical algorithm applied in China with which the rainfall intensity and hydrometeor structures can be retrieved from brightness temperatures. 