A new retrieval method for deducing path-integrated cloud liquid water content (LWC) by airborne upward-looking microwave radiometer is introduced in this paper. The coefficients of retrieval equation are obtained through the statistic regression method like the former method, but the statistical samples are produced by the one-dimensional stratiform cloud model. The new method is compared with the former one in which historic radiosonde data are employed to determine the vertical distribution of LWC by testing relative humidity above some threshold value. Besides, the detection errors are estimated through factor analysis and numerical simulation tests. The case on 8 July 2001 in Jilin Provence shows that the retrieval errors are diminished to some extent because the new statistical samples are consistent with the synoptic observations compared with the former method. The numerical simulation results indicate that the statistical relative errors decrease to 9.5%－12.7% and the retrieval accuracy is better than the former method at all altitudes. The factor analysis indicates that the errors caused by the uncertainty of the vertical distribution of LWC cannot be ignored, compared with the errors caused by the background and the instrument drift errors. In order to improve the retrieval accuracy, the authors use selected cloud samples which are supplied by a cloud model to regress retrieval coefficients according to the synoptic observation. The numerical simulation tests indicate that the retrieval accuracy can be improved at all altitudes.