L-band sounding data recorded from January 2008 to December 2009 and responding spatial and temporal Cloudsat data are used to analyze the cumulative frequency distributions of relative humidity in and near clouds. It is determined that the accuracy of cloud discrimination can reach 81% by setting the relative humidity threshold at 75%. In addition, various thresholds of relative humidity for discriminating clouds are evaluated on the bases of bias score (BS) and threat score (TS). By setting the relative humidity at 81%, the highest TS score of 0.66 is achieved. Moreover, relative humidity thresholds at various heights are also analyzed on the bases of BS and TS. The results show that the relative humidity threshold with the highest TS score among thresholds at the same cloud height decreases when the height increases. The TS score of the discriminate cloud area at these thresholds at responding altitudes is higher than 0.6, and the accuracy is more than 84%, which is significantly better result than that obtained by using a single relative humidity threshold for all heights. Finally, after optimizing these thresholds, a method of discriminating the cloud vertical structure is proposed by using L-band sounding data.