This paper investigated the differences between conventional moments (C-M) and L-moments (L-M), when being appliedto estimate parameter and design rainfalls in the Taihu Lake Basin. It introduced to the theory of L-moments, and provided theadvantages of L-moments over conventional moments by comparison. With data of the Taihu Lake Basin, the performance of thesetests was taken via a series of Monte Carlo simulation experiments. The four stations with longer rainfall data series were chosen toestimate design rainfalls by fitting the P-Ⅲ distributions based on the conventional moments and L-moments，respectively. The results show that L-M-based tests are more reasonable on annual extreme precipitation estimation and have better performance interms of unbiasedness and robustness on parameter estimation over the C-M.