The parameters of the P3 marginal distributions were estimated by L-moment method and are listed in Table 1. A chi-square goodness-of-fit and the Kolmogorov–Smirnov (K–S) test are used to test the assumption. The results in Table 2 show that the hypothesis could not be rejected at the 5% significance level because ⁠, and the P3 distributions can also pass the K–S test because D_n < D_n,0.95. The marginal distribution frequency curves of flood peak and 7-day flood volumes are shown in Figure 4, in which the line represents the theoretical distribution and the crosses represent empirical probabilities. Figure 4 indicates that these theoretical distributions can fit the observed data reasonably well.