A Simulation Study for Effects of Various Factors Based on Generalized Logistic Distribution

Abstract

The performance of the index flood method for 4 regions whose specifications are representatives of annual maximum rainfall data in South Korea, is investigated. The Monte Carlo simulation is performed to investigate the effects of at-site and regional frequency analysis, sample sizes, intersite dependence, heterogeneity, and applied frequency distributions on the accuracy of quantile estimates. Simulation experiments show that the regional frequency analysis is more accurate than the atsite frequency analysis even in the regions with moderate amounts of heterogeneity, intersite dependence, and misspecfication of the frequency distributions. Heterogeneity increases the relative bias of at-site estimates. Intersite dependence increases the variability of estimates but has little effect on the relative bias. As the nonexceedance probability increases, the performance of the regional frequency analysis over the at-site frequency analysis improves. It is also found that the misspecification of the frequency distribution is more important than heterogeneity as a source of error for large return periods.