At-Site Flood Quantile Estimation Using Multiparameter Distributions in the Gandak River, Himalayas

Modelling peak flood discharges and estimating flood frequency on a river provides essential information for designing and managing hydraulic structures. Bihar is one of the most flood-prone Indian states, with numerous Himalayan-originated rivers in Northern Bihar experiencing frequent flooding. The study aimed to analyse the at-site flood frequency to estimate peak discharge in the Gandak River basin, India. For the analysis, annual peak discharges from the Triveni gauging site, with a 46-year data record, were used. Using L-moments, the maximum likelihood estimate, and the method of moments, the parameters of ten advanced multiparameter distribution models were estimated. Quantiles of different return periods were estimated using the parameter estimation methods. Three types of goodness-of-fit tests, Kolmogorov–Smirnov, Anderson–Darling, and the Chi-squared were performed to identify the best-fit distribution among the models. The results showed that the Generalised Extreme Value (GEV) and Log-Pearson 3 distributions are the most suitable in the study area. The near-perfect positive relationship between the observed and simulated data confirmed the selection of the most suitable models. The study recommended adopting a return period of more than 500 years for the design of hydraulic structures in the Gandak River basin.