The Upper Narmada River Basin's flood frequency analysis using the GAMLSS technique

ABSTRACT Flood frequency analysis (FFA) is essential for effective flood risk management, predicting flood event probabilities over time. The Generalized Additive Models for Location, Scale, and Shape (GAMLSS) technique is used in this study to get around the problems with traditional FFA methods, which usually rely on stationarity and well-known distributions. The objectives are to conduct a comprehensive flood frequency analysis of the Upper Narmada River Basin using GAMLSS and improve flood risk assessments considering non-stationary data and covariates like rainfall, temperature, and land use changes. The study collects and analyses hydrological and meteorological data. Change points are identified using the Pettitt test, and trends are analyzed with the modified Mann–Kendall test. The Pettitt test results indicate significant change points, while the modified Mann–Kendall test reveals decreasing trends in all stations, with p-values ranging from 0.00251 to 0.6985. GAMLSS model's goodness-of-fit is assessed using various probability distributions, including gamma, Weibull, Gumbel, logistic, and log-normal. The log-normal distribution performs best in four stations, while Weibull and Gamma are optimal in others. Model selection is based on the Akaike Information Criterion (AIC), with values ranging from 508.14 to 663.