Predicition of Floods of Indus River at Guddu Barrage by Fitting Gumbel and Generalized Extreme Value Probability Distributions

Abstract

Gumbel and Generalized Extreme Value (GEV) distributions are fitted on 38 years (1962-1999) maximum flood series for river Indus at Guddu barrage, Pakistan. The distribution parameters are estimated by Method of Moments (MOM), Maximum Likelihood Method (MLM) and Power Weighted Moment (PWM). Gumbel distribution appears to be appropriate for modelling the flood data on the basis of Goodness of fit tests, Standard Errors (S.E's) of Quantile estimates coefficients, of Skewness (C.), Kurtosis (C₁) and sample L-Moment ratios and MLM is the most efficient method. The flood estimates at various return periods suggest that authorities should take appropriate measures for safety of the barrage as 100-years flood prediction at Guddu is 1.5 million cusecs, which is much higher than its existing capacity of 1.1 million cusecs