Frequency Analysis of Annual Maximum River Flow by Generalized Extreme Value Distribution with Bayesian MCMC

Abstract

The aim of this paper is to fit 9 annual maximum river flows in Sabah for a period record of over 20-48 years into the generalized extreme value (GEV) distribution. Bayesian Markov Chain Monte Carlo is employed as the parameter estimation which is believed to provide a more robust inference through prior and posterior distribution. In this study, scale parameter is being associated with the linear trend function. Based on the 95% credible interval in this study, the results suggest that the additional covariate to the model has no impact at most of the river sites. Hence, return level with 10and 100year for each river sites have been obtained by using a simple model which is urged in substituting complex models such as logistic model.