Enhancing Flood Risk Assessment in Johor River Basin through Trivariate Copula

Abstract

Abstract Copulas have gained significant prominence as a powerful statistical tool with diverse applications, particularly in the field of hydrology, where they facilitate the measurement of complex relationships among various flood characteristics. Three primary flood features are considered which are peak discharge, flood volume, and flood duration, and their interdependencies are examined using copula functions. Trivariate copula is employed to capture the interrelation between these flood variables since bivariate and univariate flood frequency analyses have several shortcomings where they are unable to consider all three crucial flood factors simultaneously. In light of the presence of extreme values in flood variables, the L-Moment is proposed to estimate the parameters of the marginal distributions. This is due to its enhanced reliability and susceptibility to outliers and extreme values, unlike the commonly used parameter estimation in flood frequency analysis, Maximum Likelihood Estimation (MLE) and Inference Function Margin (IFM). Akaike Information Criterion (AIC) was employed to identify the best fit marginal distribution and copula. The Lognormal distribution performs well in modeling peak discharge, while the Weibull and Generalized Extreme Value (GEV) distributions provide the best fits for flood volume and duration characteristics, respectively. Several widely known copula including Elliptical and Archimedean copula families are analyzed. After assessing the dependence structure between flood variables, the Clayton copula emerged as the most suitable choice. It is expected that if more flood features are combined, the return period would be higher means the event is less likely to occur if all flood factors considered simultaneously, and it was proven that the AND-joint return period has higher return periods compared to the OR-joint return period. This comprehensive analysis facilitates improved hydrological modelling and flood risk assessment in Johor River Basin, Malaysia, by employing the L-Moment method for estimating flood distribution parameters.