Modeling the dependency structure of four-dimensional flood variables using the copula approach

The four-dimensional flood variables namely peak flow, volume, duration, and peak time are modeled using a copula-based method in this study using an array of copula families which including Elliptical and Archimedean to model the dependency of the flood variables of streamflow data from Johor River Basin (JRB), Malaysia. Copula offers an adaptable approach to model the relationship between variables and complex dependence structures. Bivariate copulas were applied to address the limitation of assuming linear relationships in multivariate data analysis, enabling a more accurate representation of complex dependence structures that may not be adequately captured by traditional methods. Results indicate the Weibull distribution best fits peak flow variables with the lowest AIC values, Pearson Type-III and Gumbel distributions suit flood volume and duration, respectively, and Generalized Extreme Value (GEV) represents peak time effectively. The selection of optimal copula involves measuring AIC values to ensure the copula model can describe the dependency between these flood variables adequately and Joe copula was found to be the best copula to model the dependency between flood peak-volume variables. Due to the superior performance of the Frank copula, it was chosen as the best copula to model the bivariate distribution of the volume-duration and flood peak-duration pair. While the remaining pair were best modeled using Clayton copula. All in all, the research shows the significance of copula-based modeling techniques for complicated multivariate data and offers insights into the relationship between flood attributes that can be implemented in flood risk management plans.