Alternative parameter estimation method of extreme value copula model for skewed hydro-meteorological data

Hydrological phenomena are often multidimensional and hence several variables need to be considered at the same time. Therefore, multivariate probability model has been applied to frequency analysis for hydro-meteorological data which have two or more variables. Particularly, the copula model has been used as a useful tool for multivariate frequency analysis because the copula model has no limitation on selecting marginal distribution family. For inference of copula parameter, the maximum pseudo-likelihood (MPL) method is one of the most common method. However, skewness of variables can be suppressed because the MMPL method includes the Weibull plotting position formula in likelihood function. In this study, the alternative MPL method is presented by replacing the plotting position formula with others which can consider skewness of the data. The Monte-Carlo simulation experiment has performed for various conditions such as dependence between variables, sample size, and coefficient of skewness. Then, estimated parameters were compared to true (generated) value. Generally, the alternative MPL method was more accurate than the MPL method when correlation between variables are not too strong (less than 0.5).