Using the skew-t copula to model bivariate rainfall distribution

Abstract

We simulate monthly rainfall at two sites in the Murray-Darling Basin. In order to construct a suitable joint distribution, we model the individual totals using appropriate gamma distributions and use a multivariate skew-t distribution to construct an appropriate copula. The skew-t distribution is considered robust as it includes both skewness and tail dependence structure and allows us to model correlations. We investigate the characteristics of a bivariate skew-t distribution and show how adjusting the parameters generates simulated data which matches the observed data. © Austral. Mathematical Soc. 2010.