Modeling rainfall event characteristics using D-vine copulas

Investigation of precipitation characteristics is necessitate in understanding and predicting phenomena of precipitation such as runoff and flood. Therefore in this study, dependence among the main characteristics of a rainfall event (i.e., rainfall depth R, maximum rainfall depth M, wet period L, and dry period D) were modeled using D-vine structure. Firstly, different multivariate probability distributions were built, making all the permutations of the conditioning variables and then Archimedean and Elliptic copulas were used for fitting each pair-copula. The best copula family was selected for fitting on each pair-copula according to different criteria. In the next stage, M-R-D-L structure, i.e., with D conditioned by L, R by D and L, and M by R, D, and L, was known as the most suitable structure considering to AIC and BIC criteria. Finally, rainfall event characteristics were simulated using the selected structure. In order to evaluation of simulation accuracy of proposed model, the main statistics of simulated variables were compared with those of observed variables. The results showed that the majority of simulated statistics have good accordance with observed statistics.