A factorial Bayesian copula framework for partitioning uncertainties in multivariate risk inference

In this study, a factorial Bayesian copula (FBC) method is proposed to quantify parameter uncertainties in copula-based models and then reveal their impacts on hydrologic risk inferences within a multivariate context. In detail, Bayesian inference and factorial analysis are integrated into copula-based multivariate risk models to (1) quantify parameter uncertainties, (ii) reveal their individual and interactive effects, and (iii) identify their detailed contributions to uncertain risk inferences. Streamflow observations at Xiangxi and Wei River basins is China are used to illustrate the applicability of FBC. The results indicate that imprecise parameters in marginal distributions and the dependence structure would lead to extensive uncertainties in predictive joint return periods and failure probabilities. Also, individual and interactive effects of parameters are well revealed through multilevel factorial analysis, and the detailed contributions of one parameter to different failure probabilities under different service time scenarios are identified.