Applying copulas to predict the multivariate reduction effect of best management practices.

Abstract Best management practices (BMPs) have been widely applied to mitigate non-point source (NPS) pollution in agricultural watersheds. However, a prediction of the multivariate reduction effect of NPS pollutants by BMPs considering its stochastic nature has not been conducted. A new modeling approach combining a hydrological model and copulas was proposed to predict the multivariate effect of BMPs fully considering the stochastic characteristics of BMPs effects and the dependence structure between them. Two levels of reduction effect, i.e., the multi-indicator effect of a single BMP and the combined effect of multiple BMPs, were simulated. The approach was demonstrated in Zhangjiachong watershed, a typical small watershed in the Three Gorges Reservoir area, China. Results show that copulas can effectively simulate the dependence between the univariate effects of BMPs. The approach can accurately predict the probability to achieve the reduction objective for multiple pollutants and multiple BMPs in a watershed. It provides a stochastic way to predict the multivariate effect of BMPs and has great potential to be widely applied in BMPs related decision making.