Sediment is a major source of non-point pollution. Suspended sediment can transport nutrients, toxicants and pesticides, and can contribute to eutrophication of rivers and lakes. Modeling suspended sediment in rivers is of particular importance in the field of environmental science and engineering. However, understanding and quantifying nonlinear interactions between river discharge and sediment dynamics has always been a challenge. In this paper, we introduce a parsimonious probabilistic model to describe the relationship between Suspended Sediment Load (SSL) and discharge volume. This model, rooted in multivariate probability theory and Bayesian Network, infers conditional marginal distribution of SSL for a given discharge level. The proposed framework relaxes the need for detailed information about the physical characteristics of the watershed, climatic forcings, and the nature of rainfall-runoff transformation, by drawing samples from the probability distribution functions (PDFs) of the underlying process (here, discharge and SSL data). Discharge and SSL PDFs can be simplified into a joint distribution that describes the relationship between SSL and discharge, in which the latter acts as a proxy for different predictors of SSL. The joint distribution is created based on historical discharge and SSL data, and stores information about the discharge-SSL relationship and sediment transport process of the watershed of interest. We test this framework for seven major rivers in the U.S., results of which show promising performance to predict SSL and its likelihood given different discharge levels.
ASJC Scopus subject areas