A purely statistical discharge–sediment relationship ignores that the same discharge carries different sediment loads depending on season, on where the storm sits on the hydrograph, and on whether the load is rising or falling. Disentangling those controls at large scale had not been explored.
A stochastic sediment-yield model rooted in copula theory builds a joint distribution between discharge and suspended sediment load for storm events. The observations are first classified by the processes governing them — seasonality, hysteresis pattern, and hydrograph component — and watershed hydrology, land use, and geology are included to explain the differences. It is evaluated on 67 unregulated streams across the United States.
Sediment transport is a major contributor to a non-point source of pollution impacted by various factors that are modulated by climatic changes and anthropogenic influences. Quantifying and disentangling the contribution of these factors to sediment yield at large scales and across different flow regimes has not been fully explored. Here we present a framework to fine-tune a stochastic sediment yield model by classifying discharge and Suspended Sediment Load (SSL) observations based on the underlying governing processes in unregulated streams with various hydrological regimes. This stochastic model, rooted in copula theory, constructs a joint distribution between discharge and SSL storm events using historical time series of observations, classified based on seasonality, hysteresis patterns, and hydrograph components of the sediment transport processes. We include hydrological, land use, and geological properties of the watersheds to describe and discuss the effects of different factors on applying the underlying dynamics to enhance sediment yield estimation/prediction accuracy. We evaluated the proposed method on 67 streams across the United States. Our results show significant improvements in sediment yield modeling performance.