Improvement of Regression Simulation in Fluvial Sediment Loads

Abstract

In the mid-Atlantic region of the United States, sediment loads from stream runoff generally change more rapidly in the rising limb than in the falling limb of a storm hydrograph. As a result, sediment load reaches its peak prior to flow peak, an observation known as clockwise hysteresis. This dynamic load–flow relationship is poorly reproduced by the existing multivariate linear regression models. This paper explores regressors that attempt to incorporate observed features in a statistical model and thus improve load estimates. These included inverse discharge and flow-change regressors. The load estimates using three regression models for eight rivers are compared, and recommended regression equations are proposed.