Prediction of Suspended Sediment Using Hydrologic and Hydrogeomorphic Data within Intelligence Models

Accurate estimation of transported sediment by rivers plays an important role in water resources management. So the selection of proper methods for estimation of suspended sediment is an important objective. In this regard, application of intelligence models (e.g., ANN, SVR) have substantially improved the prediction of suspended sediment. An important step in suspended sediment modeling using these models is, the proper input selection because input vectors determine the structure of the model and, hence, can influence model results. In the most studies, only climatic and hydrological variables have been used as suspended sediment estimators using data-driven models. Therefore, this study was designed to determine effective and accessible geomorpholigical variables based on hydrologic understanding in suspended sediment estimation for the Tamar catchment. To accomplish this goal, the effect of an Index of Connectivity (IC) as a hydrogeomorphic input, in addition to the hydrologic inputs, using ANN and SVR models was investigated. Comparison of results indicated that using IC along with hydrological inputs improve models efficiency and this improvement is indicated by decrease in the root mean squared error (9.63% and 26.36%) and a noticeable increase in the Nash–Sutcliffe efficiency (25.80% and 21.85%) and in the coefficient of determination (13.20% and 45.94%) for ANN and SVR models, respectively. These results are valuable for water resources planning and management.