Bayesian Neural Networks Modeling and Sensitivity Analysis of Suspended Sediment Load (Case Study Aharchay River Iran)

The forecast of the sediment yield generated within a watershed is an important input in the water resources planning and management. In the current study the efficiency of Bayesian Neural Networks (BNN) and Support Vector Machine (SVM) models was examined for prediction of Suspended Sediment Load (SSL) of the Aharchay River. Stream flow sediment data rainfall and average temperature from 1986-2010 were used to train and test the applied SVM and BNN models. The BNN estimates were compared with SVM result in term of coefficient of determination (R2) and root mean square error (RMSE). The comparison results indicated that the BNN is superior to the SVM in estimating daily suspended sediment load. The study also includes an estimation of the relative importance of variables to determine appropriate input combinations. A method is used in this paper to calculate the relative importance of each input parameters showing that indicates that stream flow and rainfall (one previous day) are the most and least influential parameters with approximate values of %37.06 and %5.39 respectively.