Comparison of generalized regression neural network and MLP performances on hydrologic data forecasting

The estimation and forecasting of the hydrologic data carry significance for many water resources engineering problems. Establishing sediment monitoring instruments on rivers is a costly operation. The methods available in literature for sediment concentration estimation are complicated, time consuming and necessitate cumbersome parameter estimation procedures. Artificial neural networks have been applied to many kinds of hydrologic data within the last two decades. In the majority of these studies standard feed forward multilayer perceptron is employed, In this study the generalized regression neural networks are applied to the selected daily flow and sediment concentration data together with the multilayer perceptron. In both the forecasting the sediment values using the previous observed sediment values and the sediment concentration estimation using the observed river flow values the generalized regression neural networks found superior to the feed forward multilayer perceptron.