RUNOFF AND SEDIMENT YIELD MODELING USING ARTIFICIAL NEURAL NETWORK: CASE OF MUGER WATERSHED, ABBAY RIVER BASIN, ETHIOPIA

Abstract

Surface runoff and sediment loadings are immense problems that have threatened water resources development in the Muger river basin. The main objective of the study is to model runoff-sediment yield for upper Muger catchment. This paper has not included identifying the influence of topography, land use, and soil on stream flow and sediment yield of the watershed and water resource management scenario. In order to overcome the objectives of the study the whole computation was performed by using MATLAB software supporting nntoolbox were used for input data preparation, analyzing and modeling purpose of the research. The effective application of neural networks to runoff-sediment yield modelling requires; firstly, selection of an appropriate neural network type. Secondly, selection of an appropriate training algorithm and determine an appropriate network structure. Thirdly, one must decide how to pre-and post-process input-output data. The model was trained and cross validated against measured flow and sediment yield data. Statistic measures (RMSE, NSE, and R2 ) were used to evaluate the performance of the model. The results of the model training and validation showed reliable simulation of daily stream flow (R2

=0.90 and NSE=0.90) during model training and (R2 =0.88 and NSE=0.79) during model cross validation. For sediment yield, the model performance of daily sediment yield (R2 =0.98 and NSE=0.99) during model training and (R2 =0.98 and NSE=0.99) during model cross validation period. Comparison of the results reveals that, the model results showed a fairly good and satisfactory agreement between daily observed and simulated streamflow and sediment yield respectively.