Hydrological Process Modelling using RBNN - A Neural Network Computing Technique

Abstract

Severe erosion in the watersheds under Damodar Valley Corporations (DVC), Hazaribagh, Jharkhand, India has been taking place for a long time. Several soil and water conservation measures are being adopted by the Soil Conservation Department in Damodar Valley. For effective planning of soil conservation programmes, hydrologic models have been used, but with several limitations. In the present study, Nagwa watershed under DVC was selected for simulating monthly surface runoff and sediment yield using radial basis neural network (RBNN) model, which requires lesser data. Different sets of input data were employed and it was observed that only average monthly rainfall was sufficient to estimate surface runoff, while average monthly rainfall and average monthly discharge were needed for estimating sediment yield. The RBNN model was trained for the years 1991-2000 and validated for the period 2001-2007. Results indicate that coefficient of determination, R2, Nash-Sutcliffe simulation efficiency, NSE and root mean square error, RMSE (m3.s-1) values for RBNN model were 0.93, 0.92 and 1.25 during training, and 0.73, 0.73 and 0.77 during validation period, respectively. The model performed quite well for simulation of sediment yield with R2, NSE and RMSE values of 0.92, 0.92 and 1822 during training period; and 0.80, 0.70 and 2288 during validation period, respectively. It could be stated that RBNN model based on simple inputs could be a powerful tool for simulation of monthly surface runoff and sediment yield