Abstract:
Snow melt runoff is one of the main sources of streamflow in many of Himalayan Rivers. Conceptual models to simulate the snow melt runoff such as Snowmelt Runoff Model (SRM) and Snowmelt Model (SNOWMOD) require a large quantity of data which are generally not available for most locations in India. Applications of Artificial Neural Networks (ANN) in many water resources area indicate its better performance over other traditional models such as conceptual models and black box models. This paper discusses the development of ANN models for the simulation of streamflow at Rampur in Sutlej river basin. Rainfall, snowfall, temperature and discharge data of stations located at the upstream of Rampur were used as input to the models. Different combinations of significant lagged series of rainfall, snowfall, temperature and discharge data, determined from statistical parameters such as auto correlation function (ACF), partial auto correlation function (PACF) and cross correlation function (CCF), were used as input to the model. The performance of the model was evaluated using statistical criteria such as coefficient of correlation, root mean squared error (RMSE) and model efficiency. The results of the best ANN model during the calibration indicate that the all range of discharge values were simulated fairly well. However, the medium and high range values of discharge slightly deviated from the observed values during the validation of the model. The overall performance of the model, as exhibited by the various statistical criteria, indicates the suitability of ANN modelling technique to reasonably simulate the streamflow at Rampur in Sutlej river basin. Further, the development of two separate ANN models for simulating the low, medium and high did not yield better performance than the generalized ANN model with continuous data.
