Streamflow simulation using conceptual and neural network models in the Hemavathi sub-watershed, India

Abstract

Water is one of the most valuable natural resources and a major element of a state's and country's socio-economic growth. The world's water resources and India are under huge pressure because of rising demand and a limited supply. Proper water management is the only solution for ensuring a close gap between demand and supply. Hydrological modeling offers an answer to this issue by establishing relationships between different hydrological processes. Several models have been developed in the past decade to simulate the rainfall and runoff relations. Some models are simple conceptual models based on spatially distributed event-based or continuous and artificial intelligence (AI) models. This study aims to compare two conceptual daily-based models and one AI model developed for the Hemavathi sub-watershed in the Cauvery Basin (India). Two daily runoff models are implemented using conceptual models, i.e., Sacramento and the Australian water balance model (AWBM) using Rainfall-Runoff Library (RRL) tool and Feed forward Backpropagation neural network (FFBPNN) model. The models were calibrated for daily streamflow values from 1990 to 2006 and then validated from 2007 to 2015. The effectiveness of model runoff predictions is evaluated using statistical parameters such as Nash-Sutcliffe efficiency (NSE) and Correlation coefficient (CC) values. The NSE values for the FFBPNN is 0.88 (calibration) and 0.74 (validation), Sacramento model is 0.66 (calibration) and 0.48 (validation), and 0.63 (calibration) and 0.44 (validation) for the AWBM model. From the obtained results, the FFBPNN model performs well in terms of NSE and CC compared to Sacramento and AWBM models.