Scheming of Runoff Using Hybrid ANFIS for a Watershed: Western Odisha, India

Various types of techniques have been utilized for predicting runoff which involves empirical and conceptual models. Owing to complications involved in hydrological processes, accurate runoff prediction is difficult utilizing physical based watershed or linear recurrence relationships. Application of Back propagation neural network (BPNN), Adaptive neuro fuzzy inference system (ANFIS) and integration of grey wolf optimiser (GWO) with ANFIS are proposed here to predict runoff. In this work two scenarios are considered for developing the models at Baramba, Cuttack and Niali watershed, India. Scenario one is computed with inputs precipitation, minimum and maximum temperature and Scenario two is computed with scenario one input added to infiltration capacity of soil to predict runoff as output. To develop the model Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Coefficient of Determination (R2) for both scenarios are considered as model performance indicators for developing the model. For Niali watershed, ANFIS-GWO performs best with Gbell membership function having R2 value 0.9262 in training and 0.9158 in testing for scenario two than the other two techniques. Similarly at Cuttack and Baramba watershed, ANFIS-GWO model proves best value of performance with R2 are 0.97246, 0.97198 and 0.954, 0.95213 during training and testing phases respectively. In both the scenarios BPNN performs poor as compared to ANFIS and ANFIS-GWO. From comparison of results it is establish that addition of infiltration capacity of soil to the model architecture accelerates model performance. Application of back propagation neural network, adaptive neuro fuzzy inference system (ANFIS) and integration of grey wolf optimiser with ANFIS are proposed to predict runoff. In this chapter, two scenarios are considered for developing the models at Baramba, Cuttack and Niali watershed, India. Hydrological modelling is an influential method for investigating hydrologic systems essential for practising water resource engineers and research hydrologists working on design and improvement of combined procedure for water resources management. Runoff prediction is one of the most valuable processes involved in hydrological systems. Precise runoff estimation by utilizing evaporation, rainfall, and other hydrologic parameters is a significant problem in water resources engineering. Cuttack is the second largest city of Odisha, India, and one of the oldest. Back-propagation network helps i