Modeling Stage–Discharge and Runoff–Sediment Dynamics using Soft Computing and Statistical Approaches in the Nagavali River Basin, Andhra Pradesh

Abstract

Soil and water are critical natural resources underpinning agricultural productivity and ecological sustainability, particularly in monsoon-dominated regions such as the Nagavali River basin in eastern India. Effective estimation of runoff additionally and sediment yield hand in hand is essential for workable watershed planning, hydraulic structure design, and sediment management, yet remains evergreen challenging due to the stochastic and nonlinear nature of hydrological processes. The Artificial Neural Network (ANN) and Adaptive Neuro-Fuzzy Inference System (ANFIS), two data-driven soft computing techniques, are assessed in this study in conjunction with Multiple Linear Regression (MLR) and Sediment Rating Curve (SRC) methods for modeling daily stage–discharge and runoff–sediment relationships. The Water Resources Information System (WRIS) provided monsoon-season stage, discharge, and suspended sediment content data for a total of twelve years (2001–2012), with 2001–2009 used for training and 2010–2012 for testing. While ANFIS models used Gaussian and triangular membership functions with hybrid learning, ANN models used feed-forward back-propagation with Levenberg–Marquardt optimization. RMSE, correlation coefficient (r), coefficient of efficiency (CE), and pooled average relative error (PARE) were applyed to evaluate the model's performance. Results indicate that soft-computing models outperform traditional approaches for both runoff and sediment prediction. ANFIS with triangular membership functions demonstrated the highest accuracy, followed by doublehidden-layer ANN. MLR provided acceptable results, whereas SRC exhibited limited capability due to pronounced nonlinearity in sediment–runoff dynamics.