Modelling re-aeration of rivers using predictive models and developed ANN models under varying hydrodynamic conditions

Abstract

ABSTRACT Re-aeration serves as an integral part of water quality modelling, being utilized as an imperative constraint for estimating dissolved oxygen in rivers. The methods available to estimate oxygen regeneration at the air–water interface are broadly applicable and scalable. The study introduced various artificial neural network (ANN) models designed to estimate the re-aeration coefficient under varying hydrodynamic conditions. Five-year datasets of Yamuna River exemplified reduced flow and heavy organic loadings were used for performance assessment of the models. The ANN models were developed using different combinations of input parameters. The available predictive re-aeration models are compared with the developed ANN models to identify the best-performing model. The ANN models containing water quality and hydraulic parameters outperformed the predictive re-aeration models. The study reveals that both hydraulic and water quality parameters are required to analyse the re-aeration coefficient of degraded rivers.