Application of best subset method for river water quality modeling: A study on Godavari River, India

Abstract

The utilization of water quality analysis to inform optimal decision‐making is imperative to achieve sustainable management of river water quality. A multitude of research works in the past has focused on river water quality modeling. Despite being a precise statistical regression technique that allows for fitting separate models for all potential combinations of predictors and selecting the optimal subset model, the application of best subset method in river water quality modeling is not widely adopted. The current research aims to validate the use of best subset method in evaluating the water quality parameters of the Godavari River, one of the largest rivers in India, by developing regression equations for different combinations of its physicochemical parameters. The study involves in formulating best subset regression equations to estimate the concentrations of river water quality parameters while also identifying and quantifying their variations. A total of 17 water quality parameters are analyzed at 13 monitoring sites using 13 years (1993–2005) of observed data for the monsoon (June–October) period and post‐monsoon (November–February) period. The final subset model is selected among model combinations that are developed for each year's dataset through widely used statistical criteria such as R2, F value, adjusted R2a, AICc, and RSS. The final best subset model across all parameters exhibits R2 values surpassing 0.8, indicating that the models possess the ability to account for over 80% of the variations in the concentrations of dependent parameters. Therefore, the findings demonstrated the appropriateness of this method in evaluating the water quality parameters in extensive rivers. This work is very useful for decision‐making and in the management of river water quality for its sustainable use in the study area.