Artificial neural network-based assessment of water quality index (WQI) of surface water in Gwalior-Chambal region

• Published in: International journal of energy and environmental engineering 2023-03, Vol.14 (1), p.47-61
• Main Authors: Chauhan, Shyamveer Singh, Trivedi, Manoj Kumar
• Format: Article
• Language: English
• Subjects: Alkalinity; Artificial neural networks; Back propagation networks; Biological activity; Dissolved solids; Energy; Model accuracy; Monsoons; Neural networks; Neurons; Original Research; Quality assessment; Quality control; Regression models; Renewable and Green Energy; Seasons; Surface water; Technology application; Total dissolved solids; Turbidity; Water analysis; Water quality; Water quality assessments; Water quality monitoring stations; Water sampling; Wind
• ISSN: 2008-9163; 2251-6832
• DOI: 10.1007/s40095-022-00521-5

In the complex environment of physical, chemical, and biological processes in water bodies, it is always challenging to model the water quality while dealing with nonlinear behaviour of water quality variables. To assess the surface water quality of Gwalior-Chambal region, India, this paper applies the artificial neural network (ANN) process as it has been proven an efficient process to handle the nonlinear data. During the pre- and post-monsoon season of 2021, a total 49 surface water samples in each season were collected and examined for various water quality parameters such as Turbidity, pH, Total Hardness, Total Alkalinity, Total Dissolved Solids, Chloride, SO 4 , NO 3 , PO 4 , DO, BOD, COD, and MPN. Noteworthy, all surface water sources were found unfit for drinking as the MPN value of all water samples was estimated more  than zero with mean value of 481.88 and 239.20 for pre- and post-monsoon season, respectively. The value of ANN based water quality index (WQI) ranges from 27.79 to 103.42 and 33.78 to 96.70 for pre- and post-monsoon seasons, respectively. In order to develop a consistent and accurate ANN based WQI model, the paper applies Levenberg–Marquardt's three-layer backpropagation technique in ANN structures that contains 13 contributing neurons, nine hidden neurons, and one output variable. Computed values of performance statistics justify the accuracy of developed model over multiple linear regression model in predicting the WQI in both seasons. The developed model is also expected to be useful for other than Gwalior-Chambal region for surface water quality assessment.