ANN-assisted Prediction of Groundwater WQI for Drinking-water Screening in Salur Mandal, Andhra Pradesh, India
G. Rupakumari
*
Department of Civil Engineering, AU College of Engineering (A), Andhra University, Visakhapatnam-530 003, Andhra Pradesh, India.
G. V. R. Srinivasa Rao
Department of Civil Engineering, AU College of Engineering (A), Andhra University, Visakhapatnam-530 003, Andhra Pradesh, India.
B. Kalyana Ramu
Department of Chemistry, Visakha Govt. Degree & PG College for Women (A), Visakhapatnam-530020, Andhra Pradesh, India.
*Author to whom correspondence should be addressed.
Abstract
Aim: Groundwater quality monitoring in hard-rock aquifer systems is essential for sustainable resource management. This study aims to assess groundwater quality using the weighted arithmetic Water Quality Index (WQI) method and to develop an Artificial Neural Network (ANN) model for reliable prediction of WQI in the Salur region of Andhra Pradesh, India.
Study Design: A quantitative analytical study integrating WQI assessment with predictive modeling was conducted using field-monitored groundwater quality data collected over a three-year period (November 2018 to October 2021).
Methodology: A total of 468 groundwater samples was collected from 13 monitoring stations between November 2018 and October 2021 and analyzed for fifteen physicochemical parameters adopting standard procedures prescribed by APHA 2017. WQI was computed using the weighted arithmetic method to classify groundwater suitability. A feed-forward ANN model with a 15–5–1 architecture was developed using the Levenberg–Marquardt algorithm. The dataset was randomly divided into 70% training, 15% validation, and 15% testing subsets using MATLAB toolbox.
Results: WQI results revealed that 77% of groundwater samples fall within the good category (WQI = 26–50), while 23% fall under the Poor category (WQI = 51–75), indicating moderate hydro chemical variability without critical contamination. The ANN model demonstrated high predictive accuracy, with Root Mean Square Error (RMSE=0.438) and coefficient of determination (R²) of 0.998
Conclusion: The developed ANN model effectively predicts groundwater WQI with high accuracy and can serve as a robust decision-support tool for sustainable groundwater management in hard-rock aquifer regions, particularly in data-scarce rural areas.
Keywords: Water quality index, artificial neural network, hard-rock aquifer, groundwater sustainability, Levenberg–Marquardt, India