Application of ANN and MLR Models on Groundwater Quality Using CWQI at Lawspet, Puducherry in India

Abstract

With respect to groundwater deterioration from human activities a unique situation of co-disposal of non-engineered Municipal Solid Waste (MSW) dumping and Secondary Wastewater (SWW) disposal on land prevails simultaneously within the same campus at Puducherry in India. Broadly the objective of the study is to apply and compare Artificial Neural Network (ANN) and Multi Linear Regression (MLR) models on groundwater quality applying Canadian Water Quality Index (CWQI). Totally, 1065 water samples from 68 bore wells were collected for two years on monthly basis and tested for 17 physio-chemical and bacteriological parameters. However the study was restricted to the pollution aspects of 10 physio-chemical parameters such as EC, TDS, TH, , Cl-, , Na+, Ca2+, Mg2+ and K+. As there is wide spatial variation (2 to 3 km radius) with ground elevation (more than 45 m) among the bore wells it is appropriate to study the groundwater quality using Multivariate Statistical Analysis and ANN. The selected ten parameters were subjected to Hierarchical Cluster Analysis (HCA) and the clustering procedure generated three well defined clusters. Cluster wise important physio-chemical attributes which were altered by MSW and SWW operations, are statistically assessed. The CWQI was evolved with the objective to deliver a mechanism for interpreting the water quality data for all three clusters. The ANOVA test results viz., F-statistic (F = 134.55) and p-value (p = 0.000 < 0.05) showed that there are significant changes in the average values of CWQI among the three clusters, thereby confirming the formation of clusters due to anthropogenic activities. The CWQI simulation was performed using MLR and ANN models for all three clusters. Totally, 1 MLR and 9 ANN models were considered for simulation. Further the performances of ten models were compared using R2, RMSE and MAE (quantitative indicators). The analyses of the results revealed that both MLR and ANN models were fairly good in predicting the CWQI in Clusters 1 and 2 with high R2, low RMSE and MAE values but in Cluster 3 only ANN model fared well. Thus this study will be very useful to decision makers in solving water quality problems.