Prediction of chemical water quality used for drinking purposes based on artificial neural networks

The Groundwater resources generally have a good water quality and can be used for drinking purposes than water surfaces. However, the anthropogenic activities and climate change effects have been degrading the groundwater quality particularly in the arid and semi-arid areas. In addition, the monitoring of water quality in these regions is poor, as it is expensive and faces financial constraints, notably in rural areas. For this problem, we need to develop a new alternative that allows us to predict the water quality easily. Therefore, the solutions of this challenge would be to develop accurate and reliable models that would allow the prediction of chemical parameters commonly, used for evaluating the suitability of water for drinking uses. This study aims to develop Artificial Neural Networks (ANN) models for predicting the Total Dissolved Solid (TDS in mg/l), Total Hardness (TH), sulphate (SO42-) mg/l and Chloride (Cl-) mg/L parameters using Electrical Conductivity (EC), pH and Temperature as input variables. These models were developed based on the 42 samples collected and analyzed from Tanobart Groundwater in Morocco. Among the 42 samples, 30 samples were used for training of the models while the remaining data were used for the validation processes. The results showed that the ANN models are highly accurate for predicting the TDS, TH, Sulphate and of Chloride with coefficients of determination 0.962, 0.993, 0.986 and 0.957 for the TH, TDS, Sulphate and Chloride parameters respectively, for training processes. Also, the results during the calibration revealed a good accuracy for predicting theses parameters. Hence, these models can improve the water quality monitoring in rural areas to assess the chemical suitability of water for drinking purpose with low costs and in a short time.