Development of enhanced groundwater arsenic prediction model using machine learning approaches in Southeast Asian countries

Abstract

Groundwater contamination with arsenic (As) is one of the major issues in the world, especially for Southeast Asian (SEA) countries where groundwater is the major drinking water source, especially in rural areas. Unfortunately, quantification of groundwater As contamination is another burden for those countries because it requires sophisticated equipment, expensive analysis, and well-trained technicians. Here, we collected approximately 350 groundwater samples from three different SEA countries, including Cambodia, Lao PDR, and Thailand, in an attempt to quantify total As concentrations and conventional water quality variables. After that, two machine learning models (i.e. artificial neural network (ANN) and support vector machine (SVM)) were applied to predict groundwater As contamination using conventional water quality parameters. Prior to modeling approaches, the pattern search algorithm in MATLAB software was used to optimize the ANN and SVM model parameters, attempting to find the best parameters set for modeling groundwater As concentrations. Overall, the SVM showed the superior prediction performance, giving higher Nash-Sutcliffe coefficients than ANN in both the training and validation periods. We hope that the model developed by this study could be a suitable quantification tool for groundwater As contamination in SEA countries.