Pattern recognition method from hydrochemical parameters to predict uranium concentrations in groundwater

Abstract

Traditional methods for determining uranium in groundwater are spectroscopic methods that require time, money, and experienced chemists. A prediction method for predicting uranium concentration using a few hydrochemical parameters that are related to uranium was developed within this study. In this study, uranium, hydrochemical parameters, and trace elements of groundwater around Ulaanbaatar in Mongolia were measured. Inductively coupled plasma mass spectrometry was used to determine the uranium concentration in 135 samples. The relationship between uranium concentration and hydrochemical parameters was studied to determine whether the concentration in groundwater could be predicted using chemometric methods based on some hydrochemical parameters. Chemometric methods were performed using a Python programming language. A pattern recognition method for classifying groundwater samples by specific threshold uranium concentration uses a principal component analysis (PCA) and support vector machine (SVM). The average accuracy of the classification models was 88.21%. PCA is used to visualize the classification by uranium concentration and show which hydrochemical parameters are crucial to predicting uranium concentration. In this study, a regression model was developed to predict uranium concentration in groundwater using hydrochemical parameters selected by the PCA-SVM combination method. The regression method uses a combination of polynomial regression and multiple linear regression. This combined regression method has shown good results for predicting uranium concentrations based on selected hydrochemical parameters.