Prediction of Rare Earth Elements in Neutral Alkaline Mine Drainage from Razi Coal Mine, Golestan Province, Northeast Iran, Using General Regression Neural Network

The coal mining operation in the Razi mine of the Golestan Province, northeast Iran, has slightly disturbed the balance of the existing geowater system, which may have a long-term environmental impact if no attempt is made at considering such environmental problems. Prediction of rare earth elements (REEs) in mine drainages is a major task in developing an appropriate remediation strategy because high concentrations of REEs may be harmful to human beings and animals. Expert systems are broadly used in many applications. In this paper, a general regression neural network (GRNN) method has been chosen as an easy-to-use tool in order to predict REEs in neutral alkaline mine drainage (NAMD) of the Razi coal mine that contains a low concentration of REEs (varying between 2.0 and 3.09 μg/L) and high levels of sulphate (2,100 mg/L), bicarbonate (1,575 mg/L), and pH (about 9). The pH and SO42− concentration of NAMD were the main constituents used for REE prediction. To evaluate the accuracy of the GRNN method, the hydrogeochemical data measured in the study area and those data sets available in the literature were considered, and the results were compared with the ones obtained by using the multiple linear regression (MLR) technique. The results show that the GRNN method gives better and more realistic results than the MLR method. The correlation coefficients between the predicted and measured REEs by the GRNN method for train and test data sets were 0.99 and 0.99; describing the high capability of the GRNN method in REE prediction. However, the MLR method results in physically meaningless negative predictions and low correlation coefficient.