MODELING REE+ CONTENT IN A COAL ASH IMPOUNDMENT USING 3D EMPIRICAL BAYESIAN KRIGING

Identifying domestic resources for rare earth elements (REEs) is vital for the national security and economy of the U.S. Coal contains significant amounts of REEs+Sc+Y (REE+) which become concentrated during the coal combustion process (U.S. DOE, 2022). This study determined an average (N=397) REE+ content of 434 ± 153.6 ppm in coal ash impoundments in the Southeast U.S. This is significantly above the DOE cutoff grade of 300 ppm (Bagdonas et al., 2022).

To better understand the feasibility of using coal ash as a REE+ feedstock, this study assesses the use of 3D Empirical Bayesian Kriging (3D EBK), a modeling tool in ArcGIS Pro, to determine spatial variability of REE content within a coal ash impoundment. 3D EBK is an effective tool for interpolating spatial variations in continuous geologic features that are not stratified (Krivoruchko, 2012). It is also superior to other forms of kriging at modeling complex geologic features with limited data points because it uses many simulations, overlapping subsets, and model weights to increase the accuracy of predictions (Arc GIS Pro, 2022a).

This presentation will review a 3D EBK model developed for a coal ash impoundment in the southeastern U.S. Coal combusted at this site is primarily sourced from the Southern Appalachian Basin, and its resulting coal ash has an average REE+ content of 480.4 ± 83.2 ppm. Data inputs for the model include historical data (N=16) and samples collected and analyzed as a part of this study (N=46), along with depth and location data. Major element data also is available for some samples (N=20). Oxides highly correlated with REEs can be used to better constrain the model in areas where REE data are absent. The validity of the 3D-EBK model will be assessed using the minimum standards for geologic realizations and by cross-validation using the “leave one out” method (Leuangthong et al., 2004). A successful model would indicate that limited sampling can be used to model an impoundment accurately and allow for targeted extraction of predicted high REE+ concentration areas.