2015 GSA Annual Meeting in Baltimore, Maryland, USA (1-4 November 2015)

Paper No. 25-1
Presentation Time: 9:00 AM-5:30 PM

A COMPOSITIONAL DATA ANALYSIS APPROACH TO UNDERSTANDING TRACE ELEMENT DISTRIBUTION DURING BENEFICIATION OF AN ILLINOIS BASIN COAL


GEBOY, Nicholas J., Eastern Energy Resources Science Center, U.S. Geological Survey, 12201 Sunrise Valley Dr, MS 956, Reston, VA 20192, KOLKER, Allan, Eastern Energy Resources Science Center, U.S. Geological Survey, 956 National Center, Reston, VA 20192, LEFTICARIU, Liliana, Department of Geology, Southern Illinois University, Carbondale, IL 62901 and VANCIL, Gary, American Coal Company, Harrisburg, IL 62946, ngeboy@usgs.gov

This study applies compositional data analysis techniques to geochemical data for size-segregated coal and coal refuse from a coal preparation facility at an underground mine in Saline County, IL. Herrin #6 coal produced at a depth of 600 ft (183 m) fed the cleaning circuit sampled. The preparation process included separation by water, froth flotation and a series of screens, heavy media cyclones and vessels. Sampling points allowed collection of cleaned and refuse coal, each in coarse, medium, and fine size fractions. In addition, feed coal for the circuit and a final cleaned product were sampled, for a total of 8 corresponding samples. Samples were analyzed for short proximate, forms of sulfur (S) and mercury (Hg) contents. The cleaning plant feed coal contained 4.5 % pyritic-S and 236 ppb Hg. In the sized cleaned fractions the pyritic-S values were 1.3, 0.7 and 0.8 %, in descending order of size, whereas the respective refuse coal fractions had pyritic-S concentrations of 5.6, 4.4 and 3.0 %. Mercury values were similarly reduced via cleaning to 102, 79 and 71 ppb, from coarse to fine size fraction, relative to the respective refuse equivalent values of 297, 169 and 125 ppb. The final cleaned product, a combination of the three cleaned size fractions, had a pyritic-S content of 0.9 % and a Hg concentration of 86 ppb. Regression of all pyritic-S vs Hg data yields a correlation with an r2 = 0.89, supporting the understood association of Hg in pyrite in the Herrin coal. Given the compositional nature of the data, however, linear regression is an inappropriate tool. To better examine the relationship between metals and host-phases, a compositional data analysis approach, which considers the closed nature of geochemical data, was applied. A centered logratio form biplot shows obvious clustering of the cleaned vs non-cleaned coal samples in a subcompostion comprising pyritic-S, organic-S, Hg and carbon (a proxy for organic material). Following sequential binary partitioning of the data to compare the relationship of Hg to total S vs forms of S, an isometric logratio (ilr) plot also demonstrates an association of Hg to pyritic-S in the Illinois Basin coal. A linear relationship amongst the refuse samples in the ilr plot, suggests a systematic control between the ratios of Hg:Total S and pyritic-S:organic-S.