APPLICATION OF STATISTICAL TOOLS TO GULF COAST LIGNITE GEOCHEMISTRY

Statistical tools may be powerful aids in using lignite geochemical data to help decipher the geologic history of the lignites, their byproduct potential, technological performance, and potential environmental and human health impacts. Several statistical tools were applied to a data set from the U. S. Geological Survey's COALQUAL database containing about 36 parameters generated on each of 72 samples from the Wilcox and Jackson groups. As most of the data is highly skewed, a log transformation was applied to normalize the data and visualized with a box plot. The plot indicates that the Na variation is highest with some outliers. These outliers may be due to marine influence towards in the southern part of the lignite trend. The correlation matrix shows a high correlation between SiO2 and Ti and K, ash and K, Sr and Ba, Y and Be, Yb and Be, Y and Yb, Sr and Ba, F and Sr, F and Y. The data were then standardized to unit variance and zero mean before using Principal Component Analysis (PCA) and Cluster analysis for integration and to reduce the data dimensionality. The PCA analysis shows that the first four axes account for only 55% of the variance and a minimum of 18 axes have to be considered for more than 95% of the variance. Three distinct groups are visible in a biplot based on the first two principal components. The variables responsible for the groups are: 1)Ash, TiO2, Se, Be, Ba, F, Al2O3, Y, Ce, Zr, V, Li, U, Yb, SiO2, Cr, Sc, Mg, and K2O; 2) Mn, Zn, Cd, Ca, Na, Hg, As, S and P2O5; 3) MgO, SO3, Na2O, Fe2O3, and CaO. With the possible exception of Se, Group 1 appears to reflect the detrital component of the lignites. Group 2 appears to reflect epigenetic mineralization (sulfides: Zn, Cd, Hg, As, S; carbonates: Mn, Ca; silicates: K2O, Na: and phosphates: Ca, P). Group 3 is more problematic perhaps representing oxidation and the formation of sulfates (MgO, SO3, Na2O, Fe2O3, CaO). A possible fourth group may represent low-ash organically associated elements. The biplot method is very effective decomposing complex data sets and providing a visual basis for recognizing parameters that behave geologically or geochemically in similar ways.