CHARACTERIZATION AND CLASSIFICATION OF SOIL GEOCHEMISTRY WITHIN THE CONTERMINOUS UNITED STATES

A multi-element soil geochemical survey was conducted over the conterminous United States for which samples were collected from (1) a depth of 0-5 cm, (2) the soil A horizon, and (3) the soil C horizon. The survey sampled 4,857 sites representing a density of approximately 1 site per 1600 km². Each sample was analyzed for Ag, Al, As, Ba, Be, Bi, Total Carbon, Ca, Cd, Ce, Co, Cr, Cs, Cu, Fe, Ga, Hg, In, K, La, Li, Mg, Mn, Mo, Na, Nb, Ni, P, Pb, Rb, S, Sb, Sc, Se, Sn, Sr, Te, Th, Ti, Tl, U, V, W, Y, and Zn by methods that yielded the total, or near-total, elemental content. Silver and Te were dropped from further evaluation because most values were reported at less than the detection limits. Censored analytical values were replaced using the R package “robCompositions.” A log-centered transform was applied to the data followed by the application of a principal component analysis that yielded 6 significant components. Principal component analysis reveals continental-scale contrasts in soil composition that reflect distinctive geochemical features including soil horizon, bedrock source variability, and weathering. The first principal component identifies the contrast between mafic lithologies (relative Ni, Co, Cr enrichment) and felsic intrusive lithologies (relative REE enrichment). The second principal component reveals relative enrichment of Hg, Ti, Mo, Cd, As that is associated with the weathered sedimentary rocks of the southern Appalachian mountains in North Carolina, South Carolina, Georgia and northern Florida. The application of discriminant analysis provides a probability-based approach in classifying the soil geochemistry in terms of bioclimatic indices that reflect zones of humidity and temperature variability (ombrotypes, thermotypes), terrestrial ecosystems, and surface lithology. Classification accuracy is variable between the different classes of bioclimatic indices and surface lithologies; however, maps displaying the probability of occurrence are consistent with the observed data.