Building geochemically based quantitative analogies from soil classification systems using different compositional datasets

Soil heterogeneity is a major contributor to the uncertainty in near-surface biogeochemical modeling. We sought to overcome this limitation by exploring the development of a new classification analogy concept for transcribing the largely qualitative criteria in the pedomorphologically based, soil taxonomic classification systems to quantitative physicochemical descriptions. We collected soil horizons classified under the Alfisols taxonomic Order in the U.S. National Resource Conservation Service (NRCS) soil classification system and quantified their properties via physical and chemical characterizations. Using multivariate statistical modeling modified for compositional data analysis (CoDA), we developed quantitative analogies by partitioning the characterization data up into three different compositions: Water-extracted (WE), Mehlich-III extracted (ME), and particle-size distribution (PSD) compositions. Afterwards, statistical tests were performed to determine the level of discrimination at different taxonomic and location-specific designations. The analogies showed different abilities to discriminate among the samples. Overall, analogies made up from the WE composition more accurately classified the samples than the other compositions, particularly at the Great Group and thermal regime designations. This work points to the potential to quantitatively discriminate taxonomically different soil types characterized by varying compositional datasets.