A Comparison of Geographically Broad Models for Prediction of Soil Carbonate, Total Carbon, and Gypsum Using VNIR.

VNIR predictive models for soil carbonate, total carbon, and gypsum were developed from several hundred crushed, sieved, <2mm soil samples from the USDA/NRCS Soil Survey Laboratory archive. The samples covered a wide range of analyte concentrations, and were selected from a wide geographic area, covering a large range of geologic materials and geomorphic settings. The geographically broad models tend to predict more accurately for those analytes with little variation in molecular structure, such as carbonate.