Prediction of soil organic carbon content using field and laboratory measurements of soil color

The understanding, prediction, and modeling efficacy of soil organic carbon (SOC) distribution across fields and larger regions requires a large number of samples that are costly to analyze. The objective of this study was to evaluate soil color measurements to predict SOC for agriculture and prairie land uses. Munsell soil color book (B) and chroma meter (C) color readings were taken at the midpoint depth of each horizon (HB and HC) and predetermined depth increments (IB and IC) on 125 cores. Horizon matrix (HD) colors were determined by standard description. A chroma meter was used to determine the color of prepared samples, ground to < 2 mm. Both color data sets were used in a regression analysis to predict SOC content by weight and volume. The best predictors for each technique are the models that incorporate Munsell value and chroma along with the depth from which the measurement was taken. Separating samples by land use improved the prediction of SOC. Transforming SOC content by log(10) improved the coefficient of determination for nearly all models. The best predictors of SOC were HD for SOC by weight (agricultural field r(2) = 0.79, prairie r(2) = 0.53), HB for SOC by volume (agricultural field r(2) = 0.76, prairie r(2) = 0.57), and IC and IB for log-transformed SOC by weight and volume (agricultural field r(2) = 0.84, prairie r(2) = 0.62). This study indicates that while SOC content predictions can be made with field measurements, there are limitations to their predictive usefulness.