Predicting Soil Carbon Variability with a Spectral Sensor in the Laboratory.

Carbon varies spatially across agricultural in the x, y, and z dimensions. The objective of this study was to assess the potential for carbon to be predicted spatially using visible and near-infrared spectral absorbance measurements made with the VERIS NIRS sensor. Laboratory measures of spectral absorbance (350 to 2220 nm) were made for surface soil samples (n=267) collected from an agricultural field in the Western Coal Fields (WCF) physiographic region of Kentucky and from the North American Proficiency Testing Program (NAPT) dataset (n=149) at 15% moisture, 30% moisture, and 45% gravimetric moisture. Total carbon (TC) was measured for the WCF dataset and median Walkley-Black Organic Matter values were obtained for the NAPT dataset. For both datasets, 80% were randomly selected for model development and 20% were used for validation. The validation root mean squared (RMSE) for the NAPT dataset was 0.75% soil organic matter (R2 = 0.60) when all three moistures were combined. The relationship improved for individual moisture contents (0.60% WB SOM; R2=0.79). For the WCF dataset, relationships were generally poor but there was less than a 1% range in total carbon. The NIRS sensor has great potential as a tool to measure soil carbon.