Determining the clay/organic carbon ratio by visible near infrared spectroscopy

The recently presented Dexter et al. (2008) threshold (ratio of clay to organic carbon (OC) of 10 kg/kg-1) is a good indicator for numerous soil functional properties. For soils with a clay/OC < 10, non-complexed OC is available for soil structure buildup and important biophysical functions. Soils with clay/OC > 10 have critically high clay to OC ratio and reduced functionality which may cause reduced aggregate stability and potentially have a negative effect on yields of agricultural crops.
Conventional analysis of OC and clay are costly and time consuming, thus an alternative method to quantify OC, clay or clay/OC ratio directly would be valuable. Visible near infrared spectroscopy (vis-NIRS) is a cost-effective analytical method which has been applied to soil analysis since the mid- 1990’s. The absorption in the vis region are associated with minerals containing iron (realted to clay due to metal-OH bend) and with organic matter (due to color). The NIR region is dominated by bonds characteristic of both organic and inorganic matter. Thus, there is a rational for using vis-NIRS to OC and clay estimation.
Here, for the first time the feasibility of using vis-NIRS to predict clay/OC ratio was tested. We collected a total of 115 topsoil samples from two agricultural fields in western and north-western Denmark. Soil samples were analysed for OC and particle size fractions. Partial Least Squares regression (full cross-validation) was performed on 80 randomly selected samples to correlate soil spectra with OC, clay and clay/OC. The precision and robustness of calibration models was tested on the remaining 35 samples and evaluated using: root mean square error of prediction (RMSEP); R2; and a ratio of performance to inter-quartile range (RPIQ). Analysis of regression coefficients was also performed.
The soil from the two study sites vary from very sandy (90% sand) to clayey (24% clay) and from lower (1.5% OC) to high in OC content (>6%), giving clay/OC ratio between 1.20 and 10.43. Successful calibration results for OC, clay and clay/OC ratio were generated. The OC content was predicted with a RMSEP of 0.23%, R2 of 0.96 and a high RPIQ (7.5). The important wavebands were related to soil organic matter (780, 1727, 2311 and 2330 nm), Fe oxides (442, 490, 560 nm) and OH bonds (1908 and 2208 nm). The prediction of clay resulted in RMSEP=0.94%, R2=0.94, and RPIQ=6.1. The highest values of the regression coefficient in the clay model were located at 541 and 640 nm (Fe oxides) and at 1412, 1910 and 2208 nm (OH bond). The most successful predictive ability reported was interestingly of the clay/OC ratio (RMSEP=0.42, R2=0.97, RPIQ=12). The most important absorption bands were related to both clay minerals (1421, 1910 and 2206 nm – OH bonds and 429, 720 nm – Fe oxides) and organic carbon (1730, 2160 nm and 2310 nm).
The results of this study show that vis-NIRS can provide very successful and direct determination of Dexter index through its correlation to both OC and clay. Further studies across different soil types will be conducted.