Application of FTIR spectroscopy to the prediction of organic C-related soil properties from diverse Canadian agricultural regions

Infrared spectroscopy has the potential to rapidly analyse water dissolved carbon and amino sugars in soil. In this study, mid-infrared (MIR) and near-infrared (NIR) spectra collected from soil water extracts or from bulk soils, along with partial least squares regression (PLSr), were used to estimate the concentrations of water dissolved carbon and amino sugars in diverse agricultural soils collected from 5 field sites in two Western and two Eastern Canadian provinces. The MIR-PLSr models developed from soil water extract spectra estimated hot-water (100 °C) dissolved carbon (HWDC) [Coefficient of Determination (R2)=0.97-0.70, Ratio of Prediction to Deviation (RPD)=6.13-1.83] well but the MIR-PLSr models did not estimate cold-water (21 °C) dissolved carbon (CWDC) very well (R2=0.82-0.50, RPD=2.35-1.42). The models improved estimates of HWDC at the multi-site scale (pooling all samples from five sites) and for the Western Canada sites (R2=0.97-0.93, RPD=6.13-3.68) as compared to the Eastern Canadian sites (R2=0.80-0.70, RPD=2.28-1.83). The MIR- and NIR-PLSr models derived from bulk soil spectra both estimated HWDC very well at the multi-site scale (R2=0.91-0.88, RPD=2.90-2.32) and for the Western Canada sites (R2=0.90-0.87, RPD=3.18-2.89). Models developed from hot-water extract spectra and soil spectra resulted in poor estimates of soil amino sugars (R2=0.74-0.21, RPD=1.89-1.12), except for the satisfactory estimation of muramic acid by the models based on soil spectra at the Western and the multi-site scale (R2=0.80-0.82, RPD=2.21-2.33). We concluded that MIR/NIR models at regional and multi-site scales can be used as a tool to monitor HWDC but that additional research is required for estimating soil amino sugars.