Prediction of soil organic carbon stock using visible and near infrared reflectance spectroscopy (VNIRS) in the field

Considering the determination coefficient for validation (R2val), standard error of prediction corrected for bias (SEPc) and ratio of standard deviation of the validation subset to SEPc (denoted RPDval), the VNIRS prediction of SOC concentration was accurate (R2val = 0.75; SEPc = 1.6 gC kg− 1 soil < 2 mm, i.e. 13% of the mean; RPDval = 2.0). The prediction of SSOC was acceptable (R2val = 0.70; SEPc = 2.0 gC dm− 3, representing 13% of the mean; RPDval = 1.8). Part of the VNIRS prediction error could be attributed to the fact that conventional and spectral measurements were not made on the same samples. Scanning the cylinder cores would probably result in better VNIRS predictions but is not consistent with the objective of determining SSOC without tedious cylinder sampling (except for calibration). However, conventional determination of SOC concentration on auger cores (rather than on cylinder cores) could be a valuable solution for improving calibration, with the perspective to develop a rapid procedure for accurate VNIRS prediction of SOC stock in the field.