Performance of the best spectroscopic models as measured against the test set.

aSpectral transformation (SG0 = Savitzky-Golay smoothing; SG1 = Savitzky-Golay first derivative; SNV = standard normal variate); bMultivariate Calibration Model (svm = support vector machine regression; cubist = Cubist); cPredictor used in the models (spc = spectral matrix; rfe = spectral matrix with bands selected by recursive feature elimination); dStandard Deviation of the observations (g C kg−1); eRoot Mean Square Error of Prediction (g C kg−1; Eq. 4); fBias (g C kg−1; Eq. 2); gStandard Error of Prediction (g C kg−1; Eq. 3); hRatio of Performance to Deviation (Eq. 5); iNumber of validation samples.