Nitrogen, phosphorus, and potassium are some of the most important biochemical components of plant organic matter, and hence, estimation of their contents can help monitor the metabolism processes and health of plants. This study, conducted in the Yixing region of China, aimed to compare partial least squares regression (PLSR) and support vector machine regression (SVMR) methods for estimating the nitrogen (C _N), phosphorus (C _P), and potassium (C _K) contents present in leaves of diverse plants using laboratory-based visible and near-infrared (Vis-NIR) reflectance spectroscopy. A total of 95 leaf samples taken from rice, corn, sesame, soybean, tea, grass, shrub, and arbour plants were collected, and their C _N, C _P, C _K, and Vis-NIR reflectance data were measured in a laboratory. The PLSR and SVMR methods were calibrated to estimate the C _N, C _P, and C _K contents of the obtained samples from spectral reflectance. Cross-validation with an independent data set was employed to assess the performance of the calibrated models. The calibration results indicated that the PLSR method accounted for 59.1%, 50.9%, and 50.6% of the variation of C _N, C _P, and C _K, whereas the SVMR method accounted for more than 90% of the variation of C _N, C _P, and C _K. According to cross-validation, the SVMR method achieved better estimation accuracies, which had determination coefficients of 0.706, 0.722, and 0.704 for C _N, C _P, and C _K, respectively, than the PLSR method, which had determination coefficients of 0.663, 0.643, and 0.541. It was concluded that the SVMR method combined with laboratory-based Vis-NIR reflectance data has the potential to estimate the contents of biochemical components.

In constructing models for predicting soil organic matter (SOM) by using visible and near-infrared (vis–NIR) spectroscopy, the selection of representative calibration samples is decisive. Few researchers have studied the inclusion of spectral pretreatments in the sample selection strategy. We collected 108 soil samples and applied six commonly used spectral pretreatments to preprocess soil spectra, namely, Savitzky–Golay (SG) smoothing, first derivative (FD), logarithmic function log(1/R), mean centering (MC), standard normal variate (SNV), and multiplicative scatter correction (MSC). Then, the Kennard–Stone (KS) strategy was used to select calibration samples based on the pretreated spectra, and the size of the calibration set varied from 10 samples to 86 samples (80% of the total samples). These calibration sets were employed to construct partial least squares regression models (PLSR) to predict SOM, and the built models were validated by a set of 21 samples (20% of the total samples). The results showed that 64−78% of the calibration sets selected by the inclusion of pretreatment demonstrated significantly better performance of SOM estimation. The average improved residual predictive deviations (ΔRPD) were 0.06, 0.13, 0.19, and 0.13 for FD, log(1/R), MSC, and SNV, respectively. Thus, we concluded that spectral pretreatment improves the sample selection strategy, and the degree of its influence varies with the size of the calibration set and the type of pretreatment.