Measurement of soil properties using visible and short wave-near infrared spectroscopy and multivariate calibration

Abstract

Visible and short wave-near infrared spectroscopy (Vis/SW-NIRS) was investigated for measurement of soil properties. Three types of pretreatments including standard normal variate (SNV), multiplicative scattering correction (MSC) and Savitzky–Golay smoothing combining first derivative were adopted to eliminate the system noises and external disturbances. Then, partial least squares (PLS) and least squares-support vector machine (LS-SVM) methods were implemented for calibration models. Simultaneously, the performance of least squares-support vector machine (LS-SVM) method was compared with three kinds of dimension reduction inputs, including principal components (PCs), latent variables (LVs), and effective wavelengths (EWs). The results indicated that all LS-SVM models outperformed PLS models. The EWs-LS-SVM models outperformed all the other models and the determination coefficient (RPre2), and RMSEP were 0.8631 and 3.21 for organic matter (OM), 0.8203 and 17.20 for available N (N), 0.7665 and 5.50 for available P (P), 0.7273 and 15.08 for available K (K), respectively. The results indicated that Vis/SW-NIRS combined with LS-SVM could be utilised as a method for the determination of soil properties. And the EWs-LS-SVM could be very helpful for development of portable instrument of the properties of soil.