干旱区生态环境知识资源中心

Soil salinization is the primary obstacle to sustainable agricultural development in arid regions. Because total dissolved salts and soluble ion content are the primary indicators of the degree of soil salinization, their accurate estimation is essential to the determination of appropriate soil salinization remediation techniques, irrigation regimes, and the agricultural distribution layout. A total of 261 soil samples were collected from agricultural fields in the province of Xinjiang, China. A portable Fourier transform (FT) mid-infrared (MIR) spectrometer (4000-600 cm(-1)) and a visible near-infrared (VNIR) field spectrometer (350-2500 nm) were used to obtain soil spectra. We subsequently used partial least-square regression (PLSR) and support vector machine (SVM) algorithms to establish models in VNIR, MIR, and VNIR-MIR regions. The main objectives of this study are (i) to investigate the possibility of using spectroscopic techniques to predict total dissolved salts and soluble ion content; (ii) to compare the prediction accuracy of these soil properties in the VNIR, MIR, and VNIR MIR spectral regions; (3) to compare the prediction accuracy with linear and nonlinear algorithms. Our findings demonstrated that spectroscopic techniques are a promising way to predict total dissolved salts and soluble ion content. Good predictions were obtained for total dissolved salts content, HCO3-, SO42- and Ca2+, satisfactory for Mg2+, Cl-, and Na+, but poor for K. This work demonstrates the potential of portable VNIR and MIR spectrometers as proximal soil sensors for more efficient soil analysis and acquisition of soil salinity information. (C) 2016 IAgrE. Published by Elsevier Ltd. All rights reserved.