Soil pH value, organic matter and macronutrients contents prediction using optical diffuse reflectance spectroscopy

Accurate information about the variability of soil attributes and characteristics is essential for the site-specific management of agricultural inputs, also known as precision agriculture; however, the inability to obtain soil information rapidly, inexpensively and reliably remains one of the biggest challenges. Recently, visible and near infrared (VIS/NIR) diffuse reflectance spectroscopy has emerged as a rapid and low-cost tool for extensive investigation of soil characteristics, such as macronutrients contents, pH value, and organic matter content. In the present work, the potential of VIS/NIR diffuse reflectance spectroscopy to predict the contents of N, P, K and OM and the value of pH in soils was analyzed using two spectrometers: Veris VIS/NIR soil sensor from Veris Technology Inc. and MPA FT-NIR spectrometer from Bruker Optics Inc. Subsequently, different pretreatment methods were adopted to improve the correlation between soil properties and the spectra, and then principal component regression was used, with the optimum numbers of PCs were selected on the basis of PRESS value in the leave-one-out validation. The primary conclusions in our study include: (i) optical reflectance spectroscopy in visible and near-infrared regions combined with appropriate pretreatment was an ideal tool for the estimation of soil pH value and OM content, while presented poor potentials in the prediction of total N, total P and total K; (ii) the models established with spectra after the preprocessing methods include MSC and S-G filter for smooth and first-order derivative together presented preferable results than those after MSC or S-G filter for smooth and first-order derivative individually; and (iii) the prediction results of the two spectrometers with different light-splitting techniques produced similar variation tendencies among the measured soil properties. Consequently, a scanning grating spectrometer in the NIR region proves to be an effective tool to measure certain soil properties, namely OM content, pH value and total N. Moreover, compared with a FR instrument, a scanning grating spectrometer is a preferable choice in the design of an on-the-go soil sensor. (C) 2014 Elsevier B.V. All rights reserved.