Prediction and variability mapping of some physicochemical characteristics of calcareous topsoil in an arid region using Vis–SWNIR and NIR spectroscopy

Abstract

AbstractSite-specific management of soils needs continuous measurements of soil physicochemical characteristics. In this study, Vis–NIR spectroscopy with two spectroscopic instruments, including charge-coupled device (CCD) and indium-gallium-arsenide (InGaAs) spectrometers, was adopted to estimate some physicochemical characteristics of a calcareous topsoil in an arid climate. Partial least squares (PLS) as linear and artificial neural networks (ANN) as nonlinear multivariate techniques were utilized to enhance the accuracy of prediction. The best predictive models were then used to extract the variability maps of physicochemical characteristics. Diffuse reflectance spectra of 151 samples, collected from the calcareous topsoil, were acquired in the visible and short-wavelength near-infrared (Vis–SWNIR) (400–1100 nm) and near-infrared (NIR) (950–1650 nm) spectral ranges using CCD and InGaAs spectrometers, respectively. The results showed that NIR spectral data of the InGaAs spectrometer was necessary to reach the best predictions for all selected soil properties. The best predictive models based on the optimum spectral range could allow us the excellent predictions of sand (RPD = 2.63) and silt (RPD = 2.52), and very good estimations of clay (RPD = 2.35) and electrical conductivity (EC) (RPD = 2.224) by ANN and very good prediction of calcium carbonate equivalent (CCE) (RPD = 2.01) by PLS. The CCD device, however, resulted in acceptable predictions of sand (RPD = 2.13, very good) and clay (RPD = 1.66, fair) by ANN, and silt (RPD = 1.78, good), EC (RPD = 1.84, good) and CCE (RPD = 1.67, fair) by PLS. Similar variability was attained between pairs of predicted maps by best models and reference-measured maps for all studied soil properties. For clay, sand, silt, and CCE, the Vis/SWNIR-predicted and equivalent reference-measured maps had acceptable similarities, indicating the potential application of low-cost CCD spectrometers for prediction and the variability mapping of these parameters.