Estimation of soil salinity using three quantitative methods based on visible and near-infrared reflectance spectroscopy: a case study from Egypt
Soil salinization is a progressive soil degradation process that reduces soil quality and decreases crop yields and agricultural production. This study investigated a method that provides improved estimations of soil salinity by using visible and near-infrared reflectance spectroscopy as a fast and inexpensive approach to the characterisation of soil salinity. Soil samples were collected from the El-Tina Plain on the northwestern Sinai Peninsula in Egypt and measured for electrical conductivity (ECe) using a saturated soil-paste extract. Subsequently, the samples were scanned with an Analytical Spectral Devices spectrometer (350–2,500 nm). Three spectral formats were used in the calibration models derived from the spectra and ECe: (1) raw spectra (R), (2) first-derivative spectra smoothened using the Savitzky–Golay technique (FD-SG) and (3) continuum-removed reflectance (CR). The spectral indices (difference index (DI), normalised difference index (NDI) and ratio index (RI)) of all of the band–pair combinations of the three types of spectra were applied in linear regression analyses with the ECe. A ratio index that was constructed from the first-derivative spectra at 1,483 and 1,918 nm with an SG filter produced the best predictions of the ECe for all of the band–pair indices (R2 = 0.65). Partial least-squares regression models using the CR of the 400–2,500 nm spectral region resulted in R2 = 0.77. The multivariate adaptive regression splines calibration model with CR spectra resulted in an improved performance (R2 = 0.81) for estimating the ECe. The results obtained in this study have potential value in the field of soil spectroscopy because they can be applied directly to the mapping of soil salinity using remote sensing imagery in arid regions.