A Soil Quality Index Using Vis-NIR and pXRF Spectra of a Soil Profile

Abstract

A soil quality index (SQI) can be used to manage the soil for potential uses and remediate limitations of soil for sustainable agriculture and ecology. The objectives of this study were to (i) characterize vertical and horizontal variations of some soil properties and SQI in a Fluventic Haploxerepts profile in Eskis,ehir, Turkey, (ii) compare interpolation methods for mapping the distribution of soil properties and SQI in the profile wall, and (iii) assess cubist model for predicting soil properties and SQI using individual and combined Vis-NIR and pXRF spectra in the profile. The soil profile wall (1 m deep x 1 m wide) was divided into a raster of 10 x 10 cm intervals for collecting a total of 100 soil samples. The soil samples were analyzed for soil physical, chemical properties, and elemental concentrations and scanned using Vis-NIR (350-2500 nm) and pXRF (0-8 and 0-45 kV) spectrometers. Standard scoring function was used for standardization of indicators. Weighted linear combination method was used to calculate SQI and F-AHP for assigning weights of indicators. The 100 samples were split into calibration (70%) and validation (30%) datasets. Individual or combined Vis-NIR and pXRF spectra were used in cubist model to evaluate the feasibility for predicting soil properties and SQI in the Inceptisol. Results suggested that the soil profile had a considerable variation of soil properties and SQI with depth and within horizons. The A horizon had the highest SQI. In addition, Vis-NIR had better prediction performances for pH, SOM, Fe, and Mn (R-2 = 0.42-0.88), whereas pXRF had better prediction performances for EC, CaCO3, P, and Zn (R-2 = 0.18-0.90). In addition, combining Vis-NIR, pXRF8, and pXRF45 had better prediction performance for sand, silt, clay, K, Cu, and SQI (R-2 = 0.81-0.97). We concluded that Vis-NIR and pXRF spectra can be successfully used to predict SQI in Fluventic Haploxerepts. Combined Vis-NIR and pXRF spectra had better prediction performance for SQI.