Analysis of visible and near infrared spectral reflectance for assessing metals in soil.

Abstract

Visible and near infrared reflectance (VNIR; 350-2500nm) spectroscopy has greatly been used in soils, especially for studying variability in spectrally active soil components (e.g., organic carbon, clays, and Fe/Al oxides) based on their diagnostic spectral features. In recent years, this technique has also been applied to assess soil metallic ions. In this research, the feasibility of VNIR spectroscopy for determination of soil metals was investigated with two soil data sets: (i) artificially metal-spiked and (ii) in situ metal-contaminated soils. Results showed that reflectance spectra of neither metal-spiked soils with Cd, As, and Pb even at their higher concentrations of 20, 900, and 1200mgkg(-1), respectively, nor in situ metal-contaminated soils (with concentrations of 30mg Cd, 3019mg As, and 5725mgPbkg(-1) soil) showed any recognized absorption peaks that correspond to soil metal concentrations. We observed variations in reflectance intensity for in situ metal-contaminated soils only, showing higher reflectance across the entire spectrum for strongly and lower for less metal-contaminated soils. A significant correlation was found between surface soil metals' concentrations and continuum removed spectra, while soil metals were also found significantly associated with soil organic matter and total Fe. A partial least square regression with cross-validation approach produced an acceptable prediction of metals (R (2)=0.58-0.94) for both soil data sets, metal-spiked and in situ metal-contaminated soils. However, high values of root mean square error ruled out practical application of the achieved prediction models.