In this study, we demonstrate that partial least-squares regression analysis with full cross-validation of spectral reflectance data estimates the amount of polycyclic aromatic hydrocarbons in petroleum-contaminated tropical rainforest soils. We applied the approach to 137 field-moist intact soil samples collected from three oil spill sites in Ogoniland in the Niger Delta province (5.317°N, 6.467°E), Nigeria. We used sequential ultrasonic solvent extraction-gas chromatography as the reference chemical method. We took soil diffuse reflectance spectra with a mobile fibre-optic visible and near-infrared spectrophotometer (350-2500 nm). Independent validation of combined data from studied sites showed reasonable prediction precision (root-mean-square error of prediction = 1.16-1.95 mg kg(-1), ratio of prediction deviation = 1.86-3.12, and validation r(2) = 0.77-0.89). This suggests that the methodology may be useful for rapid assessment of the spatial variability of polycyclic aromatic hydrocarbons in petroleum-contaminated soils in the Niger Delta to inform risk assessment and remediation.
Keywords: Chemometrics; Hydrocarbon; Near-infrared spectroscopy; Soil.
Copyright © 2013 Elsevier Ltd. All rights reserved.