Rapid assessment of petroleum-contaminated soils with infrared spectroscopy

Abstract

Soil sensing using infrared spectroscopy has been proposed as an alternative to conventional soil analysis to detect soil contamination. This study evaluated the use of field portable and laboratory benchtop infrared spectrometers in both the near infrared (NIR) and mid infrared (MIR) region for rapid, non-destructive assessment of petroleum contaminated soils. A laboratory study of soils spiked with petroleum products showed that several factors can affect the infrared absorbance. These include soil texture, organic matter content, and the types and concentrations of contaminants. Despite these factors, infrared regions that are affected by hydrocarbon contamination can be readily found in 2990–2810cm−1 in the MIR range, and 2300–2340nm in the NIR range. Using continuum-removed spectra, the effects of soil and contaminant factors on the absorbance peaks were isolated. This study also created statistical models to predict total recoverable petroleum hydrocarbons concentration in soils by utilizing the absorption features found in the mid-infrared region spectra. Subsequently, three different approaches were tested for the prediction of Total Recoverable Hydrocarbon (TRH) concentration on 72 field contaminated samples: (i) linear regression using only 1 infrared region, (ii) multiple linear regression (MLR) using 4 regions in the MIR, and (iii) partial least square regression (PLSR) which use the whole spectra. The model created using MLR approach for portable MIR spectrometer outperformed the benchtop MIR spectrometer with a coefficient of determination (R2) of 0.71 and 0.53 respectively. While PLSR model for portable spectrometer show a better prediction for TRH prediction (R2=0.75), the MLR can also achieve a similar performance (R2=0.71) by using only 4 regions in the MIR spectra as predictors.