Rapid detection of alkanes and polycyclic aromatic hydrocarbons in oil-contaminated soil with visible near-infrared spectroscopy

Summary Recent developments and applications of rapid measurement tools (RMTs) such as visible near‐infrared (vis–NIR) spectroscopy confirmed that these technologies can provide ‘fit for purpose’ and cost‐effective data for risk assessment and management of oil‐contaminated sites. Although vis–NIR spectroscopy has been used frequently to predict total petroleum hydrocarbons (TPHs), it has had limited use for polycyclic aromatic hydrocarbons (PAHs) and there has been none for alkanes. In the present study, the potential of vis–NIR spectroscopy (350–2500 nm) to measure PAHs and alkanes in 85 fresh (wet, unprocessed) oil‐contaminated soil samples collected from three sites in the Niger Delta, Nigeria, was evaluated. The vis–NIR signal and alkanes and PAHs measured in the laboratory by sequential ultrasonic solvent extraction followed by gas chromatography‐mass spectrometry (GC‐MS) were then used to develop calibration models using partial least squares regression (PLSR) and random forest (RF) modelling tools. Prior to model development, the pre‐processed spectra were divided into calibration (75%) and prediction (25%) sets. Results showed that the prediction performance of RF calibration models for both alkanes (a coefficient of determination ( R 2 ) of 0.58, a root mean square error of prediction (RMSEP) of 53.95 mg kg −1 and a residual prediction deviation (RPD) of 1.59) and PAHs ( R 2 = 0.71, RMSEP = 0.99 mg kg −1 and RPD = 1.99) outperformed PLSR ( R 2 = 0.36, RMSEP = 66.66 mg kg −1 and RPD = 1.29, and R 2 = 0.56, RMSEP = 1.21 mg kg −1 and RPD = 1.55, respectively). The RF modelling approach accounted for nonlinearity of the soil spectral responses and therefore resulted in considerably greater prediction accuracy than the linear PLSR. Adoption of vis–NIR spectroscopy coupled with RF is recommended for rapid and cost‐effective assessment of PAHs and alkanes in contaminated soil. Highlights We evaluated the potential of vis–NIR to estimate alkanes and PAHs in oil‐contaminated soil. The prediction performance of RF models was better than PLSR models for both alkanes and PAHs. The spectral response to alkanes and PAHs in soil embodies considerable non‐linearity. Results suggest that RF‐vis–NIR is a promising tool for rapid in situ assessment of soil alkanes and PAHs.