Effective management and protection of the environment necessitate rapid assessment and remediation of hydrocarbon-contaminated soils. In response to this challenge, we have developed a binary classifier system utilising mid-infrared (Mid-IR) spectroscopy for rapid, on-site soil contamination assessment. This innovative approach can be employed in scenarios such as emergency spill response, petrol station decommissioning, and evaluation of potential environmental and health risks linked to hydrocarbon-contaminated sites.
The classifier system is based on Mid-IR spectra collected with our portable spectrometers, RemScan. The technique capitalises on the ability of mid-IR spectroscopy to directly probe the hydrocarbon absorption bands. Over 20,000 meticulously produced calibration samples, covering diverse soil properties and spiked with precise diesel concentrations, are used to train the classifier. The performance of various classifier types, pre-processing techniques, and sampling methodologies is assessed through cross-validation of the training dataset and validation on more than 1,300 field samples with laboratory reference measurements.
Preliminary results indicate classification accuracy above 80%, with Support Vector Machines and decision trees emerging as the most effective algorithms. The optimal data pre-processing technique involves targeting the specific hydrocarbon spectral fingerprint region and applying detrending algorithms to minimise baseline variations. Further advancements are anticipated with the inclusion of additional training data, algorithm refinement, and a substantial increase in field data for performance testing.
This tool can rapidly determine soil contamination above or below 1000 ppm in the field, enabling operators to make informed decisions on whether the soil requires treatment. Our research highlights the potential for rapid, accurate, and cost-effective on-site soil hydrocarbon contamination evaluation. This tool offers an economical and eco-friendlier alternative to traditional laboratory testing with lengthy turn-around times and cumbersome in-field extraction kits that require continuous solvent procurement. Such advancements are crucial for fostering sustainable environmental management and conservation.