Strategies for detecting organic liquids on soils using mid-infrared reflection spectroscopy

Stand-off monitoring for chemical spills can provide timely information for cleanup efforts, and mid-infrared reflection spectroscopy is one approach being investigated for spill detection. Using laboratory data, anomaly and target detection strategies were examined for the detection of four different low-volatility organic liquids on two different soil types. Several preprocessing and signal-weighting strategies were studied. Anomaly detection for C-H bands was good using second-derivative preprocessing and provided similar performance to that of target detection approaches such as generalized least-squares and partial least-squares, with detections at soil loads of approximately 3-6 microg/cm2 a real dosage. Good performance was also found for the detection of P=O, O-H, and C=O stretching vibrational modes, but the optimal strategy varied. The simplicity and generality of anomaly detection is attractive; however, target detection provides more capability for classification.