DETECTION AND CLASSIFICATION OF ORGANIC AND ORGANOPHOSPHORUS ANALYTES ON SOIL FROM REFLECTION-ABSORPTION SPECTROSCOPY

Abstract

Principal Components Analysis (PCA) is a common anomaly detection tool that was used in this work to detect organic and organophosphate analytes on soils using mid-infrared reflection-absorption spectroscopy. Detection is hindered by large variability in sample-to-sample soil reflectivity that is due to the random nature of the soil particle packing. Extended multiplicative scatter correction (EMSC) and Savitzky-Golay derivative preprocessing were examined as methods to reduce this variability and enhance detection capability. Second derivative preprocessing provided results that were at least as good as EMSC for detection and the simplicity of the derivative methodology makes it an attractive preprocessing approach. Typically, PCA is applied to all spectral channels and results from detection events are interrogated to identify a potential cause. In this work, PCA models were developed for specific wavenumber ranges corresponding to functional group frequencies with the objective of providing some classification capability. It was found that detection of CH 2, CH 3 and P = O stretching bands was possible; however, results for a – CH 2 scissors band was less encouraging and detection of O – H stretch, – C – C – skeletal stretch, and PO – C stretch modes was poor. Some limited classification capability may be possible, but it would be difficult to make a unique assignment of the analytes present using the strategies studied.