Chrysotile detection in soils with proximal hyperspectral sensing and chemometrics

In this work the authors present an innovative methodology, based on proximal hyperspectral sensing and chemometric techniques, aimed at detecting asbestos containing soils. Short Wave InfraRed (SWIR) reflectance spectra of reference samples containing known chrysotile fractions were collected in laboratory. Since the identification of asbestos containing soils depends on the contaminant mass percentage (weight/weight), two supervised multivariate data projection methods were evaluated for asbestos concentration prediction. The first results are reported here, together with advantages and limits of the analytical methods. Orthogonal Partial Least Squares (PLS) regression showed the lowest error in prediction and the highest coefficient of determination in prediction. This technique would support screening activities frequently conducted during environmental assessment and remediation projects.