Similarity Between Spectra and the Detection of Outliers

In digital soil spectroscopy, similarity or distance metrics between soil spectra are necessary for a large number of applications, such as for assessing the reliability of a spectrometer over repeated scans, to search for a similar soil sample based on spectra from a large database, to classify spectra into groups of similar characteristics or more generally for detecting outlier spectra. To assess similarity, we usually compare two spectra wavelength by wavelength and compute the distance between them. The distance between the two spectra is averaged into a single similarity metric.