MULTIVARIATE STATISTICAL CLASSIFICATION OF SOIL SPECTRA

The purpose oi this work was to evaluate whether AVIRIS (Advanced Visi ble Infrared Imaging Spectrometer) bands can be used to discriminate b etween soils having similar properties, as well as to compare AVIRIS s pectra with those from laboratory measurements. Multivariate analysis techniques show that two soils belonging to the same series and a thir d soil belonging to a different, but related series can be discriminat ed at a high level of accuracy using reflectance data from AVIRIS or f rom laboratory measurements. It is also shown that wavelengths importa nt in discriminating soils were highly correlated between AVIRIS and l aboratory data. The distribution of variance and weighting functions a lso show consistent patterns between these data sets.