van Ruitenbeek, Frank, van der Werff, Harald, Bakker, Wim, and van der Meer, Freek, Department of Earth Systems Analysis, International Institute of Geo-information Science and Earth Observation (ITC), PO Box, 7500 AA, Enschede, The Netherlands
Interpreting the geology from hyperspectral remote sensing imagery can be complicated because of the non-unique nature of certain minerals for specific geologic settings and the non-uniqueness of spectral signatures for certain minerals. Also, the detection of mineral end members itself is not a straightforward procedure, especially in areas with limited or no field information. In this presentation, two approaches are shown that address the non-uniqueness problems and and which reduce the uncertainty of interpretations of spectral imagery. In the first approach, the spatial context is included in the classification of hyperspectral imagery by matching a miniature image of a mineral transition to a hyperspectral image. Examples of mapping hydrothermal alteration zoning from the Pilbara VMS district of Western Australia and epithermal gold deposits in southeast Spain illustrating this approach will be shown. In the second approach, the spectral remote sensing data is integrated with airborne geophysical data and field data to resolve ambiguities in the single sensor data sets and enhance geologic interpretations. Examples using this method are the Pilbara granite-greenstone terrane in , a method for image-end member detection from hyperspectral imagery in areas without field information will be introduced. This method involves spectral feature analysis and correlation analysis for end member-selection in the Nilli Fossae area on Mars.