Use of AVIRIS data in the generation of a geological map without field data

The feasibility of using Airborne Visible/Infra-Red Imaging Spectrometer (AVIRIS) hyperspectral data to generate a geological map without the acquisition of field data was examined for data acquired over the Virgin Mountains, Nevada. Geological mapping essentially involves the location of lithological unit boundaries combined with an identification of their mineral assemblage. The spectral resolution of AVIRIS is such that reflected radiation is acquired within the visible and near-infrared (0.4-2.5 micrometer) spectral range in 224 contiguous, 10nm-wide channels. On plotting the response as a function of wavelength an effectively continuous spectral signature is generated, containing absorption troughs. These may relate to atmospheric interaction between the ground and sensor, or directly to surface composition. These atmospheric effects were removed from the data prior to quantitative analysis, so that the absorption features were assumed to relate directly to diagnostic mineral absorption feature wavelengths. Without detailed knowledge on atmospheric conditions or simultaneous ground spectral measurements, removal of these effects was performed using an Empirical Line Correction technique, through ENVI software. This Image Processing package contained a library of spectral signatures from which water was selected for direct correlation with Lake Mead, in the study region. Considering this an appropriate method of atmospheric correction, the data was then analysed using Principal Component Analysis and High Pass Filtration techniques to identify spectral variability. The spatial distribution of similar surface characteristics identified through these techniques generated a boundary map defining lithological units whose internal composition was subsequently examined through Spectral Analyst and Linear Spectral Unmixing techniques, effectively spectral shape matching algorithms. Signatures extracted from the units were compared to reference ENVI library signatures to attempt a composition analysis of each rock unit. However the inconclusive results of both reference mineral and rock composite signature matching resulted in an undefined mineral composition of the delineated units. Although the unit composition remained undefined, the spatial variability in spectral contrast enabled the production of a representative geological map of the Virgin Mountains, Nevada