Decomposition of AVIRIS spectra: Extraction of surface-reflectance, atmospheric, and instrumental components

We present techniques that use only information contained within a raw, high-spectral-resolution, hyperspectral Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) scene to estimate and remove additive components (atmospheric scattering and instrument dark current), These techniques allow normalization of multiplicative components (instrument gain, topography, atmospheric transmission) and enhancement, extraction, and identification of relative-reflectance information related to surface composition and mineralogy, Our derivation of additive components from raw AVIRIS data is based on an adaptation of Crippen's ''regression intersection method (RIM).'' As does RIM, we use pairs of surface units that are spectrally homogeneous, spatially extensive, and located in rugged terrain, However, our technique utilizes the long-wavelength spectral data of AVIRIS to derive and remove atmospheric scattering components for each unit. AVIRIS data from the Kelso Dunes and Granite Mountain areas of southern California served as spectrally contrasting, topographically modulated surfaces for illustration of this technique. For a given site and wavelength pair, subtraction of the wavelength-dependent additive component from individual bands will remove topographic shading in both sites in band-to-band ratio images, Normalization of all spectra in the scene to the average scene spectrum results in cancellation of multiplicative components and produces a relative reflectance scene. Absorption features due to mineral absorptions that depart from the average spectrum can be identified in the relative-reflectance AVIRIS product, The validity of these techniques is demonstrated by comparisons between relative-reflectance AVIRIS spectra derived from application of this technique and those derived by using the standard calibration techniques of JPL. Calibrated spectra were extracted from an AVIRIS scene of the Upheaval Dome area of Canyonlands National Park, UT, Results show that surface-reflectance information can be extracted and interpreted in terms of surface mineralogy after application of these techniques to AVIRIS data.