SPECTRAL DIMENSIONALITY COMPARISON BETWEEN MULTISPECTRAL LANDSAT 7 ETM+, ASTER, AND EO-1 ALI, AND HYPERSPECTRAL AVIRIS AND HYPERION

Most multispectral sensors such as Landsat 7 ETM+, ASTER, and ALI measure reflected energy in a few and separated wavelength bands and are thereby referred to as multispectral. In contrast, hyperspectral imaging spectrometers have a sufficient amount of continuous spectral channels to resolve the spectral variability of the earth’s surface. This study focuses on uncovering the spectral complexity of Alkali Flat and Lake Lucero, White Sands, N.M. The goal of the present study was to compare the information dimensionality limits of Landsat 7 ETM+, ASTER and ALI data with high spectral resolution, low signal-to-noise Hyperion data, using AVIRIS data as a high spectral resolution, high signal-to-noise standard. Information dimensionality thresholds constrain the number of evaporite mineral and textural endmembers that can be detected, mapped, or unmixed using standard image processing methods. In this case, eigen analysis of the Minimum Noise Fraction (MNF) transformation and Principle Component Analysis (PCA) results were used to uncover the information dimensionality limits of all these datasets. The Pixel Purity Index (PPI), and n-D Visualization were used to identify spectrally distinct endmember classes, and along with ground truth and field spectra provide insights into the origin of the spectral variability uncovered in all of the datasets. The results of this study confirm that AVIRIS and Hyperion data have higher information dimensionality thresholds exceeding the number of available bands of Landsat 7 ETM+, ASTER, and ALI data. AVIRIS has higher information dimensionality thresholds than Hyperion because of its 4-10x increase in signal-to-noise over Hyperion. Much of the information dimensionality uncovered by AVIRIS and Hyperion datasets can be due to differences in vegetation types, landcover, grain size variations, and moisture contents as opposed to mineral compositional variations. These endmember components uncovered at Alkali Flat and Lake Lucero using aforementioned datasets include: (1) dry gypsum, (2) wet gypsum, (3) water, (4) vegetation, and (5) clastic sediments containing ferric iron and calcite. The spectral signatures of these endmembers at the resolution of AVIRIS, Hyperion, ALI, ASTER, and Landsat 7 ETM+ data agreed with field and laboratory spectral analysis.