Use of Multiple Endmember Spectral Mixture Analysis and Radiative Transfer Model to Derive Lunar Mineral Abundance Maps

Introduction: Determination of mineral composition of the lunar surface is fundamental for understanding the origin and geologic evolution of the Moon. Such compositional information can be used to examine the magma ocean model of lunar crust formation and to investigate lunar crustal structure, basaltic volcanism, crater/basin structure and ejecta emplacement, as well as lunar soil mixing mechanisms. While exploration of the Moon is entering an explosive measurement phase with the ongoing flotilla of remote sensing spacecraft of this decade, lunar mineral maps will not issue from a mission because a well performing remotely mapping method is not available. Current radiative transfer modeling (e.g., Hapke's work) is time consuming when applied for both multispectral and hyperspectral mapping, and other empirical approaches (e.g. spectral band and band ratios) perform poorly for some minerals (e.g., plagioclase and olivine). The proposed approach is a combination of state of the art methods for deriving mineralogy from individual spectra of lunar materials in which spectral mixing analysis (SMA) is used to reduce time needed to derive mineral maps from large lunar data sets and a radiative transfer model (RTM) used for the individual spectrum analysis. This hybridized method is tested with lunar global Clementine ultraviolet (UV)-visible (VIS) multispec-tral (0.45-1.0 µm, 5 color) image at 1 km spatial resolution. Multiple Endmember Spectral Mixture Analysis (MESMA): SMA is a commonly used method for subpixel detection and classification of remotely sensed data[1-6]. Linear SMA assumes that the reflec-tance of each mixed pixel is a linear combination of spectra of distinct components (endmembers) with the weights representing the abundances of endmembers resident in a mixed pixel. SMA produces a series of images that contain the estimated abundance of end-member materials. The maximum number of components that SMA can map is limited by the number of bands in the image data. In relatively small regions with only a few geological components, SMA can be used to map the compositional distribution with reasonable accuracy [6, 7]. However, over large regions extending to global scales the number of lunar surface components exceeds the number of bands in Clemen-tine UV-VIS multispectral data, violating the assumptions of traditional SMA. MESMA was proposed to overcome the limitations of traditional SMA [8]. MESMA assumes that although an image contains a large number of spectrally distinct components, indi