LUNAR SURFACE COMPOSITIONAL UNITS DETERMINED BY SPECTRAL MIXING ANALYSIS OF IMAGES FROM THE MOON MINERALOGY MAPPER (M3)

Mapping-surface compositional units on large areas of the Moon is a key step for interpretating its geology. In addition, the spatial distribution and relative abundances of minerals and glasses are essential for the study of mixing processes and maturation of the soil. We are using data from the M3 imaging spectrometer [1], which was in lunar orbit onboard Chandrayaan-1 for 10 months starting in November 2008. In global observation mode [1], the spectral range is 460-2976 nm at 20 and 40 nm spectral resolution, and the spatial resolution is 140 m/pixel or 280 m/pixel. Spectral Mixing Analysis (SMA) is one method for calculating abundances of spectral components (endmembers) mixed within a surface-projected pixel [2-8]. Spectra are modeled by linear combinations of the spectral endmembers, which correspond to adjacent areas of different compositions present within the same pixel. The inversion (unmixing) of this simple physical model is convenient for an initial assessment of large data sets prior to using more sophisticated methods for compositional analysis [7-9]. In the present study, the spectral endmembers are collected from the image that will be unmixed using two ways of selection in order to arrive at the most effective ones: 1) Among the representative nearside mare compositions defined by [10], we chose the most-extreme spectra, based on their absorption bands and on the titanium content of the corresponding lunar samples. 2) Using our own iterative approach, we start with the two most-representative spectra of mature highlands and mature mare soils as input for the SMA. Then, we analyze the residuals of the SMA to define more endmembers. We perform SMA using the Multiple-Endmember Linear Spectral-Unmixing Model (MELSUM, [8, 11]) that allows limiting the number of components used in a model and guarantees positive mixing coefficients. Shade is assumed to have a neutral (flat) spectral contribution. The sum of the mixing coefficients is constrained to unity. First-resulting maps show a wealth of detail in the compositional spatial structure. Maps of mare basalt soils show variations of mafic minerals, highlands, and impact-ejecta contamination. Analysis of the residuals is efficient to identify any type of spectral endmember confined in small areas, such as spectra of fresh materials exposed by recent impacts. [1] Pieters et al., 2009, LPICo # 6002. [2] Pieters C. M. et al., 1985, JGR, 90,12,393-12,413. [3] Adams J. B. et al., 1986, JGR, 91, 8098-8112. [4] Pinet P. et al., 1993, Science, 206, 797-801. [5] Smith M. O. et al., 1985, JGR, 90, c797-c804. [6] Staid M. I. et al., 1996, JGR, 101, 23,213-23,228. [7] Li L. and Mustard J. F., 2000, JGR, 105: 20,431-20450. [8] Combe et al., 2008, 39th LPSC 2247. [9] Mustard J. F. et al., 1998, JGR, 103, 19,419-19,425. [10] Gillis-Davis et al., 2006, GCA 70, 24. [11] Combe J.-Ph. et al., 2008, PSS 56.