Mapping the Spatial Distribution, Mineralogy, and Geochemistry of Lunar Highlands

Introduction: Knowledge of the surface mineralogy and chemistry of the crust is central to understanding the evolution of the lunar crust. Using the 5-band ultravioletvisible (UVVIS) Clementine data set, Tompkins and Pieters [1] showed that approximate mineralogy could be determined from Clementine data and analyzed 109 impact craters detailing the approximate mineralogy, or rocktypes, of their central peaks. The study has been extremely helpful allowing a look at the compositional diversity of what is assumed to be immature lunar crustal material. Subsequently, these data were implemented by Wieczorek and Zuber [2] into a dual-layered crustal model that extrapolated the general rock-type proportions of the lunar crust with depth. Here we build upon Tompkins and Pieters [1] pioneering study by considering several aspects that remain unanswered. These aspects include: (1) consideration of surface maturity, (2) the distribution of spectral classes within each crater analyzed, (3) quantitative determination and variation of lunar spectral-type mineralogy, and (4) consideration of the geochemical discriminators FeO and Mg#. Here we aim to begin improving upon these aspects by recharacterizing these 109 craters using radiative transfer modeling techniques. Data: In 1992, Clementine collected 11-band multispectral data of most of the lunar surface in the ultraviolet (UV), visible (VIS), and near-infrared (NIR) portions of the spectrum [3]. From this data, the USGS produced a global mosaic sampled at 100m/pixel for the five of the bands between 0.415 to 1.0 μm [4]. Images for twelve impact craters of interest were identified and collected from this data. Method: Determination of the spatial distribution of Tompkins and Pieters [1] spectral classes within each impact crater was accomplished using a correlation algorithm detailed by Clark et al [5]. Each spectral class was digitized and mathematically compared to each Clementine region of interest pixel-by-pixel. Pixels with spectra >99% correlated with a spectral class were selected to be mapped. The spectra from these mapped regions were averaged and used to calculate mean FeO, maturity, Mg#, and mineralogy. In order to determine mineralogy and Mg# a radiative transfer model was used. A non-redundant look-up table of mineralogy was created using the minerals olivine, orthopyroxene, clinopyroxene, and plagioclase. The abundance of these minerals relative to each other was varied by 5% modal abundance (e.g. 5 ol, 25 opx, 30 cpx, and 40 pl) and was constrained to sum 100%. This created a mineral look-up table with 1771 modal mineralogical combinations. A radiative transfer model based upon work by Hapke [6-8] and Lucey [9], was used to create representative spectra for 10 maturity levels of each mineralogical combination that were characteristic of Mg#. This resulted, in 17,710 compositional and spectral variations produced between 0.75 and 1.0 μm. Ten of these databases were created based upon the variation of Mg# from 50 to 95 in increments of 5. In total, a database of 177,100 spectra was created for comparison with Clementine spectra. Both Clementine and each database of 17,710 computed reflectance spectra were normalized by dividing the reflectance by the sum of the four bands used for this model. Each Clementine reflectance spectrum was then compared to each of the 17,710 model spectra and an RMS error was taken. The chosen “best-fit” model spectrum for the Clementine pixel exhibited the lowest RMS error. This selection process occurred for each pixel ten times (for each Mg#), giving 10 best-fit possibilities for each Clementine pixel. Mg# was determined by calculating the difference between Clementine derived FeO [10] and the stoichiometric FeO estimated from our assigned model mineralogy was calculated to yield ∆FeO. For each pixel ∆FeO was calculated for each of the ten Mg# values. The Mg# chosen for a given Clementine pixel possessed the best agreement between FeO calculation methods. The mineralogy was chosen by picking the model with the ∆FeO closest to zero.