Introduction: The first stage of the exploration of the rim of Endeavour crater by the Mars Exploration Rover (MER) Opportunity concluded with the rover’s departure from the segment of the rim known as Cape York on sol 3309 of its mission. An important component of our understanding of the nature of rocks exposed on Cape York has come from analysis of multispectral images collected by the rover’s Pancam instrument. Farrand et al., 2013) described multispectral rock classes observed from Opportunity’s arrival at Cape York to the time of its winter-over at the site known as Greeley Haven. Here we describe spectral classes of rock surfaces observed from the time of Opportunity’s departure from Greeley Haven to the time of its departure from the southern point of Cape York (Fig. 1). VNIR S PECTRAL R OCK C LASSES O BSERVED BY O PPORTUNITY ’ S P ANCAM ON N ORTHERN C APE Y ORK AND ON M ATIJEVIC H ILL ON THE R IM OF E NDEAVOUR C RATER, M ARS W.H. Farrand 1, J.F. Bell 2, J.R. Johnson 3, and M.S. Rice 4 1. Space Science Institute, Boulder, CO; 2. Arizona State University, Temp, 3. Applied Physics Laboratory, Laurel, MD 5. California Institute of Technology, Pasadena, CA Acknowledgements: This work was funded through the first author’s MER Participating Scientist subcontract through JPL. B A B (c) A B C Pancam and Pancam Multispectral Imagery: Pancam has two 1024 rows by 1024 columns charge-coupled devices (CCDs) with a 30 cm stereo separation and a 0.27 mrad per pixel resolution (Bell et al., 2003). The Pancam is mounted 1.5 m above the ground on a mast (the Pancam mast assembly, or PMA). Each camera has an eight-position filter wheel. Multispectral geology observations are made with a 13 filter set including spectrally overlapping channels near 432 and 754 nm resulting in 11 spectrally unique wavelengths in the 430 to 1010 nm range. Data are calibrated to radiance factor with reference to a calibration target with dust accumulation on the target compensated for through a radiative transfer correction. Geologic Map Units: Crumpler et al. (this meeting) have mapped units on Cape York on the basis of morphology and stratigraphic position (Fig. 2). From top to bottom of the sequence, these units are the Burns, Grasberg, Shoemaker, and Whitewater Lake Formations. The Burns Fm. consist of the sulfate cemented sandstones observed by Opportunity from its landing in Eagle crater to its arrival at the rim of Endeavour crater. The multispectral reflectance properties of this unit were described by Farrand et al. (2007). The Grasberg Fm. is similar to the Burns Fm., but chemically distinct and lacks the hematitic “blueberries” found in the Burns Fm. Grasberg, and detrital materials on the bench of Cape York, host gypsum veins (Squyres et al., 2012). The Shoemaker Fm. was examined extensively during the initial exploration of Cape York and the reflectance of this and other materials was described by Farrand et al. (2013). The Whitewater Lake Fm. consists of very fine-grained materials, has patchy dark coatings, contains scattered occurrences of spherules (“newberries”), distinct from the Burns Fm. blueberries, and on the basis of orbital data is believed to contain Fe/Mg smectites. Whitewater Lake (WwL) is shot through with veins, nominally consisting also of gypsum, and in places is fractured into boxwork structures with veins between the boxwork cells. In order to determine the spectral distinctiveness and separability of these units, and their component materials, a dataset of 105 combined eye (11 band) spectra of diverse materials were examined using spectral endmember determination and clustering techniques. Spectral Endmember Determination: Spectral endmembers define the boundaries of spectral variability of a dataset (Adams and Gillespie, 2006). Endmember determination approaches resident in the commercial ENVI soft-ware were applied to the 105 spectra dataset with the identification of 4 to 5 (depending on whether or not spectra from the RAT-ground Esperance target are included) endmembers (Fig. 3). These spectra are the most spectrally unique from the dataset, but are not necessarily the only sets of spectra that are distinctive enough to be identified and distinguished from other geologic materials. To find sets of spectrally distinctive spectra, a hierarchical clustering approach was also used. Clustering of Cape York/Matijevic Hill Spectra: The hierarchical clustering methodology in the MATLAB Statistical Toolbox was used to find the hierarchy of spectral similarity in the 105 spectra dataset. Hierarchical clustering allows the analyst to view connections between spectral clusters and to distinguish between broader clusters of generally similar spectra and tighter clusters of more closely similar spectra. The result of the method is the dendrogram shown in Fig. 4. Boxes are overlaid to identify clusters representing classes of related spectra. Fig. 5 shows images of materials from which spectra shown in Fig. 6 were extracted. Examination of Parameters of Spectral Classes: Plots of spectral parameters derived from the hierarchical clustering (HC) classes provide a graphic representation of spectrally distinctive features of these units. The plot of fitted reflectance peak position vs. red/blue ratio (Fig. 7) provides a measure of relative oxidation of the surfaces with Shoemaker Fm. surfaces being minimally oxidized and vein materials more oxidized. 904 nm band depth vs. 535 nm band depth (Fig. 8) indicates the presence of potentially more crystalline ferric iron materials (high 535 nm band depth) with the Grasberg Fm. being highest in this parameter. High 904 nm band depth indicates the presence of crystalline ferric oxides and/or ferrous silicates such as low-Ca pyroxene and the hematitic blueberries are high in this parameter (high 904 nm band depth caused by hematite) Shoemaker Fm. is also high in this parameter and based on APXS results this is nominally caused by low-Ca pyroxene. Fig. 7. Plot of fitted reflect- ance peak position vs. red/blue ratio for HC classes. Fig. 9: 934 nm band depth vs. 934 to 1009 nm slope of vein classes. Fig. 1. HiRISE view of Cape York, major features and traverse. Fig. 2. Geologic traverse map of Cape York (courtesy of Dr. L. Crumpler). Fig. 5. A. Grasberg Fm. outcrop (sol 3024, P2543, L257). B. Whitewater Lake outcrop (sol 3074, P2564, L257). C. Monte Cristo (bench unit vein) (sol 2969 P2591 L257). D. RAT grind on Sturgeon River newberries (sol 3253, P2570, L357). E. Lihir boxwork (sol 3230, P2563, L357). F. Shoemaker Fm. outcrop (sol 2949, P2586, L357). Fig. 6. Representative spectra from major hierarchical clustering (HC) classes. Fig. 3. PCA/nD visualization endmembers for 105 spectra dataset from northern Cape York and Matijevic Hill Spectral Differences between Veins: The breakout of the veins into different groups by the hierarchical clustering is supported by their plotting separately on spectral parameter plots. However differences as in Fig. 8 are primarily in 535 nm band depth which is unrelated to the content of gypsum or other hydrated materials. Fig. 9 shows 934 nm “band depth” (in this case, a measure of positive convexity at 934 nm) vs. 934 to 1009 nm slope. This figure indicates a trend from steeper 934 to 1009 nm slopes and more convexity at 934 nm in the bench unit veins (Homestake, Ross, and Monte Cristo to shallower and less convexity for the Whitewater Lake veins and even more so for some of the boxwork veins. The drop in reflectance is attributed to the presence of a water overtone absorption band near 1  m that can be detected in some hydrated materials, notably including gypsum (Fig. 10), by Pancam (Rice et al., 2010). The steeper 934 to 1009 nm slope and greater convexity at 934 nm is consistent with a deeper band depth of the 1  m water overtone absorption band, and nominally greater hydration, in the bench veins vis-à- vis the Whitewater Lake and boxwork veins. Spectral Differences between Undisturbed and RAT-ground Newberries: The spectra of the undisturbed spherules in the Whitewater Lake Fm., the “newberries” are generally similar to Shoemaker Fm. spectra and were not broken out as a distinct class by the hierarchical cluster analysis. Hypotheses for the nature of the newberries include their being accretionary lapilli formed in the cloud of impact debris or from volcanic ash or that they are concretions, albeit not with the concentration of hematite observed in the Burns Fm. concretions (the “blueberries”). Concretions in terrestrial sand-stones can have concentric shells with iron oxide or oxyhydroxide cements (e.g., Chan et al., 2007). The newberries do display a concentric structure in MI images (Fig. 11) After grinding into a concentration of Newberries at the Sturgeon River target, Pancam multispectral data indicated subtle spectral differences in the RAT cuttings indicated by the fitted reflectance peak position vs. 535 nm band depth plot in Fig. 12. A longer wavelength peak position and higher 535 nm band depth in the RAT cuttings indicates that the cuttings nominally have more oxidized material- which would be consistent with the concretion hypothesis. References: Adams, J.B. and A.R. Gillespie (2006) Remote Sensing of Landscapes with Spectral Images; Bell, J.F. et al. (2003) JGR Planets, 108, /2003JE002070; Chan, M.A. et al. (2007) Geofluids, 7, 1-13; Crumpler, L.S. et al. (this meeting) Paper No ; Farrand et al. (2013) Icarus, 225, ; Farrand et al. (2007) JGR Planets, 112, E06S02, /2006JE002773; Rice, M.S. (2010) Icarus, 205, ; Squyres, S.W. et al. (2012) Science, 336, Fig. 10. Laboratory reflectance spectrum of gypsum (blue) and convolved to Pancam bandpasses (black asterisks). Fig. 4. Dendrogram resulting from hierarchical clustering of 105 spectra dataset. Fig. 8. Plot of 904 nm band depth vs. 535 nm band depth for HC classes. Fig. 11. Merge of Pancam color over MI mosaic over the newberry-rich Kirk- wood target. Fig. 12. Plot of fitted reflectance peak position vs. 535 nm band depth for undisturbed newberries and for cuttings from RAT grind into newberries.