Gray hematite distribution and formation in Ophir and Candor chasmata

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 113,, doi: /2007je002930, 2008 Gray hematite distribution and formation in Ophir and Candor chasmata Catherine M. Weitz, 1 Melissa D. Lane, 1 Matthew Staid, 1 and Eldar Noe Dobrea 1 Received 18 April 2007; revised 4 September 2007; accepted 12 November 2007; published 29 February [1] We have studied eight of the spatially largest locations in Ophir and Candor chasmata that have the highest abundances of coarse-grained gray hematite, as measured by the Mars Global Surveyor Thermal Emission Spectrometer (TES). All of the eight hematite locations correspond to relatively smooth dark units that overlie or are adjacent to lighttoned rock units. Linear modeling of TES spectra sampled in the eight hematite locations shows mostly mafic associations with the hematite, dominated by pyroxene and feldspar, with a lower abundance of sulfates. The spectral and morphologic similarity between the hematite-bearing units at Ophir and Candor to those at Meridiani Planum lead us to favor a similar aqueous origin for the hematite in Valles Marineris. Water that may have infiltrated the light-toned rocks by an increase in the water table or when lakes may have existed within the canyons could have resulted in the formation of hematite-rich grains in a scenario similar to that proposed for Meridiani Planum. Erosion by wind of the formerly overlying or adjacent friable light-toned rocks that contain the hematiterich grains enabled the more resistant hematite grains to be concentrated in sufficient abundances along the surface to be detected by TES. A more limited supply of hematite grains within the light-toned rocks, inadequate winds, subsequent burial beneath dust or sand, and fewer low-lying terrains upon which the grains could form a lag deposit may explain why there are smaller exposures of hematite within Valles Marineris compared to Meridiani Planum and Aram Chaos. Citation: Weitz, C. M., M. D. Lane, M. Staid, and E. N. Dobrea (2008), Gray hematite distribution and formation in Ophir and Candor chasmata, J. Geophys. Res., 113,, doi: /2007je Introduction [2] Gray, crystalline hematite (as opposed to fine-grained red hematite or nanophase hematite [Lane et al., 1999]) was identified by the Thermal Emission Spectrometer (TES) initially at three main locales on Mars: Meridiani Planum, Aram Chaos, and Valles Marineris [Christensen et al., 2000a, 2001]. Several origins for the coarse, gray, crystalline hematite (hereinafter referred to as hematite) were proposed by Christensen et al. [2000a, 2001], but their favored interpretations were direct precipitation of fine-grained hematite from Fe-rich circulating fluids of hydrothermal origin or low-temperature precipitation of fine-grained Fe-oxides/ hydroxides from standing, oxygenated, Fe-rich water, followed by subsequent alteration (coarsening) to gray hematite. Lane et al. [2002] hypothesized that burial metamorphism could have coarsened and preferentially aligned the hematite crystals. Origins proposed by others include: volcanism [Noreen et al., 2000; Chapman and Tanaka, 2002; Hynek et al., 2002; Arvidson et al., 2003], a high-temperature hydrothermal alteration [Catling and Moore, 2003; Hynek et al., 2002], a surface coating that 1 Planetary Science Institute, Tucson, Arizona, USA. Copyright 2008 by the American Geophysical Union /08/2007JE may not involve water [Kirkland et al., 2004], formation via the same mechanism that caused the bleached beds and hematite concretions in Utah [Chan et al., 2000, 2004; Ormo et al., 2004], or formation analogous to terrestrial banded iron formations [Calvin et al., 2003; Fallacaro and Calvin, 2003]. Glotch and Christensen [2005] provide more details on each of these proposed origins as part of their analyses of the hematite identified in Aram Chaos. [3] On 24 January 2004, the Mars Exploration Rover (MER) Opportunity landed inside Eagle crater at Meridiani Planum. Observations within Eagle crater and along the plains indicated that the source of the orbital hematite signature seen by TES was millimeter-size spherules containing hematite that were concentrated along the upper centimeters of the soil as a lag deposit [Squyres et al., 2004; Christensen et al., 2004a; Klingelhöfer et al., 2004; Rieder et al., 2004; Weitz et al., 2006a]. The hematite-rich spherules are postulated to have formed by secondary alteration of the sulfate-rich outcrop as water permeated through the rocks and produced concretions [McLennan et al., 2005]. The size of the spherules ranges from about 1 to 4.5 mm [Weitz et al., 2006a]. Within the outcrop, the spherules make up only a few volume percent of the rocks so their signature is not strong enough to be detected in the rocks themselves from orbit. It is only because erosion of the sulfate-rich host rock has concentrated the more resistant hematite spherules in larger abundances as a lag deposit in the soils that the 1of30

2 Figure 1. Thermal Emission Imaging System (THEMIS) mm (band 9) daytime IR global mosaic of the regional view of the Valles Marineris study area. White box shows where the eight patches of hematite analyzed in this paper are located. signature at Meridiani Planum was strong enough to be detected from orbit. Hence other sulfate-rich rocks with hematite spherules may exist on Mars, but the hematite may not have been detected in TES because the hematite spherules are not concentrated at high enough abundances along the surface soils. [4] Unlike Meridiani and Aram Chaos where the hematite units are confined to a specific layer or fairly continuous unit [e.g., Christensen et al., 2001; Catling and Moore, 2003; Glotch and Christensen, 2005], the gray hematite in Valles Marineris is more patchy in distribution and is scattered in separate troughs. The hematite seen within Valles Marineris is generally adjacent to or on light-toned rocks (LTR). These rocks are light toned in visible images relative to the adjacent wall rock and plains units, but they can be covered by varying amounts of darker debris making them appear as dark toned as these other two units. Thus we refer to them as light-toned units based upon clean exposures that suggest they are relatively brighter than nearby wall rock and plains units. Although not all exposures have layering, many Mars Orbiter Camera (MOC) and High Resolution Imaging Science Experiment (HiRISE) images of LTR reveal extensive layering, including repeating beds of similar thickness and erosional expressions consistent with sedimentary rocks [e.g., Malin and Edgett, 2000; Weitz et al., 2007]. The LTR have multiple proposed origins since they were initially identified in Mariner 9 images by McCauley [1978]. Origins for the LTR include deposition in standing bodies of water [Malin and Edgett, 2000; Catling et al., 2006], volcanism [Chapman and Tanaka, 2001; Hynek et al., 2002], or air fall deposition [Malin and Edgett, 2000; Catling et al., 2006]. Results from Opportunity rover suggest that the light-toned layered rocks at Meridiani Planum are sulfate-rich eolian and interdune facies derived from reworked evaporites originally formed in a playa environment [Grotzinger et al., 2005]. The detection of gypsum in Juventae Chasma LTR by the Observatoire pour la Mineralogie, L Eau, le Glace e l Acti- Figure 2. (a) Estimated hematite abundances derived by the linear modeling method (see text) and overlain on THEMIS mm daytime IR mosaic. (Note that the percentage hematite values are not absolute.) The eight patches of hematite that we have analyzed are shown by yellow ovals. Orbit-correlated noise can be seen as linear chains of pixels aligned near-vertical along some orbit tracks. (b) Estimated hematite abundance derived from the spectral index approach and overlain on THEMIS mm daytime IR mosaic. Scale bar represents hematite index value (not actual percentage) where the higher the value, the more hematite is present. (c) Viking Orbiter Mars Digital Image Mosaic (MDIM) with locations of hematite in Ophir and Candor, as determined by Christensen et al. [2001, Figure 4b]. Red ovals correspond to the largest hematite patches. Figure courtesy of P. Christensen. 2of30

3 Figure 3. Hematite distribution overlain on THEMIS mm daytime IR mosaics for (a) Ophir and Candor chasmata, (b) Aram Chaos, and (c) Meridiani Planum. The estimated hematite abundances (not absolute) were derived using the same linear modeling method, so colors can be compared between the different regions for relative hematite abundances. The location of the Mars Exploration Rover Opportunity landing site is shown in Figure 3c with a black cross. 3of30

4 Table 1. Wavelengths Used for the Definition of the 1.94-, 2.13-, and 2.4-mm Band Depths 1.94 mm 2.1 mm 2.4 mm Band center Low contiguous channel High contiguous channel vite (OMEGA) instrument on the European Space Agency s Mars Express mission led Catling et al. [2006] to exclude a volcanic origin for these rocks because gypsum only forms at low temperatures. Glotch and Christensen [2005] found that the hematite in Aram Chaos was associated with LTR and proposed that the hematite-bearing units formed in a water-rich environment. [5] In this paper, we explore the geologic context of the hematite within central Valles Marineris in order to determine if hematite is associated with a specific layer as occurs in Aram Chaos [Glotch and Christensen, 2005] or if hematite corresponds to a surface lag like at Meridiani Planum [Christensen et al., 2004a]. The study areas we discuss are shown in Figure 1 and include Ophir Chasma, West Candor Chasma, Central Candor Chasma, and East Candor Chasma. Hematite patches have also been identified in other locations of Valles Marineris [Christensen et al., 2001; Knudson, 2006], but for this paper we restrict our analyses only to locations in these two chasmata in Valles Marineris. We investigate both the geologic units where hematite occurs as well as adjacent geology that could potentially be the source of the hematite. In particular, the LTR at the Opportunity landing site are the source for the hematite-rich spherules and, consequently, we believe LTR are the most plausible source for hematite in Valles Marineris. We have therefore looked closely at images of LTR that lie within or near any patches of hematite in our study areas. After exploring the geologic units associated with each patch of hematite, we propose likely scenarios for the formation of these hematite-rich localities within Valles Marineris. Figure 4. Atmospherically corrected and continuumremoved spectral types of sulfates in west Candor Chasma (red) compared to (a) laboratory spectra of polyhydrated magnesium sulfate (white, CRISM spectral library sample CJB366) and (b) kieserite (white, CRISM spectral library sample CC-JFM-015). The observed spectra constitute an average of all the Observatoire pour la Mineralogie, L Eau, le Glace e l Activite (OMEGA) spectra (orbit 360, observation 2) presenting hydrate band depths (band depths at 1.94 and 2.1 mm for polyhydrated sulfates and kieserite, respectively) greater than 4%. This translates into 7 pixels for the polyhydrated sulfate spectrum and 20 pixels for the kieserite spectrum. The laboratory spectra were obtained from the CRISM spectral library. They have been reduced in contrast by 8, and all spectra have been offset from each other for ease of comparison. 2. Methods 2.1. TES Data [6] TES emissivity spectra were analyzed for hematite mineral detection. TES is a Fourier-Transform Michelson Figure 5. Sulfate distribution map as derived from OMEGA data (orbit 360, observation 2). Monohydrated sulfates (i.e., kieserite) are shown in green, and polyhydrated sulfates are shown in red. The OMEGA data is shown overlying coregistered High-Resolution Stereo Camera (HRSC) data (h0334_0001) of the region. 4of30

5 Figure 6. Thermal Emission Spectrometer (TES) MESMA modeling results for hematite patches 1 6. The TES emissivity spectrum (shaded line), the modeled best fit surface plus atmosphere (black line), and associated RMS errors are shown for each site. Mineral abundances obtained for each patch have been grouped by type and are presented before and after normalization. See text for further discussion of results. interferometer that covers the wavelength range from 6 to 50 mm, or 1670 to 200 cm 1 [Christensen et al., 1992]. Hematite abundance was derived in two ways. One approach was to utilize the perc_hem option in the software program Java Mission Planning and Analysis for Remote Sensing (JMARS) [Gorelick et al., 2003] to map the percentage of hematite as determined using a linear modeling method [Bandfield et al., 2000a]. Data were constrained to ocks (where orbit number = ock 1683), surface temperatures >240 K, atmospheric opacities rated in the data archive as good or questionable but not bad, and emission angles of 0 to 30 (increased from a typical maximum of 20 in order to fill in coverage gaps) where the data were high quality. The data were overlain on a Mars Odyssey Thermal Emission Imaging System (THEMIS) infrared (IR) daytime map of band 9 (12.57 mm) 5of30

6 Figure 7. Comparison of the average modeled surface spectrum from all hematite patches analyzed in Ophir and Candor chasmata relative to representative Mars surface types ST1 and ST2 [Bandfield et al., 2000a]. using the program JMARS ( A hematite distribution map created from this approach is shown in Figure 2a with a spatial resolution of 3 8 km/pixel (smeared in the along-track direction). [7] A second approach was to determine the relative hematite abundance by creating a hematite map using a derived spectral index. This index utilized the wave number positions and emissivity (e) values of several dominant bands associated with coarse, gray hematite, and the highemissivity region between them [Christensen et al., 2000a, 2001; Lane et al., 2002]. The index value was calculated according to the following equation: e 375 cm 1 ðe 315 cm 1 þ e 460 cm 1Þ where e 375 cm 1 is the average emissivity from TES bands 21 24, e 315 cm 1 is the average emissivity from TES bands 16 to 18, and e 460 cm 1 is the average emissivity from TES bands 30 to 32. The data were also constrained to the parameters listed above for the perc_hem approach (for ock, target temperature, atmospheric opacity, and emission angle). This hematite index value was mapped to a THEMIS IR daytime map of band 9 (12.57 mm) to show the location and abundance of gray hematite (Figure 2b). This index varies from that used in Christensen et al. [2000a] in that both denominator band positions were adjusted by 20 and 15 wave numbers, respectively, to optimize the index band positions to an average laboratory hematite spectrum derived by averaging 28 different hematite spectra from Lane et al. [2002, Figures 5 8]. [8] The main difference between the two methods is that the index maps are measuring the degree of inflection between two hematite bands in the spectrum associated with hematite while the perc_hem values relate to modeling with various spectral units derived by target transformation (see next paragraph). Both approaches yielded similar hematite distribution maps, although the map derived from the index approach shows spatially larger patches compared to the modeling method map. Because the index approach produced more noisy pixels in the distribution map for the eight patches, particularly evident by orbit-correlated pixels (Figure 2b), we decided to use the hematite distribution map derived from the linear modeling method in our analyses, although we refer to the index value map for clarification in some instances. [9] The percentage of hematite determined by the modeling method represents the amount of a hematite endmember used to fit the calibrated emissivity spectrum along with six other end-member spectral shapes from Bandfield et al. [2000a, 2000b], including Surface Type 1 (ST1), Surface Type 2 (ST2), atmospheric dust (during high and low opacity), and water ice (large and small particle sizes). This percentage is somewhat arbitrary, however, because it is affected by the depths of the bands of the individual endmembers. In addition, we do not know important details about the hematite grain size and spectral contrast depends on particle size. The lowest emissivity value (e min )ofthe hematite end-member is 0.89 and the e min for the other six end-members are 0.92 (ST1), 0.94 (ST2), and 0.80 for each of the four atmospheric components. [10] Because the resulting mapped hematite is shown to occur in numerous small patches within our study region, we have focused on only those locations where the hematite abundance occurs in the largest patches (i.e., across several TES orbits) and with an abundance of >18%, keeping in mind that this is a qualitative abundance, not absolute. The cutoff of 18% hematite abundance was selected because when the limit was set any lower than that, a significant portion of the entire image would get mapped as hematite where there was none (as determined by checking individual pixel emissivity spectra). The 18% cutoff therefore limited the number of falsely mapped hematite locations. [11] After restricting our hematite detection to >18% abundance, we found eight patches of hematite that we have analyzed as part of this work (Figure 2a). These eight patches were also seen by Christensen et al. [2001], although they also identified a patch 9 (Figure 2c) that we did not detect. After examining individual spectra of the patch 9 location, we found no clear spectral evidence for the presence of hematite here. [12] A similar restriction of >18% hematite when applied to Meridiani Planum and Aram Chaos (Figure 3) resulted in distribution maps similar to those found by Christensen and Ruff [2004] and Glotch and Christensen [2005], respectively, providing confidence in our technique for identifying and determining relative abundances of hematite. A comparison between the hematite maps for the three regions (Figure 3) indicates that the relative abundance of hematite (shown as a function of color) in Ophir and Candor is generally lower than in both Aram and Meridiani. Glotch and Christensen [2005] report hematite abundances between 2 and 16% for Aram Chaos while Christensen and Ruff [2004] found values between 5 and 20% across Meridiani Planum. Because only a few pixels in our hematite maps of Ophir and Candor have similar high-abundance colors matching those seen in Aram Chaos and Meridiani Planum, we infer that most of the hematite abundances within Ophir and 6of30

7 Figure 8. Ophir region with patch 1 hematite. (a) Mars Orbital Laser Altimeter topography in color with 500-m contour intervals overlain on a portion of HRSC image h0334_0001. (b) Hematite abundance (not absolute) derived from a linear modeling approach shown on same HRSC image as Figure 8a. Boxes and numbers correspond to later figures. Candor fall below 20% abundance if we use abundance estimates derived by Glotch and Christensen [2005] and Christensen and Ruff [2004]. Christensen et al. [2001] calculated hematite abundances between 2 and 15% for patches within Valles Marineris. Consequently, the 18% arbitrary abundance cutoff we use for elimination of TES pixels in our modeling method should not be considered a 7of30

8 Figure 9. HiRISE subimage of PSP_002208_1755 at 30 cm/pixel showing layering (bold arrows) in exposures of LTR. quantitative value because it is greater than the estimated abundances determined by others. [13] In each of our study areas, the hematite can be seen across multiple TES orbits, reducing the chance that the detection is a result of atmospheric or instrument artifacts. In addition, these larger patches provide the best opportunity to correlate the hematite to particular geologic units that can be identified in visible or thermal infrared images. There are other small patches of hematite within our study region but the small sizes and low percentage of hematite for these smaller locations makes them less favorable areas Figure 11. Mantle of light-toned material (bold arrows) covering portions of LTR in Ophir. Possible impact crater shown by white arrow. Portion of THEMIS visible spectrometer (VIS) image V for analyzing geologic units that correspond to the hematite. Hence although hematite is more widespread in distribution than shown in our abundance maps, we focus on the eight patches shown in Figure 2a for this paper. Figure 10. Exposure of light-toned rocks (LTR) in a dark debris unit along basin floor in Ophir Chasma. Portion of HiRISE image PSP_002208_1755. Figure 12. Southern wall rock in Ophir with exposures of LTR. A dark mantle covers both the wall rock and LTR, making a stratigraphic relationship between the wall rock and LTR difficult to determine. Portion of Mars Orbiter Camera (MOC) image S of30

9 chose to ignore the 1.4- to 1.6-mm region and define the presence of polyhydrated sulfates by the presence of its and 2.4-mm bands, and the presence of kieserite by the presence of its 2.1- and 2.4-mm bands (Figure 4). Band depths were determined as described in Noe Dobrea et al. [2008]. Table 1 lists the channels used to determine the band depths and the resulting band depth maps are shown in Figure 5. [17] We defined the relative abundance of sulfates by the band depth of the H-O-H stretch/bend-overtone (1.94- and 2.1-mm for polyhydrated sulfates and kieserite, respectively) feature and the presence of a 2.4-mm feature (as defined in Table 1). Multiple scattering by atmospheric constituents Figure 13. Features in HiRISE image PSP_002208_1755 that correspond to high hematite abundances in the Ophir basin. Small exposures of LTR are visible beneath a relatively smooth dark debris unit. Portions of the dark debris unit have brighter ripples, while other locations show evidence for cohesion along sharp edges (bold arrows) OMEGA [14] OMEGA visible near-infrared (VNIR) data were also used to generate parameter maps that quantify the distribution and relative abundance of previously detected sulfates [e.g., Gendrin et al., 2005] in the region. These maps were coregistered to a THEMIS IR daytime map of band 9 (12.57 mm) and overlaid to study the correlation of sulfates to gray hematite. Solar and atmospheric corrections of the data were performed as described in Noe Dobrea et al. [2008], and band depth maps for the relative abundance of sulfates were derived based on our observations of the spectra observed in OMEGA data (Figure 4). [15] At the time of this writing, there are two OMEGA hyperspectral cubes available in the Planetary Data System (PDS) for analyses that cover some of our study region. The first one (orbit 548, scene 3) covers the entire study region at a resolution of 3.5 km/pixel, whereas the second one (orbit 360, scene 2) covers a portion of West Candor (patch 3 and part of patch 5 in Figure 2a) at a resolution of 850 m/pixel. Unfortunately, OMEGA data from orbit 548, scene 3 has a significant contamination by water ice clouds and is therefore unreliable for our mapping purposes. However, we were able to extract good VNIR spectra for the western portion of our study area (orbit 360, scene 2) and our results are based upon this data set. [16] Polyhydrated sulfates (e.g., epsomite, hexahydrite, gypsum) are primarily characterized by a 1.4-mm stretch O- H stretch overtone, a 1.94-mm stretch/bend-overtone combination, and a 2.4-mm band that is characteristic of the sulfate ion in the presence of water. In monohydrated sulfates, such as kieserite, the O-H stretch overtone is broadened and shifted toward 1.55 mm, and the band center of the stretch/bend-overtone combination band is shifted to 2.1 mm. The response function of the OMEGA detector peaks at about 1.5 mm, which can result in saturation and nonlinear behavior in this spectral region. We subsequently Figure 14. Geologic units in southwestern Ophir basin. The dark debris unit has a cliff along the southern edge, where a valley cuts through the unit. LTR appear layered and hummocky and are covered by dark mantle. The LTR extend to the wall rock in the south. Portion of MOC image S of30

10 Figure 15. (a) Dark ripples on LTR to the west of the Ophir basin do not have corresponding hematite abundances. Portion of MOC image R (b) Debris aprons along slopes in the LTR are relatively brighter than the hematite-bearing dark debris on the basin floor. Portion of MOC image S reduces the spectral band contrast. Because we cannot determine the aerosol column density from OMEGA data, we cannot correct for the loss of spectral contrast due to multiple scattering. This precludes the derivation of actual abundances from OMEGA data. Thus it is only possible to derive relative abundances between points in the same observation. As a caveat, it is important to note that variations in grain size will affect spectral contrast as well, and therefore such variations can be misinterpreted as variations in abundance. In the OMEGA abundance map presented in Figure 5, we make the assumption that the sulfate-bearing outcrops have, on average, similar grain sizes. [18] Inspection of the OMEGA mineral maps in Figure 5 relative to our hematite distribution map (Figure 2a) shows there is no spatial correlation between the sulfates and the hematite. The 1.94-mm feature attributed to sulfates appears to be highly correlated to steep slopes along several of the LTR exposures. In contrast, as we will discuss in the following sections, the hematite is predominantly concentrated in relatively smooth dark terrains adjacent to the LTR. The lack of an exact spatial correlation between the two minerals does not imply that they could not have been derived from the same geologic units, however. Hematite and the sulfates could have formed in the same LTR initially but subsequent erosion and redistribution by the wind and mass wasting would spatially separate the two minerals. The observation that the two minerals are located near each other at the kilometer scale (the resolution of our data sets) suggests this may be the case Visible and Thermal Infrared Data Sets [19] Most of the morphology and descriptions about the hematite-bearing units were based upon observations seen Figure 16. A dark debris mantle along a wall rock spur appears competent. Erosion of the dark mantle appears to produce smaller grains that form nearby dark ripples. Portion of MOC image E of 30

11 Figure 17. (a) THEMIS mm daytime IR mosaic and (b) THEMIS mm nighttime IR mosaic for Ophir Chasma study area. Hematite region along the basin floor is shown by the letter H. in Mars Global Surveyor MOC narrow-angle (NA) images. There are three Mars Reconnaissance Orbiter HiRISE images with scales between 28 and 56 cm/pixel that we analyzed covering a portion of patches 1, 6, and 8. All MOC, THEMIS, and HiRISE images were processed from raw to geometrically calibrated, noise-reduced, mapprojected images using the Integrated Software for Imagers and Spectrometers (ISIS) software developed by the U.S. Geological Survey [Anderson et al., 2004]. Image data from MOC, HiRISE, THEMIS visible spectrometer (VIS) (band 3 centered at micron), and High-Resolution Stereo Camera (HRSC) have been stretched to varying levels so brightness differences described in the text for geologic features are qualitative. [20] THEMIS IR spatial scales are 100 m/pixel while the VIS images are 18 m/pixel resolution. Both THEMIS daytime and nighttime IR images using band 9 at mm [Christensen et al., 2004b] were mosaicked in Photoshop for each study area. Using both the daytime and nighttime THEMIS IR images aided in assessing the qualitative thermal inertia of units. For example, units that are dark in the daytime IR images and bright in the nighttime IR images are likely to have higher thermal inertia, representing rockier or less dusty surfaces Registration of the Data Sets [21] The TES hematite maps for each region, as well as the THEMIS VIS, MOC, and THEMIS IR mosaics were all combined using the software tool Adobe Photoshop and overlain on Mars Express HRSC map-projected images [Neukum et al., 2004]. Mars Global Surveyor Mars Orbital Laser Altimeter 128 ppd gridded topography data were also merged with the image data to study topography associated with hematite occurrence. Because all the data sets have been geometrically projected to the same location in JMARS, ArcGIS and ISIS, they are geometrically registered. However, the OMEGA, and HRSC data did vary in their spatial dimensions relative to the other data sets which means we had to scale these two data sets in order to have them all the same dimensions for coregistration. [22] OMEGA data was georeferenced to THEMIS daytime IR by identifying tiepoints between the THEMIS/IR map and the OMEGA reflectance image (reflectance at 1 mm), followed by a first-order transformation (linear scaling of x, y location, and zoom factor). Visual analysis was performed by blinking between the THEMIS daytime IR mosaic and the OMEGA images and it was concluded that these were coregistered to within the resolution of the OMEGA image (1 km). A similar approach was applied to the registration of HRSC data to the THEMIS daytime IR mosaic. Because the HiRISE images were much greater in spatial dimensions due to their high resolutions, we did not reduce them in size to register them with the other data sets. Instead, we visually identified the locations on the coregistered data sets where the HiRISE images would overlie TES Spectral Mixture Analyses [23] In order to further investigate the mineralogy of sites identified in the hematite maps, spectral mixture analyses of TES data from each location were performed relative to reference libraries of mineral and atmospheric end-members. Unlike the band index approach, spectral mixture analysis does not focus on mapping a single mineral, but instead models all surface components and atmospheric contributions simultaneously relative to a library of reference spectra. Because the thermal infrared spectra of mixed surfaces may be closely modeled using a linear combination of end-member spectra [e.g., Gillespie, 1992; Ramsey and Christensen, 1998], linear modeling algorithms have become a principal approach for mapping the mineralogy of Mars with TES data [e.g., Christensen et al., 2000c; Bandfield et al., 2000a, 2000b]. Extensive laboratory and field measurements support the identification of a wide range of minerals based on the application of these methods to the analysis of thermal infrared data (summaries in Kahle et al. [1993] and Christensen et al. [1992, 2000b]). [24] For this investigation, TES data have been fit using a combinatorial approach known as a multiple end-member spectral mixture analysis or MESMA [e.g., Roberts et al., 11 of 30

12 Figure 18. West Candor Chasma study region. (a) MOLA topography in color, with 500-m contour intervals overlain on a portion of HRSC image h0360_0000. (b) Hematite abundance (not absolute) derived from linear modeling approach shown on same HRSC image as Figure 18a. Boxes show locations of later figures. Yellow numbers refer to hematite patches. 1998; Li and Mustard, 2003; Staid et al., 2004]. This approach allows modeling relative to large spectral libraries, while forcing the solution to use a physically reasonable subset of minerals and atmospheric end-members. The MESMA modeling has previously been applied to a range of TES analyses and laboratory measurements in order to identify mineral components in an objective manner that does not require subsetting reference spectra based on assumptions about the surface being modeled [e.g., Gaddis et al., 2003; Staid et al., 2004; Johnson et al., 2003, 2006, 2007]. For this study, TES data over the spectral range of cm 1 and cm 1 were modeled using a spectral library of 32 minerals and glasses, six atmospheric components and a blackbody as defined by Bandfield [2002, Table 1]. The spectral range used for the modeling includes 73 channels of the TES data, eliminating a region around the atmospheric CO 2 band at 667 cm 1. The algorithm initially compares all possible combinations of four minerals plus the six fixed atmospheric end-members and a blackbody (35,960 possible eleven end-member combinations) to identify the lowest error solution with all positive mineral abundances. This solution then becomes the starting model for the remaining calculations. Additional mineral spectra are then added to this starting model to calculate improvements in the resulting fit to the TES data. The additional mineral spectrum that provides the best improvement (and maintains positive abundances for all mineral components) is then kept as an additional endmember. This procedure is then repeated until a preset maximum of twelve minerals are identified. The model produces fractional abundances of each mineral selected from the reference library along with associated atmospheric and blackbody end-member abundances. [25] The TES data of the Valles Marineris study region were initially modeled at a range of spatial resolutions. The resulting mineral maps demonstrated that less noisy data and lower-error fits were obtained through data averaging, but exposures of high-hematite areas were observed to occur at spatial scales smaller than the averaged data. Because the accuracy of the TES linear modeling can be 12 of 30

13 Figure 19. Steeper slopes along LTR have dark mantle patches (white arrows) that appear competent, with abrupt edges where the wind and mass wasting are eroding the unit. In contrast, dark debris along flatter portions of LTR (bottom of image) are continuous and do not appear to have experienced as much erosion. None of the units shown in this image have detectable hematite. Portion of MOC image M negatively affected by high atmospheric components and data artifacts, the coverage of TES data used for the spectral mixture analyses was limited by several commonly used data constraints (e.g., dust opacity <0.3 and TES orbit number constraints). Data used for the mixture analysis were limited to TES ocks below 7000 based on a reported spectral anomaly that occurs in a portion of later TES acquisitions that can lead to incorrect identification of minerals [e.g., Hamilton et al., 2003]. These data constraints resulted in reduced coverage at some sites relative to the hematite distribution maps (Figures 2a and 2b), compounding problems associated with the use of spatially averaged TES mosaics. A water ice opacity constraint was not used as this further reduced TES coverage of the study sites within Valles Marineris. [26] The TES data was not averaged by the 3 2 detector, but instead by individual detectors based on locations in the hematite index maps and available TES coverage after the dust and ock constraints. This approach was used to try and isolate the signatures of these hematite-rich regions given sparse coverage in the constrained TES data. It was observed that averaging of the entire 3 2 detectors resulted in poor sampling of the hematite-rich regions identified by the index mapping which are small relative to this spatial averaging. On the basis of the available coverage, it was decided that the most accurate models of surface mineralogy would be obtained by averaging several pixels of the full resolution TES data from each site to increase signal to noise. Individual detector measurements were subsequently selected and averaged for each site based on the hematite distributions presented in Figures 2a and 2b. The resulting spectrum of each location was then modeled as described above. Mineral abundances derived from the analysis of each site were then grouped into broad mineral classes. These mineral results as well as model fits to the TES data of each site are summarized in Figure 6. Patches 7 and 8 were not included in the spectral mixture analyses due to the dust and ock restrictions for TES mosaics. [27] Hematite was detected within all of the sites identified by the perc_hem and spectral index mapping methods that were also included in the spectral mixture analyses (i.e., Hematite patches 1 6). Unnormalized abundances determined from the MESMA spectral modeling ranged from 4 to 13% (Figure 6, patches 1 6, first column), consistent with the abundances of 2 15% observed for these deposits by Christensen et al. [2001]. Forcing the mineral components at each site to total 100% of the surface resulted in normalized abundances ranging from 7% to 35% (Figure 6, patches 1 6, second column). Minerals detected at small abundances have the lowest associated confidence due to their minor contribution to the overall fit of the TES data. Christensen et al. [2001] describes an estimated TES detection limit of several percent for hematite based on measurement uncertainties and considerations of spectral band depth. However, detection limits for the implemented modeling approach and spatial averaging are not well constrained. Hematite is most likely to have been correctly identified at sites associated with the highest unnormalized abundances of the mineral. [28] In addition to hematite, other major mineral groups were primarily composed of pyroxenes, feldspars, and, at somewhat lower abundances, sulfates, sheet silicates, and glasses. The results are consistent with a basaltic component in association with the hematite in Valles Marineris. Spectral mixture analysis results for patch 6 differed from the majority of other sites by lacking a significant pyroxene component. [29] The spectral unmixing of the TES data also provides a model of the resulting surface spectrum after removal of atmospheric components. The average modeled surface for the Valles Marineris hematite sites is plotted in Figure 7 relative to surface types 1 and 2 [Bandfield et al., 2000a]. This plot has been formatted for comparison to the spectral units presented by Glotch and Christensen [2005]. The atmospherically corrected Valles Marineris spectrum is similar to the Aram Chaos hematite rich unit presented by Glotch and Christensen [2005] and differences from ST1 and ST2 are consistent with the presence of significant hematite at these sites. 3. Description of the Eight Hematite Patches 3.1. Ophir Chasma Region Overview [30] The Ophir region studied in this paper is shown in Figure 8. There are three main geologic units in our Ophir study region that we have identified based only on the visible 13 of 30

14 Figure 20. (a) THEMIS mm daytime IR mosaic and (b) THEMIS mm nighttime IR mosaic for West Candor Chasma study area, shown in Figure 18. Hematite regions are shown by the letter H. The flat portions of Candor Mensa that have a dark debris unit appear coolest in the nighttime IR mosaic, suggesting a low thermal inertia consistent with fine-grained, loosely packed material, such as fine sand or dust. images: dark debris, LTR, and wall rock. Dark debris is generally confined to the basin but a patch of dark ripples can also be seen to the west overlying LTR. To the south of the basin is wall rock, while to the east, west, and north is LTR. Images along the floor indicate that there is some topographic variation, especially to the south where hills and mounds are evident. These topographic differences may reflect underlying topography or modification of material by wind erosion and mass wasting. [31] The hematite signatures mapped by Christensen et al. [2001] are confined to the dark debris within the basin (Figure 2c, patch 1) and two smaller patches to the west (Figure 2c, outside patch 1). Examination of TES data of these two small patches reveals no evidence for hematite in them (Figures 2a and 2b). However, we did verify the presence of hematite in the basin using the linear modeling approach (Figure 8b), and it has a similar distribution to that mapped by Christensen et al. [2001] (Figure 2c). The highest hematite abundances occur in the eastern portion of the basin, consistent with the mapping by Christensen et al. [2001]. [32] The LTR to the north of the basin are heavily eroded, with cross-cutting linear striations along much of the surface. We interpret these linear features to be yardangs produced by wind blowing downslope along the LTR. Similar yardangs have also been noted by others elsewhere in Valles Marineris in association with the interior layered deposits [e.g., Lucchitta et al., 1992]. Some lineations could also mark bedding planes that are being etched by the wind (Figure 9). The yardangs and paucity of impact craters on the LTR indicates they are friable and easy to erode by the wind. A similar observation was made of the LTR in Juventae Chasma [Catling et al., 2006], where calculated thermal inertia values of Jm 2 s 1/2 K 1 for Juventae LTR were found to be consistent with sedimentary rock covered by some sand or dust. In contrast, the wall rock does not show evidence of yardangs and has been interpreted to be more consolidated basalt units [e.g., McEwen et al., 1999]. [33] Small patches of LTR can be seen poking through dark debris in the basin (Figure 10), indicating that the LTR may be more extensive but are buried. It is possible that the 14 of 30

15 Figure 21. (a) Example of hematite-bearing unit in patch 2. The hematite corresponds to both the darker and brighter units in MOC image M (b) Hematite is located on the layered LTR in the right of this image. There is a thin layer of dark debris that mantles the LTR throughout the image. Dust devils are also visible in the image (white arrows). Portion of MOC image E LTR may have once extended all the way to the wall rock in the south but over time the wind has eroded it away in the south, leaving behind what is now the basin. Where the eroded portions of LTR have gone is subject to debate. One possibility is that the LTR erode into fine-grained dust-size particles that can be carried away in suspension by the wind. By analogy, the absence of obvious accumulations of fines derived from the outcrops at the Opportunity landing site [Yen et al., 2005] suggests that the sulfate-rich layered rocks are easily eroded by the wind into fine dust-size particles that can be transported by eolian suspension. Hence similar erosion of the LTR in Ophir may explain the formation of the basin where LTR had once existed in the past. This could be an important observation if the hematite represents a lag deposit derived from the erosion and removal of the softer light-toned layered rocks. At Meridiani Planum, the hematite-rich spherules make up only a few volume percent of the rocks [Squyres et al., 2004; McLennan et al., 2005] but erosion of the sulfate-rich rocks allows the more resistant hematite-rich spherules to be concentrated along the surface in large enough abundances to be identified from TES. Thus if the hematite patches seen in Ophir represent the same concentration of hematite-rich grains as those found at Meridiani Planum, then erosion of LTR would be a potential way to concentrate hematite in the Ophir basin. Further discussion of the formation and distribution of hematite within Valles Marineris occurs in section 4.1. [34] The LTR in the east have some yardangs but much of the upper surface appears more subdued in the images, perhaps mantled (Figure 11). TES nighttime bolometric thermal inertias of the LTR here are lowest where imagery suggests a mantle exists [Chojnacki et al., 2006]. The mantling unit appears relatively brighter compared to the dark debris seen elsewhere. A preserved impact crater and abrupt edges along the light-toned mantling unit (Figure 11) suggest that it has strength. Ripples can be seen within some of the light-toned mantling unit, but the MOC resolution is insufficient to determine if the ripples are depositional features from eolian processes or the result of erosion of the mantle by the wind. [35] Because the LTR are relatively brighter than the adjacent basaltic plains, they cannot be the source for the dark debris seen along the basin floor. While there may be debris from the LTR that is mixed into the dark debris, the source of the dark debris must be either the wall rock to the south or some other dark-toned unit within the region. The wall rock unit has the typical spur and gully morphology seen throughout Valles Marineris [Lucchitta et al., 1992], and is interpreted to be stacks of lava flows [McEwen et al., 1999]. Erosion and breakdown of these lava flows by wind and impacts would produce sand-size and smaller basalt grains that could explain the dark debris in the basin and along the LTR in the southwest. Further discussion of the dark debris source and its relationship to hematite will be covered in section 4.2. [36] Both MOC and THEMIS VIS images of the wall rock to the south of the basin show LTR along the darker wall rock (Figure 12). Unfortunately, the contact between the LTR and wall rock is obscured by a dark mantle. In other occurrences in Valles Marineris of light-toned material in contact with wall rock, there has been controversy over which unit is older. Malin and Edgett [2000] and Catling et al. [2006] interpret the exposures of LTR to be weathering out of the wall rock, and consequently represent an older 15 of 30

16 Figure 22. Hematite in patch 3 is found on both the dark debris unit and the LTR in this image. Portions of the dark debris unit are rippled, and light-toned material can be seen in some of the ripple troughs (bold arrows), suggesting that LTR may underlie the dark debris at this location. Portion of mosaicked MOC images M and E unit laid down before the lava plains that compose the wall rock. Lucchitta [2007] proposed that the LTR are superimposed on the wall rock and therefore represent a younger unit. In Figure 12, the dark mantle covers so much of the wall rock and LTR that a definitive stratigraphic relationship is impossible to decipher. In either case, the LTR could represent a similar unit to the LTR seen elsewhere in Ophir and would further support the hypothesis that the unit was once more extensive Patch 1 Hematite [37] MOC, THEMIS, and HiRISE visible images show that the hematite is confined to the basin floor and is associated with a relatively dark debris unit. The typical terrain where the hematite has been detected by TES in the eastern portion of the basin is shown in Figure 13. The terrain consists of small patches of light-toned material that is buried beneath darker debris. The patches of light-toned material are too small in area to account for the hematite, so the hematite must reside primarily in the dark debris. The dark debris unit is a mixture of light-toned ripples and smooth dark-toned debris. Small craters are visible in the dark debris unit, indicating that it may not have been active in the recent past. Portions of the dark debris unit appear etched, with jagged edges representing locations where the unit appears to be eroding away (Figure 13), presumably from the wind. The sharp edges where this erosion has occurred indicate the unit has strength at these locations. [38] The hematite in the western portion of the basin occurs in association with similar bright ripples in a dark debris unit. There are no light-toned exposures, presumably the underlying LTR, visible in these images as there were to the east. The southernmost edge of the dark debris in the western portion of the basin has a sharp contact with the adjacent hummocky LTR unit (Figure 14). This LTR unit does not have the same yardangs characteristic of the LTR to the north and it appears relatively darker. There appears to be a sharp break in topography along the southernmost edge of the dark debris unit with a valley filled with bright ripples just to the south. There are no tributaries or channels associated with this valley so it cannot be determined if it formed by water or some other processes. The sharp break and cliff along the southernmost edge of the dark debris unit indicates that it is coherent. [39] Another dark debris unit composed of ripples is located on the LTR to the west (Figure 15a) but has no corresponding hematite signature in TES spectra. In other locations, debris can be seen shedding down the spurs of LTR (Figure 15b), suggesting material is either derived from erosion of the LTR or by deposition of material farther upslope. The debris aprons associated with this downslope transport of material do not make contact with the dark debris in the basin, suggesting that they may not be the source for the dark debris and hematite within the basin floor. The debris aprons also appear brighter than the debris on the basin floor and could therefore contain dust or be covered by dust. [40] Dark debris and ripples in the southeast portion of the basin and on the LTR to the east appear to be the result of erosion of a dark mantling unit possessing strength (Figure 16). The dark mantling unit covers both the LTR and the wall rock, indicating it is one of the relatively youngest units in the region. If this dark mantling unit were once more extensive, then erosion of the unit could be the source of the dark debris in the basin. The source of the dark mantling unit remains unknown, but it has been identified elsewhere on Mars in association with other LTR [e.g., Edgett, 2002]. [41] The dark debris in the basin is relatively warmer in the daytime THEMIS IR mosaic (Figure 17a) relative to the LTR to the north and west. It appears similar in daytime brightness to the light-toned unit to the east, although this unit appears relatively colder in the nighttime THEMIS IR (Figure 17b). In the nighttime THEMIS IR mosaic, the eastern portion of the basin shows more heterogeneity than the western portion of the basin. Brighter warmer and colder darker patches occur in the basin, with the warmer patches generally corresponding to outcrops of light-toned material. TES nighttime bolometric thermal inertias determined by Jakosky and Mellon [2001] and Chojnacki et al. [2006] for material in the basin have values of Jm 2 s 1/2 K 1 consistent with mixtures of coarse sand, larger particles, or strongly crusted fines [Presley and Christensen, 1997]. 16 of 30

17 Figure 23. Hematite in patch 4 corresponds to a darktoned debris unit that has bright-toned ripples. A few bright patches, particularly along crater rims (bold arrows), suggest that LTR lie just below the dark debris in this region. Portion of MOC image R Western Candor Chasma Region Overview [42] The West Candor Chasma region studied in this paper is shown in Figure 18. The wall rock is confined to the north of our study area and the LTR dominates the central area. The LTR is a portion of Candor Mensa, which rises to about 5 km above the adjacent canyon floor. The wall rock is the southern exposure of the same wall rock studied in the Ophir patch 1 hematite and it has a similar morphology. Patches of light-toned material exposed along the northern floor and along some of the wall rock suggest LTR in Candor Chasma may extend across much of the trough floor beneath darker debris. [43] Most of the TES pixels with high hematite concentrations correspond to dark debris units (Figure 18b). Because the dark debris unit encompasses a much larger area than the same unit at Ophir, there is a greater degree of morphology seen within this unit at West Candor Chasma. Some dark debris is visible both on flat-lying portions of LTR, as well as steeper slopes along the LTR. Along the flat-lying portions of the LTR, the dark debris has a ripple texture. In contrast, dark debris on the steeper slopes of LTR appears to be a competent unit with preserved small impact craters and jagged edges where it is being eroded by the wind (Figure 19). The superposition of dark debris on the LTR in these locations, particularly at the top of Candor Mensa, indicates it is a relatively younger unit. [44] From Chojnacki et al. [2006], thermal inertias range from about 50 to 200 Jm 2 K 1 s 0.5 on flat-lying portions of Candor Mensa, compared to about Jm 2 K 1 s 0.5 along the floor to the north. The nighttime THEMIS IR mosaic (Figure 20b) supports a lower thermal inertia on the flat-lying portions of Candor Mensa. All the hematite locations (Figure 20a) correspond to higher thermal inertia locations, with values ranging from approximately 300 to 500 Jm 2 s 1/2 K 1 [Chojnacki et al., 2006]. The lower thermal inertias corresponding to dark debris on flat-lying portions of Candor Mensa are consistent with fine-grained, loosely packed material, such as fine sand or dust. [45] The hematite signatures mapped by Christensen et al. [2001] (Figure 2c) show four main patches in this study region, all associated with dark debris along the canyon floor. In our TES hematite maps derived from the linear modeling approach for this area (Figure 18b), we also identify the four large patches of hematite, but not in the exact same regions as those shown in Figure 2c. In particular, the hematite patches 2 and 3 in Figure 18b are situated on and adjacent to a portion of the LTR, whereas these two patches are located farther away from the LTR in Christensen et al. [2001] (see Figure 2c). Hematite patches 2, 3, and 5 are distributed unevenly in topography (Figure 18a), with hematite occurring both on the LTR several kilometers above hematite on the adjacent dark debris along the canyon floor. Patch 4 shows all the hematite corresponding to the dark debris unit Patch 2 Hematite [46] The hematite in patch 2 is distributed across dark debris adjacent to and superimposed on LTR. An example of hematite on dark debris is shown in Figure 21a. The hematite corresponds to both a smooth dark unit and a slightly brighter rippled unit. The darker smooth unit has a diffuse boundary and appears to be burying the underlying brighter rippled unit. The dark unit is most likely composed of basalt sand with not as much dust as the underlying rippled unit based upon its dark-toned and diffuse, smooth surface. The rippled surface could also have a component of basalt sand mixed with brighter dust. Small bright patches within the rippled unit, but not composed of ripples, suggest that underneath the brighter rippled unit is a much brighter unit that could be the same LTR seen to the north and south. [47] Hematite is also located on LTR in patch 2. Figure 21b shows extensive layering visible in the LTR, which also corresponds to several hematite pixels. Darker debris covers most of the LTR, although not at a depth that obscures the layering. The MOC image appears to have dust devils in the lower left (Figure 21b), with bright plumes spread to the southwest and dark shadows located to the northeast. There is no discernable geologic unit in either the daytime or nighttime thermal infrared images (Figure 20) that corresponds to the locations of these high hematite abundances Patch 3 Hematite [48] The hematite in patch 3 is located both on and adjacent to LTR. A MOC image showing the units associated with the hematite is shown in Figure 22. LTR with 17 of 30

18 Figure 24. Hematite in patch 5 is located on the dark debris unit and the adjacent LTR in this image. The LTR that correspond to hematite pixels have dark debris mantling much of their surfaces. A few very dark features at the top left of the image (white arrows) are interpreted as dunes. Portion of MOC image M hematite have yardangs and some dark ripples, indicating wind has modified their surfaces. Dark debris on LTR appears to occur exclusively on flatter, smooth surfaces rather than steeper slopes. The contact between LTR and dark debris along the floor is diffuse. The dark debris unit is rippled, with the ripples oriented north-south. Lighter material can be seen in some of the ripple troughs, suggesting that LTR may underlie the dark debris at this location. Moving to the north and away from the LTR, the dark debris appears as unrippled mesas with smaller ripples interspersed. Hematite corresponds to both the rippled and unrippled units. [49] LTR only a few kilometers to the south with a thin mantle of dark debris and dark ripples appear morphologically similar to those shown in Figure 22 yet do not have a corresponding hematite abundance. Much of Candor Mensa has a mantle of dark ripples and dark mesas concentrated along flat-lying portions of LTR. These dark terrains also do not have a hematite signature. These observations suggest the LTR or dark debris shown in Figure 22 are unique at a resolution scale unobservable with current MOC images to explain the hematite signature at this particular location. Even in the daytime and nighttime thermal infrared images (Figure 20) there is no recognizable feature that distinguishes the hematite-rich terrains from nearby morphologically similar units Patch 4 Hematite [50] The hematite in patch 4 is several kilometers away from Candor Mensa and it corresponds to a homogeneous dark debris unit that has widely spaced brighter ripples and numerous small (<100 m diameter) impact craters (Figure 23). A few bright patches suggest that LTR lie just below dark debris in this region. This interpretation is supported by exposures of LTR along slopes a few kilometers to the southwest that have the same surficial dark debris unit, albeit with no corresponding hematite signature. To the north of the hematite-bearing unit, the dark debris appears even darker with a higher number of ripples, and just to the west are landslide deposits. In both the THEMIS daytime and nighttime images (Figure 20), the hematite-rich unit appears more mottled in temperature with patches of warm and cold material. In contrast, the terrain to the south and east appears more homogeneous in temperature, although the difference in temperature between this terrain and those to the north and east is not readily apparent Patch 5 Hematite [51] The hematite in patch 5 is located along the western side of Candor Mensa on the chasma floor and LTR. The location is along a semicircular exposure of steep-sloped LTR with prevalent yardangs (Figure 18b). Most of the TES hematite pixels correspond to dark debris at the base of the LTR. Figure 24 provides a higher-resolution view of the hematite-rich terrain on both the LTR and dark debris. The dark debris consists of linear ripples oriented parallel to the edge of the LTR and consistent with winds blowing upslope or downslope from the LTR. There appears to be a younger darker unit superimposed on the ripples that is visible as darker patches in Figure 24. Dark streaks are seen along the edges of some of this dark patch, suggesting wind blown material. Dark dunes are visible at the top of Figure 24 and could represent basalt sand shedding off the LTR steep slopes and accumulating near the base, although it is not clear why they only appear in this particular location. The relatively dark appearance of the dunes indicates less dust than adjacent surfaces, and they are superimposed on ripples suggesting a younger age and more recent activity. [52] A few small patches of bright material can also be seen in some of the dark ripples, suggesting that LTR underlie the dark debris as an extension of the Candor Mensa LTR. The LTR that correspond to hematite pixels have dark debris mantling much of their surfaces, particularly regions that appear flat lying. Interestingly, although the dark debris continues to the west, the hematite is only concentrated in the dark debris adjacent to the edge of the LTR, perhaps because the hematite grains are too large or heavy for winds to transport them westward. In both the daytime and nighttime THEMIS IR images (Figure 20), there is no distinguishable unit that corresponds to only the hematite locations Central Candor Chasma Region Overview [53] Our study area encompasses two hematite patches (6 and 7) located on elevated LTR (Figure 25). Dark debris mantles some of the LTR, especially in flat-lying areas. 18 of 30

19 Figure 25. Central Candor Chasma study region. (a) MOLA topography in color, with 500-m contour intervals overlain on a portion of HRSC images h0334_0001 and h0515_0000. (b) Hematite abundance (not absolute) derived from a linear modeling approach shown on same HRSC image as Figure 25a. Boxes show locations of later figures. Yellow numbers refer to hematite patches. Dark dunes and brighter ripples have accumulated on the lower topography of the canyon floor in the west. The hematite does not correspond to these darker debris units in the west. A few exposures of wall rock can be seen in the southern portion of the study area. MOC images show the LTR are layered in some locations but other images do not show layering. Slopes along the LTR can have yardangs indicating slope winds. These same downslope winds, as well as gravity, have carried darker debris down some slopes where it has accumulated into debris aprons along 19 of 30

20 Figure 26. Dark dunes and debris cover much of the floor in the western portion of the study area. Debris can also be seen shedding down the steeper slopes of LTR to produce debris aprons. Hematite is not detectable in any of the units shown in this image. Portion of MOC image E the base of LTR, particularly in the west (Figure 26). The debris shedding down the slopes of the LTR are relatively brighter than the dark dunes along the floor, and neither unit corresponds to hematite. This difference in brightness may be due to more dust or a fine-grained component of LTR mixed in, thereby increasing the brightness relative to the dark sand along the floor Patch 6 Hematite [54] Although the hematite patch appears to be two distinct patches in Figure 25b, TES spectra do reveal one barely continuous patch of hematite, similar to the hematite distribution map in Christensen et al. [2001] (see Figure 2c). The hematite abundance is lower than our 18% abundance cutoff toward the center of patch 6, though, and our spectral analysis shows that the continuous portion with qualitative abundances <18% is limited to the southerly dark zone. The hematite in patch 6 correlates to dark debris in a low-lying region along the LTR. Elevations for the hematite shown in Figure 25a range from 1500 to 4500 m, indicating that it is not associated with one constant elevation. [55] Examples of hematite terrain in the western portion of patch 6 are shown in Figure 27. Both images show dark debris corresponding to the hematite TES pixels, with the terrain in Figure 27a having a slightly higher abundance of hematite than that seen farther to the west in Figure 27b. The dark debris in Figure 27a appears smooth with few ripples. The dark debris appears heterogeneous in brightness with patches of light-toned material mixed with patches of dark-toned terrain (Figure 27a). Impact craters have bright rims and dark centers. The brighter terrain does not appear to be exposures of LTR based upon its smooth surface, moderate brightness, and diffuse boundaries with the darker debris. Farther to the west, the terrain is more rippled (Figure 27b) with what appear to be small exposures of LTR. There are currently no MOC images that cover the central portion of patch 6 but THEMIS VIS and HRSC images all suggest that the terrain has more ripples than to the east and west where the hematite abundance is greater. [56] There is currently one HiRISE image (PSP_002142_ 1730) that covers the highest hematite abundance in patch 6. At a spatial scale of approximately 30 cm/pixel, the HiRISE image (Figure 28) provides details about the hematitebearing unit not resolvable in the MOC images. One observation concerns the stratigraphic relationship between the dark debris and the LTR. In Figure 28b, outcrops of LTR are clearly embayed by the dark debris unit and the sharp edge of the dark debris in several locations is consistent with a competent unit rather than loose material. Numerous small craters are also visible in the dark debris unit, also supporting the idea of a unit possessing strength that is not mobile. The thickness of the dark debris unit appears to be at least a few pixels (1 2 m) based upon where it has a sharp edge at the contact with the LTR. While ripples dominate in the southern portion of the HiRISE image, there are almost none to the north. This observation indicates that the hematite corresponds to the surface unit regardless of whether there are any ripples, an observation also noted at the Opportunity landing site where the hematite-rich spherules can be seen on both smooth unrippled plains and also on ripples [Weitz et al., 2006a]. [57] Using the TES nighttime bolometric thermal inertia mosaic from Chojnacki et al. [2006], the hematite regions correspond to higher thermal inertia values than adjacent terrains, with values approximately between 300 and 600 Jm 2 K 1 s 0.5. In the daytime and nighttime THEMIS IR mosaics (Figure 29), the highest hematite abundances appear to correlate to units that are slightly IR warmer. The central portion of patch 6 that falls below 18% qualitative hematite abundance is cooler in both daytime and nighttime THEMIS IR mosaics Patch 7 Hematite [58] This small patch of hematite is located between 3300 and 3000 km elevation along a slope of LTR. In Figure 25b, only four hematite pixels define this patch. However, examination of TES spectra for adjacent pixels as well as the hematite map produced by the spectral index approach (Figure 2b) reveal that the hematite is a larger continuous patch than shown in Figure 25b. The hematite pixels correspond to smooth dark debris that is situated on LTR. A MOC image that shows the hematite locations for the two western TES pixels is shown in Figure 30a. Bright patches in the dark unit appear to be exposures of underlying LTR. Small impact craters in the dark terrain have bright rims and dark floors. [59] The terrain just to the west that has no hematite is slightly brighter and has ripples. To the south of the hematite are steeper exposures of LTR with yardangs. To the north is a 6.3 km diameter impact crater that has layering exposed 20 of 30

21 Figure 27. Hematite-bearing units in the western portion of patch 6. (a) Hematite occurs throughout this image in both the light- and dark-toned units. Portion of MOC image M (b) The terrain has more ripples than in Figure 27a, with what appear to be small exposures of LTR within the ripples. Portion of MOC image M along the southern wall and has dark debris and ripples on the crater floor. The dark debris inside the crater appears relatively darker than the dark debris associated with the hematite location to the south (Figure 30b). [60] The hematite pixels correspond to a warmer region in both the daytime and nighttime THEMIS IR images (Figure 29), similar to that found for patch 6 hematite. The thermal inertia values taken from Chojnacki et al. [2006] are approximately around Jm 2 K 1 s 0.5, consistent with loose particles, indurated fines, and some bedrock or rock exposures East Candor Chasma Region Overview [61] The one hematite patch in this location occurs on the floor of East Candor, adjacent to layered LTR and near an exposure of wall rock detached from the northern wall rock that defines Candor Chasma (Figure 31). The LTR represents a thick sequence of sedimentary rocks that fill up the central portion of Candor Chasma. Although the LTR unit extends to the wall rock in the south, it is separated from the wall rock in the north by relatively flat canyon floor (Figure 31a). Based upon analyses of MOC images, the canyon floor in our study area is covered by a mantle of dark debris, some of which forms resistant mesas. There are few ripples and dunes associated with the dark debris unit along the floor. Thermal inertia values from Chojnacki et al. [2006] indicate high values along the northern canyon floor of Candor Chasma, consistent with coarse sand, larger particle sizes, and strongly crusted fines. In contrast, the LTR unit has lower thermal inertia values, especially on elevated, flat-lying surfaces. The LTR in the lower portion of our study area (Figure 31b) vary in morphology from smooth to heavily fractured and rough surfaces. Yardangs are visible along slopes, especially near the contact with the canyon floor. One area in the LTR shows sinuous valleys that appear distinct from the linear yardangs (Figure 32). The valleys appear to have flat floors. One valley can be seen intersecting a larger valley. The valleys could have formed by wind, mass wasting or water, although there is no evidence for the source of water to have carved these valleys Patch 8 Hematite [62] The hematite is situated along the canyon floor, adjacent to LTR (Figure 31b). This portion of the canyon floor that includes hematite and nonhematite units is complex with mixtures of dark debris, LTR, dunes, and ripples. Figure 33 shows a portion of HiRISE image PSP_002432_1730 of the terrain that corresponds to locations of TES hematite pixels. The hematite pixels are situated on LTR and smooth 21 of 30

22 Figure 28. (a) HiRISE image PSP_002142_1730, which corresponds to the highest abundances of hematite in patch 6. Hematite occurs throughout this image in association with dark debris and slightly brighter rippled terrain. The black box indicates location of Figure 28b. (b) Blowup showing outcrops of LTR that are partially buried beneath the dark debris unit. Edges along the dark debris unit (bold arrows) are sharp and appear to have topographic expression, consistent with a competent unit. dark terrain. The darker unit in the top of the image does not have hematite. It appears smooth with no resolvable impact craters and a few brighter ripples. The LTR are an underlying unit that is exposed in certain locations. Dunes with polygonal shapes cover much of an exposure of LTR in Figure 33. The crests of these dunes appear bright and define the shape of the polygons. More dunes and ripples can be seen to the east. The dunes are not as dark as the unit to the north, suggesting that either the dunes are not composed exclusively of basalt sand or that their surfaces have been covered by brighter fines, such as dust. [63] The smooth unit shown in Figure 33 corresponds to the center hematite pixel in Figure 31b and it extends to the west and east where there are other TES hematite pixels. Small bright outcrops are exposed, suggesting LTR underlie the unit. Small impact craters are visible throughout the unit, indicating that the unit is competent to preserve the crater outlines or, alternatively, there could be a meters-thick mantle of loose grains overlying competent LTR. Bright rims outline many of the impact craters, which also suggests that LTR lie only a few meters beneath the smooth hematitebearing unit. All the craters visible in the HiRISE image have dark floors relative to the adjacent plains. Darker crater floors are also noticeable in images at the Opportunity landing site and result from basalt sands and dunes that become trapped in the craters [Sullivan et al., 2005]. Ripples are only seen toward the edges of the unit. [64] Figure 34 shows an outcrop of several dozen layers that are tilted nearly vertical and now partially covered by dark debris and ripples. The western edge of the smooth hematite-bearing unit appears to be covering an underlying light-toned rippled unit. Hematite pixels do not extend on to the light-toned rippled unit. The strong correlation between TES hematite pixels and the smooth unit indicates that most of the hematite signature can be tied to the smooth unit. [65] In the daytime and nighttime THEMIS IR images, there is a strong correlation between the hematite-bearing unit and thermally distinct units seen in the IR images. The hematite corresponds to a relatively warm unit in the daytime IR mosaic (Figure 35a) and a relatively cool unit in the nighttime mosaic (Figure 35b), which suggests that the unit has a lower thermal inertia than the surrounding terrain. The patches of LTR that are to the north of the hematite along the canyon floor appear relatively cool in the daytime IR mosaic and relatively warm in the nighttime IR mosaic, indicating higher thermal inertia values than the smooth unit. Interestingly, the LTR to the south that occupy much of east Candor Chasma appear cool in the nighttime IR mosaic, except along steep slopes where yardangs are present. The cooler nighttime IR temperatures for these LTR relative to the small exposures on the floor suggest either that they are composed of different materials or that they are mantled by fine debris. The dark debris that is located to the west and northeast of the smooth hematite-bearing unit appears slightly cooler in the daytime IR mosaic but similar in temperature in the nighttime IR mosaic. 4. Discussion 4.1. Hematite Formation [66] The locations of hematite in Ophir and Candor chasmata strongly correlate to relatively dark debris units 22 of 30

23 Figure 29. (a) THEMIS mm daytime IR mosaic and (b) THEMIS mm nighttime IR mosaic for Central Candor Chasma study area. Hematite regions are shown by the letter H. that are either superimposed on or adjacent to LTR. There are numerous locations of this dark debris unit within Valles Marineris that do not have detectable hematite abundances, which indicates that the darker debris grains are not the source of the hematite but rather the hematite comes from another source that is mixed into the dark debris. [67] Figure 36 shows examples of the typical soils at Meridiani Planum that correspond to the hematite signature as measured both by TES from orbit and from Mini-TES on Opportunity. A HiRISE subimage that covers the typical terrain observed from orbit for the hematite-bearing plains at the Opportunity landing site is shown in Figure 37. The source of the hematite spherules at Meridiani Planum is a thin layer of LTR within a much thicker stack of sediments [Christensen et al., 2000a, 2001; Hynek et al., 2002]. There is still a debate as to the origin of the sulfate-rich LTR at Meridiani Planum and the hematite-rich spherules embedded within them. Origins for the exposed LTR at the landing site include impact surge deposits [Knauth et al., 2005] and a basaltic pyroclastic deposit subsequently altered through reaction with an aqueous sulphuric acid solution [McCollom and Hynek, 2005], although the geochemistry of the rocks and the identification of ripple bed forms in the rocks appear to refute these two origins [Squyres et al., 2006; Grotzinger, 2007]. The MER science team has demonstrated the outcrops at the landing site formed by erosion and redistribution of sand grains that originally were derived by chemical weathering of olivine basalt in aqueous solutions of sulfuric acid, forming sulfate salts that accumulated with finegrained silicates [Squyres et al., 2004]. After emplacement of these sulfate-rich sands, there was interaction with substantial amounts of groundwater. [68] The spherules could represent concretions formed diagenetically in migrating groundwaters that saturated the sands, perhaps as one of the last influxes [Squyres et al., 2006; McLennan et al., 2005]. Alternative origins for the spherules include impact spherules and accretionary lapilli [Knauth et al., 2005], but the Fe/Ni levels and the lack of spherules concentrated along bedding planes argues against these origins [Grotzinger, 2007]. Regardless of the origin of the hematite-rich spherules, erosion of the friable outcrops has allowed the more resistant spherules to be concentrated along the top of the soils. Perhaps only a meter or so of erosion of overlying outcrop was necessary to produce the volume of spherules seen at the landing site based upon the volume of spherules measured in the outcrop [Squyres et al., 2004]. [69] We argue that the most likely explanation for the hematite concentrations in Ophir and Candor chasmata are due to the same hematite-rich grains like those found at the Opportunity landing site for the following reasons: (1) the Valles Marineris hematite spectral emissivity signature is the same as that from both Meridiani Planum and Aram Chaos [Christensen et al., 2000a, 2001; Glotch et al., 2004] and it requires the hematite to be coarse grained and gray formed by low-temperature dehydroxylation of goethite [Glotch et al., 2004]; (2) from MOC and HiRISE images, the hematite-bearing dark plains at the Opportunity landing site look morphologically similar to the hematite-bearing dark debris units in Ophir and Candor; and (3) there are similar LTR at both locations that could be the source of these hematite-rich grains. If so, and if the hematite-rich grains represent concretions formed in water like those postulated for the Opportunity landing site, then water must have infiltrated some locations in Valles Marineris as well. [70] In the case of Aram Chaos, the highest concentrations of hematite (16%) determined by Glotch and Christensen [2005] correspond to one or more layers of a 100- to 150-mthick layer stack that is stratigraphically above lower hematite abundance units. This is not the case in either Ophir or Candor chasmata where the highest concentrations of hematite correspond to lower-lying topography. There could be similar layers that are the source for the hematite in Ophir and Candor, but they are not exposed at large enough spatial scales for TES to resolve them. [71] At both Meridiani Planum and Aram Chaos, observations support an aqueous origin for the hematite [Christensen et al., 2000a, 2001; Squyres et al., 2004; McLennan et al., 2005; Glotch and Christensen, 2005; Noe Dobrea et al., 23 of 30

24 Figure 30. MOC and THEMIS VIS images of patch 7 hematite in Central Candor Chasma. (a) Portion of MOC image M showing that the hematite (H) corresponds to a dark-toned unit. (b) Although dark debris (dd) is visible on the floor of the impact crater, at the top of the image, there is no hematite. Instead, the hematite (H) is located in slightly brighter terrain in this portion of THEMIS VIS image V ]. We favor a similar aqueous origin for the hematite in Ophir and Candor chasmata for the following reasons: the spectral similarity (i.e., band positions and spectral shape) in TES of the hematite in Ophir and Candor to that at Meridiani Planum and Aram Chaos; the identification of other minerals, including kieserite and polyhydrated sulfates, in association with LTR at Ophir, Candor, Meridiani Planum, and Aram Chaos [Gendrin et al., 2005]; the morphologic similarity, as determined from orbital visible imagery, between the hematite-bearing units at Ophir and Candor to those at Meridiani Planum; and outflow channels associated with several of the Valles Marineris chasmata, as well as at Aram Chaos, that suggest water flow emanating from the canyons. None of the alternative nonaqueous origins for coarse-grained gray hematite are consistent with all the observations we have made at Ophir and Candor chasmata. [72] Four possible sources for this water include: (1) a rising water table that caused sulfate-rich rocks (i.e., LTR) to produce hematite-rich concretions, as described by McLennan et al. [2005] for Meridiani Planum; (2) precipitation that infiltrated LTR and produced concretions; (3) melting of ground ice that then enabled water to produce hematite-rich concretions; and (4) lakes that may have once existed within parts of Valles Marineris. The hypothesis of a rising water table appears to be the best explanation for how the concretions formed at the Opportunity landing site, where the rocks record evidence in their morphologies that support infiltration of water into eolian sand dunes [Grotzinger et al., 2005]. Similar infiltration of water from below may also explain the hematite-rich grains at Ophir and Candor. [73] Ground ice is currently unstable at the equator of Mars but during different obliquity cycles it could have been stable on the LTR units in Valles Marineris [Mellon and Jakosky, 1995]. In order to produce the hematite-rich concretions, the ice would need to subsequently melt and then infiltrate the rocks in order to form hematite-rich concretions in the LTR. In Valles Marineris, volcanism from the nearby Tharsis volcanoes could have produced water by melting surface or subsurface ice that then flowed into the canyons. This water could have either risen into the LTR from below as the water table fluctuated with each pulse of water released from an eruption, or along the sides of the LTR if they were situated in a lake. Volcanic melting of ice just south of Coprates Chasma has been proposed by Weitz et al. [2006b] to explain the valley and associated terraced lobate fan deposit, suggesting that ice may have been at or near the surface in the past at the latitude of Valles Marineris. In addition, observations of fractures in Candor Chasma LTR reveal lighter haloes that are interpreted to result from circulating fluids that hardened the lining of the fractures [Okubo and McEwen, 2007]. Water flowing through these fractures could have produced these haloes and would be further evidence for water influxes into the LTR that could have also produced hematite grains. [74] It is possible that precipitation may be a source for water in Valles Marineris. Observations in favor of precipitation include a region in southwestern Melas Chasma where valley networks are dense and highly organized in support of precipitation during the Hesperian [Quantin et al., 2005]. Whether or not sufficient precipitation could have occurred throughout the canyons to produce enough 24 of 30

25 Figure 31. Eastern Candor Chasma study region. (a) MOLA topography in color, with 500-m contour intervals overlain on a portion of HRSC image h0515_0000. (b) Hematite abundance (not absolute) shown on same HRSC image as Figure 31a. Boxes show locations of later figures. 25 of 30

26 transport the grains upslope to higher elevations where it is seen today. This seems difficult to accomplish given the steep slopes and rugged terrain along the LTR. Alternatively, the sand may have been emplaced via saltation on the surfaces of former LTR of uniform topography, but subsequent erosion of the LTR has left the dark debris on top of only those areas that have not yet been eroded. The occurrence of dark mantles along some of the LTR as well Figure 32. Valleys (white and bold arrows) carve through the LTR, but they are more likely to result from wind erosion along weak zones in the rocks, rather than water flow. Portion of MOC image M water that infiltrated the LTR and formed the concretions remains unknown and speculative. [75] Had a lake once existed in Ophir or Candor chasmata, then lake water could have infiltrated the LTR and produced hematite. Because hematite is not found along all the potential low-lying canyon floors that show outcrops of LTR argues either against lake water as the means to produce the hematite, or suggests that the process whereby hematite was produced in LTR was restrictive perhaps because of compositional variations in the LTR that prevented all the LTR from forming hematite even when water was able to infiltrate the rocks. The same argument could be used to justify why not all LTR produced hematite if the water table rose throughout Valles Marineris. Clearly, hematite formation and subsequent concentration in high enough abundances to be detected by TES all factor in the current distribution of hematite within Valles Marineris Source of Hematite-Bearing Dark Debris [76] In both Ophir and Candor chasmata, there is no apparent single layer of dark rocks in the LTR that could be the source for the dark mafic debris mixed in with the hematite. More likely, the dark debris was deposited on LTR surfaces by eolian activity. If the dark debris is sand size, then saltation would have been required in the past to Figure 33. Portion of HiRISE image PSP_002432_1730 showing different geologic units in East Candor Chasma, including the hematite-bearing unit. 26 of 30

27 supply of hematite grains that would concentrate in lowlying regions. The observation that most canyon floors and low spots along the LTR do not have associated hematite (or have abundances below the minimum threshold we used in our detections) suggests that not all the LTR may contain abundant hematite grains. This means that any water that flowed through the LTR to produce hematite concretions would need to be limited in spatial extent, rather than interacting with all the rocks. It could also imply that water was not able to convert goethite to hematite in all LTR, which is a possibility given that we do not know the exact composition of the LTR in Valles Marineris and they could Figure 34. Smooth dark hematite-bearing unit surrounded by brighter rippled plains. An exposure of tilted layered deposits (bold arrows) suggests LTR exist underneath. Portion of MOC image E as the wall rock supports induration of dark debris. Erosion of these dark mantles can also have associated ripples (see Figure 16) that would support the dark mantles are the source of sand-size and finer debris seen throughout the canyon system. Such dark debris and dark mantles are not unique to Valles Marineris and have been described by others elsewhere on Mars [e.g., Edgett, 2002, 2005]. [77] The source for the basaltic sand at the Opportunity landing site also remains an enigma. There has been no basalt bedrock found by the rover yet that could weather to form sand grains. The most likely source is basalt plains farther away from the landing site. The sand grains would have eventually saltated to the landing site to produce ripples seen along the plains and dunes seen within crater floors. The wall rock of Valles Marineris is believed to be basalt plains that surround the canyons [McEwen et al., 1999] and would therefore be a plentiful source of basalt sands that compose the dark debris units Why Is Hematite Concentrated in High Abundances in Only Eight Isolated Locations? [78] So why does hematite occur as small patches in isolated locations within Ophir and Candor Chasma, as well as other locations within Valles Marineris, rather than as larger units like those at Aram Chaos and Meridiani Planum? The answer could be a combination of a more limited supply of hematite grains and low-lying terrain upon which they can form a lag deposit. If all the LTR seen within Ophir and Candor contain a few volume percent of hematite-rich grains, similar to what Opportunity images show in the LTR of Meridiani Planum, then there appears to be sufficient erosion of the LTR as evidenced by the yardangs and removal of impact craters to produce an ample Figure 35. (a) THEMIS mm daytime IR mosaic and (b) THEMIS mm nighttime IR mosaic for Eastern Candor Chasma study area. Patch 8 hematite region is outlined by the letters H. 27 of 30

28 Figure 36. (a) Opportunity Navcam Sol 359 image 1N EFF40FWP1612R0M1 illustrating the surface morphology along the plains of Meridiani Planum. Rover tracks can be seen in the soils. (b) MI image of the typical grains that compose the soils seen along the plains at the Meridiani Planum landing site. The larger grains average 1.6 mm in diameter, and most are the hematite-rich spherules that form a lag on the surface soils. The finer grains that are <100 mm in size are dust and basaltic sand. MI Sol 443 image 1M EFF55B0P2956. vary enough in composition that only some rocks would be able to form hematite-rich concretions. [79] The movement of sands and finer dust has enabled the larger millimeter-size hematite-rich spherules at the Opportunity landing site to stabilize on the surface as a lag deposit [Weitz et al., 2006a]. The combination of active winds and grain disparity is required to form the lag deposit of hematite-rich spherules on the surface, as well as desert pavements seen on Earth [e.g., Mabbutt, 1979]. Otherwise, the hematite-rich spherules would get buried by sand and dust, preventing TES from seeing their signature from orbit. [80] A similar scenario may occur in the hematite patches of Ophir and Candor chasmata. There must be winds that blow dust off the hematite grains and move the grains to keep them on the upper surface where they could be detected from orbit. There could be many more locations within Valles Marineris where hematite-rich grains formed in LTR, but either the grains were not able to accumulate in large enough abundances to be detected or the grains have been covered by dust and sands. In either case, the requirement for winds to both concentrate the grains as well as keep the grains free of sand and dust are essential for their detection and could limit the number of hematite locations. [81] The eight patches of hematite that we have analyzed in this paper all occur in relatively lower elevations than adjacent LTR. These low-lying spots would be ideal places to concentrate hematite-rich grains that weather out from the friable LTR. The identification of LTR beneath dark debris units where hematite has been found suggests that the weathering of former overlying or adjacent LTR has removed the LTR and concentrated the more resistant hematite grains. This is the same process that is now occurring at the Opportunity landing site in Meridiani Planum where perhaps a meter of LTR has been removed by the wind and the hematite-rich spherules once embedded in these rocks are now left as a surface lag [Soderblom et al., 2004; Weitz et al., 2006a]. 5. Conclusions [82] All eight of the spatially largest and highest abundances of hematite occur in association with relatively smooth dark units in Ophir and Candor chasmata. The smooth dark units all appear to have underlying LTR and several also occur directly adjacent to higher-standing LTR units. Because of the strong spectral and morphologic similarity between hematite-bearing units seen in data of Meridiani Planum and those within Ophir and Candor chasmata, we favor a similar soil lag of larger hematite grains mixed with basalt sands. The most likely source for the hematite grains is erosion of LTR that were formerly overlying or adjacent to the hematite patches in Ophir and Candor chasmata. In the case of Meridiani Planum, erosion has produced a flat surface over hundreds of kilometers upon which the hematite can be concentrated as a surface lag. The terrain in Ophir and Candor have larger topographic variations associated with the LTR which could have reduced the extent of localities where hematite could Figure 37. HiRISE subimage showing the Opportunity landing site. The lander is the bright spot within Eagle crater. The appearance of the plains at Meridiani Planum from orbital images is similar to many of the eight patches of hematite we analyzed within Ophir and Candor chasmata, suggesting comparable soil lags composed of hematite-rich grains. Portion of HiRISE image PSP_001414_ of 30