We used emissivity spectra from the Thermal Emission Spectrometer (TES) to identify the signature of crystalline gray hematite in Capri Chasma. Geologic units associated with major concentrations of hematite were then mapped using HiRISE, CRISM, and CTX images from the Mars Reconnaissance Orbiter (MRO). Along the northern portion of the Interior Layered Deposit (ILD), a lower polyhydrated sulfate (PHS) unit lies beneath a thicker kieserite unit, above which is a thinner upper PHS. An exposure at the thickest central portion of the ILD reveals additional sulfates, including a middle PHS, a mixture of lower hydration states PHS, and an intercalated unit comprised of kieserite and szomolnokite. We interpret these compositional transitions to reflect either changes in aqueous chemistry (e.g., iron levels and salinity) during groundwater upwelling events or successively buried layers of dust, ice and volcanic aerosols laid down over obliquity cycles. In addition to these sulfates, we identified a few small mounds along the chasma floor composed of either mixtures of ferric hydroxysulfate and Fe/Mg-smectites, or possible opal, leached clays, and Fe/Mg-smectites. Gray hematite is strongly spatially correlated to kieserite-bearing slopes within the ILD and mantled PHS along the northern chasma floor. These results are consistent with sulfate and hematite formation found elsewhere on Mars, including Meridiani Planum, Aram Chaos, and several other chasmata and chaos regions within Valles Marineris.
 Thick, laterally extensive sedimentary sequences recorded on Mars provide clues to the environmental conditions that prevailed when the sediments were laid down. Valles Marineris contains some of the largest exposures of these sedimentary deposits. One of the key questions is the role of water in the deposition of these materials and how the aqueous chemistry changed through time. By identifying distinct mineral phases and determining how these minerals are associated spatially and temporally, we can decipher aqueous processes through time. This paper focuses on the spectral and stratigraphic analysis of aqueous minerals in Capri Chasma, an open depression in eastern Valles Marineris that contains a diversity of minerals and sedimentary rocks.
 The light-toned outcrops observed at Meridiani Planum are sulfate-rich [Squyres et al., 2004; McLennan et al., 2005], with MiniTES data indicating the presence of both Ca- and Mg-rich sulfates [Christensen et al., 2004; Glotch et al., 2006]. The hematite spherules are also found embedded in the sulfate outcrops throughout the landing site, indicated they precipitated within these rocks [Squyres et al., 2004; Christensen et al., 2004]. The sulfates are thought to have originated as saturated sands derived from the chemical weathering of olivine basalt in aqueous solutions of sulfuric acid, forming sulfate salts that accumulated with fine grained silicates [Squyres et al., 2004]. After emplacement of these sulfate-rich sands, there was interaction with substantial amounts of groundwater as the water table fluctuated [Arvidson et al., 2006]. The hematite spherules are interpreted as concretions formed diagenetically in migrating groundwaters that saturated sands, perhaps during one of the later influxes [Glotch et al., 2006; McLennan et al., 2005]. Alternatively, Niles and Michalski  propose acid weathering inside massive ice deposits during periods of high obliquity as an explanation for the sulfates at Meridiani Planum. Erosion of the friable sulfate outcrops over time has allowed the more resistant spherules to be concentrated along the top of the soils as lag deposits, with perhaps only a meter or so of erosion of overlying outcrop necessary to produce the volume of spherules seen at the landing site based upon the density of spherules measured in the outcrop [Squyres et al., 2004; Soderblom et al., 2004].
 The size of the hematite spherules ranges from about 1 to 4.5 mm [Weitz et al., 2006; Calvin et al., 2008]. Within the outcrop, the spherules make up only a few volume percent of the rocks but their concentration as a lag deposit on the surface of Meridiani Planum has allowed their areal abundance to be high enough to be detectable from orbit. Hence, other sulfate-rich rocks with hematite spherules may exist on Mars besides those already identified by TES, but the hematite spherules are not concentrated at high enough abundances along the surface soils.
 Hematite-rich units in Meridiani Planum, Aram Chaos, Aureum Chaos, and Iani Chaos are all found in association with exposures of sulfate-bearing rocks [e.g., Christensen et al., 2001; Catling and Moore, 2003; Glotch and Christensen, 2005; Glotch and Rogers, 2007; Noe Dobrea et al., 2008; Massé et al., 2008; Sowe et al., 2012]. Similar hematite-rich units in Valles Marineris are closely associated spatially with sulfates that compose the Interior Layered Deposits (ILDs). Analysis of OMEGA data showed that finer-grained, red ferric minerals and possibly oxides occur in the ILDs and are concentrated by mass wasting in aprons at the base of slopes [Gendrin et al., 2005; Bibring et al., 2007; Mangold et al., 2008]. Weitz et al.  determined the locations of gray hematite in Ophir and Candor Chasmata strongly correlate to relatively dark mantling units that are either superimposed on or adjacent to ILDs. There are numerous other occurrences of this dark mantling unit within Valles Marineris that do not have detectable hematite signatures, indicating the hematite must be derived from another source and is then concentrated into a limited number of dark debris units.
 In this paper, we map out the distribution of gray hematite in Capri Chasma. After exploring the geologic units associated with hematite detections, we propose likely scenarios for the formation of these hematite-rich localities and hydrated units within Capri Chasma.
2. Data Sets and Methods
 TES emissivity spectra have been analyzed for hematite mineral detection. Each of the six TES detectors has an instantaneous field of view of ∼8.5 mrad, which equates to a spatial resolution of ∼3 × 8 km. TES is a Fourier-Transform Michelson interferometer that covers the wavelength range from 6 to 50 μm, or ∼1670 to 200 cm−1 [Christensen et al., 1992]. Relative hematite abundances were derived using a spectral index similar to the method we employed for Ophir and Candor Chasma [Weitz et al., 2008], which is slightly different than that of Christensen et al. . This index utilizes the wave number positions and emissivity (e) values of several dominant bands associated with coarse, gray hematite, and the high-emissivity region between them, [Christensen et al., 2000, 2001; Lane et al., 2002]. The index value was then calculated according to the following equation:
where ε375cm−1 is the average emissivity from TES bands 21 to 24, ε315cm−1 is the average emissivity from TES bands 16 to 18, and ε460cm−1 is the average emissivity from TES bands 30 to 32. This hematite index value was then mapped to a THEMIS IR daytime map to show the location and abundance of gray hematite (Figures 1 and 2). This index varies from that used in Christensen et al.  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]. The distribution of crystalline gray hematite in Capri Chasma was originally published in Christensen et al. [2001, Figure 3], and our new distribution map shows slight deviations from this older map due to improvements in determining the hematite index value.
 Because the resulting mapped hematite is shown to occur in numerous small patches within our study region, we focus on only those locations where the hematite abundance occurs in the largest patches (i.e., across several TES orbits) and with an index value of >18%, keeping in mind that this is a qualitative abundance, not absolute. The cutoff of 18% hematite index value 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. We note that a similar restriction of >18% hematite when applied to Meridiani Planum, Aram Chaos, and Central Valles Marineris (Figure 1) resulted in distribution maps similar to those found by Christensen et al. , Christensen and Ruff  and Glotch and Christensen , respectively, providing confidence in our technique for identifying and determining relative abundances of hematite. Glotch and Christensen  report hematite abundances of 2–16% for Aram Chaos while Christensen and Ruff  found values of 5–20% across Meridiani Planum. Because only a few areas in our hematite maps of Capri Chasma have similar index values (i.e., red colors in map) matching those seen in Aram Chaos and Meridiani Planum, we infer that most of the hematite abundances within Capri Chasma fall below ∼20% abundance if we use abundance estimates derived by Glotch and Christensen  and Christensen and Ruff .
 Targeted observations by the Mars Reconnaissance Orbiter (MRO) CRISM (Compact Reconnaissance Imaging Spectrometer for Mars) instrument were acquired at 18 m/pixel spatial resolution (Full Resolution Targets, FRT) or 36 m/pixel spatial resolution (Half Resolution Targets, HRL). Hyperspectral observations cover the spectral range of 0.4 to 3.9 μm over 544 channels [Murchie et al., 2007]. Typical analyses of CRISM data involve radiometric corrections (e.g., correction for the solar irradiance, de-striping, despiking), atmospheric correction, and spectral parameter mapping [Murchie et al., 2009a]. Radiometric correction is performed using a pipeline procedure that has been developed by the CRISM team, including atmospheric correction using a scaled atmospheric template derived from observations of Olympus Mons [Murchie et al., 2009a]. Spectral parameter maps corresponding to diagnostic mineralogies were then generated to map the distribution of minerals [Pelkey et al., 2007; Murchie et al., 2009b] and to aid in the selection of spectra. Spectral parameter maps shown in this paper are as follows: the olivine index (OLINDEX), which indicates a positive slope from 1 to 1.5 μm due to olivine and Fe-containing phases, is shown in red; the 1.9 μm band depth, due to combinations of H2O bending and stretching vibrations (BD1900), is shown in green; and the convexity around 2.3 μm in very hydrated phases due to strong H2O absorptions at 1.9 and 2.4 μm (SINDEX), is shown in blue.
 Corrected I/F spectra were averaged into at least 10 × 10 pixel or larger regions of interest that showed spectral features. In order to accentuate subtle mineral absorption bands and remove residual instrumental artifacts, each spectrum was divided by a spectrum of a spectrally bland region that came from the same column(s) as the spectrum of interest to account for column-dependent systematic errors [Murchie et al., 2007, 2009a]. Ratioed spectra, plotted as relative I/F values, were then compared to laboratory data of pure minerals for potential matches; however, some difficulties associated with finding a perfect match to laboratory spectra can be attributed to subpixel mixing in the CRISM data.
 Polyhydrated sulfates (PHS) (e.g., epsomite, hexahydrite, gypsum) are primarily characterized by a 1.4 μm O-H stretch overtone, a 1.94 μm stretch/bend-overtone combination, and a 2.4 μm band that is characteristic of the sulfate ion in the presence of water. Spectral similarities between PHS generally preclude identifying a specific phase at Capri Chasma [e.g., Roach et al., 2010a]. In monohydrated sulfates, such as kieserite, the O-H stretch overtone is broadened and shifted toward 1.55 μm, and the band center of the stretch/bend-overtone combination band is shifted to 2.1 μm. Kieserite, MgSO4•H2O, and szomolnokite, Fe2+SO4•H2O, are structurally similar and thus have similar combination H2O stretch and rotation vibrations at ∼2.1 μm. They can be distinguished from each other based upon an H2O combination band centered at 2.12–2.13 μm in kieserite and 2.09–2.10 μm in szomolnokite, and an additional broad absorption near 0.9 μm due to Fe2+ crystal field transitions [Cloutis et al., 2006]. Roach et al. [2010a] suggested the CRISM spectra for the Capri Chasma ILD were more consistent with kieserite rather than szomolnokite.
 Coarse grained gray hematite lacks spectral features in the 1.0–2.6 μm spectral region [Lane et al., 1999]. However, CRISM observations within Valles Marineris of ferric minerals reveal a red spectral slope from 1.0 to 1.8 μm characteristic of a number of ferric phases. Some occurrences also exhibit absorptions near 1.9 and 2.2 μm resulting from the hydrated ferric sulfate copiapite, the hydrated ferric oxyhydroxide ferrihydrite, or the oxide hematite, possibly in combination with one or more other hydrated sulfates [Roach et al., 2007; Murchie et al., 2007; Bishop et al., 2009]. Dusty regions on Mars may also exhibit a ferric absorption centered near 0.9 μm [Bell et al., 2000; Morris et al., 2000]. Because these dusty regions are commonly used in CRISM spectral ratioing to reduce instrumental noise, the absorptions due to iron in the dust need to be distinguished from crystalline ferric phases in the spectrum of interest. Roach et al. [2010a] applied a continuum removal computation to address this issue when searching for ferric oxide mineralogy in Valles Marineris, including Capri Chasma. Their results are for all ferric oxides identified in CRISM data, including both fine grained red hematite and coarse grained gray hematite. In this paper, we only discuss observations of units associated with coarse grained gray hematite detected by TES.
2.3. HiRISE and CTX
 A critical element of this study is to understand the morphological, mineralogical, and stratigraphic context for the hematite detections. To accomplish this, we have co-registered available imagery from the MRO Context (CTX) and High Resolution Imaging Science Experiment (HiRISE) cameras with the hematite abundance map in order to investigate the photogeological context of hematite occurrences and extract information concerning the origin of this oxide. HiRISE (∼25 cm/pixel spatial scale) and CTX (∼6 m/pixel) images of the hematite patches and geologic contacts with adjacent units have allowed us to determine what units correspond to the hematite and what is the stratigraphic relationship of the hematite relative to other geologic units. Color HiRISE products were also utilized and include enhanced RGB and IRB images. The IRB images consist of IR, RED, and BG channels, where RED corresponds to 570–830 nm, BG is <580 nm, and IR is >790 nm. The RGB images are composed of RED, BG, and synthetic blue images where the synthetic blue image digital numbers (DNs) consist of the BG image DN multiplied by 2 minus 30% of the RED image DN for each pixel [McEwen et al., 2007].
 Many of the geologic units are defined based upon regions where both HiRISE and CRISM data exist. Consequently, areas with coverage from both data sets will have a bias toward more geologic units defined and mapped than areas without these data sets. Nevertheless, inspection of CTX images provides confidence when mapping areas without HiRISE coverage because we can still distinguish morphologic differences at this lower resolution scale, such as evidence of layering, tonal changes, decameter-scale textures, etc.
 All data sets were registered to the THEMIS daytime IR mosaic as a base map. CRISM spectral parameter color maps were registered to HiRISE images in Adobe Photoshop using tie points. Because the registration is done manually, there could be ∼20 m or less errors in the precise placement of individual images but our interpretations are based upon observations made across larger regions and, consequently, these small placement errors should not affect our results.
 Digital Terrain Models (DTMs) with a vertical accuracy of 1 m [Kirk et al., 2008] were generated for select locations where HiRISE stereo coverage exists. DTMs are desirable for determining absolute values in elevation, unit thicknesses, and to compute slopes. CRISM spectral parameter maps were co-registered and overlain on each DTM to aide in the interpretation of stratigraphic relationships of different hydrated units. Units are defined and mapped based upon morphology and mineralogy as follows (Table 1): Light-toned Layered (LTL), Light-toned (LT), Medium-toned Layered (MTL), Medium-toned (MT), Dark Mantle (DM), and Eolian bed forms such as ripples and dunes (D). Note that the brightness of any unit is subjective, and, consequently, the term “light-, medium-, and dark-toned” do not refer to a specific albedo range but rather to a unit's reflectance relative to other units within the same HiRISE or CTX image. If a mineralogy could be identified using CRISM data, the main mineralogy is listed in association with a unit as follows: Kieserite (K), Polyhydrated Sulfate (PHS), Monohydrated Sulfate (MHS), PHS and MHS mixture (Mix), Opal (Op), Doublet absorption between 2.2 and 2.3 μm (Db), Ferric hydroxysulfate (FHS), and Fe/Mg-smectite (FeSm).
Table 1. Definition of Units and Mineralogy
Tone and Morphology
Dunes and Ripples
PHS and MHS mixture
Doublet absorption between 2.2 and 2.3 μm
3. Geologic Units Associated With Gray Hematite
 Capri Chasma is located in eastern Valles Marineris where the linear troughs of Coprates Chasma merge with outflow channels and chaos terrain. Contained within Capri Chasma is a ∼4 km tall central Interior Layered Deposit (ILD) named Capri Mensa whose surface lies ∼2 km below the adjacent plateaus. The ILD is surrounded by knobs and other collapse features that compose the chasma floor. Spectral mapping of the ILD indicates it is dominated by kieserite with a capping unit of PHS [Roach et al., 2010a; Flahaut et al., 2010a].
Figure 2 shows the footprints of HiRISE and CRISM images analyzed in this study, along with detections of hematite from TES. We analyzed all HiRISE and CRISM images where there were corresponding TES detections of hematite. Six areas that had extensive HiRISE and CRISM coverage of spatially large hematite occurrences (locations 1–6) were then analyzed in greater detail to map associated geologic units and determine stratigraphic relationships.
3.1. Location 1
3.1.1. Geologic Units
Roach et al. [2010a] mapped PHS stratigraphically above kieserite for the entire Capri Chasma ILD, but our results using DTMs combined with CRISM spectral analysis at location 1 indicate a more complex stratigraphy. Units within the thickest interior portion of the ILD are exposed along the walls of a ∼30-km diameter impact crater. Thus far, no other locations along Capri Mensa show this many sulfate units below the thick kieserite-bearing unit, LT(K), that dominates the ILD. Hence, this southeastern location along the crater wall appears to have unique and localized units due to its deeper exposure within the central portion of the ILD. Because the units show no evidence for disruption or tilting, we interpret them to be intact units rather than material brought up from depth by the impact process.
 A geologic map of location 1 and associated units is shown in Figure 3. Across much of the chasma floor and top surface of the ILD is a unit we map as Eolian and Floor Materials. CRISM spectra extracted from this unit exhibit no diagnostic spectral features in the 1–2.6 μm region, and neither HiRISE nor CTX images reveal many details about this unit other than eolian features and mounds.
 In instances where eolian debris has not obscured the upper portions of the ILD, we identify a ∼100 m thick Medium-Toned PHS, MT(PHS), unit. CRISM spectra (Figure 4) are consistent with a PHS, although the specific mineralogy cannot be ascertained. The unit appears medium-toned in brightness relative to the brighter kieserite and darker mantle units. In some instances the contact between upper MT(PHS) and lower LT(K) is sharp (Figure 5c). Previous observations of PHS and kieserite, both at Capri and elsewhere in Valles Marineris [Mangold et al., 2008], noted the PHS is generally darker than kieserite units, either due to distinct lithologic variations between the sulfates or the presence of dark sand collecting in the flatter PHS.
 Stratigraphically beneath MT(PHS) is a thicker, brighter LT(K) unit. The unit is variable in thickness, ranging from 1 km thick along the eastern portion of the crater to 1.5 km thick along the southern wall. Further to the northeast and along the ILD edge where there are no visible underlying units, LT(K) is 2.5 km thick based upon MOLA data. CRISM spectra of clean exposures of LT(K) (Figure 4) are consistent with the monohydrated sulfate, kieserite (K). The unit is exposed along much of the ILD edges, particularly in the north where there are steep, clean slopes. Layers are rarely seen and most of the material appears massive with jagged erosional edges (Figures 5c and 6j). However, to the northeast along Capri Mensa, HiRISE color images show the unit can exhibit two colors: an upper reddish color and a lower grayish color (Figure 7). CRISM spectra of the upper and lower LT(K) materials (Figure 7c) are indistinguishable, indicating the color differences are only a property in the visible wavelengths.
 DTMs covering LT(K) reveal the unit has a rougher, irregular surface relative to the underlying units (Figures 8 and 9). The average surface slope across the entire thickness of the unit is ∼15°. A dark, spectrally bland mantle obscures much of LT(K). We did not map the mantle as a separate unit, but instead interpret the mantle as a concentration of disaggregated material covering the morphology of the original terrain, in this case outcrops of LT(K).
 Along the southeastern portion of the crater wall in location 1 (Figure 3), we map units LTL(PHS), LTL(Mix), and LTL(MHS) in this localized area that corresponds to the thickest and deepest interior portion of the ILD. The location of LTL(PHS)1 shown in Figure 6k has spectral absorptions at 1.03, 1.94, and 2.4 μm (Figure 4). Murchie et al. [2009c] noted similar spectral features in association with ILDs and hematite in Candor Chasma and suggested the features are consistent with a Fe-rich PHS (i.e., hydrated ferric sulfate copiapite or hydrated ferric oxyhydroxide ferrihydrate) or a mixture of hematite plus a PHS. One layer within LTL(PHS), LTL(PHS)2 (see Figure 6k), is spectrally distinct from the other layers. Its spectrum has a feature at ∼2.11 μm that is lacking in the other layers of the unit. We interpret this layer to be an interbedded MHS within the PHS-bearing material.
 HiRISE images of LTL(PHS) indicate many layers, with the beds unequally spaced and variable in exposure. Prominent upraised marker beds occur every 10–20 m, with thinner beds between (Figure 6i). Although much of LTL(PHS) is obscured beneath dark mantle material DM1, where clean exposures are visible the beds appear bright and fractured. Faults disrupt segments of the unit (Figure 6i), but most beds can be traced around the entire exposure of LTL(PHS). The layer that corresponds to the spectrum LTL(PHS)2 is almost completely obscured beneath dark mantling debris and appears indiscernible from adjacent beds in HiRISE images.
 The transition from upper LT(K) to lower LTL(PHS) is muted by surficial debris, However, a break in slope between the two units is evident in the DTM, with the lower portion of LTL(PHS) having a slope of 16° before decreasing to 8° just below the contact with LT(K) before resuming to a slope of 15° across LT(K) (Figure 9). We interpret the break in slope to an erosional bench attributable to the more friable nature of LT(K) relative to LTL(PHS), thereby enhancing erosion and removal of LT(K) with respect to LTL(PHS) and underlying units.
 The transition from LTL(PHS) to underlying unit LTL(Mix) is also obscured by DM1 in HiRISE images, although both units can be differentiated in CRISM color parameter images (Figure 6k). Layering within LTL(Mix) is subtle and appears due to lineations and albedo differences rather than distinct beds. Fracturing is pervasive, creating angular bright blocks of variable size with dark filling material between the blocks (Figure 6h). LTL(Mix) appears to be the brightest unit in HiRISE images, either because of a true lithologic property or because of lesser amounts of dark eolian mantle. The unit is ∼400 m thick with surface slopes of 22° based upon DTM measurements.
 CRISM spectra extracted from LTL(Mix) have high reflectances in the 1.0–1.9 μm region relative to other units, with generally a strong absorption between 1.93 and 2.0 μm, a broad absorption between 2.0 and 2.24 μm, and a weak absorption or shoulder at 2.4 μm (Figure 4). Several of the spectra from LTL(Mix) are similar to LTL(PHS), although the two units are easily differentiated in both the CRISM color parameter images and in HiRISE images which is why we map them as two distinct units. A comparison to laboratory spectra indicates kieserite is sometimes a component of LTL(Mix), but the strong absorption centered at 1.93 μm is consistent with a PHS. Hence, we interpret the unit to be a mixture of intermediate hydration states between MHS and PHS. Laboratory spectra of magnesium sulfates with different hydration states confirm that the addition of water causes features to be present at 1.9, 2.1, and 2.4 μm [Dalton et al., 2005]. In particular, for MgSO4·nH2O when n = 1–2 the spectra seen in laboratory measurements from Dalton et al. [2005, Figure 2] have similar features to those sometimes seen in spectra taken from LTL(Mix).
 We map unit LTL(MHS) beneath LTL(Mix). The transition between LTL(Mix) and LTL(MHS) is subtle in morphology. Because the lower portions of LTL(MHS) are obscured beneath a dark mantle, we can only measure a minimum thickness for the unit of 360 m. Surface slopes along the unit are ∼20°.
 LTL(MHS) is distinguishable from LTL(Mix) in CRISM color parameter images by a bluer color relative to the pinker color of LTL(Mix) (Figure 6k). While LTL(MHS) has the same pervasive fracturing as LTL(Mix) (as observed in HiRISE images), the blocks creating by the fracturing appear smaller in size. Additionally, although LTL(PHS) has subtle, if any, layering, LTL(MHS) has visible layering that corresponds to albedo and color variations in HiRISE images (Figures 6f and 6g). Because the albedo and color differences extend horizontally across the entire unit and do not change in thickness or correspond to topography, we interpret the differences to represent distinct lithologies, perhaps due to variable amounts of iron. CRISM spectra support this interpretation because spectral features indicate some kieserite is present, associated with the brighter layers, while spectra extracted from the darker layers have a feature at 2.10 μm rather than 2.14 μm, consistent with szomolnokite (Figure 4). Hence, we interpret both monohydrated sulfates are interbedded and correspond to distinct brightness and color attributes. Similar alternating light and dark layers that correspond to intercalated szomolnokite and kieserite have also been identified in one of the ILD mounds of Juventae Chasma [Bishop et al., 2009].
 Lower units MT(PHS) and MT(Mix) are interpreted as material on the crater floor disrupted by the impact process. HiRISE images show heterogeneous patches of material with distinct albedo and physical properties (Figures 5d, 6c, and 6d). Possible flow structures, breccia-like textures, and faults are also found in these units. Because the units have been disturbed by the impact event, we cannot determine if they represent geologic units stratigraphically beneath LTL(MHS) or if they are mixtures of units from above. Spectra from the lower MT(PHS) are consistent with Mg-PHS due to the lack of strong ferrous absorptions (Figure 4d). MT(Mix) has absorptions at ∼1.0, 1.94, 2.4, and broader absorptions around 1.44 and 2.1 μm (Figure 4a). These features suggest a Fe-type MHS, such as szomolnokite, or a mixture of Fe-type PHS and kieserite. We did not identify these lower units at the other five locations.
 Surficial mantle units are especially well defined for areas that have CRISM coverage even though they generally appear similar in lithology in both HiRISE and CTX images. Only where CRISM data exist were we able to distinguish mineralogies and map distinct mantle units. Hence, surficial mantle units Dark Mantle 1 (DM1), Dark Mantle 2 (DM2), and Kieserite Dunes/Ripples (D(K)) may exist elsewhere at Location 1 but they could not be identified and mapped due to the absence of CRISM data.
 DM1 has spectral absorptions at 1.05–1.1, 1.92, 2.2 and a negative slope longward of 2.4 μm (Figure 4d), similar to those seen in LTL(PHS) except with an additional weak feature at 2.2 μm. We also extracted VNIR spectra of units DM1 and LTL(Mix) (Figure 4e). Both units have a broad absorption centered at 0.88 and a peak at 0.76 μm, consistent with red hematite in laboratory spectra. LTL(Mix) also has a broad absorption around 0.53 μm that can be attributed to nanophase oxides. The absorption is much weaker in DM1, perhaps because other ferric phases may occur with the gray hematite [Murchie et al., 2009c].
Roach et al. [2010a] noted that all kieserite-bearing bedrock materials at Capri Chasma have a red crystalline hematite component while the upper capping PHS does not. The spatial correlation between TES hematite pixels and units DM1 and LTL(Mix) indicates that gray hematite must be present. The VNIR spectra suggest red hematite is present as well and could be derived from the physical weathering of the coarser gray hematite [Roach et al., 2010a].
 In HiRISE images, DM1 appears as a smooth, dark unit with sharp boundaries, consistent with an indurated mantle (Figures 5e and 6f). Because we cannot see any underlying topographic expressions, the mantle must be several meters thick. Circular dark spots of DM1 could represent former impact craters that were filled with DM1 and subsequently eroded (Figure 5e). DM1 occurs on units LTL(Mix), LTL(PHS), LT(K) and LTL(MHS). A HiRISE image taken of the southwestern crater wall in Location 1 (Figure 10a) also shows lithified, thick dark mantles but without corresponding CRISM spectral information we cannot distinguish between DM1 and DM2 (see below) here.
 DM2 is spectrally unremarkable in CRISM data and it only mantles LT(K). It has sharp boundaries, appears consolidated, and obscures the underlying rough topography of LT(K) (Figure 6j). DM2 could represent lithified eolian material but it is difficult to explain its occurrence along the upper crater wall slope rather than on flatter terrain.
 Finally, we map unit D(K) in two small areas where bright eolian bed forms, such as dunes and ripples, occur along LT(K) (Figure 5f). CRISM spectra of this unit are consistent with the presence of kieserite, with a better match to the laboratory spectrum of kieserite than other kieserite-bearing units LT(K) and LTL(K). The unit is best seen in the CRISM color parameter map where it appears blue (Figure 8). We interpret the unit to represent sand-size material eroded from LT(K) that was subsequently modified into eolian bed forms by winds.
3.1.2. Correlation of Gray Hematite to Geologic Units
 With the aid of two DTMs, we identified geologic units that correspond to the detection of hematite along the southern crater wall. In Figure 8, hematite occurs in association with units LTL(Mix), LTL(PHS), and DM1. In another DTM just to the west (Figure 9), hematite corresponds to units LT(K), LTL(PHS), LTL(Mix), and LTL(MHS), and DM1. To further illustrate the association between hematite and geologic units, we produced topographic profiles at four locations (Figures 3 and 11). In all cases, hematite detections begin at upper unit LT(K) and extend down the crater wall to the lower portions of the crater floor. The three units that always occur in association with hematite include LT(K), LTL(Mix), and DM1. Based upon these observations, we interpret the hematite to be embedded within LT(K). LTL(Mix) may also be a source for hematite grains, or the hematite may constitute a lag deposit from the erosion of the overlying LT(K) unit. CRISM spectra and HiRISE images of DM1 support the claim that this unit may be a soil lag where hematite grains are concentrated. The CRISM spectra of DM1 are consistent with a Fe-bearing material while HiRISE observations show a thick surficial unit composed of dark material. Additionally, HiRISE RGB images of DM1 show a red color, also consistent with Fe-bearing material.
 Weaker detections of hematite along the northeastern chasma floor (Figure 3) are not well correlated to hydrated units. Figure 10b illustrates the terrain where hematite occurs along the chasma floor. The terrain appears bland and dark, with some dunes and ripples. Small impact craters that penetrate a few meters down appear to expose a brighter substrate, but not enough material is exposed to detect any features in CRISM data. Hematite occurs at a maximum distance of ∼10 km from the kieserite exposures along the ILD margin. It is unlikely that hematite could be weathering out of the kieserite along the ILD mound and subsequently transported by wind multiple kms, assuming the hematite grains are millimeter in size. At Meridiani Planum, the size of the hematite spherules in the soils corresponds directly to the size of the spherules seen in local outcrops [Weitz et al., 2006; Calvin et al., 2008], which suggests that once weathered out from sulfate outcrop the spherules are not rolled or moved more than a few meters by the wind. Consequently, hematite along the northeastern Capri Chasma floor may be embedded in the underlying brighter substrate where erosion and removal by the wind of formerly overlying substrate created a soil lag of hematite grains, similar to the situation at Meridiani Planum [Soderblom et al., 2004; Weitz et al., 2006]. Alternatively, if LT(K) once extended further out from the current edge of Capri Mensa but was subsequently removed by erosion, then any hematite grains that precipitated within this former LT(K) material could have collected along the floor at this location as a soil lag.
3.2. Location 2
3.2.1. Geologic Units
 Location 2 contains a relatively high concentration of hematite along the northeastern edge of the ILD (Figure 2). We have mapped four units at this location (Figure 12). Unit LT(K) is morphologically similar to unit LT(K) mapped at Location 1. Exposures show a light-toned massive unit with a jagged surface partially mantled by darker debris along the edge of the ILD (Figure 13). CRISM spectra are consistent with a monohydrated sulfate, such as kieserite (Figure 14a).
 Adjacent to the ILD edge is a smoother plains unit. To the west the unit appears dark-toned while to the east an underlying medium-toned material is sometimes visible (Figures 13c and 13d). In the CRISM color spectral parameter image (Figure 13a), the unit is reddish-yellow, indicating an Fe-bearing material is present. CRISM spectra (Figure 14a) show that the western portion of the unit, DM(H), has weak absorptions around 1.05 and 1.92 μm. The eastern portion of the unit that appears as a mixture of dark mantle on medium-toned material, DM(H)+PHS, exhibits a stronger absorption at 1.92 and a broad feature around 2.40 μm, both of which are consistent with a PHS. Consequently, we interpret the underlying medium-toned material to represent a PHS while the darker mantle is consistent with an Fe-bearing material.
 Based upon analysis of a HiRISE stereo anaglyph, unit DM(H)+PHS has a gentle downward slope moving northward away from the ILD (Figure 13g). Steeper slopes occur further to the north. Circular features shown in Figure 13i are not impact craters but rather small mounds. These mounds, as well as the larger mesas visible in Figure 13g, indicate erosion has removed overlying material at this location.
 Along the chasma floor we map Eolian and Floor Materials, which include ripples and dunes. Small mounds are seen along the chasma floor but most appear mantled by eolian debris, obscuring possible diagnostic features such as layering and mineralogy. There are a few mounds that do show light-toned layering and these we map as unit LTL. Upon closer inspection of HiRISE and CRISM images for one mound, we divide unit LTL into three subgroups: MT(FeSm), LTL(Op), and LT(Db).
 Subgroup MT(FeSm) occurs along the higher elevations of the mound and appears dark-toned with polygonal fracturing (Figures 13e and 13h). CRISM spectra are consistent with a Fe/Mg-smectite, such as nontronite (Figure 14b). Subgroup LTL(Op) represents light-toned material with polygonal fracturing and faint layering visible in some outcrops. The different layers could indicate the presence of distinct materials within LTL(Op) but the thicknesses of these layers are below the resolution of CRISM. Instead, CRISM spectra were extracted across the entire height of LTL(Op) and show possible features at 1.42, 1.91, and a broad absorption between 2.20 and 2.27 μm (Figure 14b). These features are most consistent with a hydrated silica, such as opal. We note, however, that the 2.2 μm feature exhibits a strong break in slope at 2.21 and 2.27 μm, suggesting that material from LT(Db) (see below) may form part of the unit.
 The third subgroup, LT(Db), occurs within LTL(Op). It is slightly darker in appearance than LTL(Op) but not as dark as MT(FeSm). There is no evidence of polygonal fracturing or layering (Figure 13e). A CRISM spectrum for LT(Db) reveals a weak feature at 1.91 and a doublet absorption at 2.22 and 2.27 μm. The spectrum is very similar to LTL(Op) with sharp shoulders at 2.21 and 2.27 μm suggesting that LTL(Op) may contain the same phases as LT(Db) but with a weaker doublet. This same doublet has been found elsewhere in Valles Marineris and on Mars, including Melas basin [Weitz et al., 2012], Ius Chasma [Roach et al., 2010b], Noctis Labyrinthus [Weitz et al., 2011], and Mawrth Vallis [Noe Dobrea et al., 2011]. The doublet can be explained by a mixture of gypsum and jarosite. A leached clay produced in the laboratory by acid weathering of a basalt sample also shares the doublet feature (Figure 14b) [Tosca et al., 2008a; Weitz et al., 2011]. The proximity of LT(Db) to MT(FeSm) would be consistent with leaching under acidic conditions of MT(FeSm) to form LT(Db).
 Using the HiRISE anaglyph to estimate relative topography and slopes (Figure 13h), we interpret LT(Db) and MT(FeSm) to be above LTL(Op). Erosion and mass wasting along the slopes of MT(FeSm) and LT(Db) has removed LTL(Op) material, which is why LTL(Op) is seen only at lower elevations or as small patches that drape upslope. Because HiRISE and CRISM images do not cover the other occurrences of LTL, we cannot ascertain if all three subgroups are present elsewhere or are limited to only this mound.
3.2.2. Correlation of Gray Hematite to Geologic Units
 The gray hematite detected by TES correlates strongly to unit DM(H)+PHS. A few of the TES pixels occur along the slopes of the ILD where unit LT(K) is exposed. Most of unit LTL does not correlate to any TES hematite detections. Based upon these observations, hematite is likely embedded within LT(K) and possibly DM(H)+PHS. If LT(K) extended further northward from its current location, then erosion and removal of LT(K) by the wind could have produced a surface lag of more resistant hematite grains, which now composes unit DM(H). However, we cannot rule out the possibility that the medium-toned PHS material beneath DM(H) may also be the source of some hematite if erosion of formerly overlying PHS material also contained embedded hematite grains.
3.3. Location 3
3.3.1. Geologic Units
 Hematite is detected along a large area of the chasma floor between Capri Mensa and smaller outliers of ILD material at Location 3. We map four units here as shown in Figure 15. Eolian and Floor Materials is the same unit mapped previously at the other locations. It encompasses the chaotic terrain along the chasma floor as well as eolian features and debris.
 In cases where layered material and/or medium-toned outcrops are visible, we mapped this terrain as unit DM(H)+PHS. The terrain can appear smooth and flat with ripples (Figure 16c) or rough with layering sometimes visible (Figures 17c and 17d). The unit is dark-toned if covered by dark mantle, DM(H), or medium-toned where the mantle is thin and the underlying PHS material is exposed. Where outcrops of PHS are seen in HiRISE images, the surface is polygonally fractured and meter-size blocks are evident (Figure 17c).
 CRISM spectra of DM(H)+PHS exhibit absorptions around 1.0–1.05, 1.92 and 2.40 μm (Figures 16g and 17g). These features are consistent with PHS and an Fe-bearing material, either intrinsic to the sulfate or the dark mantle. One possibility that we favor based upon the morphology and mineralogy of the unit is a PHS mixed with a hematite-bearing mantle, DM(H). This interpretation is consistent with the strong correlation of TES hematite pixels to this unit.
 Additional exposures of the unit that lack CRISM coverage have been mapped elsewhere in location 3 based upon similar morphologies seen in HiRISE images and TES detection of hematite. An example of the terrain shows underlying medium-toned polygonally fractured, sometimes layered, material interpreted to be PHS while the overlying dark mantle could represent DM(H) (Figure 18).
 LT(K) represents the same unit identified previously at locations 1 and 2. Exposures occur along the edge of the ILD and in smaller mounds, presumably outlier material of the original ILD extent. LT(K) overlies DM(H)+PHS in all occurrences where the two units are in contact. Figure 16d illustrates the lithologic change of LT(K) with elevation. The lower material is massive and homogeneous but transitions upward to a heterogeneous mixture of light and dark layered material. Spectrally, the lower and upper materials are similar.
 Near the base of LT(K) is a dark mantle composed of kieserite-bearing talus, DM(K) (Figures 16e and 16g). We interpret DM(K) to represent dark material eroded from upper LT(K) and shed downslope, with mixtures of eolian and light-toned LT(K) debris. Both Meridiani Planum and Candor Chasma also exhibit dark, sandy material emanating from discrete layers within the sulfates and redistributed by wind [Murchie et al., 2009c].
 Blue material in a CRISM color parameter image (Figure 17a) corresponds to bright eolian dunes adjacent to larger darker dunes (Figure 17e). Spectra of the bright dunes are consistent with kieserite (Figure 17g), indicating the dunes are composed of or covered by kieserite sand grains.
 The fourth unit at location 3, LT(FHS), has only been found in association with two small mounds. The unit corresponds to light-toned material within the darker DM(H)+PHS unit (Figure 16f). CRISM spectra (Figure 16h) exhibit absorptions at 1.44, 1.93, 2.24, 2.29, and 2.39 μm. These features are slightly different from those seen in LT(Db) at location 2. Flahaut et al. [2010a] interpreted the spectra to be hydrated silica, although this does not explain the 2.24 and 2.29 μm doublet. At Aram Chaos [Lichtenberg et al., 2010] and along the Juventae Chasma plateau [Bishop et al., 2009] materials with a 2.24 μm feature are attributed to the presence of ferric hydroxysulfate. LT(FHS) appears in HiRISE images to display a heterogeneous mixture of brightness and surface textures. CRISM spectra were extracted across much of the LT(FHS) exposure shown in Figure 16f, suggesting that a mixture of materials, such as nontronite and a ferric hydroxysulfate, could best explain the spectral features. The ferric hydroxysulfate at Aram Chaos is an older unit that underlies the sulfates and is thought to have formed under dry, low humidity conditions [Lichtenberg et al., 2010]. We cannot ascertain if the mounds that contain nontronite and ferric hydroxysulfate in Capri Chasma are underlying units that were buried under sulfates and then subsequently exposed by erosion, or if the materials post-date the sulfates. Alternately, it is possible that the observed relationship is not controlled by depositional age, but is instead controlled by zoning. If the ferric hydroxysulfate dissolves at a higher water activity than the polyhydrated sulfates, then it would have precipitated out at a greater depth in an upwelling groundwater model [Murchie et al., 2009c].
3.3.2. Correlation of Gray Hematite to Geologic Units
 TES hematite pixels strongly correlate to unit DM(H)+PHS at location 3 (Figure 15). Although some TES pixels encompass lower portions of LT(K), the large spatial resolution of TES means each pixel may represent a mixture of hematite in DM(H)+PHS with non-hematite LT(K) material. Consequently, it is plausible that LT(K) does not contain any hematite or at least not in abundances that could be detected by TES.
3.4. Location 4
3.4.1. Geologic Units
 Approximately 2 km of topography are exposed along the edge of the ILD at location 4. Along the top of the ILD, medium-toned polygonally fractured outcrops are sometimes visible beneath darker eolian debris (Figure 19). The outcrops are too small in size for CRISM spectra to be extracted, but their morphology is comparable to unit PHS identified just to the east at location 1. Therefore, we map and interpret the upper unit along the ILD here to be the same capping MT(PHS) identified at location 1.
 Unit LT(K) extends from upper MT(PHS) down to the chasma floor. LT(K) appears morphologically and spectrally similar to the same unit mapped at the previous locations. Where not covered by a dark mantle, the unit is bright and massive with jagged exposures. CRISM spectra are consistent with kieserite (Figure 19f). The many sulfate units below LT(K) that are exposed along the southern crater wall at location 1 are not seen along the ILD edge at location 4.
 We map the flatter terrain along the base of LT(K) as unit DM(H)+PHS. Much of the unit appears as dark eolian ripples or debris but a medium-toned, polygonally fractured, layered terrain is visible where the overlying dark mantle is thin enough to see through it (Figure 19e). Further northward, only eolian material can be seen in CTX images. The northern boundary of DM(H)+PHS is not well constrained due to the lack of HiRISE and CRISM coverage.
 Examples of CRISM spectra extracted from two locations in unit DM(H)+PHS are similar and show absorptions at ∼1.07, 1.94, and 2.40 μm (Figure 19g). As discussed for the previous locations, we interpret the unit to be a mixture of PHS and a dark mantle enriched in gray hematite. The 1.07 μm feature could be due to Fe in the PHS or in the hematite. Because gray hematite detected by TES correlates to unit DM(H)+PHS, CRISM spectra must contain a component of this hematite. However, red fine grained hematite could also be mixed into the mantle and consequently affect the Fe absorption at 1.07 μm.
3.4.2. Correlation of Gray Hematite to Geologic Units
 Gray hematite detected by TES is strongly associated with unit DM(H)+PHS (Figure 19a). Only one hematite pixel occurs on LT(K), and it is isolated in the far eastern portion of Location 4. However, there are small outlier hills of LT(K) located several kms from the ILD edge, indicating the unit was more extensive in the past but has been removed by erosion. Hematite could be embedded in either LT(K) or lower PHS rocks, but with abundances below detectability by TES given the limited exposures that correlate to hematite pixels.
3.5. Location 5
 Several TES hematite pixels cluster at location 5. Unlike the previous locations, there are no exposures of LT(K) along the ILD edge. Instead, there are a few small hills composed of kieserite that we map as LT(K) based upon spectra and morphology seen in CRISM and HiRISE images, respectively (Figure 20). Most of the terrain consists of a mixture of medium- and light-toned material covered by a darker mantle, which we map as unit DM(H)+PHS. In enhanced HiRISE RGB images, the dark mantle appears reddish-brown relative to the underlying terrain (Figure 20e). Based upon its sharp edges, the redder mantle looks lithified rather than composed of loose debris, and it appears to be eroding into small patches. The medium-toned underlying terrain is polygonally fractured and appears layered in the northern portion of location 5.
 A spectrum extracted from DM(H)+PHS from the area shown in Figure 20d has features at 1.0–1.05 and 1.92 μm, with possible small absorptions at 2.14 and 2.44 μm. As noted for previous locations with similar spectral features, the spectrum in consistent with a mixture of hematite and PHS. Gray hematite detected by TES is strongly associated with unit DM(H)+PHS. Consequently, we interpret the dark mantle to be hematite-bearing while the underlying medium-toned terrain is PHS.
 One other observation to be noted at location 5 is along the wall of a rounded depression. Several light-toned beds are exposed at the contact between the upper DM(H)+PHS and a lower layered blocky wall rock unit (Figure 20f). The light-toned beds differ in brightness, color, and erosional characteristics, indicating several lithologies are present. Because the beds are too thin to be resolved individually in CRISM images, the spectrum shown in Figure 20g (Light-toned beds) represents an average of all the beds. Spectral features appear more consistent with a PHS rather than a MHS, but a specific mineralogy cannot be inferred. Based upon physical properties seen in HiRISE images, we interpret these light-toned beds to be different from both the stratigraphically higher PHS and LT(K) units, but they could still represent sulfates or perhaps other hydrated materials.
3.6. Location 6
 Hematite detections along the far western portion of Capri Chasma at location 6 are shown in Figure 21. The hematite occurs along the southern wall of an impact crater. The wall appears to correspond to the edge of ILD material, although not the central ILD mound (Figure 2). Along the upper surface we map the same unit MT(PHS) seen elsewhere as a capping unit on the ILD. Physical features seen in HiRISE images include a medium-toned unit with polygonal fracturing beneath darker eolian debris. CRISM spectra (Figure 21g) have features at 1.04–1.07, 1.94, and 2.40 μm, consistent with a mixture of PHS and an Fe-bearing phase.
 Below MT(PHS) and exposed along the upper crater wall is unit LTL(K). The unit has spectra similar to unit LT(K) identified at the previous locations, consistent with a kieserite material. While the upper rocks appear massive, lower exposures of LTL(K) show bedding, although all beds look similar in lithology (Figure 21e). Because of the layering seen in this unit, we distinguish it from unit LT(K) which does not show similar layering.
 Hematite detected by TES correlates to LTL(K) and DM(H). In the CRISM color parameter map (Figure 21a), loose dark debris mass wasting down the crater wall appears orange (i.e., high OLIVINE index values). A spectrum taken from this debris (Figures 21f and 21g, DM(H)) has a strong feature at 1.07 and absorptions at 1.94 and 2.40 μm. The location where this spectrum was extracted also corresponds to hematite detected by TES (Figure 20b), indicating the debris contains hematite as well. Further to the west where there is no hematite signature, the same debris has similar spectral features in CRISM data (DM(PHS)) but the absorptions are stronger, indicating the presence of more PHS. While all the dark debris along the crater wall may contain mixtures of hematite and PHS, higher hematite abundances seem to correlate with lower PHS abundances. Because CRISM spectra do not show kieserite features, erosion of MT(PHS) rather than LTL(K) must be the source of the debris.
4.1. Sulfate Formation
 Numerous models have been developed to explain the deposition of sulfates within Valles Marineris. Groundwater upwelling has been proposed as a method for trapping, cementing, and lithifying layers of eolian dust and sand in Valles Marineris, Meridiani Planum, and Aram Chaos [Andrews-Hanna et al., 2007; Murchie et al., 2009c]. Modeling by Andrews-Hanna and Lewis  indicates that Capri Chasma would have experienced relatively high rates of groundwater upwelling that could have created thick sulfate deposits. A mechanism involving groundwater recharge/evaporation and multiple wetting events was favored for the formation of ferric hydroxysulfate, monohydrated sulfate, and polyhydrated sulfates in Aram Chaos [Lichtenberg et al., 2010]. Because we do not see evidence in the CRISM data for clays within the ILD, it is more plausible that we are observing deposits derived from the evaporation of magnesium and sulfur-rich brines, rather than groundwater alteration of mafic minerals.
Niles and Michalski  proposed the sediments at Meridiani Planum and in many other locations on Mars formed from acid-weathering inside massive ice-dust deposits that incorporated large amounts of volcanic sulfur gases. These massive dust-ice deposits formed through precipitation of ice condensed around dust grains and acidic aerosols during periods of high obliquity, resembling the polar layer deposits that exist today on Mars. A similar model has been suggested to explain the sulfates within Valles Marineris [Michalski and Niles, 2012].
Flahaut et al. [2010a] noted that kieserite lies between a lower and upper PHS in the Capri ILD. They proposed two models to form the sulfates: precipitation of evaporites in a shallow body of standing water or multiple groundwater upwellings altering earlier ILD volcanic material. This study only discussed formation of kieserite and PHS within the ILD, but our results indicate the sulfates within the deepest interior portion of the ILD experienced a much greater variability in hydration states relative to the outer portions of the ILD. In addition, we have identified additional minerals, including possible clays, leached clays, opal, and ferric hydroxysulfate. Hence, the depositional sequences and possible origins are more complex than previously discussed.
 We postulate the sulfates seen only at location 1 are not the result of the impact process because the increase temperatures and pressures associated with the impact event would have affected units along all crater walls yet we only see these sulfates along the southern wall, corresponding to the thickest and central portion of the ILD. The beds are continuous and flat-lying along the southern wall but do show faulting that is consistent with disruption of pre-existing materials from the impact event (see Figure 6i). The general appearance of the crater, especially the lack of ejecta and raised rims, indicates the crater has undergone significant erosion since formation.
 A MOLA topographic profile across much of Capri Mensa demonstrates the complex stratigraphy (Figure 22). The large vertical exaggeration gives the appearance of high slopes along several of the units but in reality slopes across LT(K) in profile AB are <1° and only 1.5° along the northern crater wall in profile BC. We deduce that units LTL(PHS), LTL(Mix), and LTL(MHS) pinch out because we cannot identify the same units at other outcrops in Capri Mensa, although the lack of HiRISE and CRISM coverage along the southern portion of the ILD hinders our ability to map these units here. Even though the lower PHS associated with unit DM(H)+PHS always lies stratigraphically below LT(K), the topography demonstrates that this lower PHS unit is not always below LT(K) in elevation. One possibility is that the lower PHS unit experienced erosion after emplacement, creating different elevations on which LT(K) material could then be deposited. Based upon crater counting and stratigraphic relationships, Flahaut et al. [2010b] suggested the ILD was emplaced in the Hesperian during or after formation of the chasma.
 We investigate scenarios that build upon these previous models for ILD formation using either groundwater upwelling for the water and sediment source, or ice deposits trapping dust and volcanic aerosols. Assuming the groundwater upwelling model proposed by Andrews-Hanna et al.  and Andrews-Hanna and Lewis , PHS formed under less saline conditions initially (Figure 23). With time, saline conditions rose as LT(K) was laid down across much of the chasma to form the thick sequence of kieserite that now dominates the ILD. Finally, the thin upper PHS unit was deposited as a final influx of low salinity water filled the chasma. The groundwater upwelling model can explain this sulfate sequence through multiple events of water infilling and with variable geochemistry. Tosca et al. [2008b] report increasing salinity values in order to change PHS to MHS, with a water activity (aH20) near 0.78 for epsomite, 0.62 for hexahydrite, and 0.51 for kieserite.
 Alternatively, the changes in sulfate composition may reflect successively buried layers of dust, ice, and volcanic aerosols laid down over obliquity cycles [Michalski and Niles, 2012]. As the climate changed, sublimation occurred, leaving behind a stack of crudely layered, fine-grained, altered sediments that now compose the ILD. Melting beneath or adjacent to the ice may have resulted in the clays seen at location 2, followed by acidic conditions to form the doublet unit interpreted as leached clay. The observation of darker grains embedded within the upper portions of the kieserite unit at location 3 can be explained by trapping, cementing, and lithifying layers of eolian dust and sand either in the ice or during groundwater upwelling events.
 LT(K) and the upper PHS experienced greater amounts of erosion relative to the lower PHS unit, which is why the upper PHS is only found on the ILD and kieserite is sometimes seen in smaller outlier hills that cap portions of the lower PHS. LTL(K) mapped at location 6 consists of an upper massive kieserite similar in morphology to LT(K) seen throughout Capri Mensa, as well as a lower kieserite with layering. LTL(K) may represent the same unit as LT(K) mapped elsewhere but with layering not commonly seen in LT(K).
 The mix of sulfates seen only at location 1 requires a more detailed explanation. Our observations indicate that at location 1 the dominantly PHS-bearing beds transition smoothly and gradually downward into lower hydration state sulfates that then change downward into mixed MHS. These changes could be consistent with a gradational change in phase with depth due to thermal alteration [Vaniman et al., 2004]. We note that the impact crater at location 1 exposed the thickest portion of the ILD mound, ∼4 km of topography from the top of the ILD to the chasma floor. As the central and thickest portion of the ILD, it is plausible that temperatures and pressures were relatively higher with increasing depth in the ILD, enabling diagenesis due to burial for sulfate conversion [Roach et al., 2009]. Consequently, the PHS originally deposited here may have experienced dehydration and conversion to lower hydration states with depth, which is consistent with our observations.
 However, because we see distinct layers within LTL(MHS) consisting of either kieserite or szomonolkite, it is difficult to explain these alternating Fe-mixtures by conversion of PHS to MHS. Instead, these compositional transitions are more likely to reflect primary features during MHS deposition in waters with fluctuating iron levels. If saline conditions decreased slightly, then MHS would change to lower hydration states of PHS (i.e., LTL(Mix)). Finally, hydration states had to increase during emplacement of LTL(PHS), although pulses of saline-rich waters occurred as evidenced by the MHS layer interbedded within LTL(PHS). All of this may have occurred within a smaller ancestral basin or a depression within the larger chasma because we only see these units within the center of the ILD at location 1. As the basin increased in size or the depression filled with these sulfates, the lower PHS seen along much of the floor was laid down, followed by the kieserite unit LT(K), and then finally the thin upper PHS unit (Figure 23). This scenario represents a simplified and general model to explain the sequence of observed sulfates, but environmental conditions were likely more complex.
 The ice-dust deposition model of Niles and Michalski  can also explain the unique sulfates at location 1, with earlier obliquity cycles creating the central interior sulfates seen only at location 1, while later obliquity cycles and volcanic eruptions laid down the sulfate sequence that now dominates Capri Mensa. There are two observations that could potentially be used to distinguish between the two models. First, in the ice-dust model, as the thick ice deposits sublimated away and weathered to sulfates, melting on the edges and below could create cross-bedded materials at the proximal parts of the ice where sediments eroding out of the ice would be reworked. We did not observe any evidence for cross-bedding in the sulfates, but erosion along the proximal portion of the ILD may have removed any traces of cross-bedding to test this possibility.
 A second observation where a distinction could potentially be made between the two models is the relationship between physical layering and compositional boundaries. In the case of the dust-ice model, the compositional boundaries should follow the layering. In contrast, for the groundwater model where the evaporates cement pre-existing layered material, the compositional boundaries should be controlled by equipotential surfaces (in the case of flooding) or pre-existing topography (in the case of groundwater not breaching the surface) and do not necessarily follow layering. None of our observations show clear instances where compositional boundaries cut across layers. This null observation does not negate the groundwater upwelling model, however, as the compositional boundaries would follow layering if the surface at the time of upwelling was also flat. In summary, from a morphological and geochemical perspective, the ice model and the groundwater models are nearly indistinguishable. In both cases, repeated periods of deposition/evaporation interspersed by erosional periods could explain the observed compositional interbedding and unconformities.
 After deposition of all sulfates, an impact event created the 30-km diameter crater at location 1 to expose the deepest interior portions of Capri Mensa, followed by erosion and eolian activity throughout the chasma. Although we distinguish and propose separate depositional times for units LTL(PHS) and DM(H)+PHS, we cannot rule out the possibility that the two PHS units may be coeval. Their different morphologies (i.e., prominent marker beds seen only in LTL(PHS)) would be one argument against this possibility, however.
4.2. Hematite Formation and Distribution
 A summary of the geologic units we mapped in Capri Chasma and their correlation to gray hematite is listed in Table 2. For several locations, hematite is spatially corrected to dark-toned mantling debris and talus covering underlying sulfates. Hematite detections corresponding to upslope exposures of LT(K) along the ILD indicate that some hematite grains must be embedded within and sourced from this unit. Most hematite spatially correlates to the stratigraphically lower unit DM(H)+PHS, though.
Eolian and floor materials, LT(K), DM(H)+PHS, LT(FHS), D(K)
DM(H)+PHS, possibly LT(K)
Eolian and floor materials, LT(K), DM(H)+PHS, MT(PHS)
DM(H)+PHS, possibly LT(K)
Eolian and floor materials, LTL(K), DM(H), DM(PHS), MT(PHS)
 Two scenarios for hematite distribution are possible. First, small outliers of LT(K) that superimpose DM(H)+PHS indicate that LT(K) was once more extensive across much of this area. Consequently, hematite formed in LT(K) would concentrate along the chasma floor by erosion and removal of LT(K) [Knudson et al., 2007; Chojnacki and Hynek, 2008]. Second, hematite formed within and then subsequently weathered out of DM(H)+PHS with limited contribution of hematite from overlying unit LT(K).
 At Meridiani Planum, orbital data showed the hematite correlated to a light-toned plains units, while in situ observations by the rover revealed the hematite is embedded within a jarosite-bearing sulfate unit [Christensen et al., 2000; Squyres et al., 2004]. In the case of Aram Chaos, the highest concentrations of hematite (16%) determined by Glotch and Christensen  correspond to one or more layers of a 100–150 m thick layer stack that is stratigraphically above lower hematite abundance units. In Aureum Chaos, the hematite spatially correlates to PHS and dark mantling material [Noe Dobrea et al., 2008]. Both Candor and Ophir chasmata have hematite associated with dark mantle deposits adjacent to sulfate-rich ILDs [Weitz et al., 2008; LeDeit et al., 2008].
 Diagenetic alteration of sulfates would be a plausible source for the coarse-grained hematite at Capri Chasma (Figure 24), analogous to the postulated hematite formation in Meridiani Planum sulfates [Squyres and Knoll, 2005; McLennan et al., 2005]. Observations made in situ by the rover suggest the hematite-rich concretions formed diagenetically by stagnant or slowly moving groundwaters within sandy sediments. Groundwater dissolution of jarosite is considered a likely source for the iron that precipitated the hematite [Squyres and Knoll, 2005].
 At Capri Chasma, we have not been able to detect the presence of jarosite in the CRISM data, but CRISM data of the rocks at Sinus Meridiani also do not reveal the presence of jarosite [Wiseman et al., 2010] even though jarosite was measured by the Mossbauer instrument on the Opportunity rover [Klingelhöfer et al., 2004]. Consequently, we cannot rule out the possibility that there may be jarosite mixed in with the monohydrated and polyhydrated sulfates we have identified in the CRISM spectra. If jarosite is present, then it could supply the iron needed to precipiate the hematite [Golden et al., 2008]. Alternatively, iron sourced from another location could be transported within the groundwater that later circulated within the sulfates of Capri Chasma. These Fe-rich waters would enable hematite to precipitate within the Mg-sulfate kieserite of unit LT(K). A final possibility is that the upper and lower PHS units contain iron and this allowed hematite precipitation, with dissolution of the upper MT(PHS) unit liberating iron into the lower LT(K) unit.
 (1) Based upon the stratigraphy exposed along the northern edge of the Capri Mensa ILD, the depositional sequence began with PHS, followed by a thick kieserite unit, and finally a thinner upper PHS unit. If groundwater upwelling is the source of water within the chasma, then a change in salinity due to variable water-to-rock ratios with each upwelling event can explain the observed sulfates. Alternatively, deposition of sulfur rich aerosols and dust particles inside ice deposits during obliquity cycles could lead to weathering and formation of the sulfates.
 (2) Additional sulfates are exposed along the southern wall of an impact crater that corresponds to the thickest and most central portion of the ILD. These sulfates progress from a lower intercalated kieserite and szomolnokite unit, to a middle mixture of intermediate hydration states polyhydrated sulfates, and then to dominantly polyhydrated sulfates. The compositional transitions in iron levels in the lower monohydrated sulfates unit and the general, sometimes fluctuating, progression to higher hydration states moving stratigraphically upward are most consistent with changes in aqueous chemistry during deposition rather than by diagenesis due to burial.
 (3) Other hydrated units have been identified along the chasma floor in association with small mounds near the ILD. Minerals include ferric hydroxysulfate, Fe/Mg-smectites, possible opal, and the evaporitic product of leached clays. These mounds could represent initial deposits that were subsequently eroded, or they could be localized late-stage deposits and alteration of earlier floor materials. Their small size, and in some cases distance away from hematite occurrences, indicates that these rocks are not the source of hematite within the chasma.
 (4) Gray hematite spatially correlates to sulfates and dark mantles overlying sulfates. Kieserite that dominates the rocks exposed along the ILD is one likely source of hematite material. The lower PHS seen along much of the northern chasma floor is another plausible source of hematite.
 (5) Kieserite-rich dunes and ripples occur along ILD slopes and the chasma floor, indicating that some kieserite outcrops erode into sand size particles that can then be mobilized by the wind. Some upper kieserite outcrops in the ILD contain dark material that collects as talus downslope over time as wind erodes the kieserite. This embedded dark material could result from trapping, cementing, and lithifying layers of eolian dust and sand during groundwater upwelling events or within ice deposits.
 (6) Our observations are generally consistent with other occurrences of both hematite and sulfates on Mars, including Meridiani Planum, Aram Chaos, and several other chasmata and chaos terrains within Valles Marineris.
 We thank MDAP grant NNX08AL13G for funding to CMW, MDL, and END to conduct this investigation. Many thanks to D. Berman who produced the HiRISE DTMs. Thanks also to K. Lichtenburg and D. Morris for use of their ferric hydroxysulfate laboratory spectrum. This paper was significantly improved based upon comments by Laetitia Le Deit, Tim Glotch, and Joe Michalski. We are grateful to everyone on the MRO mission, especially HiRISE, CTX, and CRISM teams who targeted and made available the data that was used in this study.