Columbus crater in the Terra Sirenum region of the Martian southern highlands contains light-toned layered deposits with interbedded sulfate and phyllosilicate minerals, a rare occurrence on Mars. Here we investigate in detail the morphology, thermophysical properties, mineralogy, and stratigraphy of these deposits; explore their regional context; and interpret the crater's aqueous history. Hydrated mineral-bearing deposits occupy a discrete ring around the walls of Columbus crater and are also exposed beneath younger materials, possibly lava flows, on its floor. Widespread minerals identified in the crater include gypsum, polyhydrated and monohydrated Mg/Fe-sulfates, and kaolinite; localized deposits consistent with montmorillonite, Fe/Mg-phyllosilicates, jarosite, alunite, and crystalline ferric oxide or hydroxide are also detected. Thermal emission spectra suggest abundances of these minerals in the tens of percent range. Other craters in northwest Terra Sirenum also contain layered deposits and Al/Fe/Mg-phyllosilicates, but sulfates have so far been found only in Columbus and Cross craters. The region's intercrater plains contain scattered exposures of Al-phyllosilicates and one isolated mound with opaline silica, in addition to more common Fe/Mg-phyllosilicates with chlorides. A Late Noachian age is estimated for the aqueous deposits in Columbus, coinciding with a period of inferred groundwater upwelling and evaporation, which (according to model results reported here) could have formed evaporites in Columbus and other craters in Terra Sirenum. Hypotheses for the origin of these deposits include groundwater cementation of crater-filling sediments and/or direct precipitation from subaerial springs or in a deep (∼900 m) paleolake. Especially under the deep lake scenario, which we prefer, chemical gradients in Columbus crater may have created a habitable environment at this location on early Mars.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Impact craters are the most common type of basin on the Martian surface in which ancient water could have ponded. Hundreds of candidate crater paleolakes have been identified based on morphologic evidence such as inlet and/or outlet valleys, fan-shaped (possibly deltaic) deposits, and putative shoreline features [e.g., Forsythe and Blackwelder, 1998; Cabrol and Grin, 1999; Fassett and Head, 2008b]. Because of their potential for habitability and preservation of biosignatures in sediments deposited in a quiescent environment, paleolakes are considered high-priority targets in the astrobiological exploration of Mars [e.g., Farmer and Des Marais, 1999; Ehlmann et al., 2008a].
 A spectral survey covering much of the southern highlands in search of new hydrated mineral exposures [Wray et al., 2009a] identified a unique group of craters in northwest Terra Sirenum that contain Al-phyllosilicates (first reported in Cross crater by Poulet et al. ) and hydrated sulfates in finely bedded deposits. The alteration mineral assemblages in these craters are reminiscent of those associated with terrestrial acid-saline lakes and groundwaters [Benison et al., 2007; Baldridge et al., 2009; Story et al., 2010]. By analogy, the Terra Sirenum crater deposits may be lacustrine evaporites; even if not, their mineralogic and morphologic properties define a distinct class of aqueous deposit on Mars [Murchie et al., 2009b]. Here we investigate in detail the morphology, thermophysical properties, mineralogy, and stratigraphy of these deposits; we then examine hypotheses for their origin to better determine their implications for ancient Martian environments. We focus first on Columbus crater (29°S, 166°W), where the greatest diversity of hydrated minerals is observed, and then look at nearby craters with similar deposits. Key data sets used in this study include the Mars Reconnaissance Orbiter (MRO)'s High Resolution Imaging Science Experiment (HiRISE) [McEwen et al., 2007], Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) [Murchie et al., 2007], and Context Camera (CTX) [Malin et al., 2007], as well as the High Resolution Stereo Camera (HRSC) [Neukum and Jaumann, 2004] on board Mars Express, the Thermal Emission Imaging System (THEMIS) [Christensen et al., 2004a] on Mars Odyssey, and the Mars Orbiter Laser Altimeter (MOLA) [Smith et al., 2001] and Thermal Emission Spectrometer (TES) [Christensen et al., 2001a] on Mars Global Surveyor.
2. Morphology of Columbus Crater
2.1. General Characteristics
 Columbus crater lies in northwest Terra Sirenum in the southern highlands of Mars. Immediately surrounding Columbus are highly cratered plains of the Npl1 unit [Scott and Tanaka, 1986], dated to the Middle Noachian Epoch [Tanaka, 1986]. Fluvial dissection is sparse here compared to other regions of the Noachian southern highlands [Carr, 1995; Hynek et al., 2010], although this may be partly due to poor preservation, as the plains ∼100 km to the east, south, and west of Columbus were largely resurfaced in the Late Noachian by materials in the Np12 unit of Scott and Tanaka .
 Columbus itself has a well-preserved rim, with no breaches by valleys entering or exiting the crater (Figure 1). A few small valleys cut the northeast inner crater wall (Figure 1b), but their alcoves do not extend beyond the crater rim and they probably did not supply substantial fluvial sediment to the crater. The crater diameter is ∼110 km, with a rim crest at 3000 ± 200 m (MOLA elevation relative to the geoid [Smith et al., 1999]): well above the surrounding plains, which span 2200–2700 m in elevation. This rim height is typical for minimally degraded large craters of this size [Garvin et al., 2003]; by contrast, Columbus's flat floor at 920 ± 30 m implies a current depth ∼1.5 km less than fresh craters of this diameter, suggesting substantial infill by sedimentary and/or volcanic materials. Further evidence for significant infill and/or erosion of the crater interior is furnished by the lack of a central peak or peak ring. A single hill complex ∼15 km NNW of crater center (Figure 1) could be a lone remnant of a peak ring (or an off-center peak), although its height (∼700 m above the floor) exceeds that typical for central peaks in craters of this size (∼440 m based on the relation of Garvin et al.  for craters up to 100 km in diameter). These hills expose megabreccia (Figure 2), like many crater central uplifts on Mars [McEwen et al., 2008].
 The flat portions of Columbus's floor have a relatively high thermal inertia (Figure 1b). We derived thermal inertia maps (∼100 m/pixel) from THEMIS nighttime infrared (IR) images [Fergason et al., 2006] using the thermal model of Putzig and Mellon . This method uses THEMIS band 9 (12.57 μm) nighttime brightness temperature images to derive best fit thermal inertia by interpolation with a seven-dimensional lookup table using season, time of day, latitude, thermal inertia, albedo, elevation, and visible dust opacity [Putzig and Mellon, 2007]. Columbus's eastern crater floor exhibits thermal inertia up to ∼510 ± 30 J · m−2 s−0.5 K−1 (units hereafter abbreviated “tiu,” after Putzig and Mellon ). For comparison, thermal inertias >386 tiu were classified as “very high” by Putzig et al.  and likely indicate high rock abundance and/or exposures of bedrock or other indurated surfaces. For example, in the Nili Patera caldera of Syrtis Major, thermal inertias of ∼500 tiu are interpreted as indicative of a relatively fresh lava flow surface underlying minor amounts of unconsolidated material [Fergason et al., 2006].
 High-resolution images of the Columbus crater floor reveal a rugged texture at decameter scales (Figure 3b), with good retention of small craters. Meter-scale boulders are also visible, further suggesting a cohesive surface material that breaks up to form the boulders; this material could be a strongly cemented sedimentary deposit or a lava flow. On the western crater floor, a series of ridges is observed (Figure 3a). These are generally segmented in planform and asymmetric in profile, with one steep scarp (most commonly west facing) and a broader, shallower rise on the other side. This morphology is characteristic of wrinkle ridges, as observed in layered materials (most commonly lava flows) on the Moon and terrestrial planets including Mars [Mueller and Golombek, 2004]. Wrinkle ridges are generally interpreted as blind thrust faults. On Mars, they are typically observed in Early Hesperian ridged plains, interpreted as low-viscosity lava flows [Scott and Tanaka, 1986]. Taken together, the wrinkle ridges and other textural and thermophysical characteristics of Columbus crater's floor are strongly suggestive of a lava flow, although we cannot rule out other possibilities such as lithified aeolian deposits. A 40 km diameter crater that cuts the northwest rim of Columbus (Figure 1) has a similarly flat floor with wrinkle ridges and a relatively high thermal inertia (∼420 ± 30 tiu), suggesting that it contains similar deposits. The source of these putative lavas is not apparent, as is typically the case for ridged plains on Mars [Greeley and Spudis, 1981].
 Other notable characteristics of Columbus crater include a graben cutting east to west across the crater floor, and an 11 km diameter crater whose radial ejecta have darkened a large portion of Columbus's southern floor (Figure 1a). These dark materials (TES albedo ∼ 0.11) form aeolian bed forms and have a thermal inertia (∼250 tiu) consistent with medium sand [Presley and Christensen, 1997], but meter-scale blocks are also present (Figure 3c). This 11 km crater and a 14 km diameter crater on the southwest floor have lobate, single-layer ejecta, which are thought to result from fluidization of subsurface volatiles in the impact target materials [e.g., Barlow and Perez, 2003].
2.2. Light-Toned Layered Deposits
Figure 1a shows outcrops of relatively light-toned materials scattered across the northeast quadrant of Columbus crater's floor, and in a discrete band approximately halfway down its walls: especially the northern and eastern walls. This band of material is more evident in Figure 1b (see also Figure 4a, inset), which shows its higher thermal inertia relative to adjacent crater wall materials, consistent with a more indurated surface. Thermal inertia of the light-toned materials is ∼290 ± 40 tiu, within the range of values found for light-toned layered deposits in Terra Meridiani [Arvidson et al., 2003] and Valles Marineris [Fergason et al., 2006; Mangold et al., 2008; Chojnacki and Hynek, 2008]. These values contrast with thermal inertias of ∼130–190 tiu for the adjacent darker crater wall materials, which are likely surfaces dominated by loose fines.
 The light-toned deposits trace a nearly continuous ring around the eastern, northern, and western crater walls, at a near-constant elevation of 1800 ± 150 m (Figure 1b). At this elevation, crater wall profiles from an HRSC digital elevation model (DEM) show a convex-up break in slope at many (though not all) azimuths (Figure 4). The ring is interrupted by the D ∼14 km crater on the southwest floor, and by the previously mentioned valleys on the northeast wall. On the southeast wall, it is poorly exposed at the resolution of Figure 1, but higher-resolution images (HRSC, CTX, HiRISE) reveal small exposures of light-toned outcrop here as well, albeit largely covered by darker surficial materials. This 360° ring of light-toned, indurated material is perhaps the most striking morphologic aspect of Columbus crater, and to our knowledge it is unique (or at least uniquely well preserved) among Martian craters.
 High-resolution images of Columbus's ring show stratification at meter scales (Figure 5). Beds are typically light-toned, but some darker beds are observed low in the section (Figure 5c) [see also Wray et al., 2009a, Figure 4c]. Successive beds exposed in cross section appear approximately parallel, and to date we have observed no unambiguous angular unconformities, although this is typical for orbital observations even at HiRISE scale (∼30 cm/pixel) and even where decimeter to meter-scale cross beds are known to exist from rover observations [e.g., Grotzinger et al., 2005]. Further insight into bed geometries can be gleaned via strike and dip measurements from a high-resolution DEM. We use a DEM with 1 m grid spacing and vertical precision ∼0.2 m produced from HiRISE images PSP_005429_1510 and PSP_005851_1510 via the techniques of Kirk et al. . These images cover a well-exposed portion of the Columbus wall ring (Figure 5a shows a subset), where some individual beds can be traced across >1 km of outcrop.
 We measured bed strike and dip angles from the DEM using multilinear regression to find the best fit plane through a set of points chosen manually along a bed. For each of twenty beds exposed within the area of Figure 5a, we selected between 8 and 28 points, which in all cases were well fit by a plane, with r2 values of 99.5% or higher. Measured dip angles generally range from 3° to 10°, with a mean of 6.4° (median 6.3°) and typical uncertainties of ∼1.0° (95% confidence level). Dip directions generally range from 125° to 235° clockwise from north, with most uncertainties <10°. Where multiple overlying beds can be measured on the same escarpment, their dips and dip directions are typically consistent with each other within the uncertainties, supporting the inference that bedding is conformable. The mean dip direction at this location is 183°, almost directly due south and toward the crater interior. These dip directions are measured from escarpments facing a range of azimuths (Figure 5a).
 DEM measurements also allow estimation of the total thickness of the Columbus wall ring deposits. Bed dips are used to calculate their true stratigraphic thickness (which is only equivalent to scarp height if bedding is horizontal), yielding an estimate of ∼20 m total thickness of light-toned beds in the deepest exposures.
 Ring bed surfaces have a range of textures, but fracturing is common, typically yielding polygonal “tiles” a few to ∼10 m across (Figures 5b and 6b). Polygon borders are typically darker than their interiors, probably due to shadowing and/or filling of fractures by fine-grained materials darker than the outcrop. However, bright polygon borders are observed in a few outcrops of the darker rocks immediately underlying Columbus's light-toned ring (Figure 7). Relative brightness may indicate that these polygon borders are intrinsically lighter-toned and/or relatively highstanding.
 Immediately downslope (toward the crater interior) from escarpments of Columbus ring materials, light-toned blocks up to several meters across are observed (Figure 6d). Many of these blocks are surrounded by “halos” of light-toned surface material, reminiscent of the “Gray Rock Soil” concentrations observed around “Gray Rocks” by the Imager for Mars Pathfinder [Bell et al., 2002]. Those soils were inferred to be flakes or spalls from the rocks that they surround, subsequently comminuted and mixed with other aeolian fines. In the rocks of the Columbus ring, pervasive fractures may promote physical weathering and rockfall, after which blocks are worn away by aeolian erosion and further weathering.
 Light-toned materials are visible (Figure 1a) not only in the ring around Columbus's walls, but also on parts of the crater floor not covered by the possible lava flows discussed previously, particularly in the crater's northeast quadrant. Figure 8a shows a representative outcrop from this area, with dozens of meter-scale beds exposed in cross section. Strata exposed on the crater floor are typically darker-toned and show less relative contrast than strata in the ring along the crater wall. Boulders sourced from these beds are rare. As with the wall ring deposit, no clear angular unconformities between successive beds are visible, although it is difficult to trace individual beds over long distances due to debris cover and erosional topography.
 Light-toned materials on the western crater floor are most commonly exposed in (typically west facing) escarpments of wrinkle ridges (Figure 3a). Figure 8b shows an example, where light-toned materials crop out beneath darker materials. Other exposures of light-toned material on the crater floor are generally limited to the ejecta and walls of small superposed craters (see section 5), and to a narrow zone at the base of Columbus's crater walls where the younger lava flow may be thin to absent. For example, Figure 6a shows two bright-ringed pits on the northwestern crater floor edge, both filled with dark materials. Color diversity in the light-toned rocks exposed in these pits (Figure 6c) suggests that diverse lithologies may be present. One final location where light-toned materials are found is in the hills ∼15 km NNW of the crater's center (Figure 8c). These hills, which have a fairly sharp crest line, appear to be predominantly composed of dark-toned material along with megabreccia (Figure 2). However, light-toned deposits that are morphologically (and mineralogically) similar to those in the Columbus wall ring are present at a range of elevations, up to the highest peak in the hills (Figures 8c–8d).
3. Mineral Identification and Distribution at Columbus
 To constrain the compositions of materials in Columbus crater (especially the light-toned deposits described in section 2.2) we used orbital infrared spectroscopy. For visible and near-infrared (NIR) wavelengths, multiple scattering dominates the reflected light signal from particulate and textured surfaces, and absorption band strengths often are not linearly proportional to mineral abundances [e.g., Clark, 1999]. This nonlinearity can allow minor mineralogic components to be detected, but it often prevents simple mineral abundance estimates via linear mixture modeling, as is commonly done at thermal infrared wavelengths. Although preliminary estimates of Martian secondary mineral abundances from NIR spectra are intriguing [Poulet et al., 2008b], we limit our scope in this section to simply identifying these minerals using CRISM data; abundances are discussed in section 4 on the basis of thermal emission spectra.
 Most CRISM science observations fall into one of two categories: multispectral survey mapping or hyperspectral targeted imaging [Murchie et al., 2007, 2009c]. In both cases, pixels having spectral absorptions characteristic of hydrated minerals can be mapped using spectral summary (i.e., mineral indicator) parameters [Pelkey et al., 2007]. Figure 9a shows maps corresponding to olivine, Al-phyllosilicates, and hydrated sulfates derived from ∼200 m/pixel multispectral mosaics. Olivine is present in the ejecta of small craters superposed on Columbus's floor (red tones in Figure 9a); in particular, the dark ejecta of the D ∼11 km crater on the southern floor contain the strongest olivine signature observed by CRISM within several hundred kilometers of this location. Al-phyllosilicates (green tones) are present in the light-toned materials widespread across the northeast crater floor, at several locations on the crater walls, and in a narrow strip on the southeast floor at the foot of the crater wall. Weak signatures indicative of hydrated sulfates are detected in some locations on the crater floor (blue tones), but they are most apparent in the light-toned ring around the crater wall, including where this ring extends onto the floor of the D ∼17 km crater shown in Figure 10. For more specific mineral identifications and correlations to surface morphology, we devote the remainder of this section to analysis of CRISM hyperspectral targeted observations.
3.1. Spectral Processing Methods
 CRISM I/F data were processed as described by Murchie et al. [2009c], including division by the cosine of the solar incidence angle and atmospheric removal via division by a scaled transmission spectrum derived from observations over Olympus Mons [McGuire et al., 2009]. Spatial and spectral noise filtering [Parente, 2008] were also applied. Spectra from many pixels were averaged to improve the signal-to-noise ratio (SNR), and the resulting average spectra were divided by a spectral average from a dusty or otherwise spectrally “neutral” region in the same CRISM scene. This spectral ratio method suppresses residual artifacts of instrument calibration and atmospheric removal [e.g., Mustard et al., 2008] while accentuating spectral signatures in the numerator spectrum that are unique relative to the denominator. Known artifacts that remain in some ratio spectra include a discontinuity near 1.65 μm due to a detector filter boundary and small features near 2.0 μm resulting from imperfect removal of atmospheric CO2 bands [Murchie et al., 2009c]; we have masked some of these known “bad bands” in the spectra plotted here.
 CRISM has a VNIR detector spanning ∼0.4–1.0 μm and an IR detector spanning ∼1.0–4.0 μm. Except for section 3.4, we focus on IR detector data because this wavelength range is less affected by ferric minerals in rock coatings and surficial dust [Swayze, 2004; Cloutis et al., 2006], and it includes diagnostic absorptions for mafic minerals, carbonates, and hydrated or hydroxylated minerals including sulfates and phyllosilicates [e.g., Ehlmann et al., 2009]. We devote most of our attention to the region from 1.0 to 2.6 μm, as beyond 2.6 μm CRISM data have lower SNR and several known instrument artifacts [Murchie et al., 2009c]. In addition, wavelengths longer than 3 μm typically include a contribution from thermal emission that reduces absorption band strengths and we have not corrected the data for these thermal effects.
 We have processed and analyzed all CRISM targeted observations of Columbus crater acquired to date, including FRTs (18 m/pixel, covering 10 km × 10 km), HRLs (36 m/pixel, 10 km × 20 km) and one HRS (36 m/pixel, 10 km × 10 km). The locations of these observations are indicated in Figure 9b, along with the secondary minerals identified in each location. Minerals were identified by examining maps of the relevant spectral summary parameters defined by Pelkey et al. , Roach et al. , and Ehlmann et al. , and then plotting CRISM spectra against laboratory spectra to confirm detections. For one CRISM observation with especially strong spectral signatures (FRT00007D87), we also used the USGS Tetracorder system [Clark et al., 2003] to search for additional spectral phases that might have been missed by our manual approach.
 In the following subsections, we provide our rationale for each mineral identification and present representative spectra. We also describe where each mineral is found within the crater.
3.2. Aluminum Phyllosilicates
 As mentioned by Wray et al. [2009a] and Murchie et al. [2009b], the most commonly detected phyllosilicate type in Columbus crater is the kaolin group (Figure 11). This group of Al2Si2O5(OH)4 polymorphs includes kaolinite, halloysite, and the less common minerals dickite and nacrite. Kaolin group clay minerals are spectrally distinct from Al-smectite clays such as montmorillonite and beidellite: the former exhibit doublets near 1.4 and 2.2 μm, whereas the latter exhibit single absorptions at these wavelengths as well as the H2O band near 1.9 μm [e.g., Clark et al., 1990; Bishop et al., 2008]. Halloysite can be difficult to distinguish from kaolinite when the latter is mixed with another hydrated mineral, but dickite and nacrite are distinguished by their more symmetric and narrower 2.2 μm doublet absorptions [Ehlmann et al., 2009]. The kaolin group clay spectra in Columbus crater are most consistent with kaolinite or halloysite.
 Tetracorder analysis of FRT00007D87 identifies not only kaolinite, but also the Al-smectite montmorillonite in some outcrops. Montmorillonite, especially when mixed with kaolinite, can be challenging to distinguish from halloysite, poorly crystalline kaolinite, or other kaolinite-smectite mixtures, because all of these have a less distinct 2.2 μm doublet than well-crystalline kaolinite. Perhaps the most diagnostic effect of adding montmorillonite to kaolinite is the broadening of the long-wavelength edge of the 2.2 μm absorption [McKeown et al., 2010; Clark et al., 2003, Figure 13a]. Spectra from locations in Columbus crater mapped by Tetracorder as montmorillonite-bearing indeed have a broad rise from 2.20 to 2.27 μm and a weak or absent 2.16 μm kaolinite doublet feature (bottom CRISM spectrum in Figure 11, top), consistent with montmorillonite being the spectrally dominant phase at these locations.
 Aluminum phyllosilicates are found in every CRISM observation of Columbus except for one covering the D ∼11 km crater on the southern floor. Kaolinite is found on both Columbus's walls and floor. Figure 12 shows a perspective view of the northern crater wall, in which the green areas contain Al-phyllosilicates. These phyllosilicates are seen directly adjacent to (and, for the most part, stratigraphically beneath) Columbus's light-toned ring deposit, as well as several km upslope and downslope. At this particular location, where montmorillonite was mapped independently from kaolinite, the strongest kaolinite signatures are adjacent to the ring deposit, whereas the largest montmorillonite-bearing exposures occur farther from the ring. However, the two Al-phyllosilicates are typically mixed in Columbus crater, with evidence for a kaolinite component in all exposures.
 Many crater wall exposures of Al-phyllosilicate are morphologically unremarkable (e.g., dark materials beneath the light-toned layers in Figure 5b), but in some cases these exposures exhibit stratification (e.g., dark layers in Figure 5c) and/or fracture patterns (Figure 7). In rare instances, Al-phyllosilicates are interbedded with the lighter-toned rocks of the wall ring (Figure 5c) [see also Wray et al., 2009a]. This interbedding likely reflects changing depositional environments or sediment sources, although it could alternatively result from in situ alteration that was strongly controlled by variations in porosity, permeability, and primary mineralogy between strata.
 Other sulfates, including Mg-sulfates, have less diagnostic near-IR spectral characteristics. Spectra with absorptions only at 1.43, 1.93 μm and an inflection at ∼2.4 μm (top two CRISM spectra in Figure 13, top) are commonly interpreted as polyhydrated sulfates [e.g., Gendrin et al., 2005], with the 2.4 μm feature attributed to an S–O overtone and/or OH/H2O-related absorption(s) [Cloutis et al., 2006]. However, caution is warranted because some nonsulfate hydrated salts [Crowley, 1991; Lane et al., 2008; Hanley et al., 2010] and some zeolites (e.g., thomsonite) [Ehlmann et al., 2009] have a similar feature at 2.4 μm. In the case of Columbus crater, the occurrence of this hydrated phase with gypsum and other sulfates described below suggests it is likely a polyhydrated sulfate, or possibly a hydrous chloride salt.
 NIR spectroscopy alone does not always allow unique identification of the cation(s) in polyhydrated sulfates, but this technique does provide some constraints. Ca and Na are unlikely because (like gypsum) bassanite (CaSO4 · 1/2H2O) and mirabilite (Na2SO4 · 10H2O) have strong bands near 1.75 μm [Crowley, 1991] that are absent from spectra of Columbus's nongypsum polyhydrate. Eugsterite (Na4Ca[SO4]3 · 2H2O) lacks a strong 1.75 μm band but absorbs at 2.48 μm [Crowley, 1991], a wavelength distinctly longer than the 2.4 μm band in our spectra. Mg and Fe are therefore the most geologically plausible candidate cations if this phase is indeed a sulfate. Either or both may be present, but we favor at least some Mg because all Fe-sulfates have a broad absorption centered near 1 μm (centered at 0.9–1.2 μm for Fe2+ or 0.8–0.95 μm for Fe3+) [Burns, 1993; Crowley et al., 2003; Cloutis et al., 2006; Lane et al., 2008]. These absorptions are not apparent in CRISM IR detector spectra (Figure 13), nor in VNIR detector spectra of Columbus's polyhydrate-bearing materials (section 3.4) [see also Murchie et al., 2009b]. Furthermore, Mg-sulfates are the most abundant salts in Meridiani bedrock [Clark et al., 2005] and in Martian soils and rock coatings at all landing sites prior to Phoenix [Vaniman et al., 2004]; they have now also been identified in Phoenix soils [Kounaves et al., 2010]. They are a major component of secondary mineral assemblages produced in laboratory experiments [Tosca et al., 2004] and geochemical models [Tosca et al., 2005] of olivine-bearing rock alteration under Mars-like conditions. Nevertheless, the lack of a strong ∼1 μm absorption in CRISM spectra is not in itself sufficient to rule out Fe-sulfates; in fact, we identify Fe-sulfates elsewhere in Columbus crater (sections 3.5 and 3.6) even though no ∼1 μm band is apparent in those cases. Therefore, we refer to this nongypsum sulfate generically as polyhydrated Mg/Fe-sulfate(s).
 The sulfate hydration state is also not well constrained by the CRISM spectra. We plot the spectrum of hexahydrite (MgSO4 · 6H2O) in Figure 13, but it is quite similar to spectra of epsomite (MgSO4 · 7H2O) pentahydrite (MgSO4 · 5H2O), and starkeyite (MgSO4 · 4H2O) [Crowley, 1991], making these sulfates difficult to distinguish with CRISM. However, the monohydrate kieserite is spectrally distinct [Cloutis et al., 2006] and inconsistent with the CRISM spectra in Figure 13. Analogously, polyhydrated Fe2+-sulfates including melanterite (FeSO4 · 7H2O) have spectra similar to rozenite (FeSO4 · 4H2O) shown in Figure 13 [Bishop et al., 2004], but the monohydrate szomolnokite is distinctive (section 3.5). In any case, these sulfates may have experienced hydration state changes since their formation [Vaniman et al., 2004].
 Polyhydrated sulfates are identified in every CRISM observation of the Columbus crater walls and in some observations of the northeast crater floor. They are the spectrally dominant phase in the finely bedded, light-toned deposits ringing Columbus's walls (e.g., Figures 5a–5b), and in a few comparably bright-toned outcrops at lower elevations within the crater (e.g., Figures 8c and 14). The Mg/Fe-sulfate is ubiquitous in these materials, with varying contributions from gypsum. Spectral parameters can be used to map the two polyhydrates independently (Figure 15a). However, inspection of colocated images reveals no clear stratigraphic relationship between the relatively gypsum-rich outcrops and gypsum-poor outcrops. In some cases, the outcrops with gypsum appear relatively highstanding, darker, and more rugged than those without gypsum (Figure 6b).
 Monohydrated sulfates such as kieserite (MgSO4 · H2O) and szomolnokite (FeSO4 · H2O) are distinguished from more hydrated sulfates by a broad absorption that is deepest near 2.1 μm. Natural samples of kieserite have the band minimum at 2.13 μm, while for szomolnokite it occurs at 2.09–2.10 μm [Crowley et al., 2003; Cloutis et al., 2006; Bishop et al., 2009]. However, pure synthetic kieserite has been observed to have a shorter-wavelength minimum coincident with that observed for szomolnokite, potentially making it difficult to distinguish between these minerals [Milliken, 2006]. Only a few other minerals have a similarly broad absorption near 2.1 μm, including some NH4-bearing minerals [e.g., Bishop et al., 2002a], but these lack the 2.4 μm absorption typical of monohydrated sulfate spectra and/or have additional absorptions not observed in hydrated sulfates (Figure 16). With band minima at ∼2.11 and ∼2.40 μm, the Columbus crater spectra in Figure 16 appear most consistent with monohydrated sulfate.
 A weaker, narrow absorption at 2.22–2.23 μm is also observed in the spectra of Figure 16 (top), especially in the bottommost spectrum. A comparably narrow absorption at 2.23–2.24 μm has been observed in light-toned layered deposits on the plains west of Juventae Chasma [Milliken et al., 2008; Bishop et al., 2009], in Aram Chaos [Lichtenberg et al., 2010], and in Cross crater (G. A. Swayze et al., manuscript in preparation, 2010). This absorption has been attributed to Fe3+-OH in the hydroxylated ferric sulfate Fe(OH)SO4, a phase that has been formed in the laboratory at temperatures >200°C via dehydration of ferric sulfates such as hydronium jarosite [Swayze et al., 2008a] or (ferri)copiapite [Milliken et al., 2008; Bishop et al., 2009], or via oxidation and dehydration of ferrous sulfates such as melanterite or szomolnokite [Morris et al., 2009; Lichtenberg et al., 2010]. In Cross crater (a mere ∼400 km from Columbus crater in northwest Terra Sirenum (see also sections 3.6 and 6.1)) the 2.23 μm band has been found to date only in association with monohydrated sulfate (G. A. Swayze et al., manuscript in preparation, 2010), consistent with where it is found in Columbus crater. If the monohydrate in both craters is (at least partially) szomolnokite, then the 2.23 μm band could imply partial oxidation and dehydration of szomolnokite to form Fe(OH)SO4.
 The strongest monohydrate + Fe(OH)SO4 signature yet observed in Columbus crater (bottom spectrum in Figure 16, top) is found in a D ∼200 m crater on Columbus's northeast floor (Figure 17a). Radial rays attest to the relative freshness of this impact crater, and exposures within the crater reveal that it excavated light-toned layered deposits. Several thin beds appear green in HiRISE enhanced color images (Figure 17b), but their exposures are too narrow to be resolved by CRISM. While this color is rare in Columbus crater and in HiRISE enhanced color images of Mars in general, we have observed it in other locations where ferric sulfates are detected from orbit (Aram Chaos and the plains surrounding Valles Marineris). HiRISE IRB color composites [McEwen et al., 2010] display IR (∼875 nm), RED (∼700 nm), and BG filter (∼500 nm) images [McEwen et al., 2007] in the red, green, and blue channels, respectively; therefore, a green hue indicates high RED I/F relative to the IR and BG, as would be expected for a ferric mineral such as Fe(OH)SO4 due to its strong electron charge transfer absorption at <530 nm and spin-forbidden crystal field transition absorption at 800–970 nm [Milliken et al., 2008; Lichtenberg et al., 2010] reducing reflectance in the BG and IR filters, respectively. Thus the green color is consistent with our CRISM-based inference of a ferric mineral such as Fe(OH)SO4.
 Most monohydrate exposures in Columbus are found on the northeast crater floor, although one outcrop has been identified at the base of the hills on the central floor (Figure 8c). Monohydrate-bearing outcrops exhibit internal stratification (Figures 8a and 17c) and, in comparison to beds within the polyhydrate-bearing crater wall ring (Figure 5), monohydrate beds are somewhat darker-toned, with weaker albedo contrasts between successive beds. The fracture patterns observed in the polyhydrate ring are less common in monohydrate-bearing outcrops, which in some cases display a “scalloped” or “reticulate” texture (Figure 17d) reminiscent of that seen on some wind-eroded surfaces elsewhere on Mars [Bridges et al., 2010]. This texture is specifically associated with monohydrated sulfate in other regions [Chojnacki and Hynek, 2008; Karunatillake et al., 2009; Lichtenberg et al., 2010]. These morphologic characteristics are shared by the majority of light-toned outcrops on Columbus's floor, many of which are unresolved or heretofore unobserved by CRISM.
 Jarosite has a nearly unique absorption at ∼2.265 μm, with additional absorptions at ∼1.5, 1.85, 2.51, and 2.62 μm [Crowley et al., 2003; Bishop and Murad, 2005; Cloutis et al., 2006; Swayze et al., 2008a]. K-jarosite has an additional band at ∼2.21 μm that is weak to absent in Na- and H3O-jarosites [Cloutis et al., 2006; Swayze et al., 2008a]. We observe one location in Columbus crater whose spectrum exhibits all of these absorptions except the ∼1.5 μm band(s), and also has a ∼1.93 μm band attributed to H2O (common in lab spectra of jarosites formed at low temperature). The 1.85, 2.51, and 2.62 μm absorptions are near the noise level but are present in both CRISM observations covering the location of interest (Figure 18). The 2.21 μm band observed in the CRISM spectra is most consistent with K-jarosite, similar to the Mawrth Vallis jarosite reported by Farrand et al. ; however, in all K-jarosite lab spectra the 2.21 μm band is significantly weaker than the 2.265 μm band, so their comparable strength in the spectrum from HRL00008565 (Figure 18) may indicate an additional absorber at ∼2.2 μm (e.g., Al-phyllosilicate). Additional hydrous minerals could also contribute to the observed 1.93 μm band, but to the extent that this band is due to H2O in jarosite, its strength suggests a relatively low formation temperature and minimal subsequent recrystallization [Swayze et al., 2008a]. Similarly, a low-temperature formation has been inferred from the spectrum of alunite in Cross crater [Swayze et al., 2008b] and for jarosite found on the plains surrounding Valles Marineris [Milliken et al., 2008].
 To our knowledge, the only common mineral other than jarosite with an absorption at ∼2.27 μm is gibbsite (Al[OH]3) [Cloutis and Bell, 2000]. While a mixture of gibbsite + montmorillonite could account for the major absorptions in our Figure 18 CRISM spectra, such a mixture would not reproduce the weak feature we observe at 1.85 μm. Gibbsite also has a very different spectral shape from 2.1 to 2.5 μm, and strong bands near 1.5 μm that are not observed in these CRISM spectra. The presence of many other sulfates in Columbus crater as well as alunite in nearby Cross crater further supports the identification of jarosite in Columbus. Alternatively, an acid-sulfate environment could have induced partial acid weathering of Fe/Mg-clay minerals (see section 3.7), resulting in silica formation and the appearance of a 2.21/2.28 μm doublet [Madejová et al., 2009]. Again, however, this would not explain the weak feature at 1.85 μm in our CRISM spectra, which is most consistent with jarosite. The one exposure of jarosite found to date is on Columbus's northeast floor; it exhibits internal bedding and lies adjacent to a circular depression containing lighter-toned polyhydrate-bearing outcrops (Figure 14).
 Alunite is distinguished by a strong, broad absorption centered at 2.17 μm, with additional bands at 1.43–1.44, 1.47–1.49, 1.76, 2.32, and 2.51–2.53 μm (those with a range of positions are at longer wavelengths in Na-alunite than in K-alunite) [Bishop and Murad, 2005; Cloutis et al., 2006]. In Cross crater, alunite has been identified as the spectrally dominant phase in some outcrops, and in others it is mixed with kaolinite [Swayze et al., 2008b]. In Columbus crater, several relatively small (up to ∼1 km wide) outcrops have spectra consistent with a contribution from alunite (Figure 19), although the signatures are not as strong as in Cross crater, and all are probably mixtures with Al- (and possibly Mg-) phyllosilicates. To varying degrees, these spectra contain absorptions at the six positions described above for alunite, and the band centers at 1.44 and 1.49 μm in some spectra are most consistent with Na-alunite. The relative weakness of the 1.76 and 2.32 μm absorptions and the presence of a band at ∼1.93 μm (if the latter is due to H2O in alunite) are most consistent with laboratory spectra of alunite formed at low temperature, similar to the alunite in Cross crater [Swayze et al., 2008b].
 It is worthwhile to consider whether the CRISM spectra in Figure 19 could alternatively be explained by mixtures of minerals identified elsewhere in Columbus. For example, gypsum mixed with kaolinite could explain bands at 1.44, 1.49, and 1.75 μm (gypsum) and a shoulder at 2.32 μm (kaolinite). However, the 2.2 μm doublet of kaolinite (or singlet of montmorillonite) cannot account for the 2.17–2.18 μm band minimum in some CRISM spectra shown in Figure 19. Aside from alunite, the phyllosilicates pyrophyllite [Clark et al., 1990] and beidellite [Kloprogge, 2006; Bishop et al., 2010] have bands centered at 2.17 and 2.18 μm, respectively. Beidellite has been previously identified elsewhere on Mars [Noe Dobrea et al., 2010]. However, the absorptions in beidellite and especially in pyrophyllite are too narrow to account for the broad band observed in some of our CRISM spectra, even if mixed with kaolinite and/or gypsum (Figure 19). Furthermore, linear mixtures of gypsum, kaolinite, and beidellite do not have absorptions with the same proportionate strengths or precise wavelengths as in alunite or as observed in the CRISM data, although intimate mixtures may have slightly different spectral properties. From spectral evidence combined with the proximity of Columbus crater to the most definitive alunite detection on Mars [Swayze et al., 2008b], we infer that an alunite component is plausible for some outcrops in Columbus.
 The outcrops with a possible alunite component are on the northeast floor of Columbus crater and on the northeast wall just below the polyhydrate-bearing ring (Figure 20a). Morphologically, these outcrops are distinguished by their smoothness at meter scales (Figures 20b–20c); in contrast to other hydrated mineral-bearing outcrops in Columbus, these strata appear more massive and lack fractures. Bedding is exposed along the edges of the alunite-bearing outcrops, but it is unclear whether all of these beds contain alunite. Outcrops of alunite-bearing material in Cross crater appear similarly smooth at meter scales (G. A. Swayze et al., manuscript in preparation, 2010).
3.7. Iron/Magnesium Phyllosilicates
 Although Al-phyllosilicates are the spectrally dominant alteration phase in most CRISM scenes in Columbus crater, Fe/Mg-phyllosilicates are the most common alteration product detected from orbit in most other regions on Mars [e.g., Bibring et al., 2006; Mustard et al., 2008]. Fe/Mg-phyllosilicates have absorptions at 2.28–2.35 μm, which in smectites occur shortward of 2.32 μm (the exact position depending on Fe versus Mg content) and with a ∼1.9 μm H2O band [e.g., Clark et al., 1990; Bishop et al., 2002b; Swayze et al., 2002]. To date, we have identified ∼10 relatively small areas on the walls and floor of Columbus that are spectrally consistent with Fe/Mg-phyllosilicates, possibly including smectites (Figure 21). Specifically, Fe/Mg-phyllosilicates are exposed in the southern wall of the D ∼17 km crater shown in Figure 10 (CRISM HRL00013FF5), in hectometer-scale resistant knobs on Columbus's northwest wall (FRT00013D1F), in materials eroding from the hills on the central floor (FRT0001663B), and in areas up to a few kilometers wide on the crater floor (e.g., HRL000062B6). Some of these materials may predate the formation of Columbus crater, but the crater floor deposits likely represent infilling materials that postdate the impact event.
 The spectra that we classify as Fe/Mg-phyllosilicates are diverse, with varying relative strengths of the 1.9 and 2.3 μm bands; the position of the latter band ranges from 2.29 to 2.32 μm. A 1.39 μm band is observed in some cases, consistent with Mg-rich phyllosilicates [Clark et al., 1990; Bishop et al., 2002b]. In most cases, a mineral identification more specific than “Fe/Mg-phyllosilicate” is not possible.
3.8. Other Hydrated Phases
 Still other locations on the walls and floor of Columbus crater have spectral absorptions at ∼1.4 and ∼1.9 μm consistent with hydrated minerals, but they lack other strong, diagnostic absorptions that would enable specific identification (Figure 22). In some cases, weak features in the 2.2–2.3 μm range are likely due to metal-OH vibrational absorptions. In particular, spectra from the D ∼11 km crater on Columbus's southern central floor (CRISM FRT0000ABF2) exhibit weak features at 2.19–2.20 and 2.27–2.28 μm (Figure 22). These wavelengths are slightly too short and too long, respectively, for jarosite (section 3.6). They are somewhat reminiscent of the 2.21/2.27 μm doublet feature observed by Roach et al. [2010a] in spectra from Valles Marineris, although the features we observe are much weaker. Roach et al. [2010a] attributed this doublet to either a mineral mixture or a poorly crystalline Fe/SiO2-bearing phase similar to that described by Tosca et al. [2008b], formed via acid weathering of Fe-bearing clays [Madejová et al., 2009].
3.9. Phases Not Observed: Carbonate, Chloride, Zeolite, Prehnite
 As described above, Columbus crater contains a wealth of phyllosilicate and sulfate minerals not commonly observed on Mars. However, several types of secondary minerals detected elsewhere on Mars by CRISM and THEMIS have not been found to date in Columbus crater. These include salts such as carbonates [Ehlmann et al., 2008b] and chlorides [Osterloo et al., 2008] as well as hydrated silicates that form under alkaline (zeolites) and/or high-temperature conditions (prehnite) [Ehlmann et al., 2009]. The potential absence of these minerals in Columbus crater would be consistent with an alteration environment of relatively low temperature and low-to-neutral pH. However, nondetection of a mineral via orbital spectroscopy does not necessarily imply the absence of that mineral [e.g., Kirkland et al., 2003].
4. Constraints on Modal Mineralogy
 Thermal emission spectra of Columbus crater enable an independent assessment of the surface mineralogy. In particular, the ∼100 m/pixel THEMIS data set is ideal for studying the small-area outcrops in Columbus crater. We have analyzed the highest-quality THEMIS observation of Columbus available to date, I07746002 (Figure 23a), which covers a ∼30 km swath across the crater, including the well-exposed sulfate-bearing ring of material on the northern crater wall.
 Emissivity spectra of the dark materials covering most of Columbus's floor (Figure 23b) are similar to that of TES Surface Type 1 (ST1; Figure 23c), the dominant spectral unit in the Martian southern highlands [Bandfield et al., 2000] that is generally interpreted as representing a basaltic composition. Spectra of the plains outside Columbus and of ejecta surrounding the D ∼11 km crater on the southern central floor are also consistent with basalt, but with a stronger absorption at ∼11 μm (THEMIS band 7). This feature, which is especially strong in the D ∼11 km crater ejecta, suggests higher olivine abundance in the ejecta relative to the rest of Columbus's floor [e.g., Hamilton and Christensen, 2005]. This is consistent with the detection of olivine in this small crater's ejecta by CRISM (Figure 9a).
 Decorrelation stretch (DCS) images show that the sulfate-bearing ring is spectrally distinct from the adjacent wall materials (Figure 23a). Where this spectral distinction is strongest (“wall lower” spectrum in Figure 23b), the slope from 9.4 to 11 μm (1070 to 910 cm−1) is greater than that seen in ST1, suggesting a greater abundance of high-silica phases that could include phyllosilicates (Figure 23c). In addition, the “wall lower” spectrum has an absorption at 8.6 μm (band 4, 1170 cm−1), consistent with the presence of sulfates. In particular, absorptions at this relatively short wavelength are most consistent with water-poor sulfates, e.g., kieserite or sanderite (MgSO4 · 2H2O), but not the more hydrated Mg-sulfates [Baldridge and Christensen, 2006; Lane, 2007]. Ca-sulfates including gypsum (which CRISM detects in Columbus's polyhydrate ring) would also be consistent with this 8.6 μm feature [Christensen et al., 2004a, Figure 7a]. In either case, this feature provides independent support for our CRISM detections of sulfates in Columbus crater.
 Linear mixing models of higher spectral resolution thermal emission data can be used to estimate mineral abundances with ∼5–15% precision [e.g., Ramsey and Christensen, 1998; Feely and Christensen, 1999]. The TES instrument is ideal for this, although its ∼3 km resolution is coarse compared to most outcrops in Columbus. We modeled a TES spectrum extracted from the “wall lower” location in Figure 23 using the standard ASU mineral library including smectites [Rogers et al., 2007], supplemented with Mg-sulfate spectra measured by Baldridge and Christensen . The results yield an estimate of roughly 40% phyllosilicates by volume, 16% hydrated sulfates, 15% olivine, and the balance in feldspars. The modeled sulfate abundance does not substantially exceed the TES detection limit of ∼10–15% [Christensen et al., 2001a]. However, the modeled spectrum was extracted from an area that includes several distinct spectral units at CRISM resolution (sulfate-bearing versus clay-bearing versus nonhydrated), so abundances of sulfates and phyllosilicates are likely higher within the light-toned outcrops specifically.
5. Stratigraphy and Chronology at Columbus
Sections 2 and 3 have described the diversity of deposits in Columbus crater. Here we describe their stratigraphic relationships and use crater counting to estimate the ages of some events in Columbus's geologic history. The diverse hydrated minerals in Columbus crater may have formed during numerous alteration events spanning significant time or during a single, geologically brief period of aqueous activity. Exposures in the walls of small craters (e.g., Figure 5c) and other steep scarps provide some insights, including the significant observation of polyhydrated sulfate-bearing beds alternating with kaolinite-bearing beds [Wray et al., 2009a, Figures 4c, DR6, DR7]. However, in many cases the stratigraphic relations are less clear; for example, Figure 14 shows polyhydrated sulfates on the crater wall, jarosite-bearing beds on the crater floor, and additional polyhydrated sulfates at still lower elevations in a depression adjacent to the jarosite. But does this topographic distribution indicate a period of jarosite formation separating two distinct periods of polyhydrate formation? Or did the jarosite form first, followed by a single period of polyhydrate formation on the crater walls and in local depressions on the crater floor? Or could the jarosite have formed diagenetically after the polyhydrates were precipitated/deposited, as in some acid saline lakes on Earth [Benison et al., 2007]? These questions are difficult to resolve based on orbital imagery alone, preventing us from constructing a simple stratigraphic column to compare to theoretical evaporite sequences [e.g., Tosca et al., 2008a; Altheide et al., 2010a].
 One constraint on the timing of aqueous activity is provided by Figure 10, which shows a D ∼17 km crater superposed on Columbus's southwest wall. The crater's flat floor and strongly degraded rim contrast with the similarly sized crater to its northeast, which retains its central peak, implying that the 17 km crater's interior has experienced significant infilling and/or erosion; the materials exposed on its modern floor therefore postdate the crater. The crater floor is at the elevation of the polyhydrate ring, and polyhydrated sulfates are indeed found in layered deposits on its floor. This implies that aqueous activity postdated the formation of the 17 km crater; therefore, at least some aqueous activity must have occurred some time after the formation of Columbus crater (i.e., not all aqueous activity predated or coincided with crater formation).
 Other observations show that the dark deposit covering much of Columbus's floor (which we have argued to be consistent with lava (section 2.1)) postdates the light-toned deposits containing hydrated minerals. As described in section 2.2, the probable lava overlies light-toned materials on the northwest crater floor (Figure 8b), and CRISM FRT00005AA4 shows these light-toned materials to be hydrated (Figure 22). Images of the northeast crater floor show the lava embaying mesas of light-toned material (e.g., Figure 24a). In addition, the D ∼11 km crater on Columbus's southern central floor exposes decameters of stratigraphy in its upper walls, with light-toned beds overlain by darker, relatively blue beds (Figure 24b). Boulders are eroding from the darker beds, which have a rougher texture and (according to CRISM) an enhanced olivine signature; we interpret these darker beds as lavas, possibly olivine-bearing basalt flows. By contrast, erosion of the lighter-toned beds appears to yield finer-grained, hydrated material that is transported downslope to form scree deposits (Figure 25). These and other probable colluvial materials mantle the lower crater walls and conceal the >1 km thickness of underlying deposits inferred to occupy Columbus's floor (section 2.1). Overlying the dark-toned beds in Figure 24b is a somewhat lighter-toned, smooth-textured deposit with sparse boulders <1 m in diameter and a relatively weak olivine (or other ferrous mineral) spectral signature. These characteristics are consistent with those of the Late Hesperian-aged “Electris deposits” of Sirenum Fossae as described by Grant et al. , who interpreted them as probable aeolian loess. From an HRSC DEM, we estimate a crater wall slope of ∼20° at the location of Figure 24b, so the lateral extent of the deposits indicates a vertical thickness of ∼10 m hydrated layered deposits, ∼15 m olivine-bearing lava, and ∼20 m Electris-like deposits at this location. While we cannot exclude the alternative that some of these layers are stratigraphically inverted ejecta from the D ∼11 km crater itself, they occur some ∼200 m below the crater rim crest and their stratigraphy is consistent with that observed elsewhere in Columbus.
 As discussed in section 2.1, a graben cuts the floor of Columbus crater approximately in half (Figure 1). The ENE–WSW orientation of this graben is similar to that of the larger Memnonia and Sirenum Fossae to the north and south of Columbus, respectively. These graben systems are in turn part of a hemisphere-wide collection of structures oriented radially to Tharsis, which formed over a large span of Martian history [Plescia and Saunders, 1982; Anderson et al., 2001]. Some graben in Terra Sirenum may date to the Noachian “stage 1” of tectonic activity, but many likely date to the Late Noachian/Early Hesperian, overlapping wrinkle ridge formation in the Early Hesperian Epoch (“stage 3” of Anderson et al. ).
 The graben in Columbus crater appears to predate the light-toned layered deposits, as revealed by the stratigraphy surrounding a D ∼2.5 km crater on the western floor (Figure 26a). Mesas surrounding this crater consist of light-toned layered deposits overlain by darker materials (Figure 26b) and extend to a height of ∼100 m above the surrounding plains, twice the typical rim height for a fresh crater of this size [Garvin et al., 2003]. These mesas are therefore probably not composed solely of crater ejecta; instead, the uppermost dark materials may be ejecta that armored underlying light-toned layered deposits against an erosional process that stripped the deposits from adjacent terrain [e.g., McCauley, 1973]. Neither the D ∼2.5 km crater nor the mesas surrounding it are visibly deformed by the graben, suggesting a sequence of (1) graben formation, (2) layered deposit formation, (3) formation of the 2.5 km crater, and (4) erosion of adjacent layered deposits. Lava deposition is inferred to postdate these events because the probable lavas embay the erosional remnant mesas. Lava would be expected to fill and bury a preexisting graben, so the fact that the graben remains visible may be explained by its later reactivation following lava flow emplacement, as evidenced by the graben's disruption of some small craters on the lava flow surface (Figure 26c).
 As mentioned in section 2.1, a few poorly developed valleys are visible on the northeast wall of Columbus (Figure 1b). The crater walls and floor near the mouths of these valleys host a dark-toned deposit with relatively low thermal inertia (∼230 ± 10 tiu) and a spectrum dominated by broad bands near 1 and 2.1–2.2 μm, consistent with a basaltic composition [e.g., Mustard et al., 1997]. This dark deposit overlies the light-toned deposits on this part of the crater floor and has partially buried the sulfate-bearing ring on the crater walls (Figure 14a). These superposition relationships suggest that gradation of the crater wall likely occurred here after the period during which aqueous minerals and light-toned deposits formed in Columbus. The dark-toned deposit may have preferentially armored light-toned deposits on this part of the crater floor against erosion. Columbus crater lies at the southern edge of the latitude band (18–29°S) in which Moore and Howard  identified alluvial fans dating to the Noachian-Hesperian boundary in 18 middle-sized to large craters. The dark deposit in northeast Columbus could be a highly degraded alluvial fan. A deltaic interpretation seems less plausible given the lack of clear stratal geometries or clay minerals; for comparison, two of the most likely deltaic deposits identified on Mars contain phyllosilicates and exhibit clear bedding [Ehlmann et al., 2008a; Milliken and Bish, 2010].
 For an estimate of the absolute age of Columbus and its interior deposits, we now consider counts of superposed impact craters. Columbus's interior and proximal ejecta represents a small area for counting craters (∼17,000 km2), so the formal statistical uncertainties (which are proportional to 1/√(n), where n is the number counted) are large. In the notation of Tanaka , in which the measured crater densities are scaled to an area of 106 km2, Columbus has N(16) = 120 ± 85 craters > 16 km in diameter per 106 km2, and N(5) = 420 ± 160. These values are most consistent with a Middle Noachian age for Columbus, although a Late Noachian age (that of the Upper Noachian units described by Tanaka ) is within the uncertainties. As stated in section 2.1, the cratered plains unit in which Columbus occurs has similarly been dated to the Middle Noachian. We also counted separately only those craters that clearly superpose (i.e., postdate) Columbus's light-toned layered deposits, yielding densities of N(5) = 280 ± 160 and N(2) = 1210 ± 340, consistent with a Late Noachian age (although Early Hesperian is within the uncertainties). Finally, the probable lava flow on the floor of Columbus has crater densities N(5) = 130 ± 130, N(2) = 920 ± 350, and N(1) = 3410 ± 670, consistent with an Early Hesperian age like most ridged plains on Mars. For each of these (Columbus crater, layered deposits, and lava), 1–3 craters were counted in the largest size category and 7–26 craters in the smaller size categories.
 In summary, Columbus crater likely formed in the Middle-to-Late Noachian, and it accumulated >1 km of fill prior to a distinct period of light-toned deposit formation and aqueous activity during the Late Noachian. A probable lava flow subsequently covered most of the crater floor during the Early Hesperian, and it was modified during this period by compressional and extensional tectonics likely related to Tharsis loading. Sediment accumulation continued at a lower rate during the Late Hesperian and possibly later, emplacing a mantle of likely aeolian origin on at least some portions of the crater interior. In the more recent Amazonian, the dominant geologic processes have likely been mass wasting and aeolian erosion of the intracrater deposits, as well as impact events that have exposed older deposits in the walls and ejecta of fresh craters such as that in Figure 25.
6. Other Aqueous Deposits of Northwest Terra Sirenum
 As reported by Murchie et al. [2009b] and Wray et al. [2009a], CTX images and CRISM multispectral data hint that several craters near Columbus in northwest Terra Sirenum may host similar layered deposits, and this region's intercrater plains also contain aqueous mineral deposits. Here we describe these deposits as regional context for our Columbus crater observations.
6.1. Intracrater Deposits
 The most striking feature of northwest Terra Sirenum in CRISM multispectral data is the presence of a ∼2.2 μm Al- or Si-OH absorption in materials occupying most of the region's large craters (Figure 27). In particular, this spectral feature is found in the region's most degraded (and therefore probably oldest) craters, which have flat floors, degraded rims, and superposed impact craters. By contrast, a D ∼80 km crater ∼300 km east of Columbus with fresh-appearing ejecta has no detectable hydrated minerals. Of the northwest Sirenum craters other than Columbus, so far the greatest mineralogic diversity has been found in Cross crater, which hosts alunite, kaolinite, and montmorillonite or hydrated silica in finely bedded deposits on its walls and floor [Swayze et al., 2008b]. Cross crater is the focus of a manuscript in preparation by G. A. Swayze et al. (2010), so we focus on other craters here.
 Of the nine craters identified with a ∼2.2 μm band in CRISM multispectral data, three are named (Columbus, Cross, Dejnev) and to the other six we assign letters (craters A–F) for convenience (Figure 27). Two other craters (G and H in Figure 27) lack CRISM coverage but have light-toned floor deposits morphologically similar to those on the floor of Columbus (see CTX P18_008237_1505 and HiRISE ESP_017573_1570, respectively). Craters A, D, E, and Dejnev have recently been observed by CRISM in hyperspectral mode, and all craters exhibiting the 2.2 μm absorption have been imaged by HiRISE. Light-toned layered deposits are observed in most of these craters (Figure 28). These deposits are especially widespread across the floor of crater E (Figure 28a), where they have likely been significantly eroded to yield the scattered mesas observed on the modern crater floor. Each hectometer-scale mesa exposes dozens of meter-scale beds in cross section (Figure 28e). The morphology and color of beds in crater F and Dejnev (Figures 28c and 28d) are reminiscent of beds on the floor of Columbus crater, including those that contain monohydrated sulfate. In HiRISE enhanced color images, the green appearance of some beds in crater B (Figure 28b) suggests a crystalline ferric mineral might be present (possibly ferric sulfate; see discussion in section 3.5), but this possibility has not yet been tested with CRISM hyperspectral data.
 As in Columbus and Cross craters, kaolinite is identified in hyperspectral images of craters A, D, E, and Dejnev (Figure 29). We have also found Fe/Mg-phyllosilicate in each of these locations except in crater D, and montmorillonite is detected in crater E. Somewhat surprisingly, these minerals generally are not found within the light-toned layered units of Figures 28c–28e. Instead, kaolinite and Fe/Mg-phyllosilicate most commonly occur in massive materials on the crater walls, and montmorillonite (possibly mixed with kaolinite) is found in the somewhat darker, rough-textured (but layered) floor of crater E (Figure 28a). Some detections of kaolinite in crater E (CRISM FRT000106E4) appear to correspond to light-toned layered deposits, but the higher spatial resolution of HiRISE is needed to confirm this. Stratigraphic relations between the kaolinite and Fe/Mg-phyllosilicates are not always clear, but in the case of crater E the latter are found eroding from the uppermost southeast crater wall (analogous to Figure 10 from Columbus), whereas Al-phyllosilicates are seen deeper within the crater.
 What, then, is the mineralogy of the light-toned beds on these crater floors? In crater E, spectra of most layered outcrops contain no obvious absorptions from 1.0 to 2.6 μm; however, these outcrops have stronger ∼3 μm bands than other materials in the same CRISM scene (Figure 28a). Absorptions at ∼3 μm are ubiquitous on Mars, though variable in strength; they are attributed to adsorbed or structural H2O and/or OH in surface materials [Jouglet et al., 2007; Milliken et al., 2007]. Many nominally anhydrous minerals have a ∼3 μm absorption in spectra obtained under terrestrial laboratory conditions [Clark et al., 2007]. Still, this absorption is typically stronger for hydrated minerals such as phyllosilicates, including those on Mars [Jouglet et al., 2007; Milliken et al., 2007], but in crater E the ∼3 μm band is weaker in the montmorillonite-bearing crater floor materials and stronger in the layered mesas. The mesas have a higher albedo, a property that is also positively correlated with ∼3 μm band depth in I/F spectra on a global scale, possibly because the greater absorptivity of darker materials results in higher daytime temperatures, driving off adsorbed water [Jouglet et al., 2007]. Alternatively, Milliken et al.  showed that the hydration–albedo correlation can result from nonlinear absorption processes, and that high albedo regions do not contain more water than dark regions once these effects are accounted for by converting spectra to single scattering albedo. Regardless, the precise mineralogy and nature of hydration in the layered mesas of crater E (and Dejnev) are currently unconstrained.
6.2. Intercrater Deposits
 Whereas Al-phyllosilicates and sulfates are the dominant secondary minerals in the craters of northwest Terra Sirenum, Fe/Mg-phyllosilicates (smectites) and chloride salts are more common on the intercrater plains, as in other regions of the southern highlands [Osterloo et al., 2008; Murchie et al., 2009b; Wray et al., 2009a; Glotch et al., 2010]. Baldridge et al. [2009, Figure 2c] identified putative chlorides in a shallow basin just ∼10 km south of Columbus crater's southwestern rim, and Figure 30a shows another location ∼60 km east of Columbus where THEMIS identifies chloride in bright outcrops; these outcrops overlie materials in which CRISM (FRT0000B59A) detects Fe/Mg-phyllosilicates. To date, chlorides have not been identified within any of the larger craters discussed in section 6.1.
 A few locations on the intercrater plains of northwest Sirenum do contain Al-phyllosilicates. Figure 27 enumerates eight intercrater sites at which a ∼2.2 μm absorption is found in multispectral and/or hyperspectral CRISM data. All except site 8 have been imaged by HiRISE, revealing a range of exposure types. At sites 1 and 4 (Figures 31a and 31b), CRISM detects kaolin group phyllosilicates in the ejecta of small impact craters (D ∼700 m and 600 m, respectively). The site 4 crater appears relatively unmodified (i.e., young), with bright rays visible in THEMIS nighttime IR images and abundant meter-scale blocks on the rim and proximal ejecta. This crater's proximal ejecta contain Fe/Mg-phyllosilicates (Figure 31a), which may underlie kaolinite-bearing materials in the subsurface here. Sites 2 and 7 each contain a cluster of small craters with kaolinite-bearing ejecta (Figures 31c and 31d); the largest of these craters exposes light-toned strata in its upper walls (Figure 31e) that may be the source of the kaolinite. At site 6, a D ∼1.4 km crater exposes light-toned materials in its rim (Figure 30b). All of these craters are probably too small to have initiated long-lived hydrothermal activity to form phyllosilicates [Rathbun and Squyres, 2002], and given the general paucity of evidence for phyllosilicate formation in post-Noachian terrains [Poulet et al., 2005; Bibring et al., 2006; Mustard et al., 2008], we infer that they likely excavated preexisting Al-phyllosilicates from the shallow subsurface.
 At intercrater Al-OH site 5, only a single hectometer-scale exposure of Al-phyllosilicate is identified in CRISM HRL00011D66, but adjacent light-toned layered deposits (Figure 30c) are spread across an area tens of km wide on the plains northwest of crater E. Similar to crater E's floor deposits, the layered deposits at site 5 have a relatively strong ∼3 μm band as their only distinguishing feature in CRISM data. These deposits provide evidence that the sedimentation and aqueous processes that occurred in the large craters of northwest Terra Sirenum were not restricted to these locations but also affected at least some portions of the intercrater plains.
 At intercrater site 3 (the only 2.2 μm site south of Columbus and Cross craters) the ∼2.2 μm band identified in CRISM mapping data is not primarily due to Al-OH, but instead to Si-OH. Although this site's relationship to the Columbus crater deposits is unclear, we describe it here in the interest of fully exploring the region's aqueous history.
 At this location (167.45°W, 33.15°S), a mound ∼3 km by 5 km wide protrudes ∼100 m from the surrounding plains (Figure 32a). The central mound has a weak spectral signature consistent with hydrated silica (i.e., opal), with stronger signatures present on its flanks and on the adjacent plains. Hydrated silica is distinguished from Al-phyllosilicates by its broader 2.2 μm band with an asymmetric long-wavelength edge extending to 2.3 μm or beyond [Milliken et al., 2008; Ehlmann et al., 2009]. In well-hydrated opaline silica this band is in fact a doublet with minima at 2.21 and 2.26 μm. Removal of H2O from opal at moderate temperature or low relative humidity causes the 2.26 μm band to disappear, weakens the 1.9 μm H2O band, and “shifts” the Si-OH overtone absorption from 1.41 μm to 1.38 μm [Anderson and Wickersheim, 1964; Milliken et al., 2008; Cloutis et al., 2009; Ehlmann et al., 2009]. Hydroxylated glasses have similar features, but their single 2.2 μm absorption is more symmetric and commonly centered at a slightly longer wavelength of 2.22–2.23 μm, and their ∼1.4 μm overtone does not shift upon dehydration [Milliken et al., 2008].
 Spectra observed on and around the mound at site 3 are most consistent with opaline silica with variable degrees of hydration (Figure 33a). H2O-poor silica is found in bright materials (Figure 33b) on the plains around the mound, whereas more fully hydrated silica occurs on the mound itself and in some intermediate-toned exposures on the plains (Figure 32a). These spectral differences are observed even when the same denominator is used for all ratio spectra in the scene. There is no obvious stratification or zoning pattern of the hydration states. The presence of variably hydrated materials in close proximity may suggest differences in formation temperature or a dehydration process driven by local thermal gradients rather than atmospheric humidity, consistent with a hydrothermal environment for silica formation; alternatively, varying physical properties could have made some silica-bearing materials more susceptible to dehydration than others. Parts of the mound and surrounding plains have a narrower 2.16/2.20 μm doublet consistent with a kaolin group clay (Figure 33a), providing a possible mineralogic link to the kaolinite-bearing materials identified elsewhere in northwest Sirenum. Finally, a relatively strong low-calcium pyroxene signature is observed in plains several km distant from the mound, but the broader geologic context (Figure 32c) suggests no obvious connection between the mound and the pyroxene-bearing unit; the latter appears to be a more widespread underlying unit.
 We propose three alternative hypotheses for this topographic mound associated with opaline silica: it could be (1) an erosional remnant mesa, (2) a siliceous spring mound [e.g., Guidry and Chafetz, 2003], or (3) a volcanic construct that has been aqueously altered, possibly under acidic conditions [e.g., Seelos et al., 2010]. Opaline silica has previously been identified adjacent to volcanic mounds both in the Columbia Hills of Gusev crater [Squyres et al., 2008] and in the Nili Patera caldera of Syrtis Major [Skok et al., 2010]. The Nili Patera mound is quite similar to our site 3 in lateral scale and in the localization of its partially dehydrated silica to bright outcrops adjacent to the mound. However, the Nili mound is ∼3 times taller than ours, is texturally massive, and sheds abundant meter-scale boulders (see HiRISE ESP_013582_1895). Fewer boulders are visible on the flanks of the mound at site 3, suggesting the material is friable and thus consistent with a sedimentary nature; the mound's lower profile, constituent silica, and bedded or terraced morphology (Figure 32d) are more consistent with silica-cemented sediments or a spring mound than with a volcanic construct. Also unlike the Columbia Hills and Nili Patera, there are no other obvious volcanic features in the vicinity of site 3; whereas pyroxene and olivine-bearing ridged materials adjacent to the Nili mound are likely lava flows [Skok et al., 2010], the pyroxene-bearing outcrops near our site 3 have a fractured texture (Figure 32b) uncharacteristic of typical Martian lavas. If site 3 is a siliceous spring mound, its size would rival the largest known spring mounds on Earth [Crumpler, 2003], but would be 1–2 orders of magnitude smaller than many Martian crater and canyon mounds that have previously been proposed as spring mounds [Rossi et al., 2008]. Alternatively, this mound could simply be an erosional remnant of a previously more widespread deposit, possibly sediments that have been cemented by silica precipitating out of groundwater. If the kilometer-wide depression near the mound's summit is a degraded impact crater, then it may have made this portion of the deposit more resistant to erosion (e.g., by armoring the surface with its ejecta). In this scenario, the process of silica formation would be poorly constrained.
7. Regional Hydrologic Modeling
 The diverse hydrated minerals found in Terra Sirenum (a region in which surface valley networks are relatively sparse [Carr, 1995; Fassett and Head, 2008b; Hynek et al., 2010]) motivate consideration of a groundwater-dominated hydrology to explain the aqueous deposits. Global-scale hydrological models representing the groundwater evolution of early Mars can explain the distribution of sulfate-rich evaporite deposits in Meridiani Planum and surrounding regions [Andrews-Hanna et al., 2007]. In those models, the region west of Tharsis (i.e., northern Terra Sirenum) is also a preferred location for a shallow water table and evaporite formation at smaller scales, driven by a combination of groundwater flow from the nearby high topography of Tharsis and the presence of a shallow topographic trough surrounding the rise [Phillips et al., 2001]. Ejection of water from aquifers buried deep beneath Tharsis may also play a role [Andrews-Hanna et al., 2007].
 In Terra Sirenum, hydrological activity in the vicinity of Columbus crater should be encouraged by a confluence of factors. These include its location roughly equidistant from the dichotomy boundary and a middle-sized basin to the south, each of which should act to draw down the water table in their immediate vicinity (so, by comparison, the water table would be relatively high near Columbus crater). Farther south, the water table drops deeper beneath the surface as a result of the southern limit of the low-latitude precipitation belt in the models and regional drawdown of the water table in high southern latitudes by the Hellas and Argyre impact basins [Andrews-Hanna et al., 2008]. For the detailed distribution of predicted evaporites in Sirenum, we turn to regional hydrologic modeling.
 We have run high-resolution (0.25 degree per pixel, corresponding to ∼15 km at the equator) local hydrological models of the region west of Tharsis, using the precipitation rates and hydrologic head from the global simulations [Andrews-Hanna et al., 2008, 2010] as boundary conditions. The precipitation is set to follow a cosine distribution between ±45° latitude, to approximately match the distribution of valley networks. The local model extends from 180° to 230°E and from 50°S to 10°N, boundaries sufficiently far from the region of interest (Figure 27) to avoid significantly affecting the results. The model assumes that groundwater evaporates upon reaching the surface, resulting in the formation of evaporites and evaporite-cemented sediments that can accumulate to substantial thickness in some regions. There is a dynamic coupling between the groundwater flow and the surface topography, in which groundwater-mediated sedimentation modifies the surface topography, which in turn modifies the paths and rates of groundwater flow. Ponding of emergent groundwater (e.g., in a crater lake) would affect the local hydrology in the same way as sediment accumulation would. The models assume a ratio between evaporated water column height and resulting sediment thickness of 50 to 1, consistent with evaporation of deep groundwater with salinity comparable to seawater and a 40% volumetric contribution from nonevaporitic sediments [Handford, 1991; Möller et al., 1997; Andrews-Hanna et al., 2007]. Higher or lower rates of deposit formation could result from increased or decreased sediment flux and fluid salinities. However, because the surface topography evolves slowly relative to the fluid flow, the system is in a state of quasi-equilibrium at any one time, and the impact of different sedimentation rates can be achieved by simply scaling the time scale for evolution. Similarly, changing the mean crustal permeability primarily changes the rate of water cycling through the hydrologic system without altering the gross distribution and flow paths of water. The model results should, therefore, yield insights into the spatial distribution of relative evaporite thicknesses even if their absolute thicknesses are poorly constrained.
 As found in the global models [Andrews-Hanna et al., 2008, 2010], the water table initially intersects the surface only within scattered impact craters and other topographic lows. The intersection of the water table with major impact craters can be clearly seen in maps of the hydraulic head at the beginning of the simulation (Figure 34b). As the simulation evolves, shallower craters fill with sediments resulting in a concomitant decrease in the rate of groundwater upwelling. In most cases, this trend progresses until the crater depth reaches some minimum value at which the groundwater flux to the surface terminates. The water table follows a smoothed approximation of the topography (Figure 34b), with the shorter-wavelength structure diminishing as the smaller craters are filled with sediments, resulting in the dominance of longer-wavelength flow paths (Figures 34f, 34j, and 34n). Nevertheless, Columbus crater still features prominently as a locus of groundwater upwelling and evaporation after the groundwater flux into many of the smaller craters has ceased (Figures 34i–34l). Although the models assume that groundwater evaporates immediately upon reaching the surface, this groundwater flux could support the formation of either playas or deeper lakes depending on the local climatic conditions.
 These models have used the present-day topography of Mars as a starting condition, though the current topography is the end product of 4.5 billion years of evolution. Although it is not possible to recreate the surface of Mars at any one time in its history, we can infer the effects of changes to the surface topography over time. It is noteworthy that the older craters in the region (Dejnev, Columbus, Cross, and others) typically contain hydrated minerals and/or layered deposits, whereas younger craters do not (Figure 27 and section 6.1). For example, the relatively fresh D ∼80 km crater approximately 300 km east of Columbus (“X” in Figure 34q) is predicted to be a site of significant groundwater inflow, but no mineralogical evidence of such hydrological activity is seen. This would be expected if this crater postdated the active hydrological cycle, as is suggested by the lack of erosion of its ejecta blanket, raised crater rim, and central peak. Ignoring this crater, the remaining craters in Figure 27 with the greatest predicted evaporite thicknesses are Columbus and Cross (Figure 34q), consistent with our identification of hydrated sulfates exclusively in these two craters to date.
 Nevertheless, the true complexity of the Martian hydrologic cycle cannot be captured in these simple models. For example, the depth of the craters present during the active hydrological cycle is uncertain; many Martian craters appear to have experienced substantial erosion in the Noachian Period, resulting in flat-floored, shallow craters without raised rims [Craddock and Howard, 2002]. Some craters may have already been substantially eroded and infilled by the time active sulfate deposition occurred in the Late Noachian to Early Hesperian, while others may have been significantly deeper than their current state. Hydrological activity may have spanned a period of hundreds of millions of years, as evidenced by the presence of large craters interstratified within the Meridiani sulfate deposits [Edgett and Malin, 2002], but might not have been continuous. Craters formed during the period of hydrological activity west of Tharsis would have intersected and drawn down the water table in their immediate vicinity, possibly cutting off groundwater flow into nearby craters. In addition, many large craters west of Tharsis (including Columbus) are crossed by graben radial to Tharsis, some of which may date to the Noachian [Anderson et al., 2001]. The faults underlying these graben would have channeled fluids along strike down the slope of the Tharsis rise, potentially enhancing groundwater flux into the craters they intersect, in contrast with the homogeneous and isotropic hydraulic properties assumed in the models. Therefore, the actual sequence of hydrological activity would have been much more complicated than the simple monotonic evolution predicted by the models.
 Although our study region west of Tharsis is a preferred location for evaporite formation, the model results suggest that intracrater hydrated mineral deposits should be relatively common in large Noachian-aged craters across much of Mars. In particular, Figures 34h, 34l, and 34p show that evaporite formation is predicted beyond the specific subregion of Terra Sirenum in which we identify hydrated minerals. The restricted distribution observed by CRISM may indicate that the hydrologic-climatic environment in northwest Sirenum was particularly amenable to formation of such deposits, or alternatively conditions here may simply favor the preservation or exposure of deposits that were originally more widespread. Specifically, pervasive dust cover north of ∼20°S in this longitude range [Ruff and Christensen, 2002] may obscure the bedrock mineralogy north of the area in Figure 27. Likewise, south of ∼30°S the Amazonian ice-dust mantle described by Mustard et al.  obscures older deposits. West of our study region, a group of larger, interconnected basins experienced a distinct hydrologic history that has been considered in detail elsewhere [Irwin et al., 2002; Noe Dobrea et al., 2008b]. The circum-Tharsis trough to the east has been completely resurfaced by Hesperian lava flows, which would have buried any older evaporites.
 In summary, both the general concentration of aqueous deposits in northwest Terra Sirenum and the specific subset of craters with sulfates detected from orbit are consistent with groundwater upwelling predicted by tested hydrologic models, although postdepositional modification of the aqueous deposits has surely also affected their observed distribution. We note that although these models have focused on deposition of evaporitic sulfates, such groundwater activity may also have altered the Noachian crust to form phyllosilicates.
 In this section, we present several hypotheses to explain the observations described above. We focus on Columbus crater but also discuss observations from elsewhere in northwest Terra Sirenum. We begin with a summary of key findings and their general implications.
 Columbus crater is a Middle-to-Late Noachian-aged crater in a Middle Noachian-aged terrain. It contains layered deposits with diverse hydrous minerals, as do at least ten neighboring craters and small areas of the plains surrounding those craters. These deposits appear to date to the Late Noachian, an epoch during which significant groundwater upwelling and evaporation is predicted in this region, with Columbus being a location of especially significant modeled evaporite deposition. Based on the 50-to-1 water-to-evaporite volume ratio assumed in our hydrologic models (section 7), forming the ∼10–20 m thickness of aqueous deposits on the walls and floor of Columbus crater (sections 2.2 and 5) via evaporation would have required ∼500–1000 m total depth of water in the crater, integrated over time; this is similar to the ∼900 m elevation difference between the modern crater floor and the sulfate ring around the crater walls.
 Thermal infrared measurements suggest that the light-toned deposits of Columbus are highly altered, with phyllosilicate and sulfate abundances in the tens of percent by volume. Visual evidence for ongoing physical weathering and erosion of these deposits (Figure 6b and 6d) argues for pervasive alteration throughout the deposits rather than a surficial alteration rind. The specific minerals observed indicate that pH, water activity (aH2O), and possibly redox conditions of the alteration environment varied in space and/or time: Fe/Mg-smectites typically form at circumneutral pH [e.g., Chevrier et al., 2007], whereas localized deposits of jarosite and alunite suggest pH < 3–4 at least locally, based on terrestrial analogs [Bigham et al., 1996; Fernández-Remolar et al., 2005; Benison et al., 2007]. Gypsum on Mars precipitates at water activity exceeding that of terrestrial seawater (0.98), whereas monohydrated Mg-sulfate precipitates at aH2O ∼ 0.5 [Tosca et al., 2008a]. Szomolnokite (suggested in section 3.5 to constitute at least part of Columbus's monohydrate) is a ferrous sulfate, whereas ferric sulfates (including jarosite) and oxides/hydroxides are observed elsewhere in Columbus, implying variable redox conditions. Columbus is also one of the few known locations on Mars with interbedded phyllosilicates and sulfates, suggesting episodic changes in environmental conditions or sediment source regions. Whereas most Martian phyllosilicates are detected in apparent isolation from the salts that must have formed with them [Milliken et al., 2009], Columbus crater (perhaps like Gale crater [Milliken et al., 2010]) may retain its full alteration assemblage.
 Terrestrial lab experiments require moderate temperatures to convert polyhydrated Mg-sulfate into monohydrate [Freeman et al., 2007] or to form Fe(OH)SO4 from hydrated Fe-sulfates [Milliken et al., 2008; Swayze et al., 2008a; Morris et al., 2009; Lichtenberg et al., 2010]. Therefore it is intriguing that the strongest monohydrate and Fe(OH)SO4 signatures in Columbus crater correspond to one of the freshest ∼200 m impact craters on its light-toned deposits (Figure 17a), whose impact event could have provided the heat needed for these mineralogic transitions. However, other monohydrate-bearing outcrops on Columbus's floor have no obvious connection to impact craters; possible origins for these monohydrates are discussed in sections 8.2 and 8.4.
 A possible geochemical analog for the deposits of northwest Sirenum is provided by Western Australian acid saline lakes and groundwaters [e.g., Benison et al., 2007; Baldridge et al., 2009; Story et al., 2010]. These playa lakes precipitate halite, gypsum, kaolinite, and ferric oxides, while associated groundwaters precipitate the same minerals in addition to jarosite, alunite, and Fe-bearing phyllosilicates. All of these minerals are identified in Columbus crater, with the single exception of halite, which (like other anhydrous chlorides) has no diagnostic infrared spectral absorptions, and therefore may be difficult to identify unless it occurs in high abundance over large areas [see Osterloo et al., 2008]. The Mg-sulfates likely present in Columbus are missing from many Western Australian lakes, but this could be readily explained by a difference in primary compositions: the bedrock of Western Australia is granitic and gneissic [Benison et al., 2007], whereas the plains outside Columbus (like much of Mars) appear to be basaltic (section 4) and thus more Mg-rich. Terrestrial acid-saline lakes and groundwaters typically exhibit substantial geochemical gradients over relatively short spatial scales, leading to variations in the precipitated minerals reminiscent of those found in Columbus crater [Baldridge et al., 2009].
 Although acid-saline lakes precipitate kaolinite directly, elsewhere on Mars kaolinite formation has been attributed to top-down (possibly acid rainfall-driven) weathering [e.g., Ehlmann et al., 2009; Noe Dobrea and Swayze, 2010], and some of the intercrater kaolinite in northwest Sirenum may have formed in this way (e.g., Figure 31a). If this weathering predated sulfate formation in Columbus crater (which may have occurred under acidic conditions as proposed for Late Noachian/Early Hesperian Mars globally [Bibring et al., 2006]) then kaolinite could have survived these conditions more effectively than other phyllosilicates (e.g., smectites) due to its greater stability at lower pH [e.g., Altheide et al., 2010b] and its comparatively slow dissolution rate [Zolotov and Mironenko, 2007]. However, the finding of kaolinite associated with a possible spring mound or silica-cemented sediments (section 6.2) supports a groundwater- (not surface weathering-) related origin for some Al-phyllosilicates in this region if these clays are authigenic. Indeed, the bright fracture fill observed in some kaolinite-bearing materials in Columbus crater (Figure 7) likely reflects mineralization from subsurface fluids migrating through the fractures [Okubo et al., 2009]. Alteration may have been greatest within the impact craters labeled in Figure 27, in which emergent groundwater could have ponded. The region's Al-phyllosilicates may thus have formed via multiple processes over a range of time.
 We will consider the origin of the Sirenum layered deposits in more detail below; for now, we make only the general point that thin, conformable, laterally continuous beds (as are found in Columbus crater; see section 2.2) are commonly cited as evidence for deposition from suspension [e.g., Wilson et al., 2007; Grant et al., 2008], either as marine/lacustrine or air fall sediment. In the case of Holden crater, the additional characteristic of a restricted elevation distribution (as is also observed for Columbus's polyhydrate ring) has been argued to support subaqueous deposition [Grant et al., 2008]. The polygonal fracture patterns observed on sulfate-bearing outcrops in Columbus crater are similar to fracture polygons observed on sulfate-bearing rocks in Meridiani Planum [McLennan et al., 2005], and on chloride- [Osterloo et al., 2008, 2010] and phyllosilicate-bearing [e.g., Wray et al., 2008, 2009b] outcrops across much of Mars. Their preferential occurrence in materials containing aqueous minerals suggests that these polygons formed via desiccation of sediments and/or dehydration of constituent minerals.
 Also during the Late Noachian Epoch, groundwater upwelling and evaporation on the other side of Mars may have produced the aqueous deposits in Meridiani Planum [Andrews-Hanna et al., 2007]. These materials share many mineralogic characteristics with those in Columbus crater, including significant abundances of Mg/Ca/Fe-sulfates and secondary aluminosilicates with minor crystalline ferric oxide and possibly Fe-phyllosilicates [Clark et al., 2005; Glotch et al., 2006]. The deposits in both regions also appear to lack carbonates. An apparent difference between Meridiani and Terra Sirenum is the presence of Al-phyllosilicates and Al-sulfates in the latter. Although modest amounts of kaolinite or alunite cannot be excluded by rover analyses of Meridiani rocks [Clark et al., 2005], there is no evidence for either in remote sensing data [e.g., Poulet et al., 2008a]. The dearth of these Al-bearing secondary phases in Meridiani and much of the rest of Mars has been cited as evidence for low water-to-rock ratios in the alteration environments [Hurowitz and McLennan, 2007]. The distinctive mineralogy of northwest Sirenum could reflect higher water-to-rock conditions, perhaps facilitated by the presence of large craters in which upwelling groundwater could pond. Similar conditions may have initially been present in Meridiani Planum, with the evidence now buried beneath the extensive younger playa deposits [Andrews-Hanna et al., 2010].
 We now consider several hypotheses for the particular sedimentary processes that may have emplaced the layered deposits of northwest Sirenum, with an emphasis on the sulfate-bearing ring around the walls of Columbus crater. We find that most hypotheses conflict with at least some of our observations and/or with other knowledge of Martian geology, with the possible exception of the deep lake hypothesis (section 8.4).
8.1. Hypothesis 1: Columbus Ring as a Preexisting Layer
 In general, layered materials at roughly constant elevation around the walls of a crater could indicate impact exposure of preexisting layers in the subsurface. In the case of Columbus crater, we feel this hypothesis is untenable for several reasons.
 First, as described in section 5, a degraded D ∼17 km crater superposed on Columbus's southwest wall contains polyhydrated sulfate-bearing layered deposits on its floor (Figure 10). These deposits are essentially contiguous with the polyhydrate ring on this portion of the crater wall, implying that the 17 km crater was present before the ring was emplaced. This in turn implies that Columbus crater predates the polyhydrate ring.
 Second, impact crater formation affects the geometry of strata exposed in crater walls and rims. Specifically, uplift of a crater rim and subsequent terrace formation via wall slumping along listric faults both cause preexisting strata to be back-tilted (i.e., to dip away from the crater interior) relative to their preimpact geometries [Melosh, 1989]. As an example, back-tilted strata are observed in the rim of Endeavour crater in Meridiani Planum, where they are inferred to predate that crater [Wray et al., 2009b]. By contrast, the beds within Columbus crater's polyhydrate ring show the opposite trend, with an average dip direction almost exactly toward the crater interior and no beds observed to dip away from the crater (section 2.2). In addition to these quantitative measurements, images such as Figure 12 give the qualitative impression that the sulfate-bearing beds onlap the crater wall.
 Additional shortcomings of the preexisting layer hypothesis include the nondetection of this layer (or any polyhydrated sulfates) at similar (or any) elevations within adjacent craters, and the hypothesis' inability to account for the layered deposits with diverse hydrated minerals on the floors of Columbus and its neighboring craters, which must postdate these craters. Because of these numerous weaknesses, we discard this hypothesis for the sulfate-bearing ring of Columbus crater, while noting that some (but probably not all) kaolinite-bearing materials in Columbus and its neighboring craters might predate these craters.
8.2. Hypothesis 2: Columbus Ring as an Erosional Remnant
 Another explanation for the ring planform of polyhydrate-bearing deposits in Columbus crater is that this ring is the erosional remnant of beds that once spanned the entire crater. A variety of processes may have emplaced these beds, although we have argued that their morphologic features in orbital images appear most consistent with deposition from suspension. In this hypothesis, hydrated mineral formation may or may not be coeval with sedimentation; i.e., the sulfates and phyllosilicates could have formed (1) elsewhere and been transported into these craters, (2) during subaqueous sedimentation, or (3) subsequently via diagenesis. A combination of these alternatives is also possible; for comparison, evaporite minerals in the Meridiani Planum sediments are inferred to have formed elsewhere, been modified during deposition in playa lakes, and later experienced multiple episodes of diagenetic overprinting [McLennan et al., 2005].
 If hydrated mineral formation predated deposition in the Sirenum craters, then the alteration environment is unknown. However, there is no known plausible source for these sulfates and Al-phyllosilicates elsewhere in the southern highlands [Wray et al., 2009a]. Alunite in particular has been identified nowhere else on Mars to date.
 Alternatively, the aqueous minerals in these craters could have formed via diagenesis of layered deposits, which could have initially been composed of relatively unaltered mineral grains. Diagenesis (which likely occurred here, whether or not it produced most of the hydrated minerals) could account for Columbus's trend of polyhydrated sulfate at higher elevations on the crater walls versus monohydrated sulfate below on the crater floor. If all the sulfates were originally polyhydrated, then increased temperatures during burial diagenesis could have enabled conversion to monohydrate in the lowermost layers; this mechanism is one of several proposed to explain the stratification of polyhydrated over monohydrated sulfates in Candor Chasma and other canyons of Valles Marineris [Murchie et al., 2009a]. However, diagenetic formation of all hydrated minerals in Columbus seems unlikely given the alternating kaolinite and sulfate-bearing beds seen in some locations (section 3.2), which are difficult to explain unless these beds originally had strikingly different primary compositions, porosities or permeabilities.
 The third possibility (that the hydrated minerals formed in shallow playa lakes coeval with sediment deposition) would be consistent with the playa environments of Meridiani Planum and Western Australia that we have cited as analogs. If Columbus crater was once filled with evaporitic sediments to at least the level of its polyhydrate ring, then these ∼900 m of evaporites would have required tens of km total equivalent depth of water to evaporate in the crater over time, according to our model assumptions.
 Late Noachian sediments filling Columbus crater would have needed to be almost entirely removed by the Early Hesperian (the age of the probable lava now spanning the crater floor), requiring average erosion rates of at least a few μm/yr. These rates are modest by terrestrial standards, but have not existed globally on Mars since the Noachian, during which they have been attributed to precipitation-driven fluvial processes [e.g., Craddock and Maxwell, 1993; Hynek and Phillips, 2001]. Such processes could not have removed sediments from Columbus crater, which has no outlet; aeolian erosion and transport is the only conceivable mechanism for sediment removal. Aeolian erosion may be much more efficient for light-toned layered deposits than for typical basaltic surface materials on Mars [Malin and Edgett, 2000], and post-Noachian rates up to ∼2 μm/yr have indeed been estimated from the lack of small craters on many such deposits [McEwen et al., 2005], although this may overestimate the steady state rate if erosion is accelerated following impacts [Golombek et al., 2010]. There is evidence for at least some erosion of light-toned layered deposits on Columbus's floor (Figure 26a and associated discussion in section 5), but the inferred eroded thickness is only a few tens of meters, over an order of magnitude less than that needed to fill the crater to the level of its wall ring.
 More broadly, there is strong evidence for substantial erosion of crater-filling layered deposits across Mars; in fact, enough examples are known that an evolutionary sequence can be defined [Malin and Edgett, 2000]. It appears that erosion typically begins along the crater walls and then proceeds inward, such that incomplete removal leaves behind intracrater mounds such as those found in Gale and Henry craters [Malin and Edgett, 2000]. To our knowledge, no Martian crater other than Columbus exhibits a remnant sedimentary ring along the crater walls instead of a central mound. Therefore, while our understanding of the erosional process(es) is sufficiently poor that we cannot rule out this scenario, it would make Columbus unusual or possibly unique among Martian craters.
8.3. Hypothesis 3: Springline “Tufas”
 If the sulfate ring at Columbus is not an erosional remnant of crater-filling sediments, but rather was deposited originally as a ring, then it could be an evaporitic deposit formed when groundwater emerged along an impermeable layer exposed in the crater walls. Carbonate deposits formed in this way on Earth are called perched springline tufa or travertine, where the latter term applies to thermal springs and the former to ambient temperature precipitates [e.g., Ford and Pedley, 1996]. In principle, similar processes could occur with sulfate-rich solutions on Mars. Given our interpretation of relatively low formation temperatures for jarosite in Columbus crater and alunite in Cross crater (section 3.6), we adopt the term tufa for the remainder of this discussion.
 Terrestrial springline tufas exhibit distributary channels and discontinuous distal fan-shaped deposits, as well as rimmed terraces in the proximal deposits where water has ponded and evaporated [Pedley, 1990; Ford and Pedley, 1996; Fouke et al., 2000]. None of these features is evident in images of Columbus's light-toned deposits. It is also unclear how the perched springline model could account for the other craters of northwest Sirenum that have laterally continuous hydrated beds spread across their floors, but no preserved ring structures on their walls. Finally, the observation most difficult to reconcile with the springline tufa hypothesis is the presence of light-toned layered deposits in Columbus crater at the summit of the hills near its center, ∼700 m above the crater floor (Figures 8c and 8d). CRISM data indicate that these deposits contain polyhydrated sulfates similar to those identified in the beds ringing the crater walls.
 Although we cannot rule out a more complex scenario combining springline tufa with other sedimentation and diagenetic processes on the Sirenum crater floors, we now turn to a single hypothesis that could account for all aqueous deposits in Columbus crater.
8.4. Hypothesis 4: Deep Lake(s)
 A deep lake filling Columbus crater to at least the level of its preserved sulfate ring could account for sediments deposited on both the crater walls and floor, including the (submerged) hills in Figure 8c. It would explain the lateral continuity and conformability of the intracrater beds, as well as the observed mineralogic sequence of polyhydrated sulfates on the crater walls versus monohydrated on the floor: the latter would have been precipitated after evaporation or freezing had lowered the lake level and yielded a concentrated brine. The observed alternation of clay versus sulfate-bearing beds could reflect alternating periods of lake level rise followed by evaporation. Highly localized deposits of jarosite and possibly alunite on the crater floor could have formed during the final stage of lake evolution in shallow ponds of highly concentrated fluid (Figure 6c may show another example of this), and/or via subsequent diagenesis as in the Western Australian acid saline systems [Benison et al., 2007].
 Terrestrial saline lakes and playas commonly display “bathtub ring” patterns of evaporite deposition, with less soluble minerals (e.g., carbonates and/or polyhydrated sulfates) precipitating early on the lake margins and soluble salts (e.g., chlorides) later on the lakebed during the final stages of evaporation [e.g., Baldridge et al., 2004]. Chlorides have not yet been identified on Columbus's floor, but this may be due to (1) the lack of diagnostic infrared spectral absorptions for anhydrous chlorides, (2) burial by younger sediments and Early Hesperian lavas, and/or (3) dissolution during later aqueous or diagenetic episodes. Indeed, multiple aqueous episodes in Columbus are suggested by the intimate association of Ca- and Mg/Fe-sulfates in its polyhydrate ring (section 3.3); in a monotonically evaporating lake, Ca-sulfate would be expected to precipitate early, and Mg/Fe-sulfates later at lower aH2O [Tosca et al., 2008a]. The fact that some outcrops in Columbus with gypsum appear as rugged, relatively high-standing masses compared to adjacent outcrops with only Mg/Fe-sulfate (Figure 6b) could indicate that more soluble Mg/Fe-sulfate has been dissolved from the upper beds during a later aqueous episode, leaving relict gypsum. In general, the minerals identified in Columbus agree reasonably well with those predicted by geochemical models to form in a deep lake on Mars [Altheide et al., 2010a], although widespread kaolinite (if precipitated from lake water) and possible alunite suggest greater dissolved aluminum than is typically assumed for Martian solutions.
 The minimum ∼900 m depth of the hypothesized Columbus paleolake is equivalent to that required by our hydrologic models (section 7) to form evaporite deposits of the thickness (∼10–20 m) observed on the crater walls and floor. This could indicate that the lake was flooded to this level only once; alternatively, the lake may have been maintained or refilled if the salinity and/or clastic sediment fraction were lower than assumed in section 7. For a crater of Columbus's size, this depth implies a paleolake volume of ∼6000 km3, fractionally larger than Earth's Lake Michigan or the inferred Holden crater paleolake on Mars [Grant et al., 2008], although the depth of the Columbus lake would have been several times greater than either. Assuming evaporation (or freezing and sublimation) as the dominant water loss process, Columbus's greater depth would have given it longer life than the Holden paleolake. For a rough estimate of Martian paleolake lifetime, several authors have assumed terrestrial evaporation rates of ∼1–10 m/yr [e.g., Lewis and Aharonson, 2006; Grant et al., 2008; Orofino et al., 2009], giving a minimum lifetime of several centuries for Columbus if the lake were filled only once. However, evaporation rates can be lowered by a factor of ∼20 for highly concentrated sulfate brines [Chevrier and Altheide, 2008], which could greatly prolong the final stages of lake evaporation.
 Alternatively, the hypothesized lake in Columbus crater may have been capped by ice. If the Late Noachian climate was similar to that of modern Mars, the surface of a lake would freeze very rapidly; even if emergent groundwater were initially warm, freezing would occur within a few years [Kreslavsky and Head, 2002]. Permafrost conditions might have posed challenges for recharge of the groundwater aquifers that we favor as a source of lake water, but a relatively thin, latitudinally restricted, and/or temporary permafrost layer would be consistent with the groundwater model. Pure water lakes on modern Mars would freeze to a depth of hundreds of meters, but under (for example) a 300 mbar atmosphere the ice cover could have been only tens of meters thick [Squyres, 1989]. Saline water (as inferred from Columbus's sulfates) would remain liquid at lower temperature, allowing thinner ice in any climate. Water would be lost via ice sublimation at rates typically ∼10–100 times slower than evaporation of liquid water [McKay et al., 1985; Squyres, 1989], although these rates are strongly temperature-dependent. Detailed thermal models of lakes initially ∼1000 m deep yield estimated lifetimes of ∼105–106 years under present Martian conditions [Moore et al., 1995; Rivera-Valentin et al., 2010]; of course, Noachian Mars could have been quite different [e.g., Craddock and Howard, 2002].
 A deep lake in Columbus crater could have produced concentrated nearshore deposits of sulfate-bearing sediment in several ways. On Earth, lacustrine tufas form near the margins of large, deep lakes; Zimbelman et al.  have noted the potential utility of such tufas for identifying the level of Martian paleoshorelines. On Earth these structures (sometimes termed “freshwater reefs”) appear to be largely biogenic in origin [Pedley, 1990], but abiotic salt precipitation may also be enhanced in shallow water near the shore. In Columbus, crater wall profiles (Figure 4) typically have surface slopes <10° at the elevation of the sulfate ring and up to 20–30° below this level. Evaporites may have accumulated on a shallow water topographic “platform” here when the lake level was just above this point. Greater input of clastic sediment along the lake margins (e.g., from mass wasting or slope wash off the steep upper crater walls) may also have led to thicker and/or more resistant deposits forming there. This effect would have been strongest if the lake were ice covered: many lakes in the Antarctic dry valleys experience summer melting in a narrow “moat” around the lake margins [e.g., Nedell et al., 1987]. Evaporation (and evaporite deposition) would then be limited to this melted zone. Of course, some light-toned layered deposits did also form on Columbus's central floor.
 Historically, large paleolakes on Mars have been identified almost exclusively via morphologic criteria, such as inlet or outlet channels, fan/delta deposits, and/or possible shoreline features such as wave-cut terraces [e.g., Cabrol and Grin, 1999; Fassett and Head, 2008b]. For Columbus crater (and possibly some of the neighboring craters) we propose a groundwater-fed lake, which would not require inlet channels or deltas. As for a shoreline, the break in slope observed at the elevation of Columbus's polyhydrate ring (Figure 4) could be interpreted as an imperfectly preserved shoreline, and the lack of other paleoshorelines recording lower lake levels would be consistent with terrestrial experience that only the high-standing shoreline is well preserved in some paleolakes [Zimbelman et al., 2009]. Of course, terraces are common features of impact craters even on dry bodies such as the Moon (although such slump terraces typically do not have a constant elevation around the crater walls), and where shorelines have been identified previously on Mars, this interpretation is generally not unique [e.g., Malin and Edgett, 1999; Leverington and Maxwell, 2004]. Furthermore, laboratory experiments [Lorenz et al., 2005] and modeling results [Kraal et al., 2006] suggest that waves needed for shoreline development could not form in the tenuous atmosphere of modern Mars. And if the hypothesized lake in Columbus crater was covered by ice, then terrestrial polar beach analogs suggest that many traditional shoreline morphologies would never have developed [Rice, 1994]. Ice push ramparts might have developed at the lake margins, but these are rarely preserved in paleolakes [Gilbert, 1890, p. 72].
 Even if beach morphologies were once more apparent in Columbus crater, they have been attacked by gradational processes for over three billion years. Columbus is 5 orders of magnitude older than terrestrial paleolakes such as Bonneville [Gilbert, 1890] and Lahontan [Russell, 1885] to which we often appeal as morphologic analogs. Columbus crater has experienced volcanic resurfacing, tectonics, impacts by ∼km-scale bolides, aeolian erosion, and mass wasting since its period of aqueous activity. In contrast, there is little remote sensing evidence for widespread mineralogic alteration (i.e., formation of hydrous minerals) outside the north polar region during much of this time [Bibring et al., 2006; Murchie et al., 2009b]. Therefore, it is not entirely unsurprising that evidence for Noachian/Hesperian paleolakes might be better preserved in their mineralogic stratigraphy than in their morphologic features.
8.5. Astrobiological Implications
 If a deep, groundwater-fed paleolake did exist in Columbus crater (or anywhere else on ancient Mars) then such an environment may have been promising for habitability and fossilization of prospective Martian life forms. The detection of gypsum in Columbus crater indicates a high water activity during part of its aqueous history, although other identified salts are consistent with a lower water activity that may have been detrimental to life [Tosca et al., 2008a]. Temperatures might have been low, but hypersaline environments on Earth host organisms capable of growth at temperatures below 0°C [Niederberger et al., 2010]. The acid saline lakes in Western Australia to which we have appealed as possible geochemical analogs are not only inhabited, but contain diverse microbial populations [Mormile et al., 2009]. These lakes preserve microfossils and organic matter within crystals of gypsum and halite [Benison et al., 2008], and similar preservation has been observed in Mg-sulfate crystals precipitated from groundwater-fed lakes in British Columbia [Foster et al., 2010].
 The thick ice covers that may have been present on Martian paleolakes would have presented a challenge for photosynthetic life, as opaque aeolian materials accumulating on the ice would substantially reduce light flux into the lake. Under this scenario, chemosynthetic life may have been more plausible in Martian paleolakes, and from this perspective groundwater-fed lakes may be more promising than those fed by surface runoff. In Earth's Lake Huron, for example, redox gradients between oxygenated lake water and anoxic waters emergent from sinkhole plumes are exploited by sulfate-reducing chemoautotrophs living at depths of <100 m [Biddanda et al., 2006]. Redox gradients between groundwater and the surface/atmosphere have been proposed for ancient Mars [Hurowitz et al., 2010], and terrestrial experience suggests that the interface between these two volatile reservoirs can be a source of chemical energy for life.
8.6. Future Investigations
 Future work will enable additional tests of the hypotheses described above. The Mars Odyssey spacecraft has recently transitioned to an early afternoon orbit that will enable higher SNR observations for THEMIS, potentially yielding stronger constraints on the mineralogy of light-toned deposits in Columbus and other Sirenum craters. Expanded CRISM coverage may also lead to the identification of additional minerals; this is especially true for the nine craters other than Columbus and Cross craters marked in Figure 27, which have minimal coverage to date. Our work described here has focused on Columbus crater but has benefited greatly from consideration of the regional context. Future orbital studies of comparable detail focused on craters such as Dejnev or crater E (Figure 27) may lead to evolution of some interpretations made here. Ongoing geochemical and physical modeling of Columbus crater [Altheide et al., 2010a; Rivera-Valentin et al., 2010] will further inform our interpretations.
 Many questions about the deposits in Columbus crater could be most effectively answered by a landed mission. High-resolution imaging on the surface could resolve sedimentary textures, stratigraphic contacts and mineralogic boundaries, providing new insights into the depositional and diagenetic history of the light-toned layered deposits. The diverse secondary minerals identified from orbit would allow a rover or sample return mission to probe a range of chemical conditions and explore variations in ancient habitability over space and time.
 More broadly, the mere possibility that Columbus crater once hosted a groundwater-fed lake highlights the value of mineralogy for identifying Martian paleolakes. Previous studies have focused on morphologic indicators such as channels and deltas that may provide more definitive evidence for ponded water but that cannot be used to identify groundwater-fed paleolakes. As the CRISM global mapping data set nears completion, broader searches for evaporites in paleolakes will be possible. Exploring the range of evaporites in Martian paleolakes could reveal not only how lake chemistry varied in space and time, but it may also constrain the composition of the ancient Martian atmosphere [e.g., Moore et al., 2010; Wray et al., 2009c].
 Columbus crater in the Terra Sirenum region of Mars contains light-toned layered deposits with diverse secondary minerals. Gypsum and polyhydrated Mg/Fe-sulfate are found in a discrete ring around the crater walls, and in some locations these sulfates are interbedded with kaolinite. Modeling of thermal emission spectra suggests that abundances of these minerals are in the tens of percent by volume. Crystalline ferric oxide/hydroxide also appears to be eroding from the crater wall ring deposit. Light-toned outcrops on the crater floor contain additional minerals, including monohydrated sulfates, jarosite, and possibly alunite. Additional Al and Fe/Mg phyllosilicates are found in scattered outcrops on the crater walls and floor.
 Beds in Columbus's sulfate-bearing ring dip gently toward the crater interior and appear to postdate the crater itself. Some bed surfaces exhibit polygonal fracture patterns, consistent with desiccation or dehydration of constituent minerals. Crater counts suggest that these and the other light-toned deposits in Columbus formed during the Late Noachian Epoch; subsequently, the crater floor was largely resurfaced by a darker, rough-textured deposit that we interpret as Early Hesperian lava flows.
 We have surveyed the region surrounding Columbus crater and found that ∼10 nearby middle-sized craters also contain layered deposits and/or Al-phyllosilicates, although sulfates have only been found in one of these to date [Swayze et al., 2008b]. The intercrater plains of the region also contain scattered exposures of Al-phyllosilicates and one isolated mound with opaline silica, in addition to more common Fe/Mg-phyllosilicates with chloride salts. The scarcity of fluvial dissection in this region (and around craters such as Columbus in particular) suggests a groundwater origin for the region's aqueous deposits. Regional hydrologic modeling results reported here confirm that Late Noachian groundwater upwelling is a plausible explanation for the evaporites found in Columbus and Cross craters.
 Based on these observations, we suggest that a lake of at least ∼900 m depth may have occupied Columbus crater during the Late Noachian and that some of the observed minerals may have precipitated out of solution during evaporation or freezing. Geochemical gradients in this hypothesized lake may have been capable of supporting putative chemosynthetic life forms at some point in its existence. Alternatively, the sulfate ring around Columbus's walls could be a springline tufa/travertine-like deposit, or an erosional remnant of layered deposits that once filled the crater, although each of these hypotheses has shortcomings. The possibility of a deep groundwater-fed lake on ancient Mars motivates future spectroscopy-based searches for evaporite deposits in other Martian craters, which would be targets well suited for future in situ investigation.
 We thank S. Mattson and A. Dumke for their efforts producing HiRISE and HRSC DEMs, respectively. Early reviews by J. K. Crowley, G. A. Desborough, and S. A. Wilson Purdy as well as discussions with J. F. Mustard, M. P. Golombek, N. A. Cabrol, J. A. Grant, D. J. Des Marais, V. F. Chevrier, T. S. Altheide, and S. Karunatillake improved the paper. We thank R. P. Irwin III and an anonymous reviewer for their thorough attention to the manuscript. J.J.W. thanks the Fannie & John Hertz Foundation and the NSF Graduate Research Fellowship for support. We thank the HiRISE and CRISM science and operations teams for acquiring the data most critical to our observations and interpretations.