By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Please be advised that we experienced an unexpected issue that occurred on Saturday and Sunday January 20th and 21st that caused the site to be down for an extended period of time and affected the ability of users to access content on Wiley Online Library. This issue has now been fully resolved. We apologize for any inconvenience this may have caused and are working to ensure that we can alert you immediately of any unplanned periods of downtime or disruption in the future.
McDonnell Center for the Space Sciences, Department of Earth and Planetary Sciences, Washington University in St. Louis, St. Louis, Missouri, USA
Corresponding author: Y. Liu, McDonnell Center for the Space Sciences, Department of Earth and Planetary Sciences, Washington University in St. Louis, Campus Box 1169, One Brookings Dr., St. Louis, MO 63130, USA. (firstname.lastname@example.org)
 The DISORT radiative transfer model was used to retrieve Lambert albedos from 0.4 to 4.0 μm over hydrated sulfate deposits in Aram Chaos for the Mars Express OMEGA instrument. Albedos were also retrieved for a relatively anhydrous area to the north to use as a control for comparison to the hydrated sulfate spectra. Atmospheric gases and aerosols were modeled, along with both solar and thermal radiance contributions and retrieved Lambert albedos are similar for multiple OMEGA observations over the same areas. The Lambert albedo spectra show that the control area is dominated by electronic transition bands due to nanophase iron oxides and low-calcium orthopyroxenes, together with the ubiquitous 2.98μm band due in part to water adsorbed onto particle surfaces. The retrieved Lambert albedos for Aram Chaos show an enhanced 2.98 μm water band and bands located at 0.938, 1.46, 1.96, and 2.41 μm. We infer the presence of nanophase iron oxides, schwertmannite, and starkeyite based on consideration of these band locations, inferred electronic and vibrational absorptions, stability under Mars conditions, and pathways for formation. This mineral assemblage, together with gray, crystalline hematite previously detected from TES data (Glotch and Christensen, 2005), can be explained as a result of iron oxidation and evaporation of iron-, magnesium-, and sulfur-rich fluids during periods of rising groundwater.
 The Observatoire pour la Minéralogie, l'Eau, les Glaces et l'Activité (OMEGA) imaging spectrometer onboard Mars Express was inserted into Martian orbit in December 2003 [Bibring et al., 2004]. OMEGA is a visible-near infrared hyperspectral imager, operating in the spectral range 0.38–5.1μm with 352 contiguous spectral channels spread over three spectrometers: the visible near-infrared wavelength (VNIR) spectrometer with 96 channels from 0.38 to 1.05μm and a spectral sampling interval of 7 nm, the C spectrometer with 128 channels from 0.97 to 2.73 μm and a spectral sampling interval of 13 nm, and the L spectrometer with 128 channels from 2.55 to 5.1 μm and a spectral sampling interval of 20 nm [Bibring et al., 2004]. The spatial sampling ranges from 300 m/pixel to 4.8 km/pixel, depending on orbital altitude and emergence angle. Analyses of the OMEGA data reveal a diverse and complex Martian surface mineralogy, including evidence of hydrated sulfates and phyllosilicates [e.g., Bibring et al., 2005; Gendrin et al., 2005; Arvidson et al., 2005; Poulet et al., 2005].
 Thus far identification of hydrated phases on Mars has been performed mainly using the relatively short wavelength spectrometer data (∼0.4 to 2.6 μm) and the simple “volcano scan” correction method of removing atmospheric gas absorptions, to retrieve surface spectra. The “volcano scan” correction attempts to minimize atmospheric gas absorption features by dividing each I/F spectrum (radiance detected by instrument/solar radiance divided by π) by a scaled atmospheric transmission spectrum. The atmospheric spectrum was derived by dividing spectra acquired at the top of Olympus Mons by spectra acquired at the base of the volcano [e.g., Bibring et al., 2005; Langevin et al., 2005a; Mustard et al., 2005]. Vincendon et al.  developed a method to recover surface reflectance for the bright ice deposits within a north polar crater by evaluating and removing atmospheric aerosol contributions based on Monte Carlo approach, but this method is limited to model aerosols. Arvidson et al. [2005, 2006], Wiseman , and Cull et al.  used the Discrete Ordinate Radiative Transfer (DISORT) code [Stamnes et al., 1988; Wolff et al., 2007] to separate atmospheric and surface contributions from OMEGA and CRISM I/F spectra. DISORT is a one dimensional radiative transfer model with vertical discretization of surface and atmospheric properties [Stamnes et al., 1988]. The model was used to simulate I/F values at the top of Martian atmosphere by including radiative contributions from both the surface and atmosphere, including gas and aerosol absorption, scattering, and emission. The model results were used to retrieve surface reflectance values from measured OMEGA I/F data using a lookup table approach. The purpose of our paper is to retrieve Lambert albedo spectra using DISORT for hydrated sulfates in Aram Chaos for the OMEGA spectral region from 0.4 to 4.0 μm for detailed mineralogy analyses and interpretations. Lambert albedo is the directional-hemispherical reflectance of a surface in which the scattered light has a radiance that is dependent only on the incidence angle of the incoming radiation [Hapke, 1993]. Arvidson et al. [2005, 2006] and Wiseman  showed that Lambert albedo is a reasonable approach for retrieving surface reflectances. We note that based on analyses conducted for this paper the uncertain wavelength calibrations, multiple gas bands, and competing influences of solar reflected and thermal emission preclude detailed spectral retrievals for the 4 to 5 μm OMEGA wavelength region.
 The focus of our analysis is Aram Chaos, which is located in a 280 km wide crater centered at 2.5° N, 21.5° W (Figure 1). The area has been extensively studied using data from the Mars Global Surveyor (MGS) Thermal Emission Spectrometer (TES) [Christensen et al., 2001; Glotch and Christensen, 2005] and both the OMEGA and the Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) instruments [Gendrin et al. 2005; Massé et al., 2008; Lichtenberg et al., 2010]. Gendrin et al.  and Massé et al. , in particular, used the OMEGA data over Aram Chaos to identify hydrated minerals based on ∼1.4 and ∼1.9 μm absorption features and a minor contribution from monohydrated sulfate minerals based on ∼1.6, 2.1, and 2.4 μm absorption features. Lichtenberg et al.  used CRISM data to identify hydrated sulfates and a hydroxylated ferric sulfate (Fe3+(SO4)OH), with the latter identification based on a unique combination of bands at 1.49, 1.82, 2.238, and 2.38 μm. Crystalline ferric oxides (as opposed to nanophase iron oxides ubiquitous in Mars spectra [e.g., Morris et al., 1993]) were also identified in Aram Chaos using TES data [Glotch and Christensen, 2005] over the same areas as the hydrated sulfate detections [Bibring et al., 2005]. Aram Chaos provides a target area for which hydrated sulfates, hydroxylated sulfates, and iron oxides have all been identified and mapped and thus provides an excellent target for detailed surface spectral retrievals using explicit radiative transfer modeling. This paper includes retrieval of OMEGA Lambert albedos at longer wavelengths than has been done in previous studies in Aram Chaos, and explicitly searches for correlations between short wavelength water and metal-OH bands in OMEGA C data and enhanced (relative to surrounding areas) water and metal-OH bands in L data in the 3μm wavelength region, together with searching for evidence for C-O, S-O, and N-O vibrational features associated with carbonate, sulfate, and nitrate minerals [Bibring et al., 2004]. For example, carbonates have an absorption feature located at ∼3.4 μm wavelength region and sulfates have a ∼4.5 μm absorption feature due to sulfur-oxygen vibrational overtones.
2. OMEGA Data Description and Reduction Methodology
 The primary data set used in this paper for detailed spectral processing and analyses is OMEGA ORB0401_3. This scene was selected because it completely covers Aram Chaos and regions to the north (Figure 1). Two sites within this OMEGA scene were selected as study areas (Figures 1 and 2 and Table 1). Study site A is located in Aram Chaos and contains deep absorption features associated with hydrated sulfates as indicated by previous studies using the OMEGA C spectrometer data and “volcano scan” processing [e.g., Gendrin et al., 2005; Massé et al., 2008]. Study site B is located to the north of Aram Chaos and is the control area, exhibiting signatures of pyroxene, but not hydrated minerals [e.g., Combe et al., 2008]. Both sites exhibit an absorption band in the 3.0 μm wavelength due to water; another application of the OMEGA retrievals is the comparison of the band depths for this feature for the hydrated and control areas.
Table 1. OMEGA Data and DISORT Model Input Parametersa
OMEGA Observation IDs
0401_3 Site A
0401_3 Site B
Note: Lighting and viewing angles are from OMEGA geometric files. Atmospheric and surface data are from historical and climatology of TES data [Smith, 2004]. Surface temperatures were derived from the thermal model of Mellon et al. .
Solar longitude (Ls)
Incidence angle (degree)
Emergence angle (degree)
Phase angle (degree)
Surface pressure (millibar)
Surface temperature (K)
Dust optical depth
Ice optical depth
Water vapor abundance (precip-micrometers)
 Three other OMEGA observations (ORB0353_2, 1326_1, and 2240_3) were used to evaluate the modeling approach through comparison of retrieved Lambert albedos for the same areas (Figure 1 and Table 1). These scenes were selected because they were taken at different times, atmospheric conditions, and lighting and viewing geometries, and overlap coverage with ORB0401_3 data (Figure 1). Note that these observations will be only used for validating the modeling approach but not for detailed spectral analyses and interpretations, since they cover limited areas of Aram Chaos regions relative to ORB0401_3 (Figure 1).
 The approach is to use DISORT to simulate I/F values at the top of the atmosphere and retrieve surface Lambert albedos with spectral corrections for gases and aerosols for the wavelength region from 0.4 to 4.0 μm. For each OMEGA channel I/F was modeled as a function of Lambert albedo, atmospheric pressure and temperature as a function of height above the surface, dust and ice aerosol optical depths (scaled to the OMEGA wavelength ranges), atmospheric water vapor abundance, surface temperature, surface pressure, geometric parameters (incidence, emergence, and phase angles), and solar irradiance scaled to the appropriate heliocentric distance for the given observation (Table 1). Temperatures for each layer of the atmosphere, dust and ice aerosol radiative properties, and water vapor abundance were estimated from historical climatology trends based on TES data at the appropriate latitude, longitude, and solar longitude (Ls) [Conrath et al., 2000; Smith, 2002, 2004]. Surface pressure was based on Viking Lander measurements, and the pressure for each atmospheric layer was calculated by integrating the hydrostatic equilibrium equation [Conrath et al., 2000]. Dust and ice aerosols have strong wavelength dependences on single scattering albedos and single particle phase functions [Clancy et al., 2003; Wolff et al., 2009]. In the retrievals dust single scattering albedos for 1.5 μm radius particles (typical grain size for non-storm dust) were used and phase functions were modeled as a wavelength dependent Legendre polynomial with coefficients derived from the analysis of CRISM hyperspectral data [Wolff et al., 2009]. Water ice aerosols were modeled with single scattering albedos for 2.0 μm radius particles from Warren  and phase functions as a Legendre polynomial with coefficients derived by Clancy et al. .
 Both DISORT model output spectra and OMEGA I/F spectra show the dominance of CO2 gas bands and minor CO and H2O gas absorption bands for C data and the effects of aerosols for the VNIR data (Figure 3). The deep CO2 triplet at ∼2.0 μm is particularly useful for validating the Lambert albedo retrievals by ensuring residuals for these features were minimized, which was accomplished by running multiple simulations in which the atmospheric pressures at all altitude levels were adjusted to minimize these residuals. The model I/F spectra for this spectral range and resultant relationships between Lambert albedos and I/F values have been discussed in detail by Arvidson et al. . A lookup table was generated for each wavelength of measured data, relating model I/F (with atmospheric and surface radiances included) and input flat Lambert albedos (Figure 4). To retrieve surface Lambert albedo from OMEGA observations, at each wavelength, the observed I/F is replaced with the corresponding Lambert albedo using the lookup table.
 The DISORT modeling approach has been tested by comparing retrieved Lambert albedos from the different OMEGA observations over the Aram Chaos regions (Table 1 and Figure 1) and evaluating differences in the shape and absolute magnitude of retrieved spectra. For the OMEGA observations used in this study, TES-based climatology data show that variations of dust and water ice opacities for a given Ls through 4 Martian years (24–27) are relatively small (Table 2) and sensitivity tests using DISORT show that this range will not influence the shape of the retrieved spectra and lead to at most a few percent variation in absolute values. Also note that ORB0353_2 and 0401_3 data were acquired on 30 April and 15 May, 2004 (Ls = 26.84° and 33.43° in MY 27), respectively, when TES was still operating. For those two observations actual TES-based retrievals were used in Lambert albedo retrievals. For ORB1326_1 and 2240_3, the climatology data from a randomly picked Martian year were used (MY 26) (Tables 1 and 2).
Table 2. Dust and Water Ice Opacity Variations Over 4 Martian Years for a Given Observation (or Ls)a
Data ID (MY)
0401_3 (Site A)
0401_3 (Site B)
Note: τdust and τice represent dust and water ice opacities, respectively, which have been retrieved from TES spectra [Smith, 2004], covering almost three complete Martian seasonal cycles, from Ls = 141° in MY 24 through Ls = 82° in MY 27.
 The spectral range from 1.0 to 2.6 μm (i.e., C data) was used to compare retrievals for the four OMEGA scenes because as noted this wavelength range contains atmospheric gas bands, effects due to aerosols, and distinct diagnostic features due to hydrated minerals and basaltic materials and thus provides a good test for the ability of the method to retrieve similar spectra from different observations. Spectra were retrieved for overlapping areas for the four scenes. Note that although the I/F values are different for different observations because they have different atmospheric conditions and lighting and viewing geometries, the retrieved Lambert albedos are comparable both in shapes and absolute values (Figure 5). For example, the I/F values from ORB0401_3 and 1326_1 data have differences of as much as 8% at some wavelengths whereas the retrieved Lambert albedos have differences that are only 0.2% (Figure 5b). Extensive sensitivity tests using DISORT for varying atmospheric opacities and atmospheric pressures that cover and exceed the range of variations within the TES climatology database for the observations confirm that the methodology used in this paper for retrieval of Lambert albedos leads to repeatable results. This result is similar to what was found by Wiseman  in her extensive sensitivity analyses for retrieving Lambert albedos using DISORT. Further retrievals and analyses of Lambert albedos for this paper are restricted to ORB0401_3 since these data cover both the Aram Chaos and anhydrous control areas.
 For radiative transfer modeling using OMEGA L data thermal emission must be added to the model-based retrievals, since both reflected solar light and thermal emission contribute to I/F spectra at wavelengths greater than 2.8μm. OMEGA L data are also strongly influenced by deep 2.65–2.85 μm and 4.15–4.50 μm atmospheric CO2 bands. The deep atmospheric bands and roles of reflected and emitted radiation are illustrated in Figure 6, in which solar-only, thermal-only, and both contributions to I/F spectra are modeled. Depending on temperature, there is a “crossover” wavelength at which thermal and reflected radiance are impossible to separate and Lambert albedo cannot be retrieved.Figure 7shows examples of model I/F spectra for two different surface kinetic temperatures. Surface temperature-dependent crossovers are evident at wavelengths of 5.02 and 4.55μm, respectively, for the assumed surface temperatures of 250 K and 270 K (Figure 7). The crossover is a consequence of Kirchhoff's Law in that I/F values increase with increasing spectral reflectance in the solar-dominated wavelengths whereas thermal emission dominates the spectra at longest wavelengths and increases as emissivity increases at a given wavelength. Assuming isothermal conditions dominate, Lambert albedo is the complement of directional emissivity. The crossover occurs when the combination of solar surface reflectance and thermal emission, modulated by the atmospheric spectral radiance, leads to equivalent solar and thermal radiance contributions.Figure 8 is an example of relationships between model I/F and input Lambert albedos for the OMEGA L data at 270 K, showing positive slopes for shorter wavelength data and negative slopes at wavelengths >4.55 μm, where thermal effects dominate. Changing albedo causes corresponding change of emissivity by Kirchhoff's law, since radiance contains both reflected light and thermal emission from the surface and it is possible that changing albeo will not cause the change of radiance received by detector. Thus the band located at the crossover wavelength corresponds to a vertical line (Figure 8) whereby albedo (or emissivity = 1-albedo) cannot be retrieved using our modeling approach because radiance is independent of the albedo value.
 For wavelengths outside of the crossover region, modeling thermal emission requires that either the surface kinetic temperatures or the Lambert albedo at a given wavelength must be known. We have chosen to use the wavelengths at ∼5.0 μm to separate surface temperatures from albedo due to dominance of emitted radiance and relatively low atmospheric absorption in this spectral region. Empirical equations developed by Jouglet et al. [2007a] and Milliken [2006, chap. 5] were used to predict the albedo at ∼5.0 μm. Following Milliken's [2006, chap. 5] approach, we used the 2.4877 μm albedo retrieved from the OMEGA ORB0401_3 C data to estimate the albedos at 5.015, 5.0335, 5.0518, and 5.0708 μm. Several of the long wavelength bands were used, to compensate for uncertainties in band pass locations and to reduce statistical scatter. The predicted albedos at these wavelengths and an array of assumed surface temperatures were then used with DISORT to calculate the I/F values at these wavelengths that would be observed by the OMEGA instrument. This approach allowed retrieval of surface temperatures by searching for the temperatures that best matched the 5-μm I/F values. Jouglet et al. [2007a] and Milliken [2006, chap. 5] used a similar method to estimate surface temperatures, but atmospheric absorption, scattering, and emission were not considered. The OMEGA-based temperature maps for sites A and B retrieved from ORB0401_3 data using DISORT are shown inFigure 9. Temperatures retrieved based on the DISORT model are systematically ∼3 K warmer than those derived using the approach developed by Jouglet et al. [2007a] and Milliken [2006, chap. 5]. This difference in resulting temperatures is attributed to the addition of atmospheric attenuation of surface emission in the calculations that improves the accuracy of the model.
 For comparison with retrieved temperature maps surface temperatures were estimated across the Aram Chaos regions using the Mars surface thermal model of Mellon et al. . This model accounts for solar heating, thermal radiation to space, subsurface heat conduction, as well as contributions from the atmosphere and surface frost to compute the diurnal and seasonal variations in surface temperatures. TES-based mapped values of surface thermal inertia and albedo [Putzig and Mellon, 2007] and MOLA-based elevations [Smith et al., 1999] at 1/20° spatial resolution were used to estimate temperatures from the model. The model-based temperature map for site A is shown inFigure 9b. The OMEGA-based and the model-based temperatures are similar, i.e., the average of the OMEGA-based temperatures is 249.0 K with standard deviation of 5.6 K, whereas the average of model-based temperatures is 249.1 K with standard deviation of 3.5 K.
 As a final check on processing methodology, OMEGA ORB0401_3 L data I/F values for study sites A and B were reduced to Lambert albedos and displayed as RGB composites using wavelengths of 3.98, 3.46, 2.92 μm (Figure 10). With thermal effects removed less reddening should be seen for the Lambert albedo images (Figures 10b and 10d), relative to images derived from I/F data (Figure 10a and 10c). This was observed, except for the hydrated sulfate deposits in Aram Chaos which are interpreted to have an enhanced 3.0 μm water or OH band. We conclude that our overall approach for estimating temperatures from OMEGA data produces consistent and reasonable results. Extensive modeling of the 4 to 5 μm region as part of this paper show that retrieval of Lambert albedo spectra is notional at best and interpreted to be a consequence of overlapping gas bands, a relatively poor understanding of instrument band positions and shapes, and the rapid increase in thermal emission with increasing wavelengths. This leads to very large errors in spectral retrievals. We thus only show Lambert albedos from 0.4 to 4.0 μm. The use of the instrument as a “radiometer” at ∼5 μm is still valid because of the dominance of thermal emission and the lack of gas bands at this long wavelength location.
3. Spectral Feature Descriptions and Mapping
 Lambert albedo spectra for the OMEGA ORB0401_3 spectral range from 0.4 to 4.0 μm are shown in Figure 11, both for original retrievals and for spectra smoothed using an eleven-band sliding Savitzky-Golay filter [Savitzky and Golay, 1964]. The spectra were extracted from regions of interest (ROIs) covering five lines and five samples (25 pixel boxes) (Figure 2). As noted the OMEGA L data are strongly influenced by deep 2.65–2.85 μm atmospheric CO2 bands (e.g., Figures 6 and 7) and Lambert albedos could not be recovered from this spectral region. As noted the site B spectrum has a different slope and spectral features as compared to the site A spectrum. Key absorption bands are labeled on the spectra in Figure 11and band center wavelengths for these features derived by inspecting continuum-removed spectra are given inTable 3. Both of the two spectra show the abrupt increase in albedo with increasing wavelengths from 0.4 to ∼0.75 μm (0.495 μm band center) that is ubiquitous to Mars and interpreted to be a ferric absorption edge associated with nanophase iron oxides [e.g., Morris et al., 1997]. The spectrum from site B has a reflectance maximum at 0.77 μm, along with absorption features centered at 0.945 and 1.89 μm. This is consistent with the spectral character of low-calcium orthopyroxenes [e.g.,Adams, 1974; Burns, 1993]. The site A spectrum exhibits a shorter wavelength reflectance maximum (0.74 μm) as compared to site B, a band minima centered at 0.938 μm and steep increases in reflectance from ∼0.95 to ∼1.38 μm. These features are interpreted to be ferric electronic transition features, a topic to be explored more fully in the next section of this paper.
Table 3. Wavelength Band Center Assignments for Lines A–H in Figure 11
Wavelength centers (μm)
 In the vibrational mode wavelength regime, the site A spectrum shows reflectance minima centered at 1.46, 1.96, and 2.4 μm. These features have been interpreted previously for Aram Chaos to be due to overtone and combination absorption bands associated with water in hydrated sulfates [e.g., Gendrin et al. 2005; Cloutis et al., 2006; Massé et al., 2008; Lichtenberg et al., 2010]. These features are not observed in spectra of study site B. The spectra of both sites show a broad, asymmetrical band with a minimum at 2.98 μm. This feature is ubiquitous on Mars and interpreted to be due to hydrated minerals and/or molecular H2O adsorbed onto grain surfaces [Jouglet et al., 2007a; Milliken et al., 2007]. We also searched for the presence of a 3.4 μm carbonate band using procedures developed by Jouglet et al. [2007b]with a negative identification for either study area. As noted we could not search for S-O or N-O features associated with sulfates and nitrates, respectively, because of the inability to retrieve spectra from 4 to 5μm.
 In order to illustrate the areal distribution of the absorption band depths for the spectral features characterized in this study, integrated band areas (IBA) were computed for the OMEGA ORB0401_3 data based on the differences between the spectral continuum slopes and the Lambert albedos at each wavelength as
where R(λ) is the albedo at the wavelength λ and Rc(λ) is the albedo of the continuum fit at the same wavelength. Linear continua were used between 0.75 to 0.98 μm for the IBA centered at 0.938 μm that is evident in the site A spectrum. The long wavelength side of this IBA estimation was constrained by low S/N for the OMEGA VNIR spectrometer at longer wavelengths. Additionally, the distribution of the low-calcium orthopyroxenes was mapped using the CRISM spectral parameter summary products developed byPelkey et al. , modified to OMEGA wavelengths as follows
 For the vibrational features evident in the site A spectrum the following continua values were used: 1.36 to 1.57 μm for IBA at 1.46 μm, 1.83 to 2.07 μm for IBA at 1.96 μm, 2.33 to 2.49 μm for IBA at 2.41 μm, and 2.86 to 3.84 μm for IBA at 2.98 μm. Results of the mapping are shown in Figures 12–13. The maps show the lack of low-calcium pyroxene and relatively high values of the H2O vibrational band maps for the hydrated sulfate-rich area, including the 2.98μm band depth. Note that the relatively small low-calcium pyroxene values seen in site B on the south and southwest sides of craters are consistent with maps of increased olivine for these types of deposits for a region just to the south of site B as mapped byCombe et al. . Overall the parameter maps show that the spectral features seen in Figure 11 for site A are representative of an extended area. We thus use the spectra shown in Figure 11 as type spectra for mineralogical interpretations.
4. Mineralogical Interpretations
 In this section of the paper we summarize mineralogical inferences based on comparisons of OMEGA data to hundreds of laboratory spectra (as spectral albedos and as continuum removed spectra), consideration of the strengths and wavelengths of electronic and vibrational absorption features, and evaluations of the stability of candidate minerals under Martian current environmental conditions. Based on previous work for Aram Chaos and our examination of a variety of laboratory spectra we focus on hydrated ferrous sulfates, ferric sulfates, magnesium sulfates, together with iron oxides, oxyhydroxides, and hydroxides as candidates to explain the spectra for Aram Chaos (Figures 14–18). Key candidate minerals in our “short list” and the sources for the spectra were summarized in Table 4.
Table 4. Compositions of Candidate Minerals and Sources of Spectral Reflectance
 We used the “Spectral Feature Fitting” techniques in ENVI's Spectral Analyst Tool, which is an absorption-feature-based method to directly identify a material by fitting unknown image spectra to the laboratory reference spectra using a least square fit. Visual examination and rank-ordering the sums of squares of deviations between the Aram Chaos spectrum and spectra for minerals listed inTable 4 show that schwertmannite and ferrihydrite are the best matches for Aram Chaos spectrum (Figures 14 and 15). This is the case for the overall spectral region under consideration and for the shorter wavelength region (<1.3 μm) dominated by electronic transition features. For this region only the inflection at 0.485 μm as evident in the schwertmannite spectrum is not evident in the Aram Chaos spectrum (or anywhere on Mars as far as we know for either OMEGA or CRISM data) (Figure 14). This is likely due to the obscuring effects of the ubiquitous nanophase iron oxides, combined with decreasing S/N of the OMEGA data at the shortest wavelengths.
 Schwertmannite has been identified as a likely mineral to be present on Mars [Burns, 1993; Bishop and Murad, 1996] and has been suggested as a possible surface component at Aram Chaos by Massé et al. using OMEGA data and the volcano scan technique to separate atmospheric gases from surface radiative streams. Schwertmannite is also a plausible candidate for the Fe3D3 Mössbauer feature observed by Opportunity in sulfate-rich rocks at Meridiani Planum [Morris et al., 2006b]. Schwertmannite is predicted to be the major initial iron oxide precipitate from acidic (pH 2.8 to 4.5) sulfate-rich aqueous fluids interacting with an oxidizing environment [Hurowitz et al., 2010] and may form by itself or as an admixture with ferrihydrite up to pH 6 [Raiswell et al., 2009and references therein]. Thermodynamic data indicate that in sulfate-bearing brines, such as fluids in contact with magnesium sulfate minerals, schwertmannite is more stable than ferrihydrite up to mildly alkaline conditions (pH 8) [Majzlan et al., 2004]. Once formed, schwertmannite will not transform to ferrihydrite, instead undergoing recrystallization to goethite upon wet aging [Schwertmann and Carlson, 2005; Burton et al., 2008; Kumpulainen et al., 2008]. It may also convert to a jarosite-goethite mixture under acidic pH conditions [Tosca et al., 2008]. Schwertmannite recrystallization under Mars-relevant conditions (low temperature and low activity of water) is currently unknown as all studies to date have been conducted at either elevated temperature or in the presence of liquid water; schwertmannite will likely show substantial preservation potential under cold and dry conditions [Raiswell et al., 2009]. We thus conclude that based on spectral matches, likely formation pathways, and stability schwertmannite is the more likely mineral controlling the overall properties of site A spectrum.
 The Aram Chaos spectrum has an absorption feature at 2.4 μm which is not present in the spectrum of schwertmannite. Hydrated iron sulfates have such a band, but are precluded from consideration because of the poor fits in the electronic (<1.3 μm) wavelength region (Figure 16). Polyhydrated magnesium sulfates do have an absorption feature at 2.4 μm, in addition to the absorption bands at ∼1.46 and 1.96 μm (Figure 17). Further, hydrated magnesium sulfates do not exhibit electronic transition bands but rather are spectrally neutral shortward of the H2O vibrational bands. Rank ordering of fits between the Aram Chaos spectrum and synthetic hydrated magnesium sulfates manufactured and measured by Wang et al.  in both albedo and continuum removed modes shows that sanderite (2 water molecules per unit cell) and starkeyite (4 water molecules per unit cell) provide good fits longward of 1.3 μm (Figures 17 and 18). Wang et al.,  show that under Martian current climate conditions starkeyite should be the stable phase across most of the Mars. We thus conclude that this phase is the most likely one to accompany schwertmannite in explaining the OMEGA spectra over Aram Chaos.
 Finally we note that the 2.98 μm band present in spectra for both Aram Chaos and the control area have similar shapes, but with a deeper band depth over Aram Chaos. The shape is best matched by the presence of water as opposed to metal-OH band(s), which would lead to diagnostic features dependent on the exact metal-OH band and crystal chemistry. Both schwertmannite and starkeyite exhibit the band shapes evident in the OMEGA data and we interpret the deeper band depth in this area as consistent with the presence of these hydrated phases, in addition to the ubiquitous adsorbed water or hydrated phases evident everywhere on Mars.
5. Conclusion and Implications
 Based on detailed reduction of OMEGA spectral data, focusing on use of DISORT-based radiative transfer procedures to model solar reflected and thermally dominating spectral regions, together with stability and formation pathway considerations, we conclude that the spectral features in Aram Chaos are dominated by the presence of nanophase iron oxides, schwertmannite, and starkeyite. Our modeling approach was validated by retrieving similar Lambert albedos from four different OMEGA observations at the same areas, and by the retrieving comparable surface temperatures from OMEGA data and the TES-based surface thermal model ofMellon et al. . Gray, crystalline hematite has been previously detected using TES data [Glotch and Christensen, 2005]. An assemblage of nanophase iron oxides, schwertmannite, starkeyite, and hematite is expected to be thermodynamically stable on Mars and is consistent with formation by iron oxidation and evaporation of iron-, magnesium-, and sulfur-rich fluids during periods of rising groundwater [Hurowitz et al., 2010].
 We would like to thank Ralph Milliken for helpful discussions and Scott Murchie for his great comments. We also thank Mathieu Vincendon and Deborah Domingue for their insightful reviews. We are grateful to the OMEGA Team for data acquisition and reduction and NASA for support for this work.