Chemically striking regions on Mars and Stealth revisited

Authors


Abstract

[1] The Mars Odyssey Gamma Ray Spectrometer Suite has yielded global chemical information for Mars. In this work, we establish regions of unusual chemical composition relative to average Mars primarily on the basis of Ca, Cl, Fe, H, K, Si, and Th. Using data from Mars Odyssey; the Mars Exploration Rovers; the Mars Reconnaissance Orbiter Imaging; and 3.5 cm and 1.35 cm radar observations from Earth, we examine a chemically striking ≈2.E6 km2 region and find it to overlap significantly with a radar Stealth region on Mars. It is remarkably enriched in Cl and depleted in Fe and Si (along with minor variations in H, K, and Th) relative to average Mars. Surface dust observed at the two rover sites mixed with and indurated by Ca/Mg-bearing sulfate salts would be a reasonable chemical and physical analog to meter-scale depths. We describe potential scenarios that may have contributed to the unique properties of this region. The bulk dust component may be an air fall deposit of compositionally uniform dust as observed in situ. Hydrothermal acid fog reactions on the flanks of nearby volcanoes may have generated sulfates with subsequent deflation and transport. Alternatively, sulfates may have been produced by low-temperature, regional-scale activity of ground ice–driven brine and/or regional-scale deposition of acidified H2O snowfall.

1. Introduction

[2] Recent remote sensing and in situ observations of Mars have revealed significant mineralogic and chemical variations at both regional and local spatial scales. Regional-scale mineralogic heterogeneities have been observed with the Mars Global Surveyor Thermal Emission Spectrometer (MGS-TES) [e.g., Rogers et al., 2007a; Ruff and Christensen, 2007]. More localized variations have been highlighted with the Mars Odyssey Thermal Emission Imaging System (THEMIS) [e.g., Osterloo et al., 2008], the Mars Express Observatoir pour la Minéralogie, l'Eau, les Glaces et l'Activité (MEX-OMEGA) [e.g., Poulet et al., 2007], and the Mars Reconnaissance Orbiter Compact Reconnaissance Imaging Spectrometer for Mars (MRO-CRISM) [e.g., Milliken et al., 2008]. Meanwhile, the Mars Exploration Rover (MER) Spirit's instrumentation has helped to identify particularly striking mineralogic and chemical diversity across a kilometers-scale traverse from the “basaltic plains” to the “Columbia Hills” within Gusev crater [e.g., Arvidson et al., 2006].

[3] The Mars Odyssey Gamma Ray Spectrometer (GRS) instrument suite complements TES, THEMIS, OMEGA, and CRISM observations in two important ways: greater sampling depth by several orders of magnitude (tens of centimeters versus tens of microns) and direct estimation of elemental mass fractions [e.g., Boynton et al., 2004]. The complementary nature of the two classes of instruments (neutron and γ photon versus infrared photon sensors) has provided greater insight into and established major constraints on the origin of some TES-derived mineralogic types [e.g., Karunatillake et al., 2006; Wyatt, 2007]. The GRS data also indicate significant regional-scale chemical variations in the midlatitudes of Mars. Motivated by previous analyses that highlighted such heterogeneities [e.g., Newsom et al., 2007; Keller et al., 2006], our work and two companion papers [Gasnault et al., 2009; G. J. Taylor et al., Mapping Mars geochemically, submitted to Geology, 2009] explore the diversity of the Martian surface from a chemical perspective.

[4] The primary goal of our work is twofold: to identify regions on Mars that are chemically striking and to describe one of them as a case study motivating future analyses of the remaining regions. We define a chemically striking region (CSR) as one where the mass fractions of two or more elements are significantly different from their midlatitudinal averages. We delineate these regions with the global midlatitudinal maps of elemental mass fractions generated by the Mars Odyssey team. Our case study also exploits the synergy with other missions: MER, MGS-TES, and MRO High Resolution Imaging Science Experiment (HiRISE) key among them.

[5] A thorough description of the primary instrumentation and data generation by the Mars Odyssey GRS is provided by Boynton et al. [2004, 2007], Evans et al. [2006], and Karunatillake et al. [2007] along with key statistical techniques to analyze remote sensing data on a global scale (S. Karunatillake et al., Discovering correlations without ignoring uncertainties: A Martian overview of multivariate methods, submitted to Earth, Moon, and Planets, 2009a; Recipes for spatial statistics with global data sets: A Martian case study, submitted to Journal of Scientific Computing, 2009b). The GRS is the first instrument suite to provide global data of elemental concentrations as mass fractions in the Martian subsurface to several tens of centimeter depths. This suite consists of a High Energy Neutron Detector (HEND), Neutron Spectrometer (NS), and Gamma Subsystem (GS). The GS's footprint, defined as the nadir-centered region within which ≈50% of the signal originates, is 3.7° arc radius (corresponding to 220 km linear radius) [Boynton et al., 2007; Karunatillake et al., 2007]. The three instruments, particularly the NS and GS, are complementary in their determination of H, with different sensitivities to mass fraction variabilities as well as different sampling depths [Boynton et al., 2004]. While the NS is capable of indirectly estimating the mass fractions of elements that affect the neutron energy spectrum, only the GS provides direct estimates for multiple elements by means of characteristic energies of γ photons emitted during nuclear deexcitations. For this reason and the focus of this work on defining CSRs, we rely solely on elemental mass fraction maps generated with the GS data.

2. Delineating Chemically Striking Regions

[6] We delineate chemically striking regions using the mass fractions of the seven elements for which uncertainties are reasonably small. These are Ca, Cl, Fe, H2O (as the stoichiometric equivalent of the GS's H estimates), K, Si, and Th. While the uncertainties in the values and methodology of Al mass fraction estimates are being refined, we present regions that include Al for future study. The CSRs are delineated in two steps. First, significant deviations from bulk Mars for a single element (a gaussian tail cluster (GTC)) is defined. Second, regions of overlap among GTCs that exceed an area threshold are identified.

2.1. GS Mapping Summary

[7] GS data subjected to temporal cumulation, multiple processing steps, and a mean filter to maximize signal-to-noise ratios, yield global maps of elemental mass fractions as described by Boynton et al. [2007], Evans et al. [2006], and Karunatillake et al. [2007]. The spatial extent of these maps is limited to the midlatitudes, since polar regions of high H concentration are subject to both mass dilution effects and spectrum-to-mass fraction conversion difficulties as discussed by Boynton et al. [2007]. We utilize the most extensive cumulation period currently available consisting of epochs 1 and 2 corresponding to the combined primary and extended mapping periods from 8 June 2002 (0000:00) UTC to 2 April 2005 (2020:00) UTC, and 30 April 2005 (0000:00) UTC to 22 March 2006 (0754:00) UTC, respectively. We choose a bin size of 5° × 5° for the mass fraction maps to partially address spatial uncertainty in the form of spatial autocorrelation (Karunatillake et al., submitted manuscript, 2009a) introduced primarily by the mean filter, the arc radius of which varies as shown in Table 1.

Table 1. Arc Radius, Linear Radius, and Approximate Surface Area of the Mean Filter Used to Generate the Global Map of Each Element/Oxidea
Element/OxideArc Radius (deg)Linear Radius (km)Surface Area (km2)
  • a

    The linear radius and surface area assume Mars to be an exact sphere and utilize the MGS95J model 3.396E3 km planetary radius [Konopliv et al., 2006].

Al, Ca, Si158.9E22.5E6
Cl, Fe, H2O, Th105.9E21.1E6
K53.0E22.8E5

2.2. Step 1: Delineating Gaussian Tail Clusters

[8] An unfortunate consequence of the nearly gaussian distribution of the elemental mass fractions (Karunatillake et al., submitted manuscript, 2009b) is that robust outliers do not exist in the distributions. However, spatially extensive regional variations in GS chemical maps make a strong case for treating them as such, the lack of outliers notwithstanding. Therefore, we utilize an improvised test parameter, t, as a measure of deviation from the bulk of Mars for each element. For any given element, the test parameter evaluated at the ith bin is

equation image

where ci is the mass fraction of the element at the ith bin, m is the global arithmetic mean mass fraction, sm,i is the numerical uncertainty of ci, and s is the standard deviation of the mass fractions. The key difference between ti and the commonly used Student's t parameter [e.g., Helsel and Hirsch, 2002, p. 126; Press et al., 2007, p. 728] is the inclusion of sm,i in the denominator. This term ensures that the significance of deviations from the global mean is evaluated in the context of numerical uncertainties. Since the GS data follow gaussian distributions to first order (Karunatillake et al., submitted manuscript, 2009b), our parameter is effective at identifying data in the distributional tails, i.e., the gaussian tail clusters (GTCs) of an element. In addition, the standard deviation (s) exceeds the root-mean-square uncertainty (srms) by more than 20% for the observed elements (Karunatillake et al., submitted manuscript, 2009b), confirming that numerically meaningful deviations from the mean exist.

[9] Due to the lack of outliers in the data and the use of sm,i to enhance rigor, the magnitude of ti does not exceed 3 in any of the elemental mass fraction maps. Therefore, the thresholds we utilize to identify GTCs are ti no less than 1, 1.5, 2, and 2.5 which correspond to better than 1s, 1.5s, 2s, and 2.5s confidence, respectively. Corresponding statistical confidence based on a Student's t distribution is listed in Table 2.

Table 2. Statistical Significance of a Given Deviation From the Mean Computed as the Cumulative Tail Probability of a Student's t Distributiona
tDeviation ExceedsPmag (%)P (%)
  • a

    The significance of the directional deviation (P), which is more relevant in the context of gaussian tail clusters (GTCs), is generally greater than that of the magnitude of deviation (Pmag). Here s denotes the standard deviation, and t is the test parameter in equation (1).

11s6884
1.51.5s8794
22s9598
2.52.5s9999

[10] The GTCs that result from two of the t thresholds are illustrated in Figures 12. The preservation of GTC interiors at t ≥ 1 (e.g., Figure 1, left) when the threshold is increased to 2 (e.g., Figure 1, right) is evidence that the GTCs are spatially meaningful. Those of Cl have been analyzed by Keller et al. [2006], with particular emphasis on the Cl-enriched regions that extend west from the Tharsis area and overlap considerably with the Medusae Fossae formation. As discussed by Karunatillake et al. [2006], GTCs that mark the enrichment of K and Th overlap strikingly with higher areal fractions of surface type 2 material [Rogers et al., 2007a; Rogers and Christensen, 2007]. Our work reinforces the spatial coincidences of GTCs with secular units as discussed by Hahn et al. [2007] and with Northern lowlands as discussed by Dohm et al. [2009]. Given our emphasis on the spatial overlap of GTCs to define CSRs, we do not discuss GTCs of individual elements further in this work.

Figure 1.

Gaussian tail clusters (GTCs) for two of the confidence thresholds used in this work for Al, Ca, Cl, Fe, and H2O. Al is illustrated solely to motivate future investigations as the underlying data are being refined. Remaining elements shown in Figure 2. (left) A confidence of t ≥ 1 (Table 2) and (right) t ≥ 2. Green indicates depletion, and red indicates enrichment. As discussed in section 2.1, data are restricted to the midlatitudes with the exception of K and Th. Even for these two elements, we arbitrarily exclude latitudes more extreme than ±75° to minimize mass dilution effects. As discussed in the text related to Figure 4, overlap among GTCs yields the CSRs.

Figure 2.

GTCs for K, Si, and Th as described in Figure 1.

2.3. Step 2: Spatial Overlap of GTCs and Area Threshold

[11] As described in the introduction, the CSRs are defined to be those of overlap among the GTCs of multiple elements. Even when underlying GTCs are spatially expansive, uncertainties in the form of spatial autocorrelation (e.g., Karunatillake et al., submitted manuscript, 2009a) may place their physical significance in doubt. As an analogy, any region that is smaller than the areal extent of the approximately 7.4° GS footprint may be considered as tenuous as a feature smaller than the 1.5 pixel point spread function (PSF) of an image generated by the HiRISE [e.g., McEwen et al., 2007, paragraph 18 and Figure 9]. For any given set of elements, we reduce the impact of this concern conservatively by identifying the CSRs that equal or exceed the area of the largest mean filter (Table 1). The area calculation is approximate, not exact, since it does not account for topography and assumes Mars to be a sphere. Furthermore, as Gasnault et al. (submitted manuscript, 2009) discuss, even our conservative constraints may not completely localize the proper area for study. Nevertheless, even a small CSR contained within those that show consistent chemical trends and satisfy area thresholds may prove insightful, as we will demonstrate below with our case study.

[12] For the sake of consistency and simplicity, we delineate CSRs for each of the 247 possible sets of elements taken two or more at a time, including Al, using GTCs defined with the same statistical threshold, such as t = 1. Only seven sets, all two elements each, survive the thresholds while highlighting numerous areas of the surface as given in Table 3. We do not impose additional constraints (such as eliminating sets of elements that are unknown to show chemical covariability on Earth) in order to identify CSRs free of terrestrial bias.

Table 3. Key to the Numerical Code of Chemically Striking Regions in Figure 3a
KeyValue
  • a

    Each chemically striking region (CSR) is denoted by the corresponding set of elements in curly braces, confidence (Table 2) as an approximation to a multiple of the standard deviation (s), enrichment (E) and/or depletion (D) in element order, and arc radius of the area threshold (Table 1). For example, {Cl, Si} 1.5s ED 15° would denote a bin belonging to a single CSR marked by the enrichment of Cl and depletion of Si at better than 1.5s confidence and exceeding a 15° radius area. On the other hand, {Cl, H} 1s EE 15° {Cl, Si} 1s ED 15° identifies a bin of overlap between two CSRs: One {Cl, H} 1s EE 15° and the other {Cl, Si} 1s ED 15°. Note that such bins generally do not delineate a sufficiently large contiguous area to be classified as a CSR in its own right. CSRs of Si and Th overlap completely with the CSRs of K and Th albeit at different statistical confidence levels. The one region on the basis of Al is solely to motivate future investigations as the Al map is being refined. Higher numerical uncertainties and weak correlation with other elements caused the absence of Ca-based CSRs.

Unclassified0
{Al, Fe} 1s ED 15°5
{Cl, H} 1s EE 15°10
{Cl, H} 1s EE 15° {Cl, Si} 1s ED 15°15
{Cl, Si} 1.5s ED 15°20
{Cl, Si} 1s DE 15°25
{Cl, Si} 1s DE 15° {K, Th} 1s EE 15°30
{Cl, Si} 1s ED 15°35
{Fe, Th} 1s EE 15°40
{Fe, Th} 1s EE 15° {K, Th} 1.5s EE 10°45
{Fe, Th} 1s EE 15° {K, Th} 1.5s EE 10° {Si, Th} 1s EE 15°50
{Fe, Th} 1s EE 15° {K, Th} 1s EE 15°55
{Fe, Th} 1s EE 15° {K, Th} 1s EE 15° {Si, Th} 1s EE 15°60
{Fe, Th} 1s EE 15° {Si, Th} 1s EE 15°65
{H, Si} 1s DE 15°70
{H, Si} 1s DE 15° {K, Th} 1s DD 10°75
{K, Th} 1.5s EE 10°80
{K, Th} 1.5s EE 10° {Si, Th} 1s EE 15°85
{K, Th} 1s DD 10°90
{K, Th} 1s DD 15°95
{K, Th} 1s EE 15°100

[13] The CSRs subsequent to area threshold application are illustrated in Figure 3. Comparisons with Figures 1 and 2 reveal that while GTCs overlap spatially across most statistical confidence thresholds for sets of two elements, few actually exceed the area of the largest corresponding mean filter (Table 1). As a case in point, Figure 4 illustrates the region delineation steps for Cl and Si at t ≥ 1.5. We also make a minor refinement to the area threshold method by discarding any bins with fewer than three edge or corner sharing neighbors. This effectively eliminates any narrow regions that are only a bin across with consequently higher spatial uncertainties. The delineation of GTCs, delineation of overlap, elimination of narrow regions, and the application of area thresholds for the 247 sets of elements were all implemented algorithmically without manual intervention.

Figure 3.

Chemically striking regions (CSRs) overlain on a geographically labeled (F, Fossae; P, Planitia/Planum; T, Terra) MOLA topographic map and cases of overlap among them numbered as given in Tables 3 and 4. Note that we are unable to classify areas of overlap as CSRs themselves, since they are generally smaller than the conservative area thresholds corresponding to the largest mean filter (Table 1) for each set of elements that we use in this work. Landing sites marked as Op, Opportunity; PF, Pathfinder; Ph, Phoenix; Sp, Spirit; VL1, Viking Lander 1; VL2, Viking Lander 2.

Figure 4.

Illustration of the CSR delineation steps described in section 2.3 using (left) Cl snd (right) Si mass fraction and uncertainty maps at t ≥ 1.5 as an example. Note how the element with the smaller GTCs constrains the areal extent of the resulting CSRs.

3. Overview of the Chemically Striking Regions

[14] To first order, two other independent methods of defining regions on the basis of elemental mass fractions (one a combination of principal component analysis, cluster analysis, hierarchical modeling, and a field-of-view filter (Gasnault et al., submitted manuscript, 2009) and the other a cluster analysis implemented in ENVI (Taylor et al., submitted manuscript, 2009)) also highlight the same broad regions of the planet as our approach. As discussed in the two companion papers, this reinforces the geochemical significance of CSRs in spite of spatial uncertainties.

[15] As evident in Figure 3, several CSRs marked by enrichments involving Fe, K, Si, and Th occupy Chryse Planitia northward to Acidalia. Several other CSRs, involving the enrichments of K, Si, and Th, and depletion of Cl, exist in the regions NE of both Isidis and Arabia and the western perimeter of Utopia into Vastitas Borealis. Collectively, these regions are suggestive of strong chemical variations marking the lowlands proximate to the lowlands margin. Accordingly, they may help to constrain models that relate lowland geochemistry dominantly to aqueous alteration [e.g., Dohm et al., 2009] or dominantly to igneous processes [e.g., Karunatillake et al., 2006].

[16] Several regions of high southern latitudes, mostly beyond the midlatitudinal constraint of the other elements (section 2.1), are marked by the mutual depletion of K and Th which may be influenced in part by mass dilution effects of H enrichment closer to the polar regions. Two areas of the midlatitudinal southern highlands are also highlighted by the CSRs. One lies immediately south of Valles Marineris overlapping with Syria, Solis, and Thaumasia plana corresponding to CSRs delineated by the enrichment of Si and depletion of H, K, and Th. The second, marked by the mutual enrichment of K and Th, occupies the vicinity of Sirenum Terra and Terra Cimmeria. Tentatively, a third area delineated by the depletion of Fe and enrichment of Al on the NW perimeter of Hellas may provide useful insight once the Al map is refined further.

[17] SE Elysium lava flows constitute nearly 70% of the underlying surface in one of smallest CSRs, delineated by the simultaneous depletion of K and Th, as mentioned later in section 3.1. Abutting this region is the last broad area on Mars that is identified by the CSRs. It extends westward from the Tharsis bulge, through the Medusae Fossae formation, and into Elysium Planitia. As mentioned in section 2.2, the key underlying GTCs for this broad expanse are those of Cl enrichment. The individual CSRs essentially identify chemical heterogeneities within it. Those of Si depletion are limited to the Tharsis construct, while the western section is marked by H enrichment.

[18] Do the CSRs conform with the global correlations that Karunatillake et al. [2006] identified in multivariate space? To first order, they do. For example, the presence of CSRs delineated by the enrichment of both K and Th and the absence of those marked by the enrichment of one and depletion of the other is consistent with the strong positive correlation that exists globally between them as discussed by Karunatillake et al. [2006, also submitted manuscript 2009a]. Global correlations and the CSRs are similarly consistent even for elements that are not as strongly correlated, such as Cl and Si. As was anticipated by the anticorrelation between Cl and Si [Karunatillake et al., 2006; Keller et al., 2006], all of the CSRs delineated with Cl and Si show enrichment of one and depletion of the other. The CSRs delineated by Cl and H, such as those in the western Medusae Fossae formation area, are consistent with their positive correlation [Karunatillake et al., 2006; Keller et al., 2006] in multivariate space as well. Last, Ca, which shows only weak multivariate correlations with others in preliminary analyses (e.g., Karunatillake et al., submitted manuscript, 2009a), does not yield any CSRs that satisfy our area thresholds.

[19] Summary comparisons with the global distributions of mapped geologic units, thermal inertia, albedo, and rock areal fractions in the following sections complete our overview of the CSRs. The CSRs that we have delineated are consistent with the surface type 2 observations by Karunatillake et al. [2006]. Nevertheless, as anticipated by Karunatillake et al. [2006, paragraph 49], they reveal regional exceptions to these general trends, particularly in equatorial regions where the areal fractions of surface types 1 and 2 are both high relative to their global distributions. A detailed investigation comparing and contrasting the eleven potential [Rogers et al., 2007a] mineralogy type distributions [e.g., Rogers and Christensen, 2007] with these and other CSRs may yield a better understanding of surficial processes and their variation at depth as Rogers et al. [2007b] and Wyatt [2007] demonstrated prefatorily. Taylor et al. (submitted manuscript, 2009) provide an overview of such comparisons in the context of vastly different sampling depths of the underlying instruments: the GRS at tens of centimeters, and the TES from 50 μ m [Christensen et al., 2004, section 3.1.4 paragraph 2] to probably no more than the Visible Near Infrared (VNIR) limit of 100 μ m [Poulet et al., 2007, paragraph 21].

3.1. Geologic Overview

[20] It is intriguing that the CSRs do not appear to follow the spatial patterns of mapped geologic units [e.g., Skinner et al., 2006] in spite of local contributions [Newsom et al., 2007; Gasnault et al., submitted manuscript, 2009; Taylor et al., submitted manuscript, 2009] and chemical trends at the planetary dichotomy [e.g., Dohm et al., 2009]. Since the GS is only sensitive to compositions at tens of centimeter depths, this may indicate that the surficial processes and compositions reflected in the GS data are not closely linked to the underlying geology. If so, future investigations with CSRs may reveal overlap with surficial features controlled by the surface-atmosphere interface instead, such as gullies [e.g., Levy et al., 2009, Figure 3] and dissected mantles [e.g., Milliken and Mustard, 2003]. Even though such overlap may be tenuous except where features are abundant at regional scale, as we discuss in the context of reticulated bed forms of our case study in section 4.3, chemical constraints established at regional scale may be valid for localized occurrences of similar features.

[21] Given the GS's inherently coarse resolution (section 2.1), we compile the areal fractions of mapped geologic units at 0.5° × 0.5° spatial resolution within each CSR to evaluate the extent of direct spatial overlap between CSRs and particular geologic units. Since the CSRs are binned at 5° × 5°, we minimize the loss of geologic information by simply rebinning them at 0.5° × 0.5°. We use a sinusoidal equal area projected digital atlas of geologic units with a maximum resolution of 0.0625° × 0.0625° at the equator (E. Guinness, personal communication, 2003) linearly interpolated by row to a 0.5° × 0.5° equirectangular grid [e.g., Snyder, 1987, p. 248] at zero interpolation order.

[22] Our geologic atlas is based on the Viking legacy I-1802 series (available online http://webgis.wr.usgs.gov/pigwad/down/mars_geology.htm) and lacks recent updates/revisions with MOLA and other data sets [e.g., Skinner et al., 2006]. However, preliminary comparisons indicate that at the coarse spatial resolution of the GS, area calculations with the updated maps would converge with the old. Therefore, we present our results based on the I-1802 series in Figures 67 both for relative age and mapped geologic units.

[23] As evident in Figure 6, one age unit typically dominates areally over others within most CSRs. However, we do not observe a consistent association of particular sets of elements or of particular CSRs with specific age units. Given the contiguous spatial extent of each age unit, this suggests that the apparent areal dominance of one within a particular CSR may just be coincidental instead of reflecting chemical processes representative of a particular era. Nevertheless, the dominant geologic age group may provide general constraints on surficial processes in detailed investigations, such as the area we selected for our case study (Figure 5).

Figure 5.

Sketch of our case study region (sky blue outline) that is marked by Cl enrichment and Fe, Si depletion along with the CSRs that surround it: {ClSi ED 1.5s 15°} in lime and {ClSi ED 1s 15°} in purple. Consistent chemical trends of Cl enrichment and Si depletion highlight our region. CSR to the west outlined in red is {ClH EE 1s 15°}. Overlain on the MOLA elevation map from PIGWAD at 1: 2.8E7 lateral scale. Reddish hues indicate higher elevation, while bluish hues indicate lower elevation in the MOLA map. HiRISE images that are used to characterize the surface in section 4.3 are also indicated by numerical tags corresponding to Table 5. The possibility of HiRISE sampling bias is noted in section 4.3.

[24] As shown in Figure 6 the areal fractions of geologic units and secularly categorized volcanic units are even less insightful than the relative ages since the areal dominance of a single unit within a CSR is quite rare. In fact, most CSRs contain similar areal fractions of many different units. The {K,Th DD 1s 10°} CSR in the vicinity of Elysium (Figure 3) is a possible exception where nearly 90% of the surface is underlain by Amazonian units of which ≈80% consists of Elysium lava flows (Figures 6 and 7). Similarly, nearly 90% of the {K,Th DD 1s 10°} CSR in the vicinity of Syria-Solis-Thaumasia planae is underlain by units interpreted to be Hesperian volcanics. These are also the only CSRs in which the three geologic units with the greatest areal fractions constitute nearly 90% of the total area (Figure 7). It is possible that detailed analyses of such CSRs would provide insight into igneous processes. The CSR in the vicinity of Thaumasia may be particularly insightful, as relatively low albedo (Figure 8) along with high thermal inertia (Figure 10) and rock areal fractions (Figure 9) suggest that the underlying bedrock may contribute meaningfully to the GS signal. In addition, comprehensive geologic investigations analogous to the work by Dohm et al. [2008] may reveal trends that are not apparent in our overview.

Figure 6.

The relative areal fractions of secular and volcanic units within each CSR according to the I-1802 series. “N” indicates Noachian, “H” indicates Hesperian, and “A” indicates Amazonian. The prefix “V” indicates the corresponding volcanic units. The area of the volcanic units corresponding to each secular unit is shown in a lighter fill color. The total area often does not sum to 100% since some units are uncategorized. Data source and processing are described in section 3.1.

Figure 7.

The three mapped geologic units with the highest areal fractions are shown for each CSR, with the total areal fraction of remaining units indicated in purple. The geologic unit notation of the I-1802 series is utilized as described in Figure 6.

Figure 8.

Qualitative comparison of the average albedo of each CSR with the global albedo distribution. Data sources are identified in sections 3.2. The “global” box plot marks the 25th percentile, median, and 75th percentile of the global distribution. Regional averages at or above the 75th percentile of the global distribution are identified by black columns, intermediate are identified by light gray, and at or below the 25th percentile are identified by solid outlines. Each CSR is also tagged with the name of a nearby geographic feature to facilitate identification in Figure 3. Column charts are omitted for CSRs that lie mostly beyond the areal bounds of corresponding data sets. Similarly, Figures 9 and 10 compare rock areal fraction, surface type 1 areal fraction, surface type 2 areal fraction, and thermal inertia.

Figure 9.

Qualitative comparison of the (top) average rock areal fraction and (bottom) surface type 1 areal fraction within each CSR with the corresponding global distributions as described in Figure 8.

3.2. Overview of Thermally Derived Attributes

[25] Remote sensing observations at Thermal InfraRed (TIR) wavelengths by the Viking missions first enabled detailed characterization of key surficial properties of the Martian surface, including albedo and thermal inertia. The latter was used in conjunction with radar reflectivity variations to estimate the areal fraction of fragments no smaller than 0.1 m, termed “rocks” [Christensen, 1986]. While the spatial resolution and estimation methods have been refined with more recent instruments such as the MGS-TES [e.g., Putzig et al., 2005; Christensen et al., 2001], the resulting maps are generally consistent with their predecessors. In light of this and the context of our general discussion at the coarse spatial resolution of the GS, we use thermal inertia, albedo, and rock areal fractions derived with the Viking InfraRed Thermal Mapper (IRTM) [Christensen, 1986] as described by Karunatillake et al. [2006]. The average value of each (computed as the arithmetic mean) relative to the corresponding global distribution is qualitatively illustrated in Figures 810.

Figure 10.

Qualitative comparison of the (top) average surface type 2 areal fraction and (bottom) average thermal inertia within each CSR with the corresponding global distributions as described in Figure 8.

[26] In general, apparent rock areal fraction, thermal inertia, and albedo values within CSRs are consistent with the observations in multivariate space [e.g., Karunatillake et al., 2006; Keller et al., 2006; Karunatillake et al., submitted manuscript, 2009a]. For example, the northern CSRs with K and Th enrichment also feature higher rock areal fraction and thermal inertia values, along with low-albedo values (Figures 8, 9, and 10). On the other hand, the CSRs in SW Tharsis with Cl enrichment and Si depletion have high albedo, low thermal inertia, and low rock abundance. It is also intriguing, and perhaps indicative of primary mineralogic effects, that the majority of the CSRs are relatively high thermal inertia and low-albedo regions (Figures 8 and 10). Our subsequent case study involving the region of Cl enrichment, Fe depletion, and Si depletion (Figure 5) demonstrates how these attributes (specifically thermal inertia) can act as important constraints for candidate surficial processes. Nevertheless, it is important to recall that even the areas with the highest rock abundances are dominated by fine material, and as such, the chemistry may be more indicative of variations in the fine component than of underlying bedrock [Newsom et al., 2007].

4. Region Among Volcanic Edifices: A Case Study

[27] In our case study, we focus on the chemical signatures evident in the CSR delineated as {ClSi 1s ED 15°} (Table 3 and Figure 3) denoting (E)nrichment of Cl and (D)epletion of Si at 1s confidence exceeding the area of a 15° radius cap (Table 1). Abundances of these elements differ from global averages by over 1.5s throughout most of its SW portion. In turn, within the S portion of the {ClSi 1.5s ED 15°} CSR lies a region also depleted in Fe at better than 1s confidence. In spite of being smaller than the area threshold (Table 1) by ≈20%, we choose this smaller region for our case study due to two key reasons: (1) Consistent chemical trends in a spatially nested pattern that highlight the region and (2) the marked depletion of both major elements detected by the GS, Fe and Si, making the region particularly unusual. This region, along with the larger CSRs that contain it, is illustrated in Figure 5. As one that does not quite satisfy our conservative area thresholds (Table 1), the insight we glean from it further reinforces the utility and significance of the (larger) CSRs in general. Given its proximity to Daedalia Planum and Tharsis volcanoes, we refer to it (outlined by the sky blue contour in Figure 5) as Region Among Volcanic Edifices (RAVE) throughout the rest of this work. We begin our discussion with the insight into RAVE from other instruments and missions that would be relevant at the tens of centimeter GS sampling depths.

4.1. Radar Stealth and Bulk Density

[28] Even though we selected RAVE (Figure 5) solely due to spatially convergent chemical signatures across several CSRs, it overlaps notably with a region highlighted by two independent data sets: radar reflectance observations from Earth at 1.35 cm [Ivanov et al., 1998] and 3.5 cm [Edgett et al., 1997] wavelengths. At both wavelengths, much of this region is an efficient absorber producing the classic signature of Stealth. The visually impressive nature of this overlap illustrated in Figure 11 is particularly relevant as the free space vertical resolution at 1.35 cm and 3.5 cm wavelengths yields a coarser effective resolution that is comparable to the GS sampling depths. In fact, the bulk of the Stealth region at 3.5 cm overlaps much more compellingly with RAVE than it does with the easternmost Medusae Fossae formation (Figure 11). In contrast, the Stealth features apparent across the broad Medusae Fossae formation at much longer wavelengths of 12.6 cm [Harmon et al., 1999], 15 m [Carter et al., 2008], and ≈150 m [Picardi et al., 2005] (corresponding to effective vertical resolutions greater than the GS sampling depths) do not overlap with RAVE.

Figure 11.

Sketch of RAVE (sky blue) overlain on Stealth region at 3.5 cm and mapped geologic units (adapted from Edgett et al. [1997, Figure 4]). Note the striking spatial overlap between RAVE and Stealth of greatest confidence (hatched region) relative to the Medusae Fossae formation units: Amm, Amu, and Aml. The surrounding CSRs include {ClSi ED 1.5s 15°} in lime and {ClSi ED 1s 15°} in purple. Surficial chemical differences between eastern and western Medusae Fossae are revealed by {ClH EE 1s 15°} CSR (red) to the west and {ClSi ED 1s 15°} to the east.

[29] The spatial overlap between RAVE and 3.5 cm Stealth [Edgett et al., 1997] has significant implications for the physical properties of the bulk material. Specifically, bulk density constrained by the real component of the dielectric constant is just 0.4E3 kg m−3 [Ivanov et al., 1998]. This is consistent with the 1.9E3 kg m−3 upper bound for bulk density in the Medusae Fossae formation [Watters et al., 2007; Keszthelyi and Jaeger, 2008]. Such low bulk densities, particularly the former, are indicative of high porosity at GS sampling depths [e.g., Watters et al., 2007; Keszthelyi and Jaeger, 2008]. For comparison, porosity of terrestrial sedimentary rocks is typically ≈20% but can be as high as 50% [Chang et al., 2006, Figures 1, 2, 3].

4.2. Thermal Observations

[30] The bulk physical properties inferred for RAVE with radar reflectivity may be characterized more descriptively with observations at Thermal Infrared (TIR) and VNIR wavelengths. As expected, the TIR-derived rock areal fraction, albedo, and thermal inertia maps, which we discussed in section 3.2, highlight RAVE and the surrounding CSRs as those dominated by fine material. In particular, RAVE is contained entirely within a low thermal inertia/high-albedo unit as delineated with the TES (Figure 12) [Putzig et al., 2005]. On the basis of apparent thermal inertia, Putzig et al. [2005, section 3.2.1, Figure 5 unit A] infer the fine material on the RAVE surface to be particles no greater than 40 μ m across. Such material, classified as coarse-to-medium silt on the Wentworth scale [e.g., Lewis, 1984; Encyclopedia Britannica online, http://search.eb.com/eb/art-61244], is termed “dust” by us.

Figure 12.

Sketch of RAVE (sky blue line) overlaid on the thermal inertia/albedo unit map (adapted from Putzig et al. [2005, Figure 5] with permission from Elsevier). RAVE is contained entirely within the low thermal inertia/high-albedo unit (blue) as is the bulk of the two surrounding CSRs. These are {ClSi ED 1.5s 15°} outlined in lime and {ClSi ED 1s 15°} outlined purple. The {ClH EE 1s 15°} to the west is outlined red and its southern portion is filled in green indicating high thermal inertia and medium albedo.

[31] Diurnal and seasonal variations in apparent thermal inertia have indicated that hardpan, and perhaps even duricrust, is ubiquitous on the planet at ≈3 km spatial resolution occuring vertically layered or laterally mixed with other material [Putzig and Mellon, 2007, section 3.1, Figures 15 and 16]. Model fit uncertainties preclude a clear determination of these combinations within RAVE, but marked seasonal variation in apparent thermal inertia is evident [Putzig and Mellon, 2007, Figure 7 (bottom) and section 3.1]. Vertical layering involving duricrust/hardpan buried shallower than the seasonal skin depth is a distinct possibility. The thermal skin depth is highly sensitive to the degree of cementation; Putzig and Mellon [2007, section 2.3 and Table 2] estimate seasonal skin depth to be on the order of 0.2 m (diurnal: 8 mm) for dust and 2 m (diurnal: 9 cm) for (sulfate cemented) hardpan. For example, if the GS signal over RAVE were dominated by indurated bed forms as opposed to the fine material mantle and if the two were chemically distinct, we would anticipate the former to be buried shallower than the tens of centimeter GS sampling depth.

4.3. Surficial Morphology

[32] We utilize HiRISE images to test the hypothesis that RAVE consists of indurated material shallowly buried beneath dust with low density (section 4.2). The spatial distribution of the 50 HiRISE images that we viewed within RAVE is shown in Figure 5 along with identifiers in Table 4. Even though HiRISE images are available over most GS bins within RAVE, the sampling frequency is nonuniform (Figure 5) as the targeting was guided not by our work, but by research interests of the HiRISE team. Targeted imaging at additional locations within RAVE in the future should reveal whether there has been a resulting sampling bias.

Table 4. HiRISE Image Identifiers, Approximate Coordinates, Image Tags Used in All Figures and Tables of This Work, and Map-Projected Scalea
HiRISE IdentifierLatitude (deg)East Longitude (deg)Image TagResolution (cm/pixel)
  • a

    North is up in all HiRISE images.

PSP_001511_1730−7−129.6125
PSP_001656_18353.3−126.3225
PSP_002157_1715−8.4−119.9350
PSP_002289_1725−7.3−123.8425
PSP_002764_18000−133.1525
PSP_002922_1725−7.3−123.8625
PSP_003410_18403.8−129.9750
PSP_003647_1745−5.5−118.6825
PSP_003832_18403.8−129.9950
PSP_004056_1735−6.4−125.41025
PSP_004201_0735−6.6−123.91125
PSP_004346_0755−4.3−121.81225
PSP_004412_0715−8.6−123.61325
PSP_004570_0705−9.6−116.91450
PSP_004689_0765−3.6−125.81525
PSP_004702_0705−9.4−120.61625
PSP_004768_0705−9.4−1211725
PSP_004781_0710−9−117.51850
PSP_004847_0745−5.5−118.61925
PSP_004913_0735−6.3−1212025
PSP_005058_0720−7.8−119.72125
PSP_005071_0775−2.7−1152225
PSP_005124_0705−9.2−120.32325
PSP_005203_0730−6.7−119.52425
PSP_005256_0735−6.6−123.92525
PSP_005296_0705−9.3−137.82625
PSP_005349_0665−13.4−143.92750
PSP_005375_0675−12.5−134.82825
PSP_005414_0735−6.5−1202925
PSP_005467_0700−9.8−125.73050
PSP_005612_0700−9.9−125.53150
PSP_005625_0730−6.7−119.53225
PSP_005665_08000−133.13325
PSP_005691_0685−11.3−121.13425
PSP_005770_0745−5.5−118.63525
PSP_005783_0775−2.7−1153625
PSP_005836_0735−6.3−1213725
PSP_005902_0700−10−1223850
PSP_005915_0720−7.8−1173925
PSP_005916_0665−13.3−143.74025
PSP_005942_08252.5−135.94125
PSP_005995_0700−9.7−141.84225
PSP_006192_0700−9.8−120.44325
PSP_006601_08252.4−129.34425
PSP_006680_0740−6.2−1254525
PSP_006693_0755−4.3−120.84625
PSP_006759_0700−9.9−121.14725
PSP_006773_0735−6.4−144.24825
PSP_006878_08050.3−132.14925
PSP_006904_0755−4.3−120.85025
PSP_003331_0580−21.6−1305150

[33] The morphology of the broader region (Figure 5) has already been discussed by others, including Bridges et al. [2007, 2008] and Keszthelyi et al. [2008]. In general, the tens-to-hundreds of meter-scale topography of the broader area is obscured by a mantle of material consisting of particles that are too fine to be resolved by HiRISE [Bridges et al., 2007; Keszthelyi et al., 2008] and its predecessor, the Mars Orbiter Camera (MOC) [e.g., Bradley et al., 2002].

[34] A bed form that is ubiquitous in the broader area (Figure 13) is coined a “reticulate bed form” [Bridges et al., 2009]. Ridges delineate nested patterns, leading to reticulation at meter, tens of meter, and hundreds of meter size scales as described by Bridges et al. [2008]. This bed form type is present almost exclusively within low thermal inertia regions that include RAVE, and resembles the surficial morphology of some wind-eroded features such as ones in Valles Marineris and White Rock within Pollack Crater [Bridges et al., 2008; Keszthelyi et al., 2008]. The association with low thermal inertia surfaces and wind erosion has been interpreted to indicate an eolian origin for reticulate bed forms [Bridges et al., 2008]. The similarity with wind erosion features and some light-toned bedrock [Bridges et al., 2009] would be consistent with induration of the reticulate bed forms. While present throughout RAVE, we find this bed form to be pervasive and more easily discernable primarily on the flanks and caldera of Arsia Mons. Farther out, it is overlain by a veneer of fine material as shown in Figure 13.

Figure 13.

Examples of reticulate bed forms within RAVE in grayscale. Image tags identify the approximate locations in RAVE as shown in Figure 5, with corresponding HiRISE image details in Table 5. Parenthetic coordinates of each excerpt indicate its approximate location within the larger HiRISE image. Solid black circles are 10 m across, while the glassy circles are 20 m across. Excerpt tagged 40 shows the generally subdued nature of reticulate bed forms away from Arsia, unlike surfaces on the flanks and caldera shown in the excerpts tagged 3, 8, and 16. Excerpt 3 is also a good example of prominent “honeycomb” shaped reticulate bed forms, while excerpt 8 shows the “accordian” distortion. The close spatial association of potential star dunes and reticulate bed forms on a crater (560 m across) floor just south of RAVE is apparent in the excerpt tagged 51.

[35] A second type of bed form, more common within RAVE than reticulate bed forms, consists of en echelon hollows, such as the examples in Figure 14. We term these “lenticular bed forms.” If eolian in origin, these could be deflation hollows. Surfaces dominated by lenticular bed forms sometimes have a braided appearance, remarkably similar to those in the Medusae Fossae formation that Bradley et al. [2002, section 5, paragraphs 22 and 26, Figure 10] interpreted as bidirectional yardangs (Figure 14, bottom).

Figure 14.

Examples of lenticular bed forms that occur at varying size scales within RAVE. Scale indicated next to image tags by the solid black circle is 20 m across, the glassy circle is 40 m across, and the dimpled solid circle is 160 m across. For correspondence, shaded circles identify the same location in the (top, left) fine- and (top, right) coarse-scale excerpts. Excerpt tagged 7 illustrates the transitions among typical (triangular marker 1), small (marker 2), and smallest (marker 3) lenticular bed forms. The surface texture of the same excerpt at coarse scale on the right is comparable to the bidirectional yardangs in the Medusae Fossae formation as viewed by the MOC [Bradley et al., 2002, Figure 10] and indicates that the long axes are oriented parallel to the fossae. (bottom) Excerpt tagged 26 contains some of the largest lenticular bed forms in RAVE. Yellow crosses are processing artifacts.

[36] In addition to the reticulate and lenticular bed forms, we are able to identify three other types of surface morphologies within RAVE. All three bear a strong similarity to terrestrial eolian formations (linear, barchanoid/transverse ridges, and ripples) as illustrated in Figure 15. The ripples are particularly difficult to discern due to their small size (ripple wavelengths average 1.7 m) except where illuminated obliquely.

Figure 15.

Examples of additional surface morphologies within RAVE. Solid black circles are 10 m across. Excerpt tagged 20 shows tentatively linear bed forms, 44 shows barchanoid ridges, and 10 shows ripples. The distinction between excerpts 20 and 10 is subtle, suggesting that what we consider to be linear bed forms may in fact be large-scale indurated ripples instead of true linear dunes. Yellow crosses are processing artifacts.

[37] Could all five types be primarily eolian in origin? The last three appear most clearly eolian due to their strong similarity to terrestrial dune formations, with the barchanoid/transverse ridges in particular comprising the largest equatorial dune field on the planet [Edgett, 1997, Figure 1, Item B]. These three are sometimes interspersed with the two dominant bed forms, lenticular and reticulate, as in Figure 16. Such a transition from barchanoid, to ripples, to reticulate bed forms reinforces the possibility that reticulate bed forms are eolian in origin, which is likewise supported by a transition between linear and reticulate bed forms (Figure 16). Last, the transition between lenticular and reticulate bed forms across a narrow gap connecting two pits shown in Figure 16 is consistent with an eolian origin to the lenticular bed forms. Such gaps are particularly likely to exert local topographic control on eolian turbulence.

Figure 16.

Potential transitions among bed forms. Solid black circles are 10 m across. Excerpt tagged 20 indicates a potential transition between reticulate and linear bed forms, and excerpt tagged 22 indicates a potential transition between lenticular and reticulate bed forms across a topographic gap (imperceptible at fine resolution). Excerpt tagged 44 is a potential transition area among barchanoid ridges with ripples on their lee sides, and reticulate bed forms, with the barchanoid type dominating toward the NW beyond the excerpt border, while reticulate dominates to the SE.

[38] The distribution and variations of lenticular bed forms lend additional support to an eolian origin. Where present among fossae, their long axes are generally oriented parallel to topographic ridges (Figure 14), consistent with topographic control of wind direction. Their seamless transition from typical sizes to smaller sizes (Figure 14) may also reflect eolian modification at several different spatial scales. Variations of reticulate bed forms, typically between symmetric “honeycomb” and distorted “accordion” shapes (Figure 13) [Bridges et al., 2008] as well as potential similarities with star dunes further reinforce an eolian origin. In fact, a crater floor about 7° south of the RAVE perimeter shows tentative evidence of a transition between star dunes and reticulate bed forms (Figure 13).

[39] Even though all five bed form types may be consistent with eolian origins, whether the requisite eolian conditions have existed at these high elevations (mostly >2.5 km) is less clear. If the bed forms were to consist entirely of (unconsolidated) dust, current models of wind speed versus grain size to initiate saltation [e.g., Almeida et al., 2008; Merrison et al., 2007] pose a significant hurdle to forming any bed forms. Electrostatic aggregates are also unlikely to saltate, as they usually disaggregate upon entrainment [e.g., Sullivan et al., 2008, paragraphs 67 and 75] and suspend in the atmosphere. A reasonable alternative is that dust grains cemented into aggregates, e.g., by salts, saltated to generate the bed forms. Higher wind speeds needed in the low-density atmosphere of RAVE to initiate saltation [Greeley et al., 1976] may be achieved by a combination of turbulence due to local topography, katabatic winds from nearby volcanoes [e.g., Bridges et al., 2009], and equatorial east-to-west winds [e.g., Benson et al., 2006, section 3.2]. Nevertheless, recent models indicate that once initiated, saltation of sand-sized particles/aggregates may be sustained at speeds as low as 1 ms−1 [Almeida et al., 2008; Merrison et al., 2007].

[40] Accepting an eolian origin to all five bed forms in RAVE, are they currently active or inactive? Edgett [1997] inferred that the barchanoid/transverse dune field within RAVE is inactive due to burial by dust. Alternatively, exhaustion of particles suitably sized for saltation [e.g., Sullivan et al., 2008] may have inactivated these bed forms. Obviously, a spatially varying combination of indurated bed forms, inactivated dunes, and active dunes is also possible.

[41] However, we suggest that a different scenario is more likely within RAVE for some bed forms: Induration as opposed to just inactivation. One indication of this possibility is the similarity of reticulate bed forms to those found on surfaces such as White Rock that has been interpreted by Ruff et al. [2001] as indurated eolian sediment. More compelling is the resistance of the bed forms to disruption by meter-scale craters that postdate them (Figure 17). Where bed forms are disrupted by fresh impact craters, arcuate overhanging ridges are generated sometimes (Figure 17), suggesting induration to meter-scale depths. These potentially indurated bed forms are unlikely to be thicker than 2 m, as Fergason et al. [2006, section 5.1], infer given the visibility of underlying decameter-scale degraded impact craters. This is consistent with the protrusion of meter-scale blocks in places, particularly where the bed forms appear to postdate block slides as shown in Figure 18.

Figure 17.

Examples that illustrate the strength of hardpan/duricrust within RAVE. Solid black circles are 10 m across, and the glassy circle is 160 m across. (top) Excerpts tagged 2 and 28 show bed forms remaining intact subsequent to the formation of fresh impact craters (crater cluster in 2; crater ≈5 m across in 28). Even when bed forms were disrupted by meter-scale crater forming impacts (bottom left, tagged 41 at coarse scale), overhanging arcuate ridges had formed (bottom right, at fine scale), indicating significant induration at meter-scale depths.

Figure 18.

Example of (left) block slides that predate bed form formation and (right) blocks that protrude from the bed forms. Solid black circle is 10 m across, and the glassy solid circle is 20 m across. The typical size of blocks constrains the thickness of the bed forms to meter scale.

[42] Induration would certainly not be unusual, since duricrust in excess of many meters [Pain et al., 2007] has been inferred across a variety of Martian locations. Local relief related to duricrust has been identified at places as disparate as the flanks of Olympus Mons, Valles Marineris, and Arabia [Pain et al., 2007, Figure 3]. As we discussed in section 4.2, such observations are bolstered by TIR observations as well.

[43] We posit further that what we infer to be indurated bed forms are likely buried beneath a veneer of dust, as evident in the low apparent thermal inertia of the RAVE surface (section 4.2). The presence of km-scale albedo banding, kilometers long potential dust devil tracks (active dust devils have been imaged in the general area as described by Cantor et al. [2006] and Cantor [2007]), and slope streaks that do not disrupt the bed forms make a strong case for such a veneer. As evident in Figure 19, these features generally fail to show discernable topographic effects in HiRISE images, suggesting that the dust veneer is unlikely to be thicker than some fraction of the tens of centimeter scale HiRISE image resolution (section 2.3).

Figure 19.

Examples of surficial features distinguishable only by albedo with little if any distinctive textures in HiRISE images within RAVE: Albedo banding oriented W–NW and E–SE visible in excerpt tagged 1, dust devil tracks in excerpt 45, and slope streaks in excerpt 48. However, potential avalanche scars that are distinct in topography but not albedo are also present as identified in excerpt tagged 2. Solid black circles are 10 m across, the dimpled solid circle is 40 m across, and the glassy solid circle is 160 m across.

[44] The surficial morphology that we have discussed so far with HiRISE images appears to converge with inferences made with radar reflectivity (section 4.1) and TIR observations (section 4.2) that RAVE may consist mostly of one to two meter thick indurated bed forms that are buried by dust shallower than the tens of centimeter sampling depth of the GS. In subsequent sections, we seek to constrain induration processes and the origin of the overall subsurface within RAVE by considering potential roles of volcanism, glaciation, and climate in the context of GS-derived chemistry.

4.4. Volcanism

[45] As shown in Figure 5, the closest volcanoes to RAVE are Arsia Mons, Biblis Patera, Pavonis Mons, Ulysses Patera, Olympus Mons, and Ascraeus Mons, in the order of increasing distance. With the exception of Biblis Patera and Ulysses Patera, which are inferred to be no younger than Hesperian in age (older than ≈1.9 Ga) [Plescia, 1994], volcanism in the remaining edifices may be as recent as Late Amazonian [e.g., Neukum et al., 2004]. However, late stage volcanism was probably not as extensive or voluminous as it was in late Hesperian/early Amazonian [Dohm et al., 2007; Scott and Dohm, 1997]. RAVE is essentially surrounded by the largest cluster of the highest volcanoes on the planet. Given the low areal density of impact craters [e.g., Edgett, 1997, p. 108] within RAVE, the younger Amazonian volcanic events may be more relevant to the evolution of its surface.

[46] In fact, most of the underlying mapped geologic units of RAVE are Amazonian in age and volcanic in origin (Figures 6 and 11) with the possible exception of a slide deposit on the W flank of Arsia Mons interpreted by some to be of an Amazonian glacial origin (section 4.5) suggesting that the surficial bed forms overlying them are even more recent. Specifically, sections of the Olympus Mons and Ascraeus caldera floor complexes may be as young as 100 Ma, and that of Arsia ≈130 Ma [Neukum et al., 2004]. While significant uncertainties exist [e.g., McEwen et al., 2005], some of the lava flows on the Olympus Mons scarp have been estimated to be as recent at 2.5 Ma [Neukum et al., 2004]. The large-scale shape of most of these volcanic edifices and lava flow morphologies suggest that they are shield volcanoes with low-viscosity lava flows akin to the Hawaiian type [e.g., Dohm et al., 2008, section 2.3].

[47] However, a host of features including putative cinder cones at the Pavonis Mons summit [Mouginis-Mark, 2002, paragraph 3] and southern flank [Keszthelyi et al., 2008, paragraphs 30–31 and Figure 6c], pit craters at elevations 5 km–7 km below caldera rims [Mouginis-Mark, 2002, Table 1] (Figure 20), and edifice morphometry along with theoretical implications of magmatic gas expansion in a low-density atmosphere suggest that the Tharsis/Olympus volcanoes may be composite volcanoes (synopsis by Hiesinger et al. [2007]). In addition, deposits within the Medusae Fossae formation, Candor chasm, Ophir chasm, and Arabia Terra have been interpreted to be constructs of Tharsis basaltic plinian eruptions occurring as recently as the late Amazonian [Hynek et al., 2003]. Olympus Mons eruptions may have also transitioned from less viscous, stable, and long-lived tube-forming systems to more viscous, episodic, and less stable channel-forming systems in the late Amazonian [Bleacher et al., 2007]. A higher abundance of channels relative to tubes is often characteristic of pyroclastic eruptions entailing greater volatile content [Bleacher et al., 2007, paragraph 30].

Figure 20.

Examples of pit craters within RAVE, about 314 m across excerpt tagged 46 and ≈200 m across excerpt tagged 20. Solid black circles are 20 m across.

[48] If explosive basaltic volcanism were a main contributor to the surficial material within RAVE, theoretical clast sizes upon eruption would be tens of μ m to a few mm, while accretion (along with significant hydration) could lead to sizes between 0.1 and 1 mm upon deposition with less welding than on Earth [Wilson and Head, 2007, sections 5 and 7]. The scoriaceous tephra/ash would grade to more sorted finer and thinner deposits with distance from the vent or eruptive column with dispersal distances varying from 20 km (for mm clasts) to >1E4 km (for clasts <50 μ m) [Wilson and Head, 2007, sections 6 and 7]. Even though source vents have not been identified, Hynek et al. [2003] infer the presence and thinning of such layers westward from Tharsis. Another possible source of surficial material could be similar fragments produced through the massive flank failure processes which are interpreted to have occurred on Olympus Mons flanks producing its aureole deposits [Morgan and McGovern, 2005; McGovern et al., 2004; Tanaka, 1985]. Such massive movements would produce tremendous amount of material of the spatial extent discussed earlier, which could then be transported aerially to our region of interest.

[49] Even if only some of the preceding interpretations were correct, significant pyroclastic deposits such as scoria and basaltic ash may have been present surficially, reworked chemically and mechanically into fine material, and distributed regionally from the flanks of the volcanoes into our region. While such surficial volcanic contributions may be possible, it is important to remember that the volcanic material underlying RAVE (Figure 11) is probably too deeply buried to generate a signature in the γ or TIR spectra (sections 4.2 and 4.3). Nevertheless, chemical composition of the overlying deposits may be influenced by such material [cf. Newsom et al., 2007].

4.5. Glaciation

[50] Akin to volcanic units in RAVE, any glacial units are likely buried too deep to be evident as a strong enrichment in the GS stoichiometric H2O map. In fact, RAVE is only slightly enriched in H2O (Figure 3) as we discuss later in section 5, and only a tenuous spatial link exists between putative relict glaciers and interpolation of NS-based H maps [Elphic et al., 2005, Figure 1]. Nevertheless, any relict glaciers that are present nearby may have indirect chemical and physical effects though perhaps not as pronounced as inferred on the basis of long-wavelength radar reflectivity at km depth scales for the Medusae Fossae formation (section 4.1).

[51] Relict glaciers have been hypothesized to exist primarily NW of each of the volcanoes, consistent with precipitation under higher obliquity [e.g., Forget et al., 2006, Figure 1] and with compelling morphologic similarities to Antarctic piedmont glaciers [e.g., Shean et al., 2007; Russell and Head, 2007, p. 327]. The largest among them, extending 350 km [e.g., Head et al., 2005, p. 348 bottom] along the NW flank of Arsia, lies in the SW portion of RAVE and may have formed as recently as 65 Ma ago [Shean et al., 2007]. The corresponding surface area and volume are estimated to be 166E3 km2 and 3E5 km3, respectively [Shean et al., 2007, paragraph 16]. Incidental evidence for much larger ice-rich deposits encompassing the Medusae Fossae formation include pedestal craters, radar loss tangent values, and layered terrain: all of which show more than a passing similarity to their counterparts in the polar layered terrain [e.g., Schultz, 2007].

[52] Large-scale flow features associated with the largest graben in RAVE (on the NW flank of Arsia Mons) have been interpreted by Shean et al. [2007, Figure 4], to indicate underlying relict piedmont glaciers. They also estimated that a meters thick debris cover could have shielded an underlying ice matrix as thick as 300 m from sublimation over the last 50 Ma [Shean et al., 2007, paragraph 65]. Features on the flank of Olympus Mons have also been interpreted as evidence of hydrothermal activity in the presence of ice [Neukum et al., 2004, p. 977].

4.6. Climate

[53] Glacial hypotheses are underpinned by the Martian climate cycle which is in turn significantly affected by the obliquity cycle. While the temporally chaotic nature of Martian obliquity prevents precise retrograde modeling to Hesperian and older eras, variations within the last tens of Ma relevant to our investigation of RAVE (section 4.4) may be modeled quite precisely [Laskar et al., 2004]. Such modeling indicates that obliquities of 42° and higher were more likely than the current value, with the most recent such occurrence ≈5 Ma ago [Laskar et al., 2004, sections 3.2.1 and 3.2.2, Figures 9 and 10a]. Assuming an atmospheric volatile budget including polar volatiles similar to that of current Mars [Forget et al., 2006, p. 370], global circulation models estimate that 20 μ m–50 μ m scale H2Os (cf. 6 μ m–8 μ m in current Tharsis clouds) would precipitate at column rates of 30 mm a−1–70 mm a−1. Such high rates of precipitation would be capable of generating hundreds of meters thick glaciers within a few thousand years [Forget et al., 2006, p. 370]. Global Circulation Models (GCMs) would be consistent with the hypothesized relict glaciers particularly since the predicted precipitation is localized over them [Forget et al., 2006, p. 370]. Given these retrograde predictions of high precipitation rates, it is plausible that regionally pervasive ground ice would have accumulated throughout RAVE in addition to the spatially localized glaciers.

[54] In conjunction with high precipitation, instances of higher obliquity on Mars could have caused net deposition of atmospheric dust throughout RAVE and the broader Tharsis region (Figure 5) as predicted by GCMs [Haberle et al., 2006, paragraph 14]. Along with dust deposition, high precipitation may have contributed to glaciation, aqueous chemical processes, and the formation of complex eolian bed forms such as those potentially buried beneath a veneer of dust that we discussed in section 4.3. Such formation would have been facilitated by a potentially denser atmosphere, if the thick CO2 inventory model applies [Manning et al., 2006], lowering the threshold speeds for entrainment and saltation of particles [Greeley et al., 1976] under higher obliquities as well.

[55] The current 25° Martian obliquity does not lead to high precipitation within or in the neighborhood of RAVE. Nevertheless, present atmospheric conditions result in perennial H2O(s) cloud cover over the SW flank of Arsia Mons [Noe Dobrea and Bell, 2005], orographic H2O(s) clouds over neighboring Olympus, Pavonis, and Ascraeus summits [Benson et al., 2006], ground H2O fog [Feldman et al., 2005, paragraph 27], and GCM prediction of light H2O(s) precipitation [Feldman et al., 2005, Figure 5]. Consistent with the persistence of regional E to W winds [e.g., Benson et al., 2006, section 3.2], the H2O(s) clouds are distributed W–NW over Olympus, Ascraeus, and Pavonis Mons showing interannual and seasonal variability (cloud activity between Ls = 0, Northern spring, and Ls = 220, before winter, with peak area near Ls = 100 just after summer solstice) [e.g., Benson et al., 2006, section 3.2]. The collective effect of these conditions, though probably insufficient to generate pervasive aqueous solutions within the surficial material of RAVE, may nevertheless be sufficient to create occasional concentrated brines.

[56] In contrast to the net deposition of atmospheric dust over Tharsis at higher obliquities, GCMs predict a somewhat complicated dust exchange between the surface and atmosphere under the current obliquity. In these models, the flanks of the Tharsis volcanoes, of Olympus Mons, and the broader Tharsis low thermal inertia region are net deflation areas at about 1 μ m a−1 [Kahre et al., 2006, paragraph 46], while RAVE is mostly a net deposition area [Kahre et al., 2006, paragraph 44 and Figure 7]. Based on results at 22.5 ka and 72.5 ka, GCMs also predict these conditions to prevail over the orbital precession cycle on the order of 50 ka [Haberle et al., 2006].

[57] In summary, the climatic conditions during the late Amazonian are likely to have produced complex eolian and aqueous processes. Overall, the eolian processes may have led to net dust deposition on most RAVE surfaces, with perhaps some of the dust derived regionally via deflation from the flanks of nearby volcanoes and surroundings. While localized H2O glaciers and regionally pervasive ground ice may have persisted before 5 Ma, the recent climate has been mostly arid though moderated by ground fog and local frost.

4.7. Synopsis of Overviews

[58] The preceding overviews of geology (section 4.4), thermally derived attributes (section 3.2), surficial morphology at high resolution (section 4.3), ground ice/glacial conditions (section 4.5), and climatic conditions (section 4.6) over RAVE are consistent with indurated bed forms that may be a meter or two thick overlain by a veneer of dust.

[59] Our overview effectively constrains neither the composition of the material cementing the bed forms, nor the processes that formed them. In fact, a broad range of processes could yield indurating salts, such as leaching alteration of dust by low-pH brines derived from volcanic degassing (i.e., acid fogs); eolian deflation of salts that were generated under hydrothermal acid fog conditions on the flanks of volcanoes and their subsequent deposition in RAVE; significant concentrations of salt in atmospheric dust; local production of salts via aqueous processes facilitated by buried ground ice, relict glaciers, dehydration of sulfates [e.g., Tosca et al., 2008, Figure 18], or a combination thereof as H2O sources; and local production of salts in isochemical alteration of basaltic lapilli deposited from plinian eruptions.

[60] Clearly, additional information is needed to elucidate and evaluate candidate processes for the origin of RAVE. To this end, we discuss chemical considerations in the succeeding sections, which also indicate that alternatives to salts as indurating agents, such as Fe-bearing or siliceous minerals [e.g., Blatt et al., 1972, pp. 348–368; Encyclopedia Britannica online, http://search.eb.com/eb/article-9022044], are unlikely due to the depletion of Fe and Si in this region.

5. Origin of RAVE: What Is the Bulk Component?

[61] The chemical constraints on the bed form material and cementing agents just discussed can be established by considering the distribution of GS-derived elemental mass fractions within RAVE. The first-order chemical properties are those of Cl enrichment and Fe and Si depletion relative to the global average. Tentatively, this may reflect a mass dilution effect since Cl can be a proxy for salts, at least on Earth, and shows a statistically significant anticorrelation with both thermal inertia and Si mass fractions at global scales [Karunatillake et al., 2006; Keller et al., 2006]. However, mass dilution by sulfates instead of halides is strengthened by the potential dominance of sulfates under Martian low-pH aqueous conditions [e.g., Tosca et al., 2005, p. 129] and the lack of a strong correlation of Cl with potential cations at either Meridiani [Clark et al., 2005, sections 4.2.3 and 4.2.4] or Gusev [Clark et al., 2007a]. It is also possible that a significant fraction of the Cl substitutes into sulfate or mixed anion salts rather than forming pure chlorides. While halides have been identified tentatively with TIR spectra [Osterloo et al., 2008, Figure 2], their distribution appears decoupled from and their spatial extent minute relative to the GS Cl-enriched regions.

[62] Since the primary chemical signature of RAVE does not lead us conclusively to either minerals or chemical processes, we compare elemental mass fraction ratios within RAVE with those for the Rest of Mars (ROM) and for several types of material that have been analyzed in situ (e.g., Figure 21). We retain the rest of Mars distribution as a point of reference in comparisons involving all other types. The in situ types we use are those classified as “soils” at both MER sites, including surface dust; the rocks at both MER sites; and the Shergottite-Nakhlite-Chassignite (SNC) meteorites. To ease detailed comparisons, we divide each type of material into the classes and subclasses defined by the MER team.

Figure 21.

Scatterplots of oxide to SiO2 mass fraction ratios versus the SiO2 mass fractions for all soil classes at the MER sites, RAVE (legend, ClFeSi), and the rest of Mars (ROM). Remaining legends and sampling sols are listed in Table 5. GS-derived SiO2 mass fractions have been renormalized to H2O-free to enable direct comparison with APXS values. Sample error bars are shown on null values for RAVE, dust, and ROM to one standard error (1sm) for each data set computed as the root-mean-square of numerical uncertainties. Scatterplots on the right highlight the consistent difference between RAVE ratios and dust, which is not observed for the other samples.

[63] We use ratios of mass fractions in lieu of the mass fractions themselves to address systematic differences between the GS and MER Alpha Particle X-ray Spectrometer (APXS) data [e.g., Karunatillake et al., 2007]. To facilitate comparisons with other chemical analyses in the literature, such as Total Alkali Silica (TAS) diagrams [Bas, 2000], we use SiO2 (renormalized to an H2O-free composition for the GS) as the abscissa in our comparative plots. Unfortunately, robust comparisons are currently limited to CaO, Cl, FeO, and K2O as the ordinates. We do not compare Al2O3 since its GS-derived values are being refined, H2O as it has only been estimated indirectly by the MER mission, and Th as it is undetectable by the MER APXS. The lack of robust Mg and S estimates with the GS and Th estimates with the APXS are particularly challenging, since they are necessary to evaluate the possibility of aqueous processes [cf. Taylor et al., 2006a].

5.1. Soils of Mars

[64] The soil classes that we use, the sols on which they were sampled by the MER APXS, and the literature upon which the sample selection is based are listed in Table 5. It is important to note that the chemical differences among some of these classes, such as surface dust [Yen et al., 2005] and Laguna class/Panda subclass soil, are subtle [e.g., Gellert et al., 2006, sections 11.2–11.4]. The work by Gasnault et al. (submitted manuscript, 2009) also suggests that RAVE, Meridiani, and Gusev may all be chemically similar at the GRS spatial resolution. Nevertheless, the classification appears reasonable given its utilization of the combined constraints from Mössbauer spectra, Miniature Thermal Emission Spectrometer (MiniTES) spectra, and MI textural information, as has been used successfully for rocks [e.g., McSween et al., 2008].

Table 5. Soil Classes Used, Sols on Which They Were Sampled by the APXS, and Legend Key for Figure 21
Soil ClassLegend KeyAPXS Sampling Sola
Berry class, Mooseberry subclassBMSoilM80,b M91,b M100,b M416,b M420,b M420B,b M443b
Berry class, Nougate subclassBNSoilM023,b M090,b M369,b M509b
Eileen Dean classESoilG1239,c G1246c
Gertrude Weise classGSoilG1190,c G1194,c G1199c
Laguna class, Boroughs subclassLBSoilG113,b G114,b G140,b G141b
Laguna class, Doubloon subclassLDSoilG502,c G611c
Laguna class, Liberty subclassLLSoilG47,d G135,d G280,d G315,d G428,d G477,d G814,c G847,c G831,c G1017,c M123,b M368b
Laguna class, Panda subclassLPSoilG43,d G49,d G50,d G74A,d G158,d G167,d G342,d G457,d G709,c G710,c M11,b M166,b M237A,b M237B,b M249b
MER surface dustDustG14,e G65,e G71,e G126,e G823,c G1352,c M25,e M60,e M90,e M123e
Paso Robles classPSoilG401,d G427,d G723,c G1013,c G1098c
GS validation trenchesGRSSoilG049,f G050,f G115,f M081,f M368f
AveragesLPPFSoilLaguna,c GPanda,b MPanda,b Pathfinderg

[65] We make two key observations after considering all in situ soil classes, including average Pathfinder soil and the trench samples that were used to validate the GS estimates: (1) The absence of a consistent overlap with any class (or even with a single sample) across the four ratios highlights the chemical uniqueness of RAVE; and (2) the strongest evidence for a simple mass dilution scenario is the comparison between RAVE and MER surface dust (denoted simply as “dust”). Figure 21 emphasizes this consistency by juxtaposing the comparison of all classes with one of only MER surface dust and the rest of Mars.

[66] Observation (2) highlights that while strikingly enriched in Cl relative to the rest of Mars, RAVE's Cl mass fractions are generally less than those of MER surface dust. In other words, the similar Cl/SiO2 ratio between the two, which, along with the similar FeO/SiO2 (Figure 21) and K2O/SiO2 (Figure 21) ratios and depletion of SiO2 in RAVE would be consistent with a simple mass dilution effect on MER surface dust as discussed in detail below. The significantly higher CaO/SiO2 ratio in RAVE (Figure 21) could act as a proxy for sulfates, such as CaSO4.2H2O (gypsum) or CaSO4 (anhydrite). Hydrated salts would also explain the slight enrichment of H2O in RAVE relative to the rest of Mars that is evident in Figure 22. In addition, the preliminary GS-derived S map suggests S enrichment in this area. These observations help to constrain our conjectures on the origin of RAVE in section 7.

Figure 22.

Modified box plots (refer to S. Karunatillake et al., submitted manuscript, 2009b) comparing the RAVE distribution to the rest of Mars distribution for each element and for ratios of particular elements of interest. Section 5.1 summarizes the features of the plot and the significance of the error bars.

[67] Since observation (2) favors mass dilution of MER surface dust by potentially sulfate-bearing salt, we explore its feasibility further with first-order estimates of the mass fraction of salts in a mixture that would yield the typical (median) mass fraction of FeO, Cl, K2O, and SiO2 within RAVE. With mass balance as the sole constraint, the mass fraction of salt in the mixture would vary between 5% and 15% depending on the oxide, with surface dust samples that appear to be composed of the finest material in Microscopic Imager (MI) images (e.g., sol 60 at Meridiani) yielding the greatest convergence. Across all samples, the mass fraction of CaSO4 only needs to be 3%–5% to account for the enrichment of Ca in RAVE relative to surface dust. That would amount to physically feasible bounds of 20%–100% of the diluting salt, with the rest unconstrained. For comparison, the areal fraction of cementing salts in the surface of White Rock duricrust, which we mentioned as a possible analog to the RAVE bed forms in section 4.3, is estimated to be less than 15% [Ruff et al., 2001, section 3 paragraph 1].

[68] A possible alternative to MER surface dust as the major component of RAVE is the average Martian crust [e.g., Taylor et al., 2006b] as represented by GS data over the rest of Mars. Given the large number of rest of Mars data, we may supplement the qualitative information in the ratio scatterplots (Figure 21) with a more quantitative comparison of distributions using modified box plots as discussed by Karunatillake et al. (submitted manuscript, 2009b). In essence, we compare the low values within RAVE with the high values of the rest of Mars (25th percentile of RAVE divided by the 75th percentile of the rest of Mars), high values of RAVE with low values of the rest of Mars (75th percentile of RAVE divided by the 25th percentile of the rest of Mars), and typical RAVE with typical rest of Mars (median RAVE by median rest of Mars). We estimate error bars conservatively as the per bin RMS uncertainty of RAVE and the rest of Mars propagated for the ratio of medians. The resulting plots are shown in Figure 22.

[69] The qualitative comparison between RAVE and the rest of Mars (Figure 21) suggests that CaO/SiO2 and K2O/SiO2 are similar, while the RAVE FeO/SiO2 values are significantly low. Even more striking, though already discussed in the delineation of RAVE, are the remarkably low SiO2 mass fractions and remarkably high Cl mass fractions in RAVE relative to the rest of Mars. These observations are consistent with the quantitative comparisons (Figure 22) suggesting a typical (median) enrichment of Cl by about 50%, depletion of Si by about 8%, and depletion of the Fe/Si ratio by ≈10%.

[70] As discussed earlier, the enrichment of Cl and depletion of Si could be attributed to mass dilution by halides, much as we inferred dilution by (potentially sulfate bearing) salts for MER surface dust in the case of Ca and Si. However, cations for potential halides are poorly constrained as the enrichment of K is tentative and Ca is, if anything, depleted (Figure 22). In addition, as discussed initially and elaborated further in section 6.1, sulfates, not halides, are expected to be the most common salts in Martian surficial material.

[71] The biggest challenge of all to a dilution model based on the average crust is the lower Fe/Si ratio in RAVE. While the preferential leaching of Fe under low-pH conditions (section 6.2) is a possible resolution, the similarity of K/Th between RAVE and the rest of Mars (Figure 22) remains difficult to explain with aqueous processes under any pH [Taylor et al., 2006a]. Last, the rest of Mars is chemically, geologically and mineralogically heterogeneous. It seems unlikely that such a heterogeneous surface could be the source for the remarkably coherent chemical signature within RAVE and its neighborhood. Given these reasons we do not consider the rest of Mars to be a viable alternative to MER surface dust either in a mass dilution scenario or in a chemical alteration scenario.

[72] In summary, our comparison of the chemical distribution within RAVE with the various in situ soils of Mars and the rest of Mars hints that MER surface dust (and Laguna class soil) may be a reasonable analog to the bulk material of RAVE. Dilution of such a composition by sulfate salts (<15%) with a Ca cation component would yield the RAVE composition to first order. Such salts may also be hydrated, given the typical enrichment of H2O in RAVE relative to the rest of Mars by ≈10% (section 7). If instead the average Martian crust were the primary source material, two key disparities that are somewhat difficult to resolve would arise: a lower Fe/Si ratio and a similar K/Th ratio (Figure 22). Therefore, we pursue the MER surface dust analog in our conjectures on the origin of RAVE (section 7), but for thoroughness we also consider the in situ rocks observed by the MER mission and SNC meteorites as potential bulk candidates.

5.2. MER Rocks and SNCs

[73] The rock classes that we use, the sols on which they were sampled by the APXS, and the literature upon which the sample selection is based are listed in Table 6. Our ratio scatterplots of rock classes in Figure 23 do not show the consistent differences that were apparent for the case of MER surface dust (Figure 21). For example, a cursory look suggests that Adirondack class with its lower Cl/SiO2 and K2O/SiO2 ratios could be diluted by KCl to give the composition of RAVE. However, FeO/SiO2 of the Adirondack class is at the high end of RAVE, leading to issues similar to that for the rest of Mars in section 5.1. Unlike for soils, the CaO/SiO2 in RAVE also appears elevated relative to most rocks (including the sedimentary Burns formation) to the point of making the region an outlier.

Figure 23.

Scatterplots of oxide mass fraction to SiO2 mass fraction ratios versus the SiO2 mass fraction for rock classes at the MER sites and RAVE (legend, ClFeSi) with the rest of Mars (ROM) and dust included for reference. Legends are identified in Table 10. GS-derived SiO2 mass fractions have been renormalized to H2O-free to enable direct comparison with APXS values. Unlike dust (Figure 21), the rock classes do not show consistent differences with RAVE. Error bars as in Figure 21.

Table 6. Rock Classes Used, Legend Key for Figure 23, and the Sols on Which They Were Sampled by the APXS
Rock ClassLegendAPXS Sampling Sola
Algonquin classAlgonG660,b G675,b G688,b G700b
Adirondack classAdiroG34,b G60,b G86,b G100,b G1341b
Backstay classBackG511b
Barnhill class, Barnhill subclassBaBaG754,c G763,c G764c
Barnhill class, Pesapallo subclassBaPeG1206,c G1209,c G1211,c G1216c
Burns classBurnsM139,d,e M145,d,e M147,d,e M149,d,e M153,d,e M155,d,e M162,d,e M178,d,e M180,d,e M184d,e
Clovis class, Clovis subclassClClG216,f G225,f G231,f G291,f G300,f G304f
Descarte classDescG553c
Elizabeth Mahon class, Elizabeth Mahon subclassElElG1216,c G1226c
Elizabeth Mahon class, Innocent Bystander subclassElInG1251,c G1252c
Independence class, Independence subclassInInG429B,c,g G542,c,g G533c,g
Irvine classIrviG600,b G1055b
Peace classPeacG374,f G377,f G380,f G385Bf
Montalva classMontG1072,c G1079c
Other class, Pot of Gold subclassOtPoG172f
Torquas classTorqG1143c
Watchtower class, Watchtower subclassWaWaG416,c,f G417,c,f G496,c,f G499c,f
Watchtower class, Keel subclassWaKeG646c
Wishstone classWishG335,b G357b

[74] While complex alteration processes in conjunction with mass dilution may be contrived to infer a genetic association of RAVE with some igneous rocks, we do not find any parallels to the simpler possibilities that exist in the case of soils. We may have perhaps anticipated this given the dominance of fine material over rocks in the near surface of RAVE (section 4.3). If a rock analog is sought in spite of these issues, Adirondack class (picrobasalt-basalt in the TAS diagram [McSween et al., 2006, Figure 7]) and the Barnhill-Pesapallo class of putative pyroclasts [Ming et al., 2008] would be reasonable choices.

[75] For the case of meteorites, we use the summary tabulation of chemical compositions of the Martian Meteorite Compendium (http://curator.jsc.nasa.gov/antmet/mmc/index.cfm) by Charles Meyer, limiting ourselves to the samples for which bulk compositions have been reported. The corresponding source files are listed in Table 7. As expected of minor elements in igneous material, meteoritic bulk compositions have far too low Cl/SiO2. In addition, the SNC meteorites (possibly sourced from a depleted mantle) [e.g., Taylor et al., 2006a, sections 5 and 7.2] have very low K2O/SiO2 ratios (Figure 24) for a mix with salt to be plausible, unless mineral assemblages present in alteration veins [Rao et al., 2005, 2008] were considered representative of the bulk.

Figure 24.

Scatterplots of oxide mass fraction to SiO2 mass fraction ratios versus the SiO2 mass fraction for SNC meteorite classes and RAVE (legend, ClFeSi) with the rest of Mars (ROM) and dust included for reference. Legends are identified and data sources listed in Tables 12–13. GS-derived SiO2 mass fractions have been renormalized to H2O-free to enable direct comparison with APXS values. Unlike dust (Figure 21), the SNCs do not show consistent differences with RAVE.

Table 7. SNC Classes and Corresponding Legend in Figure 24a
SNC GroupLegendSampleSource
  • a

    File names and tables of bulk compositions in the Martian Meteorite Compendium by Charles Meyer. For example, the URL for the file “any.pdf” would be (http://curator.jsc.nasa.gov/antmet/mmc/any.pdf) The last column also includes an abbreviated reference to analysts who determined the composition.

Basaltic ShergottiteBSEETA79001_BLodders [1998, Table 4, Cl from Table 3]
  Los AngelesXVLosAngeles03.pdf; p. 5/5; A. E. Rubin (2000) 207mg
  QUE94201que94201.pdf; p. 5/13; P. H. Warren (1999)
  Shergottyshergot.pdf; p. 7/12; J. C. Laul (1986) 139.8mg
  ZagamiZagami.pdf; p. 5/11; K. Lodders (1998)
  Dhofar378dho378.pdf; p. 3/4; Y. Ikeda (2006) 550 mg
Clinopyroxenite (Nakhlite)CGovernadorGovVal.pdf; p. 4/5; F. Burragato (1975)
  LafayetteLafay.pdf; p. 5/9; K. Lodders (1998)
  MIL03346MIL03346e.pdf; p. 6/10; J. A. Barret (2006)
  NakhlaNakhla.pdf; p. 9/14; K. Lodders (1998)
  Y000593XXII_Y000593.pdf; p. 5/5; Y. Oura (2003)
Dunitic shergottiteDChassignyChassig.pdf; p. 7/8; K. Lodders (1998)
  NWA2737Beck et al. [2006, Table 2] (FeO computed and normalized to 100.15%)
Lherzolitic shergottiteLALH77005Lodders [1998, Table 4, Cl from Table 3]
  LEW88516Lodders [1998, Table 4, Cl from Table 3]
  Y793605Lodders [1998, Table 4]
OrthopyroxeniteOALH84001Lodders [1998, Table 4, Cl from Table 3]
Olivine-Orthopyroxene shergottiteOODarAlGani476XIV-Dar%20al%20Gani.pdf; p. 5/6; J. Zipfel (2000) 232 mg
Olivine-phyric shergottiteOPDhofar019XVII-%20Dhofar.pdf; p. 5/5; L. A. Taylor (2002) calculated
  EETA79001_ALodders [1998, Table 4, Cl from Table 3)
  SayhAlUhaymir005Dreibus et al. [2000, Table 1] (K2O and Cl from INAA)
  Y980459XXVII_Y980459.pdf; p. 4/4; N. Shirai (2004)
  RBT04262RBT04261.pdf; p. 3/4; M. Anand (2008)

[76] Even if we disregard Cl as a mobile component and K in SNCs as unrepresentative of the crust [Taylor et al., 2006a], a simple scenario of diluting the substantially higher mass fraction of SiO2 in the majority of SNCs by salts would be difficult. Others such as lherzolites, that have SiO2 content comparable with RAVE, have higher FeO/SiO2 ratios posing the same challenges as for the rest of Mars (section 5.1). Nevertheless, the igneous evolution of possible SNC parent magmas has been considered, and may possibly yield the Fe and Si mass fractions observed in RAVE as discussed in detail by El Maarry et al. [2009].

[77] We feel that of all three candidate material types just discussed, MER soils in general and MER surface dust in particular are the optimal analog(s) for the bulk component of RAVE, given consistent chemical differences, morphology particularly involving fine particle sizes at GS sampling depths (sections 4.2, 4.1, and 4.3), and the relative simplicity of processes that need to be invoked.

6. Origin of RAVE: What Is the Minor Component?

[78] On the basis of scatterplot comparisons of elemental ratios between RAVE and the rest of Mars, Martian in situ soils, in situ rocks observed by the MER mission, and SNC meteorites in the preceding sections, we concluded that the most reasonable chemical analog for the bulk component of RAVE would be MER surface dust. The bulk analog also requires a minor component of mass dilution. As summarized in section 4.7 and reiterated in section 5, the surficial properties are consistent with a salty cementation matrix as the diluting component. The diluted nature of RAVE relative to our analog(s) also requires that any chemical processes involved in the origin of the minor salt component not be isochemical within the GS sampling depths.

6.1. Could It Be Sulfate(s)?

[79] The higher CaO/SiO2 ratio in RAVE relative to MER surface dust and the ≈10% enrichment of median H2O mass fraction in RAVE relative to that of the rest of Mars (Figure 22) hint that a hydrated Ca-bearing salt may be a significant part of the cementing salts. In fact, our first-order estimate in section 5.1 indicated that anhydrite could constitute ≈20% of the unknown salt. Nevertheless, given the possibility of systematic differences between the GS and APXS data [cf. Karunatillake et al., 2007] as well as the lower availability of Ca relative to Mg discussed below, it is conceivable that the salt component may be dominated by Mg instead of Ca. Unfortunately, we are unable to evaluate this quantitatively, since the distribution of Mg cannot be estimated with the GS.

[80] Comparing RAVE surface textures to other Martian regions inferred to contain sulfates provides additional evidence that Mg sulfates may be the dominant variety in RAVE. Figure 25 shows a locale in Aram Chaos determined by orbital near-infrared spectroscopy [Lichtenberg et al., 2009] to contain hydrated sulfate(s). The surface texture resembles that of surfaces in RAVE dominated by reticulate bed forms, and also appears similar to White Rock which was discussed earlier. Spectral features of most spectra at this location are consistent with a monohydrated sulfate [Lichtenberg et al., 2009] for which the geologically most plausible candidate is MgSO4.H2O (kieserite) because, according to recent experiments [Vaniman et al., 2009], CaSO4.H2O is likely to be unstable at the low Martian surface temperatures relative to CaSO4.2H2O (gypsum) and CaSO4.1/2H2O (bassanite) [Vaniman and Chipera, 2006]. We have made an initial survey of other equatorial sulfate exposures identified from orbital near-infrared spectra, and similar reticulate surfaces are seen in many of these regions, almost exclusively in monohydrated sulfate-bearing materials. Irrespective of the cation types, sulfate is likely the primary anion of the salt mixture due to two reasons. First, sulfates have been detected remotely on the surface as summarized by Chevrier and Mathé [2007, section 2.4 paragraph 2] and Gendrin et al. [2005]. Such identifications include gypsum-bearing outcrop surfaces and one dune field (Olympia Undae), the latter of substantial spatial extent [Fishbaugh et al., 2007, paragraph 6]. All surface missions have detected sulfates in situ: Viking, Pathfinder, and MER [Chevrier and Mathé, 2007, section 2.4, paragraph 2]. More specifically, Meridiani is a site of sulfate-rich outcrops [Clark et al., 2005, Figure 12]. Mineral models for the Meridiani outcrops suggest ≈7% CaSO4 [e.g., Clark et al., 2005, Figure 12], with similar/greater enrichments in Boroughs soil within ≈10 cm depth [Haskin et al., 2005], Clovis outcrop, Peace outcrop, and Paso Robles soil [Ming et al., 2006, Table 5]. S enrichment can also be associated with surface dust [e.g., Knoll et al., 2008, paragraph 9].

Figure 25.

Example of an Aram Chaos locale where reticulate bed forms and NIR spectral features of monohydrated sulfate(s) overlap spatially. Implications of this association are discussed in section 6.1. Solid black scale circle is 20 m across. Excerpt from HiRISE image with ID PSP_010025_0835.

[81] Second, the identification of sulfates (in significant amounts) across widely separated localities is consistent with the higher concentration of S in the Martian primitive mantle relative to Earth [Dreibus and Wanke, 1985]. Mass-independent depletion of 33S in SNC meteorites also suggests sustained S cycling between the atmosphere and the crust [Farquhar et al., 2000] making sulfate anions globally available for surficial processes over geologic time even in the absence of sustained volcanic exhalations. Furthermore, Halevy et al. [2007] estimate (volcanic - primarily Tharsis) S outgassing to be roughly twice that of Earth and suggest the control of aqueous conditions on early Mars by S-driven chemical pathways leading to the widespread formation of sulfites, which would cause a surficial prevalence of sulfates under the geologically more recent oxidizing atmosphere [cf. Gaillard and Scaillet, 2009]. Impact processes may have also contributed to geologically recent recycling of sulfates [McLennan and Grotzinger, 2009].

6.2. How Would Sulfates Form?

[82] Given the collective reasons to favor sulfates as cementation agents and the key minor component within RAVE, an evaluation of chemical processes that form sulfates locally may help us identify reasonable conjectures for the genetic processes of RAVE. As summarized by [Wang et al., 2008, paragraph 82], generally only a minor proportion of Ca would be present in sulfates produced by low-pH aqueous alteration due to the slower dissolution rate of plagioclases relative to olivines [Hurowitz and McLennan, 2007; Tosca et al., 2004]. Additionally, Ca sulfates tend to precipitate first making them relatively less mobile in solution [Tosca et al., 2005].

[83] Experimental alteration of Hawaiian (plagioclase feldspar rich) basaltic tephra and (olivine rich) sands by S-rich acidic vapors under hydrothermal conditions (145°; simulating an acid fog scenario with very low water:rock ratios) causes sulfates, primarily Mg and Ca sulfates, to precipitate and forms amorphous silica [Golden et al., 2005, section 5.1, paragraph 46]. In contrast, high water:rock ratios under high-temperature acid fog conditions would cause significant leaching of cations from the host material [Golden et al., 2005, paragraph 30]. Of the Fe3+ sulfates that form under these conditions, jarosite is usually the only one to remain in the residue [Golden et al., 2005, Table 2].

[84] The results of hydrothermal high water:rock acid fog experiments are consistent with low-temperature experiments that Tosca et al. [2004] conducted with synthetic Martian basalts, as well as theoretical analyses by Tréguier et al. [2008]. In particular, Ca, Fe, and Mg cations are released into solution [Tosca et al., 2004] with subsequent precipitation of sulfates at low pH (1–3) and modeled solubilities increasing in the order Ca < Fe < Mg [Tosca et al., 2005, pp. 124–130, Figures 3–8]. As expected, the sole exception is jarosite which precipitates first.

[85] McAdam et al. [2008, p. 93] consider the possibility that the snow, ice, and dust deposition under higher obliquities that we discussed in section 4.6 may have effectively scavenged acidic aerosols from volcanic exhalations. If so, the resulting ground ice could have facilitated low-pH alteration of fine material and rocks by acidic thin brine films more effectively than by acid fog alone [McAdam et al., 2008, p. 93].

[86] Even in the absence of acid fog related aerosols, Chevrier and Mathé [2007, section 2.4] hypothesize that chemical alteration mediated by thin films of water could yield sulfates as long as sulfide-bearing minerals (e.g., pyrrhotite) and suitable cation (e.g., Ca) bearing minerals (e.g., diopside) are available. However, aqueous conditions with high water:rock ratios [Chevrier and Mathé, 2007, Figure 8] are probably necessary to generate sulfate deposits on the scale of the Olympia Undae dune formation regardless of whether they formed by precipitation from percolating groundwaters [Fishbaugh et al., 2007, paragraphs 32 and 33] or some other mechanism. If deposited in place, the required Ca could still be locally derived from high-Ca pyroxenes or calcic plagioclase feldspars [Fishbaugh et al., 2007, section 3.2.2], of which the former is also more likely to dissolve under low-pH conditions [McAdam et al., 2008, p. 93].

[87] Perhaps more feasible at the large spatial scale of RAVE would be mediation of low-pH chemical processes in fine material by a sustained S cycle between the atmosphere and the near surface of Mars (section 6.1). For example, as Tosca et al. [2008] discuss in the Meridiani context, oxidation and dehydration processes involving hydrous Fe sulfates may help to sustain low-pH brine films in Martian dust/soil. The acid fog driven processes that we discussed at the beginning of this section could then occur even without volcanic exhalations.

7. Synthesizing the Origin of RAVE

[88] First-order estimates of volume and mass may help develop a sense of scale for the salts and fine material involved in the formation of RAVE. RAVE has a surface area of roughly 2.E6 km2, approximating that of the Medusae Fossae formation [Bradley et al., 2002, section 7]. Assuming a 2 m depth (section 4.3), 1.2E3 kgm−3 bulk density (section 4.1), and 12.5% mass fraction of salt (section 5.1) yields a total 6.E14 kg mass of salts, roughly equivalent to a 2.E2 km3 volume if it were a mixture of gypsum (30% mass fraction) and kieserite (70%). The total mass of salt would exceed that of gypsum at Olympia Undae by a factor of 10, the volume by a factor of 8, and the area by a factor of 100 according to the Olympia Undae mass, volume, and area estimates by Fishbaugh et al. [2007, paragraphs 50 and 51].

[89] While the total volume of RAVE is smaller than what Bradley et al. [2002, section 7] estimate for the Medusae Fossae formation by a factor of 1E3, RAVE is still enormous by terrestrial standards at roughly 20% of the land area of the USA. Terrestrial analogs even at the scale of Olympia Undae are rare, with the White Sands dune formation only 7E2 km2 in extent [Fishbaugh et al., 2007, paragraph 4]. Sulfate deposits associated with volcanism, such as the Julcani (Peru) and Creede (CO) formations [e.g., Rye, 2005] are even less than 1E2 km2 in extent. These bulk comparisons suggest that the chemical processes that helped to form the cementation salts in RAVE are likely to have been regional, rather than local, in scale.

7.1. General Inferences

[90] While our chemical observations are currently too limited to fully constrain potential formation scenarios for RAVE, the preceding discussions may be used to guide a few conjectures. These generally share the same inference on the production of bed forms and the veneer of dust: surficial salts in association with fine material were mobilized by thin films of water within RAVE during typical 42° and higher obliquities that last occurred ≈5 Ma ago (section 4.6). The evaporation of resulting brine films aggregated fine particles to sufficiently large sizes to saltate (section 4.3), perhaps under denser atmospheres at higher obliquities (section 4.6). Regional winds, driven in part by katabatic winds at nearby volcanoes, formed the complex bed forms from the aggregated particles (section 4.3).

[91] Cementation by salts over time indurated most of the bed forms, while others either may have been inactivated by ongoing air fall dust or are still evolving albeit at slower rates than can be observed over the lifetime of current missions (section 4.3). The present veneer of dust (sections 4.2 and 4.3) has accumulated mostly under recent low-obliquity hyperarid conditions (section 4.6) that do not mobilize salts as effectively.

[92] Most of the conjectures also rely on the chemically and physically reasonable assumption that the bulk of RAVE as seen by the GS is constituted of indurated fine material (sections 4.3 and 4.7) with a composition similar to MER surface dust (section 5.1). The minor (mass fraction typically <15%) salt component that helped to form the bed forms via aggregation and indurated them subsequently is taken to be mostly sulfates with some amount of Ca (section 6.1). The scenarios discussed below differ from one another in two key areas: (1) The primary source of the sulfates and (2) the processes that produced them.

7.2. Less Viable Scenarios

[93] We consider three distinct scenarios: First, sulfate-enriched atmospheric dust; second, regional production of sulfates by large-scale volcanism; and third, localized production of sulfates by volcanism with subsequent transport.

[94] The first and simplest conjecture is that atmospheric dust similar in composition to MER surface dust is deposited across much of the region through obliquity cycles (section 4.6), and subsequently diluted by sulfate salts formed in other locations and transported to RAVE. Known plausible source localities for the sulfates that mix with and dilute the atmospheric dust are mostly low-elevation regions with the nearest such deposit no closer than the interior of Valles Marineris more than a thousand kilometers away [e.g., Gendrin et al., 2005]. In spite of the simplicity of this model, transport of the sulfates from distant low-elevation sources would be a major challenge that is difficult to overcome.

[95] The second scenario emphasizes the location of RAVE surrounded by some of the largest volcanic edifices on the planet (section 4.4). We may envision a few basaltic plinian eruptions or numerous fire fountain and strombolian eruptions during the late Amazonian depositing massive beds of unconsolidated material throughout RAVE including reticulite, lapilli, and scoriaceous ash (section 4.4). Salts produced by acid fog alteration of such deposits on the ground mixed with air fall dust could generate the observed chemical signature of the region. Unfortunately, the feasability of this model is undermined by the tenuous evidence for geologically recent explosive volcanism of such magnitude (section 4.4).

[96] Alternatively, smaller magnitude volcanism, comparable to Hawaiian type fire fountains or strombolian eruptions, could have deposited large beds of scoriaceous ash and lapilli on the flanks of the Tharsis volcanoes. Local acid fog under hydrothermal conditions (section 6.2) could then yield sulfate beds. These (particularly easily friable sulfates such as gypsum) [e.g., Fishbaugh et al., 2007, paragraph 4] could have been preferentially deflated and deposited in the intervening topographic lows that include our region, perhaps throughout the last 0.1 Ma (section 4.6). Meanwhile, ongoing deposition of atmospheric dust, chemically similar to MER surface dust, would form the bulk component. However, bulk considerations, preferential deflation, and the tenuous detection [Cooper and Mustard, 2002] of residual sulfate deposits on neighboring volcanoes argue against this possibility.

7.3. More Likely Scenario: Air Fall Dust, Ground Ice, and Widespread Acid Fog

[97] The scenario we consider to be more likely invokes the chemical alteration of air fall dust by the activity of thin films of brine under higher obliquities. Such brines would have been sustained mostly by regional-scale ground H2O ice more than 5 Ma ago (section 4.6), perhaps with a minor contribution from deeply buried relict H2O glaciers that formed as recently as 65 Ma ago on the Arsia Mons flanks (section 4.5). The low-to-moderate pH alteration mediated by this sustained source of groundwater formed regionally widespread sulfate salts that migrated to the surface by evaporative wicking-up effects [e.g., Gellert et al., 2006; Haskin et al., 2005]. Alternatively, sulfates may have formed via the chemical alteration of air fall dust by regional- scale acid fog [McAdam et al., 2008, p. 93]. The process may have been accelerated by the scavenging of acidic aerosols by H2O snow at higher obliquities (section 6.2). In either case, salts that had migrated to the surface mixed with continuing air fall dust that was chemically similar to MER surface dust, eventually accumulating the meter-scale beds that we now see. An important constraint in this model is that the source material of cations, such as Ca, has to be buried deeper than the GS sampling depths to yield the apparent Ca enrichment within RAVE. We infer that the chemically weathered material beneath the surficial layers would be depleted in Ca relative to MER surface dust.

[98] The current conjecture, while invoking one additional step (the formation of salts driven by regional-scale groundwater, acid fog, dehydration/oxidation of salts, or a combination thereof) relative to the first model (section 7.2) avoids compositional pitfalls. Much like the simplest, the regional scale of the current conjecture also accounts for the necessary bulk. However, a drawback is that chemical alteration mediated by thin films of brine and/or acid fog at low temperature may require longer time scales than afforded by the <65 Ma geologic age (section 4.5) of RAVE. In addition, it is poorly known whether regional-scale acid fogs ever formed on Mars and whether regional-scale salt precipitation could occur via upward migration [e.g., Amundson et al., 2008, pp. 15–16] of brine films. Apart from such concerns, we find our conjecture to be viable.

8. Conclusions

[99] The Chemically striking regions (CSRs) that we have delineated on Mars may provide significant insight into surficial processes on the planet, and perhaps even into deep seated igneous processes in areas such as Elysium and low-albedo surfaces in the North. They also show that the near surface of contiguous geologic units, such as the Medusae Fossae formation, may nevertheless be chemically heterogeneous. The CSRs represent the synthesis of all chemical maps that have been finalized with the GS data, and supplement region delineations with Principal Component-based cluster analyses of the companion papers. Even more important, they demonstrate that the intensity and areal extent of chemical differentiation in the Martian surface is sufficient to guide future explorations and comparisons with other data sets.

[100] We have demonstrated the synergy of the CSRs and other data sets with a case study involving the one region of the planet, RAVE, that is enriched in Cl and depleted in both Fe and Si relative to the average Martian crust. The strong spatial overlap of RAVE with an area of Stealth in radar reflectivity as observed from Earth validates the complementary nature of Martian remote sensing observations. Meanwhile, the chemical constraints that we were able to establish with MER data sets demonstrate the utility of combining remote sensing and in situ data. Thermally derived attributes, observation of morphology with HiRISE, and climate models enabled us to further constrain potential formation scenarios of RAVE.

[101] The RAVE surface is dominated by indurated eolian bed forms. In addition to barchanoid, ripple, and linear forms that are common on Earth, unusual reticulate and lenticular forms are also present. Chemical comparisons indicate that, contrary to initial expectations from the apparent Cl enrichment relative to the average crust, the bulk material is similar to MER surface dust diluted by Ca-bearing salts. Such salts may also be the key indurating agent. Additional chemical and/or mineralogic information may be needed to fully constrain the processes that formed RAVE. Nevertheless, we favor an origin involving the interactions of a MER surface dust analog, ground ice, and acid fog, which may help guide future investigations. The refinement of GS-derived Al and S maps, as well as upcoming analyses of chemical layering with the NS data [Diez et al., 2008] may be particularly useful in this regard.

Acknowledgments

[102] We thank the Mars Odyssey Mission for both collegial and financial support. Michael J. Finch, Daniel M. Janes, Kristopher E. Kerry, and Remo Williams, in particular, ensured the availability of robust GS data sets. Linda Martel and Chris Okubo contributed to our effort with targeting and analysis of HiRISE images. John Keller and two anonymous reviewers enhanced the clarity, brevity, and accuracy of the paper. Jeevak Parpia, Veit Elser, Robert W. Kay, Deanne Rogers, Tim Glotch, Leslie Looney, and Ed Sutton guided us with incisive queries. We also thank Nicole Button, Ryan Yamada, Soshanna Cole, Ryan Anderson, Briony Horgan, Melissa Rice, William Woerner, Amy Lien, and Nicholas Hakobian for numerous (and fruitful) discussions.