4.2. Proposed Sequence of Events
 In summary, we propose the following generalized geologic history of the region, which incorporates TES- and THEMIS-detected spectral units (Figure 24). Following formation of the earliest exposed crust, the surface was degraded via impact, aeolian, and/or fluvial processes,generating regolith from bedrock (unit 1) which may be indurated in some places. These processes may have resulted in mechanical and/or chemical alteration of the original surface composition; however, the original lithology (for example, the olivine content relative to younger units) is not well constrained.
Figure 24. Cartoon depiction of the proposed sequence of events for units 1–3. The proposed sequence is as follows: Impacts (b) into ancient highlands surface (a) formed a cratered surface with intercrater plains. Portions of the intercrater plains and some crater floors are later resurfaced by volcanic infilling (c). These materials are enriched in olivine relative to the older units. It is unclear how much of the compositional difference between the two units is due to alteration of the older unit and how much is due to a primary difference in crust-forming magma compositions. During and after the emplacement of units 2 and 3, impacts continue to form crater ejecta (d) which partially bury the margins of older units. Some of these later impacts exposed unit 2 materials in the crater walls, which later eroded to partially fill some crater floors (e). Impacts into unit 2 excavate unit 1-like compositions from beneath (f); smaller impacts into unit 2 only expose unit 2 materials in their crater ejecta (g). Portions of unit 2 either are altered to become less mafic and similar to unit 1 or are covered with sediments derived with unit 1 (h). Unit 4 is a surficial unit only and therefore is not included in this diagram. The origin of unit 5 is poorly constrained and thus is also not included (see text).
Download figure to PowerPoint
 Later, widespread volcanism resulted in partial filling of intercrater plains (unit 2) and possibly crater floors (unit 3). Some impacts occurring during or after this time may have exposed unit 2-like materials in crater walls, which were then distributed across the crater floors. All three units (1, 2 and 3) continued to be reworked and resurfaced, resulting in partial burial of units 2 and 3 with locally derived sediment as well as older (and perhaps regionally derived) sediment from unit 1. Chloride deposits formed in association with unit 2-like surfaces after substantial modification (to reduce the strength of the TES 507 cm−1 index) and perhaps burial (Figure 22). The fraction of sediment most conducive to saltation then formed dunes and other bed forms (unit 4).
 The formation mechanism(s) of unit 5 is/are so ill-constrained that it is difficult to incorporate it into the sequence with any reasonable certainty. It may represent isolated areas that underwent additional chemical alteration to produce slightly more silicic/felsic surfaces. In this case, the enhanced alteration may have been enabled by underlying differences in primary lithology or grain size; or was perhaps assisted by local ground waters or hydrothermal systems and later exposed. Alternatively, it may represent a less altered version of unit 1 that has been exposed by erosion (section 4.1.5).
4.3. Relationship to Similar Units in Other Noachian Highlands Regions
 Units 1, 2, and 3 bear strong resemblance to units observed in Mare Serpentis, a low-albedo heavily cratered region northwest of Hellas Basin. Rogers et al.  reported high thermal inertia, resistant mafic units interspersed throughout less mafic, lower thermal inertia degraded plains. They suggested that the mafic units likely represented volcanically emplaced materials that postdate older, less mafic plains units. As described here for unit 1, it is not clear if the primary lithology of the older units was originally less mafic or if the present-day olivine-deficient composition was achieved through alteration processes [Rogers et al., 2009]. Similar to this study region, the resistant mafic materials in Mare Serpentis are also found in crater floors. The major difference between Mare Serpentis mafic units and units 2 and 3 from this study are the overall thermal inertia ranges; mafic units in Mare Serpentis reach thermal inertias as high as 500–1200 J m−2 K−1 s−1/2, compared to a maximum of ∼420 J m−2 K−1 s−1/2 for this study region. The difference could simply be related to efficiency of sediment removal between the two regions; local wind patterns could facilitate increased areal coverage of sediment-free surfaces in Mare Serpentis.
 Elsewhere in the highlands, high thermal inertia, flat-floored craters have been examined in detail [Mest and Crown, 2005; McDowell and Hamilton, 2007; Edwards et al., 2010]. In contrast with the unit 3 surfaces presented here, craters in Margaritifer Terra do not typically appear spectrally unique from surrounding materials in THEMIS data [McDowell and Hamilton, 2007]. However, preliminary examination of some of those craters using the 507 cm−1 index indicates that some may at least have spectral distinctions at long wavelengths in TES data. The geomorphology of Millochau Crater was examined in detail by Mest and Crown . The crater floor partially consists of similar material to unit 3 (mapped as “rugged material” by Mest and Crown); they report a gentle sloping from the edge of the rugged material downward toward the crater center, as observed here for some occurrences of unit 3. Mest and Crown  suggest that the rugged material may consist in part of mass-wasting material, though they note that the absence of transverse ridges and other morphological features associated with gravity-driven flow are puzzling. They also suggest that the topographic characteristics are consistent with postmodification of near-horizontal strata. The characteristics they describe are similar to our observations of unit 3. Preliminary work by Edwards et al.  suggests that the relative increase of mafic minerals found in flat-floored high thermal inertia Intracrater deposits is a widespread trend throughout the highlands. They suggest that the crater filling process may have been dominated by impact-driven volcanism, such as that proposed for some lunar craters [Schultz, 1976].
 Bandfield  described high-silica deposits in Hellas Basin, consisting of ∼80% high-silica phases such as amorphous silica or zeolite; the identification was made using both THEMIS and TES data. Additionally, bright soils uncovered by the Mars Exploration Rover Spirit in Gusev Crater were found to consist of >80% SiO2 [Squyres et al., 2008]. Areas described as unit 5 differ from the Hellas deposits in both THEMIS and TES data. Unit 5 surfaces do not exhibit the low surface emissivity at ∼8.5 μm (THEMIS band 4) nor the relatively featureless emissivity at wavelengths >∼25 μm (<400 cm−1) in TES data. Thus, although unit 5 is relatively silicic/felsic compared to unit 2, it is not similar to Hellas or Gusev soils, which are overwhelmingly dominated by silica.
4.4. Outstanding Questions Related to Major THEMIS Signatures in the Highlands
 Results from this work and other studies highlight a few outstanding questions regarding highland surface compositions, and in particular, regarding spectral unit signatures that are commonly observed in THEMIS data. The first question relates to the true composition of the heavily degraded unit 1. There are clearly two dominant spectral units in the intercrater plains, represented here as units 1 and 2 and by Rogers et al.  as less mafic, low thermal inertia plains and mafic rocky surfaces. These differ from each other in relative age, composition, and thermophysical properties. It is not clear if the older, less mafic unit (unit 1 in this work) represents a truly different lithology from the younger units, or if the units were actually similar in composition, but the older unit has experienced more alteration. If unit 5 represents a less altered version of unit 1, then this would suggest that the primary lithology of unit 1 truly was and is olivine-deficient. Rogers  reported that crater ejecta in this region commonly exhibit increased abundances of plagioclase and/or high-silica phases and decreased abundances of pyroxene, relative to the target materials. Rogers suggests, among other scenarios, that the craters are either exposing (1) more pristine or (2) more altered materials than the surrounding target surface. The increased abundance of high-silica phases in the ejecta materials, however, more strongly supports the latter scenario. The ejecta materials found within unit 1 are not deficient or enriched in olivine relative to unit 1, but are deficient relative to unit 2. If the craters are exposing more altered material, the implication is that the olivine abundance associated with unit 1 was likely higher at the time of crystallization, but has been decreased because of alteration. (Olivine is more susceptible to dissolution than pyroxene; thus if pyroxene is decreased because of alteration, then olivine would also have been decreased). However, the subsurface alteration scenario favored by Rogers  depends heavily on the observation of slightly higher abundances of high-silica phases in ejecta materials, which could potentially represent glass or shocked materials instead of aqueous alteration phases. More importantly, the slight difference in high-silica abundance was not considered to be statistically separable for many of the craters examined [Rogers, 2011]. Thus, the possible clues from the crater ejecta are equivocal and the question remains: at the time of crystallization, were these older materials deficient in olivine relative to younger units?
 Answering this question has important implications for understanding the role of aqueous alteration in soil formation as well as the expected global volume of Mg-bearing alteration products. For example, Bandfield et al.  recognize a global trend whereby rocky surfaces typically contain higher abundances of olivine than lower thermal inertia surfaces. They pose the hypothesis, drawing from geochemical observations in laboratory [e.g., Tosca et al., 2004] and in situ Mars surface studies [Hurowitz et al., 2006], that this trend can be explained by olivine dissolution by aqueous alteration during the formation of Martian soils from bedrock. In areas where olivine-deficient sediment can be clearly traced to olivine-rich bedrock (such as in Argyre Basin [Bandfield and Rogers, 2008]), the hypothesis is firmly supported. However, for much of the Martian surface, the source of soils is not well constrained. It is possible that the lower olivine abundance in most Martian soils (relative to younger bedrock) is due to derivation from older, olivine-deficient bedrock that has now been obscured by olivine-deficient regolith derived from that older bedrock. The olivine-enriched bedrock sources (units 2 and 3 in this work) cover a small area relative to olivine-deficient surfaces, thus one would expect olivine-deficient soils to be volumetrically dominant over olivine-enriched soils in any case. High abundances (>30%) of Mg sulfate are observed in strata in isolated areas on Mars [e.g., McLennan et al., 2005; Glotch et al., 2006] and could be ubiquitously present at lower abundances (5–10%, below the TES detection limit) in Martian soils [e.g., Bandfield, 2002; Rogers and Aharonson, 2008]. Additionally, Fe- and Mg-bearing clay minerals are some of the most commonly detected phyllosilicates to date in CRISM and OMEGA data [e.g., Murchie et al., 2009; Milliken and Bish, 2010]. Dissolution of olivine is thought to be a primary source for the Mg cations needed to form these minerals [e.g., Tosca et al., 2004; Hurowitz et al., 2005]. Thus, determining the expected volume of Mg-bearing secondary minerals in Martian soils should depend on the starting compositions of ancient crustal materials. In summary, results from this and previous work demonstrate that the original olivine abundance of degraded intercrater plains surfaces is not well constrained. This question might potentially be addressed through detailed, systematic studies of exposed subsurface compositions in impact materials.
 A second question relates to the origin of less mafic material commonly observed overlying resistant olivine-enriched intercrater plains (e.g., Figure 1). In some cases, a strong thermal inertia distinction from unit 2 is not observed (Figure 1), whereas in other cases, the thermal inertia is decreased relative to unit 2 exposures, indicating variable thickness. We interpret this material as either sediment derived from unit 1, or material directly derived from alteration of the underlying unit 2. Distinguishing between these interpretations has important implications for the role and timing of olivine alteration in Martian soils. For example, if the deposits are material directly altered from unit 2, then it implies that conditions conducive to olivine alteration continued at least until the early Hesperian (based on age estimates for Hesperian “smooth units,” designated Hpl3 by Greeley and Guest , which sometimes coincide with unit 2 occurrences). The alteration scenario would also raise the question of what property (e.g., grain size, porosity, lithology) of unit 2 is changing to enhance alteration where the overlying material is found.
 A third question relates to understanding the dominant origin of the high-thermal inertia, olivine-enriched crater floor materials (unit 3 in this work). As discussed by Edwards et al. [2009, 2010], high thermal inertia, flat-floored craters are widespread in the Martian highlands. In some cases, they are not compositionally distinct from the surrounding degraded plains materials [e.g., McDowell and Hamilton, 2007]. In other cases, they are distinctly enriched in olivine and are spectrally similar to olivine enriched intercrater plains [Rogers et al., 2009; this work]. It is not clear if these differences are due to true lithologic differences between the deposits, or to differences in alteration/resurfacing between them. Furthermore, what is the origin of the crater filling material? As discussed by McDowell and Hamilton  and in section 3.3, these crater floors are too shallow to represent impact melt/breccia material; this leaves volcanic and/or sedimentary interpretations. In the craters studied here, the compositional similarity to unit 2 argues against sedimentary infill from outside the crater (section 4.1.3); however, it is possible that impacts into olivine-enriched materials (like unit 2) could have exposed these materials in crater walls, which were then subsequently eroded and distributed across the floor of the crater. A volcanic origin also cannot be ruled out (section 4.1.3). Understanding the origin of these widespread materials has important implications for the volcanic and sedimentary history of Mars; namely, constraining the degree of and timing of volcanic resurfacing and/or sediment generation from unit 2-type surfaces.
 A fourth question relates to the dominant origin of unit 5 surfaces, which are spectrally distinct from the dominant unit in the region, unit 1, and are likely more felsic/silicic in composition than unit 1. The origin of these materials, and their relationship to other units, are poorly constrained. As discussed in section 4.1.5, they may represent altered versions of units 1 or 2, or, could represent less altered versions of unit 1. Distinguishing between these scenarios has important implications for the nature of the true composition of ancient crustal materials, and relates back to the first question above: were older unit 1 materials deficient in olivine at the time of crystallization? If unit 5 is representative of the true composition of unit 1, then it would suggest a true deficiency in olivine relative to younger units. If, on the other hand, unit 5 represents an altered version of unit 1, it may imply local groundwater or hydrothermal systems that led to enhanced alteration, then later exposure by erosion.