4.1. Models for Emplacement of the Mantling Unit
 As outlined in the background discussion, the hypotheses presented for the origin of the fill material included aeolian deposition (e.g., dust or loess) [Grant and Schultz, 1990; Moore, 1990], pyroclastic emplacement [Moore, 1990; see also Hynek et al., 2003], volcanic intrusion [Wilhelms and Baldwin, 1988] or deep-water sedimentation (such as from lakes or an ocean [e.g., Edgett and Parker, 1997]). The new observations in this study help constrain the processes responsible for the mantling unit's emplacement.
 The hypothesis that the layered mantle formed from sediment deposited via oceanic sedimentation is motivated by the fine-grained composition of the mantling unit, its smoothness, layering and areal extent [e.g., Edgett and Parker, 1997; see also Moore, 1990]. However, many other characteristics of the mantling unit seem inconsistent with this model. Exposures of the unit are found over a wide range of elevations (>∼3000 m), and the highest elevation mantling deposits are at elevations >1500 m above the datum (Figure 2). Many well-preserved valley networks are found at elevations well below the maximum elevation of the mantling unit in the study area (Figure 2), as well as across Mars [Carr, 2002]. If a deep ocean deposited the mantle, submarine valley networks would have to survive largely unmodified. Furthermore, if an ocean covered this region to the 1500 m contour, it would cover 70% of the surface of Mars, equivalent to a ∼2.4 km average global layer. The water volume of such an ocean would be comparable to terrestrial surface inventories [Carr and Head, 2003]. This calculation assumes that the present topography is essentially the same as that at the end of the Noachian; this appears to be a reasonable assumption since the inferred geometry of Noachian valley networks is consistent with current topography [Phillips et al., 2001]. Evidence for such a sizable ocean is controversial [Carr and Head, 2003], especially at a time period as late as the Noachian/Hesperian boundary. Finally, within this study area, the thickness variation and elevations range of the mantling deposit are inconsistent with a model of oceanic sedimentation (Figure 13).
 An alternative mechanism for depositing the mantling unit in deep water is lacustrine sedimentation, which commonly produces layered deposits in topographic depressions on Earth. However, the detailed relationships of valleys with craters that contain the mantle cast doubt on whether this is a reasonable possibility. The unconformable emplacement of the mantling unit on preexisting terrain, and the lack of valley incision into the unit, suggest that it was emplaced after widespread fluvial processes operated in this region. Most outcrops of the mantling deposit lack direct connection to valley network-incised terrain (e.g., Figure 3a). The volume of the mantling unit is at least 103 km3, and if this volume of sediment had a local source, it would require erosion of tens to hundreds of meters from the neighboring highlands at the time of the emplacement of the mantling unit. Finally, craters that are isolated and distant from highlands watersheds have fill deposits as widespread and thick as those in more favorable locations for fluvial sedimentation. Taken together, these arguments suggest that sedimentation from fluvial action in the highlands cannot explain most of the observed characteristics of the mantling deposit.
 Explanations that invoke local volcanic processes, either surface flows or exposed intrusions (as proposed by Wilhelms and Baldwin ), are motivated by the strength of the mantling unit and observation of potential dikes which are exhumed in a few locales in the study area (see section 3.1; Figures 5 and 11). However, widespread evidence suggests that the mantling unit was emplaced unconformably on its substrate (e.g., Figure 3b), which is inconsistent with formation of the mantling deposit via exhumation of intrusive structures. Both intrusive and extrusive volcanic flows are equally problematic for explaining the emplacement of the mantling unit, its erosional style, and the evidence that it is composed of indurated fine-grained particles.
 Moore  argued that the mantling unit was deposited from the atmosphere because of its fine-grained nature, horizontal layering, areal extent, and apparent unconformable emplacement upon preexisting topography. A wide range of new evidence in our study supports the atmospheric dust or ash deposition mechanism as originally proposed by Moore , including (1) meter-scale horizontal layering of the mantling unit, apparent in MOC images (Figure 9); (2) a broad trend in the thickness of the deposit as a function of location (Figure 13); and (3) thickening of the deposit in the deepest portion of preexisting lows.
 The new observation of meter-scale, horizontal or subhorizontal layering, coupled with the large areal extent of the mantling unit, strongly suggests deposition of material from suspension. Atmospheric transport of dust or ash has the potential for producing thick, regional layered sequences of fine-grained material [e.g., Moore, 1990]. The repetitive layering observed on mantle outcrops suggests that this deposition occurred in a sequence of discrete events. If the mantling material was deposited as dust, this could be due to periodicity driven by climate; alternatively, the fine-scale layering could result from ash transport and deposition from a series of distant eruptive events.
 Both the broad trends in the thickness of the mantling deposit and its significant local variability in thickness help constrain its origin. The existence of a broad trend in thickness implies that the deposition of the mantling unit was a regional-scale phenomenon, not dependent on local sources of material for its deposition. The thickening of the deposit in the deepest portions of preexisting topography appears to have resulted from the ability of these areas to act as a trap during sedimentation or as the material was reworked after its deposition. The preservation of the thickest portions of the mantle in craters may have been aided or enhanced by induration in these locations (see below), as well as by the protection from erosion afforded by crater walls.
 Given that atmospheric sedimentation from suspension seems the most plausible mechanism for emplacing the mantle deposit, either ash or dust could be the dominant mantling material. Climate modeling suggests that Arabia Terra is a region that experiences minimal dust lifting (weak winds) in a wide range of scenarios, including under different obliquity conditions [Haberle et al., 2003; Newman et al., 2005]. This is consistent with a hypothesis that Arabia Terra can act as a dust sink. Indeed, on the basis of the dustiness of Arabia Terra at thermal skin-depth scales, it appears to be a locus of dust deposition today [e.g., Christensen, 1986].
 The mantling deposit appears to be at least locally resistant to erosion, and the deposit must be indurated [Moore, 1990] sufficiently to support 100–300 m high scarps (Figures 3a and 12). The inversion of terrain also requires that some of the mantling unit was more resistant than surrounding terrain. What agents could have caused this induration? One plausible mechanism is incorporation or alteration of the mantling unit by a volatile (most likely H2O). This could have happened in a primary manner, with ice nucleation on dust grains, or as a secondary process, as groundwater passed through the mantling unit. The coupled deposition of volatiles and dust seems especially plausible in light of the following observations:
 1. Deposition of volatile/dust mixtures are known to occur on the Martian surface at the poles [Thomas et al., 1992]; upon sublimation, disaggregation may result.
 2. Cyclic deposition and removal mediated by the climate and atmosphere is consistent with the observed areal extent and layering of the mantling unit.
 5. The spin-axis orbital parameters of Mars are known to be highly variable in the recent past, and this variability has been shown to extend into the planet's earlier geological history as well [Laskar et al., 2004]. Statistical study suggests that the most probable obliquity value over the past 4 Gyr for Mars is 41.8° (much greater than the present value of ∼25.2°), and that there is a ∼90% chance that obliquities greater than 60° occurred at some point over that last 3 Gyr [Laskar et al., 2004].
 An alternative model proposed for past co-deposition of volatiles and dust in modern low-latitude regions is true polar wander [Schultz and Lutz, 1988]. In this model, locales presently in the low latitudes were at a rotational pole at some point in the past; thus these regions were once focal point for deposition of layered material similar to what is found at the modern pole (the polar layered deposits). As polar wander progressed, these regions were left with paleo-polar deposits [Schultz and Lutz, 1988]. However, both geophysical [Grimm and Solomon, 1986] and geological [Tanaka, 2000] analyses cast doubt on whether polar wander on Mars occurred, at least as late as when the mantling unit and other friable layered materials were deposited. Most potential paleo-polar or friable layered deposits on Mars, including the mantling deposit discussed here (see section 3.4), appear to be Late Noachian or younger [Hynek et al., 2003], after Tharsis was mostly in place [Phillips et al., 2001]. Moreover, recent studies imply that transport of water ice to near-equatorial latitudes can occur in the present polar configuration due to obliquity variations alone [e.g., Mischna et al., 2003]; thus characteristics of deposits thought to require polar wander may have a simpler explanation.
 Incorporation of water ice during deposition or subsequent infiltration of groundwater can lead to substantial alteration of the material in which it is found. One process important for the evolution of the mantling deposit is cementation of dust or ash grains, which may have indurated the unit and protected it from removal. Cementation resulting from water-sediment interactions are known to have altered sedimentary material at the Meridiani landing site [Squyres et al., 2004], though the indurating or cementing agent and environmental conditions for the northeast Arabia Terra deposit are unknown. Another potentially important process linked to volatiles is direct removal of material from the mantling unit to the atmosphere, which might help explain unusual erosional characteristics of the deposit (see next section).
 If the deposit is made of ash rather than dust, a plausible local source of ash exists just east of the study region. Syrtis Major has an Early Hesperian surface age [Hiesinger and Head, 2004] essentially consistent with the mantling unit's apparent emplacement age. The broad structure of Syrtis is similar to other highland volcanic provinces thought to involve abundant pyroclastic volcanism [Greeley and Crown, 1990; Crown and Greeley, 1993]. Wilson and Head  suggest that explosive volcanism is potentially a more important process for all Martian volcanoes (including basaltic shields) than for those on Earth, due to the lower atmospheric pressure and greater eruption speeds on Mars. Deposits from explosive eruptions might be recognized by mantling deposits over tens to hundreds of kilometers of terrain [Wilson and Head, 1994; Hynek et al., 2003]. Moreover, recent work with Mars General Circulation Models (GCMs) suggest that over a wide range of orbital parameters and climate scenarios, east-to-southeast winds from Syrtis Major toward the study area are common, due to the Isidis basin [e.g., Madeleine et al., 2007; see also Fenton and Richardson, 2001]. Wind predominantly from the southeast across Syrtis Major (centered at ∼67°E, ∼9°N, see Figure 1a) would lead to the observed thickness distribution of the mantling deposit, with the thickest portions near 50°E, 20°N and thinning significantly to the south of 15°N (Figure 13). These factors suggest that Syrtis Major is an excellent potential source for the mantling unit.
 Recently, Wilson and Head  have suggested that more water may be exsolved during a Martian pyroclastic eruption than is typical on Earth, which would provide a potential source of volatiles that may become incorporated in the ash deposit. In this model, volatiles incorporated into the mantling unit may be related to the volcanic event itself. In the future, we hope to construct a forward model for an eruption on Syrtis Major, utilizing the new parameterization developed by Wilson and Head  for expected particle size distributions and a Mars GCM to directly trace ash transport following an eruption. Further comparison of this sort of modeling with the geological record will help constrain the viability of such a scenario. Distinguishing between dust or ash deposition for the origin of the mantling unit will require both further modeling and new observations; indeed, a combination of factors may be involved.
4.2. Models for Erosion and Terrain Inversion
 From our observations, the mantling unit appears to have been initially widespread, and subsequently has undergone erosion and inversion of relief. These observations also suggest that the magnitude of removal must have been extensive, since terrain inversion has left isolated mesas, buttes and knobs hundreds of meters high (Figure 3), and the mantling material is deeply fractured and pitted (Figures 5 and 6). Crater counts suggest that this may have occurred in a geologically short period of time, perhaps less than a hundred million years (Figure 14). Combining this magnitude of erosion and time estimate suggests an average erosion rate of ∼1 m/Myr.
 On the basis of analysis of Viking data [e.g., Greeley and Guest, 1987; Moore, 1990], the preferred hypothesis for erosion of the mantling unit in northeast Arabia Terra was aeolian erosion. No evidence exists that the mantling unit was modified by fluvial erosion or transport, and no reasonable pathways of sediment out of the low-lying basins of the study area exist.
 A difficulty for the aeolian removal model is that recent GCM analyses predict very low deflation potentials in the study area under present conditions (deflation potential is defined as a depth of dust that can be removed from a surface during a given period of time, and is primarily a function of exposure to winds sufficient for dust-lifting) [Haberle et al., 2003]. Low deflation potential also appear to exist in the study area over a variety of spin-axis scenarios [Haberle et al., 2003]; indeed, in many scenarios, net deposition of dust in Arabia Terra seems likely [Newman et al., 2005]. Thus, along with observations that suggest it is a dust sink today [e.g., Christensen, 1986], this modeling suggests it may also have been a dust sink in the past as well.
 One mechanism that may explain the apparent rapid erosion of the mantling deposit is if some of the material lost from the mantle was due to direct loss of a volatile component to the atmosphere via sublimation. Sublimation of surrounding material to the atmosphere has been proposed as a mechanism for forming pedestal craters [e.g., Barlow, 2006], which are perched above their surroundings like the inverted craters seen in this study area (though pedestal craters are less modified than the inverted craters seen here). Sublimation may also help explain the formation of large pits, moats, and fractures in the mantle material (Figures 5 and 6), which are hard to explain by aeolian erosion alone. However, even if direct loss of volatiles to the atmosphere played a role in modifying the mantling deposit, a significant amount of aeolian transport of disaggregated material away from the northeast Arabia deposit must have occurred, eroding not just portions of the mantling unit but also preexisting highlands materials. This may require different climate conditions and deflation potential in the Early Hesperian from what is observed and modeled today. One candidate scenario is a thicker early atmosphere.
 The terrain inversion that is observed in northeast Arabia Terra is striking, but inversion of relief of this sort is not unique to this location on Mars. Pain and Ollier  provide type examples of terrestrial environments where inversion of relief is known to have occurred in a variety of circumstances, such as alluvial deposits in Australia where terrain inversion has been driven by duricrust formation. On Mars, terrain inversion has been observed over a variety of scales in other locations on the surface [see, e.g., Malin and Edgett, 2003, footnote 13; Williams and Edgett, 2005; Burr et al., 2006]. This is likely to be due to the confluence of many factors on Mars which create a potential for terrain inversion [Pain and Ollier, 1995], including duricrust formation [e.g., Mutch et al., 1976], surface armoring [e.g., Conca, 1982], and aeolian erosion [Greeley et al., 1992]. Combined with the lengthy exposure of material on the surface due to low rates of surface change [Golombek and Bridges, 2000], these slow processes, acting over time, may have led to much of the inversion of relief observed on Mars.
 The sequence of processes that led to the inversion of valleys in northeast Arabia Terra (Figure 4) seem to differ from what caused inversion of the sedimentary deposits in Eberswalde crater, however [Malin and Edgett, 2003; Moore et al., 2003]. In Eberswalde, the inversion of relief is believed to have resulted from preferential preservation of relatively coarse grained channel deposits upon post-depositional aeolian erosion. We interpret the inversion of valleys in northeast Arabia Terra to result from deposition of mantling material that filled in and preserved valleyforms, rather than preferential preservation of fluvial sediments related to the valleys themselves. Thus, when small valley-forms are found exhumed out of the rock record in other environments on Mars [e.g., Williams and Edgett, 2005], it is possible that the materials marking the past presence of valleys are not necessarily fluvially derived sediments themselves. The cause of terrain inversion in a given location needs to be examined on a case-by-case basis.