4.1. Wind Regime From Dune Slipfaces
 Fenton et al.  found that the dunes of Proctor Crater (labeled “PC” in Figure 1) displayed three slipface orientations, labeled “primary,” “secondary,” or “tertiary” according to the size of the area in which they were found within the dune field. The primary slipfaces are oriented toward the east-northeast, the secondary slipfaces are oriented toward the west-northwest, and the tertiary slipfaces are oriented toward the west-southwest [see Fenton et al., 2003, Figure 5]. This multidirectional wind regime corresponds well with the observed combination of reversing transverse, double-sided barchan, and star dunes that are present in the Proctor Crater dune field. If these dunes were formed by regional winds (i.e., winds that extend across a large portion of Noachis Terra or more), then it is reasonable to expect a similar distribution of dune slipface orientations in the measurements made in this study. Conversely, if the Proctor Crater dunes were formed by winds derived from local topography (e.g., slope winds unique to the crater walls), then slipface orientations would probably vary from one dune field to the next, particularly for those dune fields not located in craters (such as those on the intercrater plains and off the edge of Hellas Planitia).
 Crater Np is located ∼150 km north-northwest of the center of Proctor Crater, and it contains the dune field closest to that of Proctor Crater. Its slipfaces (see Figure 4b) correspond well to two of those found in the Proctor Crater dunes. The most plentiful slipfaces, facing west-southwest, are similar in direction to the tertiary slipfaces from Proctor Crater. A few remaining slipfaces, facing east-northeast, are similar to the primary slipfaces. This correlation strongly suggests that at least two winds that influence the dunes in Proctor Crater are also found in crater Np.
 The dune field of Rabe Crater is located ∼300 km northeast of the center of Proctor Crater, the next closest dune field to Proctor Crater discussed in this work. The dominant slipface orientation in the Rabe Crater dunes is to the northwest (see Figure 10f). It is possible that these slipfaces are created by the same winds that produce the secondary slipfaces in Proctor Crater, although they are rotated ∼45° clockwise relative to the Proctor Crater slipfaces. If these slipfaces are produced by the same regional wind, they may have been deflected by local topography, such as the large pit that is unique to the Rabe Crater floor. Because these slipfaces are so prevalent throughout the dune field, it is possible that a southeast wind was responsible for transporting sand into the dune field. However, no sand streamers, dunes, or drifts indicative of a transport pathway are visible on the southern rim or directly south of Rabe Crater (although high resolution images here are sparse). If such a transport pathway ever existed, it is now buried or eroded. The minor slipfaces oriented to the northeast and southwest in Rabe Crater (see Figure 10f) may correspond, respectively, to the primary and tertiary slipfaces observed in Proctor Crater. As with crater Np, the dune slipfaces of Rabe appear similar to those of Proctor Crater.
 The next closest dune field to the Proctor Crater dunes is that on the intercrater plains, located ∼350 km northeast of Proctor Crater and ∼40 km west of the edge of Hellas Planitia. The most common slipfaces, oriented to the southwest (see Figure 6c), may correspond to the tertiary slipfaces observed in Proctor Crater, although they are rotated relatively ∼30° counterclockwise. The other dominating slipface orientation in the intercrater plains dunes compares very well with the primary slipfaces observed in Proctor Crater. As with the intracrater dunes discussed above, the intercrater plains dunes show similarities to those measured in Proctor Crater. This is perhaps the best evidence that the dune-forming winds are regional rather than local; their influence of dune morphology both on the floors of craters and out on the intercrater plains strongly suggests that these winds cannot be driven from local topography alone.
 The dunes on the western edge of Hellas Planitia are located ∼650 km northeast of Proctor Crater. The dominant slipface orientation ranges from west-southwest to northwest (see Figure 8c). As discussed in section 3.3.2, this spread in slipface orientation likely reflects two factors: deflection from local topography and two or more dune-forming winds that overlap on the rose diagram. This dominant orientation compares well with both the secondary and tertiary slipfaces observed in Proctor Crater. The second most dominant slipfaces in the Hellas dunes are oriented to the northeast, possibly corresponding to the primary slipfaces from the Proctor Crater dunes. The third most common slipfaces are oriented to the south-southwest; these could be produced by the same winds that create the tertiary slipfaces in Proctor Crater, although they are rotated relatively ∼45° counterclockwise and it is difficult to explain how two distinct winds observed in Hellas dunes could turn into the single tertiary slipface orientation observed in Proctor Crater dunes. Finally, a minor component of slipfaces oriented to the east-southeast are present in the Hellas dunes, but not found in any other dunes measured in this study. It is likely that the extreme topography of the Hellespontus Montes area and slope winds known to dominate circulation in the Hellas basin [Joshi et al., 1997] affect local wind speed and direction, possibly contributing to the high variability of slipface orientations observed in the Hellas dunes. It is unclear whether the winds that produce Hellas dune slipfaces truly correspond to those that produce the Proctor Crater dune slipfaces, but their similarity is suggestive.
 Each of the four sets of dune fields discussed above displays slipfaces that appear to correspond to those found in the Proctor Crater dune field. This strongly supports the idea that the dunes in Noachis Terra are formed by regional winds rather than local winds driven by small-scale topography. However, the proportion of these slipface orientations differs from one dune field to the next. For example, the dominant (what would be considered the “primary”) slipface orientation in Rabe Crater corresponds best to the “secondary” slipfaces of Proctor Crater. Furthermore, in some cases the winds appear to rotate by as much as 45° relative to those in Proctor Crater. Thus these regional winds vary spatially in relative strength and direction, probably as a result of a number of factors: both local and regional topography, variations in albedo and thermal inertia, position relative to upwelling and downwelling Hadley currents, influences of local and regional dust storms, and possibly others.
4.2. Short Transport Pathways
 In Noachis Terra, the aeolian sedimentary history is different from that typically observed on the Earth. In terrestrial deserts, sand transport pathways can often be traced “upstream” from a dune field to a sand streamer or sand sheet that feeds the dune field. Sometimes the sand transport pathway can be traced upwind beyond this point to further dune fields and sand sheets and possibly finally to a sand source. This is not the case with MOC WA, daytime IR, or nighttime IR mosaics, which reveal the Noachis Terra dunes to be isolated accumulations of sand, with no connections from one dune field to another.
 Although some sand sheets and streamers are visible in the MOC WA and daytime IR mosaics, they are relatively short (on the scale of the dune field width) and provide no clear link to upwind sources of sand. No sand sheets are apparent in any of the nighttime IR mosaics, where the dunes themselves are often difficult to identify. This lack of visibility in nighttime IR mosaics suggests that the sand sheets identified in MOC WA and daytime IR mosaics are thermally thin (less than a few centimeters thick). MOC narrow-angle images provide the best information on sand transport pathways.
 These images often confirm the presence of dark sand sheets where they are inferred in the mosaics. The presence of boulders or knobs appearing through sand cover in many MOC NA images of dark sandy areas is consistent with relatively thin sand sheets proposed from observations of daytime and nighttime IR mosaics. However, dark sand is more difficult to identify in winter MOC NA images, in which a lower sun angle (i.e., the sun is lower in the sky) and possible frost cover both reduce the contrast of dark sand with the surrounding terrain (see discussion in section 2.2, Figure 2). Furthermore, the coverage of MOC NA images is far from complete in most areas on Mars, including Noachis Terra. These limitations inhibit efforts to map out transport pathways of dark sand.
 The general absence of observable sand transport pathways in Noachis Terra has implications for the nature of the pathways themselves. Fenton et al.  proposed that the lack of a sand transport pathway for the Proctor Crater dunes indicates that any previously existing pathways have been either eroded away or buried. No such pathways are evident from data sets that span Noachis Terra, such as thermal inertia maps [Putzig et al., 2005] or MOC WA mosaics (see Figure 1). No sand streamers or small sand drifts are visible in the vast majority of the MOC NA images that (sparsely) sample the intercrater plains of Noachis Terra. However, it is possible such pathways were once widespread, actively carrying sand across hundreds of kilometers, with sinks in the present-day dune fields. The present-day wind may be strong enough to prevent the dune fields from becoming eroded or buried, but too weak to maintain the transport pathways. Alternatively, the sand supply feeding these potential transport pathways may have been limited, and any such transport pathways have disappeared because they “ran dry.” If this were the case, then these pathways must now be inactive, allowing more recent surface processes to erode or bury them, obscuring them from view. Although this option cannot be ruled out, another possibility exists: the transport pathways may simply be very short, and the dune sand may be derived from nearby local and/or regional sources, either of which could explain the observed situation in Noachis Terra. If the sources of sand are local, then the transport pathways leading from the sand sources to the dune fields may be very short. If the sand sources are regional, then the dunes will accumulate in areas where the winds converge. In this second case, transport pathways may be difficult to detect if they are ubiquitous (i.e., they may cover so much of the surrounding terrain that they appear indistinct).
 Not all dune fields in Noachis Terra are contained within craters. One dune field is located on the intercrater plains, and several dune fields are clustered on the western rim of Hellas Planitia, in the Hellespontus Montes. The possibility of short sand transport pathways suggests that sand sources for these dune fields are located on the intercrater plains. Sand sources for intracrater dune fields may come from within the crater, or perhaps on the nearby intercrater plains (or possibly both).
 The presence of dunes on the intercrater plains also suggests that dune fields do not accumulate in craters solely because sand is trapped inside them, prevented from escaping by steep crater walls, as suggested by Christensen . If this were the case, dune fields would be unable to accumulate outside of craters. Other factors, such as sand supply and wind regime, must also dictate why dune fields are common to crater floors. Sand sources preferentially located within craters could explain why few dunes are found on the intercrater plains. Additionally, similar morphology and dune slipface orientations among all of the dune fields studied in this work, regardless of where they are located, suggests that the same dune-forming winds blow across most of Noachis Terra. It is likely that dune fields accumulate where these regional winds, possibly funneled or impeded by local topography, balance one another.
4.3. Possible Regional Deposition
 Although sand transport pathways are difficult to detect in the low resolution mosaics, there are a few places in MOC NA images that suggest a possible nearby source for dune sand. Both instances occur within craters: Rabe Crater and crater Xn. These two craters are separated by a distance of more than 1200 km, roughly half the mean width of the Noachis Terra quadrangle, but the similarity of the two cases is striking.
 In both crater Xn and Rabe Crater, dark dunes are located near the walls of large pits that have similar physical characteristics. In both cases, dark material interpreted as sand is visible in gullies in a lower, night-dark (cool) layer (unit 1 in Rabe Crater), and that dark material is connected to dark dunes on the pit floors. It is possible that a stratigraphically higher night-bright layered unit (unit 2 in Rabe Crater) is eroding, shedding dark sand that is carried downward by way of the gullies and becomes incorporated into the dune fields. Although the dunes are dark in visible images, the layered units have roughly the same albedo as the surrounding crater floors (see Figures 10a and 11a). Therefore if the dark sand in these two craters is indeed derived from the resistant, night-bright layers exposed in the pit walls, the sand cannot be the sole component of the material in the layers.
 Likewise, unit 3 contains accumulations of dark material interpreted as sand (see Figure 10d). In Rabe Crater this unit lies stratigraphically and topographically above unit 2, the layered unit thought to be shedding dark sand directly into the dune field below. The accumulations of dark sand on unit 3 may also be derived from the underlying unit 2, perhaps blown up-strata by the wind. Alternatively, unit 3 may also contain some amount of dark sand that is accumulating in place as this unit erodes.
 The similarity between the two sets of wall units in both MOC NA images and nighttime THEMIS IR images is clear: a local sand source is eroding and contributing to the nearby dune field. The fact that the pattern is consistent for two craters that are separated by a large distance from one another (∼1200 km), and that the craters are quite different from one another morphologically, is noteworthy. The larger craters (>∼20 km diameter) in Noachis Terra also have bright floors in the nighttime IR mosaics, and these floors may be composed of a material similar to the night-bright layers in Rabe Crater and crater Xn.
 For example, crater Np has a fairly bright floor (in visible light) with a rough surface in MOC NA images (see mottled terrain in Figure 4c and interdunes in Figure 4a). This rough surface may be a material similar to Rabe unit 4. The terrain in the large pit in crater Np has an etched appearance (see Figures 2 and 4d), similar to that of Rabe unit 3. There are accumulations of material interpreted as dark sand in the crater Np pit (see Figure 4d), just as there are accumulations of material interpreted as dark sand in parts of Rabe unit 3 (see Figure 10d). It is possible that the sequence of units in Rabe Crater is repeated in crater Np, and that beneath the floor and pits in crater Np (possibly units 4 and 3, respectively), there is unexposed material similar to Rabe units 1 and 2.
 This raises the question of how extensive the Rabe units are across Noachis Terra: could the pit in Rabe Crater be exposing a sequence of sediments that were deposited across all or part of Noachis Terra, or does each crater contain its own distinct set of sedimentary layers?
 The sequence of Rabe units is not always consistent with what is observed in other craters. For example, although pits in crater Xn show units very similar to the Rabe units 1 and 2 (gullied and layered units, respectively), the surface of the flat crater floor above these units is dark in nighttime IR mosaics (see Figure 11c), unlike the nighttime-bright Rabe units 3 and 4. However, this disparity may have an explanation that is consistent with a hypothesis of regional sedimentation: it is possible that units similar to Rabe units 3 and 4 never formed in crater Xn (i.e., these units are not as spatially extensive as units 1 and 2), or that they did form and are now completely eroded away.
 Another inconsistency between Rabe Crater and crater Xn must be explained: the difference in the sizes of the dune fields. It is not hard to believe that the small dunes in crater Xn are composed of locally derived materials. If the dune sand was derived from outside of crater Xn, then it would have had to travel across one of the deep pits ringing crater Xn, and then across the flat crater floor to reach its current location. It is more probable that the dune sand is locally derived from pit and fracture walls, suggesting that the sand area has not traveled far from its source region. However, the Rabe Crater dune field is quite extensive (∼50 × ∼35 km wide, up to 500 m high), and it covers a higher proportion of the crater floor than most other dune fields in Noachis Terra (see Figure 1). Yet the dune sand could easily fill the cavity volume of the pit, so it is possible that the Rabe Crater dune sand is derived entirely from pit materials. Furthermore, no accumulations of sand (sand streamers or drifts) are visible in images southeast of (or on) the Rabe Crater rim, the most plausible upwind source direction based on dune morphology. Although a distant sand source for the Rabe Crater dunes cannot be ruled out, a local source is consistent with observations.
 The hypothesis of a regional set of deposits may also be consistent with the observed correlation of crater size and nighttime temperatures. In most areas discussed above, craters with diameters less than ∼10 km all have floors that are dark (cool) in the nighttime IR mosaics. Most craters with diameters more than ∼20 km have floors that are bright (warm) in the nighttime IR mosaics. There are a few possible explanations for this pattern. One is that the larger bolides impacted farther into the southern highland surface, exposing underlying rock that has different physical properties than the more shallow rock exposed by smaller bolides. However, the exposure of layered material in pits within crater floors indicates that some amount of sediment has filled in the larger craters, suggesting that the nighttime bright material in the larger crater floors is not the original crater floor but more recent infill. For example, Fenton et al.  estimated that Proctor Crater had ∼450 m of sediments filling its floor.
 Another explanation for the correlation of nighttime temperature and crater size relates to crater age. Most of the larger craters in Noachis Terra date to the Noachian epoch [Petersen, 1977], but some of the smaller craters may be much younger. It is possible that units such as those observed in Rabe Crater predate the formation of many of the younger craters in Noachis Terra. Thus the sedimentary units may never have been deposited in the smaller craters because they did not yet exist when deposition was taking place.
 Finally, the idea of widespread sedimentation across Noachis Terra may explain the many spots on the intercrater plains that appear bright (warm) in the nighttime IR mosaics. These nighttime-bright areas appear throughout most of the study area (see Figures 3c, 5c, 7c, and 9c). If the nighttime-bright units in Rabe Crater represent sedimentary units that were deposited over all or part of Noachis Terra, then the bright areas on the intercrater plains may be erosional remnants of similar (perhaps the same) units. The intercrater plains have poor coverage in MOC NA imagery, so that a systematic morphologic study of the night-bright spots is difficult. However, part of one of these bright spots is shown in the intercrater plains. The western edge of Figure 6b captures the eastern wall of a round mesa that is bright in the nighttime IR mosaics (see “5B” in Figure 5c). The top of the mesa has a dark mottled appearance that was interpreted as dark sand partly covering a bouldery or knobby surface. The wall of the mesa has bright lineations that may be outcrops of layers. The night-bright mesa has a texture similar to that of Rabe unit 4. If the night-bright areas in Noachis Terra are erosional remnants of sedimentary units, then they may be considered potential sources of dark dune sand. If so, then these nighttime-bright spots on the intercrater plains may be the sand source for the intercrater dune fields.
 If the nighttime-bright spots on the intercrater plains do represent erosional remnants of a formerly more extensive set of sedimentary layers, then it is possible to test the hypothesis that small (<∼10 km diameter) craters postdate the deposition of these layers. There are few examples of THEMIS VIS images (not shown) of small nighttime-dark craters on the intercrater plains, but they do indicate that ejecta blankets of such craters overlap adjacent nighttime-bright spots. In some cases where ejecta blankets have been destroyed, it is clear that the small nighttime-dark craters formed from impacts into the nighttime-bright spots. However, there are a few places where the stratigraphy is unclear, and it is possible that nighttime-bright material on the intercrater plains buried some small nighttime-dark craters, and these craters are now exposed by erosion. It seems that in most cases, small nighttime-dark floored craters postdate the deposition of sedimentary materials on the intercrater plains.
 The concentration of dune fields in crater floors (rather than on the intercrater plains) may be dictated in part by the amount of remnant sedimentary layers still present within craters relative to that remaining on the intercrater plains. The amount of sedimentary material removed by erosion and the amount of cratering on that which still remains indicates that the formation of the sedimentary layers, and much of their subsequent erosion, took place long ago. These layers (and possibly the dunes) are not a product of recent climate changes that are considered responsible for polar layered terrains and other younger sediments.
 The hypothesis of regional sedimentation provides a potential source of sand for the dark dunes that is consistent with the lack of observed sand transport pathways in Noachis Terra. In this case, the sand source is a set of regional-wide deposits that may be exposed only locally (either in crater floors or on the intercrater plains), leading to dune formation not far from each set of local sources. A regional set of sedimentary layers would have draped the underlying topography, covering low-lying crater floors, plains, and mountain ranges on the intercrater plains. The characteristics of these potentially sand-bearing sedimentary units provide information regarding their origin. For example, they could not be purely loess deposits, formed from dust that has settled from suspension, because the sand grains they contain would be too large to be carried into suspension in the first place. These units are also unlikely to be purely lacustrine deposits because of their potential extent out on the intercrater plains: while lakes may have formed in local topographic lows such as on crater floors, it is unlikely that lakes could have formed high up on the intercrater plains where the surface does not form a closed basin. It is possible that some of the deposits in some places may be either loess or ancient lake beds, but they cannot form the bulk of the sedimentary material. Other sources of material, such as volcanic ashfall or glacial till, cannot be ruled out by their spatial extent or the inclusion of sand grains in these deposits. In the future, better coverage with MOC NA and THEMIS images can help to connect a sequence of sedimentary units in one crater with those in another crater, or with those on the intercrater plains, allowing for further interpretation of sand sources.