4.4.1. Dark Filamentary Streaks
 Of the 26 MOC NA images taken of Proctor Crater before the end of the mapping mission, nine contain dark filamentary streaks that we interpret as dust devil tracks. One summertime image in Proctor Crater (MOC NA E1101316, not shown) shows a small bright spot ∼50 m across that is not present in an earlier, overlapping image. This bright spot has been interpreted as an active dust devil by K. Edgett (personal communication, 2003).
 Like the features described by Malin and Edgett , the dark filamentary streaks are long and thin, up to at least a few kilometers long and 10–50 m wide. Grant and Shultz  mapped many similar features on the floor of Proctor Crater, attributing them to “tornado-like tracks”. Dust devils are local vortices produced in unstable atmospheric conditions. The Martian surface, which warms up significantly more than the overlying atmosphere, is an ideal place for the forming of dust devils, particularly during the summer when surface heating is at its peak.
 These streaks are very similar to the “long dark filaments” first described by Cutts and Smith , although they did not attribute them to dust devils. Grant and Shultz  mapped these features in a portion of the Proctor Crater floor in both Mariner 9 and Viking images, noting that they appear in midsummer and disappear in the fall, and that their positions change from one year to another. These tracks almost certainly provide information on the orientations of winds of moderate strength. Weaker winds would not have the strength to move dust devils downwind [Metzger et al., 1999], and very strong winds remove kinetic energy from the boundary layer, preventing dust devils from forming.
 All the dark filamentary streaks in Proctor Crater are dark features. If they are dust devil tracks, then their lower albedo indicates that the dust devils that formed them removed or disrupted a relatively bright material from the surface as they passed by. The tracks cross dark sand, bright dune forms, and seemingly bare surfaces. Although there are tracks on dark sand sheets at the edge of the dune field, most dark dunes are free of these features, even in images where the tracks are dense just off the edge of the dune field. This suggests that if dust devils do pass over dark dunes, there is little bright dust present on the surface of the sand to be removed by a passing dust devil. With two exceptions (described below), the tracks are found in images ranging in season from Ls = 223°–354°, or from mid spring through late summer. This is the same time of year that they were observed by Grant and Shultz  in Mariner 9 and Viking Orbiter images. As discussed in section 4.4.2, seasonal frost begins to appear in Proctor Crater partway through autumn, and remains on shady slopes until late winter. Thus dust devils appear to form and create erosive tracks during the warmest time of year when surface heating is at a maximum and frost is absent.
 Figure 3g shows an example of the dark tracks on the wall of the western pit. In this image the lower left corner is high ground. The layers are visible in a steep wall that crosses the frame from the upper left to the lower right. The upper right corner is low ground, and largely covered in bright dust that has been removed in places by dust devils. The dark splotches on the hillside are interpreted as accumulations of dark sand that has been transported into the region from the southwest (lower left), much like the falling dunes of Figure 3e. Some of the dust devil tracks appear to originate in the dark sand, and then to move downwind to the ENE. The surface heating over dark sand will be greater than that over bright dust, and so it is probably easier for dust devils to form over dark sandy surfaces than elsewhere. Further discussions below of dust devils in the vicinity of the dark dune field are consistent with this idea.
 Figure 12 shows several examples of dust devil tracks in Proctor Crater. Figure 12a shows the locations of each example. Figure 12b shows several dust devil tracks that are quite apparent when crossing over a featureless surface, but appear only faintly over a nearby outcrop. A few possible scenarios can explain this observed difference. The outcrop may be relatively free of a dust cover that is being removed elsewhere by passing dust devils. Alternatively, the outcrop may be comprised of material with an albedo similar to that of a regional dust cover, so that any removal of dust leaves behind no obvious trail.
Figure 12. Dust devil tracks on the Proctor Crater floor. (a) Context for Figures 13b–13h. (b) Tracks crossing a featureless plain but not across a bright outcrop (MOC NA M11-03806). (c, d) Overlapping images in the late winter and late spring, demonstrating the appearance of a dust devil track (MOC NA M03-00338, M09-06250). (e, f) Overlapping images from the late spring, showing a growing number of dust devil tracks (MOC NA M08-02629, M10-01334). (g, h) Images showing potential dust devil tracks in the winter time (MOC NA M03-03087, MOC NA M07-01445), although the faintness of the tracks (images have been greatly stretched) may indicate that these tracks are remnant features from the previous summer season.
Download figure to PowerPoint
 There are two examples of MOC NA frames that overlap, allowing for a study of dust devil development through the spring and summer. Figures 12c and 12d show an example from inside the western pit. In Figure 12c, an image from late winter, there are no signs of dust devil tracks. In Figure 12d, there is one faint track marked by white arrows. It is oriented ENE-WSW. The black arrow shows a remnant patch of seasonal ice on a shady slope that has disappeared by late spring, when the second frame was taken.
 The second set of overlapping images is shown in Figures 12e anf 12f, which are located just south of the edge of the dark dune field. In late spring, in Figure 12e, the dust devil tracks are not very abundant and seem to be concentrated near the edge of the dune field. Later in the summer, in Figure 12f, the dust devil tracks are much more prevalent. Most of the tracks in the overlapping portion of the springtime image are still present in the summertime image, indicating that they are not often erased by strong summer winds. The average orientation of dust devil tracks does not appear to change from one frame to the next, indicating no net seasonal shift in wind direction. This growth in density from spring to summer suggests that the dust devil tracks develop each year and do not persist from season to season. In addition, the dust devils that form early in the season may be more common on the dark dunes than on the rest of the crater floor because the low albedo of the dune sand increases surface heating, possibly encouraging dust devil formation as well. The fact that springtime tracks are not erased by summer winds may indicate either that summertime winds are not strong enough to erode away spring tracks, or that the winds may indeed be strong, but that there is no loose sand available to erode older tracks.
 There are two examples of dark streaks observed before the start of the dust devil season, shown in Figures 12g and 12h (white arrows). In each case the image contrast was stretched a great deal to show these features. Both sets of streaks are aligned with most of the other dust devil tracks (see discussion below). Because of the time of year at which these images were taken and the faintness of the features, we suggest that these streaks are old dust devil tracks from the previous summer. If this is the case then they have recently become defrosted. It may be that over the course of the year, dust fallout obscures dust devil tracks so that each spring, the surface presents a “blank slate” for dust devils to erode. Perhaps such a process is not always complete, sometimes leaving old tracks incompletely obscured. Alternatively, dust fallout may continue after the time at which these images were taken, obscuring them before the summer season commences.
 Only one summertime image on the crater floor contains no dust devil tracks (not shown). This is in the same area north and east of the dune field in which the small bright dune forms appear rounded and eroded (see Figure 11c). It is not clear why dust devil tracks are not produced in this area. If this is a place of constant wind erosion, then there may be no thin layer of settled dust to be lifted by a passing dust devil. Either this is a place where dust is prohibited from settling, perhaps from strong winds, or it is quickly remobilized with a mechanism other than dust devils, such as saltation in local sand sheets. Another explanation may be that there is dust on the surface here, but that there is little contrast between it and the underlying surface. However, thermal inertia calculations, discussed in section 4.3.3, are high enough that there can be very little dust on the surface in this area, supporting the proposal that this area does not accumulate fines. Not surprisingly, the effectiveness of wind action appears to vary spatially across the crater floor.
 To determine the daytime wind regime during the summer, the orientations of the most prominent dust devil tracks were measured using an application in Arcview. Figure 13 shows a rose diagram (i.e., a histogram on a polar plot) of the mean orientation of 196 measured dust devil tracks. Because determining the upwind versus downwind direction is impossible from observing most dust devil tracks, all directions shown have been restricted to 0° to 180°. Tracks oriented at 0° or 180° are oriented north-south, and tracks measured at 90° are oriented east-west. One modal direction is evident in Figure 13, with a spread from 60°–100°, or generally ENE-WSW. This is the same main orientation of dust devil tracks mapped by Grant and Shultz  in both Mariner 9 and Viking Orbiter images. Because of the dust devil tracks appearing to initiate on the sand patch shown in Figure 3g, we consider these dust devil tracks to be formed by winds from the WSW. This direction also corresponds to the dune orientations observed by Cutts and Smith , as well as those measured in this work, as discussed in section 4.1.2.
Figure 13. Rose diagram of dust devil track orientations. Orientations of 0° or 180° indicate a north-south alignment, while an orientation of 90° indicates an east-west alignment. Because of an upwind versus downwind measurement ambiguity, all orientations are restricted to an orientation of 0°–180°. Note the two concentrations of orientations at 60°–80° and 90°–100°.
Download figure to PowerPoint
 The orientation measurements of dunes and dust devil tracks have several implications. The persistence of dust devil track orientations from one mission to another indicates that daytime summer winds are very consistent in direction. Because the dunes in Proctor Crater are so large, they have a long memory, and thus reflect the prevailing winds over at least several decades, and possibly over the last million years (assuming that they are active). The correspondence of dust devil tracks to dune slip face directions indicates that these wind directions have been very typical of this area in the summer for some time. Furthermore, dunes require winds above the saltation threshold to shape them, but dust devils may move under lighter winds (Metzger et al.  estimates that ∼5 m/s winds will move dust devils), and so this alignment indicates that both strong and moderate winds blow in this direction.
4.4.2. Frost Features
 Malin and Edgett  studied the frosting and defrosting of dunes at both poles. They found that the dunes are generally the first features to develop frosted surfaces during autumn, and they are the last features to lose frost in the spring. They suggest that Martian dunes may trap volatiles, much as Sharp  found that the Kelso Dunes of the Mojave Desert trap water. In addition, Malin and Edgett  found that defrosting tends to begin with small dark spots commonly located at the dune margins, which enlarge slowly and coalesce until the entire dune surface is defrosted. They proposed that dark sand beneath thin bright frost accumulations warms more quickly than in other areas, causing the frost to sublimate in patches. Little is known about the relationship between frost and dune sand on Mars, but the MOC narrow-angle images indicate that the interactions are complex and certainly worth further study.
 Although Proctor Crater is located in the midlatitudes (47°S), rather than near the poles, the region is covered in seasonal frost each winter. Thus the numerous images of the dune field provide an opportunity to compare how these dunes frost and defrost relative to the polar dunes. MOC narrow-angle images show that frost cover begins in mid southern autumn, at approximately Ls = 50°, and remains in patches on the dunes until late winter at approximately Ls = 165°. Because Proctor Crater is located in the midlatitudes, the seasonal frost cover does not last as long as it does closer to the poles. Image coverage during the southern fall is too sparse to determine if the dunes acquire frost before the remainder of the Proctor Crater floor does, as is expected. However, poleward facing slopes on the dunes are certainly the last surfaces to defrost in late winter, consistent with the observations of Malin and Edgett [2000b].
 It is not clear from MOC NA images whether the seasonal frost is CO2 or H2O ice, or some combination thereof. Figure 14 shows spectrally derived surface temperatures from TES over the Proctor Crater dune field within the solar longitude range that frost is apparent in MOC NA images. Bandfield and Smith  show that these temperatures in a fairly nondusty atmosphere should be no more than ∼2 K lower than the actual surface temperature. Nighttime temperatures in Figure 14 remain close to the CO2 frost point until about Ls = 135°, indicating that CO2 frost probably forms on the surface of the dunes during the winter night. However, the daytime temperatures only drop down to 170 K, right at the winter solstice (Ls = 90°), indicating that CO2 frost is unstable on the dunes during the daytime. We interpret these data to indicate that the frost that is visible in the MOC NA images, which were obtained during the warmest time of day, most likely show a thin veneer of H2O frost that persists throughout the day, while CO2 frost forms at night and sublimates the following day.
Figure 14. Spectrally derived surface temperatures from TES over the Proctor Crater dune field. The temperatures are separated into day and night measurements, and they are plotted as a function of season during the winter. Nighttime temperatures indicate the presence of CO2 frost, but daytime temperatures are too warm for CO2 frost. Thus the seasonal frost visible in MOC NA images is interpreted as H2O frost that survives during the day.
Download figure to PowerPoint
 Figures 15a and 15b show overlapping images during the late winter from two consecutive years. Each image shows partially defrosted dunes within the Proctor Crater dune field with dark spots similar to those found by Malin and Edgett [2000b]. Interpreting images with partial frost cover can be tricky, because features can appear bright from either frost albedo or incident sunlight. Determining whether an area is inherently bright or only apparently so from shading differences requires experience with viewing several MOC images, preferably over different seasons and under different lighting conditions. Figure 15c shows an interpretation of the features shown in Figure 15b, built from such a knowledge base. Dune crestlines refer to linear peaks of dunes. Slip faces generally begin from these crestlines, but not all slopes that reach to the crest are necessarily slip faces. Bright dune forms are located in the interdune flats and thus represent low-lying areas. Slope or slip face adjustments, as defined here, are avalanches of sand that has been oversteepened by the wind at the brink of a dune. Such slope adjustments are generally oriented downhill and are good indicators of the local gradient. The patchy seasonal frost remains in low-lying areas and on the southern and western sides of dunes. Shadows are located southeast of dune crests, created by a low winter afternoon sun to the northwest.
Figure 15. Frost in the dark dune field. (a, b) The same area nearly 1 Martian year apart. Defrosting spots occur in the same locations, suggesting that they are produced by some underlying persistent aspect of the dunes (MOC NA M02-02711, E03-01039). (c) A diagram of the features in Figure 15b. (d, e) The same region nearly 1 year apart, with new slip face adjustments formed within the year (MOC NA M02-02711, E03-01039). (f–h) Sequence of images over the same area of the dune field, each showing slope adjustments that persist for at least 1 year. Figure 15f is a late winter image of year 1, where such adjustments appear to be superimposed on seasonal frost, giving the impression that they are fresh features (MOC NA M02-02711). Shortly after the frost has disappeared from the slope (Figure 15g), the same adjustment scars are still visible (MOC NA M03-03088). Nearly a year after the first frame (Figure 15h), when the frost has reformed, the slope adjustment scars are still visible, giving the impression that they are recent movements over the seasonal frost. Rather, they must be somewhat older features that inhibit frost formation or allow frost to sublimate more easily (MOC NA E03-01039).
Download figure to PowerPoint
 The Proctor Crater dunes develop dark spots as they defrost that persist in location from year to year (e.g., compare Figures 15a and 15b). The repetitiveness of dark spot locations indicates that their position is dependent on some relatively stable aspect of the dune surface. Unlike the dark spots on polar dunes, the dark spots on the Proctor Crater dunes are concentrated on steep slopes, rather than along the dune margins. In addition, these features rarely appear on hills facing any direction but toward the pole. The largest spots in Figures 15a–15c contain bright cores, which have not been observed in polar dunes. The cores are brighter than frost on the surrounding slope, indicating that this material is not simply a remnant frost patch from the previous uniform cover. Bridges et al.  found that dark spots located in small gullies on Mars, similar to the spots on the dunes, are aligned with the local dip and channel trend (i.e., downhill along the channel). In contrast, the spots in the Proctor Crater dunes appear to be aligned either along the strike of the south facing dune slope or parallel to the crest.
 There are a few possible explanations for the presence and location of the dark spots. The spots may be associated with granule ripples that can form on dune slopes. However, in areas where there is summertime coverage of slopes that form dark spots while defrosting, there is no evidence for any features on the smooth dune slopes. Furthermore, there is no physical reason why granule ripples would preferentially form on pole-facing slopes, as the dark spots do. Another idea is that the dark spots are small avalanches over the seasonal frost cover, although no known mechanism causes slope adjustments to form in the same locations in the middle of dune slopes year after year. Finally, it may be that the dark spots are concentrated along interfaces between exposed dune strata. Wind erosion on terrestrial dunes often exposes the internal dune strata. Furthermore, water ice tends to accumulate along strata interfaces in terrestrial dunes. Snowfall can form icy strata within the dune that persists from year to year [e.g., Calkin and Rutford, 1974]. A similar process could occur on Mars with either H2O or CO2 ice. If this occurred, then insolation would warm the exposed sandy strata more than the nearby exposed icy strata. This could easily cause reprecipitation of sublimating frost onto the exposed icy strata, creating the bright cores of the dark spots seen in Figure 15b. This mechanism is similar to that proposed by Malin and Edgett [2000b] to explain bright frost halos around dark sublimation spots.
 Alternatively, the dark spot cores could brighten as wind blows frost from elsewhere into small cracks that are exposed by defrosting, such as the process described for polar “spiders” by Piqueux et al. , although summertime images of dune slopes show a surface devoid of any obvious roughness that could trap windblown ice grains. The melting of snow lenses trapped in terrestrial dunes is known to create unusual surface features, such as small sinkholes and tensional cracks [Koster and Dijkmans, 1988]. However, there is no such known process on Mars, where ice trapped in dunes may never melt.
 Since these dark spots mostly form on the shaded slopes and manage to persist much longer into the season than on other slopes, the frost may have enough time to undergo sintering caused by compaction and grain growth in a process modeled by Eluszkiewicz . Thus preferential sublimation may occur from differential ice grain growth, in which transparent areas absorb more heat from insolation and sublimate faster than opaque areas. This process may in turn be influenced by ice trapped in underlying exposed dune strata. For example, the weight of frost overlying an exposed icy layer between sandy strata may compact some of the upper icy layer enough that it sinters into larger grains, which become more transparent to sunlight as they grow. As the frost slowly thins from sublimation, this relatively transparent layer becomes more and more exposed to daily heating until enough heat is collected to begin localized sublimation. There is no reported terrestrial analog for this process, or for defrosting spots on sandy surfaces such as these. This process is similar to that suggested by Malin and Edgett [2000b] and Bridges et al.  to explain dark sublimation spots. In the case of the Proctor Crater dunes, sintering in icy dune strata may explain the alignment of dark spots along the slope of the dune. The presence of ice strata may also explain the bright cores, as icy strata may be cold enough to cause reprecipitation of locally sublimating frost, once the initial wave of defrosting has occurred and swept outward from the center of the spots.
 Edgett and Malin  show slip face adjustments in Proctor Crater dunes and propose that they are fresh movements over seasonal frost. However, as they consider, and as discussed below, such slope adjustments may not be as fresh as they appear. Fortunately, there is evidence suggesting that the dunes in Proctor Crater are active from year to year, although the season in which these movements occur is not constrained. Figures 15d and 15e show the same area roughly 1 year apart. Figure 15d is the same area shown in Figure 17 of Edgett and Malin . In the first frame a frost-covered slope shows superimposed dark tongues of sand that are typical of dune slip face adjustments from easterly winds. In the second frame, taken nearly 1 year later, seasonal frost has since sublimated away and reformed on the dune field. The slip face features of the previous year are barely discernable beneath the recent year's frost accumulation. However, new dark tongues of sand have formed farther south, overriding some older scars that are faintly visible in the first frame. In both cases the sand-moving winds come from the east. These dunes are without question active with fresh slip face scars forming each year. However, these movements do not appear to be very common, as not all of the slope adjustments of the previous year are buried by subsequent activity.
 There is enough coverage over enough time to show that some slip face adjustment scars remain for at least a year, and that they therefore may superimposed on annual frost cover as suggested by Edgett and Malin . Figure 15f–15h show the same area at three different times, each showing the same set of slip face adjustments. Figure 15f was taken during the time of partial frost cover, in late winter. Arrows point to a prominent dark slope adjustment lobe that appears to have formed over the frost. Figure 15g shows the same area later that year, when most of the frost has sublimated away. In this frame, most of the bright surfaces are sunlit slopes, whereas in Figure 15f most of the bright surfaces are frosted slopes. Even though the frost is gone, the dark stripes from slip face adjustments are still visible. The following winter, in Figure 15h, the same dark slope adjustments are still visible, although a new larger and darker scar has formed nearby in the intervening year. It appears that although these features appear to be superimposed on seasonal frost, they are not. Some aspect of the nature of these scars inhibits frost from forming, and/or allows frost to sublimate more easily than on the surrounding slopes, making them appear dark relative to the ice-covered slopes. Slip face adjustments loosen the surface material, and it may be that these less densely packed surfaces expose more surface area to the air and thus allow frost to sublimate more quickly than on a more densely packed surface. The assessment of frost features on the Proctor Crater dunes reveals more about the nature of the dunes than it does about the frost that covers them.