Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
 An accurate self-consistent way of coregistering the imaging and topographic data sets of the Mars Global Surveyor mission was developed and used to begin a stratigraphic analysis of the northern polar region. A distinct change in the layering style exists at a definite stratigraphic horizon near the base of the north polar layered deposits. Occurrences of the contact between two distinct layered units can be mapped hundreds of kilometers apart at nearly the same Mars Orbiter Laser Altimeter (MOLA) elevation. The lower layered unit has a consistent association with sand dunes, leading to the conclusion that it is an eroding sand-rich deposit that predates most of the overlying north polar layered deposits, which exhibits the expected features of a dust-ice mixture. These results suggest that an areally extensive erg was in existence before the present ice cap and that the present circumpolar erg is likely composed of material reworked from this older deposit. The volume of this lower unit is estimated to be on the order of 105 km3. The presence of this deposit implies that there existed a period in Mars' history when there was no icy polar cap. A dramatic climatic change leading to the deposition of the upper layered (icy) unit in the present-day polar layered deposits represents a major event in Mars' history. However, owing to uncertainties in the mechanics of layered deposits formation, such an event cannot be dated at this time.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The northern polar layered deposits are additionally complicated by the large sand sea encircling them [Tsoar et al., 1979; Dial, 1984; Tanaka and Scott, 1987] which is mostly composed of the Olympia Planitia dunefield. Many authors have noted that erosion of the layered deposits may be a source of the dark sand-sized material [Breed et al., 1979; Thomas, 1982; Blasius et al., 1982; Lancaster and Greeley, 1990; Thomas and Weitz, 1989] which makes up the circumpolar erg. This would seem to pose a problem for the accepted formation scenario of layered deposits, which relies on aeolian deposition of fine dust in accumulating ice. Sand is not expected in the layered deposits because the polar cap rises over a kilometer above the surrounding plains and sand-sized particles (at least in the present atmosphere) cannot be carried by suspension. It has been suggested that the dark saltating material could be composed of dust aggregates originating in the polar layered deposits [Greeley, 1979; Saunders et al., 1985; Saunders and Blewett, 1987]. However, Thomas and Weitz  noted that the Viking color and albedo values derived for the north polar dunes do not differ significantly from dark dunes anywhere else on the planet. In contrast, Viking-based thermal inertia results suggest that the circumpolar erg has a lower bulk density than dunefields at lower latitudes [Herkenhoff and Vasavada, 1999; Paige et al., 1994], which lends support to the idea that the dunes are composed of dust aggregates or perhaps unweathered, basaltic fragments.
 The Mars Global Surveyor (MGS) spacecraft has been in polar mapping orbit since March of 1999 [Albee et al., 1998, 2001]. Instruments on board used in this study are the Mars Orbiter Camera (MOC) [Malin et al., 1992] and the Mars Orbiter Laser Altimeter (MOLA) [Zuber et al., 1992]. Over a year's worth of topography measurements and acquired images have been released to the public. Owing to the spacecraft's near-polar orbit, both the imaging coverage and the topographic measurements are densest in the polar regions. The combination of this new topographic knowledge and the ability to accurately locate high-resolution images relative to that topography makes possible for the first time a more detailed stratigraphic analysis of the stack of layers which makes up the polar deposits at both poles. It seems certain, now that this kind of detailed analysis is possible, that many stratigraphic studies, previously impossible, will be undertaken. Here a surprising result, which was apparent once the technical hurdles of precisely coregistering these data sets were resolved, is reported.
 The focus of this paper is to document a distinct change in the style of layering exposed in the many troughs and chasms within the northern cap at a consistent elevation. Layers above this horizon have the expected features of the canonical dust-ice mixture [Thomas et al., 1992; Toon et al., 1980; Cutts et al., 1979; Squyres, 1979; Cutts, 1973b]. Layers below this horizon differ markedly in albedo, in morphology, and in resistance to erosion from those above. A strong association of exposures of this lower section with occurrences of dune material is demonstrated, leading to the conclusion that it is this specific section within the north polar layered deposits that is the source of the material comprising the current circumpolar erg.
2. Data Preparation
 To facilitate this and future work, high-resolution digital elevation models (DEM) of both polar regions were constructed following the method of Neumann et al.  using the Generic Mapping Tools package [Wessel and Smith, 1998]. The northern DEM was constructed by fitting a continuous curvature surface to ∼28 million independent MOLA altimetry measurements. Preprocessing of the data helped reduce the volume of points to be fit and short-scale aliasing problems. Data points were initially selected to satisfy certain quality requirements (G. Neumann, MIT, personal communication, 2001) judged by parameters such as along- and across-track shift and crossover residuals. Off-nadir tracks were excluded, leaving no topography information within ∼180 km of the pole (poleward of 87° north); this region was separately dealt with. The spatial resolution was chosen to be 200 m, and the region covered is roughly 1200 × 1200 km centered on the pole. Tracks visibly offset from their surroundings were removed, and the surface was regenerated.
 The derived shaded relief map, at 200 m per pixel, has a higher resolution than either the MOC wide-angle camera or U.S. Geological Survey (USGS) Mars Digital Image Mosaics (MDIMS). In addition, shaded relief has advantages over visible imaging in that there is little pixel to pixel smearing and lighting and atmospheric effects are not an issue. For these reasons this shaded relief product was chosen as our basemap, although individual MOC wide-angle images continue to be used for albedo information where needed.
 MOLA points acquired simultaneously with a MOC narrow-angle image can be located within the image by line and sample number using spacecraft time and information on the coalignment of the narrow-angle CCD and MOLA boresight (S. Anderson, A. Ivanov, JPL, personal communication, 2000). Similarly, those line and sample positions can be related to MOLA-derived latitude and longitude points, which are superior to those obtained from normal spacecraft orbital information since they include the MOLA crossover correction [Neumann et al., 2001]. This allows the narrow-angle image to be map projected in a best fit way to ensure that those MOLA points project to the correct position on the MOLA-derived basemap. In this way, MOC images can be placed relative to each other and their surroundings in an accurate self-consistent way. Where narrow-angle images overlap the mutual offset is commonly observed to be a few tens of meters. Radiometric calibration of the narrow-angle images was performed with the USGS ISIS software package.
 The imaging and topographic data sets were combined in a geographic information system package (Arcview, by ESRI) modified by us for use in the Martian polar regions. This makes spatial relationships between images (in three dimensions) clear, allows different data products to be overlain, and permits easier distinction of topographic and albedo effects.
3. Change in Layering Style
 Sections of the layered sequence within the cap and layered deposits are exposed in the many troughs and arcuate scarps both at the cap edge and in the interior. The purpose of this paper is to document a clear division of this layered sequence into two parts at a definite stratigraphic horizon. The upper, younger sequence is finely layered with smooth outcrops showing individual layers visible down to the limit of the camera resolution (hereinafter referred to as the finely layered unit). The lower, older sequence is distinctly different in layer morphology, albedo, and material properties. It has a characteristic irregular thick-plate-like structure in exposures, such that individual layers are less uniform both vertically and laterally (it is hereinafter referred to as the platy unit). This section has lower albedo than the overlying finely layered section. Layers in both sections appear to be close to flat lying; however, this is less certain in the platy unit owing to the more confusing morphology. Hereinafter, the two sections of the northern polar layered deposits described will be referred to as distinct geologic units. This division of the polar layered deposits was noted by Malin and Edgett  in their review of the first year of MOC observations.
 The transition between these units is sharp and distinct. Figure 2a shows some examples of the contact between the units discussed in MOC narrow-angle images. Figure 2b shows how this contact is identifiable in MOLA-derived shaded relief maps; the lower unit protrudes out of the bottom of scarps as a prominent step. We used the shaded relief images to map the contact between available narrow-angle MOC frames. This contact is visible over a large area; Figure 1c shows where the contact can be mapped from a combination of MOC narrow-angle frames and the MOLA shaded relief map in the 120°–240°W region of the cap. Figure 1b shows that occurrences of this contact can be found as far away as the head of Chasma Boreale over 600 km distant. The minimum area of a circle needed to enclose all the mapped occurrences of this contact is roughly 0.4 million km2, so assuming that the lower unit is continuous and has a near-circular shape, it is at least that extensive.
 Significantly, wherever this contact is visible in the 120°–240°W region, it is observed to occur in the elevation range from −4400 to −4200 m. The large horizontal distances and narrow vertical distribution of instances of this contact lend credence to the interpretation that it is a widespread and definite change in deposition style and material composition contained within the stratigraphic record rather than isolated patches of unusual layering with no relevance to each other. Figure 3 shows a topographic profile taken from the DEM along the line indicated in Figure 1c. This line was chosen as it crosses a number of locations where the surface intersects the elevation range mentioned and also because at these locations narrow-angle MOC frames are available to confirm the continued presence of the contact. In each case the contact appears at the expected elevation.
 The lower platy unit also seems to have different material properties than the overlying finely layered unit. On scarps where both units are exposed, breaks in slope are commonly observed at the contact, as illustrated in Figure 4. In other instances the lower platy unit is observed to protrude from the base of the scarp as a step; see, for example, Figure 5. In all cases a different composition for the lower platy unit, which is more resistant to mechanical or thermal erosion, can be inferred.
 The differences in resistance to erosion, morphological appearance, and albedo strongly indicate that the lower unit differs in both composition and deposition style. The sharp appearance of the contact could indicate that this change was rapid, at least compared to the timescale of layer formation, or that there may be an erosional unconformity separating the two units, representing an unknown amount of time.
 No exposures of the platy unit at the cap edge in the longitude ranges of 0°–70°W and 225°–360°W are seen, although Chasma Boreale contains exposures within these ranges. The quality of the current narrow-angle coverage was unfortunately adversely affected by a dust storm, which occurred during the M02 and M03 mapping phase (H. Wang, Caltech, personal communication, 2001). Many of the narrow angles acquired in this period contain very little surface information. Imaging conditions for the next Martian year will hopefully be clearer. It is possible that as more high-resolution imaging is acquired and released for pubic use, exposures of this unit will be discovered in other regions.
4. Correlation With Duneforms
 One of the most important observations of the lower platy unit is the fact that exposures of this unit are highly correlated with the nearby occurrence of dune bed forms. Dunes require saltating particles to form, which usually implies sand-sized material but conceivably could also be sand-sized aggregates of dust [Greeley, 1979; Saunders et al., 1985; Saunders and Blewett, 1987]. The dunes observed here will be referred to as sand dunes; however, the possibility that sand-sized dust aggregates may play a role will be left open.
 Many of the exposures of the contact between the two units discussed in section 3, in the longitude range 140°–240°W and at the cap edge, occur in closed depressions next to arcuate scarps. Although these closed depressions are “uphill” from, and their bases are higher in elevation than, the main circumpolar erg, deposits of dune-forming material are present. There is no evidence of dunes in the region between these depressions and the main erg, which would have implied poleward or equatorward transport of dune material. Thus it appears that the dune material was derived in situ. Figure 6 shows some examples of dune material located adjacent to exposures of the lower platy unit. Dunes are also visible in Figures 4, 8, and 9 next to the exposed platy unit. Thomas and Weitz  also noted that dunes appear to have restricted sources within the polar layered deposits. Although with Viking imagery they were unable to see the platy unit, they realized that the sources of dune material are probably in the stratigraphically older deposits.
 In another area of the cap edge at roughly 125°W (illustrated in Figure 1d), there is a large reentrant with an associated train of dunes, which are migrating southwest toward, and eventually join, the main circumpolar erg. If formed from the material eroded at the scarp, these dunes have traveled from at least where the scarp edge used to be at the time of erosion to the edge of the erg, a minimum distance of 130 km. This large minimum distance, their dark color, and the inherently violent nature of the saltation process cast doubt on the possibility of these dunes being composed of sand-sized dust aggregates which may not be strong enough to survive intact. Saunders et al. [1985, 1986] have suggested that dust aggregates can survive saltation up to a few tens of kilometers and that they can acquire their dark color by gathering carbonaceous meteoritic dust as they saltate. In this case, however, the minimum saltation distance exceeds their quoted survivability distance, and the amount of dark carbonaceous meteoritic dust must be low since this material saltates over an otherwise high-albedo surface. The thermal inertia results [Herkenhoff and Vasavada, 1999; Paige et al., 1994], combined with the above discussion, lend support to the idea that the dunes are composed of unweathered basaltic fragments. However, it remains unclear as to the applicability of these thermal inertia results. High local slopes within dunefields have been ignored in all thermal models, and the high emission angle of the Viking observations in this area makes it likely that thermal measurements have been dominated by the “hot” side of these dunes. These two facts combined could possibly lead investigators to infer an incorrect value of thermal inertia.
 Although in general dune bed forms are not resolved by the MOLA data, dunefields are easily identifiable from the derived shaded relief images as areas of apparently rough texture. Figure 1d shows the correspondence of rough texture in shaded relief maps to dark albedo regions in MOC wide-angle images commonly associated with dunes. Where higher-resolution narrow-angle images exist, they confirm that the dune-covered areas can be correctly identified from the MOLA data as illustrated in Figure 1d (iii).
 A section of MOLA profile 13387 over the steepest portion of the scarp is shown in Figure 7; its position relative to the scarp is shown in Figure 1d. The maximum scarp slope along the MOLA track is 60.2°. When this is adjusted to take into account the nonperpendicular strike of the MOLA trace to the scarp, a 67.4° downhill slope is inferred. Even higher slopes may be possible over length scales shorter than the 300 m altimeter shot separation. The extremely steep slopes, arcuate shape, and interior ridges, which parallel the scarp edge and are visible in Figure 1d (i), of the main reentrant strongly indicate an origin through successive landsliding. We interpret these interior parallel ridges to be remains of the lobate deposits expected of landslides. Why landslides should have occurred in this area is unclear to us. However, given the sharpness of the scarp and the possibility of glacial-like flow or at least relaxation discussed by many authors [Nye, 2000; Fisher, 1993; Zuber et al., 1998], it seems likely that at least the last of these landslides was a recent occurrence.
Figure 8 shows a fortuitously placed MOC narrow-angle frame, whose position is also marked in Figure 1d (i). Although this image is taken in poor illumination and atmospheric conditions, the platy unit can clearly be seen protruding from the base of the scarp. Also of interest and perhaps more clearly visible in the Figure 1d (i) context image is the 6.4 km crater immediately adjacent to the scarp which appears to have been uncovered recently by its retreat. Either an impact or some endogenic process may have formed the crater. The size of the feature lends support to the theory that a long period of time may be represented by an unconformity between the finely layered and platy units. In the wall of this crater is a clearly visible exposure of the platy unit. Dunes cover the eastern part of the crater, providing yet another example of the correlation between contemporary dunes and the platy unit.
 Immediately to the northeast of the main reentrant is another arcuate scarp, and again the platy unit is visible along with associated dunes; see Figure 9 (outline of this image is also marked in Figure 1d). The dunes have collected in one corner of the depression. There is a ridge ∼100 m high separating this depression from the main reentrant. However, there is a route where slopes are low enough so that the dunes derived from the material within this depression can climb up on top of this ridge and escape down the other side into the main reentrant and ultimately the circumpolar erg. The morphology of the dunes is consistent with saltation in this direction. Although this ridge is only a few pixels wide in the MOLA shaded relief image, it has the heavily textured appearance associated with dune cover in other locations. Wide-angle views of the same location also show a thin dark streak connecting the two areas.
 Chasma Boreale provides an exposure of over a kilometer of relief, cutting directly into the polar cap. It intersects the level at which one would expect to find the contact between the platy and finely layered units. Figure 10a shows the contact at the eastern head scarp of the chasam, occurring close to the expected elevation hundreds of kilometers away from the previous examples and the other side of the rotational pole. The elevation of the contact in this one image can be seen in Figure 10a to be roughly −4600 m. This is at a lower elevation than exposures at the cap edge in the vicinity of 180°W. It could be either that the platy unit is thinner here or that the basement topography is lower. Chasma Boreale provides an opportunity to probe the location of the bottom of the platy unit. Figures 10b and 10a show the platy unit continuing down to the bottom of the chasma at −4900 m, leaving the question of how deep this layer extends unresolved. Dunes fill the chasam at many locations, and while it is difficult to argue that these dunes must have been derived in situ, it is certainly consistent. The chasma contains two distinct heads (see Figure 1b); the eastern head scarp of the chasma at the higher elevation contains the contact (Figure 10a). The western head scarp also shows an exposure of the platy unit (Figure 10b); however, it appears that the upper finely layered unit has been completely stripped off in this location. The eastern head scarp is cut off from the rest of the chasm by the western head scarp, a sharp cliff with slopes well above the angle of repose, so it is reasonable to assume that the dune material in the portion of the chasam between the eastern and western head scarps must have been derived in situ and have not migrated in from the outside.
 Troughs further poleward are higher in the stratigraphic sequence of layers, and so the lower platy unit is not exposed. These toughs do not have any associated dunes, and so the finely layered unit has been ruled out as a significant source of dark sand-sized material.
5. Discussion and Conclusions
 Observations of a distinct change in layering style that is evident over at least half the northern polar cap at a nearly constant elevation have been presented. Exposures of the lower platy unit are associated with the presence of nearby dune fields. This leads to the conclusion that the dune material is being derived from the erosion of the lower platy unit.
 The elevation of the lowest observed outcrop of the platy unit is −4900 m, although it may extend deeper. This is only ∼200 m above the lowest terrain surrounding the polar cap. The basement topography under the cap is largely dependent on whether the cap is compensated or not [Johnson et al., 2000]. However, it is likely that the basement cannot be more than a few hundred meters below the lowest observed occurrence of the platy unit. The highest observed occurrence of this unit in the same region (head of Chasma Boreale) is at −4600 m elevation, implying that the platy unit is at least a few hundred meters thick in this locality.
 We are of the opinion that this unit is rich in, and perhaps entirely composed of, frozen sand. Owing to the difficulties in codepositing an ice-dust mixture with large amounts of sand [Herkenhoff and Murray, 1990], it seems likely that this unit was deposited in a very different environment than that needed to deposit the upper unit or that of present-day Mars. The Olympia Planitia dunefield, which sits atop the Olympia lobe, ranges in elevation from roughly −4200 to −4700 meters. This is the same elevation range over which exposures of the sand-rich deposit are seen. We suggest that the Olympia lobe is composed of the same sand-rich, platy unit which extends under the north polar layered deposits. The Olympia Planitia dunefield and the rest of the circumpolar erg could then be composed of material eroded from this platy unit. This would vastly increase the minimum size of the platy unit initially estimated in section 3.
 A possible geologic history is outlined in Figure 11, which represents a cross section through the polar deposits from 75°N, 0°W (left) through the pole to 75°N, 180°W (right). A large sand or sand-rich deposit is postulated to have formed before the existence of the present-day north polar cap. It may be that sand-sized particles migrated to this locality through aeolian action from elsewhere on the planet, and the broad topographic low centered roughly on the north pole seemingly served as an effective sink for this material. From mapping the contact elevations in Chasma Boreale and the arcuate scarps at the cap edge, the center (thickest part) of the deposit is interpreted to be offset from the rotational pole along the 180°W meridian. This interpretation is dependent on having reasonably flat basement topography. A shift in environmental or climatic conditions caused the present ice cap to begin forming. This may be related to some climatic change triggered by a chaotic obliquity swing [Touma and Wisdom, 1993] or perhaps even to the end of large-scale volcanism on the planet. The accepted mechanism of deposition of ice with varying concentrations of dust could create the finely layered ice cap. This icy cap may have been much larger in the past and only more recently retreated to its current extent [Fishbaugh and Head, 2000]. This climatic shift need not have been sudden; a relatively shallow covering of ice could stabilize any dunes, thus allowing the rest of the cap some time before starting to form. Finally, at some point in the past, Chasma Boreale and the spiral troughs formed, cutting through the cap into the platy deposit underneath (where the cap is thin enough). A highly porous sand deposit could have stored the large amounts of groundwater needed for the catastrophic outburst theorized to have formed Chasma Boreale. Groundwater in the platy unit would not be molten under present conditions, and so this mechanism would require some form of heat injection such as that theorized by Benito et al.  or much higher planetary heat flow in the past. In this scenario, since the icy portion of the cap formed symmetrically about the rotational pole, the upper finely layered unit drapes completely over the lower unit in the vicinity of 0°W (left side of Figure 11). For this reason no exposures of the platy unit are expected to be observed at the edge of the cap in this region.
 Accumulation of the platy unit may have occurred over a considerable period of time with many episodes of dune migration involved. Modest obliquity changes could have mildly affected polar climate and insured that each episode of dune deposition could contain differing quantities of ice, leading to stratigraphic variations in erosional resistance. Incorporation of bright dust in such units can be ruled out, however, as saltating particles would kick dust back into suspension [Herkenhoff and Murray, 1990]. The irregular outcrops of the platy unit are consistent with laterally variable resistance to erosion, perhaps due to variable dune thickness. A large enduring change in mean obliquity, as modeled by Touma and Wisdom , could have ended this depositional period and begun that of the overlying ice-dust mixture of the northern polar layered deposits.
 If Figure 11 is an accurate representation of the current situation, the volume of material contained within the platy unit can be approximated. Taking it to be a section of a sphere with a radius of 540 km and height of 600 m, the volume occupied is 0.27 million cubic kilometers. For comparison the volume of the total north polar cap is 1.2–1.7 million cubic kilometers [Zuber et al., 1998], and the volume of sand in the present-day erg has been estimated at 1158 cubic kilometers [Lancaster and Greeley, 1990]. Even assuming a large percentage of pore space (which is presumably filled with ice), e.g., 50%, the volume of rock eroded to produce this material is on the order of 105 km3.
 At the cap edge in the vicinity of 180°W (right side of Figure 11) the Olympia Planitia dunefield could have once extended farther equatorward. However, since it lacks the protective cover of the ice cap, it would have been subject to erosion and would have possibly liberated large quantities of sand. This sand could be that which forms the remainder of the circumpolar erg. It is difficult to estimate a timescale over which this process would occur since it is an open question as to how much of the year these dunes are free to saltate [Ward and Doyle, 1983]. Previous wind regime studies [Thomas and Gierasch, 1995] concluded that the current circumpolar erg is confined to a narrow latitude band by winds created in part by its own low albedo. Thus this dark, sand-sized material would not be expected to have been redistributed over a large fraction of the planet.
 The major implication of this newly recognized deposit is climatic in nature. The fact that the icy part of the northern polar cap was absent at some point indicates a much warmer polar environment than at present. The northern cap is also the largest known reservoir of water on the planet. Where this water was in the past if not locked up in the polar cap is a puzzling mystery. It is possible that the water was concentrated at lower latitudes filling craters and is responsible for the apparent sedimentary exposures reported by Malin and Edgett . A secondary implication is that no exotic polar dust aggregates are needed to explain how sand can possibly be weathering out of the current layered deposits. The low thermal inertia results [Herkenhoff and Vasavada, 1999; Paige et al., 1994] still prove to be an unresolved issue though.
 High-resolution Thermal Emission Imaging System (THEMIS) data from the upcoming Odyssey mission will likely be able to definitively say whether the layered unit in question has the same infrared spectral signature as the circumpolar erg. These data may also resolve the difference in temperature between the “hot” and “cold” sides of the dunes, leading to a more accurate value of thermal inertia. The near-complete 20 m per pixel multiband coverage expected from the THEMIS visible camera will be used to search for exposures of this contact in areas where MOC has not imaged and look for characteristic color ratios unique to each unit.
 We are especially grateful to the MOLA and MOC teams for providing their high-quality data in such a prompt manner. We would like to especially thank Anton Ivanov (master of all things MOLA) for his help in acquiring MOLA data in an accessible form and Lori Fenton for her help in understanding dunes and all the cool stuff they can do. We would also like to thank Arden Albee, Andrew Ingersoll, Mark Richardson and Ashwin Vasavada for their comments (and encouragement). Finally we thank Ken Herkenhoff for his comments as a reviewer; the paper is very much improved as a result.