The authors are indebted to L. Dupeyrat, W.K. Hartmann and R. Schultz and two anonymous reviewers for helpful discussions and careful comments about the manuscript.
1The surface of Mars is covered by aeolian dust that forms thick and widespread deposits that are consolidated over time by weathering and diagenesis (Kahn et al., 1992; Christensen et al., 1992). Dust on Mars is transported by regional and/or global dust storms. The grain size of Martian dust, measured by photometric data, is less than 1 mm in diameter (Kahn et al., 1992). Therefore, dust on Mars is a very thin material, even thinner than the loess on Earth. At present, dust accumulates preferentially at the polar ice caps and at high elevations where saltation is reduced, likely forming mantles up to a few meters thick (Arvidson et al., 1989; Greeley et al., 1992). The geographic distribution of the dust mantle for the top decimeter was mapped with thermal inertia cartography obtained by the Viking Infrared Thermal Mapper (IRTM) and the MGS Thermal Emission Spectrometer (TES) (Christensen, 1986; Jakosky et al., 2000). Low thermal inertia regions interpreted as dust deposits were found in areas such as Tharsis Montes, Amazonis Planitia, Elysium Mons, and Arabia Terra.
2Dust often buries small impact craters and as a result modifies their distribution. For example, the image in figure 1 displays very few craters. Among these craters only the smallest are fresh and recent (f), whereas (ld) indicates a low degradation, (sd) a strong degradation, and (g) a ghost crater, i.e. a crater completely filled which is almost not identifiable. The different stages represent the progressive infill of craters by dust, with the smallest craters being the quickest buried.
3The study presented here aims to determine the modification of small impact crater distributions using Mars Observer Camera (MOC) images in order to provide mapping of dust mantle thickness. The method described in this manuscript was applied to the region of Arabia Terra where the observed dust mantle has been well documented since the Viking missions and later Mars Global Surveyor (e.g., Zimbelman and Greeley, 1982; McEwen et al., 1988; Edgett and Malin, 2000; Edgett, 2002). In Arabia Terra, it seems likely that other external phenomena did not influence the aeolian deposits during recent geological times, as no evidence of recent volcanic or hydrologic activity is superimposed on the area (Greeley and Guest, 1987).
Fig. 1 – MOC image R1100390 (26.5°E, 27.6°N) in Arabia Terra with examples of more or less degraded craters.
Fig. 1 – Image MOC R1100390 (26,5°E, 27,6°N) montrant la région d’Arabia Terra avec quelques exemples de cratères plus ou moins dégradés.
Crater (f) is fresh, crater (ld) displays low degradation, crater (sd) is strongly degraded, and crater (g) is a ghost crater almost fully obliterated. North is up.
Le cratère (f) est frais, le cratère (ld) montre une faible dégradation, le cratère (sd) montre une forte dégradation, et (g) est un cratère presque totalement recouvert, souvent nommé « cratère fantôme ». Le Nord est en haut.
4The study area of Arabia Terra extends over 6.5 million km2 (0° to 30°N in latitude and 300° to 360°W in longitude, fig. 2 and fig. 3) in the Martian highlands, with elevations ranging from -2 to 4 km. Arabia Terra was mainly shaped by thermal inertia below 250 tiu (tiu refers to the thermal inertia unit: Jm-2 K-1 s-1/2). Soils with values <250 tiu are generally interpreted as being composed of dust or indurated dust in the top ten centimeters (e.g., Putzig et al., 2005). In this study, dust thickness was estimated and mapped from crater distributions observed in MOC and Viking images for 94 sites in Arabia Terra.
Fig. 2 – Location of the studied area and an example of a crater count.
Fig. 2 – Localisation de la zone d’étude et exemple de comptage de cratères.
(a) Location of the study region in Arabia Terra on a MGS MOC Wide Angle photo mosaic map of Mars with cylindrical projection (NASA/JPL/MSSS). (b) Example of crater counts in the Arabia Terra region on a 3m per pixel MOC image (M1000717, 23.22°N, 308.24°W, NASA/JPL/MSSS) and Viking images. Counts are plotted in an incremental logarithmic diagram containing W.K. Hartmann isochrons (1999; dashed lines) revised by B.A. Ivanov (2001) with uncertainty in the range of a factor of two (Hartmann and Neukum, 2001). This diagram contains the distribution curve for saturated surfaces (Hartmann, 1984; black solid curve) with an uncertainty about plus or minus a factor of two (Hartmann and Gaskell, 1997) (gray area around black solid curve). The site has a crater distribution shape and surface morphology representative of Arabia Terra. The counted surface is of 30 km2. Error regarding the number of crater counts is represented by bars surrounding the plots. Craters in the 125-177m bin to the 250-350m bin follow isochrons, but craters smaller than 125 m show a flat distribution, a turndown typical of modification by aeolian processes. D indicates the diameter of this turndown which is used as the critical size for craters not obliterated by dust mantle. The depletion of craters becomes more and more important for small crater diameters due to their quicker obliteration. (c) Part of the counted area on the MOC image M1000717 displays a smooth surface with poor cratering typical of aeolian mantle.
(a) Localisation de la zone d’étude d’Arabia Terra sur la mosaïque grand angle MOC. (b) Exemple d’un comptage de cratère dans Arabia Terra sur une image MOC (M1000717, 23.22N, 308.24W, NASA/JPL/MSSS) et des images Viking. Les comptages sont représentés sur le diagramme des isochrones de W.K. Hartmann (1999; lignes pointillées) révisées par Ivanov (2001) avec une incertitude d’un facteur deux (Hartmann et Neukum, 2001). Ce diagramme contient aussi la distribution des surfaces saturées en cratères (courbe pleine). La surface comptée est de 30 km2. Les barres d’erreur du comptage de cratères figurent autour des points de mesures et sont calculées par écart-type (racine carré de la densité de cratère divisée par la surface). Les résultats montrent que les cratères entre les intervalles de 125-177 m et 250-350 m suivent une isochrone, tandis que la distribution des cratères inférieurs à 125 m s’en écarte. Ce changement de pente est typique d’une modification par des processus éoliens. Le diamètre critique D indiqué par la flèche correspond à la première classe de diamètre de cratères dont la densité n’a pas été affectée par le recouvrement éolien, et donc celle qui servira pour l’estimation de son épaisseur. En effet, les cratères inférieurs à cette taille critique ont été touchés par le recouvrement éolien : la courbe marque un déficit de ces cratères, sinon la courbe continuerait de suivre un isochrone; ce déficit croissant pour les petits cratères signifie que les petits cratères ont été oblitérés d’autant plus qu’ils sont petits. (c) Zoom sur l’image MOC utilisée en b) montrant peu de cratères en surface et une surface lisse typique d’un recouvrement éolien fin et continu.
Fig. 3 – Thermal inertia and hydrogen map in the studied region of Arabia Terra.
Fig. 3 –Inertie thermique et distribution de l’hydrogène dans la région d’étude d’Arabia Terra.
(a) A map of thermal inertia in the Arabia Terra region as obtained from the N.E. Putzig et al. (2005) dataset. Dark blue regions display low inertia typical of dust mantle. (b) Map of the same area of Mars with the water equivalent abundance of hydrogen in weight percent, as published by W.C. Feldman et al. (2004) using the Neutron Spectrometer onboard Mars Odyssey.
(a) Carte de l’inertie thermique d’Arabia Terra d’après N.E. Putzig et al. (2005). Les régions en bleu sombre correspondent à de faibles inerties typiques de recouvrement éolien. (b) Carte de la même zone de la Mars montrant la teneur d’hydrogène en proportion massique équivalent d’eau, comme publiée dans W.C. Feldman et al.(2004) en utilisant le Spectromètre Neutron de Mars Odyssey.
5Arabia Terra is also an interesting region when one considers hydrogen concentrations as revealed by the Neutron Spectrometer onboard Mars Odyssey (e.g., Feldman et al., 2004). These data are obtained by the orbital detection of neutrons ejected by atomic interactions between cosmic rays and hydrogen atoms from the Martian surface. Neutron data provide input for a map of water-equivalent hydrogen concentrations within approximately one meter down the surface. Three major reservoirs can be readily distinguished (Feldman et al., 2004). Two of the reservoirs, which have relatively rich deposits of water-equivalent hydrogen (>20%), surround the poles. The third reservoir, of which Arabia Terra is a part, is an equatorial/mid latitude reservoir with a range of water-equivalent hydrogen reaching locally 12%. Since free water should not exist at the Martian surface in equatorial latitudes, questions remain regarding whether water is present in minerals of the bedrock, in minerals contained in aeolian deposits, or as adsorbed water originating from atmospheric deposition (e.g., Feldman et al., 2005). Therefore, understanding aeolian material thickness is helpful for determining hydrogen sources.
6An understanding of the main sinks of Martian dust is also an important parameter for climate models and for understanding the role of dust in the atmospheric circulation. Dust thickness is currently only estimated from thermal infrared data in the top ten centimeters of the regolith. Therefore, large regions determined to be “dusty” using thermal IR data can result from the recent deposition of dust deposits less than 1 m thick, whereas other regions with similar inertia may be a sink for dust for billions of years with thickness over 100 m. In the results that follow, we will show that Arabia Terra is likely a main sink for dust on a geological scale, and plays a role in global estimates of total dust volume.
7Distinct Martian chronological models based on impact crater distribution, which were obtained from Viking and Mariner 9 imagery, were proposed by several authors including L.A. Soderblom et al. (1974), G. Neukum and D.U. Wise(1976), W.K. Hartmann et al. (1981), and G. Neukum and K. Hiller(1981). In this study, we use distribution curves plotted with a logarithmic incremental diagram, as established in the last Hartmann’s chronological model (Hartmann, 1999), and revised by B.A. Ivanov(2001). Because cumulative plots tend to hide subtle variations in the distribution (Hartmann and Neukum, 2001), we use an incremental. Impact crater distributions were obtained with statistical counts of visible impact craters on MOC and Viking images. We developed a computer program to help to determine crater counts. We plotted three points on the rim of each visible crater, and calculated and classified the diameter of each crater. Obtained data were then plotted with a Hartmann logarithmic incremental histogram, which showed the number of impact craters per km2 versus their diameters (fig. 2b). Diameter precision was determined to be approximately one pixel for each impact crater. Below six pixels in diameter it became difficult to identify impact craters, thus explaining an impact crater deficit for craters with diameters less than or equal to 6 pixels (e.g., a crater diameter of 18 m on a MOC image of 3m per pixel, or a crater diameter of 1.5 km on a Viking image of 250 m per pixel). Due to this limitation, craters less than or equal to 6 pixels were not taken into account in crater count curves. The large difference in resolution between MOC and Viking images also created a gap for craters approximately 500 m in diameter (fig. 2b), despite the fact that high resolution Viking (typically 40-50 m per pixel) images were used when available. In future studies, visible THEMIS (Thermal Emission Infrared Spectrometer) images (19 m per pixel) and HRSC (High Resolution Stereo Camera) images (12 m per pixel) will significantly contribute to fill this gap. Here we focus only on MOC imagery and Viking imagery.
8W.K. Hartmann (1999) proposed theoretical distribution curves or “isochrons” for Martian terrain of specified ages not modified by surface processes. A plot of a crater density with bins of increasing diameter size can provide the age of the surface. For a fresh surface, not influenced by erosion, deposition, or crater degradation, crater density should follow isochrons corresponding to the age of the formation of the terrain regardless of the range of the chosen crater diameter. The interpretation of these curves is more complex when surface processes occur, as is the case for most of the surface of Mars. When surface modifications do occur, the age is referred to as the “crater retention age”, since the determined age can correspond to a degradation process rather than to a formation age. When influenced by aeolian processes, the smallest impact craters are the first to be buried. Therefore, the curve does not follow the isochrons for all crater diameters, and shows the turn down of small craters, which plot cross isochrons toward younger ages. As a consequence, the diameter of the largest completely buried craters can be obtained by comparing the distribution of measured craters and the theoretical Hartmann isochrons (revised by B.A. Ivanov, 2001). Buried craters are also usually buried by dust, but not by enough dust to be obliterated, whereas craters from the diameter bin below that critical value display a small depletion, identified by the departure of the plot from the isochron (see the example in next section).
9The critical diameter allowed us to estimate the average of the minimum aeolian dust thickness covering the surface around the crater using the geometric relationship between the impact crater diameter (D) and the rim height (H; Garvin et al., 2002), as described in relationship (1) below:
10H=0.07D0.52 (for D < 7km) (1)
11Rim height (H) is proposed as an appropriate measurement for the minimum dust thickness (T) needed to completely cover and obliterate the impact crater, assuming that the crater was not previously eroded and that aeolian erosion did not occur in the past (fig. 4). The relationship (1) only provides heights for diameters below 7 km. All crater sizes obtained during this study were below 7 km, enabling us to use this single relationship. The method provides a measurement for the minimum dust infill in the plains. The infill inside the craters can be much higher due to the lack of erosion. The exact thickness cannot be estimated by the method and requires a more detailed procedure as proposed by N.K. Forsberg-Taylor et al. (2004). The described method is also subject to a series of approximations. Error bars, usually obtained from the root square of the crater density divided by the area (e.g., Hartmann et al., 1981), were used for crater counts, and did not modify the critical diameter obtained. The use of the critical diameter depends upon the bin ranges used for crater counts, which decrease by factors of √2 (e.g. from the bin (352-500 m), to the bin (250-352 m). Thus, a difference of √2 in the value of this diameter would result from a doubt regarding the bin diameter of this critical thickness. As an example, the difference between a critical diameter taken at 500 m, or from the bin below at 352 m, corresponds to a final difference in thickness of 20% using the relationship (1). Therefore, this method should be used as a rough estimate, with error of approximately 20 of the value of the individual thickness extracted. The number of thickness values extracted from individual images, however, was large enough to provide a good statistical sampling in the studied region and allowed us to establish a regional thickness, as well as to distinguish regional variations.
Fig. 4 – Dust thickness (T) overlaying a crater of diameter (D).
Fig. 4 – Epaisseur de poussière T recouvrant un cratère de diamètre D.
The rim height (H) gives an estimate of the minimum dust thickness needed to completely cover and obliterate a crater.
L’épaisseur du rempart H fournit une estimation minimale de l’épaisseur de poussière nécessaire pour remplir et oblitérer le cratère de la surface.
12The study area in Arabia Terra exhibits smooth morphological surface characteristics typical of an aeolian dust mantle (fig. 1 and fig. 2c). Figure 2b shows results for crater counts on a site located at approximately 23°N and 308°W, with a close-up of the image in figure 2c. This site is representative of Arabia Terra surface morphologies, as well as for the shape of its crater distribution. The distribution for large craters counted on Viking images (>1 km in diameter) does not fit with the Hartmann isochrons. A major surficial process seems to modify the distribution curve for large craters, which was described and interpreted as hydrologic and/or volcanic activity in previous works (e.g., Hartmann, 1973). This process may have stopped or declined approximately 3 to 1 Gy ago. Craters from 125 to 500 m in the MOC image M1000717 follow isochrons pretty well over three bin sizes, indicating that this crater size has not been modified for ≈1 Gy. In contrast, a depletion of small craters occurs between 22 m and 177 m (fig. 2b) implying a process of obliteration that modifies the distribution of small impact craters. This depletion of small craters was interpreted to result from obliteration by aeolian deposits observed at the surface, since no evidence of aeolian erosion (such as yardangs) or volcanic and hydrologic activity exists in all of the studied deposits in the last 3 Gy. Determined ages suggest a nearly continuous process between about 1 My and 1 Gy (Amazonian epoch), confirming this interpretation. The largest craters completely buried by dust are in the diameter range of 125 m to 177 m wide, as indicated by the arrow in figure 2b. Taking the mean of the diameter bin at 151 m, and using the relation (Eq. 1), this method gives for this example an estimated thickness of roughly 26 m for this location.
13We estimated a minimum dust thickness using the largest buried crater diameter for all 94 MOC images (fig. 5a). Many images (34 from the 94) did not provide a real minimum because the largest craters at the MOC scale did not reach a point for which the depletion of small craters begins, at the difference of the 177 m for figure 1b, due to the resolution gap. Therefore, for these 34 images the effective value of the minimum thicknesses is underestimated. In order to address this issue, crater distribution curves were extrapolated using a linear regression between MOC and Viking counts when necessary. Among the 94 sites, 60 distribution curves provided reasonable linear regression to obtain their diameters. We use these 60, better constrained curves, to provide a map of minimum thickness on Arabia Terra by interpolating the 60 data points (fig. 5b). Using this method, minimum dust thickness varied from 10 m to 60 m. In more than half of the studied areas, the thickness was found to be 20 m to 40 m. The thickest dust accumulation is located in the Cassini crater where it reached a height of approximately 60 m (dashed circle in figure 5b). We observed an increasing gradient of dust thickness from the edge of the studied region into Cassini area and slight regional variations. Aeolian mantle seems present all over the entire study area, in contrast to previous studies, which identified an Arabia mantle only in the northern and western sides (Greeley and Guest, 1987; Schultz and Lutz, 1988; dashed line, fig. 5a).
Fig. 5 – Maps of dust thickness throughout Arabia Terra.
Fig. 5 – Cartes de l’épaisseur de poussière dans Arabia Terra.
(a) The distribution of minimum dust thickness in meters estimated locally at the 94 studied sites in the region of Arabia Terra. The dashed line shows the maximum southern extent of the Arabia mantle of P.H. Schultz and A.B. Lutz (1988). (b) Interpolation map of the estimated dust thickness using the 60 best crater plots. The interpolation is based on a local smoothing technique with polynomial regression (IDL based). The dashed circle is the location of the Cassini crater. The strong increase or decrease at the edges of the studied area (especially in the northern and the southern parts) is an edge effect of the program (no measurements as seen from dots reported on the map).
(a) Distribution de l’épaisseur minimum de poussière en mètres estimée localement en 94 endroits de la région d’Arabia Terra. La ligne pointillée indique la délimitation antérieure de la zone de couverture éolienne établie par P.H. Schultz et A.B. Lutz (1988). (b) Interpolation des épaisseurs estimées en utilisant les 60 données les plus fiables. L’interpolation est basée sur une technique de lissage par régression polynomiale (dans le logiciel IDL). Le cercle pointillé est le cratère d’impact Cassini. Les fortes valeurs en bordures sont des effets de bord dus au programme (aucun point de mesures).
14Crater counts on Viking images in different parts of Arabia Terra suggest that its mantle is one of the oldest geological units on Mars, dating to the Noachian to Hesperian epochs (around 3 Gy) (Greeley and Guest, 1987). Those previous studies interpreted the age of the Arabia mantle without evaluating the difference between buried formations and surficial ones because Mariner and Viking data did not have sufficient resolution to study small craters (<1 km in diameter). Schultz and Lutz(1988) estimated that the mantle is up to 1 km thick. However, their data were determined using Viking scale images where only large structures are observed, and where surficial deposits are not significant. Our study, which is based on high resolution MOC images analysis, shows that an Amazonian dust mantle, tens of meter thick, covers the older geologic formations previously described in Arabia Terra. This dust deposit is estimated to be 1 Gy old at a maximum (plus or minus a factor of 2). Dust accumulation was still active 1 My ago, and is possibly still active at present time, which explains the low measured thermal inertia (Christensen, 1986).
15Regional variations in dust thickness apparently exist. For instance, dust accumulation between Syrtis Major, on the eastern side of the study area, and in the center of Arabia Terra increases westward. Such a finding may be consistent with the observation of more aeolian erosion to the east, as revealed by wind streaks in the Syrtis Major region. The same gradient has been observed between the center of the study area and the two regions of Acidalia Planitia and Terra Meridiani, to the west and the southwest, respectively, of the study area. Such variations are not visible in the thermal inertia map, which is not surprising since it cannot account for thickness >1 m.
16Estimates for Arabia dust thicknesses are interesting in regards to Neutron Spectrometer data interpretation. Regions with water equivalent hydrogen abundance higher than 8% are all contained in regions of dust mantle >20 m thick. We observed differences in maximum hydrogen concentrations and dust mantle: the regional maximum of dust is located close to the Cassini crater, whereas the maximum abundance of hydrogen is located close to the equator. Due to edge effects, we cannot know the dust mantle thickness in this location. Nevertheless, due to the one meter limit of penetration of the neutron spectroscopy, our results show that the detected hydrogen is likely contained inside the >20 m thick dust mantle for most of the Arabia Terra region. Therefore, the presence of the hydrogen detected in hydrated minerals within the bedrock (such as hydrated sulfates or clays) is unlikely. Based on our dust mantle thickness map, the hydrogen should be present either as hydrated minerals present in dust, or as adsorbed water deposited with dust during last excursions of Mars’ rotational axis at high obliquity (Feldman et al., 2005). The first hypothesis, hydrated minerals in dust, is not consistent with near infrared spectral data, which does not reveal any hydration in the surface of these dust deposits (e.g., Poulet et al., 2007). Nevertheless, this does not rule out possible hydrated minerals buried beneath the first centimeter of desiccated ground that is not accessible by near infrared imagery. One critical issue regarding the second hypothesis, that is the presence of hydrogen as water in dust, is the high hydrogen content south of the equator, south of the low inertia region (fig. 3b). If hydrogen is present in the dust, this could mean that a dust mantle is also present south of the equator, contrary to thermal inertia values. However, a more indurate dust mantle, as seen from thermal inertia values >200 tiu might be possible, since this region may be an older part of the dust deposits studied in our work. For future work, this region south of Arabia may be a good test region for this hypothesis, as well as for further investigations using the method developed in this study coupled with THEMIS or HRSC images.
17This study presents a method for obtaining estimates of dust thickness over the Martian surface using the distribution of small impact craters (<1 km) observed in high resolution Mars Orbiter Camera (MOC) images. Using the method, the distribution of small craters was found to be different from the theoretical distribution due to the meteoritic impact flux since many craters are progressively buried by dust deposits. The rim height of the largest buried impact crater was determined to approximate the minimum dust thickness blanketing the surface. The method was applied to the region of Arabia Terra, which appears to be covered by a dust mantle with a minimum thickness of 20 m. The results show that Arabia Terra was a regional sink for dust during most of the Amazonian epoch. The results also show that the hydrogen present in Neutron Spectrometer data should belong to the dust and not the underlying bedrock. Further improvements in the mapping of aeolian dust can be accomplished in the future with additional crater counts and new data from THEMIS imagery and Mars Express high resolution images. Regional maps of dust thickness will improve our knowledge of the Martian surface and aeolian dynamics controlled by climatic changes during its evolution.