Martian gullies are found on steep slopes of all origins, on all sorts of terrains of all ages, scattered across nearly all of Mars. Gullies are observed on all manner of substrates (layered, massive, shattered, rubble), with or without nearby mantling deposits. Gullies are most common in the southern midlatitudes but also occur in the northern hemisphere, in near polar terrain, on equatorial volcanoes, and on northern plains. Most gullies in the southern hemisphere are on south-facing slopes, but they occur on slopes of all orientations. Gullies are among the youngest features on Mars but locally are overlain by eolian deposits and cut by faults. Old or eroded gullies are rare, and those found have been partially stripped from slopes, leaving no rock debris behind. Most gully deposits contain no detectable rocks. These data are inconsistent with published hypotheses of gully formation, including seeps and breakouts of water or brine, hydrothermal activity, cryovolcanism, and breakouts from liquid carbon dioxide. The data are consistent with gullies being dry flows of eolian material (dust and silt), comparable to climax snow avalanches on Earth. Eolian sedimentation should be correlated little with underlying geology: cause of slope, age of terrain, type of terrain, or the nature of the rocks. Eolian sedimentation should be correlated with wind deceleration (which will cause suspended sediment to drop), and areas with common gullies are those with strong wind deceleration (predicted by global circulation model). In such areas, sediment will be deposited preferentially in the lee of obstacles; for the gully-rich areas of the southern midlatitudes, winds blow from the NNW, so that sediment is deposited on SSE-facing slopes (i.e., poleward). These predictions are in accord with observations.
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.
Malin and Edgett  reported the existence of gullies on Martian slopes, and inferred that they formed from water-rich flows of rock debris. Their report was widely hailed as proof of the abundance of water on Mars, that it was active recently, and that it is easily available for living organisms (indigenous Martian or human). However, liquid water is not thermodynamically stable at or near the Martian surface because present-day temperatures and atmospheric water vapor pressure are too for low pure liquid water to have formed or persist [Ingersoll, 1970]. And only in a few areas at the surface of Mars (the bottom of the Hellas basin) is the pressure high enough to stabilize liquid water at all, if it were protected from exchange with the atmosphere. These problems with production and retention of liquid water engendered several theoretical models of how it could be produced and retained long enough via: obliquity variations, igneous heating near the surface, geothermal heat at depth, and abundance of dissolved salts. Other workers, recognizing the difficulties with liquid water, suggested that liquid carbon dioxide could have produced the gullies, and invoked creative scenarios to produce and stabilize it near Mars' surface.
 These models relied almost entirely on the geological data of Malin and Edgett , which is (of necessity) terse and limited. Malin and Edgett  gave more geological data, but important issues were not addressed. Thus, this work was begun in order to explore the geological settings and relations of the Martian gullies.
 As work progressed, it was apparent that the geologic relations of the gullies were difficult to reconcile with origins from either liquid water or liquid carbon dioxide. Another hypothesis was needed. The observed similarities between Martian gullies and massive snow avalanches suggested that they could have formed similarly - as flows of dry granular material. On Mars, the flows would likely be of dust and silt, deposited by wind.
1.1. Philosophical Approach
 This work is principally a compilation of observations, dictated by the imagery available and the geology they show. Its inferences rely mostly on geomorphology and careful analogies with terrestrial geology and laboratory experiments, and it does not develop theoretical or mechanical models. Gullies were identified solely by the morphologic criteria given below, as given by Malin and Edgett [2000, 2001].
 The intended scope of this work is synoptic - to consider all gullies on Mars as manifestations of a single general process. Although a few gullies in a single location may show a particular relationship clearly, the relationship will be generalized as representative of all gullies. As an example, the best example of eroded gully deposits is on the central peak of the Li Fan impact crater. This erosion style is inferred to be representative of all gully deposits, not only those on central peaks nor only those in Li Fan.
 Narrow angle images from the Mars Observer Camera (MOC) were obtained from the Internet as distributed from the MOC operator, Malin Space Science Systems, through the NASA Planetary Data System (PDS). The images included all those taken through August 1, 2001. Several thousand such images were examined for gully features; in some regions (e.g., 155W 30S to 180W 60S), every narrow angle image was examined. Images are accompanied by coordinate and orientation information, which were verified against context images and maps of Mars. Scenes shown in the figures are subimages of the full MOC frames, taken from “view” jpeg images provided on the Internet site. These images are not calibrated radiometrically nor geometrically. The uncalibrated images are adequate here, because neither quantitative radiometric nor geometric inferences are made.
Figures 1 through 9 show areas of MOC images, selected to show specific features of Martian gullies. Each images has been rotated from its original orientation so that North is approximately to the top, except as noted. Brightness and contrast were adjusted as needed for display.
3. Geology of Gullies
3.1. Defining Characteristics
 Martian gullies include three distinct, connected landforms [Malin and Edgett, 2000]: alcoves, channels, and depositional aprons, here called cones (Figures 1a and 1b). Alcoves are depressions into the slope, generally constricting downhill, where slope material appears to have been removed. Alcoves can be large or small compared to the channels and deposits (Figures 1a and 1b), show a range of tapering (nearly straight up the hill to widely flaring [Malin and Edgett, 2000, Figures 3C and 3F]), and be filled or not with later material [Malin and Edgett, 2000, Figure 3D]. Channels extend from the bottom of the alcove down slope, and are incised into slope materials. They commonly have distinct raised levees and a sinusoidal planform, which probably represents meanders. The channels incise into, debouch into, and gradually vanish into their deposits, which form a tapered “cone” of material extending downslope. Depositional cones commonly have distinct raised margins, which can be rounded or lobate planform. Commonly, the depositional cones extend beyond the gully slope onto the adjacent flat plains.
3.2. Gully Ages
 The ages of gullies are a critical constraint on their origin. Gullies are among the youngest features on Mars: they have sharply defined topography and albedo contrasts [Malin and Edgett, 2001, Figure 102]; they are superposed on other young surfaces (e.g., eolian landforms [Malin and Edgett, 2000, Figure 9; Malin and Edgett, 2001, Figure 101]); and they are rarely superposed by other structures, like impact craters [Malin and Edgett, 2000, 2001]. The absence of old gullies, despite the observation that they form on surfaces of all ages (described below) is puzzling, and two explanations seem possible. First, gullies appear to be young because they are all young - gully-forming processes did not operate in Mars' distant past. Second, gullies appear to be young because older examples are rapidly eroded away - despite the tediously slow average erosion rate on Mars [Golombek and Bridges, 2000].
 Available MOC images favor the latter hypothesis - gully landforms did form in Mars' past, and can be eroded away rapidly. In fact, gully formation appears to be an ongoing process, concurrent with recent eolian processes that form dunes and mantle deposits. Rare MOC images show partially eroded gully forms.
 In nearly all areas, gullies do appear to be the youngest geological landforms. However, gully formation itself spans some time, and other landforms are rarely superposed on gully deposits. Evidence that gully formation spans some time includes: gully debris cones on earlier cones (Figures 1a and 2a) [see Malin and Edgett, 2001, Figure 103]; gully alcoves transected by other alcoves (Figure 2b); debris cones transected by alcoves (Figure 2c); and fractures that cut some gully deposits and channels, and are overlain by others (Figures 1a and 2a). Malin and Edgett [2000, 2001] have emphasized that gully landforms overly (and so are younger than) other landforms. However, gully forms are uncommonly overlain by, or are superposed by, other features: dark sand dunes (Figure 3a), light dunes (Figure 3b), featureless mantling material (Figure 3c), and crater fill material (Figure 3d).
 Having shown that gully formation spans some time, and is contemporaneous with other geological processes (Figures 2 and 3), one can ask where have all the gullies gone? In careful examination of MOC images, a few partially eroded gullies have been found (Figure 4). In these cases, gully deposits and their channels are partially stripped off a surface as a layer. In some cases, the leveed meanders of an original gully channel remain standing (Figures 4a and 4b). In other cases, the whole gullied slope is partially stripped, leaving remnant layers with gully landforms in swales and channels (Figures 4c).
3.3. Rarity of Rocks
 The gully landforms are transportational features - material was transported from high on slopes toward their bottoms. It is important to know what kind of material was transported in the gullies, but that knowledge is limited by the spatial resolution of the MOC images (no better than 1.7 meters/pixel, more commonly 3.4–5 meters/pixel) and the lack of spectral information (only relative albedo in a single channel). Despite these limitations, the gully material has been inferred to consist primarily of rock debris. “[A gully channel]…transports the accumulated debris downslope out into the apron as a slurry flow of ice, liquid and rock debris” [Malin and Edgett, 2000, p. 2334]. “…rapid vaporization of the liquid CO2 with entrainment of rock and clathrate-hydrate ice…” [Musselwhite et al., 2001, p. 1283]. “…the gullies were probably created by debris flows of liquid H2O mixed with rocks and residual water ice” [Costard et al., 2002, p. 110]. No direct evidence has been offered about the nature of the transported material, and data available suggests that rock is a minor component of the gully debris.
 First, identifiable rocks are uncommon in the debris cone deposits of gullies. Distinct rocks are abundant in the alcove regions of many gullies, but are never observed to be abundant in the deposits, and are commonly absent (Figure 5) [Malin and Edgett, 2001, Figure 102]. One should expect many rocks and boulders in gully deposits, given the proposed violence of gully initiation [Malin and Edgett, 2000; Musselwhite et al., 2001; Costard et al., 2002]. However, rocks are rarely visible and are not abundant when seen (Figure 5c) [Wynn-Williams et al., 2001, Figure 4]. Similarly, rock is generally darker than surrounding surficial materials, and so a rock-bearing gully deposit would be darker than it surroundings (Figure 5d). This evidence is ambiguous, however, because smaller boulders might not be imaged by MOC, and large boulders might be buried under finer-grained material.
 More telling, gully deposits appear to erode easily and leave no detectable residue behind (described above, see Figure 4). There are few signs that the gully deposits were eroded by vigorous processes (e.g., floods, or landslides) so the absence of rocks and boulders where gully deposits have been eroded implies that none were present initially.
 In summary, there is no direct evidence that gully deposits contain rocks in any abundance. Gullies do not resemble rockfalls, and show no evidence in their structure, strength, or albedo for a significant proportion of rocks. Of course, rocks smaller than about 5 meters cannot be detected in available imagery.
3.4. Local Geology
 Many factors of the local geology have been implicated in the formation of gullies: e.g., subsurface layering, competent bedrock, mantling deposits, and age of local rocks. Clearly, a steep slope seems required for formation of gullies, but it is important to consider: the distribution of gullies across Martian geologic and chronological units; whether gullies are related to the origin of the slope they occur on; whether layered rocks are required; and whether there are consistent relationships between the elevations of alcoves and the surrounding geology.
 The elevations of alcoves and alcove heads, relative to surrounding topography, have figured prominently in some hypotheses of gully formation. In many areas, gully alcoves consistently head at cliff-like layers (e.g., Figures 1b, 6b, and 9b) [Malin and Edgett, 2000, Figures 3B, 3E, 3F, and 5; Malin and Edgett, 2001, Figures 99 and 104]. However, in areas where layering is absent or weak, gullies start below cliffs and steep pitches of any sort (Figures 8a, 8b, and 9a). Similarly, transitions from gully alcoves to channeled debris cones can be observed at several elevations (i.e., different layers) on slopes (Figure 8b). The common factor here is not layering but steep walls. Rock with alternating strong and weak layers tends to erode into cliffs [Howard and Selby, 1994], and this process alone could explain the statistical association of gully forms with layered bedrock.
 In summary, no geologic characteristic except a steep slope is required for gully formation. Gullies have developed on geologic units of all ages and all origins, of a wide range of physical characteristics, and of all available types of layering. The cause of the steep slope also does not control gully formation. Nor is there a necessary relationship between the elevations of gully alcoves and local geology. Thus, gully formation appears independent of local geology, and its formation mechanisms cannot rely on specifics or particulars of local geology.
3.5. Slope Orientation
 Gullies are reported to be more abundant on pole-facing slopes: ∼71% of gullies counted by Malin and Edgett  were on such slopes. Several of the hypotheses for gully formation have been tailored to fit pole-facing slopes, couched in terms solar heating on those slopes during the current or a past obliquity of Mars' rotation axis [Costard et al., 2002]. Study of thousands of MOC images confirms that gullies are more abundant on pole-facing slopes, and that many lie on slopes with other orientations. In several locations, gullies occur on adjacent slopes facing in several directions (Figure 9). In these cases, it is clear that slope orientation is not an important factor in gully formation. Thus, the relation between gully formation and slope angle cannot be so regular and unchanging as solar insolation, but must be subject to caprices that are not wholly dependent rotation axis and obliquity.
4. Other Hypotheses
 Hypotheses offered to explain the Martian gullies have involved liquid water or liquid carbon dioxide, and have invoked special or specific geological circumstances to emplace either liquid near Mars' surface. However, the required special circumstances are not fulfilled by many gullies, so the hypotheses (at least as general explanations) are put into doubt. It appears that none of the published hypotheses for gully formation is consistent with the geological constraints developed above.
4.1. Groundwater Seeps and Eruptions
 In their original description of the gullies, Malin and Edgett  suggested that they were formed by seepage of groundwater because comparable gully landforms are found on Earth where water-laden debris flows down steep slopes [e.g., Lee et al., 2002; Hartmann et al., 2002]. Groundwater is highly problematic so near the Martian surface. In most areas and at most times, the surface and near subsurface is too cold to permit liquid water [Kieffer et al., 1977]. Similarly, the atmospheric pressure is too low to allow liquid water to persist indefinitely at or near the surface [Ingersoll, 1970]. Thus, Malin and Edgett  had to appeal to special circumstances for the accumulation and seepage of groundwater.
 They proposed that groundwater aquifers were present in layered bedrock. The water was concentrated in porous layers, which were sealed against evaporation to the atmosphere by overlying impermeable layers (aquicludes). The water flowed in the porous layers to cliff faces, undermining them and producing piles of debris at their bases. The water collected in or near the debris behind an ice barrier, through which it eventually burst to initiate water-lubricated debris flows. Thus, gully landforms start at the bases of cliffs. Gullies were restricted to pole-facing slopes at midlatitudes because only those slopes received the right amount of solar heat to permit production of liquid water. Equator-facing and low-latitude slopes presumably receive too much solar heat, causing the water present to evaporate or sublime into the atmosphere. High-latitude slopes remain too cold for production of liquid water.
 Several aspects of this hypothesis are inconsistent with observed geology. First, many gullies formed on unlayered rocks, both massive and rubbly (Figures 4a, 4b, 7b, 7c, 8a, and 9a). It is not clear how aquifers in such rock could be restricted to limited elevations or horizons. Nor is it clear how aquifers could be formed in some gully settings, like isolated knobs, central peaks of impact craters, and the crests of impact basin walls (Figures 4a, 4b, 6c, 7b, and 8a). Second, it is not obvious that impermeable rock layers (aquicludes) could be present near impact craters (e.g., Figures 1–4), which are typically surrounded by broken and pulverized rock [Melosh, 1989]. Third, the inference that the gully deposits are rock-filled is inconsistent with evidence above that rocks are uncommon (e.g., Figure 5). And fourth, more than a quarter of gullies occur on found on slopes that do not face poleward or otherwise would not have the proper solar heating (Figure 9) [Malin and Edgett, 2000]. Significant here are the gullies in the calderas of the large Martian volcanoes (Olympus, Pavonis, Arsia, and Ascreus Mons; Figure 6b), which are near-equatorial and so receive equal heating on all slopes. Also significant are gullies in the polar pits, which are poleward of 72°S (Figure 7a). Only during significant obliquity excursions might these pit walls receive enough solar insolation to permit gully formation by the Malin and Edgett  mechanism [Costard et al., 2002].
 The Malin and Edgett  hypothesis was modeled numerically by Mellon and Phillips , who found that “even under the most optimistic conditions, near surface ground ice cannot reach the melting point of pure water (273K) by solar heating alone” (p. 23, 171). They proposed a related scenario: water vapor rises from depth, during high-obliquity periods, and condenses as liquid below near-surface aquicludes. There, it is kept from freezing by an overlying, highly insulating layer of eolian dust. As obliquity and solar heating diminish, the water slowly freezes and increases the pressure on the remaining liquid. Now under pressure, that liquid erupts explosively and produces gullies.
 This scenario runs afoul of several geological observations. First, as above, many gullies are not on slopes that would receive adequate solar heating, whether in normal or extreme obliquity conditions. Second, gullies are found commonly without insulating dust layers adjacent (Figures 6c and 7c). And third, many gullies occur on massive or rubbly materials that would not readily contain impermeable layers (Figures 7b, 7c, 8a, and 9a).
4.3. Geothermal Heating
 Recognizing the problems in gully formation via solar heating, Hartmann  proposed that the heat required to make liquid water was geothermal - derived from recent shallow intrusions of magma. Otherwise, his scenario is similar to that of Malin and Edgett . The paucity of gullies on equator-facing slopes is explained by solar heating - equator-facing slopes are warmed so much by the sun that the water evaporates or sublimes, and does not collect to initiate gullies [Malin and Edgett, 2000].
 This scenario is inconsistent with several geological constraints, and a recent geophysical observation. First, most gullies occur in areas with no indications at all of recent magmatic activity (Gorgonium Chaos, Figures 1a and 6d; Noachis, Figures 3a, 3d, 5a, 5b, and 7c). Second, many gullies are found in areas that show indications of ground ice, like on patterned ground (Figure 6e) and in polar pits (Figure 7a). If magma had intruded shallowly in such areas, one might additional signs of magma-water interaction like maar craters, tuff rings, liquefaction surfaces, or catastrophic water outflows; none of these are observed. Third, it is not clear why gully forms should start only on steep slopes if they reflect an internal heat source. Fourth, if shallow igneous intrusions have been so abundant in the recent geological past as to produce gullies across the globe, it would be reasonable to expect at least one such intrusion to still be warm. Yet no warm spots have been reported from the TES instrument on the MGS spacecraft [Christensen et al., 2001], nor so far in preliminary THEMIS instrument images, reports, and press releases.
Gaidos  suggested a different mechanism to avoid the need for liquid water to persist near Mars' surface - that water responsible for gully formation originated at depth and moved rapidly from there to the gully sites. Water is generated at depth by geothermal heat, cools to freezing, and is pressurized (as in the Malin and Edgett  model) by confinement by impervious rock and growing ice plugs. When confinement is breached, the water acts as if it were an igneous body; driven by its own buoyancy, it intrudes upwards as a dike. Reaching its neutral buoyancy level, the “dike” spreads out as a “sill” and “erupts” from steep wall, carrying rock debris downslope.
 This model can explain several geological observations, like the presence of gullies at the calderas of the Tharsis volcanoes (Figure 6b), but is not so helpful with others. It does obviate the need to fill aquifers on isolated peaks (e.g., Figures 4a, 4b, 6c, 8a, and 9a), but does not explain why cryovolcanic eruptions should favor these peaks and not the surrounding lowlands. As with other water-breakout models, it does not explain the absence or paucity of rocks in the gully deposits. Similarly, many gullies formed on unlayered rocks, both massive and rubbly (Figures 7b, 7c, 8a, and 9a). It is not clear how aquifers in such rock could be restricted to limited elevations or horizons, nor why gullies should occur at several levels in a single wall, nor why water should erupt at restricted horizons from the broken and shattered rock on the walls of impact craters.
4.5. Brine Seeps
 Recognizing the difficulties in production and maintenance of pure water near Mars' surface, three groups suggested independently that gullies might be formed by brines [Doran and Forman, 2000; Knauth et al., 2000; Wynn-Williams et al., 2001; Knauth and Burt, 2002]. Groundwaters that interact extensively with rock become briny solutions, which can have freezing temperatures as low as ∼200K and very low vapor pressures. Such brines, trapped in near-surface aquifers, could migrate to cliff walls, and there produce gullies as was suggested for pure water [Malin and Edgett, 2000].
 Invoking concentrated brines rather than relatively pure water avoids many problems of the Malin and Edgett  mechanism, notably the need for temperatures and pressures above ambient. Some issues remain unresolved. First, as with most other mechanisms, concentrated brine scenarios do not explain the paucity of rocks in gully deposits. Second, these scenarios do not explain the presence of gullies in areas without layered bedrock (Figures 7b, 7c, 8a, and 9a), nor near impact craters where rock is likely to be broken and pulverized [Melosh, 1989]. Similarly, the scenarios do not explain how water, briny or not, came to be sequestered near the tops of isolated peaks (Figures 4a, 4b, 6c, 7b, and 8a). It is reasonable that groundwater would evolve to become concentrated brine solutions, but this fact does not seem to explain all the geological features of the gullies.
4.6. Liquid Carbon Dioxide Eruptions
Musselwhite et al. , Draper et al. , and Hoffman [2000, 2001] suggested that the fluid responsible for the gully outflows was not water but carbon dioxide liquid and vapor. In their hypotheses, carbon dioxide from the atmosphere or at depth is concentrated or sequestered in porous rock layers. The CO2 would be trapped above and below by impermeable rock layers. As temperatures cool, CO2 ice forms nearest the ground surface and seals the impermeable layer with an ice plug. Compressed by this growing plug and by overlying rock, pressure increases to the point of liquefaction. Gullies are initiated when an impermeable layer fails, and allows the CO2 liquid to flow toward the ground surface and flash-vaporize into gas. This gas cloud is able to suspend rock debris and form a density flow, comparable to an ash-flow (nuee ardent) from a silicic volcano.
 These hypotheses invoking CO2 liquid in the Martian subsurface have been criticized on theoretical grounds by Stewart and Nimmo . They concluded that neither CO2 nor CO2-bearing clathrates could accumulate in sufficient quantities, nor over sufficient areas, to support the formation of individual gullies nor the concentrations of gullies observed in some areas.
 Several geological constraints are also at variance with the CO2 liquid hypothesis, in part because its physical mechanism of gully formation is similar to that of the liquid water hypotheses. Many gullies formed on unlayered rocks, both massive and rubbly (Figures 7b, 7c, 8a, and 9a), and it is not clear how CO2 could be trapped, liquefied, and retained in these systems. Second, it not obvious that impermeable rock layers could be present near impact craters (e.g., Figures 1–4), which are typically surrounded by broken and pulverized rock [Melosh, 1989]. Third, the inference that the gully deposits are rock-filled is inconsistent with evidence above that rocks are uncommon (Figure 5). Thus, liquid CO2 is as unlikely as liquid water to have been the cause of Martian gullies.
5. Gullies As Large Granular Flows
 The geologic relations and spatial distribution of gullies are explained inadequately by available hypotheses, so other hypotheses should be considered. Here, it is suggested that the Martian gullies may not be related to liquids of any sort, but are massive avalanches of fine granular material - e.g., dust to sand-sized particles.
 For this hypothesis to be taken seriously, several point need to be demonstrated. It should be shown that: (1) the Martian gully deposits are composed principally of fine granular material, such as could be transported and deposited by wind; (2) that similar granular material can and does accumulate where gullies have formed; (3) that the spatial distribution of gullies is related to Martian wind patterns; (4) that wind patterns predict the lack of relationships between gully presence and the underlying geology or the cause of the slope; (5) that flows of dry granules can behave as if they were liquids; and (6) that flows of dry granules can produce landforms like those of the gullies. These issues are considered in turn here, and all shown true or plausible.
5.1. Fine Granular Material
 Here, it is shown that the gully deposits are fine granular materials, eroded by wind, similar in texture and color to eolian mantle deposits and bright dunes, and available near the gullies. Gully deposits contain few if any rocks or boulders because: they rarely contain discernable rocks (Figure 5), erode away without leaving a lag of discernable rocks (Figure 4), and are not so dark colored as rocks in the overlying cliffs.
 The most likely agent in the erosion of most gully deposits is wind. As shown above, gully deposits can be eroded (see above, Figure 4), leaving no obvious remnant debris, no signs of channels or other transportation conduits, and no signs of collapse to the subsurface. Thus, one cannot invoke mass wasting, flow of liquids, nor subsidence as erosive mechanisms. Only the mechanism of wind abrasion remains, and other effects of wind are abundant in and around gullies and their host slopes. If wind abrasion is the erosion mechanism, then the particles eroded from the gully deposits are likely to be silt- or dust-sized (e.g.,< ∼100 μm, Greeley et al. ).
 The gully deposits are similar in color (i.e., brightness in MOC imagery) to the light-colored dune and mantling material found across Mars. The similarity in color is evident qualitatively on many images that show both gully forms and dunes or mantles above or below the gullies (Figures 5a,b,d, 6a,b,d,f, 7a; Figures 3A, 4 of Malin and Edgett ; Figures 99, 101, 102A, 104 of Malin and Edgett ). This proposed similarity has not been tested quantitatively, given the lacks of firm radiometric calibration of MOC images [Malin and Edgett, 2001], and of a complete phase function for mantle and bright dune materials.
 Mechanical properties of the gully deposits are also similar to those of the mantling deposits and bright dunes, at least so far as can be discerned in MOC imagery. Mantling deposits are easily eroded, as shown by the common presence of deflation pits. Gully deposits are also easily eroded, as described above. Mantling deposits and bright dunes are mechanically weak, as shown by their rounded topography compared to rugged topography developed on bedrock (Figures 5a, 5b, 5d, and 7c) [Malin and Edgett, 2000, Figure 4; Malin and Edgett, 2001, Figures 99 and 101]. Gully debris shows evidence for similar weakness - commonly it is cut by slump blocks and by graben perpendicular to the steepest slope (fall line), suggesting that the debris is weak enough to slide downhill and break apart readily (Figure 2a).
 Finally, light-colored dune and mantling materials are present and common near gullies and in their alcoves, and so are available to participate in gully formation. A few of the many images that show dune and mantling material near gullies include Figures 5a, 5b, and 5d [Malin and Edgett, 2000, Figures 3A and 4; Malin and Edgett, 2001, Figures 99, 101, 102A, and 104].
 Thus, by weathering style, similarity, and proximity, the material of mantling deposits and bright dunes is a reasonable candidate for the material transported downhill to form gully deposits. The mantling and dune material was deposited by wind in and near the gully alcoves, and then flowed downhill through the alcoves to produce the gully channel and debris cone.
5.2. Spatial Distribution
 The spatial distribution of gullies is critical for earlier hypotheses of gully formation: more abundant in southern hemisphere than the northern, most abundant in midlatitudes, concentrated in distinct regions within those latitude belts, and predominantly on pole-facing slopes [Malin and Edgett, 2000, Figure 1; Malin and Edgett, 2001, Figure 100]. Examination of additional MOC images confirms this general pattern, but also reveals many gullies that are exceptions (e.g., Figure 6b). Thus far, the spatial pattern has been explained in terms of solar insolation and temperature. However, the pattern can also be explained by Martian surface winds, as predicted by the Geophysical Fluid Dynamics Laboratory global circulation model [Fenton and Richardson, 2001].
 If the Martian gullies represent eolian sediment accumulated on slopes (and then avalanching down), one should look for areas where sediment deposition is most likely, i.e., where wind decelerates significantly. By the GFDL-GCM maps, the greatest wind decelerations are in the southern midlatitudes, from ∼30°–50°S [Fenton and Richardson, 2001, Figure 2], coincident with the belt of most abundant gullies [Malin and Edgett, 2001, Figure 100]. Within this belt, the area of greatest predicted deceleration is between 120°W and 180°W, the Gorgonium/Newton region where gullies are most abundant [Malin and Edgett, 2001, Figure 100]. In the northern hemisphere, the area of greatest wind deceleration is east of Tempe Terra and in western Chryse, one of the concentrations of gullies [Malin and Edgett, 2001, Figure 100].
 Gullies are more abundant on pole-facing slopes [Malin and Edgett, 2000, 2001], and this observation is readily explained in terms of the CGM wind field. According to the GCM, Mars' winds blow into southern-hemisphere gully belt from the north-north-west (on average). When sediment-laden wind passes an obstruction (like an impact crater) it will tend to drop the sediment in the lee of the obstruction. The lee sides of southern-hemisphere craters, the north-north-west slopes, are also the pole-facing slopes. Local irregularities in the wind flow, for instance in complex systems of mountains or troughs, may yield sediment deposits on slopes of many orientations, and may produce gullies on all the slopes (Figure 9). The wind pattern in the northern hemisphere is much less consistent than in the southern [Fenton and Richardson, 2001], and it is observed that northern hemisphere gullies show much less regularity in their orientation.
 Thus, the spatial distribution of gullies across Mars is broadly consistent with eolian sedimentation as inferred from the Global Circulation Model. However, the correspondence is not perfect, and suggests areas for future research. Gullies near the Elysium Montes (30°N, 190–240°W; Figure 7b) and in the south peri-polar pits (∼72°S, 0°W; e.g. Figure 7a) are not in mapped zones of wind deceleration, and some zones of wind deceleration are not associated with gullies (e.g., NE of Syrtis). Of course, the rarity of steep slopes in the northern lowlands does contribute to the rarity of gully features there.
5.3. Local Geology
 Above, it was shown that the presence and orientation of gullies is unrelated to local geology, including: ages of surrounding land surface, the type of surface (e.g., volcanic versus heavily cratered), the origin of the slope on which the gully developed (e.g., crater walls versus sinuous channels), and layering properties of the material in which the slope is developed. Models that invoke liquid water or carbon dioxide in the subsurface do not predict this lack of correlation between gullies and local geology.
 However, the dry avalanche hypothesis is consistent with a lack of relation between gullies and underlying geology. It predicts that gullies can develop wherever eolian sediments are deposited on steep slopes. Eolian sedimentation is related superficially to local geology - only in the sense that geology influences land forms, which can serve as sources or sinks for sediment. Thus, one should expect sediment (dust) to fall where it may, and show limited preferences for local geology. This lack of relationship is exactly as observed.
5.4. Dry Granular Flows
 The idea that gullies formed via water-rich debris flows [Malin and Edgett, 2000, 2001] is an inference from geomorphology to rheology to composition - water-rich debris flows have the proper fluid properties to excavate alcoves, carve leveed channels, and produce deposits with sharp margins [Johnson, 1984; A. M. Johnson and R. W. Fleming, Landslides, flowslides and mudflows, 2001, available at http://www.eas.purdue.edu/engeomap/1_introduction.htm (hereinafter referred to as Johnson and Fleming, 2001)]. However, water-rich debris flows are not required - any fluid with a similar rheology will do.
 The rheological characteristics of water-rich debris flows have been determined in laboratory experiments and inferred from field observations [Johnson, 1984]. Broadly speaking, debris flows act as Bingham fluids: finite yield strength and a (relatively) constant Bingham viscosity once flowing [Johnson, 1984; Johnson and Fleming, 2001]. In detail, their flow laws are significantly more complex [Bulmer et al., 2002; Johnson and Fleming, 2001], but the complexities do not affect these arguments.
 Dry granular flows generally also have Bingham rheologies, with finite yield strength, for the same reason as do wet debris flows. For a stationary mass of particles (wet or dry) to come into motion, its particles must be able to move past each other by moving out of each other's ways (dilatation [Bagnold, 1966]). This motion increases the volume of the mass, and so (with gravity) implies an activation volume or energy for flow; this activation volume translates to a yield strength. In a very rapid granular flow, individual grains saltate - they bounce off each other and fly until the next collision [Campbell, 1990; Drake, 1990]. This motion can be facilitated by entrained gas and by acoustic energy [Melosh, 1983; Druitt, 1998; Draper et al., 2000].
 Large flows of dry granular materials are uncommon on Earth, in part because water is so abundant. Dry, wind-deposited sediment can yield dry granular flows. The largest such known on Earth, the 1920 landslide in Gansu, China [Bolt et al., 1975; Derbyshire et al., 2000], was activated by a nearby earthquake. Dry masses of large rock particles can also flow as rock avalanches, sturzstroms [Hsü, 1975]. Dry flows of snow can range from small sluffs (comparable to slope streaks on Mars [Sullivan et al., 2001]) to catastrophic avalanches (e.g., Figure 10a) [McClung and Schaerer, 1993]. Pyroclastic flows, like nuee ardents, consist of volcanic ejecta and hot gas, and can be considered dry granular flows [Druitt, 1998].
5.5. Geomorphology of Granular Flows
 Dry granular flows can produce landforms with the characteristics of Martian gullies, and the closest terrestrial analogs are “climax” snow avalanches. In climax avalanches, a significant portion (or all) of the snowpack moves and flows down mountain slopes as a granular mass [McClung and Schaerer, 1993]. Liquid water is not volumetrically important in most snow avalanches, but is present in most as intergranular films. Climax snow avalanches commonly release to a headscarp at the top of an alcove in the mountainside, and the flow is commonly funneled into a gap below the alcove (Figure 10a) [Perla, 1978, Figure 10; McClung and Schaerer, 1993, Figures 4.29 and 4.38]. The snow is deposited near the slope base in steep cones (Figure 10a) or as lobes [McClung and Schaerer, 1993, Figure 4.39]. The deposits commonly have sharp margins, reflecting the Bingham rheology of the flow (Figure 10b) [e.g., Sullivan et al., 2001, Figure 11; McClung and Schaerer, 1993, Figure 8.1]. Channels can develop in the deposits (Figure 10a), and the channels can be leveed (Figures 10a and 10b) and can meander [McClung and Schaerer, 1993, Figure 5.30]. It should be admitted that no dry snow avalanches (in images available to the author) show the deep entrenched meanders of the Martian gullies (e.g., Figures 1a, 5c, and 5d). Yet, formation of meanders requires only flowing fluid and erodable walls [e.g., Sparks, 1986] so that meandering channels should be expected.
 Other types of dry granular flows show some resemblances to the Martian gullies. Pyroclastic volcanic flows obviously cannot have alcoves and headscarps like gullies, but their deposits can show sharp raised margins, digitate toe lobes, and distributary channels [Druitt, 1998, Figure 6a; see Hoffman, 2000]. Similarly, rock avalanches commonly leave alcove-like scars at their sources, and their deposits can be leveed and show sharp margins [Blair, 1999].
 The salient features of the Martian gullies [Malin and Edgett, 2000, 2001] are consistent with their origin as dry flows of eolian sediment: gully deposits are fine granular material (erodable by wind); eolian sediment are available where gullies form; the distribution of gullies are consistent with deposition of sediment from wind; and the orientations of gullies are similarly consistent with wind patterns. Further, it is clear that granular materials can flow as if they were Bingham liquids, and granular flows can produce landforms with all of the geomorphic features of Martian gullies. No known data concerning the gullies (chronological, geomorphic, or geologic) falsify this hypothesis, so it is worth further investigation.
 The morphology and distribution of gullies on Martian slopes have been attributed to the action of liquid water at the surface and in the near subsurface [Malin and Edgett, 2000]. However, evidence for involvement of water was circumstantial and by analogy with Earthly landforms that are mediated by liquid water. Production and retention of liquid water is difficult in the cold and low pressure of Mars' near-surface environment, and several ingenious explanations have been offered [e.g., Malin and Edgett, 2000; Knauth et al., 2000; Gaidos, 2001; Hartmann, 2001; Mellon and Phillips, 2001]. Other students forswore water in favor of liquid carbon dioxide, despite the known difficulties of producing and retaining it in Mars' near-surface environment [Hoffman, 2000; Musselwhite et al., 2001]. In either case, the gullies were interpreted as signs that Martian geology and history are far different than had been inferred from earlier investigations.
 Here, it is shown that dry flows of eolian dust or silt can have all of the geological features observed in the gullies, and that their spatial distribution on Mars is consistent with atmosphere circulation models. Many details of this hypothesis remain to be investigated, but it is wholly consistent with earlier inferences interpretations of the present-day Martian environment: no liquid water, and dominated by eolian processes.
 I thank D. Clifford, D. Draper, W, Kiefer, and D. Musselwhite for their assistance and helpful suggestions. R. Day and H.E.F. Amundsen, to whom I am grateful, provided the images of Figure 10, for this paper only. This work would not have been possible without MOC images, made available through the NASA PDS (Planetary Data System), with instant access provided by the U.S. Geological Survey Internet site http://ida.wr.usgs.gov/. It would also have been impossible without the continuing efforts of Malin Space Science Systems in camera targeting and data processing. The suggestions of S. Clifford, W. Kiefer, T. Perron, and S.H. Williams were most tolerant and helpful. Supported by contract funds of the Lunar and Planetary Institute, which was operated by USRA under NASA Contract NASW 4574. Lunar and Planetary Institute contribution 1147.