Research Article

Evidence for Recent Groundwater Seepage and Surface Runoff on Mars

See allHide authors and affiliations

Science  30 Jun 2000:
Vol. 288, Issue 5475, pp. 2330-2335
DOI: 10.1126/science.288.5475.2330


Relatively young landforms on Mars, seen in high-resolution images acquired by the Mars Global Surveyor Mars Orbiter Camera since March 1999, suggest the presence of sources of liquid water at shallow depths beneath the martian surface. Found at middle and high martian latitudes (particularly in the southern hemisphere), gullies within the walls of a very small number of impact craters, south polar pits, and two of the larger martian valleys display geomorphic features that can be explained by processes associated with groundwater seepage and surface runoff. The relative youth of the landforms is indicated by the superposition of the gullies on otherwise geologically young surfaces and by the absence of superimposed landforms or cross-cutting features, including impact craters, small polygons, and eolian dunes. The limited size and geographic distribution of the features argue for constrained source reservoirs.

Present-day Mars is a desert world on which liquid water is not likely to be found at the surface, because average temperatures are below 273 K and atmospheric pressures are at or below water's triple-point vapor pressure of 6.1 mbar. However, in 1972 the Mariner 9 orbiter photographed evidence—in the form of giant flood channels and arborescent networks of small valleys—that liquid water (1) might have been stable in the surface environment at some time in the past (2–8). Analysis of Mars 4 and 5 (1974) and Viking orbiter images (1976–80) contributed to these views (9–14), as did observations of flood terrain by Mars Pathfinder in 1997 (15–17).

The Mars Global Surveyor (MGS) orbiter reached the planet in 1997 (18). One of the most important early results of the Mars Orbiter Camera (MOC) investigation was the absence of evidence for precipitation-fed overland flow. For example, there are no contributory rills, gullies, and/or small channels associated with the martian valley networks (19–21). MOC images instead suggest that most of the form and relief of the valley networks was created by collapse along paths dictated by slope and structure, but without obvious gradient continuity at the surface. It is possible—perhaps even likely—that surficial mantling and/or erosional processes have hidden or erased conduits and other surface manifestations of overland flow. Whatever the explanation for the absence of these features, the possibility that liquid water flowed across the martian surface in a sizeable volume for an extended period of time, and especially in the recent past, seems quite remote.

However, equally early in the MGS mission, MOC observations hinted at a more complex story regarding groundwater seepage. A few impact crater walls seen in images with spatial resolution of ∼20 m/pixel displayed features resembling morphologies that, had they occurred on Earth, might be attributed to fluid seepage (22). Here we report on new high-resolution (2 to 8 m/pixel) images that more strongly indicate the occurrence of seepage-fed surface runoff.


Since the start of the mapping phase of the MGS mission (18), the MOC has acquired >20,000 images at scales of 1.5 to 12.0 m/pixel. The images cover from <1 km to a maximum of 3 km across, with lengths from a few to several hundred kilometers. With its body fixed to the spacecraft and without any independent means to point, the camera acquires nadir observations of selected features that pass below it (23).

As of January 2000, ∼150 MOC images of ∼120 locales (Fig. 1) show what we interpret as the results of fluid seepage and surface runoff. Three attributes of the observed landforms—the head alcove (representing a “source” morphology), the main and secondary channels (defining transportation conduits), and the depositional apron (Fig. 2)—are consistent with fluid-mobilized mass movement processes.

Figure 1

Simple cylindrical equal angular map of Mars (centered on 0° latitude, 0° longitude) showing locations of gully landforms as found in images acquired through the end of January 2000.

Figure 2

Example of landforms that comprise the martian gullies. The three main geomorphic elements of these features are a theater-shaped alcove that tapers downslope, one or more major or secondary channels that continue downslope from the distal apex of the alcove, and triangular aprons that broaden downslope and appear to be material deposited after transport through the channel system. Illuminated from the upper left, this is a subframe of MOC image M03-00537, located near 54.8°S, 342.5°W.

Geographic relations. Of the observed locales, about one-third are on interior walls or central peaks of impact craters, a quarter are on walls of distinctive pits in the south polar region, and a fifth are on walls of two major martian valley systems: Nirgal Vallis and Dao Vallis. Slightly less than 50% of all occurrences are found on south-facing slopes and ∼20% face north. Features occur about two and a half times more often on poleward-facing slopes in a given hemisphere than they do on equator-facing slopes. More than 90% occur south of the equator, and with the exception of Nirgal Vallis (located at 27° to 30°S latitude), all are poleward of 30° latitude. Concentrations are found along the two valleys, in the polar pits, in craters and troughs associated with Gorgonum Chaos, and within the craters Hale, Maunder, Newton, and Rabe (Fig. 1).

Head alcove. The head alcove consists of the head- and sidewall escarpments, located below the brink of a slope, and the region of disrupted topography bounded by these scarps (Fig. 2). Alcoves are theater-shaped depressions within the wall or slope on which they have formed. Often, colluvium appears to have collected at the base of these escarpments, forming a wedge of debris climbing part of the way back up the head- and sidewall slopes. The bounding escarpments and mantling colluvium display rills with orientations that generally parallel slopes and gravity. Irregular mounds within the colluvium suggest that slumping and other, more surficial, mass movements contribute material from the slopes onto the floor of the alcove. In some areas, particularly south of 60°S, alcoves are commonly littered with large boulders (several meters to several decameters in diameter). The derivation of these boulders appears to be characteristic of the surface layer (or a near-surface layer) unique to this area. Rills in the colluvium contribute, across the floor of the alcove, toward the head of the main channel.

Four types of alcoves have been recognized: lengthened, widened, occupied, and abbreviated (Fig. 3). Lengthened alcoves are longer (downslope) than they are wide and commonly have subsidiary or contributing secondary alcoves or chutes (Fig. 3A). In the case of the polar pits, the abundance and distribution of elongated alcoves produce a “badlands” morphology (Fig. 3B). Widened alcoves are broad in transverse dimension (cross-slope) and can consist of more than one smaller alcove focusing to a single downslope outlet (Fig. 3C). Occupied alcoves are filled or partly filled with material that might be mobile as a unit, but not as disaggregated debris. The type examples of this form occur throughout the length of Dao Vallis (Fig. 3D). Abbreviated or attenuated alcoves are typical of areas of strong topographic or stratigraphic control of landform location. For example, in areas where a resistant rock unit creates an overhang, the alcove may be small or even missing (Fig. 3E). Other occurrences include some impact crater inner walls (Fig. 3F), where particularly steep topography may contribute to rapid downslope transport of material and upslope propagation of the landform.

Figure 3

Examples of alcoves. Lengthened alcoves (A) are longer (downslope) than they are wide; where they occur in close proximity, they form characteristic badlands (B). Widened alcoves (C) are broad in transverse dimension and may include more than one smaller alcove. Occupied alcoves (D) are those that appear to be filled with material. Abbreviated alcoves (E and F) show strong topographic or stratigraphic control of landform location, which limits the extent of the alcove. Each picture is a subframe of a MOC image illuminated from the upper left, except for (C), which is illuminated from the upper right, and (F), which is illuminated from the left: (A) image M07-05535 (54.8°S, 342.5°W); (B) image M03-02709 (70.8°S, 355.8°W); (C) image M03-02214 (58.9°S, 336.0°W); (D) image M11-01601 (33.1°S, 266.8°W); (E) image M07-02909 (38.5°S, 171.3°W); and (F) image M11-00944 (41.1°S, 159.8°W).

The overall impression of the uppermost zone of these landforms is that a headland has experienced undermining and collapse across a confined zone, with downslope movement focused at or toward a point, and material moving under the force of gravity downslope toward an apex (the head of the main channel). These characteristics are shared by terrestrial seepage gullies and classic landslides (24–29).

Main and secondary channels. The most striking geomorphic aspect of these landforms is the occurrence of entrenched, steep-walled, V-shaped channels emanating from the downslope apex of the head alcove. In relation to other martian valleys and slope features, these channels are distinctive: They generally start broad and deep at their highest topographic position and taper downslope and distally (Figs. 4 and5). Their courses are influenced by variations or undulations in surface relief and by subsurface processes; a small number are discontinuous (Fig. 6). Lineations adjacent to the main channel suggest secondary channels or alternative courses, now abandoned. The dominance of a single, deep, narrow path suggests that entrenchment mitigates channel migration. The transition between head alcove and channel can be abrupt or extended. In those areas where it is extended, there is in some places good evidence of the contributory nature of the process: “streamlining” behind obstacles, anastomosing patterns, and rapidly decreasing order (Fig. 7).

Figure 4

Channels start broad and deep at their highest topographic position and taper downslope and distally. Many slopes show secondary channels or alternative courses, although one channel from each alcove clearly dominates. Illuminated from the upper left, this is a subframe of MOC image M07-01873, near 37.3°S, 168.0°W.

Figure 5

In addition to contributory upper reaches, channels on aprons display distributary networks. Image and insets (Aand B) are illuminated from the left and are subframes of MOC image M11-00944, located in a crater at 41.1°S, 159.8°W.

Figure 6

Discontinuous reaches of some gullies (arrows) suggest that subsurface flow (or piping) may contribute to gully development. Illuminated from the upper left, this is a subframe of MOC image M03-02709, located at 70.8°S, 355.8°W.

Figure 7

Contributory anastomosing pattern in portions of some gullies suggests the confluence of several streams of flow from the alcove into the main channel. Illuminated from the upper left, this is a subframe of MOC image M07-03298, near 46.9°S, 51.0°W.

Depositional aprons. Surfaces downslope from the distal apex of the alcove display two distinct appearances. In the first of these forms, a crudely triangular zone or apron broadens downslope from the apex and continues to, and in some places beyond, the base of the slope hosting these features (Fig. 2). The surface of this zone may be smooth at the decameter scale, but it displays smaller swells and swales oriented downslope. These surfaces “host” the main and secondary channels, in the sense that the channels appear to cut down into the aprons. Many of the small swells are themselves wedge shaped, with their apices upslope. Swells often occur at the mouths of channels and are highest at the mouth of the channel and along a center line downslope from there, and their relief tapers marginally and distally. No lateral or distal terminal escarpments have been observed.

The second type of downslope pattern consists of distributary networks of small subsidiary channels (Fig. 5) and digitate flows (Fig. 8). Anastomosing channels, piracy, superposition, levees, distal and marginal escarpments, distal thickening, relief transitions from negative (channels) to positive (flows), and other attributes attest that the channels acted as conduits for the transport of material and that some may have been repeatedly occupied.

Figure 8

Example of an apron with digitate flows. Illuminated from the left, this is a subframe of MOC image M04-02479, near 40.0°S, 251.8°W.

Age Relations

The proposed seepage landforms are not cratered: Only one apron in >150 examined has any craters at all visible in images with an average scale of 3 m/pixel. The absence of impact craters suggests that these landforms are geologically young (30). No examples of old, degraded seepage sites have been recognized. Rather, nearly all have other properties that also suggest relative youth, including superposition of aprons on eolian bedforms in Nirgal Vallis (Fig. 9A), superposition on polygonally patterned ground and the absence of rejuvenated polygons (Fig. 9B), and sharply defined topography and albedo patterns within the alcoves of some of the landforms (Fig. 9C). These relations provide evidence that the seepage landforms are quite young, because superposed surfaces also lack impact craters and the superposed landforms are either ephemeral and/or dynamic (31).

Figure 9

Age relations between gullies and other landforms. (A) The best example of an apron that partly covers and obscures eolian bedforms in Nirgal Vallis. Illuminated from the upper left, this is a subframe of MOC image M07-00752, near 29.4°S, 39.1°W; north is toward the upper right. (B) Superposition of apron material on polygonal-patterned ground. Illuminated from the upper left, this is a subframe of MOC image M03-00537; north is toward the upper right. (C) Unmantled, bouldery, and mottled surface of a large alcove in the upper western wall of a crater at 65°S, 15°W. Illuminated from the upper right, this is a subframe of MOC image M11-00530; north is toward the lower right.


The martian landforms resemble terrestrial gullies, which form by a combination of processes, including overland flow, headward sapping (32), debris flow, and other mass movements (33)—all processes that, on Earth, involve the action of water. Stimulated by the problems raised by maintaining water as a liquid in the environments in which these features occur on Mars (34), we have sought evidence of other origins. However, it is unlikely that “dry” mass movement processes (such as granular flows or avalanches fluidized by atmospheric or soil gases) can explain the observed landforms. There are many instances of mass movements on slopes on Mars—indeed, every crater and depression on Mars seen by the MOC or earlier spacecraft exhibit such landforms—but the martian “gullies” described here are rare and easily differentiated from most of the other martian slope landforms. In many cases, the more common forms of mass movements are found on equator-facing slopes of craters that have alcove, channel, and apron landforms on their pole-facing slopes. These other landforms share many characteristics with landforms on the moon and Earth related to dry mass movements, reinforcing the impression that these gullies are different.

Other attributes of these landforms provide evidence for fluid erosion. Head alcoves are often associated with distinct layers within a cliff (e.g., Fig. 3, A and E); where they occur in groups on a wall with a single prominent layer, all the alcoves head within this layer. The distal, tapered portion of the alcoves (and the head of the main channel) occurs at the base of the bed, suggesting that this point is the location where undermining and collapse are initiated and from which material has been transported downslope in a manner that facilitates entrenchment. The main and secondary channels display characteristics of fluid flow, including sinuous paths, branched and/or anastomotic reaches, levees, streamlining, super-elevated banking, and incision. Many aprons show patterns in morphology that mimic those seen on terrestrial alluvial fans, including a filigree of distally diverging medial to marginal lineations (suggesting flow lines) and channels, lobate margins, and distal thinning. Another observation of note is that the channel heads and interiors, where visible, do not appear to be choked or clogged with debris, implying efficient transport throughout the conduit. The correlation between latitude and occurrence of these landforms, and their preponderance on poleward slopes, also suggests that there is a relation between the existence of these landforms and the presence of a volatile, such as water, that responds to solar insolation.

Despite the remote possibility that some unknown fluid might be responsible for these gullies, their attributes are most consistent with the action of water. The absence of geographic association of these features with volcanic terrains (Fig. 1) or obvious thermal sources (35), although at the same time showing latitudinal and azimuthal orientation relations, suggests that surficial environmental factors and not geothermal processes control their development.

Alcoves that are high on the walls of craters (but below the level of the preimpact surface), or high on the walls of grabens or other depressions, suggest that the water involved in creating these gullies must occur at shallow depths beneath the martian surface. Examination of available Mars Orbiter Laser Altimeter data shows that the seepage locations are only a few hundred meters or less below the surface.

The volume of water involved in the formation of the gullies is unknown. Conservative speculative scaling calculations (36) suggest that the formation of each apron may have involved a few thousand cubic meters of water. In some locations, the number of gullies and the area covered by their aprons suggest that the abundance of water might have been greater by a factor of 100 or more. Clusters of sites with limited spatial extent (as seen in grabens near Gorgonum Chaos, within Newton crater, in the pitted plains of the south polar region, and along Nirgal Vallis and Dao Vallis) argue for restricted source regions (reservoirs or aquifers).

A simplistic model of the processes by which these gullies form would involve groundwater moving within and along bedrock layers until it was exposed at the surface within depressions whose walls truncate the beds. Undermining and collapse of a region above the site of seepage would promote headward extension of an alcove, and surface runoff (and perhaps piping within colluvia) would remove debris accumulating at the seepage site. The resulting landform combines elements of a “classic” landslide promoted by groundwater and soil water with elements of surface runoff from a limited groundwater source. The location of the gullies (both in latitude and on nonequator-facing slopes) implies that the gradual percolation of groundwater to the surface and its evaporation as it emerges must be impeded, perhaps by an ice barrier that accumulates over time. In this model, a reservoir of liquid forms beneath this barrier, and the eventual, short-lived outburst from this ephemeral reservoir transports the accumulated debris downslope and out into the apron as a slurry flow of ice, liquid, and rock debris. The limited runout distance at the base of the slopes supports this idea. In this model, higher temperatures (closer to the equator and on equator-facing slopes) quickly evaporate emergent groundwater, preventing the persistence necessary to create gullies. Although the available evidence suggests that the processes that created these landforms acted in the relatively recent past and could even be contemporary, the absence of old, degraded, or cratered examples remains a mystery.


Stay Connected to Science

Navigate This Article