Geomorphology of Ma?adim Vallis, Mars,
and associated paleolake basins
Rossman P. Irwin III
Center for Earth and Planetary Studies, National Air and Space Museum, Smithsonian Institution, Washington, DC, USA
Department of Environmental Sciences, University of Virginia, Charlottesville, Virginia, USA
Alan D. Howard
Department of Environmental Sciences, University of Virginia, Charlottesville, Virginia, USA
Ted A. Maxwell
Center for Earth and Planetary Studies, National Air and Space Museum, Smithsonian Institution, Washington, DC, USA
Received 27 April 2004; revised 7 October 2004; accepted 26 October 2004; published 30 December 2004.
[1] Ma?adim Vallis, one of the largest valleys in the Martian highlands, appears to
have originated by catastrophic overflow of a large paleolake located south of the valley
heads. Ma?adim Vallis debouched to Gusev crater, 900 km to the north, the landing site
for the Spirit Mars Exploration Rover. Support for the paleolake overflow hypothesis
comes from the following characteristics: (1) With a channel width of 3 km at its head,
Ma?adim Vallis originates at two (eastern and western) gaps incised into the divide of the

1.1 M km2 enclosed Eridania head basin, which suggests a lake as the water source.
(2) The sinuous course of Ma?adim Vallis is consistent with overland flow controlled by
preexisting surface topography, and structural control is not evident or required to explain
the valley course. (3) The nearly constant 
5 km width of the inner channel through crater
rim breaches, the anastomosing course of the wide western tributary, the migration of
the inner channel to the outer margins of bends in the valley?s lower reach, a medial
sedimentary bar 
200 m in height, and a step-pool sequence are consistent with modeled
flows of 1?5  106 m3/s. Peak discharges were likely higher but are poorly constrained by
the relict channel geometry. (4) Small direct tributary valleys to Ma?adim Vallis have
convex-up longitudinal profiles, suggesting a hanging relationship to a valley that was
incised quickly relative to the timescales of tributary development. (5) The Eridania basin
had adequate volume between the initial divide and the incised gap elevations to carve
Ma?adim Vallis during a single flood. (6) The Eridania basin is composed of many
overlapping, highly degraded and deeply buried impact craters. The floor materials of the
six largest craters have an unusually high internal relief (
1 km) and slope (
0.5?1.5
)
among degraded Martian craters, which are usually flat-floored. Long-term, fluvial
sediment transport appears to have been inhibited within these craters, and the topography
is inconsistent with basaltic infilling. (7) Fluvial valleys do not dissect the slopes of these
deeper crater floor depressions, unlike similar slopes that are dissected at higher levels in
the watershed. These characteristics (6, 7) suggest that water mantled at least the lower
parts of the Eridania basin floor throughout the period of relatively intense erosion early in
Martian history. The lake level increased and an overflow occurred near the close of
the Noachian (age determined using >5 km crater counts). Initially, the Eridania basin
debouched northward at two locations into the intermediate basin, a highly degraded
impact crater 
500 km in diameter. As this intermediate basin was temporarily filled with
water, erosion took place first along the lower (northern) reach of Ma?adim Vallis,
debouching to Gusev crater. The western overflow point was later abandoned, and erosion
of the intermediate basin interior was concentrated along the eastern pathway. Subsequent
air fall deposition, impact gardening, tectonism, and limited fluvial erosion modified the
Eridania basin region, so evidence for a paleolake is restricted to larger landforms that
could survive post-Noachian degradation processes. INDEX TERMS: 6225 Planetology: Solar
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 109, E12009, doi:10.1029/2004JE002287, 2004
Copyright 2004 by the American Geophysical Union.
0148-0227/04/2004JE002287$09.00
E12009 1 of 33
System Objects: Mars; 5415 Planetology: Solid Surface Planets: Erosion and weathering; 1824 Hydrology:
Geomorphology (1625); 1821 Hydrology: Floods; KEYWORDS: flood, lake, valley networks
Citation: Irwin, R. P., III, A. D. Howard, and T. A. Maxwell (2004), Geomorphology of Ma?adim Vallis, Mars, and associated
paleolake basins, J. Geophys. Res., 109, E12009, doi:10.1029/2004JE002287.
1. Introduction
[2] The discovery of large outflow channels and valley
networks on Mars during the 1971?72 Mariner 9 mission
[McCauley et al., 1972; Milton, 1973] suggested that Mars
once had abundant surface water, much of which is still
present as polar and ground ice (summarized by Carr
[1996]). Pressure-driven groundwater originating from frac-
tures or chaotic terrain appears to have formed most of the
outflow channels in the Chryse region of Mars [Carr, 1979],
but several wide, deeply incised highland valleys do not
exhibit chaotic terrain at their source. Some of these larger
highland valleys occupy ridge-bounded drainage basins and
have tributary networks of small valleys, suggesting forma-
tion by fluvial runoff and/or groundwater sapping [Sharp
and Malin, 1975; Grant, 2000; Irwin and Howard, 2002;
Aharonson et al., 2002; Hynek and Phillips, 2003]. Other
valleys originate at gaps incised into the drainage divides of
large enclosed basins or impact craters, suggesting a water
supply from overflowing paleolakes [Goldspiel and
Squyres, 1991; Cabrol and Grin, 1999; Clifford and Parker,
2001]. The cratered highland landscape was favorable for
ponding of water at many sites, but evidence for overflows
(which would require high water tables or relatively abun-
dant inflows) is less common. Overflow gaps originating
above an undissected basin floor are, however, the most
definitive evidence that paleolakes did occur on Mars,
particularly in basins where other surface features support
this conclusion. Paleolakes could be favorable sites for
hosting and preserving evidence of former life, if it existed
on Mars [e.g., Cabrol and Grin, 1999], so a thorough
characterization of these sites is beneficial for targeting
future human and robotic exploration.
[3] At 8?25 km wide, Ma?adim Vallis in Terra Cimmeria
(Figure 1) is similar in size to the outflow channels, but it
formed earlier than did most of the large valleys on Mars
[Masursky et al., 1977, 1980; Carr and Clow, 1981] and
contemporary with small valley network activity near the
Noachian/Hesperian boundary [Irwin et al., 2002] (
3.7 Ga
before present from Hartmann and Neukum [2001]). We
previously concluded that Ma?adim Vallis was carved by an
overflowing paleolake located immediately south of the
valley head, on the basis of several lines of morphologic
evidence [Irwin et al., 2002] that are detailed and reinforced
with new analysis in this paper. Alternative published
hypotheses involve an outflow or long-term fluvial activity,
perhaps with structural control of groundwater flows, and
these concepts are also addressed here. While other process
models may yet be developed for this or other features in
the Martian landscape, we show that the available data is
consistent with the origin of Ma?adim Vallis by paleolake
overflow, followed by a limited period of fluvial activity to
partially incise its tributary network. We identify the fea-
tures that can be used to identify a source paleolake in this
region, we describe the rapid development of Ma?adim
Vallis during a catastrophic overflow, and we constrain
the magnitude of the flood. The morphology of Ma?adim
Vallis and its tributary valleys is also provided as evidence
for a paleolake overflow. These observations provide an
integrated history of the Eridania basin, Ma?adim Vallis, and
Gusev crater; offering a set of features that may be used to
recognize paleolakes elsewhere on Mars. This study has
direct relevance to the Mars Exploration Rover (MER)
mission, as Gusev crater is the landing site for the Spirit
rover [Golombek et al., 2003; Grant et al., 2004].
1.1. Description and Terminology of the Ma?adim
Vallis System
[4] Ma?adim Vallis is among the largest valleys on Mars,
and it originates full-born, with an upstream inner channel
width nearly equal to its downstream dimensions (use of
the term ??channel?? is justified in section 4.2 in reference
to the wide, commonly flat cross section of the valley
floor). The inner channel is 2.5?4 km wide along a steep
reach near the source, and a width of 3?5 km is common
along lower-gradient reaches downstream. The overall
length of the valley is 920 km, from the source northward
to the Gusev crater rim entrance breach where the valley
ends (Figure 1). The volume of Ma?adim Vallis and its
tributaries, below the sharp break in slope at the top of the
valley sidewalls, is 
14,000 km3 [Goldspiel and Squyres,
1991; Cabrol et al., 1996; Irwin et al., 2002]. Transporting
this volume of sediment would require at least 35,000 km3
of water given Komar?s [1980] maximum possible water/
sediment volume concentration of 0.4, or an order of
magnitude larger water volume for more typical fluvial
sediment concentrations.
[5] Incision of Ma?adim Vallis connected three originally
enclosed basins, (1) the Eridania basin (Figure 2); (2) the
intermediate basin, a 500 km, highly degraded impact crater
(Figure 1); and (3) Gusev crater at the valley terminus, in
addition to several smaller craters that were breached. The
divides of Gusev crater and the Eridania basin are only
partially incised, so volumetrically reduced central basins
still exist. In contrast, the northern divide of the intermediate
basin was downcut to below its floor level, opening the
basin completely for through-flowing drainage. The contig-
uous Ma?adim Vallis watershed network (not counting
valleys in the Eridania basin watershed) occurs within the
intermediate basin and on the northward-sloping intercrater
plains between the intermediate basin and Gusev crater.
[6] Ma?adim Vallis originates 
25 km south of the
breached intermediate basin rim within the enclosed, irreg-
ularly shaped Eridania basin (Figure 1). This large basin
contains many highly degraded impact basins and craters on
an otherwise near-level highland plateau (Figure 2). Within
the eastern part of the Eridania basin, six deep subbasins
(Gorgonum, Atlantis, Ariadnes and three unnamed basins)
are highly degraded impact craters, where large segments of
the crater rims are missing. The plains within these larger
concave subbasins have up to 1.5 km of internal relief,
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
2 of 33
E12009
unlike the nearly flat ridged plains and degraded crater
floors that are common elsewhere on Mars. To the east and
southeast of the Eridania basin, Newton and Copernicus
craters also exhibit concave floors like the larger craters in
the Eridania basin. We refer to the subaerial Eridania basin
watershed separately from the paleolake basin itself. The
watershed (using direct surface flow paths) typically
extends only 10?100 km above the inferred paleolake
Figure 1. Ma?adim Vallis in a cylindrical projection of MOLA topography (128 pixel/degree, 50 m
contour interval). The valley crosscuts a circular, highly degraded impact basin (black circle), identified
as the intermediate basin. The locations of the knickpoint (K), Gusev crater (G), the Gusev exit breach
(E), and chaotic terrain north of the valley terminus (C) are indicated. The lower reach of the valley lies
north of the knickpoint, the intermediate reach extends along the rest of the intermediate basin floor, and
the upper reach occurs along its southern interior slope. Ma?adim Vallis originates full-born at two source
spillways (O1 and O2) in the intermediate basin divide. Another low point (P1) contains a degraded
crater, the rim of which may have formerly raised the divide above the 1250 m level of O2.
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
3 of 33
E12009
F
ig
u
re
2
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
4 of 33
E12009
highstand of 1100 to 1250 m. The Eridania basin and its
watershed have a combined area of 
2.08  106 km2, not
counting adjacent, low relief, topographically isolated crater
or intercrater basins that may have contributed groundwater.
The possible contributing area within a higher hydrologic
divide surrounding the Eridania basin is 
3  106 km2.
[7] The longitudinal profile of Ma?adim Vallis (Figure 3)
is subdivided into high- and low-gradient reaches. From the
Figure 3. (a) Longitudinal profile of the main channel of Ma?adim Vallis from 128 pixel/degree MOLA
topography. The steep upper reach of the channel transitions to the lower-gradient intermediate reach,
where the large tributary meets the main valley. A constriction in the valley cross section and a knickpoint
in the profile mark the location of a possible resistant outcrop. The Gusev crater paleolake had a level of
1,550 m and a depth of 
400 m, suggested by the northern Gusev spillway and the elevation of mesas
that are possible remains of a Ma?adim delta. Chaotic terrain development north of Gusev has obscured
any evidence of flow into the lowlands. (b) Vertical exaggeration of the longitudinal profile of the lower
reach and part of the intermediate reach, showing step-pool topography with pools up to tens of meters
deep. The gray line approximates the inner channel thalweg.
Figure 2. MOLA topography of the Eridania basin and vicinity, in a cylindrical projection bounded by 10
S, 60
S,
140
E, and 210
E. The image is 2964 km from north to south. White lines are major drainage divides that delimit direct
surface flow paths. The paleolake basin that supplied Ma?adim Vallis is contained within interconnected subbasins SW,
NW, E, and EG. The blue line is the 1100 m contour, which approximately delimits the Eridania basin from its watershed.
Around Newton Crater (NC) the blue line is the 1300 m contour, the (possibly unoccupied) overflow level for that basin.
Drainage divides for small crater interiors and neighboring enclosed drainage basins are not shown. The drainage basins
used for hypsometric analyses and crater counts are labeled with black letters: eastern subbasins (E + EG), eastern
subbasin?s Gorgonum Crater subset (EG), northwest subbasin (NW), southwest subbasin (SW), Newton Crater
interior (NC), Newton Crater watershed (ND), Copernicus Crater and watershed (C), Ma?adim Vallis watershed (MV),
and Al-Qahira Vallis watershed (AV, not used for hypsometry). The locations of high-resolution images of the Eridania
basin used in subsequent figures are indicated with white numbers. Also in white numbers are the locations of undissected
low points in the Eridania basin divide, P1 and P2, as discussed in the text.
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
5 of 33
E12009
eastern outlet, the steep upper reach of the valley descends
from 950 m elevation into the intermediate basin interior. On
the intermediate basin floor, the intermediate reach is wider
and has a lower gradient than does the upper reach. This
intermediate reach contains the confluence of the Ma?adim
Vallis main channel and its anastomosing principal tributary,
which originates at the western outlet of the Eridania basin. A
prominent knickpoint with 300 m of relief separates the
intermediate and lower reaches of the valley. As the lower
reach leaves the intermediate basin, Ma?adim Vallis widens
from 
10 to 
20 km, and terraces separate a wider outer
valley from an inner channel that is similar in width to the
intermediate and upper reaches of the valley. Despite local
slope irregularities discussed in section 4.2, the lower reach
maintains a nearly constant overall gradient of 0.0012 (0.07
)
to the 160 km Gusev crater terminal basin, where mesas
(which also record this channel gradient) at1550 m may be
remnants of a former delta [Schneeberger, 1989; Grin and
Cabrol, 1997]. The northern rim of Gusev crater is also
breached, so flows probably continued into the lowlands, but
later chaotic terrain development and volcanism north of the
crater precludes an evaluation of these possible flows.
1.2. Previous Investigations
[8] Ma?adim Vallis was first identified from Mariner 9
imaging in 1972, when it was referred to as the Rasena
Channel, and several early papers considered its origin
and age in the context of more general valley network
studies [Milton, 1973; Sharp and Malin, 1975; Malin,
1976; Masursky et al., 1977, 1980; Carr and Clow, 1981;
Brakenridge, 1990; Carr, 1996]. Sharp and Malin [1975]
provided the first detailed description of the valley and
tributaries, noting the ??irregularly constricted?? reach of the
channel that we now recognize as a knickpoint (Figure 3).
They classified the valley as a runoff channel, which Malin
[1976] supported by noting (as shown in section 3.2) that
Ma?adim Vallis had attacked craters ??through both grazing
and head-on erosion.?? Craddock et al. [1987] used Viking
Orbiter Infrared Thermal Mapper-derived thermal inertias to
conclude that 0.25 to 0.5mm aeolian sands had resurfaced the
floor of Ma?adim Vallis, resulting in low rock abundances of
<5%. This sediment transport was attributed to high-velocity
valley winds [Craddock et al., 1987], which have since been
modeled and correlated with aeolian surface features in
Gusev crater [Greeley et al., 2003].
[9] In a significant advance, Schneeberger [1989] noted
the mesas at the terminus of Ma?adim Vallis, where it
debouches to Gusev crater, and suggested that the mesas
may be eroded remnants of a delta that formed in a Gusev
paleolake. This observation corrected the Mariner 9-based
view that Ma?adim Vallis terminated in a 30 km crater, now
recognized as two overlapping craters provisionally named
NewPlymouth andDowne, which are actually superposed on
the Gusev crater rim and crosscut by the valley [Landheim,
1995; Milam et al., 2003]. Subsequent base level declines
within Gusev crater could be responsible for dissection of the
former delta [Schneeberger, 1989; Grin and Cabrol, 1997].
Goldspiel and Squyres [1991] supported that interpretation,
estimating that 13,000 km3 of material had been eroded from
Ma?adim Vallis and deposited in Gusev crater. Carter et al.
[2001] have since compared theMars Orbiter Laser Altimeter
(MOLA) topography of Gusev to the younger Bakhuysen
Crater, estimating that 15,100 km3 of fill materials occur in
Gusev. Prior to the Mars Global Surveyor (MGS) mission,
depth and gradients in Ma?adim Vallis were calculated from
atmospheric pressure [Scott et al., 1978], shadows [Lucchita
and Dembosky, 1994], digital elevation modeling from stereo
images [Thornhill et al., 1993a], and Earth-based radar
[Goldspiel et al., 1993]. Thornhill et al. [1993b] also used
a shape-from-shading algorithm to calculate the discharge
given arbitrary flow depths, which generated values from
104 to 106 m3/s.
[10] More recently, Landheim [1995] suggested that
Ma?adim Vallis is an outflow channel originating from
etched and chaotic terrain located several hundred kilo-
meters south of the valley head. Later episodes of flow were
interpreted as possible but uncertain. Of the previous studies,
the Landheim [1995] thesis bears the closest resemblance to
the model we propose here, except that we invoke a
paleolake rather than an outflow as the valley source for
reasons described below, and details of the valley?s incision
history are now better defined with new topography. Cabrol
and her colleagues have studied Ma?adim Vallis and Gusev
crater in detail, and they have used their work to successfully
advocate Gusev crater as a landing site for the first MER
[Cabrol et al., 1994; Landheim et al., 1994]. In a series of
papers [Cabrol et al., 1996, 1997, 1998a, 1998b; Grin and
Cabrol, 1997], they suggested that Ma?adim Vallis was
supplied by groundwater or drainage from the chaotic
terrain or Sirenum Fossae south of the valley head [see also
Landheim, 1995; Kuzmin et al., 2000; Gulick, 2001]. Crater
counts suggesting a young surface mantle in Gusev crater,
terraces within Ma?adim Vallis, and morphology of
the terminal delta supported their view of a prolonged
period of episodic activity spanning 0.7 to 2.0 Ga. They
suggested that in this interval, impact craters twice inter-
rupted Ma?adim Vallis, such that intravalley lakes formed at
two sites. They attribute the terraces within Ma?adim Vallis
to these ponding episodes and subsequent catastrophic
flooding as the intravalley lakes were breached. They
suggest that the morphology of the Gusev floor reflects the
former paleolake [Grin and Cabrol, 1997], and that knobs in
the floor are related to emergence of ice as pingoes [Cabrol
et al., 2000]. Our identification of an enclosed basin at the
head of Ma?adim Vallis, a reevaluation of crater counts and
new topographic data, and new detailed visible and thermal
imaging of the valley cast much of that proposed history in
doubt, as we describe herein. Aharonson et al. [2002]
examined the irregular longitudinal profile of Ma?adim
Vallis, noting that the valley and its tributary networks are
poorly graded (they do not exhibit the usual downstream
decline in gradient). This observation suggested that the
drainage network was underdeveloped if formed by runoff
erosion, and that it may be more consistent with sapping.
Aharonson et al. [2002] plotted the valley head east of
the location identified by Scott et al. [1978] and Irwin
et al. [2002], which may have resulted from the superposi-
tion of a fresh crater on the upper reach of Ma?adim Vallis,
which diverts a computationally derived flow path up an
eastern tributary.
[11] Geologic maps at 1:5M scale were completed of the
contributing drainage basins by De Hon [1977], Scott et al.
[1978], Mutch and Morris [1979], and Howard [1979].
These earlier mappers recognized the dissected parts of the
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
6 of 33
E12009
landscape, eroded knobby and chaotic terrains, several
units of smoother plains (some interpreted as volcanic),
and aeolian mantles, but with variability in mapped units
and interpretations. Scott and Tanaka [1986] andGreeley and
Guest [1987] remapped the equatorial region of Mars at
1:15M scale, attributing the Eridania basin plains to volcanic
resurfacing. Kuzmin et al. [2000] produced a detailed
1:500,000-scale map of geologic materials in and around
Gusev crater, with interpretations similar to Cabrol et al.
[1996, 1998a, 1998b]. Milam et al. [2003] remapped the
Gusev floor using detailed MGS imaging and thermal inertia
products, acknowledging a possible diversity of volcanic,
fluvial, and aeolian materials in the crater.
2. Paleolake Indicators
[12] In this section, we describe the evidence for an
extensive paleolake, based on the topography of the basin
at the head of Ma?adim Vallis, the distribution and origin of
erosional features within and around this basin, and the
likely origin of the water.
2.1. Valley Head Morphology
[13] The Ma?adim Vallis head is among the most defin-
itive indications that an overflowing paleolake carved the
valley. Figure 1 illustrates the eastern valley and the large
anastomosing tributary that enters from the southwest, both
of which originate abruptly at gaps in the divide (degraded
crater rim) separating the intermediate and Eridania basins.
At the source for the eastern (main) valley, Ma?adim Vallis
flows incised the drainage divide from its original 
1,200 m
elevation and displaced the channel head 25 km southward
to its present elevation of 950 m on the Eridania basin
plains. The western gap is the source for the main valley?s
large tributary (Figure 4), which is discussed in more detail
in section 3.4. Determining the initial elevation of the
western gap is less straightforward, as a 28 km fresh crater
formed adjacent to the gap after fluid flow had ceased in the
tributary (Figure 4a). Fluidized ejecta from this crater
partially filled the uppermost reach of the tributary valley
and locally raised the elevation of the drainage divide.
Modern topography suggests that the tributary source oc-
curred at 
1320 m, but the precrater level can be calculated
by subtracting the ejecta thickness d, which thins with
distance r from the center of a crater with radius R (in an
ejecta blanket, r > R meters). McGetchin et al. [1973]
developed an ejecta thickness function for lunar craters:
d ? 0:14R0:74 r=R? ?3: ?1?
We estimate that the ejecta thickness declines to 
70 m at a
distance of 34 km from the crater center, so that the original
Figure 4. (a) The major tributary system to Ma?adim
Vallis, with THEMIS daytime IR images overlaid on the
Viking Orbiter Mars Digital Image Mosaic Version 2.1 and
colorized with MOLA topography as in Figure 1. The valley
originates at the intermediate basin drainage divide (1) and
immediately divides into two anastomosing channels (2).
Downslope bifurcations occur at sites marked by arrows.
The western tributary courses (3) were abandoned early as a
crater (4) captured the flows. Upon emerging from this
crater, the flow again bifurcated along the intermediate basin
floor to incise the deep tributary channels 5 and 6, which are
graded to the Ma?adim Vallis floor. (b) Longitudinal profile
of the easternmost flow path, which was the last to be
utilized. The upper reach is incised only 100?200 m below
the plateau surface (given as the highest point within 4 km of
the channel thalweg), whereas the lower tributary reach was
graded to the Ma?adim Vallis main channel floor, where it
incised up to 600 m.
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
7 of 33
E12009
divide occurred at 
1250 m relative to the modern datum,
close to the level of initial overflow of the eastern gap. This
approximation is imprecise given that the crater occurs on
Mars, where the gravity-controlled ballistic component of
ejecta transport is shorter, but fluidized flows may extend
the ejecta blanket.
[14] The morphology of the Ma?adim Vallis source points
is consistent only with origin by paleolake overflow. Ele-
vations in the Eridania basin decline to the south of the two
valley heads, so no smaller valley networks could have
supplied either of the Ma?adim Vallis source gaps directly,
nor are such networks evident on the Eridania basin floor.
Malin and Edgett [2000a] suggested that, in some areas,
significant changes in topography occurred during the
Noachian due to aeolian fill and removal. However, we
find no constructive support in the geologic record (e.g.,
taller remnant mesas or pedestal craters) for ancient Eridania
basin topography that would have allowed surface drainage
to Ma?adim without ponding. Some aeolian deflation is
evident in the Eridania basin, mainly of Electris air fall
materials [Grant and Schultz, 1990] that appear to have
been deposited after Ma?adim Vallis formed, as discussed
below. Remnant mesas suggest that these sediments previ-
ously filled only a small portion of the total Eridania basin
floor relief (Figure 5).
[15] Many valleys on Mars are attributed to sapping
processes due to their poorly developed tributary networks
and theater headwalls that may lack upslope tributaries [e.g.,
Baker, 1982; Malin and Carr, 1999; Gulick, 2001]. Where-
as some terrestrial [Howard et al., 1988; O?Connor, 1993]
and Martian [Irwin and Howard, 2002] drainage divides
appear to have been breached by headward migration of a
sapping headwall, fed by surface water or groundwater on
the opposite side of the divide, no headwall scarp occurs at
the atypically wide Ma?adim Vallis heads, and no contrib-
uting aquifer (except the paleolake that we propose) oc-
curred above the ridge crest where the valley source points
occur. Furthermore, sapping without subsequent paleolake
overflow is inconsistent with the discharges suggested by
valley morphology, based on calculations presented In
section 4.2. The western tributary developed an anastomos-
ing course (Figure 4), which can occur when the topography
cannot confine the volume of water it is required to deliver
(as in the Channeled Scabland, Washington, U.S.A. [Baker,
1982, pp. 147?148]), or when aggradation of a channel
floor forces the flow out of the channel [Ritter et al., 1995,
p. 221]. In this case, the large flow alternative is most
reasonable, as no upstream surface was available to supply
sediment to aggrade the floors of these large tributaries. The
western paleolake outlet is more favorable to anastomosing
channels relative to the eastern outlet, because fewer topo-
graphic obstructions were available to confine the western
flow.
[16] Both the main and tributary channels are full-born,
i.e., their 
3 km width at the source is similar to down-
stream channel dimensions. Sites of locally decreased
channel width are associated with increases in slope as
shown on the longitudinal profile (Figure 3) and remain
consistent with a constant discharge in the down-valley
direction. In contrast, Al-Qahira Vallis (Figure 6) is a
valley network with a similarly wide lower reach, located

400 km west of Ma?adim Vallis. Both valleys cross the
dichotomy boundary and are incised a kilometer or more
into cratered materials. Unlike Ma?adim, Al-Qahira Vallis
originates through the confluence of narrow, ridge-bounded
tributary networks. Ma?adim Vallis maintains its width even
where it crosscuts impact craters, whereas crater breaches
at Al-Qahira Vallis have variable width. The wide lower
reach of Al-Qahira Vallis maintains a low, nearly constant
gradient (Figure 6a), suggesting that the valley widened
only after a stable base level was acquired.
[17] Although most Martian and terrestrial valleys origi-
nate immediately down-slope of their drainage divide,
Ma?adim Vallis incised its initial headward divide to

250 m depth, so that the present divide at the channel
head is lower and projects 25 km southward into the
Eridania basin. Headward displacement of the drainage
divide is typical of terrestrial basin overflows. An analogous
but much smaller catastrophic flood occurred on July 15,
1982, at Lawn Lake, Larimer County, Colorado, U.S.A. At
this site, glacial scour of bedrock and development of a
terminal moraine formed a natural basin of 66,370 m2. An
earthen dam was built on the moraine in 1903, artificially
enlarging the precursor basin. Overflow of the composite
dam stripped the unconsolidated dam and underlying glacial
till down to bedrock, incising an outlet channel into the till
and displacing the overflow spillway headward from its
earlier position [Jarrett and Costa, 1986] (Figure 7). This
overflow is particularly relevant to Mars. Before the flood,
stream flows occurred on the surface of glacial till in the
valley, as water may have flowed on the surface of similar
Martian crater ejecta. With increased discharge during the
Lawn Lake flood, a critical shear stress was reached to
mobilize the bouldery glacial till, rapidly producing the
modern deeply incised outlet valley. A much larger paleo-
lake overflow could mobilize the Martian megaregolith
similarly, as demonstrated in section 4.3, carving Ma?adim
Vallis over a short time. A comparable butmuch larger natural
dam failure occurred at Red Rock Pass, southern Idaho,
U.S.A., on the northeastern margin of paleolake Bonneville
Figure 5. Morphology of positive-relief chaotic terrain and a remnant mesa of the Electris deposit in the deep crater
subbasins. (a) The rough, high-albedo material that makes up the mesas is capped by a darker, more durable, thin cap rock
(subset of MGS Mars Orbiter Camera (MOC) image M04-03292, centered at 35
S, 173
E). (b) MOLA track 14244,
showing the positive relief of chaotic terrain relative to smooth plains on the depression floors. Dissected terrain occurs
only above 
700 m, whereas lower-lying plains are smooth and undissected despite their steep slopes. Chaotic terrain is
offset from the subbasin center. (c) MOLA track 16803, showing a remnant mesa of the Electris deposit overlying the
Eridania basin floor plains. The Electris deposit did not fill the Eridania basin completely to the 950?1250 m elevations of
the Ma?adim Vallis heads. The track also crosses two degraded impact craters with floor materials that slope toward the
Eridania basin interior. (d) MOC Geodesy Campaign Mosaic for context. White line, 1100 m contour; black line, 700 m
contour.
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
8 of 33
E12009
Figure 5
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
9 of 33
E12009
about 14,500 years before present. The natural dam at this site
was an alluvial fan, which impounded the lake at its highest
(Bonneville) stage for several hundred years, yielding a lake
of 51,530 km2. The divide ultimately failed due to over-
topping by the lake or headward growth of sapping channels
(supplied by groundwater from the elevated lake) that devel-
oped in the unconsolidated fan material. A 108m drop in lake
level followed as the overflow channel stripped the fan down
to bedrock, yielding a discharge of 
106 m3/s [O?Connor,
1993]. Lake Bonneville experienced a rapid decline in level
shortly after its single overflow, as the climate became
progressively more arid [Currey, 1990].
2.2. High Internal Relief, Undissected Eridania Basin
Impact Crater Floors
[18] The topography of the Eridania basin floor is atypical
of basin plains in the Martian highlands, and characteristics
of the floor topography suggested the occurrence of isolated
paleolakes [Howard, 2000] before an association was made
between a larger, integrated paleolake and Ma?adim Vallis
[Irwin et al., 2002]. Six highly degraded impact craters
(180?240 km diameter) within the Eridania basin have
atypically concave interiors, with high internal relief (up to
1 km) and relatively steep, inward-declining slopes (com-
monly 0.5?1.5
) (Figures 5 and 8). In contrast,most degraded
impact craters >60 km in diameter on Mars have nearly flat
floors, a sharp break in slope at the base of the crater wall,
and decreasing floor relief with increasing crater diameter
(Figure 9) [Howard, 2000; Craddock and Howard, 2002].
[19] Noncumulative hypsometric curves, or plots of the
total area within elevation intervals, illustrate the concavity
of these crater floors relative to normal Martian impact crater
interiors at progressive states of degradation (Figure 9,
Table 1). The depth of an impact crater usually declines
with advanced degradation, as the rim is lowered and the
interior fills. Hypsometric curves of typicalMartian degraded
craters peak at the lowest elevations (Figure 9), representing
the large area of the nearly flat crater floors. This is contrary
to the trend in the Eridania subbasins and in the nearby
Newton and Copernicus craters, where the infilling material
appears concentrated around the margins of the crater floors,
so that the internal relief at the center was largely maintained.
The high floor relief and morphology of these unusual crater
floors can be recognized in the broad bench on the lower side
Figure 6. Al-Qahira Vallis. (a) Longitudinal profile of the main channel, which crosscuts a small crater
(C) near the valley head. The valley is incised up to 700 m at I. The valley widens abruptly below S,
where it acquires its base level at 
1500 m. (b) Drainage basin of Al-Qahira Vallis, with major
tributaries (black lines) and drainage divides (white lines) indicated. In contrast to Ma?adim, Al-Qahira
Vallis grows from a space-filling tributary network rather than a large point source, and it is wide only in
its lower reach.
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
10 of 33
E12009
of the hypsometric curve (Figures 9h and 10). The Eridania
basin hypsometry also differs from that of large fresh craters,
which exhibit neither the bench nor a marked peak at the
lower end of the curve (Figure 9e).
[20] The Eridania basin floor has been interpreted in part as
volcanic plains [De Hon, 1977; Howard, 1979; Scott and
Tanaka, 1986; Greeley and Guest, 1987], however the large
relief within the plains is inconsistent with the flat or radially
outward-sloping topography of more widely recognized
volcanic landscapes on Mars (e.g., Solis, Lunae, or Hesperia
Plana). Volcanic plains do not preserve concave depressions
with deep central low points, as lava preferentially fills low
areas in the topography. All degraded craters on the Eridania
basin floor have concave floor deposits or floors that slope
downward toward the center of the larger subbasin in which
they are located (Figures 5c and 8). This contrasts with lava-
embayed crater floors, which are typically flat or have a
monotonic upward slope toward the breach through which
lava entered the crater. As the Eridania basin craters are not
embayed, the larger subbasins were not filled deeply or to a
near-level surface, and no positive-relief volcanic constructs
are evident in the area, volcanic resurfacing likely played a
minor role, if any, in the present basin shape.
[21] The large relief of the Eridania basin floor suggests
that the lower parts of the basin were shielded from fluvial
erosion during the Noachian. As valley networks have
formed on most >0.5
 slopes in the region, it is likely that
such valleys would have transported sedimentary materials
toward the center of the Eridania basin over time, filling
the depressions similarly to most other degraded craters
[Craddock et al., 1997; Craddock and Howard, 2002].
However, most of the large impact craters that form the
subbasins have developed wide gaps in their original crater
rims (Figure 8), so we must attribute some modification and
infilling of the large craters to erosive processes rather than
simple mantling. To collect the infilling material primarily
around the margins of the craters, rather than in the center,
sediment transport processes must have been relatively
inefficient within the interiors of the large Eridania basin
craters, consistent with the presence of standing water.
2.3. Absence of Valley Networks at Low Elevations
[22] Inhibited fluvial sediment transport within the deeper
subbasins is consistent with the distribution of valley net-
works. In the deeper eastern subbasins, valley networks do
not extend to the bottom of the lower-lying concave crater
floors, despite the occurrence of regional slopes of 0.5?1.5

below 700 m elevation (Figures 5 and 11). Valley networks
in the Eridania basin watershed terminate over a range of
elevations, with most ending at the plains boundary between
1100 and 950 m (the elevation of the Ma?adim Vallis head).
In localized, steeper sections of the Araidnes, Atlantis, and
Gorgonum subbasin interiors, shallow valleys extend in-
ward only to near the 700 m contour, whereas the low points
of these subbasins occur at 200, 500, and 450 m,
respectively. Dissection of steep slopes below 700 m occurs
only in two localized areas, along the eastern margin of
Ariadnes basin and the northern margin of Atlantis basin,
where a few valley networks terminate at or above 300 m.
In the eastern subbasins, valleys dissect the 1100 m contour
along 
90% of its length, but only 
15% of the 700 m
contour is crosscut, and no valleys extend below 300 m.
Figure 7. (a) Concept sketch of a paleolake overflow. The
paleolake overtops the dividing ridge (black line) at the L1
level. Floodwaters downcut the divide with a relatively high
stream power on the steep S1 slope, regrading the outlet to
the lower S2 slope and releasing a large volume of water
stored between the L1 and L2 levels. Further incision of the
outlet below the L2 level suffers from lower slope and
stream power, plus lower discharge released from the lake
per unit of material that must be removed from the outlet
valley. Ultimately, incision terminates and a remnant lake is
impounded below the L3 level. (b) Lawn Lake, Colorado,
showing the predamburst shoreline (black line) and the
remnant lake (gray), adapted from Jarrett and Costa [1986].
Dashed lines show the inlet channel and the wide outlet
channel carved during the flood. In this case, the incising
channel encountered bedrock at the L2 level and no further
incision was possible. The modern spillway occurs head-
ward of the original dam crest as shown in Figure 7a. (c) The
modern remnant lake and light-toned, exposed lake bed,
facing down-valley to the southeast.
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
11 of 33
E12009
Undissected relief within the Eridania basin floor therefore
amounts to 500 to 1200 m depending on the subbasin. In
the northwestern subbasin, no valleys were observed below
770 m. In high-resolution imaging, some rare, shallow, late-
stage gullies dissect crater rims and other steep slopes across
a wide range of elevations [Malin and Edgett, 2000b], but
these gullies represent little geomorphic work.
[23] These observations suggest that fluvial erosion af-
fected only the higher levels of the Eridania basin and that
only spatially and temporally limited water sources were
available to dissect topographic levels below 
700 m. In
contrast, valley networks are common and extend to the
drainage divides above 950 m. This pattern of elevation-
controlled fluvial incision is consistent with the hypothesis
that the deeper subbasins were occupied by standing water
during much of the Late Noachian and Early Hesperian
Epochs. Fresh impact crater walls and Sirenum Fossae
graben provide a view of layered stratigraphy beneath the
Eridania basin floor, where thin, blocky outcrops alternate
with more friable layers (Figure 12). The more durable
layers are resistant to aeolian erosion, and boulders are
common in talus slopes at these sites. Layered deposits in
the highlands are certainly not unique to a paleolake floor,
but are consistent with the hypothesis. The thin, durable
layers may have formed in response to cementation of
sediments by precipitated minerals, although a definitive
lithology cannot be determined with available data.
2.4. Indications of Former Lake Levels and
Post-Noachian Adjustments
[24] Three types of landforms suggest a shoreline for the
Ma?adim Vallis head paleolake, including (1) the overflow
Figure 8. MOLA topography (60 pixels/degree, 100 m interval) of the eastern part of the Eridania
basin. The MOLA tracks shown in Figure 5 are indicated. The large impact craters in this region
maintained an unusually large floor relief and interior slope below 700 m elevation (dark blue line),
relative to smaller degraded impact craters at higher elevations. The light blue line is the 1100 m contour,
which is generally the plains/watershed boundary. One crater (B) has a wide breach in its rim. Some
impact crater floors (S) above 700 m elevation slope downward toward the Eridania basin interior,
demonstrating control of sediment transport by local slopes rather than a regional directionality as with
aeolian deposition. These craters are not embayed by lavas from the basin interior, which is low-lying.
Some impact crater interiors (R) have been reintegrated with surface drainage patterns, whereas other
low-relief intercrater plains (P) were not integrated. Cratering generally created a multibasinal landscape.
Named subbasins are indicated by (1) Araidnes, (2) Atlantis, and (3) Gorgonum.
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
12 of 33
E12009
Figure 9. (a?d) Cross-sectional profiles of three peak-ring crater interiors at progressive states of
degradation (Figures 9a?9c) compared to the interior of Newton crater (Figure 9d) without its exterior
watershed. Newton is interpreted to have contained a separate paleolake. Most impact craters on Mars
developed flat floors during degradation, so much of the floor area is concentrated within a narrow
elevation interval. (a) Fresh Lowell crater. (b) Degraded Kepler crater. (c) Highly degraded Flaugergues
crater, where the central peak ring has been largely removed or buried. (d) Newton crater, where infilling
of the crater floor took place only around the margins of the floor, such that the original interior relief is
largely preserved. (e?h) Noncumulative hypsometric curves (100 m elevation bins) of the degraded
crater floors in (Figures 9a?9c), respectively, compared to Newton crater in Figure 9d. Newton exhibits a
broad bench in the curve at lower elevations, instead of the usual peak in the curve at low elevations.
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
13 of 33
E12009
points, (2) the boundary of the Eridania basin interior plains,
and (3) the termination of valley networks.
2.4.1. Overflow Points
[25] Two apparent overflow points occur on the Eridania
basin divide, where the divide is breached to initiate
Ma?adim Vallis. The western outlet occurred at 
1250 m
and the eastern (main) outlet has been downcut from

1200 m to 950 m. The original elevation of the eastern
outlet is uncertain due to later incision, but it likely did not
exceed 1200 m. Two other low points on the Eridania basin
divide, at elevations of 1125 and 1170 m (points P1 and P2
on Figures 1 and 2, respectively), exhibit no evidence of
overflow. These other low points require that the source
paleolake level did not exceed 1125 m, if modern topogra-
phy is an exact representation of its previous state [Irwin et
al., 2002]. However, the western outlet is 125 m and 80 m,
respectively, higher than the elevation of the unused passes
P1 and P2. Some postoverflow topographic movement is
therefore necessary to account for the western tributary
overflow. Here we examine candidate processes that have
affected the Eridania basin topography.
[26] The P1 low point (elevation 1125 m) is only 200 km
east of the eastern (main) Ma?adim Vallis source gap
Table 1. Fresh and Degraded Impact Basins Used for Comparison
With Eridania Basin Hypsometry
Crater Location Diameter, km
Lowell (fresh) 52
S 278.5
E 210 km
Kepler (degraded) 46.5
S 145
E 230 km
Flaugergues (degraded) 16.5
S 19
E 250 km
Newton (paleolake) 40
S 201
E 280 km
Figure 10. Noncumulative hypsometric curves (100 m elevation bins) of the Eridania subbasins and
nearby Newton and Copernicus crater watersheds, as delimited in Figure 2. (a) Eastern subbasins.
(b) Northwest subbasin. (c) Southwest subbasin. (d) Newton crater. (e) Copernicus crater. (f ) Gorgonum
crater and its watershed, which is one of the eastern subbasins. Curves are described in the text. The
700 m level is indicated in the Eridania basin, where valley networks commonly terminate despite steeper
slopes below.
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
14 of 33
E12009
Figure 11. Valley network terminations in the Eridania
basin shown in the Mars Digital Image Mosaic (MDIM)
2.1. Valley networks commonly terminate at the plains
boundary between the 1100 m contour (black line) and the
950 m contour (white line), although valley networks
locally extend down to the 700 m contour as in Figure 5d.
Figure 12. Layered stratigraphy in the Eridania basin
floor. (a) Layering in a crater wall (MOC E11-00845,
centered at 31.6
S, 193.6
W). (b) In another crater wall in
the same image, layers are evident where high-albedo
materials overlie a darker, thin layer. (c) High-albedo
material overlying thin, dark layers in the Sirenum Fossae
graben, which crosscut the Eridania basin floor (MOC M02-
01583, centered at 38.9
S, 172.9
W, as shown in Figure 2).
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
15 of 33
E12009
(original elevation 
1200 m) (Figure 1). At P1, a 29 km,
rimless degraded crater occupies the entire width of the
pass. It is possible that during the outflow to Ma?adim
Vallis, the rim of this crater still filled the pass, raising the
present gap. On the basis of MOLA data analysis, Garvin et
al. [2003] give the rim height h (m) for fresh complex
craters of diameter D (m) as
h ? 0:02D0:84: ?2?
Assuming that the crater widened 10% with degradation
[Grant and Schultz, 1993; Craddock et al., 1997], a 26 km
fresh crater would have a rim height of 
300 m, so this
crater would provide the necessary blockage at a fresh or
reduced state of degradation relative to its present form. As
adequate fluvial erosion occurred after the main Ma?adim
Vallis flow to incise its watershed tributary network,
adequate time was available to erode this small crater to
its present rimless form. Terrestrial bedrock erosion rates are
typically highest on steep slopes, regardless of whether
slope or fluvial processes are active [Howard, 1998], so
ancient Martian crater interiors should have experienced
high erosion rates relative to intercrater plains.
[27] To account for the 80 m discrepancy in elevation
between the western source point and the P2 undissected
pass, a westward tilt or declivity of 0.0001 (0.006
) would
have to develop across the Eridania basin after the paleolake
dissipated. Isostatic rebound may have influenced modern
basin topography somewhat, but we have not modeled its
effect. The eastern half of the Eridania basin contains Early
Hesperian, Sirenum Fossae Tharsis-radial graben [Anderson
et al., 2001], and larger compressive ridges that are gener-
ally circumferential to Tharsis occur throughout the basin,
so a growing Tharsis volcanic load may have adequately
flexed the lithosphere after the overflow. Phillips et al.
[2001] suggested that the Tharsis load (some of which must
be post-Noachian to account for the young surface age)
deformed the elastic lithosphere globally, but they did not
specify the magnitude of the deflection. This load created
uplifts at 80
E and 260
E, with corresponding topographic
troughs centered at 170
 (within the Eridania basin) and
350
E. The ridge of mountains located east of Newton
crater (upper-right corner of Figure 2) may be a thrust belt
of Late Noachian to Early Hesperian age, suggesting that
crustal shortening and associated loading occurred in that
area [Schultz and Tanaka, 1994]. By any of these means, a
minor flexure of 
100 m amplitude could have been
applied to the Eridania basin, accounting for the observed
relationship between dissected O1 and O2 and the undis-
sected P2 low point in the Eridania basin divide. At the
present stage of understanding of Martian global geody-
namics, we are unable to fully evaluate these possibilities.
2.4.2. Boundary of the Eridania Basin Interior Plains
[28] The boundary between the Eridania basin plains and
higher dissected surfaces is displayed in noncumulative
hypsometric curves in Figure 10. We delimited the water-
shed using MOLA topography (64 cells/degree), supported
by the mapped extents of contributing valley networks.
Peaks in the hypsometric curves correspond to low-gradient
plains, which are elevation intervals containing relatively
large areas. Within the Eridania basin, the subbasin curves
exhibit three dominant features: (1) an exponentially de-
creasing area with increasing elevation above 1300?1400 m
(typical of both eroded and cratered landscapes [Grant and
Fortezzo, 2003]); (2) a break in slope below 1100?1200 m,
transitioning to the Eridania basin floor plains that lie
approximately between 700 and 1100 m; and (3) deep
subbasins below 700 m in the east. The northwestern and
southwestern subbasins have more level floor plains rather
than deep depressions, so the lower ends of their hypso-
metric curves are steeper than is the curve is for all eastern
(deeper) subbasins. The upper plains boundary in all of
these subbasins is 
100 m below the elevation of the
Ma?adim Vallis source gaps prior to incision of the valley,
suggesting localization of sedimentary deposition below
1100 m. If flooding occurred to 1100?1250 m relative to
present topography (the range of elevations incorporates
possible later tilting), the entire Eridania basin would have
contained a contiguous water body. The drainage basins of
two large, concave floored, degraded craters adjacent to the
Eridania basin, Newton and Copernicus, exhibit broad
plains at higher elevations (1100?1400 m). The base level
for erosion in these craters may have been somewhat higher
than in the Eridania basin. Alternately, air fall deposits of
100?400 m thickness have also mantled these more south-
erly drainage basins [Grant and Schultz, 1990], raising the
elevation of modern topography above the level of under-
lying plains.
2.4.3. Termination of Valley Networks
[29] As stated above, Hesperian valley networks that
incise the Electris deposit do not continue into the deep
subbasins within the Eridania basin, below about 700 m
elevation. Other valley networks terminate at higher eleva-
tions where they encounter the Eridania basin plains
(
1100 m) or other low-slope surfaces.
[30] On the basis of the hypsometric and morphologic
evidence discussed above, we suggest that a paleolake
occupied variable levels in the Eridania basin from 
700
to 1100?1250 m during the Late Noachian and Early
Hesperian Epochs. During this time the paleolake inhibited
fluvial erosion perpetually below 700 m elevation in the
Eridania basin and occasionally at higher elevations. Sedi-
ments were generated from the degradation of topographic
highs in the surrounding watershed, and these sediments
were likely deposited and reworked within the Eridania
basin as the water level rose and fell, over a long period or
possibly in episodic or periodic cycles.
2.5. Discussion
[31] As liquid water does not occur on the Martian
surface at present, identification of paleolakes must rely
on related landforms or topography that remained preserved
to the present time. Erosion rates declined in the equatorial
Martian highlands after the Noachian [Craddock and
Maxwell, 1993], so many features of that age are preserved,
but 
3.7 billion years [Hartmann and Neukum, 2001] of
post-Noachian processes now dominate the surface at meter
to hectometer resolutions, or lower resolutions in some
regions. Given that any Martian paleolakes probably
had variable levels (as do terrestrial lakes) and must
have ultimately dissipated with time, all levels of such
water bodies were ultimately exposed to impact gardening
[Hartmann et al., 2001], aeolian processes, tectonism,
and progressively limited fluvial and lacustrine processes.
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
16 of 33
E12009
Preservation of paleolake features therefore depends on
large landforms relative to the thickness of younger deposits
and the efficacy of more recent erosional processes. Shore
landforms like bars, spits, and overwash are particularly
susceptible to later erosion, or they may not form at all if a
lake is perennially ice-covered [French, 1976, pp. 206?
208]. Larger features and the morphology of Ma?adim Vallis
provide the support for our inferences about the origin of the
valley.
[32] As for all fluvial features on Mars, a debate continues
over whether atmospheric or subsurface processes supplied
the water. Landheim [1995] and Cabrol et al. [1997]
proposed that Ma?adim Vallis was supplied by groundwater
from Sirenum Fossae or from chaotic or knobby terrains,
which occupy parts of the deep Eridania subbasin floors
(Figure 5). However, the Hesperian graben [Anderson et al.,
2001] crosscut all other major geomorphic features in this
area, and one smaller aligned graben crosscuts Ma?adim
Vallis, so the timing does not support that hypothesis.
Furthermore, the chaotic terrains lie 
1 km below and at
least 370 km from the head of Ma?adim Vallis, so they
could not supply water to the valley heads without first
forming a lake. We also have no definitive indication that
the chaotic terrain was a water source. The chaotic terrains
do not occupy the full extents of their respective subbasins,
and the groups of mesas are not always centered within the
basins, so origin of the very deep basins by collapse is not
suggested by the chaotic terrain morphology. It may be
simply an in-place, eroded sedimentary deposit [Lucchita,
1982; Grant and Schultz, 1990; Irwin et al., 2002; Moore
and Howard, 2003].
[33] The second possibility is that precipitation supplied a
source paleolake for Ma?adim Vallis. Fluvial erosion of the
cratered highland landscape is indicated primarily by de-
velopment of widespread branching valley networks that
often head near sharp ridge crests (where groundwater
sources are unlikely) [Masursky et al., 1977; Irwin and
Howard, 2002], locally basin-filling drainage networks
[Grant, 2000; Irwin and Howard, 2002; Hynek and Phillips,
2003], degradation patterns of impact craters [Craddock et
al., 1997] and other basins [Irwin and Howard, 2002],
ubiquitous loss of small Noachian craters (including on
interfluves) [Craddock and Maxwell, 1990, 1993; Strom et
al., 1992], control of erosion by slopes on the order of 
1
,
and the requirement for aquifer recharge to support sapping
erosion [Howard et al., 1988; Goldspiel and Squyres, 1991;
Grant, 2000; Gulick, 2001]. For these reasons, the occur-
rence of a paleolake is consistent with other evidence for
surface water summarized by Craddock and Howard
[2002].
[34] Precipitation is the ultimate water source for peren-
nial lakes on the Earth occurring above sea level, but these
lakes are almost always connected with subsurface aquifers.
Although a lake can initially form over a frozen layer in the
substrate, a thermal gradient that is adequate to maintain the
lake (including those with a seasonal or perennial ice cover)
will ultimately provide heat to melt the underlying perma-
frost. In this case, the lake creates a connection to the
groundwater system where one may not have existed
beforehand [French, 1976, pp. 122?125], and the lake
surface will ultimately respond to local hydraulic head as
well as precipitation/evaporation controls. Alternately, if the
climate is cold enough for the freezing front to extend below
the lake floor, both the lake and subsurface will freeze.
Salinity can impede wholesale freezing as a relatively fresh
ice layer develops on the surface. In this manner, paleolakes
in the Eridania basin may have maintained their connection
with groundwater sources during less clement time periods.
The Eridania basin crater depressions, and the similar con-
cave floors of nearby Newton and Copernicus craters, are the
lowest points on the Terra Cimmeria/Sirenum plateau. If an
unconfined water table occurred anywhere within 1.5 km of
the mean plateau surface (given frequent impact cratering,
long-term confinement is unlikely), groundwater should have
emerged in the deeper depressions.
[35] In this study, we are discussing a region in the
midlatitudes, which would have been generally cooler than
the equatorial regions except during periods of high obliq-
uity [Ward, 1992]. The actual climate at the time and place
of the Ma?adim Vallis outflow is uncertain, although annual
temperatures near 0
C would minimize evaporation and
help to maintain a paleolake on a relatively (to Earth) arid
planet, particularly if it had a seasonal ice cover. Terrestrial
lakes are most common in latitudes of 40?60
 for this
reason, relative to multibasin landscapes in warmer climates
[Cohen, 2003]. It is also possible that melting polar ice
helped to raise water tables over much of the highlands
[Clifford and Parker, 2001], including in the Eridania basin.
Basal melting of south polar ice sheets may have occurred
over the same range of elevations that were occupied by the
paleolake discussed here [Head and Pratt, 2001; Carr,
2002]. However, basal melting was not the only source
for water, as many valley networks in the Eridania basin
watershed and elsewhere originate at elevations of 2000?
3000 m [Carr, 2002].
[36] We can only speculate on the cause of the Noachian/
Hesperian boundary overflow as definitive indicators are
not apparent. A period of greater water influx or reduced
evaporation with a cooling climate may have allowed the
level to rise. O?Connor [1993] suggests that groundwater
sapping of alluvial fan material in southern Idaho initiated
the Lake Bonneville flood, which then carved its own
channel through a preexisting drainage basin. The sapping
explanation seems unlikely in this case, however, owing to
the contemporary development of two outlets. Two outlets
are more reasonable for a large contributing head basin than
for a small one. Unless the two outlets occurred initially at
very similar elevations by chance, the lake would need to
infill at a rate adequate to raise its level despite a small
initial overflow discharge, activating the second overflow
point. A large area or an abundant influx would make this
possible. It is also possible that a small impact into the lake
created a wave, which breached the Eridania basin divide
more rapidly. A long-lived lake is expected to have sus-
tained an impact flux comparable to outlying surfaces.
3. Basin Overflow and Development of Ma?adim
Vallis
[37] The morphology of Ma?adim Vallis provides addi-
tional support for the overflow hypothesis. We begin with a
discussion of the timing of the overflow, followed chrono-
logically by controls on the valley course, influences on its
cross-sectional and longitudinal profiles as it incised, and
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
17 of 33
E12009
the history of tributary development during and after the
flood. We also calculate the flood discharge and briefly
describe the deposits in Gusev crater.
3.1. Timing of Initial Overflow
[38] The timing of the paleolake overflow can be esti-
mated using fresh crater counts on the Eridania basin,
its watershed, and the watershed within the intermediate
basin that supplied Ma?adim Vallis directly. Fresh craters
overlie the widespread valley networks, so fresh crater
counts provide a lower bound on the age of fluvial events
[Craddock and Maxwell, 1990]. Several previous studies
have produced variable crater counts for Ma?adim Vallis,
which may reflect different data sources, crater diameters
counted, or criteria for delimiting the valley [Malin, 1976;
Masursky et al., 1977, 1980; Carr and Clow, 1981; Cabrol
et al., 1998b]. Including the area between the top of the
valley sidewalls, Ma?adim Vallis has 11 superposed fresh
craters of >2 km in diameter, in an area of 15,100 km2,
but such small crater populations generate statistically
insignificant age dates. In this study, we minimized the
error inherent to small surface areas by dating the Ma?adim
Vallis watershed?s small tributaries (the contributing area
within the intermediate basin and other small contributing
watersheds) in addition to the trunk valley itself. The
tributaries crosscut (are younger than) the trunk valley
walls, so dating the tributaries maximizes the available area,
minimizes the error, and allows both N(2) and N(5) counts
to be used. The crater retention age, N(x) ? error, is the
number of craters y greater than x km diameter in a specified
area, ?
??
y
p
error, with count and error normalized to an area
of 106 km2.
[39] Another apparent discrepancy is found in correlating
crater counts of different diameter to the Tanaka [1986]
stratigraphic scheme, whereby highland N(2) fresh crater
counts can yield younger stratigraphic ages than N(5)
counts from the same crater population. On this issue the
reader may refer to discussion by Tanaka [1986] and
Hartmann and Neukum [2001], and counts by Grant and
Schultz [1990], Tanaka [1997], Irwin and Howard [2002],
and Irwin et al. [2002]. The issue is that small crater counts
on Tharsis lava flows, on which the stratigraphic scheme is
based, apparently exhibit larger populations at <5 km than
do highland units or the Vastitas Borealis Formation
[Tanaka, 1986]. The discrepancy in small diameter popula-
tions may result from rapid degradation of small highland
craters or a different population of impactors during heavy
bombardment [Barlow, 1990], and more work is needed to
define the scope and implications of the problem. Our N(2)
fresh crater counts in this study area yield a late Hesperian
age, whereas N(5) counts yield an age near the Noachian/
Hesperian boundary on the stratigraphic scheme of Tanaka
[1986] (Table 2, modified from Irwin et al. [2002]). Cabrol
et al. [1998b] interpreted the young N(2) ages as indicating
prolonged, episodic fluvial activity over 0.7 to 2.0 Ga in
Ma?adim Vallis, starting in the Late Noachian Epoch and
continuing throughout the Hesperian period. However,
when the discrepancy in the highland crater data relative
to the Tanaka [1986] scheme is recognized, the Gusev crater
N(2) stratigraphic ages are actually consistent with the older
ages suggested by larger-diameter highland fresh crater
populations (Table 2), and the counts do not uniquely
indicate a long period of activity. Cabrol et al. [1998b]
and Kuzmin et al. [2000] also dated geologic units using
crater counts on units with small surface areas. As shown in
Table 2, the large errors inherent to such counts can result in
widely variable stratigraphic ages, even when the discrep-
ancy in the stratigraphic scheme is not considered.
[40] Regardless of which crater diameter is used to
determine age, Table 2 shows that our N(2) crater counts
on the Eridania basin floor are well within the error of the
Cabrol et al. [1998b] results on Ma?adim Vallis and the
Gusev crater floor. In Table 2 we have not included fresh
crater populations for the southwestern subbasin, as pristine
craters are rare south of 
40
S, where they have been more
substantially degraded by recent air fall deposits or ice-
related processes [Jankowski and Squyres, 1993]. Crater
degradation and valley network activity appear to have
declined at around the same time in the Eridania, interme-
diate, and Gusev basins, and the Eridania basin floor deposit
and the Ma?adim Vallis tributary networks appear to be
contemporary features (Tables 2 and 3). This timing is
consistent with the regional decline in valley network
activity and crater degradation that has been dated to near
the N/H boundary using N(5) [Maxwell and McGill, 1988;
Craddock and Maxwell, 1990; Grant and Schultz, 1990;
Irwin and Howard, 2002]. With supporting evidence dis-
cussed below, the scenario of a long-lived Ma?adim Vallis
can now be simplified by a brief flow near the N/H
boundary (again using N(5) ages), with limited fluvial
incision and resurfacing beyond that time.
[41] The longevity of the source paleolake is less certain.
Whereas fresh craters provide a lower bound on its age, the
only available upper bound is the middle Noachian age of
the Npl1 (cratered) unit in which the paleolake occurred
[Scott and Tanaka, 1986; Greeley and Guest, 1987]. The
Table 2. Comparison of N(2) and N(5) Impact Crater Counts for
Floors of the Eridania Basin, Ma?adim Vallis, and Gusev Crater
Geologic Unit Count (y) N(2)a N(5)b
Eastern, NW subbasins
<1100 m elevation
732 633 ? 27 192 ? 16
Gusev crater floorc 24 541 ? 221 none
Ma?adim Vallis floorc 5 708 ? 306 none
Ma?adim deltac 10 619 ? 206 none
aAbsolute age interpretation from Hartmann and Neukum [2001]: Late
Hesperian, 
2.93.6 Ga.
bAbsolute age interpretation from Hartmann and Neukum [2001]: Late
Noachian/Early Hesperian boundary, >3.5?3.7 Ga.
cData from Cabrol et al. [1998b]. The large error in their results is due to
small area of Gusev crater and small sample size.
Table 3. Fresh Crater Retention Ages for Hydrologic Units in the
Ma?adim Vallis Region
Hydrologic Unit
Count (n)/Area,
km2 N(5)
Ma?adim watershed 41/208,500 197 ? 31
Main source subbasin (all) 199/1,089,000 183 ? 13
Main source subbasin watershed (>1,100 m) 83/463,300 179 ? 20
Main source subbasin floor (700?1,100 m) 66/363,500 182 ? 22
Main source subbasin floor (<700 m) 50/262,300 191 ? 30
NW source subbasin (all) 57/285,500 200 ? 26
NW source subbasin watershed (>1,100 m) 22/123,500 178 ? 38
NW source subbasin floor (<1,100 m) 35/162,000 216 ? 37
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
18 of 33
E12009
paleolake was therefore present during a fraction of the

250 Ma included in the Middle and Late Noachian Epochs
[Hartmann and Neukum, 2001]. At a minimum, the lake
was present from near the time of formation of the larger,
deep subbasins, which were not extensively modified below
the 700 m level by fluvial sediment transport or valley
incision. Fresh crater retention ages at different levels of the
Eridania basin floor are similar to those of the higher-
standing watershed and the dissected intermediate basin
(Table 3), so the larger source paleolake did not persist
for long after the valley network activity ended in the
Ma?adim Vallis watershed. Some water may have collected
in the Eridania basin depressions episodically into the late
Hesperian [Howard and Moore, 2004].
3.2. Development of Flow Path
[42] We tested the overflow hypothesis by reconstructing
the topography of the Ma?adim Vallis region prior to incision
of the valley, to determine whether overland flow would
pursue the present course given point sources at the valley
heads. Given that Ma?adim Vallis typically has a sharp break
in slope at the top of the valley sidewalls, and only minimal
erosion appears to have occurred after the flow, the valley
remains sharply defined, and backfilling produces a close
approximation of the antecedent surface. We deleted the
valley from the 64 pixel/degree MOLA digital elevation
model (DEM) and refilled the DEM by kriging (Figure 13).
The volume added to fill the valley in this process,

14,000 km3, closely matches independent estimates of
the valley?s volume calculated from individual cross sections
[Goldspiel and Squyres, 1991; Cabrol et al., 1996; Irwin et
al., 2002]. The reconstruction is therefore generally accurate,
but it is not precise at all points, as some grid cells are filled
somewhat too little or too much by this method.
[43] The Ma?adim Vallis course favors the overflow
hypothesis. The reconstruction of prevalley topography
shows that antecedent topographic gaps and impact craters
would have controlled overland flow (arrows in Figure 13),
producing the modern sinuous valley course exactly as
shown in Figure 1. Water originating from the outlets
(points A in Figure 13) was directed into the intermediate
basin by the topography of several impact craters and other
preexisting drainage divides (points B?E) near the inter-
mediate basin rim. Water overflowed the northern interme-
diate basin divide (point F) and was channeled toward
Gusev crater by two ridges (points G and H) that lie
perpendicular to the dichotomy boundary. Along this lower
reach, the Ma?adim flow exploited an older drainage basin,
similar to those occurring to the east and west along the
dichotomy boundary slope (Figure 1). The ridges delimiting
this old drainage basin (G and H) directed water across the
lower-standing, eroded rim of overlapping New Plymouth
and Downe craters (C1), which are superposed on the
Gusev rim. Down-cutting along this topographically con-
trolled course created the entrance breach to Gusev crater.
[44] Impact crater rims appear to have resisted erosion
more than did other landforms. The flow crosscut five
impact craters (Figure 13, points X, C1, and C2) only
where other ridges, standing higher than the previously
eroded crater rims, directed flows through a crater. Other-
wise, the rims of impact craters as small as 17 km diameter
diverted the valley course around the crater. This control of
the valley course by prevalley topography (without excep-
tion) invokes overland flow of water rather than headward
erosion by structurally controlled groundwater sapping.
[45] A tectonic origin of the valley [Brakenridge, 1990] is
considered unlikely on the basis of the following observa-
tions: (1) Ma?adim Vallis is a sinuous valley rather than a
linear feature, it changes its longitudinal direction liberally,
and the valley is comparably wide in its north-south, east-
west, and intermediate orientations. (2) Extension related to
an east-west principal stress is not evident in other land-
forms in the area, and no observed extensional faults radiate
from the head, terminus, or bends of Ma?adim Vallis
(mapping by Brakenridge [1990] shows this lack of graben
with appropriate orientations). While the lack of observa-
tions does not preclude the existence of appropriate exten-
sional fault structures, we are unable to validate a tectonic
contribution to the valley. (3) Some impact crater rims are
breached, but the rims are not displaced to widen the crater
in the cross-valley direction. A ??left-laterally displaced??
crater discussed by Brakenridge [1990] is actually two
overlapping craters that have been collectively breached
[Landheim, 1995; Milam et al., 2003]; generally, very little
evidence of strike-slip motion on Mars has been observed.
(4) Valley bends are curved as would be expected for a wide
flow, not intersecting lineations as would be expected for a
structural valley. (5) We have demonstrated that the course
of the valley is entirely consistent with control by prevalley
topography. A tectonic hypothesis fails at this time for a
lack of constructive support, in addition to inconsistencies
with observations given here. These observations are also
inconsistent with structural control of groundwater flow
[Brakenridge, 1990]. Aeolian and volcanic hypotheses for
the origin of Martian valley networks and outflow channels
have been discussed in depth and refuted in earlier literature
[e.g., Carr, 1974; Baker, 1982], so such a discussion need
not be repeated here. The paleolake overflow concept, in
contrast, offers an internally consistent story that explains
the observations well.
3.3. Incision of the Ma?adim Vallis Lower Reach
[46] Overland flows from the Eridania basin were initially
impounded in the adjacent intermediate basin (Figures 1 and
13). The intermediate basin?s overflow point at 1000 m (on
its northern side) suggests a basin volume of 
60,000 km3,
of which <10% is the eroded volume of the upper and
intermediate reaches of Ma?adim Vallis. The intermediate
basin may have filled slowly prior to the Eridania basin
overflow, by atmospheric sources and/or northward seepage
Figure 13. (a) Reconstruction of topography prior to the Ma?adim Vallis outflow, from the 32 pixel/degree MOLA grid.
Arrows indicate the flow paths, where preexisting ridges and craters channeled overland flow from the Eridania basin. Sites
are described in the text. (b) The lower reach of Ma?adim Vallis, from the knickpoint through the Gusev exit breach.
THEMIS daytime infrared mosaic overlaid on the Viking Orbiter MDIM 2.1 and colorized as in Figure 13a with MOLA
topography.
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
19 of 33
E12009
Figure 13
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
20 of 33
E12009
of groundwater, or overland flow from the Eridania basin
may have filled it rapidly. As we discuss below, constrained
estimates of the total volume discharged would allow either
of these scenarios. Being initially covered with standing
water, the floor of the intermediate basin could not be
eroded until the lower reach of Ma?adim Vallis (north of
the intermediate basin) was incised, allowing the basin to
drain.
[47] Inward-sloping terraces along the wide lower reach
of Ma?adim Vallis, from the intermediate basin exit breach
to Gusev crater (Figure 14), may represent a wider channel
floor that was active during a peak discharge from the
intermediate basin. The terraces are likely not resistant
outcrops, as they do not constrict the channel relative to
its width in the intermediate reach, and they do not affect
the longitudinal profile. Including the width of the terraces,
the lower reach?s valley floor is wider (
9 km) than are the
floors of the intermediate and upper reaches (
5 km),
although the lower reach?s inner channel has a similar width
of 
5 km. Given that the lower reach was incised first, its
width may reflect earlier flow conditions. Channel width
increases with the square root of discharge [Knighton,
1998]. Filling the intermediate basin to its overflow point
(
1000 m) would not impede further influxes from the
Eridania basin, where water stood at a higher level (origi-
nally 1100 to 1250 m). Therefore, as the northern divide of
the intermediate basin was incised, water would be released
from storage in the intermediate basin, while water would
simultaneously flow from the Eridania basin into the inter-
mediate basin, temporarily increasing the total discharge,
erosive power, and width of the Ma?adim Vallis lower
reach.
[48] Another important influence on paleolake discharge
would be the incision rate at the valley head gaps. For an
Eridania basin paleolake of 
1.1  106 km2 (the area below
the 1100 m contour) some 
104 km3 of water would be
available for discharge per each 10 meters of outlet incision.
If the outlet incision was initially more rapid than the
decline of the Eridania basin?s water level (e.g., if surface
materials were poorly consolidated relative to underlying
bedrock), the discharge would increase for some time, and a
peak flood would occur. Increasing resistance of deeper,
better-consolidated materials would then retard the incision
rate, and the paleolake?s water level could decline, reducing
the flow depth and discharge. The lower reach terraces
occupy a lower level (300 to 700 m) than the interme-
diate basin floor (+400 m), so the high discharge from the
Eridania basin probably outlived the peak discharge from
the intermediate basin.
3.4. Development of Western Anastomosing Tributary
[49] The simultaneous activity of two overflow points,
the Ma?adim Vallis main channel and its large anastomosing
tributary, contributed to the peak discharge, but more rapid
incision of the eastern outlet ultimately left the western
outlet abandoned above the falling lake surface. The large
western tributary originated at the western gap in the
Eridania basin rim (Figure 1, point O2; Figure 4, point 1).
This western overflow immediately divided into two anas-
tomosing channels near the ridge crest (Figure 4, points 2),
which suggests that the sloping topography at this site was
unable to confine a large flow. Tributary flows initially ran
out onto the intermediate basin floor, incising the western-
most paleochannels (points 3), and the water also over-
topped the rim of a degraded 21 km impact crater on the
intermediate basin floor (point 4). This crater captured the
tributary flow, so that the westernmost tributary course was
abandoned early. Upon exiting this small crater, unconfined
water again flowed across the intermediate basin floor
toward the site of the main Ma?adim channel flows, and
Figure 14. Terraces (T) and erosion of the lower reach of
Ma?adim Vallis in THEMIS daytime IR imaging. The rim of
the crater G was eroded during the terrace-level flooding
and then the rim remnant confined the flow as the valley
incised its present inner channel. Gullies dissect the terraces
and valley walls, and space-filling valley networks (V)
dissect sloping terrain in the watershed. Plains surfaces (P)
are poorly dissected, but they collected water to incise stem
valleys headward. Flat areas remained undissected (U). The
crater C2 is rimless, which allowed the crater to be breached
widely by the inner channel (C). The former C2 rim did not
divert flows around the crater, as nearby low-lying ground
at U is undisturbed.
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
21 of 33
E12009
two lower tributary reaches were incised (points 5 and 6).
As these reaches down-cut and advanced headward, they
confined the flows crosscutting the 21 km crater. The
eastern lower tributary reach (point 6) crosscuts the western
lower reaches (points 3 and 5), suggesting that the eastern
reach ultimately captured the entire flow. Layered bar
deposits (Figure 15) are preserved mainly at the confluence
of this easternmost tributary reach and the main valley,
supporting its relatively late activity. In this manner, one
source point fed anastomosing tributary channels with three
outlets into Ma?adim Vallis.
[50] As the larger tributary valleys are graded to the
present floor of the main channel at 400 m elevation,
the Ma?adim lower and intermediate reaches must have
incised to this level either before the tributary flow oc-
curred, or contemporary with the tributary development.
Furthermore, a depositional bar and a streamlined ridge on
the main channel floor appear to have been influenced by
large flows from both the main channel and eastern distrib-
utary simultaneously (Figure 15). These observations sup-
port the overland flow sequence whereby the lower reach of
Ma?adim was incised first, draining water from the inter-
mediate basin and then allowing the upper reach of the main
channel and its broad western tributary to incise. At this
point, tributary flow declined and the western overflow
point was abandoned. The other possibility is that the
western tributary flow initiated after the eastern outlet
discharge had already done most of the work, but this
scenario requires a rapidly increasing Eridania paleolake
level during the overflow flood and/or some means (e.g.,
ice) of temporarily damming the eastern outlet. The added
complexity of such a scenario does not seem required by
observations, but it cannot be ruled out at present.
3.5. Late-Stage Incision
[51] The final stage of Ma?adim Vallis incision involved
only water from the eastern source point (O1 in Figure 1),
which cut the upper reach of the main valley 
300 m below
the upper reaches of the western tributary. This late stage
also incised and regraded the lower reach to a 1550 m
terminus at Gusev crater. As is typical of craters emplaced
on a regional slope (the dichotomy boundary), the southern
rim of Gusev crater was higher than its northern, downslope
rim, which stood at <1250 m elevation prior to the flood.
A Gusev crater paleolake, supplied by Ma?adim Vallis,
could only provide base level control once the lower reach
of Ma?adim Vallis (including the entrance breach of Gusev
crater), had incised to 1250 m. Exit breaching of Gusev
crater was required prior to the final 
300 m of base level
decline and incision of the Ma?adim Vallis inner channel.
The modern exit breach on the northwestern side of Gusev
is incised to1550 m, which corresponds with the1550 m
level of the putative Gusev delta [Schneeberger, 1989;
Grin and Cabrol, 1997]. The 1550 m base level at Gusev
and a postpeak discharge were active as the wide inner
channel was graded to its nearly constant slope of 0.0012
(Figure 3).
3.6. Alternative Histories
[52] Cabrol et al. [1996] proposed a sequence of events
for Ma?adim Vallis involving a long-lived valley and mul-
tiple disruptions by impacts. In their model, Ma?adim Vallis
Figure 15. (a) Medial bar at the confluence of the main
valley and western tributary in mosaic of THEMIS
V05612003 and V05250002. The bar is 1?2 km wide
and stands 120 m above the floor of the valley, which is

700 m deep at this location. (b) At the white arrow in
Figure 15a, a 1.4 km impact crater excavated part of the bar,
exposing visible layers at the 3.5 m resolution of MOC
image E05-01177.
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
22 of 33
E12009
Figure 16
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
23 of 33
E12009
first debouched to Gusev crater along its western rampart
before entering Gusev at the present entrance breach. Present
topography show that this scenario is not possible, as no
breach occurs through the confining ridge G in Figure 13b.
After the valley was incised, they suggest that two 30 km
craters (C1 and C2 in Figures 13 and 16) interrupted
Ma?adim Vallis, creating temporary intravalley lakes. Given
the detailed imaging and topography that is now available,
we find no evidence that these craters are younger than
Ma?adim Vallis. In the case of crater C1 (Downe/New
Plymouth), the higher preexisting ridges G and H in
Figure 13 would simply direct water across the C1 crater
rim and into Gusev, so there is no need for a putative
disrupted Ma?adim Vallis to deliver water to breach this
crater rim. In the case of the crater C2, its rim (in a fresh
state) would have redirected flow through a broad low point
to the west of the crater (near U in Figure 14), but the site is
undissected. As flows in fact crosscut the crater rather than
diverting to the west, the C2 crater rim must have already
been eroded when overland flow reached this site. The
modern C2 crater profile is rimless (Figure 14). Cabrol et
al. [1996] used radar data to determine the valley?s longitu-
dinal profile, which appeared to have a reverse slope along
part of its length. They attributed this reverse slope to
sedimentation near the crater dam. In MOLA data, this
anomaly does not occur (Figure 3), and sedimentation
initially occurs primarily at the head of a lake, not near its
dam. For these reasons and given consistent crater counts,
we find no compelling evidence that Ma?adim Vallis evolved
over a long period, with interruptions as described in earlier
papers. The relationship between the valley course and other
topography, including breached impact craters, is entirely
consistent with brief development during an overland flood.
3.7. Gusev Crater Deposits
[53] Thorough descriptions of the Gusev crater floor
morphology have been prepared for the MER mission,
detailing the physiographic units [Kuzmin et al., 2000;
Milam et al., 2003] and features related to the crater?s wind
regime [Greeley et al., 2003]. We therefore limit our
discussion to the floor stratigraphy. The exposed floor
materials contain at least two major layers, which are
evident from high-resolution visible and infrared imaging
(Figure 16). The upper layer is discontinuous and has a
variable thickness up to tens of meters. Wind has recently
stripped this apparently friable unit in some places to form
etched terrain, with enclosed depressions some tens of
meters deep, and at other locations it may have been
removed completely. In etched terrains, the surface layer
appears to have a more resistant cap, which favors degra-
dation by scarp backwasting prior to deflation of the
material. The thin surficial layer may be a lacustrine, ash,
or other air fall deposit.
[54] Removal of this layer and gardening by craters of
>200 m diameter has exposed a deeper layer with higher
thermal inertia (Figure 16). The gardened material has
capped the friable material to form a partial pedestal crater
at one site (Figure 16b). This underlying layer is thick
enough to make a significant contribution to crater ejecta,
but it is too coarse to be deflated significantly by wind. This
deeper material may be a coarse deposit from Ma?adim
Vallis, a layer of basalt, or one material overlying the other
as shown in Figure 16d. Early results from the Spirit rover
suggest basalt at the surface [Grant et al., 2004], which may
locally overlie the friable layer, or the friable material may
be stripped at the landing site to expose underlying basalt.
These hypotheses were considered among others at the
landing site selection meetings before the MER launch, as
the Apollinaris Patera volcano is located immediately to the
north of Gusev. Regardless of the nature of the surface layer,
the volume of sediment deposited in Gusev crater
(
15,100 km3 [Carter et al., 2001]) closely matches the
volume eroded from Ma?adim Vallis (
14,000 km3), so
fluvial sediments may comprise the greater part of the Gusev
fill material. Gusev likely acted as a terminal basin or a
runoff detention and settling pond for any flows that entered
the crater [Schneeberger, 1989; Grin and Cabrol, 1997].
4. Flood Magnitude
4.1. Total Volume Discharged
[55] The Eridania basin overflow can be characterized by
the total volume of the flood and by the dominant discharge
that was responsible for the valley geometry. Although the
Eridania basin topography might allow a straightforward
calculation of total overflow volume, tectonic activity, air
fall deposits, and aeolian deflation (discussed below) have
modified the basin after the flood, and a precise indication
of the former shoreline (our estimate is 
1100?1250 m as
discussed above) is not available given subsequent resur-
facing. Here we bracket the total discharge that would be
available from a single overflow of the Eridania basin to test
the hypothesis that the valley could be carved in a single
flood, although multiple overflows may have occurred.
These results offer greater precision than those presented
by Irwin et al. [2002], and we now incorporate a revised
estimate of lake levels associated with the western anasto-
mosing tributary.
[56] For a conservative estimate of the total volume
discharged from the Eridania basin, we assume that the
Figure 16. The lower reach of Ma?adim Vallis and Gusev crater. (a) THEMIS night IR image I01511006, showing
patches of coarse (thermally bright) materials in the ejecta of craters superposed on the Gusev floor. Patches of coarse
material are also exposed to the east of an impact crater (white arrow). (b) These exposures of coarse material correspond to
the low areas within etched terrain, where a superjacent layer tens of meters thick has been stripped by wind (white arrow;
MOC image E17-00827). (c) Context of Gusev crater in daytime infrared, showing the location of thermally dark etched
terrain shown in Figures 16a and 16b (white arrow). This coarse material is also exposed in a broad area in the southeast
part of the crater, again corresponding to the location of etched terrain [Milam et al., 2003]. These images have not been
radiometrically or geometrically calibrated. Craters C1 and C2 from Figure 13 are indicated for reference. (d) Inferred
stratigraphy of Gusev crater where the etched layer is present. A thin surface mantle, tens of meters thick, locally overlies
coarser (volcanic possibly overlying fluvial) materials that cannot be transported by wind.
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
24 of 33
E12009
lake initially occupied a uniform level of 1100 m, relative to
present topography, and that the intermediate basin was
initially dry. At this level, surface water would be available
to Ma?adim Vallis from the eastern, northwestern, and
southwestern subbasins (Figure 2). The southwestern sub-
basin would become isolated as the level declined below
1050 m, the northwestern subbasin would be closed at
1035 m, and the eastern subbasins alone would provide
87,000 km3 of water to carve Ma?adim Vallis before the
950 m level was reached (Table 4). By this method, a total
of 114,000 km3 of water was delivered from the Eridania
basin, consistent with a water/sediment volume ratio of 8:1
if a single flood excavated the total Ma?adim Vallis volume
of 14,000 km3. More conservatively, if contemporary
topography was such that the three largest subbasins were
not integrated at the time of the overtopping, and if the
intermediate basin were initially dry, enough water was
available in the eastern subbasins alone to carve Ma?adim
Vallis, with a water/sediment ratio of at least 6:1 by volume,
above the minimum ratio of 2.5:1 determined by Komar
[1980].
[57] A more liberal estimate of flow volume assumes that
the intermediate basin also contained a paleolake, with up to
60,000 km3 of water below its 1000 m overflow level, and
that the water level was as high as 1250 m in the eastern
subbasins, relative to present topography. Adding this
150 m thick layer of water onto the 1100 m contour in
the eastern subbasins (area 
625,000 km2), yields an upper
limit of 94,000 km3 of additional water. Therefore, depend-
ing on assumptions, 87,000 to 268,000 km3 of water was
available to carve Ma?adim Vallis, yielding a reasonable
water/sediment volume ratio on the order of 
10:1. Aquifer
contributions may have also provided water to the eastern
subbasins, extending the tail end of the flood.
4.2. Inner Channel Discharge
[58] As discussed in section 3.3, the discharge from the
Eridania basin was likely variable during the overflow. Peak
flood discharge calculations would have considerable error,
as later incision of the inner channel has altered the
geometry of Ma?adim Vallis, and stage indicators have not
been observed for the peak flow when much of the
excavation may have occurred. The inner channel is a more
suitable site for discharge calculations, as the inner margins
of the lower reach terraces provide an upper limit for the
depth of confined flows. The nearly constant gradient of the
lower reach (Figure 3) suggests that discharge was constant
with distance from the head, consistent with a paleolake
overflow. In channels where discharge increases with con-
tributing area, the sediment transport capacity also increases
with area, and progressively lower gradients are required in
the down-valley direction to transport the load [Gilbert,
1877; Davis, 1902].
[59] Cabrol et al. [1998b] concluded that the wide
Ma?adim Vallis floor was not a channel, because large
flows were inconsistent with their interpretation of the
valley?s longevity. Cabrol et al. [1996] did, however, allow
for some large discharges associated with the release of
water from putative intravalley lakes. Landheim [1995] left
open the possibility of a single large flow with later activity
of tributaries. We find no compelling mechanism to explain
the 3?5 km width of the valley floor unless a flow occupied
its entire width. Fluvial dissection of the valley sidewalls
was limited during and after incision of Ma?adim Vallis, in
contrast to the Grand Canyon in the southwestern U.S.,
which is of similar size but has deeply dissected walls. This
poor dissection makes it unlikely that water generated
within the intermediate basin widened a narrow precursor
Ma?adim Vallis over time to 15?25 km width. Few mass
movements crosscut the usually shallow sidewall gullies, so
postfluvial mass wasting does not explain the width, and
furthermore the terraces and medial floor ridges are well
preserved (Figures 14 and 15). The valley floor maintains
much of its 
5 km width as it crosscuts impact crater rims,
which are relatively resistant materials that have diverted
the valley course somewhat in other locations (Figures 14
and 16c). Somewhat wider areas of the valley floor are
associated with likely less resistant sedimentary materials,
such as crater interior deposits (see C2 in Figure 16).
Ma?adim Vallis narrows only where its slope increases,
including along the upper reach and at the knickpoint, as
expected for a channel that could freely adjust its cross-
sectional area to suit a constant discharge and variable flow
velocity. The western tributary is anastomosing without a
sediment source above its head (Figure 4), suggesting a
flow that the antecedent topography could not confine. The
main valley and tributary are both 
3 km wide at their
heads, unlike other valleys in the region. A layered medial
bar at the confluence of the main valley and tributary stands

120 m above the valley floor (Figure 15). The inner
channel migrates to the outer margin of most bends in the
valley?s lower reach (Figure 13b), as expected for a channel
incising along the thalweg of the terrace level (peak flood
stage). Thalwegs are typically located along the outer
margin of channel bends. The lower reach exhibits a step-
pool sequence, with pools up to tens of meters deep
(Figure 3b), which would require deep flows for excavation.
The possible deltaic deposit in Gusev crater is small relative
to the volume of Ma?adim Vallis (Figure 16b). Catastrophic
floods typically do not leave deltas, as their velocities are
adequate to maintain sediment transport well into the
interior of their depositional basins, so the delta develop-
Table 4. Estimated Volumes Discharged to Ma?adim Vallis From the Eastern, Northwest, and Southwest Paleolake
Subbasins Filled to 1100 m and the Intermediate Basin Filled to 1000 m
Subbasin
Volume at 1,100 m
or Full Level, km3
Volume at 950 m Level
or When Isolated, km3
Volume Available
for Discharge, km3
Eastern subbasins 305,000 218,000 (at 950 m) 87,000
Northwest 58,000 48,000 (at 1035 m) 10,000
Southwest 139,000 122,000 (at 1050 m) 17,000
Intermediate basin 60,000 (at 1000 m) 0 (breached) 60,000
Total 562,000 388,000 174,000
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
25 of 33
E12009
ment probably occurred only near the end of the flood. The
low settling velocity on Mars relative to flow velocity
[Pieri, 1980] would facilitate transport of suspended load
into Gusev relative to a similar flow on Earth. The inner
channel width is therefore likely related to the same flows
that transported material through the valley.
[60] The above comparisons are qualitative, but we can
test the paleolake overflow hypothesis by quantitatively
comparing the inner channel of Ma?adim Vallis to terrestrial
river channels. First, terrestrial river valleys have variable
width/depth ratios, but an approximation can be made using
empirical relationships between width W, depth H, and
bankfull discharge Qb (m
3/s) as
Qb ? W=Kw? ?2? H=Kh? ?2:8; ?3?
where Kw and Kh are coefficients with one standard
deviation ranges of 2.5?4.8 and 0.26?0.56, respectively
[Knighton, 1987]. By taking the midpoints of these ranges,
Kw = 3.65 and Kh = 0.41, a relationship between W and H
can be determined:
H ? 0:16W 0:72: ?4?
Equation (4) suggests that a channel 5 km wide should have
a depth of approximately 70 meters, give or take a factor of

2, so it would be 
70 times as wide as it is deep. The
inner terrace margins, where scarps bound the inner
channel, are as low as 
100 m above the valley floor and
are appropriate to confine such a flow. Bend wavelength l
varies linearly with channel width as
l ? KlW ; ?5?
where Kl 
 7?14. The shortest bend wavelength in
Ma?adim Vallis is 
37 km, in the intermediate reach of the
valley, so Kl > 7, which is consistent with the range of l/W
observations from terrestrial channels. The interior floor of
Ma?adim Vallis therefore has geometry consistent with a
fluvial channel.
[61] Initial estimates of discharge were made using em-
pirical relationships from terrestrial channels, where bank-
full discharge (Qb) is the dominant control on channel width
[Knighton, 1998]. Here we make the assumption that a large
flow could freely modify its channel banks, so a relationship
between width and discharge would pertain to the inner
channel of an entrenched valley [see Montgomery and
Gran, 2001]. This assumption is supported by the nearly
constant width of the inner channel. Discharge can be scaled
to Mars using a unit-balanced version of equation (3) from
stability analyses of terrestrial meandering rivers [Anderson,
1967; Parker, 1976], where Kw has units of (s/m)
0.5 and g
and H are relative gravity and depth, respectively:
Qb ?
W
Kw
 2
gH? ?1=2: ?6?
Depth scales with the 0.23 power of gravity (G. Parker,
University of Minnesota, personal communication, 2003),
so Q  1  106 m3/s. A combination of the Manning and
continuity equations can also be used for a discharge
estimate in channels that are much wider than deep, with the
assumption of steady, uniform flow:
Q ? H5=3S1=2g1=2Wn1; ?7?
where H is depth, S is slope (for the lower reach, S =
0.0012), g is relative gravity, and n is the Manning
roughness coefficient. We assumed n = 0.035, comparable
to that used by O?Connor [1993] for the Bonneville
outflow. When combined with the gravitational scaling
term in equation (7), this value of n is consistent with the
results of Wilson et al. [2004] (0.035/0.381/2 = 0.0565, much
like their result of 0.0545), and the natural range of n
would not alter the results by more than a factor of two in
either direction. Discharge would be approximately 2?7 
106 m3/s for postulated flows of 50?100 meters deep. At
the confluence of the main and large tributary valleys, a
sedimentary bar is located within the channel, which
also appears to have confined main valley flows. Using
equation (7), we confirmed that a 5  106 m3/s discharge
would pass east of the bar, approaching the crest elevation
without overtopping it.
[62] Finally, we used the step-backwater method [e.g.,
Webb and Jarrett, 2002; Brunner, 2002] to model flows of
this order of magnitude, beginning with 30 cross sections
along the lower reach. The computational procedure is as
follows:
[63] 1. Assume a water surface and an appropriate dis-
charge at the downstream cross section 1 and upstream
cross section 2.
[64] 2. Measure hydraulic radius and cross-sectional area
of the flow.
[65] 3. Calculate velocity (equation (8)), conveyance
(equation (9)), and energy gradient (equation (10)).
[66] 4. Determine the energy loss between cross section 2
and cross section 1 (equations (11) and (12)).
[67] 5. Calculate the water surface elevation for the
upstream cross section.
[68] 6. Vary the assumed upstream water surface itera-
tively to approach the calculated value.
[69] The inner margins of lower reach terraces, the medial
bar elevation, and the empirical comparisons constrained
the water surface elevation and discharge assumptions. For
planetary applications, the step-backwater method suffers
from poorly constrained estimates of water surface elevation
and roughness, but the method can confirm that for a given
roughness and discharge, the proposed channel is able to
dissipate gravitational energy.
[70] Mean flow velocity V (m/s) is calculated from
assumed Q and measured cross-sectional area A (m2) by
V ? Q=A: ?8?
The conveyance C is
C ? 1=n? ?A R2=3g1=2: ?9?
where R (m) is the hydraulic radius (R = A/P where P is the
wetted perimeter). The energy gradient S between upstream
(cross section 2) and downstream (cross section 1) points is
S ? Q2=C2: ?10?
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
26 of 33
E12009
The energy equation predicts head losses between cross
sections:
WS2 ?
a2V 22
2g
? WS1 ?
a1V 21
2g
? He; ?11?
where at the upstream cross section, WS2 is the water surface
elevation, a2 is a scaling coefficient assumed equal to 1
given no overbank flow, V2 is mean velocity, and g is
gravitational acceleration. The right side of the equation
represents values for the downstream cross section, where
He is energy head loss (m) between the sections:
He ? LS ? Kc
a2V 22
2g
 a1V
2
1
2g
 
; ?12?
L is the distance between cross sections (here 5 km, the
approximate channel width), and Kc is an expansion or
contraction loss coefficient, equal to 0.1 for contracting
reaches or 0.3 for expanding reaches [Brunner, 2002].
[71] For modeled flows of 
2  106 m3/s, the lower
reach geometry affords a reasonable energy head loss with
distance, such that the predicted water surface increases at
upstream cross sections, without requiring upstream
declines in gradient to balance energy. At the steps in the
lower reach (Figure 3b), the model fails as the Froude
number F rises above one, where flow becomes supercrit-
ical (rapid) given these discharges.
F ? V= gH? ?1=2: ?13?
The fact that the model predicts rapid flow at steps
(followed by pools) in the longitudinal profile is another
consistency with channel morphology created by a large
discharge. It is important to note that these steps are likely
not sapping headwalls. They are not scarps and do not
appear in imaging, but are higher-than-average gradient
reaches that were identified only in the topographic data.
The succession of enclosed pools below these steps is
inconsistent with origin by small flows. We did not model
flow headward of the lower reach, as this discharge would
become supercritical again over the knickpoint, and the
upper reach lacks terraces to constrain the flow depth.
[72] Discharges of 1?5  106 m3/s are on the low end of
estimates for the Martian outflow channels (commonly 1?2
orders of magnitude greater) [Carr, 1979, 1996; Baker,
1982] and similar to the terrestrial Lake Missoula floods
[Baker, 1982] and Athabasca Vallis [Burr et al., 2002],
but they are somewhat higher than the 1  106 m3/s
peak discharge calculated for the Lake Bonneville flood
[O?Connor, 1993]. The Bonneville flood is a good example
of an overflowing lake producing discharge on this order of
magnitude, however. Peak discharge, at the terrace level of
the Ma?adim Vallis lower reach, may have been consider-
ably higher than this range. At a possible average 5 
106 m3/s, 87,000 to 268,000 km3 of water would be
discharged from the Eridania basin in 
201 to 620 days.
The actual duration of the flood depends on precise knowl-
edge of paleolake volumes and the change in discharge with
time, but the flood may have lasted less than one Martian
year. This estimate does not constrain the duration of low-
discharge fluvial activity within Ma?adim Vallis, which was
supplied mainly by the tributary network within and north
of the intermediate basin. This later fluvial activity did not
significantly modify the trunk valley shape or its interior
landforms, however.
4.3. Sediment Transport
[73] The near-constant slope of the lower reach (
0.0012)
and width of the inner valley (
5 km) suggest that Ma?adim
Vallis was able to conform its channel geometry to flow
conditions. This requires that the substrate be easily eroded
by the available discharge, except at the knickpoint, where a
likely bedrock outcrop resisted erosion. Sediment transport
rates and the diameter of transportable particles are func-
tionally related to bed shear stress t (N/m2),
t ? rgRS; ?14?
where r is fluid density (kg/m3). For Q = 5  106 m3/s, bed
shear stresses on the order of 
400 N/m2 are averages for
the low-gradient intermediate and lower reaches. Pools
would have lower shear stresses of 100?300 N/m2, whereas
shear stresses for steps, the steep upper reach and knickpoint
(Figure 3) occupied the range of 500?1000 N/m2. For the
low-gradient reaches, particles <
1 cm diameter would be
carried as suspended load, and particles as large as 
50 cm
would be carried as bed load [Komar, 1980]. In high-
gradient reaches, particles as large as one meter could be
carried as bed load, although considerable uncertainty
applies to the bed load/suspended load threshold for
>10 cm particles [Komatsu and Baker, 1997]. With the
flood able to move sediment of this size, only large bedrock
blocks along the Ma?adim Vallis course would be able to
resist the flow, which could otherwise modify its channel
relatively freely.
[74] Presently, the channel floor is relatively flat in cross
section along the lower reach. Steeper reaches of the
channel, including the upper reach, knickpoint, and smaller
steps, typically have concave-up floor cross sections. The
knickpoint probably represents a bedrock outcrop, as its
almost V-shaped cross section resembles that of terrestrial
bedrock valleys. The overall form and longitudinal profile
are very similar to knickpoints developed experimentally in
uniformly moderately resistant material, where actual shear
stress is only moderately greater than critical shear stress for
erosion [Gardner, 1983]. Otherwise, Ma?adim Vallis appears
to be incised into less well consolidated materials, including
impact ejecta, megaregolith, and sedimentary deposits.
5. Post-Noachian Geomorphology
5.1. Eridania Basin
[75] Post-Noachian air fall materials profoundly affected
the Eridania basin morphology, but cratered highland sur-
faces immediately to the north (and south of Medusae
Fossae materials) were not significantly altered. The youn-
gest of these materials is a 
1?10 m thick air fall deposit
that mantles the impact-gardened surface south of 
35
S
[Mustard et al., 2001], including the southern portion of the
Eridania basin. Partial deflation of this layer has created
patches of smooth and pitted surfaces, which dominate
the meter-scale surface texture where the deposit occurs
(Figure 17). This layer is thought to have originated in the
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
27 of 33
E12009
last 105 years as ice-rich loess [Mustard et al., 2001], with
similar cycles of aeolian deposition and erosion extending
over at least 106 to 107 years. By calculating differential
slopes along MOLA data tracks, Kreslavsky and Head
[2000] noted that mantling deposits had smoothed this
region on 0.6 km length scales, but not adequately to
modify roughness on 19.2 km scales, relative to equatorial
highland topography. Mantling materials in this area were
originally recognized by Soderblom et al. [1973] at kilo-
meter resolution, coinciding approximately with the air fall
materials identified in high-resolution imaging [Mustard et
al., 2001] and topography [Kreslavsky and Head, 2000].
[76] In the Eridania basin, the most significant of these
layers is the Hesperian ??Electris deposit?? mapped by Grant
and Schultz [1990], which overlies the southern margin of
the Eridania basin watershed and parts of the basin floor
(see their Figure 3). This layer is at least 
100?400 m
thick, based on MOLA topographic profiles where its lower
contact appears in crater walls or at the base of eroded
scarps or mesas (Figure 18). As it mantles the Eridania basin
and watershed surface irrespective of elevation, the Electris
deposit is likely composed of air fall materials [Grant and
Schultz, 1990]. The surface of the deposit appears durable at
present, as it does not exhibit yardangs or other landforms
attributable to aeolian deflation, and superposed valley
networks often exhibit knickpoints where they compromise
its surface. These valley networks are sparse, with broad
undissected interfluves (Figure 18), in contrast to unmantled
slopes north of 
30
S, where networks incised into older
slopes are commonly denser. As erosion of the Electris
material did not leave pedestal impact craters, Grant and
Schultz [1990] concluded that the deposit was emplaced and
partially stripped during a short interval, dated to the middle
of the Hesperian period using N(5) crater counts.
[77] In the eastern subbasins, the air fall materials are
often exposed as a scarp overlying the plains of the Eridania
basin floor (Figure 18a). The deepest parts of the Eridania
basin depressions may have experienced a late episode of
lacustrine activity (or a terminal stage of the paleolake we
discuss here) near the close of the Hesperian, when flat
terraces were deposited around the center of the Gorgonum
subbasin at a constant level of 300 m. These terraces
overlie or onlap the eroded Gorgonum chaotic terrain mesas
and maintain a pristine appearance [Howard and Moore,
2004]. Evidence for post-Hesperian aqueous processes is
limited in this area. Very recent, fresh-appearing gullies
occur along the sides of some chaotic terrain mesas, fresh
crater walls, and graben in the Eridania basin [Malin and
Edgett, 2000b], but these gullies are spatially confined and
have not had much geomorphic effect. As Noachian surfaces
are saturated with impact craters of 
40 m diameter, impact
gardening has reworked all Noachian surfaces to at least

8 m depth [Hartmann et al., 2001] (Figure 12). This
diffusive process would be effective at removing all
Noachian landforms on the order of 
100 m width and
Figure 17. Subkilometer morphology in the Eridania
basin south of 
35
S, representing post-Noachian resurfac-
ing (subsets of MOC image M02-03267, centered at
39.18
S, 196.04
W). The left image is adjacent to and
north of the right image. (a) Note the sparse small craters
and rough, pitted surface, in contrast with the impact-
gardened surface in Figure 12. Fresh gullies occur in some
crater interiors, while older valleys are degraded where they
debouched to the Eridania basin plains (bottom right).
(b) Enlargement of white inset box in Figure 5a. The rough,
pitted texture may be related to episodes of deposition and
subsequent deflation of ice-rich loess [Mustard et al., 2001],
obfuscating any older, small-scale features.
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
28 of 33
E12009

10 m thickness. Tectonism also significantly modified the
basin topography after the Noachian. In addition to the
Sirenum Fossae and subperpendicular compressive ridges,
circular ridges developed on the Eridania basin floor are
thought to reflect compaction of plains over buried crater
topography [Watters, 1993].
5.2. Ma?adim Vallis Watershed
[78] Longitudinal profiles of small Ma?adim Vallis
tributaries (within the intermediate basin) suggest that
fluvial activity modified the landscape somewhat, but not
significantly, after the overflow. Where the surrounding
terrain maintains adequately long and steep slopes, dense
drainage networks exist down to the limit of THEMIS
resolution (100 m/pixel; Figure 14) and extend to drainage
divides, likely indicating that precipitation and runoff mod-
ified this landscape. Low-gradient surfaces are more com-
monly undissected over a range of elevations. Several of
these tributaries have incised deep canyons along their
lower reaches, which dissect the terraces and down-cut to
the Ma?adim Vallis floor (Figure 14). This incision therefore
postdates the complete incision of the main valley. At least
three tributaries within the intermediate basin are hanging,
in that they emerge well above the floor of the valley and
exhibit a scarp but no incised knickpoint at the confluence.
These tributaries, including the westernmost flood tributary
course (the western point 3 in Figure 4a), are likely
abandoned paleochannels that developed when the interme-
diate basin was draining, but before the intermediate reach
was fully incised to its present level.
[79] Ideally, mature tributaries would develop graded
(concave-up) longitudinal profiles where discharge increases
with contributing area (i.e., in humid areas), such that
downstream reaches have progressively greater discharge
and can transport the sediment load under lower gradients
[Gilbert, 1877; Davis, 1902]. The younger valley networks
draining the intermediate basin (tributaries directly to
Figure 18. The 
100?400 m thick, Hesperian, Electris air
fall deposit [Grant and Schultz, 1990] dominates kilometer-
scale morphology along the southern margin of the Eridania
basin watershed, where it has buried Noachian landforms.
Note the relatively low density of small degraded craters
and valley networks. (a) The margin of the deposit where it
overlies the more durable ridged plains of the Eridania basin
floor (subset of THEMIS daytime IR image I01168002,
centered at 38.2
S, 174.5
E). Broad valleys were incised
into the deposit, supplied by drainage from narrow upstream
channels located atop the deposit. A Sirenum Fossae graben
crosscuts the ridged plains at the top of the image. (b) More
advanced erosion of the Electris deposit, where retreat
of valley walls has isolated mesas (subset of THEMIS
daytime IR image I03490002, centered at 38.8
S, 168.8
E).
(c) Another example of narrow upstream valleys supplying
broad lower reaches (subset of THEMIS daytime IR image
I01168002, centered at 45.2
S, 173.5
E, located at the white
inset box in Figure 18d). (d) The floor elevation of these
broad reaches corresponds with terraces in nearby degraded
crater walls, suggesting that the terraces reflect the base of
weaker Electris materials (subset of MOC Geodesy
Campaign Mosaic).
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
29 of 33
E12009
Ma?adim Vallis), in contrast, are poorly graded (Figure 19).
The 17 tributaries we examined have convex-up longitudi-
nal profiles, which commonly develop in fluvial systems in
response to base level decline along the master stream. This
is reasonable for Ma?adim Vallis if the valley incised
rapidly relative to contemporary tributary incision rates,
or if it was emplaced in a very short time with limited
fluvial erosion afterward. These postflood tributaries may
have supplied a small channel that is incised into the floor
of Ma?adim Vallis near the mouth (across C1 in Figure 16c).
This postflood small channel appears only near Gusev and
is incised to below 1550 m, suggesting that the water
level had by then declined in the crater.
6. Conclusions
[80] The excavation of Ma?adim Vallis, Mars, is best
explained by (and probably requires) a catastrophic over-
flow of a 
1.1 M km2 paleolake located south of the
valley?s source spillways. The paleolake overflow hypoth-
esis is supported by the morphology and morphometry of
Ma?adim Vallis and its head basin, flow modeling, and
crater retention ages. This evidence includes the following:
[81] 1. Ma?adim Vallis originates at two incised gaps in
the rim of a 500 km degraded crater, identified as the
intermediate basin (Figure 1), which the valley crosscuts.
The eastern source point, which is more deeply incised, is
displaced 
25 km south of the original divide into the
plains of the enclosed, 1.1  106 km2 Eridania head basin.
This displacement is consistent with terrestrial damburst
floods and requires a large ponded water source within the
Eridania basin. No sapping headwall, elevated aquifer, or
contributing valley networks occur south of the valley
heads, which are located at a topographic crest.
[82] 2. At its headward termination in the Eridania basin,
Ma?adim Vallis has a floor width of 2.5?4 km, which is
commensurate with downstream dimensions and suggests a
large flow. The course of the western tributary is anasto-
mosing, indicating that the landscape was unable to confine
the large flows that breached the divide crest. The valley
walls are poorly dissected, so water generated within the
intermediate basin likely did not widen the valley over time,
and postfluvial degradation processes have not substantially
modified the valley?s interior terraces and gullies.
[83] 3. Flow discharge was estimated by channel width
and step-backwater modeling of a water surface below the
inner margins of terraces, all of which yield discharges on
the order of 1?5  106 m3/s. A peak flood with greater
discharge may have been responsible for eroding a wide
valley at the terrace level before the inner channel was
incised. Valley morphology and morphometry supports this
use of the valley floor as a channel. All craters that have
been breached by the valley have atypically wide gaps in
their rims, as would be necessary to accommodate a large
flow. Decreases in width are associated with increases in
gradient. Interior ridges and a layered medial bar of 120 m
height are consistent with flows that occupied the entire
floor width. The width/depth ratio, width/bend wavelength
ratio, and capacity to dissipate gravitational potential energy
are consistent with terrestrial channels. The gradient of the
lower reach is nearly constant, consistent with a constant
discharge in the down-valley direction, with the exception
of local steps and pools. The depth of these pools, up to tens
of meters, suggests a deep flow.
[84] 4. When the topography of the region is recon-
structed to before the incision of Ma?adim Vallis, the valley
Figure 19. Typical convex-up longitudinal profiles of
tributaries to Ma?adim Vallis. The tributaries become
deeply incised only near the confluence with Ma?adim. In
both cases, the channel ??head?? incises deeply into a
plateau, but shallow contributing valleys are present
upslope. Steps in the longitudinal profile appear where
raw MOLA shots are not available; these are not real
features of the valley. (a) Upstream tributary. (b) Down-
stream tributary. (c) Location of the tributaries (solid black
lines). The dashed line is the Ma?adim main channel, and
the white line is the drainage divide.
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
30 of 33
E12009
course is shown to be entirely consistent with flow spilling
over the preexisting landscape, exploiting low points in the
topography and allowing the intermediate basin to be
crosscut. No structural control of the valley course is
evident.
[85] 5. The tributaries to Ma?adim Vallis are poorly
graded to the main channel floor, as they have convex-up
longitudinal profiles. These characteristics are consistent
with rapid incision of the main valley relative to tributary
incision rates, or a sudden incision of the valley followed by
limited time for the adjustment of tributaries to the lowered
base level.
[86] 6. The Eridania basin could contain adequate water
(87,000 to 268,000 km3) to erode the 14,000 km3 volume
of Ma?adim Vallis, with a physically plausible water-to-
sediment volume ratio on the order of 10:1.
[87] 7. Crater counts suggest a similar age for the Eridania
basin floor and for fluvial activity in the Ma?adim Vallis
watershed (within the intermediate basin). Supported by
other observations, the crater counts do not require the
prolonged development of Ma?adim Vallis proposed by
Cabrol et al. [1998b] and Kuzmin et al. [2000]. The
paleolake overflow was contemporary with extensive fluvial
erosion of the surrounding landscape, which probably
required precipitation up to the Noachian/Hesperian bound-
ary (using N(5)) [Craddock and Maxwell, 1990, 1993;
Grant, 2000; Craddock and Howard, 2002; Irwin and
Howard, 2002; Hynek and Phillips, 2003]. Since the
Eridania basin is enclosed, lakes would be reasonable in
the lower parts of the basin, which are the lowest areas on
the Terra Cimmeria/Terra Sirenum plateau.
[88] 8. The Eridania basin contains six highly degraded
impact craters of 180?240 km diameter with concave
floors. Most Martian degraded craters have flat floors,
which developed as the craters infilled [Craddock and
Maxwell, 1993; Craddock and Howard, 2002], but the
highly degraded Eridania basin depressions retain much of
the craters? original interior relief. The topography of these
depressions is similar to the other nearby large craters,
Newton and Copernicus. The inward-sloping crater floors
suggest a less efficient sediment transport process below

700 m elevation, as occurs in standing water bodies. The
relief in these concave crater floors is inconsistent with
known volcanic surfaces on Mars, which are nearly flat or
slope away from a central high point.
[89] 9. Nearly all valley networks in the Eridania basin
watershed terminate between 1100 and 
700 m elevation,
approximately 1 km above the low points in the respective
drainage basins. The concave crater floors in the Eridania
basin have slopes of 
0.5?1.5
 below 700 m elevation, but
the slopes are not dissected by valley networks. Similar
slopes above 1100 m elevation are commonly dissected.
[90] 10. The boundary between the watershed and floor
plains of the Eridania basin is similar to the elevation of the
Ma?adim Vallis source breaches (
1100?1250 m), as might
be expected for a paleolake surface.
[91] The Eridania basin?s hypsometry and floor morphol-
ogy suggest that a paleolake occupied temporally variable
levels of 700?1250 m (modern topography may be some-
what modified by post-Noachian tectonism) during the time
of widespread fluvial activity. The rims of large craters in
the Eridania basin were extensively eroded above 
700 m,
leaving large gaps in the rims, but topographic relief from
700 to 400 m lacks signs of incision and efficient
sediment transport. We attribute the absence of valley
networks in these crater floors to protection by standing
water during the time that relatively intense fluvial erosion
occurred.
[92] The Ma?adim Vallis development can be summarized
in discrete stages as follows:
[93] 1. Near the Noachian/Hesperian boundary (using
>5 km craters), water levels in the Eridania basin rose due
to increased precipitation, reduced evaporation, and/or
influx of groundwater from melting circumpolar ice or
snow. A short-lived, contemporary lake may have formed
in the intermediate basin.
[94] 2. The Eridania basin overflowed to the intermediate
basin at two points at 
1200 to 1250 m elevation. The two-
point overflow was possible because the original gaps likely
had similar levels, and because available water from the 2?
3  106 km2 Eridania drainage basin exceeded the ability of
either of the outlets (which were not yet fully incised) to
discharge it. An impact into the lake may have facilitated
breaching of the divide at two locations. Early in the flood,
standing water protected the intermediate basin interior, so
only limited incision of the upper reach occurred before the
lower reach was incised.
[95] 3. The intermediate basin overflowed at 
1000 m
elevation, and the lower reach of Ma?adim Vallis was
incised to the terrace level (
500 m), below the level of
the intermediate basin floor (400 m). The intermediate reach
of the main valley and the lower reaches of the western
tributary were then incised.
[96] 4. Incision of the eastern outlet resulted in abandon-
ment of the western outlet. Continued flows predominantly
from the eastern outlet incised a constant-width inner
channel in the lower reach that matches the width of the
intermediate and upper reaches. This late stage discharge
was 
1?5  106 m3/s. The upper reach of the main channel
was fully incised to its present 950 m spillway, while the
western tributary, having been abandoned, did not incise to
the same extent. When the Eridania basin water declined to
this level, the flow terminated.
[97] 5. After the flood ceased, drainage networks reac-
tivated in the intermediate basin as tributaries to Ma?adim
Vallis. Overall dissection of the Ma?adim Vallis walls was
temporally limited, so that terraces and interior flow features
are well preserved, and the tributaries are poorly graded to
the main valley floor.
[98] 6. The Eridania basin paleolake level declined, and
the 100?400 m thick Hesperian Electris air fall deposit
[Grant and Schultz, 1990] was deposited along much of the
Eridania basin floor and the southern margin of its water-
shed. Hesperian valley networks dissected the Electris
deposit, which was then deflated from much of the Eridania
basin floor. Subsequent impact gardening and thinner air fall
deposits [Mustard et al., 2001] dominate the present mor-
phology of the Eridania basin.
[99] Acknowledgments. We gratefully acknowledge the constructive
dialogue and support from Bob Craddock, John Grant, and Tom Watters.
Ken Tanaka and Robert Brakenridge provided valuable reviews that
improved the manuscript, although Brakenridge prefers a mechanism to
form the valley involving a significant role for tectonism. The first author
thanks Clyde Brown for assistance with field work at Lawn Lake. This
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
31 of 33
E12009
project was supported by a NASA Mars Data Analysis Program grant to
Maxwell and a Planetary Geology and Geophysics grant to Howard.
References
Aharonson, O., M. T. Zuber, D. H. Rothman, N. Schorghofer, and K. X.
Whipple (2002), Drainage basins and channel incision on Mars, Proc.
Natl. Acad. Sci U. S. A., 99, 1780?1783.
Anderson, A. G. (1967), On the development of stream meanders, paper
presented at 12th Congress of the International Association for Hydraulic
Research, Fort Collins, Colorado.
Anderson, R. C., J. M. Dohm, M. P. Golombek, A. F. C. Haldemann, B. J.
Franklin, K. L. Tanaka, J. Luis, and B. Peer (2001), Primary centers and
secondary concentrations of tectonic activity through time in the western
hemisphere of Mars, J. Geophys. Res., 106, 20,563?20,585.
Baker, V. R. (1982), The Channels of Mars, Univ. of Tex. Press, Austin.
Barlow, N. G. (1990), Constraints on early events in Martian history as
derived from the cratering record, J. Geophys. Res., 95, 14,191?14,201.
Brakenridge, G. R. (1990), The origin of fluvial valleys and early geologic
history, Aeolis quadrangle, Mars, J. Geophys. Res., 95, 17,289?17,308.
Brunner, G. W. (2002), HEC-RAS River Analysis System Hydraulic Refer-
ence Manual, Version 3.1, CPD-69, Hydrol. Eng. Cent., U.S. Army
Corps of Eng.
Burr, D. M., J. A. Grier, A. S. McEwen, and L. P. Keszthelyi (2002),
Repeated aqueous flooding from the Cerberus Fossae: Evidence for very
recently extant, deep groundwater on Mars, Icarus, 159, 53?73.
Cabrol, N. A., and E. A. Grin (1999), Distribution, classification, and ages
of Martian impact crater lakes, Icarus, 142, 160?172.
Cabrol, N., R. Landheim, R. Greeley, and J. Farmer (1994), Fluvial pro-
cesses in Ma?adim Vallis and the potential of Gusev crater as a high
priority site (abstract), Proc. Lunar Planet. Sci. Conf. 25th, 213?214.
Cabrol, N. A., E. A. Grin, and G. Dawidowicz (1996), Ma?adim Vallis
revisited through new topographic data: Evidence for an ancient intraval-
ley lake, Icarus, 123, 269?283.
Cabrol, N. A., E. A. Grin, and G. Dawidowicz (1997), A model of outflow
generation by hydrothermal underpressure drainage in volcano-tectonic
environment, Shalbatana Vallis (Mars), Icarus, 125, 455?464.
Cabrol, N. A., E. A. Grin, R. Landheim, R. O. Kuzmin, and R. Greeley
(1998a), Duration of the Ma?adim Vallis/Gusev crater hydrogeologic
system, Mars, Icarus, 133, 98?108.
Cabrol, N. A., E. A. Grin, and R. Landheim (1998b), Ma?adim Vallis
evolution: Geometry and models of discharge rate, Icarus, 132, 362?
367.
Cabrol, N. A., E. A. Grin, and W. H. Pollard (2000), Possible frost mounds
in an ancient Martian lake bed, Icarus, 145, 91?107.
Carr, M. H. (1974), The role of lava erosion in the formation of lunar rilles
and Martian channels, Icarus, 22, 1?23.
Carr, M. H. (1979), Formation of Martian flood features by release of water
from confined aquifers, J. Geophys. Res., 84, 2995?3007.
Carr, M. H. (1996), Water on Mars, Oxford Univ. Press, New York.
Carr, M. H. (2002), Elevations of water-worn features on Mars, J. Geophys.
Res., 107(E12), 5131, doi:10.1029/2002JE001845.
Carr, M. H., and G. D. Clow (1981), Martian channels and valleys: Their
characteristics, distribution, and age, Icarus, 48, 91?117.
Carter, B. L., H. Frey, S. E. H. Sakimoto, and J. Roark (2001), Constraints
on Gusev basin infill from the Mars Orbiter Laser Altimeter (MOLA)
topography, Proc. Lunar Planet. Sci. Conf. 32nd, abstract 2042.
Clifford, S. M., and T. J. Parker (2001), The evolution of the Martian
hydrosphere: Implications for the fate of a primordial ocean and the
current state of the northern plains, Icarus, 154, 40?79.
Cohen, A. S. (2003), Paleolimnology: The History and Evolution of Lake
Systems, 500 pp., Oxford Univ. Press, New York.
Craddock, R. A., and A. D. Howard (2002), The case for rainfall on a
warm, wet early Mars, J. Geophys. Res., 107(E11), 5111, doi:10.1029/
2001JE001505.
Craddock, R. A., and T. A. Maxwell (1990), Resurfacing of the Martian
highlands in the Amenthes and Tyrrhena region, J. Geophys. Res., 95,
14,265?14,278.
Craddock, R. A., and T. A. Maxwell (1993), Geomorphic evolution of the
Martian highlands through ancient fluvial processes, J. Geophys. Res.,
98, 3453?3468.
Craddock, R. A., R. Greeley, and P. R. Christensen (1987), Martian outflow
channels: IRTM and visual observations (abstract), Proc. Lunar Planet.
Sci. Conf. 18th, 203?204.
Craddock, R. A., T. A. Maxwell, and A. D. Howard (1997), Crater mor-
phometry and modification in the Sinus Sabaeus and Margaritifer Sinus
regions of Mars, J. Geophys. Res., 102, 13,321?13,340.
Currey, D. R. (1990), Quaternary palaeolakes in the evolution of semidesert
basins, with special emphasis on Lake Bonneville and the Great Basin,
U.S.A., Palaeogeogr. Palaeoclimatol. Palaeoecol., 76, 189?214.
Davis, W. M. (1902), Base-level, grade, and peneplain, in Geographical
Essays, vol. 18, pp. 381?412, Ginn, Boston, Mass.
De Hon, R. A. (1977), Geologic map of the Eridania quadrangle of Mars,
U.S. Geol. Surv. Misc. Invest. Ser., Map I-1008, scale 1:5M.
French, H. M. (1976), The Periglacial Environment, 309 pp., Addison-
Wesley-Longman, Reading, Mass.
Gardner, T. W. (1983), Experimental study of knickpoint and longitudinal
profile evolution in cohesive, homogeneous material, Geol. Soc. Am.
Bull., 94, 664?672.
Garvin, J. B., S. E. H. Sakimoto, and J. J. Frawley (2003), Craters on Mars:
Global geometric properties from gridded MOLA topography, in 6th
International Conference on Mars, abstract 3277, Lunar and Planet. Inst.,
Houston, Tex.
Gilbert, G. K. (1877), Report on the geology of the Henry Mountains, U.S.
Geogr. Geol. Surv. Rocky Mountain Reg., U.S. Govt. Print. Off.,
Washington, D. C.
Goldspiel, J. M., and S. W. Squyres (1991), Ancient aqueous sedimentation
on Mars, Icarus, 89, 392?410.
Goldspiel, J. M., S. W. Squyres, M. A. Slade, R. F. Jurgens, and S. H. Zisk
(1993), Radar-derived topography of low southern latitudes of Mars,
Icarus, 106, 346?364.
Golombek, M. P., et al. (2003), Selection of the Mars Exploration Rover
landing sites, J. Geophys. Res., 108(E12), 8072, doi:10.1029/
2003JE002074.
Grant, J. A. (2000), Valley formation in Margaritifer Sinus, Mars, by
precipitation-recharged ground-water sapping, Geology, 28, 223?226.
Grant, J. A., and C. Fortezzo (2003), Basin hypsometry on the Earth, Mars,
and the Moon, in 6th International Conference on Mars, abstract 3050,
Lunar and Planet. Inst., Houston, Tex.
Grant, J. A., and P. H. Schultz (1990), Gradational epochs on Mars:
Evidence from west-northwest of Isidis Basin and Electris, Icarus, 84,
166?195.
Grant, J. A., and P. H. Schultz (1993), Degradation of selected terrestrial
and Martian impact craters, J. Geophys. Res., 98, 11,025?11,042.
Grant, J. A., et al. (2004), Surficial deposits at Gusev crater along Spirit
rover traverses, Science, 305, 807?810.
Greeley, R., and J. E. Guest (1987), Geologic map of the eastern equatorial
region of Mars, U.S. Geol. Surv. Misc. Invest. Ser., Map I-1802-B, scale
1:15M.
Greeley, R., R. O. Kuzmin, S. C. R. Rafkin, T. I. Michaels, and R. Haberle
(2003), Wind-related features in Gusev crater, Mars, J. Geophys. Res.,
108(E12), 8077, doi:10.1029/2002JE002006.
Grin, E. A., and N. A. Cabrol (1997), Limnologic analysis of Gusev crater
paleolake, Mars, Icarus, 130, 461?474.
Gulick, V. C. (2001), Origin of the valley networks on Mars: A hydrolo-
gical perspective, Geomorphology, 37, 241?268.
Hartmann, W. K., and G. Neukum (2001), Cratering chronology and the
evolution of Mars, Space Sci. Rev., 96, 165?194.
Hartmann, W. K., J. Anguita, M. A. de la Casa, D. C. Berman, and E. V.
Ryan (2001), Martian cratering 7: The role of impact gardening, Icarus,
149, 37?53.
Head, J. W., III, and S. Pratt (2001), Extensive Hesperian-aged south
polar ice sheet on Mars: Evidence for massive melting and retreat,
and lateral flow and ponding of meltwater, J. Geophys. Res., 106,
12,275?12,299.
Howard, A. D. (1998), Long profile development of bedrock channels:
Interaction of weathering, mass wasting, bed erosion, and sediment trans-
port, in Rivers Over Rock: Fluvial Processes in Bedrock Channels, Geo-
phys. Monogr. Ser., vol. 107, edited by K. Tinkler and E. Wohl, pp. 297?
319, AGU, Washington, D. C.
Howard, A. D. (2000), Ancient crater basins: A triad of morphologies, Eos
Trans. AGU, 81(48), Fall Meet. Suppl., abstract P62C-06.
Howard, A. D., and J. M. Moore (2004), Scarp-bounded benches in
Gorgonum Chaos, Mars: Formed beneath an ice-covered lake?, Geophys.
Res. Lett., 31, L01702, doi:10.1029/2003GL018925.
Howard, A. D., R. C. Kochel, and H. E. Holt (Eds.) (1988), Sapping
Features of the Colorado Plateau: A Comparative Planetary Geology
Field Guide, NASA Spec. Publ., NASA SP-491, 108 pp., NASA,
Washington, D. C.
Howard, J. H. (1979), Geologic map of the Phaethontis quadrangle of Mars,
U. S. Geol. Surv. Misc. Invest. Ser., Map I-1145, scale 1:5M.
Hynek, B. M., and R. J. Phillips (2003), New data reveal mature, integrated
drainage systems on Mars indicative of past precipitation, Geology, 31,
757?760.
Irwin, R. P., and A. D. Howard (2002), Drainage basin evolution in
Noachian Terra Cimmeria, Mars, J. Geophys. Res., 107(E7), 5056,
doi:10.1029/2001JE001818.
Irwin, R. P., T. A. Maxwell, A. D. Howard, R. A. Craddock, and D. W.
Leverington (2002), A large paleolake basin at the head of Ma?adim
Vallis, Mars, Science, 296, 2209?2212.
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
32 of 33
E12009
Jankowski, D. G., and S. W. Squyres (1993), ??Softened?? impact craters
on Mars: Implications for ground ice and the structure of the Martian
regolith, Icarus, 106, 365?379.
Jarrett, R. D., and J. E. Costa (1986), Hydrology, geomorphology, and dam-
break modeling of the July 15, 1982 Lawn Lake dam and Cascade Lake
dam failures, Larimer County, Colorado, U. S. Geol. Surv. Prof., 1369.
Knighton, A. D. (1987), River channel adjustment?The downstream di-
mension, in River Channels: Environment and Process, edited by K. S.
Richards, pp. 95?128, Basil Blackwell, New York.
Knighton, D. (1998), Fluvial Forms and Processes: A New Perspective,
Oxford Univ. Press, New York.
Komar, P. D. (1980), Modes of sediment transport in channelized water
flows with ramifications to the erosion of the Martian outflow channels,
Icarus, 42, 317?329.
Komatsu, G., and V. R. Baker (1997), Paleohydrology and flood geomor-
phology of Ares Vallis, J. Geophys. Res., 102, 4151?4160.
Kreslavsky, M. A., and J. W. Head (2000), Kilometer-scale roughness of
Mars: Results from MOLA data analysis, J. Geophys. Res., 105, 26,695?
26,711.
Kuzmin, R. O., R. Greeley, R. Landheim, N. A. Cabrol, and J. D. Farmer
(2000), Geologic map of the MTM -15182 and MTM -15187 quadran-
gles, Gusev crater?Ma?adim Vallis region, Mars, U. S. Geol. Surv. Geol.
Invest. Ser., Map I-2666, scale 1:500K.
Landheim, R. (1995), Exobiology sites for Mars exploration and geologic
mapping of Gusev crater-Ma?adim Vallis, M.S. thesis, 167 pp., Ariz.
State Univ., Tempe.
Landheim, R., N. A. Cabrol, R. Greeley, and J. D. Farmer (1994), Strati-
graphic assessment of Gusev crater as an exobiology landing site (ab-
stract), Proc. Lunar Planet. Sci. Conf. 25th, 769?770.
Lucchita, B. K. (1982), Preferential development of chaotic terrains on
sedimentary deposits, in Reports of the Planetary Geology Program,
NASA Tech. Memo., NASA TM-85127, 235?236.
Lucchita, B. K., and J. Dembosky (1994), Scarp heights of Martian chan-
nels from shadow measurements (abstract), Proc. Lunar Planet. Sci.
Conf. 25th, 811?812.
Malin, M. C. (1976), Age of Martian channels, in Studies of Martian
Geology, Ph.D. dissertation, chap. 2, pp. 48?73, Calif. Inst. of Technol.,
Pasadena.
Malin, M. C., and M. H. Carr (1999), Groundwater formation of Martian
valleys, Nature, 397, 589?591.
Malin, M. C., and K. S. Edgett (2000a), Sedimentary rocks of early Mars,
Science, 290, 1927?1937.
Malin, M. C., and K. S. Edgett (2000b), Evidence for recent groundwater
seepage and surface runoff on Mars, Science, 288, 2330?2335.
Masursky, H., J. M. Boyce, A. L. Dial, G. G. Schaber, and M. E. Strobell
(1977), Classification and time of formation of Martian channels based
on Viking data, J. Geophys. Res., 82, 4016?4038.
Masursky, H., A. L. Dial Jr., and M. E. Strobell (1980), Martian channels?
A late Viking view, in Reports of the Planetary Geology Program, NASA
Tech. Memo., NASA TM-82385, 184?187.
Maxwell, T. A., and G. E. McGill (1988), Ages of fracturing and resurfac-
ing in the Amenthes region, Mars, Proc. Lunar Planet. Sci. Conf. 18th,
701?711.
McCauley, J. F., M. H. Carr, J. A. Cutts, W. K. Hartmann, H. Masursky,
D. J. Milton, R. P. Sharp, and D. E. Wilhelms (1972), Preliminary
Mariner 9 report on the geology of Mars, Icarus, 17, 289?327.
McGetchin, T. R., M. Settle, and J. W. Head (1973), Radial thickness
variation in impact crater ejecta: Implications for lunar basin deposits,
Earth Planet. Sci. Lett., 20, 226?236.
Milam, K. A., K. R. Stockstill, J. E. Moersch, H. Y. McSween Jr., L. L.
Tornabene, A. Ghosh, M. B. Wyatt, and P. R. Christensen (2003),
THEMIS characterization of the MER Gusev crater landing site, J. Geo-
phys. Res., 108(E12), 8078, doi:10.1029/2002JE002023.
Milton, D. J. (1973), Water and processes of degradation in the Martian
landscape, J. Geophys. Res., 78, 4037?4047.
Montgomery, D. R., and K. B. Gran (2001), Downstream variations in the
width of bedrock channels, Water Resour. Res., 37, 1841?1846.
Moore, J. M., and A. D. Howard (2003), Ariadnes-Gorgonum knob fields
of north-western Terra Sirenum, Mars, Proc. Lunar Planet. Sci. Conf.
34th, abstract 1402.
Mustard, J. F., C. D. Cooper, and M. K. Rifkin (2001), Evidence for recent
climate change on Mars from the identification of youthful near-surface
ground ice, Nature, 412, 411?414.
Mutch, T. A., and E. C. Morris (1979), Geologic map of the Memnonia
quadrangle of Mars, U.S. Geol. Surv. Misc. Invest. Ser., Map I-1137, scale
1:5M.
O?Connor, J. E. (1993), Hydrology, hydraulics, and geomorphology of the
Bonneville Flood, Spec. Pap. Geol. Soc. Am., 274.
Parker, G. (1976), Cause and characteristic scales of meandering and braid-
ing in rivers, J. Fluid Mech., 76, 457?480.
Phillips, R. J., M. T. Zuber, S. C. Solomon, M. P. Golombek, B. M.
Jakosky, W. B. Banerdt, R. M. E. Williams, B. M. Hynek, O. Aharonson,
and S. A. Hauck II (2001), Ancient geodynamics and global-scale hy-
drology on Mars, Science, 291, 2587?2591.
Pieri, D. C. (1980), Geomorphology of Martian valleys, in Advances in
Planetary Geology, NASA Tech. Memo., NASA TM-81979, 1?160.
Ritter, D. F., R. C. Kochel, and J. R. Miller (1995), Process Geomorphol-
ogy, 3rd ed., McGraw-Hill, New York.
Schneeberger, D. M. (1989), Episodic channel activity at Ma?adim Vallis,
Mars (abstract), Proc. Lunar Planet. Sci. Conf. 20th, 964?965.
Schultz, R. A., and K. L. Tanaka (1994), Lithospheric-scale bucking and
thrust structures on Mars: The Coprates rise and south Tharsis ridge belt,
J. Geophys. Res., 99, 8371?8385.
Scott, D. H., and K. L. Tanaka (1986), Geologic map of the western equa-
torial region of Mars, U. S. Geol. Surv. Misc. Invest. Ser., Map I-1802-A,
scale 1:15M.
Scott, D. H., E. C. Morris, and M. N. West (1978), Geologic map of
the Aeolis quadrangle of Mars, U. S. Geol. Surv. Misc. Invest. Ser.,
Map I-1111, scale 1:5M.
Sharp, R. P., and M. C. Malin (1975), Channels on Mars, Geol. Soc. Am.
Bull, 86, 593?609.
Soderblom, L. A., T. J. Kreider, and H. Masursky (1973), Latitudinal dis-
tribution of a debris mantle on the Martian surface, J. Geophys. Res., 78,
4117?4122.
Strom, R. G., S. K. Croft, and N. D. Barlow (1992), The Martian impact
cratering record, in Mars, edited by H. H. Kieffer et al., pp. 383?423,
Univ. of Ariz. Press, Tucson.
Tanaka, K. L. (1986), The stratigraphy of Mars, Proc. Lunar Planet. Sci.
Conf. 17th, Part 1, J. Geophys. Res., 91(suppl.), E139?E158.
Tanaka, K. L. (1997), Sedimentary history and mass flow structures of
Chryse and Acidalia Planitiae, Mars, J. Geophys. Res., 102, 4131?
4149.
Thornhill, G. D., D. A. Rothery, J. B. Murray, A. C. Cook, T. Day, J. P.
Muller, and J. C. Iliffe (1993a), Topography of Apollinaris Patera and
Ma?adim Vallis: Automated extraction of digital elevation models,
J. Geophys. Res., 98, 23,581?23,587.
Thornhill, G. D., D. A. Rothery, J. B. Murray, T. Day, A. C. Cook, J.-P.
Muller, and J. C. Iliffe (1993b), Discharge rates in Ma?adim Vallis, Mars
(abstract), Proc. Lunar Planet. Sci. Conf. 24th, 1429?1430.
Ward, W. R. (1992), Long term orbital and spin dynamics of Mars, in Mars,
edited by H. H. Kieffer et al., Univ. of Ariz. Press, Tucson.
Watters, T. R. (1993), Compressional tectonism on Mars, J. Geophys. Res.,
98, 17,049?17,060.
Webb, R. H., and R. D. Jarrett (2002), One-dimensional estimation tech-
niques for discharges of paleofloods and historical floods, in Ancient
Floods, Modern Hazards: Principles and Applications of Paleoflood
Hydrology, Water Sci. Appl., vol. 5, edited by P. K. House et al.,
pp. 111?125, AGU, Washington, D. C.
Wilson, L., G. J. Ghatan, J. W. Head III, and K. L. Mitchell (2004), Mars
outflow channels: A reappraisal of water flow velocities from water
depths, regional slopes, and channel floor properties, J. Geophys. Res.,
109, E09003, doi:10.1029/2004JE002281.

A. D. Howard, Department of Environmental Sciences, University
of Virginia, 291 McCormick Road, Clark Hall, P.O. Box 400123,
Charlottesville, VA 22903, USA.
R. P. Irwin III and T. A. Maxwell, Center for Earth and Planetary Studies,
National Air and Space Museum, Smithsonian Institution, MRC 315,
Washington, DC 20013-7012, USA. (irwinr@nasm.si.edu)
E12009 IRWIN ET AL.: GEOMORPHOLOGY OF MA?ADIM VALLIS, MARS
33 of 33
E12009