R E P O R T
Sur?cial Deposits at Gusev Crater Along
Spirit Rover Traverses
J. A. Grant,1* R. Arvidson,2 J. F. Bell III,3 N. A. Cabrol,4 M. H. Carr,5 P. Christensen,6 L. Crumpler,7 D. J. Des Marais,8
B. L. Ehlmann,2 J. Farmer,6 M. Golombek,9 F. D. Grant,6 R. Greeley,6 K. Herkenhoff,10 R. Li,11 H. Y. McSween,12
D. W. Ming,13 J. Moersch,12 J. W. Rice Jr.,6 S. Ruff,6 L. Richter,14 S. Squyres,3 R. Sullivan,3 C. Weitz15
The Mars Exploration Rover Spirit has traversed a fairly ?at, rock-strewn terrain
whose surface is shaped primarily by impact events, although some of the
landscape has been altered by eolian processes. Impacts ejected basaltic rocks
that probably were part of locally formed lava ?ows from at least 10 meters
depth. Some rocks have been textured and/or partially buried by windblown
sediments less than 2 millimeters in diameter that concentrate within shallow,
partially ?lled, circular impact depressions referred to as hollows. The terrain
traversed during the 90-sol (martian solar day) nominal mission shows no
evidence for an ancient lake in Gusev crater.
Gusev crater is 160 km in diameter, is of
Noachian age, and lies at the terminus of the
900-km-long branching Ma?adim Vallis. The
crater is partially filled by Hesperian-aged
materials (1) and was selected as the landing
site for the Mars Exploration Rover Spirit to
search for evidence of previous liquid water
flow and/or ponding that may be responsible
for the crater infilling (1?8).
Landing occurred on a generally flat plain
(14.5692?S, 175.4729?E) characterized by
approximately circular, shallow depressions

20 m in diameter [hollows (9)] (Fig. 1A)
and poorly defined ridges up to hundreds of
meters long and a few meters high (Plate 1).
The 210-m diameter Bonneville crater (9) is
300 m northeast of the lander, and surface
albedo increases from 0.19 to 0.26 to-
ward its rim as a result of increased dust
mantling (10). Bonneville and its ejecta de-
posits comprised the primary exploration tar-
gets along the 506-m traverse during the
nominal mission reported here.
The largest rocks within 20 m of the
lander are 
0.5 m in diameter, smaller than
the largest rocks at the three previous Mars
landing sites (11, 12). Rocks 1 cm cover
about 5% of the surface, and the area covered
by fragments 10 cm is 50% of the total
rock-covered area (12). The size-frequency
distribution of rocks1 cm generally follows
the exponential model distribution based on
the Viking Lander and Mars Pathfinder land-
ing sites for 5% rock abundance (1, 11).
Most rocks 1 cm are angular to suban-
gular (13) and of variable sphericity (13), and
almost none display obvious rounding (14).
The majority of rocks are intrinsically dark
gray in color, but some exhibit variable dust
coverings and possibly associated weathering
coatings or rinds that impart a light-toned
and/or reddish color, especially apparent on
the lowermost 5 to 15 cm of their surfaces
(10?12, 15, 16) (Plate 9). Bulk-rock compo-
sitions are consistent with picritic (olivine-
rich) basalt (16).
Rocks 15 cm and within 20 m of the
lander are mostly around hollows or near drift
deposits, whereas they are largely absent
within hollows (Fig. 1A). Rocks sitting ex-
posed or perched on the surface are up to 10
times as numerous around the immediate ex-
terior of hollows as elsewhere (Fig. 1A).
Fractured and split rocks are also concentrat-
ed around the hollows, but lighter toned (red-
der) rocks are often closer to eolian drifts.
Faceted rocks are five to eight times as abun-
dant away from hollows (Fig. 1B) and are
likely the result of wind erosion (15).
There are six times as many rock frag-
ments 
4 cm in diameter per m2 as at the
Pathfinder landing site and 36 times as many
per m2 as at either of the Viking Landers
(17). Rocks 
2 mm comprise regolith and
local drift deposits covering much of the
surface. These fine sediments are characterized
by bimodal size distributions having modes
between 1 and 2 mm in diameter and less than
about 0.5 mm in diameter (18). The coarser 1-
to 2-mm grains are often rounded (14), perhaps
the result of abrasion during transport. Grains

0.2 mm in diameter comprise about 30% of
the samples, are typically basaltic in composi-
tion and devoid of clays and aqueous weather-
ing products (16), but are too small to be char-
acterized more completely (18).
The largest rocks within 10 m of the
traverse to Bonneville?s rim increased by a
factor of five to at least 2.5 m in diameter
across the crater?s outer discontinuous ejec-
ta (where the deposit incompletely buries
the preimpact surface) and more proximal
continuous ejecta (where the deposit com-
pletely buries the preimpact surface), which
starts about 205 m and 175 m from the rim,
respectively (Fig. 2). Partial burial of some
of the largest rocks along Bonneville?s rim
precludes accurate measure of their maxi-
mum diameter, but those that are visible
confirm an exponential increase in number
with decreasing size. By contrast, the aver-
age size of rocks 
10 to 15 cm increases by
a factor of less than two, from 1.75 to 2.95
cm, along the traverse while their relative
sorting decreases (Table 1).
The rim of Bonneville crater (Plates 3 and
5) rises 6.4 m above the surrounding plain,
and the crater averages 10 m deep, with a
maximum floor-to-rim-crest relief of 14 m.
Wall slopes measured along 32 evenly spaced
lines by using two different stereo panoramas
average 11?, with a range between 6? and
16?. Although a cover of eolian drift mantles
the crater floor and portions of the walls (10,
15), rocks poking up through the drift in a
few locations suggest that most of the depos-
its are relatively thin (less than a meter or two
thick). Crater walls are locally steepest and
convex along the southeast wall, where some
of the largest boulders protrude and where
several small craters (12 to 15 m in diameter)
excavated the eastern wall.
Several lines of evidence suggest that
Bonneville was formed into unconsolidated
1Center for Earth and Planetary Studies, National Air
and Space Museum, Smithsonian Institution, Wash-
ington, DC 20560, USA. 2Department of Earth and
Planetary Sciences, Washington University, St. Louis,
MO 63130, USA. 3Department of Astronomy, Space
Sciences Building, Cornell University, Ithaca, NY
14853, USA. 4NASA Ames/SETI Institute, Space Sci-
ence Division, MS 245-3, Moffett Field, CA 94035,
USA. 5U.S. Geological Survey, 345 Middle?eld Road,
Menlo Park, CA 94025, USA. 6Department of Geolog-
ical Sciences, Arizona State University, Tempe, AZ
85287, USA. 7New Mexico Museum of Natural Histo-
ry and Science, 1801 Mountain Road NW, Albuquer-
que, NM 87104, USA. 8NASA Ames Research Center,
Moffett Field, CA 94035, USA. 9Jet Propulsion Labo-
ratory, California Institute of Technology, Pasadena,
CA 91109, USA. 10U.S. Geological Survey, Flagstaff,
AZ 86001, USA. 11Department of Civil and Environ-
mental Engineering and Geodetic Science, Ohio State
University, Columbus, OH 43210, USA. 12Department
of Earth and Planetary Sciences, University of Tennes-
see, Knoxville, TN 37996, USA. 13NASA Johnson Space
Center, Houston, TX 77058, USA. 14DLR Institute of
Space Simulation, Linder Hoehe, D-51170, Cologne,
Germany. 15NASA Headquarters, 300 E Street S.W.,
Washington, DC 20560, USA.
*To whom correspondence should be addressed. E-
mail: grantj@nasm.si.edu
S P I R I T A T G U S E V C R A T E R
www.sciencemag.org SCIENCE VOL 305 6 AUGUST 2004 807
S
PEC
IA
L
S
EC
TIO
N
blocky debris. First, the low wall slopes in-
dicate that coherent strata did not impede
slumping during final stages of crater forma-
tion. Second, the largest rocks appear jum-
bled and reflect local transport after disrup-
tion of rocks up to 6 m in diameter during the
late stages of crater formation (19). Third, the
small craters formed into the eastern wall of
Bonneville do not expose bedrock. Finally,
the sharp contrast between Bonneville?s ejec-
ta deposits (Fig. 2) and surfaces beyond the
ejecta (e.g., surrounding the lander) indicate
that surfaces of the crater facies remain rela-
tively pristine and are not mantled by young-
er debris that might mask detection of bed-
rock in the crater walls. We conclude that the
crater formed largely in loose rubble but that
the largest observed ejecta blocks may be
derived from locally more competent rocks
(e.g., a lava flow).
Despite shallow wall slopes relative to
many pristine primary craters (20), the ab-
sence of debris chutes and subjacent talus at
their base and uniformly distributed rocks on
the walls imply that there has been little
modification of the walls by mass wasting. In
addition, relatively drift-free sections of the
walls (Plates 3 and 5) retain low slopes,
which demonstrates that eolian infilling is not
responsible for masking steeper walls. In-
stead, the well-preserved character and low
depth-to-diameter ratio of 0.07 suggests
that Bonneville crater experienced meters or
less of erosion and may have formed during a
secondary cratering event (21).
Hollow morphology and sizes (Fig. 3)
argue for an impact origin. Their spatial
distribution across most surfaces is random
[figure 1A in (22)] and characterized by a
population that increases exponentially
from 20 m down to less than 1 m (Fig. 3),
consistent with that expected for small cra-
ters (23?25). Hollows have fairly uniform
morphology, and the increased perched,
fractured, split, and sometimes radial dis-
tribution of rocks around their margins
(Fig. 1C) is consistent with emplacement as
ejecta. Although the number density of hol-
lows decreases within about 100 m of the
rim of Bonneville, their detection may be
impeded by relief associated with the in-
creased number of rocks 50 cm. Increas-
ing rock size and abundance near the rim of
Bonneville likely creates a clast-supported
substrate (20) that may be more difficult to
excavate during small impacts and leads to
limited expression of near-rim hollows.
The pristine appearance of Bonneville?s
ejecta deposits, however, makes it unlikely
that near-rim hollows were formed and
completely removed by erosion.
Once formed, hollows are rapidly mod-
ified to their present form. Their excavation
during an impact is accompanied by em-
placement of a surrounding ejecta deposit
with widely varying grain size and frac-
tured rocks (19). Surface roughness across
these landforms would be in disequilibrium
with the eolian regime (15 ), leading to
deflation of ejected fines (fragments 
2
mm in diameter), exposing fractured rocks,
and creating a population of perched coars-
er fragments. At the same time, hollow
interiors would be filled as transported
fines are trapped within the depression.
Trenching with Spirit?s wheels in Laguna
hollow near the edge of the Bonneville
ejecta exposed unaltered basaltic fines (16,
22) capped by a thin layer of brighter, finer,
globally pervasive dust. The absence of
dust interbeds or any chemical signature of
dust in sediment filling the hollows, cou-
pled with their uniformly filled appearance,
implies rapid modification to their current
more stable form.
Fig. 1. Example rock types around the lander. (A) Perched (blue), split (red), and fractured
(purple-pink) rocks are 2 to 10 times as numerous around hollows as elsewhere (azimuth shown is
160? to 195?). (B) Faceted (green) and light-toned (yellow) rocks are concentrated around drift,
with faceted rocks ?ve to eight times as numerous (azimuth shown is 70? to 95?). (C) Radial
accumulation of blocks (light blue) around hollow (azimuth shown is 260? to 285?). Mapped on
Pancam Mission Success panorama (mspan_2X_?nal-A1OR1_br, approximate true color based on
a scaling of 750-nm, 530-nm, and 480-nm ?lter data as RGB) (see Plate 1) and emphasized blocks
10 cm in diameter up to 20 m from the lander. Near-?eld scales are approximate.
Table 1. Small rock size and sorting. Rocks
10 to
15 cm per m2, sorting based on comparison of
measured variance.
Location* Average longaxis (cm)
Variance
(cm)
Spirit landing site 1.75 0.99
Bonneville
discontinuous ejecta
1.96 1.68
Bonneville
continuous ejecta
2.95 1.91
*Measured in 1 m2 grids with standardized Pancam and
end-of-drive clast surveys.
S P I R I T A T G U S E V C R A T E R
6 AUGUST 2004 VOL 305 SCIENCE www.sciencemag.org808
SP
EC
IA
L
SE
C
TI
O
N
Surficial deposits created by impact and
eolian processes are recognized in Gusev cra-
ter. Impact craters and associated ejecta de-
posits dominate the landscape and account
for disruption of the upper 10 m of the
volcanic subsurface. The predominant role
played by cratering processes in shaping the
surface landforms is underscored by the im-
pact origin of the hollows.
The shape, size, and clearly defined distri-
bution of ejecta blocks 20 cm around Bon-
neville is generally consistent with observations
around pristine terrestrial craters (20, 26). Mi-
nor differences between observed rock popula-
tions and that expected from a single impact
event are likely the result of comminution dur-
ing multiple impacts that mix dark gray and
lighter toned (presumably coated and/or more
weathered) rocks (10, 16, 27). Although the
hollows are modified, their diameters and asso-
ciated implied depths (21) suggest meters or
less of infilling. Some local redistribution of
rocks 
10 cm and/or winnowing of fines to
form pavements [figure 4 in (28)] may account
for lesser along-traverse variation in these size
fragments but is also consistent with only
meters of gradation at most. Hence, relief on
the plains appears to be created by rubble ac-
cumulated in the aftermath of multiple impact
events, and morphology diagnostic of nonim-
pact processes is not preserved.
All of the surficial deposits are basaltic, and
their uniform composition (16) argues for a
local volcanic source. If the rocks at the landing
site were delivered by large impacts outside of
Gusev or alluvial transport out of the Ma?adim
Vallis watershed [which exceeds 200,000 km2
along the main channel alone (7)], they would
sample a range of terrains and have variable
chemistry (at least in trace elements), which is
not seen. Moreover, the size, poor sorting, and
rounding of the largest rocks makes their arrival
as alluvium from Ma?adim Vallis unlikely. Es-
timated discharge (29) and derived flow ve-
locity (30) for Ma?adim Vallis indicate that
rocks 1 m could not have been transported
tens of kilometers to the landing site by clear
water or hyperconcentrated flows. Finally, it is
unlikely that the numerous rocks larger than 1 to
2 m could be ejecta coming from outside of
Gusev, because they would not survive ballistic
sedimentation (19). While no obvious fissure or
other volcanic edifice is visible on the floor of
Gusev crater, plains to the northeast have been
interpreted as volcanics (31) that may have ex-
tended into the landing region but were broken up
during subsequent impacts.
Local eolian deposits are inactive in the
current setting (15). Although past eolian
activity must have accounted for drift migra-
tion, formation of numerous faceted rocks,
and infilling of hollows, and is another means
of creating light-toned rocks (15), the local
distribution of eolian deposits indicates that
eolian processes are subordinate to impact
processes in shaping the surface.
While conclusive evidence for ancient
water-lain deposits in Gusev remains elusive,
the volcanic materials and subsequent action
of impact and eolian processes may mask
their signature across terrains traversed dur-
ing the nominal mission. Nevertheless, the
preservation of the landscape is testament to
the minimal gradation occurring since Hes-
perian times in Gusev that may be roughly
comparable to post-Hesperian gradation af-
fecting the Mars Pathfinder landing site (32).
References and Notes
1. M. P. Golombek et al., J. Geophys. Res. 108, 8072 (2003).
2. J. A. Grant et al., Planet. Space Sci. 52, 11 (2004).
3. N. A. Cabrol, E. A. Grin, G. Dawidowicz, Icarus 123,
269 (1996).
4. N. A. Cabrol, E. A. Grin, R. Landheim, Icarus 132, 362
(1998).
5. N. A. Cabrol, E. A. Grin, R. Landheim, R. O. Kuzmin, R.
Greeley, Icarus 133, (1998).
6. R. O. Kuzmin, R. Greeley, R. Landheim, N. A. Cabrol, J.
Farmer, U.S. Geol. Surv., Misc. Geol. Invest. Map
I-2666 (2000).
7. R. P. Irwin, T. A. Maxwell, A. D. Howard, R. A. Crad-
dock, D. W. Leverington, Science 296, 2209 (2002).
8. J. W. Rice Jr. et al., Lunar Planet. Sci. [CD-ROM],
abstract 2091 (2003).
9. Names have been assigned to areographic features
by the Mars Exploration Rover (MER) team for plan-
ning and operations purposes. The names are not
formally recognized by the International Astronomi-
cal Union.
10. J. F. Bell III et al., Science 305, 800 (2004).
11. M. P. Golombek, D. Rapp, J. Geophys. Res. 102, 4117
(1997).
12. M. P. Golombek et al., J. Geophys. Res. 108, 8086
(2003).
13. M. R. Leeder, Sedimentology, George Allen and Unwin
Ltd. (London, 1982).
Fig. 2. Distal margin of
Bonneville?s ejecta de-
posit (black dashed line)
from ?Middle Ground?
hollow. Mosaic of Pan-
cam images 2P131526-
662FFL1155P2437 and
2P131526768FFL115-
5P2437 with the 430-
nm blue ?lter.
Fig. 3. Hollow size-frequency distribution along
the traverse to Bonneville crater. Hollows are
plotted in square root of 2 bins for the size range
20 m down to 1 m and reported as the number
per km2. Includes 85 hollows 0.02 km2.
S P I R I T A T G U S E V C R A T E R
www.sciencemag.org SCIENCE VOL 305 6 AUGUST 2004 809
S
PEC
IA
L
S
EC
TIO
N
14. R. L. Folk, Petrology of Sedimentary Rocks, Hemphill
Publishing (Austin, TX, 1980).
15. R. Greeley et al., Science 305, 810 (2004).
16. H. Y. McSween et al., Science 305, 842 (2004).
17. M. P. Golombek et al., Lunar Planet. Sci. [CD-ROM],
abstract 2185 (2004).
18. K. E. Herkenhoff et al., Science 305, 824 (2004).
19. H. J. Melosh, Impact Cratering, Oxford Univ. Press
(New York, 1989).
20. J. A. Grant, P. H. Shultz, J. Geophys. Res. 98, 11025
(1993).
21. M. Hurst, M. P. Golombek, R. Kirk, Lunar Planet. Sci.
[CD-ROM], abstract 2068 (2004).
22. R. E. Arvidson et al., Science 305, 821 (2004).
23. W. K. Hartmann, J. Anguita, M. A. de la Casa, D. C.
Berman, E. V. Ryan, Icarus 149, 37 (2001).
24. W. K. Hartmann et al., Nature 397, 586 (1999).
25. F. Horz, M. J. Cintala, W. C. Rochelle, B. Kirk, Science
285, 2105 (1999).
26. J. A. Grant, P. H. Schultz, J. Geophys. Res. 98, 15033
(1993).
27. P. R. Christensen et al., Science 305, 837 (2004).
28. S. W. Squyres et al., Science 305, 794 (2004).
29. R. P. Irwin, A. D. Howard, T. A. Maxwell, Lunar Planet.
Sci. [CD-ROM], abstract 1852 (2004).
30. G. Komatsu, V. R. Baker, J. Geophys. Res. 102, 4151
(1997).
31. K. A. Milam et al., J. Geophys. Res. 108, 8078 (2003).
32. M. P. Golombek, N. T. Bridges, J. Geophys. Res. 105,
1841 (2000).
33. Research was supported by NASA through the Mars
Exploration Rover Project. Our sincere thanks go to
the Mars Exploration Rover management, staff, and
engineering teams for their outstanding support and
operation of Spirit.
Plates Referenced in Article
www.sciencemag.org/cgi/content/full/305/5685/807/
DC1
Plates 1, 3, 5, and 9
3 May 2004; accepted 28 June 2004
R E P O R T
Wind-Related Processes Detected by the Spirit
Rover at Gusev Crater, Mars
R. Greeley,1 S. W. Squyres,2 R. E. Arvidson,3 P. Bartlett,4 J. F. Bell III,2 D. Blaney,5 N. A. Cabrol,6 J. Farmer,1
B. Farrand,7 M. P. Golombek,5 S. P. Gorevan,4 J. A. Grant,8 A. F. C. Haldemann,5 K. E. Herkenhoff,9 J. Johnson,9
G. Landis,5 M. B. Madsen,10 S. M. McLennan,11 J. Moersch,12 J. W. Rice Jr.,1 L. Richter,13 S. Ruff,1 R. J. Sullivan,2
S. D. Thompson,1 A. Wang,3 C. M. Weitz,14 P. Whelley,1 Athena Science Team
Wind-abraded rocks, ripples, drifts, and other deposits of windblown sediments are
seen at the Columbia Memorial Station where the Spirit rover landed. Orientations
of these features suggest formative winds from the north-northwest, consistent with
predictions from atmospheric models of afternoon winds in Gusev Crater. Cuttings
from the rover Rock Abrasion Tool are asymmetrically distributed toward the
south-southeast, suggesting active winds from the north-northwest at the time
(midday) of the abrasion operations. Characteristics of some rocks, such as a
two-toned appearance, suggest that they were possibly buried and exhumed on the
order of 5 to 60 centimeters by wind de?ation, depending on location.
In the current environment of Mars, wind ap-
pears to be the most frequent agent of surface
modification, resulting in albedo patterns that
change on time scales as short as a few weeks
(1). Abundant dune forms, mantles of wind-
blown deposits, and wind-eroded features are
seen from orbit in many parts of Mars, includ-
ing the three previous sites where successful
landings have occurred. Understanding the pro-
cesses that form aeolian (wind-related) features
provides insight into the evolution of the mar-
tian surface, including rates of erosion and dep-
osition. The Mars Exploration Rover (MER)
Spirit landed near the middle of Gusev Crater
(2?5) in a relatively low-albedo zone (6)
considered to be a track left by the passage of
dust devils that removed bright dust to expose
a relatively darker substrate (Fig. 1A). Com-
parison of orbital images taken of the same
area from July 2003 to January 2004 show
changes in the tracks, indicating that dust
devils were recently active.
Here, we describe initial analyses of aeo-
lian features during the first 90 sols (7) of
operation. Wind-related features include sed-
iments (some of which are organized into
bedforms such as ripples), wind-abraded fea-
tures on rocks, eroded zones around rock
edges, and features generated by the rover
during operations that suggest active winds.
The surface at Columbia Memorial Station
consists of rocks, regolith, dark granules, and
fine-grained material, including dust (8, 9).
Patches of red regolith range in size from 0.5 m
across to as large as 15 m across. Bonneville
(10) crater and many of the small depressions,
called hollows, are partly filled with regolith
deposits. Light-toned material was inferred to
be dust, the upper surfaces of some rocks, the
rover solar panels, and the Panoramic Camera
(Pancam) calibration target (11). Although dust
grains are too small to be resolved by the
Microscopic Imager (MI) (12), previous esti-
mates suggest that martian dust is a few mi-
crometers in diameter (13, 14).
An MI image of a regolith patch shows a
bimodal size distribution of particles (Fig.
1B) that includes coarse (1 to 3 mm) grains
and finer grains smaller than a few hundred
micrometers in diameter. Although some
coarse grains are subangular, most are round-
ed, suggesting erosion during transport. We
propose that the dark coarse particles are
lithic fragments on the basis of their basaltic
composition (16, 17) and appearance in the
MI images (18).
Aprons of granular debris occur as isolat-
ed patches on the regolith and around some
rocks. For example, the rock Adirondack (2)
has an encircling debris apron that extends 5
to 20 cm from the edge of the rock. The
aprons consist of coarse grains that have
spectral properties similar to those of Adiron-
dack and the other basaltic rocks in the area
(11, 19). Therefore, some of the more angular
coarse-grained material is probably derived
1Department of Geological Sciences, Arizona State
University, Box 871404, Tempe, AZ 85287?1404,
USA. 2Department of Astronomy, Cornell University,
428 Space Sciences Building, Ithaca, NY 15853?1301,
USA. 3Department of Earth and Planetary Sciences,
Washington University, One Brookings Drive, St. Lou-
is, MO 63031?4899, USA. 4Honeybee Robotics, 204
Elizabeth Street, New York, NY 10012, USA. 5Jet
Propulsion Laboratory, 4800 Oak Grove Drive, Pasa-
dena, CA 91109?8099, USA. 6Ames Research Center,
Moffett Field, CA 94035?1000, USA. 7Space Science
Institute, University of Colorado, Boulder, CO 80301,
USA. 8Center for Earth and Planetary Studies, Nation-
al Air and Space Museum, Smithsonian Institution,
Washington, DC 20560?0315, USA. 9U.S. Geological
Survey, 2255 North Gemini Drive, Flagstaff, AZ
86001?1698, USA.10Niels Bohr Institute for Astrono-
my, Physics, and Geophysics, Center for Planetary
Science and ?rsted Laboratory, University of Copen-
hagen, Universitetsparken 5, DK-2100 Copenhagen,
Denmark. 11Department of Geosciences, State Uni-
versity of New York at Stony Brook, Stony Brook, NY
11794?2100, USA. 12Department of Earth and Plan-
etary Sciences, University of Tennessee, 1412 Circle
Drive, Room 306, Knoxville, TN 37996, USA. 13Deut-
ches Zentrum fu?r Luft-und Raumfart (German Aero-
space Center)?Institute of Space Simulation, Linder
Hoene, D-51170 Cologne, Germany. 14National Aero-
nautics and Space Administration (NASA) Headquar-
ters, Washington, DC 20546?0001, USA.
*To whom correspondence should be addressed. E-
mail: Greeley@asu.edu
S P I R I T A T G U S E V C R A T E R
6 AUGUST 2004 VOL 305 SCIENCE www.sciencemag.org810
SP
EC
IA
L
SE
C
TI
O
N