Periglacial landforms at the Phoenix landing site

and the northern plains of Mars

Michael T. Mellon,1 Raymond E. Arvidson,2 Jeffrey J. Marlow,2,3 Roger J. Phillips,2,4

and Erik Asphaug5

Received 13 November 2007; revised 3 May 2008; accepted 30 May 2008; published 5 November 2008.

[1] We examine potentially periglacial landforms in Mars Orbiter Camera (MOC)and High Resolution Imaging Science Experiment (HiRISE) images at the Phoenixlanding site and compare them with numerical models of permafrost processes tobetter understand the origin, nature, and history of the permafrost and the surface of thenorthern plains of Mars. Small-scale (3–6 m) polygonal-patterned ground is ubiquitousthroughout the Phoenix landing site and northern plains. Larger-scale (20–25 m)polygonal patterns and regularly spaced (20–35 m) rubble piles (localized collections ofrocks and boulders) are also common. Rubble piles were previously identified as‘‘basketball terrain’’ in MOC images. The small polygon networks exhibit well-developedand relatively undegraded morphology, and they overlay all other landforms. Comparisonof the small polygons with a numerical model shows that their size is consistent with athermal contraction origin on current-day Mars and are likely active. In addition, theobserved polygon size is consistent with a subsurface rheology of ice-cemented soil ondepth scales of about 10 m. The size and morphology of the larger polygonal patterns andrubble piles indicate a past episode of polygon formation and rock sorting in thermalcontraction polygons, while the ice table was about twice as deep as it is presently.The pervasive nature of small and large polygons, and the extensive sorting of surfacerocks, indicates that widespread overturning of the surface layer to depths of many metershas occurred in the recent geologic past. This periglacial reworking has had a significantinfluence on the landscape at the Phoenix landing site and over the Martian northernplains.

Citation: Mellon, M. T., R. E. Arvidson, J. J. Marlow, R. J. Phillips, and E. Asphaug (2008), Periglacial landforms at the Phoenix

landing site and the northern plains of Mars, J. Geophys. Res., 113, E00A23, doi:10.1029/2007JE003039.

1. Introduction

[2] Permafrost on Mars today is global in extent. Belowabout a centimeter of depth at the equator and below thevery surface at higher latitudes the soil temperatures remainbelow the freezing point of water year-round. At higherlatitudes soil temperatures are cold enough that even inthe dry Martian climate subsurface ground ice has beenpredicted to be stable [e.g., Leighton and Murray, 1966] andindeed the presence of ground ice has been inferred from thedetection of abundant subsurface hydrogen [e.g., Boynton etal., 2002]. It is therefore not surprising that ground ice

would have an influence on the geomorphic character of thelandscape. One of the primary goals of the Mars ScoutMission Phoenix is to land in a high-latitude region withabundant ground ice to examine the environment, thelandscape, and the subsurface ice itself. To this end alocation has been chosen in the northern plains of Mars(near 68�N � 233�E) where ground ice is expected to beabundant a few centimeters below the surface [Arvidson etal., 2008]. The desire to land in a region with shallowground ice is necessarily coincident with the desire to landin a region exhibiting abundant periglacial landforms. Anabsence of these landforms could suggest an absence ofground ice and would be of concern to the primary scientificobjectives of the Phoenix mission.[3] Perhaps the most widespread landforms in terrestrial

permafrost are thermal-contraction fractures and associatedpolygonal-patterned ground. These patterns form whencohesive permanently ice-cemented soil undergoes seasonalthermal contraction resulting in tensile stresses that exceedthe strength of the frozen ground, thus developing ahoneycomb network of fractures to relieve this stress [e.g.,Leffingwell, 1915; Lachenbruch, 1962]. Individual fracturesmay only open millimeters in one season, but over many

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 113, E00A23, doi:10.1029/2007JE003039, 2008ClickHere

for

FullArticle

1Laboratory for Atmospheric and Space Physics, University ofColorado, Boulder, Colorado, USA.

2Department of Earth and Planetary Sciences, Washington University,St. Louis, Missouri, USA.

3Now at the Department of Earth Science and Engineering, ImperialCollege London, London, UK.

4Now at Southwest Research Institute, Boulder, Colorado, USA.5Department of Earth and Planetary Science, University of California,

Santa Cruz, California, USA.

Copyright 2008 by the American Geophysical Union.0148-0227/08/2007JE003039$09.00

E00A23 1 of 15

years of repeated cracking at the same previously fracturedzones of weakness, surface material may enter the crack anddevelop a subsurface wedge. In climates where seasonalsurface thaw and runoff are available, such as in Arctictundra, a subsurface ice wedge may form [Black, 1976]. Inarid regions analogous to Mars, such as the Antarctic DryValleys, loose surface sand and dust enter the crack forminga subsurface sand wedge [Pewe, 1959, 1974]. The growthof the wedge and subsequent seasonal thermal expansionresult in the development of surface topography in the formof a perimeter trough over the crack and wedge wheresurface material is being consumed, and uplift within part orall of the polygon interior.[4] Another common feature in terrestrial permafrost is

the self organization of rocks, i.e., cobbles and boulders.Most often this sorting occurs through freeze-thaw cycles inthe active layer above the permafrost [e.g., Washburn, 1956,1980; Hallet and Prestrud, 1986], where frost heave drivesan upward and outward convection. Sorting may also occurin permafrost undergoing thermal contraction, where nofreeze-thaw cycle is present and rocks slump into thepolygon trough driven by gravity [Washburn, 1956; Pewe,1959] or combined with differential sublimation [Marchantet al., 2002]. Sletten et al. [2003] suggested such a convec-tion like cycle in some Antarctic sand-wedge polygons,where long-term wedge growth drives fine material inwardand results in upward deformation of the polygon centerfollowed by surface creep toward the troughs. Indeed, theyfound sand-wedge growth in the oldest polygons resulted incomplete replacement of the polygon interior subsurfacewith sand-wedge material. In these cases rocks larger thanthe typical thermal-contraction-crack width of millimetersbecome trapped in the trough, unable to be incorporated atdepth in the wedge. In general, rocks can be moved by aconvective overturning in the soil, cryoturbation, and canresult in observable patterns in the surface rock population.While the term cryoturbation usually refers to a freeze-thaw-driven churning of soil materials, in the case ofthermal-contraction polygons it describes convection-likereworking of the surface layer in the absence of freeze-thawcycles.[5] Climate conditions on Mars are notably different from

those on Earth. In particular the annual mean temperaturesare significantly lower on Mars, with a global annual meanof 204 K and a range from 234 K at the equator to 159 K atthe poles [Mellon et al., 2004]. Such low mean temperaturesmakes the potential for freeze-thaw processes in the currentclimate extremely unlikely. In addition, the rheologicalresponse of the permafrost to induced stresses, such asthermal contraction, differs from that on Earth in that atthese low temperatures the ice-rich material is more resistantto viscous relaxation, allowing high tensile stresses todevelop [Mellon, 1997]. Under such conditions thermal-contraction cracks easily form in cohesive ice-rich perma-frost and thus polygonal-patterned ground is expected to bewidespread.[6] Analysis of periglacial features in general are useful

in two ways. These features tell us something about thepresent state of ice in the permafrost and, since thesefeatures are long-lived in their development, they can offera window into the history of the permafrost and of theclimate. The purpose of this paper is to examine the

potential periglacial geomorphology at the Phoenix landingsite and associated areas in the northern plains, focusing onpolygonal-patterned ground and rock self-organization.From observations of these features and comparison withnumerical models we will show that they are consistent witha recent long period of shallow ground ice, but that theground ice distribution may have been different in thegeologic past.

2. Periglacial Landforms in the Northern Plains

[7] Periglacial landforms and patterned ground have beenobserved on Mars since the Viking mission [e.g., Lucchitta,1981; Squyres and Carr, 1986]; however, those observa-tions were limited to only the largest scale of features thatwould be found in terrestrial permafrost. Mars GlobalSurveyor (MGS) significantly improved this view withnearly an order of magnitude higher resolution than thebest Viking provided. With image resolution approachingmeter-scale and the ability to distinguish multimeter fea-tures, more periglacial landforms on Mars could be ob-served. Indeed, a number of studies examined theoccurrence and distribution of patterned ground [e.g., Malinand Edgett, 2001; Seibert and Kargel, 2001; Mangold,2005]. However, with the best available resolution stillbeing larger than a meter and the need to frequently binimages at high latitude owing to low illumination, it hasbeen difficult to identify the breadth of periglacial featuresand diagnostic details. Nevertheless polygonal-patternedground has been found to be relatively common in thenorthern plains and observed in a notable fraction of theMars Orbiter Camera (MOC) images. With the arrival ofthe High Resolution Imaging Science Experiment (HiRISE)on Mars Reconnaissance Orbiter (MRO), submeter imagingand high signal-to-noise ratios at low illumination levels[McEwen et al., 2007] has allowed for meter-scale featuresto be identified. Indeed, we find that polygonal-patternedground is far more widespread over the high latitudes andthe northern plains than was previously thought.[8] In examining the Martian northern plains for an

optimal Phoenix landing site we began our survey ofpotentially periglacial features with a study of MOC imagesover the range of 60�–77�N, slightly broader than thoselatitudes accessible to the Phoenix mission, 65�–72�N. Onthe basis of the previous studies we set out to identify andmap the geographic distribution of images that containedspecific geomorphic signatures that were considered to belikely periglacial in origin. These features included polyg-onal patterned ground, ‘‘basketball terrain’’ and organizedstripes. Polygonal patterns were identified by distinct mul-tisided boundaries of troughs in a range of sizes down to thelimit of the resolution. Basketball terrain refers to a regularknobby or stippled texture reminiscent of the surface of abasketball [Malin and Edgett, 2001; Head et al., 2003]. Insome cases basketball terrain exhibited a polygonal appear-ance at MOC resolution and likewise the smallest-scalepolygons observed in MOC images exhibited a basketballtexture; separation of these two features was sometimesblurred. Stripes consist of parallel to subparallel continuousand discontinuous ridges, and include aligned knobs similarto basketball knobs.

E00A23 MELLON ET AL.: PERIGLACIAL LANDFORMS AT PHOENIX SITE

2 of 15

E00A23

[9] We examined 3141 images from the MGS primaryand extended missions of which 2063 images were consid-ered of medium-to-high quality, sufficient to confidentlyidentify these features and base statistical analysis. A fewimages exhibited resolutions as high as 1.6 m pixel�1,but most images were 3.2 m pixel�1 and larger. Figure 1shows the geographic distribution of the occurrence andnonoccurrence of these features, and Figure 2 shows thefrequency of occurrence with respect to latitude. Over theselatitudes of the northern plains polygonal patterns werethe most abundant, identified in 45% of the images andoccurred geographically nearly uniformly distributed.Basketball terrain occurred in 36% of the images also nearlyuniformly distributed, but with slightly greater geographic

variability. Stripes occurred in 21% of the images, includingthe northern portions of ‘‘region D’’ near the selectedPhoenix landing site, and in contrast to other features areconcentrated mostly in the highest latitudes that we exam-ined and with notable longitudinal variability. Some imagesexhibited more than one type of feature and only 26% of theimages show none of these feature at MOC resolutions.[10] With the arrival of MRO, HiRISE images were used

to examine the detailed morphology and rock populations toassess potential landing sites [Arvidson et al., 2008]. Con-cerns about rocks as a landing hazard lead to a wide zonalsurvey of the latitude range 65�–72�N. In all 157 HiRISEimages were collected, of which 23 were acquired in theregion of the selected landing site. Throughout this latitude

Figure 1. Polar orthographic maps showing the distribution of potentially periglacial landforms foundin MOC images: (a) polygonal patterned ground, yellow dots; (b) ‘‘basketball terrain,’’ green dots; and(c) striped terrain red dots. Blue dots indicate images that lack features in the associated category. Theoutlined box indicates the location of region D, in which the Phoenix landing site has been selected.Examples of each feature are shown: (d) polygons in MOC image E03-00299 (64.68�N � 67.08�E);(e) basketball terrain in MOC image E01-01867 (73.25�N � 319.6�E); and (f) stripes in MOC imageE23-00658 (74.62�N � 292.47�E).

E00A23 MELLON ET AL.: PERIGLACIAL LANDFORMS AT PHOENIX SITE

3 of 15

E00A23

zone the terrain appear quite similar, dominated by small-scale polygons with varying degrees of topographic sharp-ness, and with surface rocks and boulders in varyingconcentrations.[11] The remainder of this paper will focus on the

examination of the potentially periglacial landforms seenin HiRISE images in the selected landing-site region.Figure 3 shows the location of the 23 HiRISE images inthe region of the selected landing site acquired at resolutionsof about 31 cm pixel�1 and 62 cm pixel�1. For the purposesof quantitative analysis we focused on four of these images(see also Table 1) chosen for their low emission angles tominimize potential atmospheric effects and distortions ofthe smallest-scale features, and to cover eastern, western,and central potions of the landing ellipse. Given theenormous amount of high-resolution real estate in any oneHiRISE image, three to six sample regions were selectedfrom each image for further quantitative and statisticalmeasurements representing various type examples of theterrains in each image.[12] Four dominant types of potentially periglacial

geomorphic characteristics are observed throughout thePhoenix landing-site region.[13] 1. Small-scale polygons, less than 10 m in scale, are

ubiquitous and blanket all other geologic features.

Figure 2. Histograms of the features mapped in Figure 1as a function of latitude. Generally, polygonal and basket-ball terrains are abundant at all the latitudes examined. Aweak anticorrelation of occurrence between the two featuresmay be related to difficulty distinguishing between them insome of the images. Stripes occur primarily at higherlatitudes.

Figure 3. A map of HiRISE image coverage at the location of the selected Phoenix landing site. Imagesoutlined are accompanied by the MRO orbit number for reference. Black outlines indicate full-resolutionimages (�31 cm pixel�1), and white outlines indicate 2 � 2 binned images (�62 cm pixel�1). Imagesoutlined in red are at full resolution, as noted in Table 1, and were used for quantitative analysis. Thebackground is a MOLA elevation base map with latitudes and longitudes shown at 1� intervals.

E00A23 MELLON ET AL.: PERIGLACIAL LANDFORMS AT PHOENIX SITE

4 of 15

E00A23

[14] 2. Larger polygons, greater than 10 m in scale, arealso observed in some locations.[15] 3. Small concentrated piles of rocks and boulders

(termed here ‘‘rubble piles’’) are found in many locations,particularly in the western portions of the landing site regionand in other locations throughout the northern plains.[16] 4. Scattered throughout the polygonal patterns are

larger-scale contiguous trough-and-ridge features that donot generally form closed polygonal shapes.[17] Small-scale polygonal patterns cover nearly 100% of

every image acquired at these latitudes. Examples areshown in Figure 4. Polygons are outlined by a perimetertrough on the order of a meter wide. Troughs vary inappearance from shallow and gently sloping, to distinctand V-shaped. Occasionally longer and wider troughs occurthat connect several polygons in line and sometimes exhibita thin depression along the center of the trough (e.g.,Figures 4a and 4f). Finer details in the small-polygontroughs are limited by the resolution of the HiRISE images(31 cm pixel�1). Together these troughs outline largelyhexagonal or irregular polygon shapes with occasionalsquare shapes. The majority of trough intersections are

three-way junctions closer to equiangular than right angles.The polygon interiors vary (see Figure 4) from generally flatwith only a slight central uplift, to uplifted and domed, todiscontinuous uplifted ridges adjacent to the trough andwith centers slightly lower than the ridges. Stereo analysissuggests topographic relief over the scale of a polygon is nomore than a few tens of centimeters [Arvidson et al., 2008].Variation in the clarity of troughs and the definition of thepolygons occur throughout each image on horizontal scalesas small as hundreds of meters.[18] In addition to the morphology, a measure of the

effective diameter of polygons is a valuable parameter togauge the origin and environmental conditions under whichthese features formed. To determine an effective diameterwe measure and averaged the two perpendicular axesrelative to the image coordinates from trough center totrough center through the central potion of each polygonregardless of the specific shape or orientation. For each ofthirteen 100 � 75 m sample areas from the four studyimages analyzed (see Figure 4 for examples) we firstidentified the troughs. Only polygons that were clearlyenclosed by troughs were measured. In most images troughsoriented perpendicular to the solar azimuth were more easilyrecognized compared to troughs oriented parallel to thesolar azimuth, however in all but one of these sample areas60–100 enclosed polygons could be identified. To evaluatemeasurement biases, in two sample areas we also measuredthe long and short axis of each polygon regardlessof azimuth and we measured the perpendicular spacingbetween adjacent troughs regardless of enclosure. Thesetwo alternate methods produce nearly identical results to the

Table 1. HiRISE Study Imagesa

Image Latitude Longitude

PSP_001959_2485 68.159� 232.634�PSP_002025_2485 68.166� 230.871�PSP_002170_2485 68.255� 232.873�PSP_002249_2485 68.205� 234.337�

aImages listed were acquired at a resolution of 31.2 cm pixel�1.

Figure 4. Samples of small-scale polygons from each of the study images in Table 1 and for differentterrain types. Some morphological variation is apparent. However, sizes are extremely consistentfrom area to area. Each sample is from HiRISE images (a) PSP_001959_2485, (b) PSP_002249_2485,(c) PSP_001959_2485, (d) PSP_002249_2485, (e) PSP_002025_2485, and (f) PSP_002170_2485. Allsamples scenes are 100-m wide. North is roughly down, and the illumination is from the upper right.Figures 4a and 4b also contain larger-scale polygonal patterns. Figures 4c and 4e also contain rubblepiles.

E00A23 MELLON ET AL.: PERIGLACIAL LANDFORMS AT PHOENIX SITE

5 of 15

E00A23

initial method, and well within the standard deviation ofeach method. In all we measured 915 polygons in the 13sample areas. Figure 5 shows a histogram of effectivediameters. The mean effective diameter of all measurementswas 4.6 m with a standard deviation of 1.1 m. This valuevaried only slightly (4.3 to 5.1 m) between the sample areasof from all four study images over a horizontal distance ofabout 75 km.

[19] Distinctly larger polygonal patterns are observed andcan be separated into two morphological types. The firsttype occurs as regular circular-to-hexagonal ring depres-sions that appear distinct from the smaller polygons, thoughthe smaller polygons are superposed on the larger ones(Figures 4b and 6). The second type is delineated bytopographically exaggerated troughs (larger than typicalsmall-polygon troughs) that mark one side of the smallpolygons and that connect to form a larger pattern enclosingtens of smaller polygons (see Figure 4a). While thesepolygons are fairly regular in size and similar in size tothe first type of large polygon, they are more irregular inshape and are frequently not fully enclosed. The first type ofcircular-to-hexagonal polygons exhibits a preponderance ofequiangular three-way intersections and mostly straightsides. The troughs, on the order of 4-m wide, are subduedand gently sloping, lacking sharp depressions. The smallpolygons within the larger troughs also appear subdued,possibly mantled, in contrast to the well-defined smallpolygons that superpose the large polygon interior. Theinterior portions of these large polygons are generallydomed upward and uniformly. In addition, larger rocksand boulders resolvable by HiRISE are virtually absentfrom the large polygons terrains. Measurements of theeffective diameter are shown in Figure 7. The averageeffective diameter of 45 large circular-to-hexagonal poly-gons in image PSP_002249_2485 is 22.1 m with a standarddeviation of 3.2 m. The three other study images listed inTable 1 do not show these type of larger polygonal patterns,though other HiRISE images in this landing site region doexhibit the larger polygonal patterns concentrated exclu-sively in eastern images in this region. The more irregular-shaped type of large polygons are observed throughout theregion and mainly in areas with few rocks and boulders.

Figure 5. A histogram of small-scale polygon sizesmeasured from 915 polygons. The mean polygon size is4.6 m with a standard deviation of 1.1 m, with mostindividual polygon sizes between 3 and 6.5 m. Sizes weremeasured for 13 sample areas from the images listed inTable 1. The size of the small polygons varies littlethroughout the region.

Figure 6. Examples of larger-scale polygons exhibiting circular-to-hexagonal perimeter depressions.Image PSP_002249_2485 is shown. Scene width is 500 m. Illumination is from the upper right.

E00A23 MELLON ET AL.: PERIGLACIAL LANDFORMS AT PHOENIX SITE

6 of 15

E00A23

[20] In images covering the western terrain of the landingsite region, rocks and boulders are clearly evident andorganized into regularly spaced concentrated piles or‘‘rubble piles’’ (Figures 4c, 4e, and 8). The rubble pilesconsist of a distribution of large and small rocks, usuallyclustered in an area 5–10 m across with relatively few rocksobserved in the interrubble-pile areas. Occasionally rubblepiles are connected by arcs or bands of rubble. Both rubblepiles and arcs tend to sit atop slight topographic highs,either mounds or ridges typically 50 cm higher thanthe intermound terrain [Kirk et al., 2008]. The boundarybetween regions dominated by rubble piles and rubble-pile-free regions is usually fairly distinct with an abrupt eleva-tion drop of less than a few meters down to the rubble-free

regions over horizontal distances tens to hundreds of meters[see also Seelos et al., 2008]. Small polygons are alsoubiquitous throughout the regions dominated by rubble piles(e.g., Figures 4c and 4e). The polygons are quite distinctbetween rubble piles. While the polygon troughs appears tocross-cut the rubble piles and arcs in many locations, thetroughs are typically not distinguishable through the rubbleat HiRISE resolution. In the study images rubble piles arenot aligning linearly but staggered, however, in otherregions of Mars the rubble piles can be aligned in linearto sublinear rows. At MOC resolution the rubble pile terrainhas been previously identified in this region as basketballterrain [e.g., Malin and Edgett, 2001]. Figure 9 compares aMOC image of basketball terrain with a coincident HiRISEimage where the rubble piles are clearly visible at highresolution. For comparison with the polygonal terrains wemeasured the spacing between rubble piles in terms of thedistance from one rubble pile to three to five of its nearestneighbors. We measured 452 spacing distances between277 rubble piles in four sample areas of the western threestudy images. Results are shown in Figure 10. The meanspacing was 27.5 m with a standard deviation of 7.2 m.[21] Contiguous ridge-and-trough features are also

observed (Figure 11). These features consist of a long(sometimes 1=2 to 3=4 km) sublinear ridge with a distinctU- and V-shaped trough to one or both sides. Within thetrough a thin linear depression running most of the length ofthe trough is often observed. The ridge and trough togetherare typically 5–10-m wide. Small polygons overlay theridge and are aligned to share the same trough. These ridge-and-trough features can be oriented in any direction buthave a strong preference to a NW–SE trend. They aresparsely distributed in most of the images with a few areasof concentration. They occur both in terrains dominated byrubble piles and those without rubble piles, and occasionallycross cut the boundaries between these terrains. In a fewcases circular and concentric ridge-and-trough orientations

Figure 7. A histogram of larger polygons such as thoseshown in Figure 6, including different locations in imagePSP_002249_2485.

Figure 8. An example of rubble pile terrain (also called basketball terrain at MOC resolution). ImagePSP_002025_2485 is shown. Scene width is 500 m. Illumination is from the upper right.

E00A23 MELLON ET AL.: PERIGLACIAL LANDFORMS AT PHOENIX SITE

7 of 15

E00A23

are observed that indicate the presence of a heavily modi-fied ghost crater.

3. Modeling Permafrost Polygons on Mars

[22] The observed polygonal patterns are proposed tohave originated from thermal-contraction cracking ofpermafrost on the basis of their morphological similaritiesto terrestrial permafrost and the knowledge that theseregions on Mars are cold and ice rich. However, thedominant polygons are notably smaller that those typicallyfound in terrestrial permafrost (around 10–35 m are typicalin some of the Antarctic Dry Valleys) [Pewe, 1959; Black,1976; Marchant et al., 2002; Sletten et al., 2003]. Numer-ical models of the physical process of thermal contraction inMartian permafrost and polygon behavior can be used toevaluate the observational characteristics, such as polygonsize, relative to the environmental conditions on Mars incontrast to those on Earth. Mellon [1997] used an analyticsolution to a viscoelastic model of unfractured permafrostand showed that under current climate conditions wintertensile stresses would greatly exceed those needed toproduce polygonal fractures in ice or ice-cemented soil,and that polygonal-patterned ground could easily form.However, this solution did not account for the stress reliefcaused by a fracture and therefore could not predict how faraway from an initial fracture that a second fracture couldform, nor predict the size of a typical polygon underMartian environmental conditions.[23] To estimate the polygon sizes expected on Mars, a

finite element model is needed to determine how the stressfield near and far from a fracture responds to the fractureopening and closing during the seasonal temperature cycles.Then by estimating the peak tensile stress at the center of apolygon of a given size, it can be determined for a polygonof that size whether a new fracture will form to subdivide it

or if the existing fractures will adequately relieve the stressto prevent subdivision. To evaluate the stress field infractured permafrost we utilized a finite element model,TECTON [Melosh and Raefsky, 1980, 1981] with a smalladaptation to account for seasonal thermal contraction.TECTON is a fully viscoelastic finite element model thatwas developed to study terrestrial plate tectonics and hasbeen successfully used in planetary applications such ascrater relaxation on icy satellites [Hillgren and Melosh,1989]. We use the flow laws for ice and icy soil inTables 2 and 3 [Durham et al., 1997]. (See also Mellon[1997] for a review of viscous and elastic parameters for ice

Figure 9. A comparison of basketball terrain as seen in (a) MOC (R22-00099 at 3.35 m pixel�1) and(b) HiRISE (PSP_001404_2490 at 31.6 cm pixel�1) images near 68.99�N � 254.01�E. (c) The close-upHiRISE view shows the presence of rubble piles which exhibit a basketball-like texture at lowerresolution. (d) A small impact crater that has been significantly degraded and muted is shown.Illumination is from the upper right.

Figure 10. A histogram of the spacing between rubblepiles. Spacing is measured from each rubble pile to three tofive of its nearest neighbors in samples from the western-most three images listed in Table 1. Most rubble piles arespaced 20–35 m apart.

E00A23 MELLON ET AL.: PERIGLACIAL LANDFORMS AT PHOENIX SITE

8 of 15

E00A23

and icy soils, and of viscoelastic theory.) In the currentapplication we model the ductile strain rate as proportionalto the deviatoric stress multiplied by the effective stressto the power of n-1, which better mimics the conditionsunder which the flow law was derived [Durham et al.,1997] and thus provides improved extrapolation of theviscous response under nonlaboratory conditions.[24] Figure 12a shows a schematic model of the rheolog-

ical behavior of the permafrost. In the presence of fixedboundary conditions, thermal contraction initially leads toan elastic stress response followed by a ductile, or non-Newtonian viscous, relaxation of this stress. In the presenceof fractures, additional strain (or motion) of the permafrostmatrix will also relieve stress. Since TECTON does notincorporate thermal contraction, we preloaded the stressfield with the addition of elastic stress proportional to theamount of isotropic thermal contraction at the beginning ofeach temperature time step (once per Martian day) andallowed TECTON to compute the strain and viscousresponses. This procedure was repeated for three Mars yearsto eliminate initial conditions. Comparison of TECTON-based stresses with the analytic solution from Mellon [1997]shows excellent agreement with this approach. Some finiteelement nodes at the horizontal limits of the model wereallowed to move toward the polygon center to simulate theplacement of cracks and the associated relief of stress. Initialcrack formation and propagation is beyond the scope of thismodel, and is generally abrupt in terrestrial permafrost fromthe speed of sound [Leffingwell, 1915] to meters per hour[MacKay, 1993]. Once the initial fractures form, the sea-sonal strain cycles of these open cracks proceed indepen-dently of how they initially propagated. Crack closure wasmodeled by applying a resistive force at the crack nodes toprevent hyperextension. Temperatures were computed usinga standard Mars thermal model which included a two layersubsurface (ice-free soil atop saturated ice-rich material)

[Mellon et al., 2004] where the depth of the ice table wasallowed to vary. All temperatures were calculated for68.25�N and thermal properties of the surface consistentwith the landing site location. Figure 12b shows an exampleof temperature at the top of the ice table (the boundarybetween ice-free soil and the ice-rich layer), and the result-ing horizontal stress at the center of a 10-m diameterpolygon. We examine the stress at the ice table, becauseat this shallow depth the ice experiences the greatestseasonal temperature cycles and the greatest stresses.During fall and winter (Ls 180 and later) tensile stressesincrease rapidly owing to thermal contraction, reaching apeak in early winter. As the rate of cooling slows throughthe winter tensile stresses relax gradually. As the seasonsmove to spring and summer, surface CO2 frost accumulatedin winter dissipates and rapid warming and thermal expan-sion drives a change to compressive stress.[25] Figure 12c shows the wintertime tensile stress as a

function of horizontal distance from crack to crack across apolygon at the ice table for a few crack spacings (effectivepolygon diameters). For smaller polygons the stress isrelieved by crack strain and the central stress is low. Forlarger polygons the center of the polygon senses less of thepresence of the crack and the stresses remain high. If thetensile stress at the center of the polygon is higher than

Figure 11. Example of large-scale ridge-and-trough patterns. Ridges tend to exhibit small-scalepolygonal patterns aligned with the ridge and sharing the trough. In some cases, thin linear depressionscan be seen in wider troughs (arrows mark a few examples). Image PSP_002170_2485 is shown. Scenewidth is 500 m. Illumination is from the upper right.

Table 2. Ductile Flow Law Parameters for Polycrystalline Icea

DuctileRegime Temperature (K) A (Pa�n s�1) n Q (J mol�1)

A 273–234 10�12.2 4 91 � 103

B 234–195 10�18.9 4 61 � 103

C 195–150 10�39.8 6 39 � 103

aSeeDurham et al. [1992, 1997], where the ductile flow lawe =Asne�Q/RT

generally represents the relationship between strain rate e and deviatoricstress s.

E00A23 MELLON ET AL.: PERIGLACIAL LANDFORMS AT PHOENIX SITE

9 of 15

E00A23

the tensile strength, the polygon will need to subdivide byforming new cracks. The tensile strength of ice and ice-cemented soil was reviewed by Mellon [1997] and wasestimated to be between 2 and 3 MPa at Martian temper-atures and for length scales comparable to permafrostpolygons. By comparing the tensile stress at the center ofa polygon of different sizes with the tensile strength of theice-rich permafrost, the effective size of polygons expectedon Mars can be estimated.[26] Figure 13 show the peak seasonal horizontal stress at

the center of the a polygon for a range of crack spacings,and compares various parameters. In Figure 13a we com-pare pure ice and ice-rich soil by varying the rheology as inTable 3. Regardless of the specific rheology, the behavior issimilar. The smallest crack spacing (polygons size) produ-ces the least peak tensile stress. Increasing the crack spacingresults in increasing the peak stress owing to the decreasinginfluence of the cracks. As the crack spacing approaches50 m, the peak tensile stress approaches the maximum thatwould occur in unfractured material. Differences are alsoevident between the two materials. Tensile stresses aregenerally higher in ice-rich soil because the viscous relax-ation of stress by ductile creep is impeded by flow of icearound soil grains [Durham et al., 1997]. It follows that fora given tensile strength of the ice-rich matrix, pure poly-crystalline ice will result in larger polygons. For example, ifwe assume a minimum strength of 2 MPa for both materials,a polygon in ice-rich soil larger than 6 m would subdivide,while a polygon 5 m or smaller would not. If the permafrostconsisted of pure ice at and below the ice table, the polygonwould subdivide if it were much larger than about 10 m,almost twice the size of the of the ice-rich-soil polygon.[27] Figure 13b shows the effects of crack depth on the

peak stress and polygon size. In general, shallower cracksresult in higher stresses for a given crack spacing. Thisresult is due to a reduction in crack strain and less influenceon the stress field in the polygon interior. Also evident isthat for a given crack spacing, there exists a depth beyondwhich deeper penetration of the crack has no influence onthe stress at the polygon center. This behavior results in theinflections seen also in Figure 13a around 7 and 12 m.While shallower cracks result in smaller polygons, theshallowest cracks may not be realistic. Crack propagationis rapid and depends on factors such as the energy releaseby the formation of new surface area at the crack tip. Peakstresses at the crack tip as computed by TECTON provide agauge of whether or not a particular crack depth is reason-able. For shallow cracks and small polygons this crack-tipstress typically exceeds the tensile strength. For the smallest

polygons 3–5-m crack depths would be needed for thecrack-tip stress to be low enough to not drive cracks deeper.For polygons smaller than about 15 m, crack depth does notmatter beyond about 3-m depth (Figure 13b).[28] Figure 13c shows the effects of ice-table depth on the

peak stress and polygon size. Ice-table depth can have thegreatest influence on crack spacing and polygon size, andindeed can determine if polygons will form at all. Shallowice tables result in the highest stresses because the ice-richmaterial experiences the largest seasonal temperature oscil-lations which drive the thermal-contraction stress. Thedeeper the ice table, the less stress is produced. For ice-table depths approaching 20 cm the peek tensile stress nolonger exceed the minimum tensile strength for any crackspacing, thermal-contraction cracks will not form, andpolygonal-patterned ground will not develop in the presentMars climate. Conversely, for an ice-table depth less thanabout 6 cm polygons will form cracks with about 5-mspacing. Intermediate ice-table depths around 10 cm wouldresult in polygons around 15 m in diameter.

4. Discussion

[29] Small-scale polygons of order 3–6 m in diameter areubiquitous throughout the northern plains and the Phoenixlanding site. On Earth polygonal patterns can form from avariety of processes; lava cooling, desiccation, freeze-thaw‘‘convection,’’ and thermal contraction in permafrost. Thecold and ice-rich Martian environment present today at thislocation lends strong support to the interpretation that thepolygonal pattern ground is due to thermal contraction inpermanently frozen soil. This present climate is not condu-cive to a freeze-thaw process. Surface and subsurfacetemperatures at this latitude do not exceed the freezingpoint of water and temperatures of the ice at the ice table donot exceed about 212 K at any time during the year [see alsoMellon et al., 2008]. During geologically recent climatecycles ground temperatures and atmospheric conditionswould have differed [e.g., Toon et al., 1980; Mischna etal., 2003] affecting the stability of ground ice. However,under these expected climate conditions the temperatures atthe ice table are still not expected to reach melting [Mellonand Phillips, 2001]. An exception is that thin films of liquidwater may have occurred for brief periods in the upper 1–2 cm of the surface [Zent, 2008], but these quantities anddepths scales are insufficient to create the observed land-forms. Desiccation cracks (mud cracks) are similarly prob-lematic in the absence of liquid water. Lava cooling canunder the right circumstances produces polygonal fractures

Table 3. Viscoelastic Parameters

Polycrystalline Ice Icy Soil

Young’s modulus (E) 2.339 � 1010–6.48 � 107 T a same as polycrystalline iceb

Poisson’s ratio (n) 0.4c same as polycrystalline iceb

Thermal expansioncoefficient (a)

2.47 � 10�7 T–1.17 � 10�5 c 1.11 � 10�7 T–1.42 � 10�6 d

Flow law see Table 2 see Table 2; A = Ae�8bn e

aSee Gold [1958].bSee Mellon [1997] for a discussion of these parameters and differences and similarities between icy soil and pure ice.cSee Hobbs [1974].dAssuming a linear mixture of pure ice and basalt [see Mellon, 1997].eSee Durham et al. [1997]. Here b = 2, and 8 is the soil fraction. T is temperature in K.

E00A23 MELLON ET AL.: PERIGLACIAL LANDFORMS AT PHOENIX SITE

10 of 15

E00A23

[Peck and Minakami, 1968], but the geomorphology in thisregion lacks any evidence of a volcanic origin [Seelos et al.,2008].[30] Estimates of the current depth of the ice table are

about 2–6 cm based on a wide array of remote-sensingmethods and quantitative analysis, and for a range of spatialscales [Mellon et al., 2008]. While these methods andestimates are largely sensitive to the upper meter of thesurface, the presence of polygons is consistent with acohesive icy layer extending to at least 10-m depth. Thesize scale of these small polygons is remarkably consistentwith the numerical predictions of thermal-contractionstress for polygons on Mars with an ice table up to 6 cmdeep, and inconsistent with a deeper ice table or the absenceof ice altogether. In addition, these results indicate thatthe rheological properties over the upper meters to decame-ter of the surface are more consistent with the rheology ofice-cemented soil than that of pure ice (Figure 13a).Uncertainties in the numerical model and in the laboratorymeasurements of the rheology of icy soils under Mars-likeconditions make it difficult to precisely constrain the soil-to-ice ratio. However, the dependence of creep strength on thesoil content of the ice is exponential [Durham et al., 1997],indicating that 10–20% soil in the ice would be insufficient.This conclusion would seem in conflict with gamma ray andleakage neutron observations that indicate the ice content ofthe surface layer greatly exceeds the pore volume ofnominal soils and is as high as 85% ice by volume [e.g.,Feldman et al., 2008]. However, gamma ray and leakage-neutron observations are largely sensitive only to the uppercentimeters of the icy layer when such high concentrationsof ice are present. Polygonal ground forms in response torheological properties over many meters depth. Therefore,this result would suggest that the high concentration of iceobserved is confined to the upper reaches of the icy-soillayer and that the bulk of the subsurface material consists ofice-cemented soil (see Mellon et al. [2008] for furtherdiscussion).[31] The morphology of the small polygons is consistent

with a geologically recent period of long-term development.An abundance of equiangular three-way intersections onMars may be indicative of advanced network developmentand age. Sletten et al. [2003] present observations whichsuggest that in antarctic permafrost younger thermal-contraction polygons begin with orthogonal trough inter-sections and minimal surface relief, while over time thesepatterns evolve into equiangular intersections and developsubstantial topographic relief. Alternatively, the evolutionfrom orthogonal to equiangular (hexagonal) networks maybe related to the progressive subdivision of patterns asnetworks age [Black, 1976], or to an increase in sinuosityof the primary fractures [Plug and Werner, 2001] related tosubsurface heterogeneity. In addition, the superposition ofsmall polygons over all other features (including all smallcraters; see Figure 9d) suggests that they are the most recentdevelopment on this surface. Other features such as rubblepiles and larger polygonal patterns do not appear to disruptthe small polygon. The small polygons do not appear to beeroded and many of the troughs are distinct and sharp withoccasional smaller thin depressions within the troughs,which may indicate active consumption of surface soildirectly over an open fracture. While the absence of ice in

Figure 12. (a) Schematic of a rheological model ofpermafrost, shown with ridged boundaries before fracture,while postfracture would include a component of strain.(b) Seasonal temperature and horizontal stress (sxx) at theice table as a function of season at the center of a 10-mpolygon, assuming an ice-table depth of 4 cm. Temperaturesare determined form a standard thermal model (see text)assuming 68.25�N, a surface thermal inertia of 200 J m�2

s�1/2 K�1, an albedo of 0.2, and an atmospheric pressure of7.24 mb consistent with the elevation at the Phoenix landingsite. Tensile stresses greater than about 2–3 MPa willfracture the permafrost (see text). (c) Winter horizontaltensile stress at the ice table across the polygon (crack-to-crack) for 5-, 10-, and 20-m-diameter polygons. Stress isrelieved near the crack with the least effect (most stress) atthe polygon center. The peak central stress increases as thepolygon diameter (crack spacing) increases.

E00A23 MELLON ET AL.: PERIGLACIAL LANDFORMS AT PHOENIX SITE

11 of 15

E00A23

a once ice-rich region that developed extensive polygonswould leave behind a remnant polygonal surface expressionthat we would see today, the measured presence of ice todaysuggests that these polygons are currently active.[32] The development of thermal-contraction polygons is

a slow process. In any one year the crack may only open afew millimeters and many thousands of years are required todevelop the topography that we observe [Lachenbruch,

1962; Black, 1976]. However, these polygons also appeardistinct and undegraded, suggesting that the ice table haspersisted at this shallow depth for some time withoutmoving significantly under the influence of recent climateoscillations. Combined with observations of the currentdepth distribution of ice, this morphological observationsupports the idea that the current distribution of ground iceand the current Martian climate have been stable for manythousands of years or longer. If the climate were to haverecently changed dramatically causing the ice table tomigrate to shallower and deeper depths, the size anddevelopment of polygons would be irregular. A period ofrecent stability is consistent with the most recent quiescentperiod in the obliquity which has lasted about 300 ka [e.g.,Ward, 1992; Laskar et al., 2002].[33] The wide spread uniformity of small polygons in

both size and morphology suggest a uniform distribution ofthe ice table and physical characteristics of the ice-cementedlayer. Although, some morphological difference are appar-ent (see Figure 4), these may be related to local differencesin the rate of sand-wedge growth or local dust mantling andaeolian reworking, rather than variations in ice-table depthor ice content.[34] The presence of regularly spaced rubble piles and of

rubble arcs connecting the piles (Figure 8) is suggestive ofperiglacial self organization. However, their specific originlacks a good terrestrial analog. On Earth, rock sorting cantake place in a freeze-thaw environment creating circularand polygonal rock patterns through convection-like over-turning of the soil driven by differential frost heave [Halletand Prestrud, 1986]; geologically recent surface temper-atures on Mars are not expected to support such a process.Alternatively, the long-term growth of thermal-contractionsand-wedge polygons in permanently frozen environmentsmay also drive overturning (cryoturbation) and self organi-zation of rocks. Sand-wedge growth will continually pushfine grained material into the subsurface of the polygoninterior [e.g., Pewe, 1959], uplifting the surface, followedby surface creep back toward the troughs [Sletten et al.,2003]. Cobble and boulder sized rocks may slump into thepolygon trough by gravity driven creep [Pewe, 1959;Washburn, 1956], where they would become trapped abovethe thin contraction cracks. The specific details of thisproposed mechanism are not well understood and terrestrialanalogs are rare. Even the oldest terrestrial sand-wedge

Figure 13. Peak stress at the center of a polygon as afunction of crack spacing for (a) pure ice and ice-rich soilrheology (see Table 3); (b) ice-rich soil and a range of crackdepths (0.5–7.25 m); and (c) ice-rich soil and a range of ice-table depths (1–30 cm). Plusses in Figure 13a indicateactual model crack spacing nodes. Dotted lines show thestress that would occur if no fractures are allowed to occur.Assuming a minimum tensile strength of 2 MPa, for thecase of ice tables 2- to 6-cm deep, the nominal crackspacing is close to 5 m, similar to the observed small-polygon size. Larger 20-m polygons would be consistentwith an ice table slightly greater than 10 cm. If the ice tablewere as deep as 20 cm or more, thermal contraction stresseswould be insufficient to form fractures and polygons wouldnot develop.

E00A23 MELLON ET AL.: PERIGLACIAL LANDFORMS AT PHOENIX SITE

12 of 15

E00A23

polygons are young relative to the Martian environment.Perhaps the best example is in Beacon Valley Antarcticawhere the polygonal-patterned surface may be as much as8 million years old [Sugden et al., 1995] and the slowprocess of sand-wedge growth has consumed the polygoninteriors [Sletten et al., 2003]; in this location some sortingmay be occurring (Figure 14b) where cobbles and bouldersare accumulated directly over the cracks and relativelyabsent from the trough shoulders.[35] In this scenario of thermal-contraction-based cryo-

turbation and overturning of the upper meters of the surfacelayer, the scale of the rubble-pile patterns would suggest alarger 20–35-m scale polygon network. The observed 20–25-m circular-to-hexagonal depression-type of large poly-gons (Figure 6) may be related this process, though norubble or rubble sorting has yet been observed in theselarger patterns. However, Figure 14a shows an example ofrocks and boulders sorted into the polygon troughs ofsmaller 3- to 6-m polygons at similar latitudes to Phoenix,but in a location exhibiting generally higher rock abundan-ces. Many of these troughs contain no rocks or boulderswhich indicates downward consumption of surface finesand active thermal contraction is the driver for the polygonformation. This example shows that a similar mechanismfor formation of the rubble piles and arcs is indeed viable.[36] Rubble piles and arcs tend to sit atop the local high

ground (mounds and ridges typically about 50 cm higherthan the intermound terrain [Kirk et al., 2008]). Thisplacement suggests that either they formed in the centralportions of the polygons and the now degraded troughs liein the lower intermound areas, or that they formed as rubblecollected in the polygon troughs and trough intersectionsfollowed by subsequent erosion of the surface (e.g., aeoliandeflation) removing material between the rubble piles. Thelater scenario would be consistent with the rubble pilesprotecting the underlying surface from erosion or eventrapping windblown material, with the rubble-pile-freeregions eroding to slightly lower in elevation. In this waythe rubble collected in the polygon troughs may have

formed rubble arcs, similar to Figure 14, while rubbleconcentrated at the trough intersections formed the piles.Subsequent erosion of surface soils coincident with orfollowed by extensive small-polygon formation may haveresulted in the landscape we now see. The larger hexagonal-to-circular-depression polygons that we see in the Phoenixlanding site region (Figure 6) may be a remnant of thisepisode of polygon formation but in an area that lackedlarge rocks and boulders to sort and was not subjected to theerosion seen in other areas. An observational test of thisscenario would be to locate examples of rubble-pile terrainwhere the larger polygonal pattern remains intact. Addition-ally, the Phoenix Lander can image the local small-scalepolygon troughs to look for rock sorting at a scale notobservable by HiRISE.[37] The presence of contiguous ridge-and-trough fea-

tures suggests cyclic fracturing on long horizontal scalesor an alignment of smaller polygons along an existing zoneof weakness. While larger-scale fracturing could be theresult of long-term climate cycles in temperature andcontraction stress, several lines of evidence contradict thishypothesis. The surface temperature cycles caused byorbital oscillations such as obliquity are actually relativelysmall (<10 K) at these latitudes and smaller still at the depthif the ice table [Mellon and Jakosky, 1995; Mellon andPhillips, 2001]. Combined with the long 105-year timescaleof these oscillations, stresses cannot adequately build tofracture the permafrost. In addition, thermal contraction istypically isotropic and would form enclosed polygonalpatterns rather than the typical subparallel patternsobserved. Their NW-SE trend and occasional concentricghost-crater alignment suggests that some topographic ortectonic control is helping to align and exaggerate the ridgesand troughs. The horizontal component of gravity on slopescan help align small polygon fractures and exaggerate thetroughs to best relieve local stresses.[38] The inferred larger-scale polygonal patterns pro-

posed to have controlled the formation of rubble piles andthe observed larger-scale polygons suggest a period of

Figure 14. Self-organization of rocks and boulders on Earth and Mars. (a) Meter-scale and smallerrocks and boulders are collected into polygonal patterns. (Image PSP_001381_2485 is shown at68.428�N � 164.183�E, at 31.5 cm pixel�1). The presence of rock-free troughs suggests an underlyingpattern of thermal contraction polygons may be controlling the rock distribution. Illumination is from theupper right. (b) Rocks and boulder sorting in sand-wedge polygon trough in Beacon Valley Antarctica(Image credit: Mellon. 77.8�S � 160.7�E). In this example the rocks are largely absent from the shouldersof the troughs and are collected along the center of the troughs directly over the open cracks, suggesting amigration of rocks to the center of the trough (see text). Scene width is about 6 m.

E00A23 MELLON ET AL.: PERIGLACIAL LANDFORMS AT PHOENIX SITE

13 of 15

E00A23

polygon formation and cryoturbation in which the ground-ice table was perhaps 10–12 cm deep depending on thetensile strength (see Figure 13c), only slightly different thantoday. During periods of high obliquity Mellon and Jakosky[1995] showed that increased global atmospheric humidityand decrease equatorial ground temperatures allow groundice to become stable in equatorial regions. However, at thePhoenix landing site latitudes increased ground temperatureat higher obliquity causes the ice table to recede to slightlydeeper depths despite the increase in humidity This situationwould be consistent with the formation of larger polygons.[39] A problem with the interpretation that the polygon

size is related to the obliquity-driven ice-table-depth varia-tions, is that this depth is expected to vary smoothly overthe course of 105 year climate cycles and polygon sizeswould be expected to vary producing irregular networks.While the most recent quiescent period in the obliquityhistory (the last 300 ka) is consistent with a stable ice-tabledepth and small polygon development, higher obliquitiesare typically achieved in alternation between high and lowvalues at times longer than 300 ka ago. Coincident cycleswould occur in the ice-table depth. Two possible explan-ations are that: (1) a bimodal distribution of polygon sizesare active throughout time and one size or the otherbecomes active during a particular state of ice-table depth;and (2) that the polygon formation process shuts down atlow obliquity owing to the lack of strong seasonal temper-ature cycles.[40] In general, the pervasive presence of polygons at all

scales, the degree of polygon development, and the sortingof rocks into rubble piles indicate a long-term activity ofperiglacial processes. In particular, these features suggestmassive cryoturbation and reworking in the upper meters ofthe surface over recent and possibly longer-term geologichistory at the Phoenix landing site and over the greaternorthern plains. Such overturning of the surface soil layerwould act to raise rocks and boulders to the surface and sortthem, reduce general topographic relief, erase small impactcraters (e.g., Figure 9d), and mute the topography of largecraters producing ghost craters. In addition, any period ofperiglacial activity prior to the formation of large polygonsand rubble piles has likely been erased by the most recentreworking of the surface by the present polygons.[41] The Phoenix mission can provide a number of

key observations that can help evaluate these hypotheses.Imaging of the local polygon topography to examine thescale of polygonal patterns at the landing site can help:determine if the 3- to 6-m polygons are indeed the smallestscale or if smaller-scale features are present below theresolution of HiRISE; examine the nature of uplift andtrough morphology; and allow more precise measurementsof polygon size through stereo imaging and provide groundtruth for HiRISE polygon size interpretations. Additionally,Phoenix will be able to examine the distribution of decime-ter-scale rocks, which are below HiRISE resolution, andlook for evidence of sorting within polygon troughs relativeto the polygon interiors. Phoenix will dig into the permafrostand determine the actual depth of ground ice; knowledge ofthe ice-table depth will help constrain the model of perma-frost stress. Phoenix will examine the nature of surface soils,ground-ice textures, gradients or uniformity in soil chemis-try, and the D/H ratio in the subsurface ice relative to the

atmosphere, which may provide clues about the effects ofcryoturbation and the general mixing of the surface soils.

5. Summary

[42] We have analyzed landforms at the Phoenix landingsite and at similar latitudes in the northern plains of Marsand have found several types of feature that are analogous tothose found in terrestrial permafrost. Polygonal patterns arethemost common consisting of regular networks of perimetertroughs and uplifted portions of the polygon interior. Small-scale 3- to 6-m polygonal-patterned ground is ubiquitous inthe Martian northern plains and at the Phoenix landing site.Numerical models of the process of thermal contraction inMartian permafrost, combined with numerous predictionsthat the ice table in this region is only 2–6 cm below thesurface, indicate that these small polygons are consistent withan origin by thermal-contraction cracking in ice-rich soil.These small polygons appear to be the most recent feature bytheir well developed morphology, and by their consistencywith the current distribution of ground ice and the process ofpolygon formation. The distribution of these features alsoindicates that ground ice is uniformly shallow throughout theregion to within a few centimeters depth.[43] Regularly spaced rubble piles (localized collections

of rocks and boulders) and connecting rubble arcs andbands are also abundant and may be due to an earlierepisode of larger polygon formation that sorted rocks ontothe surface and into local collections. The spacing of thesepiles (20–35 m) is similar to larger-scale (20–25 m)polygonal patterns that are also observed but devoid ofrubble. Some areas outside the Phoenix landing site at thesame latitude are dominated by the small-scale polygonsthat exhibit evidence of rock and boulder sorting into thepolygon troughs. This observation supports the notion thatpermafrost polygons can sort rocks and boulders into therubble piles in the absence of any freeze-thaw cycles.[44] The likely history of this region begins with a period

of ground ice depth around 10–12 cm and large polygonformation, concurrent with sorting of surface rocks andboulders into piles and arcs. Such a period is largelyconsistent with the occurrence of higher obliquity in thepast. This epoch was followed by a period of small-scalepolygon development with a shallower and stable 6-cm orless ice-table depth, leading to the present day. This overallhistory is largely consistent with the most recent quiescentobliquity history (the past 300 ka) preceded by a long periodof higher obliquity, though the effects of large-amplitudeobliquity cycles prior to 300 ka ago is not understood. Abetter understanding of this history and its relation to orbitaloscillations may come from specific modeling of the rate ofpolygon formation and development and of rock sortingunder various Martian climate conditions. Any earlierhistory of polygonal-patterned ground and periglacial pro-cesses has likely been erased by the most recent activity.[45] The rock sorting and widespread polygon develop-

ment at all scales indicates that a massive reworking of theupper meters of the surface has occurred on timescales ofthe past few million years, which has significant implica-tions for the development of the landscape of the northernplains and specifically the Phoenix landing site. In particularthis geologically recent process has likely resulted in an

E00A23 MELLON ET AL.: PERIGLACIAL LANDFORMS AT PHOENIX SITE

14 of 15

E00A23

overall muting of topographic features such as small andlarge craters, and the sorting of rocks and boulders onto thesurface and in some cases into distinct localized piles.

[46] Acknowledgments. We wish to thank Jay Melosh for graciouslyproviding the finite element model TECTON for use in this research. Wewould also like to thank the HiRISE and MRO teams for their support inacquiring the Phoenix landing site images used in this study.

ReferencesArvidson, R. E., et al. (2008), Mars Exploration Program 2007 Phoenixlanding site selection and characteristics, J. Geophys. Res., 113, E00A03,doi:10.1029/2007JE003021.

Black, R. F. (1976), Periglacial features indicative of permafrost: Ice andsoil wedges, Quat. Res., 6, 3–26, doi:10.1016/0033-5894(76)90037-5.

Boynton, W. V., et al. (2002), Distribution of hydrogen in the near surfaceof Mars: Evidence for subsurface ice deposits, Science, 297, 81–85,doi:10.1126/science.1073722.

Durham, W. B., S. H. Kirby, and L. A. Stern (1992), Effects of dispersedparticulates on the rheology of water ice at planetary conditions, J. Geo-phys. Res., 97, 20,883–20,897.

Durham, W. B., S. H. Kirby, and L. A. Stern (1997), Creep of water ices atplanetary conditions: A compilation, J. Geophys. Res., 102, 16,293–16,302, doi:10.1029/97JE00916.

Feldman, W. C., M. T. Mellon, O. Gasnault, S. Maurice, and T. H. Prettyman(2008), Volatiles on Mars: Scientific results from the Mars OdysseyNeutron Spectrometer, in The Martian Surface: Composition, Mineralogy,and Physical Properties, edited by J. F. Bell, pp. 125–148, CambridgeUniv. Press, New York.

Gold, L. W. (1958), Some observations of the dependence of strain in stressfor ice, Can. J. Phys., 36, 1265–1275.

Hallet, B., and S. Prestrud (1986), Dynamics of periglacial sorted stonecircles in western Spitsbergen, Quat. Res., 26, 81–99, doi:10.1016/0033-5894(86)90085-2.

Head, J. W., F. Mustard, M. A. Kreslavsky, R. E. Milliken, and D. R.Marchant (2003), Recent ice ages on Mars, Nature, 426, 797–802,doi:10.1038/nature02114.

Hillgren, V., and H. J. Melosh (1989), Crater relaxation on Ganymede:Implications for ice rheology, Geophys. Res. Lett., 16, 1339–1342,doi:10.1029/GL016i011p01339.

Hobbs, P. V. (1974), Ice Physics, Oxford Univ. Press, New York.Kirk, R. L., et al. (2008), Ultrahigh-resolution topographic mapping ofMars with MRO High Resolution Imaging Science Experiment stereoimages: Meter-scale slopes of candidate Phoenix landing sites, J. Geo-phys. Res., doi:10.1029/2007JE003000, in press.

Lachenbruch, A. H. (1962), Mechanics of thermal contraction cracks andice-wedge polygons in permafrost, Geol. Soc. Am. Spec. Pap., 70, 1–69.

Laskar, J., B. Levard, and J. F. Mustard (2002), Orbital forcing of theMartianpolar layered deposits, Nature, 419, 375–377, doi:10.1038/nature01066.

Leffingwell, E. K. (1915), Ground-ice wedges: The dominant form ofground ice on the north coast of Alaska, J. Geol., 23, 635–654.

Leighton, R. B., and B. C. Murray (1966), Behavior of carbon dioxide andother volatiles on Mars, Science, 153, 135–144, doi:10.1126/science.153.3732.136.

Lucchitta, B. K. (1981), Mars and Earth: Comparison of cold-climatefeatures, Icarus, 45, 264–303, doi:10.1016/0019-1035(81)90035-X.

Mackay, J. R. (1993), The sound and speed of ice-wedge cracking, ArcticCanada, Can. J. Earth Sci., 30, 509–518.

Malin, M. C., and K. S. Edgett (2001), Mars Global Surveyor MarsObserver Camera: Interplanetary cruise through primary mission,J. Geophys. Res., 106, 23,429–23,571, doi:10.1029/2000JE001455.

Mangold, N. (2005), High latitude patterned ground on Mars: Classifica-tion, distribution and climate control, Icarus, 174, 336–359, doi:10.1016/j.icarus.2004.07.030.

Marchant, D. R., A. R. Lewis, W. M. Phillips, E. J. Moore, R. A. Souchez,G. H. Denton, D. E. Sugden, N. Potter, and G. P. Landis (2002), Formationof patterned ground and sublimation till overMiocene glacier ice in BeaconValley, southern Victoria Land, Antarctica, Geol. Soc. Am. Bull., 114,718–730, doi:10.1130/0016-7606(2002)114<0718:FOPGAS>2.0.CO;2.

McEwen, A. S., et al. (2007), Mars Reconnaissance Orbiter’s High Resolu-tion Imaging Science Experiment (HiRISE), J. Geophys. Res., 112,E05S02, doi:10.1029/2005JE002605.

Mellon, M. T. (1997), Small-scale polygonal features on Mars: Seasonalthermal contraction cracks in permafrost, J. Geophys. Res., 102, 25,617–25,628, doi:10.1029/97JE02582.

Mellon, M. T., and B. M. Jakosky (1995), The distribution and behavior ofMartian ground ice during past and present epochs, J. Geophys. Res.,100, 11,781–11,799, doi:10.1029/95JE01027.

Mellon, M. T., and R. J. Phillips (2001), Recent gullies on Mars and thesource of liquid water, J. Geophys. Res., 106, 23,165 – 23,180,doi:10.1029/2000JE001424.

Mellon, M. T., W. C. Feldman, and T. H. Prettyman (2004), The presenceand stability of ground ice in the southern hemisphere of Mars, Icarus,169, 324–340, doi:10.1016/j.icarus.2003.10.022.

Mellon, M. T., W. V. Boynton, W. C. Feldman, J. L. Bandfield, T. N. Titus,D. A. Paige, A. Hock, N. E. Putzig, R. E. Arvidson, and H. G. Sizemore(2008), A prelanding assessment of the ice-table depth and groundice characteristics in Martian permafrost at the Phoenix landing site,J. Geophys. Res., doi:10.1029/2007JE003067, in press.

Melosh, H. J., and A. Raefsky (1980), The dynamical origin of subductionzone topography, Geophys. J. R. Astron. Soc., 60, 333–354.

Melosh, H. J., and A. Raefsky (1981), A simple and efficient method forintroducing faults into finite element computations, Bull. Seismol. Soc.Am., 71, 1391–1400.

Mischna, M. A., M. I. Richardson, R. J. Wilson, and D. J. McCleese (2003),On the orbital forcing of Martian water and CO2 cycles: A generalcirculation model study with simplified volatile schemes, J. Geophys.Res., 108(E6), 5062, doi:10.1029/2003JE002051.

Peck, D. L., and T. Minakami (1968), Formation of columnar joints inthe upper part of the Kilauea Lava Lake, Geol. Soc. Am. Bull., 79,1151–1169, doi:10.1130/0016-7606(1968)79[1151:TFOCJI]2.0.CO;2.

Pewe, T. L. (1959), Sand-wedge polygons (tessellations) in the McMurdoSound region, Antarctica: A progress report, Am. J. Sci., 257, 545–552.

Pewe, T. L. (1974), Geomorphic processes in polar deserts, in PolarDeserts and Modern Man, edited by T. L. Smiley and J. H. Zumberge,pp. 33–52, Univ. of Ariz. Press, Tucson, Ariz.

Plug, L. J., and B. T. Werner (2001), Fracture networks in frozen ground,J. Geophys. Res., 106, 8599–8613, doi:10.1029/2000JB900320.

Seelos, K. D., F. P. Seelos, R. E. Arvidson, S. Cull, S. L. Murchie,P. McGuire, and T. J. Parker (2008), Geomorphologic and mineralogiccharacterization of the northern plains of Mars at the Phoenix Missioncandidate landing sites, J. Geophys. Res., 113, E00A13, doi:10.1029/2008JE003088.

Seibert, N. M., and J. S. Kargel (2001), Small-scale Martian polygonalterrain: Implications for liquid surface water, Geophys. Res. Lett., 28,899–902, doi:10.1029/2000GL012093.

Sletten, R. S., B. Hallet, and R. C. Fletcher (2003), Resurfacing time ofterrestrial surfaces by the formation and maturation of polygonal patternedground, J. Geophys. Res., 108(E4), 8044, doi:10.1029/2002JE001914.

Squyres, S. W., and M. H. Carr (1986), Geomorphic evidence for thedistribution of ground ice on Mars, Science, 231, 249 – 252,doi:10.1126/science.231.4735.249.

Sugden, D. E., D. R. Marchant, N. Potter, R. A. Souchez, G. H. Denton,C. C. Swisher III, and J.-L. Tison (1995), Preservation of Miocene glacierice in east Antarctica, Nature, 376, 412–414, doi:10.1038/376412a0.

Toon, O. B., J. B. Pollack, W. Ward, J. A. Burns, and K. Bilski (1980), Theastronomical theory of climate change on Mars, Icarus, 44, 552–602,doi:10.1016/0019-1035(80)90130-X.

Ward, W. R. (1992), Long-term orbital and spin dynamics of Mars, inMars,edited by H. H. Kieffer et al., pp. 298–320, Univ. of Ariz. Press, Tucson,Ariz.

Washburn, A. L. (1956), Classification of patterned ground and review ofsuggested origins, Geol. Soc. Am. Bull., 67, 823–865, doi:10.1130/0016-7606(1956)67[823:COPGAR]2.0.CO;2.

Washburn, A. L. (1980), Geolcrology: A Survey of Periglacial Processesand Environments, John Wiley, Hoboken, N. J.

Zent, A. P. (2008), An historical search for habitable ice at the Phoenixlanding site, Icarus, 196, 385–408.

�����������������������R. E. Arvidson, Department of Earth and Planetary Sciences, Washington

University, Campus Box 1169, 1 Brookings Drive, St. Louis, MO 63130,USA.E. Asphaug, Department of Earth and Planetary Science, University of

California, Earth and Marine Sciences Building, Santa Cruz, CA 95064,USA.J. J. Marlow, Department of Earth Science and Engineering, Imperial

College London, South Kensington Campus, London, SW7 2AZ, UK.M. T. Mellon, Laboratory for Atmospheric and Space Physics, University

of Colorado, 392 UCB, Boulder, CO 80309, USA. (michael.mellon@lasp.colorado.edu)R. J. Phillips, Southwest Research Institute, 1050Walnut Street, Suite 300,

Boulder, CO 80302, USA.

E00A23 MELLON ET AL.: PERIGLACIAL LANDFORMS AT PHOENIX SITE

15 of 15

E00A23