Distribution, morphology, and origins of Martian pit crater chains

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 109,, doi: /2004je002240, 2004 Distribution, morphology, and origins of Martian pit crater chains Danielle Wyrick and David A. Ferrill CNWRA, Southwest Research Institute, San Antonio, Texas, USA Alan P. Morris Department of Earth and Environmental Science, University of Texas at San Antonio, San Antonio, Texas, USA Shannon L. Colton and Darrell W. Sims CNWRA, Southwest Research Institute, San Antonio, Texas, USA Received 21 January 2004; revised 18 March 2004; accepted 21 April 2004; published 11 June [1] Pit craters are circular to elliptical depressions found in alignments (chains), which in many cases coalesce into linear troughs. They are common on the surface of Mars and similar to features observed on Earth and other terrestrial bodies. Pit craters lack an elevated rim, ejecta deposits, or lava flows that are associated with impact craters or calderas. It is generally agreed that the pits are formed by collapse into a subsurface cavity or explosive eruption. Hypotheses regarding the formation of pit crater chains require development of a substantial subsurface void to accommodate collapse of the overlying material. Suggested mechanisms of formation include: collapsed lava tubes, dike swarms, collapsed magma chamber, substrate dissolution (analogous to terrestrial karst), fissuring beneath loose material, and dilational faulting. The research described here is intended to constrain current interpretations of pit crater chain formation by analyzing their distribution and morphology. The western hemisphere of Mars was systematically mapped using Mars Orbiter Camera (MOC) images to generate ArcView TM Geographic Information System (GIS) coverages. All visible pit crater chains were mapped, including their orientations and associations with other structures. We found that pit chains commonly occur in areas that show regional extension or local fissuring. There is a strong correlation between pit chains and fault-bounded grabens. Frequently, there are transitions along strike from (1) visible faulting to (2) faults and pits to (3) pits alone. We performed a detailed quantitative analysis of pit crater morphology using MOC narrow angle images, Thermal Emission Imaging System (THEMIS) visual images, and Mars Orbiter Laser Altimeter (MOLA) data. This allowed us to determine a pattern of pit chain evolution and calculate pit depth, slope, and volume. Volumes of approximately 150 pits from five areas were calculated to determine volume size distribution and regional trends. The information collected in the study was then compared with non-martian examples of pit chains and physical analog models. We evaluated the various mechanisms for pit chain development based on the data collected and conclude that dilational normal faulting and sub-vertical fissuring provide the simplest and most comprehensive mechanisms to explain the regional associations, detailed geometry, and progression of pit chain development. INDEX TERMS: 5475 Planetology: Solid Surface Planets: Tectonics (8149); 6225 Planetology: Solar System Objects: Mars; 6207 Planetology: Solar System Objects: Comparative planetology; 8149 Tectonophysics: Planetary tectonics (5475); KEYWORDS: chains, Mars, pits Citation: Wyrick, D., D. A. Ferrill, A. P. Morris, S. L. Colton, and D. W. Sims (2004), Distribution, morphology, and origins of Martian pit crater chains, J. Geophys. Res., 109,, doi: /2004je Introduction [2] Pit crater chains are linear features consisting of collapse pits and troughs. They have been identified on Earth [Okubo and Martel, 1998], Venus [Bleamaster and Hansen, 2001], Moon [Head and Wilson, 1993], Mars Copyright 2004 by the American Geophysical Union /04/2004JE [Banerdt et al., 1992] and small planetary bodies such as Phobos [Thomas, 1979], a Martian satellite. Pit craters, pit crater chains, and troughs are common on Mars and are found in a range of settings from the flanks of large volcanoes (e.g., Alba Patera) to flat-lying floors of broad basins (e.g., Utopia Planitia). It is generally agreed that these features are formed by collapse into a subsurface cavity or by explosive eruption [Tanaka and Golombek, 1989; Spencer and Fanale, 1990; Davis et al., 1995; Mège 1of20

2 Figure 1. MOC narrow angle images showing stratification and differential erosion of the pit walls. (a) Pits aligned with normal faults in the Tractus Catena region, center of image approximately 28.5 N, W. Image credit NASA/JPL/MSSS, MOC image E (b) Pits in the Tractus Fossae region show stratification and differential erosion, center of image approximately 23.5 N, W. Image credit NASA/JPL/MSSS, MOC image M Stars with arrows show illumination direction. and Masson, 1996, 1997; Tanaka, 1997; Liu and Wilson, 1998; Mège, 1999a; Montesi, 1999; Mège et al., 2000; Scott et al., 2000; Gibbons et al., 2001; Mège et al., 2002, 2003; Scott and Wilson, 2002; Wilson and Head, 2002; Ferrill et al., 2003; Wyrick et al., 2003]. The reliance on remote sensing imagery, with its limited resolution and 3-D imaging capabilities, has made it difficult to analyze these relatively small features (1 4 km wide) and pit craters remain poorly understood. Our work constrains interpretations of pit crater chain formation by analyzing their distribution and morphology, using combined data sets (MOC, MOLA, THEMIS) along with GIS and 3D software (ArcView TM, Earth Vision TM, Imagine TM ). On the basis of pit morphology, volume calculations, structural associations (e.g., normal faults, grabens), vertical offset of pits and general lack of associated volcanic features, our results indicate that most pit chains in the western hemisphere of Mars form over dilational normal faults or extension fractures. 2. Pit Crater Morphology [3] Pit craters are collapse structures that lack the elevated crater rim, ejecta deposits or lava flows that are typically associated with impact craters, phreatic volcanic eruptions, or calderas. Individual pits typically have a conical shape with or without a flat floor and commonly occur in linear chains. Pit walls reveal stratification and are differentially eroded; gullies are common in the upper flanks of pit craters (Figure 1). Pits in many cases have an elliptical plan shape with the long axis parallel to the chain orientation. There are numerous examples of coalesced pits forming scalloped troughs. Pits are often bounded by a graben, although they are also found apart from grabens. Where associated with a graben, pits commonly occur along the bounding faults. 3. Proposed Mechanisms [4] Hypotheses regarding the formation of pit crater chains require a substantial subsurface void to accommodate collapse. Suggested mechanisms of formation include (Table 1): collapsed lava tubes (e.g., NASA/JPL, available online at dike swarms [Mège and Masson, 1996, 1997; Liu and Wilson, 1998; Mège, 1999a; Montesi, 1999; Scott et al., 2000, 2002; Gibbons et al., 2001; Wilson and Head, 2001, 2002; Scott and Wilson, 2002]; dike swarms possibly associated with collapsed magma chambers [Mège et al., 2000, 2002, 2003]; karst dissolution [Spencer and Fanale, 1990]; fissuring beneath loose material [Tanaka and Golombek, 1989; Banerdt et al., 1992; Tanaka, 1997]; and dilational faulting [Ferrill and Morris, 2003; Ferrill et al., 2003; Sims et al., 2003; Wyrick et al., 2003]. We analyzed these models in light of the most recent data from Mars, terrestrial examples and physical analog modeling to determine their viability Lava Tubes [5] Lava tubes are frequently cited on NASA websites as the fundamental mechanism of pit chain formation (e.g., NASA/JPL, available online at gov/catalog/?idnumber=pia03836; Figure 2a). Terrestrial lava tubes form in fluid basaltic lava flows. Conduits form within the flow and channel the lava much like a river. As the lava overhead is cooled from exposure to the surface air, it forms a roof above the lava tube [Bardintzeff and McBirney, 2000]. When the lava supply ceases or is diverted elsewhere, 2of20

3 Table 1. Proposed Mechanisms for Pit Chain Formation Proposed Mechanism Reference Description Characteristics Comments Lava Tubes NASA/JPL websites Lava tube roof collapse Volcanic (basaltic) lava flows Cannot explain pit chains that Sinuous in plan view Pit chains and troughs oriented down slope propagate perpendicular to the topographic slope; most pit chains in study were linear Dikes (Hydrosphere/Cryosphere Interaction) Mège and Masson [1996, 1997], Mège [1999a], Montesi [1999] Propagating dike interacting with subsurface groundwater/cryosphere Trace to magma source, likely in radial and/or concentric orientations Evidence of eruption (cinder cones, lava flows, fallout deposits) Evidence of water/ice source No direct evidence of cinder cones, lava flows, maar eruption deposits Dikes (Plinian Eruption) Scott et al. [2000, 2002], Gibbons et al. [2001], Scott and Wilson [2002] Collapse after dike erupts in Plinian-type eruption with high dispersal rates Trace to magma source, likely in radial and/or concentric orientations Evidence of eruption (cinder cones, lava flows, fallout deposits) Consistent morphology as described by authors (e.g., craters occur in regularly repeated cycle of sizes; large pits always within graben) No direct evidence of cinder cones, l ava flows, pyroclastic fallout/flow deposits; morphologies did not consistently match prediction Dikes (Volatile Exsolution) Liu and Wilson [1998], Scott and Wilson [2002], Wilson and Head [2001, 2002] Collapse into void by leakage of volatiles at top of dike Trace to magma source, likely in radial and/or concentric orientations Evidence of eruption (cinder cones, lava flows, fallout deposits) Volatile volume versus pit volume No direct evidence of cinder cones, lava flows; volumes inconsistent with large pit formation Collapsed Magma Chamber Mège et al. [2000], Mège et al. [2002, 2003] Collapse of elongate magma reservoirs Trace to magma source, likely in radial and/or concentric orientations Evidence of eruption (cinder cones, l ava flows, fallout deposits, small scale volcanic edifices) Evidence of reverse faulting along centerline, steep-walled or terraced pits No direct evidence of cinder cones, lava flows; pits were not terraced, steep walled or bounded by reverse faulting Karst Dissolution Spencer and Fanale [1990] Chemical dissolution of soluble (e.g., carbonate) rock Terrestrial karst-like topography with structurally controlled dissolution, sinkholes Evidence for an active hydrological cycle Evidence for soluble (e.g., carbonate) rocks Pit chains found in non-equatorial regions; spectroscopic evidence for carbonate is lacking Tension Fractures Tanaka and Golombek [1989], Banerdt et al. [1992], Tanaka [1997] Tension fractures from extension; may be enhanced by weathering & erosion Extensional stress field Model fits some pit chains, but difficult Opening of fractures, enlargement by weathering and erosion Evidence for extension (e.g., normal faults) to incorporate into large scale pit chains with vertical offset Dilational Faulting Ferrill and Morris [2003], Ferrill et al. [2003], Sims et al. [2003], Wyrick et al. [2003] Tensile or hybrid failure along normal faults resulting in dilational segments Opening of dilational segments by displacement along normal faults Evidence of near surface mechanical stratigraphy Evidence for extension (e.g., normal faults) Model fits most pit chain morphologies; consistently found with tectonic structures (e.g., normal faults) 3of20

4 Figure 2. Schematic illustration of proposed mechanisms for pit chain formation. (a) Lava tubes. (b) Dike with hydrosphere/cryosphere Interaction (modified from Wilson and Head [2002]). (c) Dike with exsolved volatiles (based on Liu and Wilson [1998]). (d) Dike with Plinian-style eruption (based on Scott and Wilson [2002]). (e) Collapsed magma chambers (modified from Mège et al. [2003]). (f ) Karst (modified from Spencer and Fanale [1990]). (g) Extension fractures (modified from Tanaka and Golombek [1989]). (h) Dilational faulting (modified from Ferrill et al. [2003]; after Ferrill and Morris [2003]). 4of20

5 the lava conduit drains, leaving an empty tube. Often these underground voids cannot support the overlying rock and collapse [Bardintzeff and McBirney, 2000]. Although lava within a lava tube can travel far from its source, a tube is typically confined to a single lava flow and is rarely, if ever, utilized by more than one flow [Bardintzeff and McBirney, 2000]. Lava tubes tend to be oriented in a downslope direction and are non-linear in plan view. Pit chains formed from collapsed lava tubes would be expected to be confined to basaltic rock layers, oriented down topographic slope, traceable to their volcanic source, and have non-linear surface traces Dikes [6] Radial and concentric pit chain formation in the Tharsis region has been interpreted as evidence for subsurface dikes [Mège and Masson, 1996, 1997; Liu and Wilson, 1998; Mège, 1999a, 1999b; Montesi, 1999; Mège et al., 2000; Scott et al., 2000, 2002; Gibbons et al., 2001; Wilson and Head, 2001, 2002; Scott and Wilson, 2002]. In this interpretation, a rising magma plume reaches a neutral buoyancy level and begins to propagate laterally and vertically as a dike. As the dike propagates upward, strain is concentrated at the dike tip. The strain produces fracturing in the host rock and subsidence above the dike in the form of a graben [Rubin, 1992; Wilson and Head, 2001, 2002]. Igneous dike injection is cited as a controlling mechanism in several explanations of pits and pit chains: (1) dikes interacting with groundwater or permafrost/cryosphere [Mège and Masson, 1996, 1997; Mège, 1999a; Montesi, 1999] (Figure 2b) may induce water or vapor release, producing void space into which overlying rock can collapse, or the dike may come into direct contact with groundwater producing a phreatomagmatic eruption; (2) others [Liu and Wilson, 1998; Scott and Wilson, 2002; Wilson and Head, 2001, 2002] have proposed that void spaces are formed at the tips of the dikes by concentrations of volatiles created by convection within the dike (Figure 2c); (3) a third scenario is used to describe large pits [Scott et al., 2000, 2002; Gibbons et al., 2001; Scott and Wilson, 2002] which form by collapse into void space created by large Plinian (i.e., explosive eruption of fragmented magma and magmatic gas characterized by large volumes of tephra and tall eruption columns) eruptions (Figure 2d). Although dikes may propagate for thousands of kilometers laterally [Ernst et al., 1995], they should be traceable to their source region. Evidence for the presence of dikes may include coeval volcanic formations, such as fissure eruptions, lava flows, pyroclastic fallout, or maar explosions (e.g., Craters of the Moon, Idaho [Kuntz et al., 1992]) Magma Chamber Collapse [7] Mège et al. [2002, 2003] proposed a mechanism in which dikes in a rift zone propagate upward from an elongate magma reservoir (Figure 2e). When the reservoir subsequently deflates, either by magma migration or eruption, surface rock collapses along the dikes forming pit chains. Substantiation for this interpretation would be similar to the coeval volcanic features for dike injection Karst Formation [8] Pit chain formation by chemical dissolution of soluble (e.g., carbonate) rock has been proposed by Spencer and Fanale [1990] (Figure 2f). Karst topography is common on Earth, characterized by sinkholes and solution valleys. Terrestrial examples of karst formation are often controlled by geologic structures such as faults and fractures [Palmer, 1991]. Groundwater flow along fracture zones and subsequent collapse along these tectonic features has been invoked to explain the structural associations of pit chains in the Valles Marineris [Spencer and Fanale, 1990]. Early Martian greenhouse conditions could have favored carbonate deposition in equatorial regions [Spencer and Fanale, 1990] and possibly in basins of the northern hemisphere and major impact basins [Morse and Marion, 1999]. Dissolution by an active hydrologic cycle, magmatic acids (CO 2, S, Cl), or high-silica-content carbonates produced during high heat flow [Spencer and Fanale, 1990] could produce the high porosity needed to create pits and troughs. Pit and trough distribution confined to regions of favored carbonate deposition, coupled with evidence of an active hydrologic cycle, would support this interpretation Extension Fractures [9] Collapse over extension fractures has been proposed as a mechanism for pit chain formation [Tanaka and Golombek, 1989; Banerdt et al., 1992; Tanaka, 1997] (Figure 2g). Horizontal extension can result in extension fractures in basement rock that propagate upward producing a graben at the surface. Further extension would cause collapse along the bounding faults into the void caused by the fissure at depth [Tanaka and Golombek, 1989]. Tanaka [1997] expanded on this interpretation, calling on fluidization and erosion through subsurface conduits and subsequent collapse to form the larger coalesced troughs. Evidence for this formation mechanism would be a strong correlation between pit chains and extensional features such as grabens and possibly evidence of an active hydrologic cycle Dilational Faulting [10] Dilational segments on normal faults can occur where the fault traverses mechanically strong stratigraphic layers or where tensile or hybrid mode failure occurs under conditions of low differential stress [Ferrill and Morris, 2003] (Figure 2h). We envision that lava flows on Mars form competent layers that are interlayered with less competent layers composed of materials such as volcanic ash and cinders, or eolian sediment. Deformation of such a heterolithic stratigraphic sequence in an extensional setting would produce refracted fault geometries and dilational fault segments in competent layers beneath unconsolidated surficial material. These dilational fault segments produce a void into which the overlying unconsolidated material drains, producing individual pit craters and pit chains. In the Martian gravity, these steep fault segments could extend to depths of 5km[Ferrill et al., 2003]. Dilation of these steep segments would provide a large volume in the upper crust to accommodate the formation of a subsurface cavity. Evidence for extensional tectonic regimes and a stratigraphy of varying mechanical strength would support this interpretation. 4. Methodology 4.1. Systematic Mapping [11] Previous interpretations of pit crater chains focused on individual chains, located mainly in the Alba Patera 5of20

6 region of Mars [Liu and Wilson, 1998; Montesi, 1999; Scott et al., 2000; Gibbons et al., 2001] or the Valles Marineris [Tanaka and Golombek, 1989; Spencer and Fanale, 1990; Banerdt et al., 1992; Tanaka, 1997]. We have performed systematic mapping of the western hemisphere of Mars to locate all visible pit crater chains, map their orientations, and determine associations with other structures, if any. Images used in mapping were obtained from the Mars Orbiter Camera (MOC) (Malin Space Science Systems, MOC Geodesy Image Mosaics, Wide Angle red images NASA/JPL, 2003 s Planetary Photojournal available at gov/) aboard the Mars Global Surveyor (MGS). These images are available in two resolutions, wide-angle and narrow angle images. We use MOC wide-angle images as the basis for global mapping to provide a consistent scale of resolution. These images have an average resolution of 232 meters per pixel and have nearly global coverage (the MOC wide-angle images do not cover the far south polar region, from 60 to 90 S). Images were assembled into a global mosaic by Malin Space Science Systems with a simple cylindrical projection (areographic). We imported them into ArcView TM GIS, and performed detailed viewing and mapping of pits, pit chains and troughs. The images were viewed at a consistent view scale of 2 2. Completed areas were tagged to keep track of progress. We used the following guidelines for determining a pit crater: a pit crater does not have a raised rim (to distinguish from impact craters); a chain must consist of at least three separate pits in alignment or two coalesced troughs in alignment; pits that occur within a collapsed trough are considered the same pit crater chain. Once identified, an ArcView TM georeferenced coverage was constructed. The coverage consists of two categories: (1) pit crater chains and troughs and (2) possible pit crater chains (used for small features near the limit of image resolution) Analysis of Pit Crater Morphology [12] Pit crater morphology was studied using different types of data. MOC narrow angle images provide resolution up to 1.5 meters per pixel, but have limited coverage of the planet. Thermal Emission Imaging System (THEMIS) visible spectrum images have an average resolution of 19 meters per pixel. These images were used to visually assess the pits for their detailed characteristics. The Mars Orbiter Laser Altimeter (MOLA) [Zuber et al., 1992; Smith et al., 1999] data comes in two varieties, both of which were used in this research. The Mission Experiment Gridded Data Record (MEGDR) is a digital terrain model at 1/128th degree pixels. The MEGDR data was used to determine the overall topography of pit crater chains. The Precision Experiment Data Record (PEDR) was used to acquire the most accurate depth measurements of pits by extracting orbital tracks that pass over selected pit craters. The PEDR was used to determine crater depth, wall slope and crater profiles Pit Volume Calculation [13] Conceptually, there are two fundamental controls on the maximum size a pit can attain (assuming a given angle of repose for the near surface material): (1) the thickness of the overlying draining material and (2) the amount of subsurface accommodation space. Volume calculations help constrain the volume of collapsed material and give some idea of the dimensions of the underlying void space. To simplify, we calculated volumes of pits that are circular in plan view and have not coalesced into other pits. We limited our study to five regions around Alba Patera and used a simple formula to calculate volume based on a circular right cone (Figure 3). Although many pits in the study appear to have conical shapes, others appear to have shallower floors. In these cases, the calculated slope values may not represent the slope of the entire pit. The apex of the cone may project below the actual floor of the pit, thus the volume estimate, assuming a circular right cone geometry, is likely to be an upper bound in many cases. This source of error could potentially be mitigated by a qualitative analysis of pit shape based on shadow geometry, however image resolution is not high enough at this time to apply this analysis to all the pit craters in the study. PEDR data was not utilized for pit volume calculations because of low spatial resolution and difficulty in co-registering MOC and MOLA data sets. [14] Pit diameter and shadow length were measured using ArcView TM GIS with images projected in Mercator with the equator set to image center. Wide-angle MOC images were used, with an average resolution of 232 meters per pixel. Utilizing the sun inclination and azimuth, we determined volumes based on the following approach (Figure 3): [15] We used sun inclination (a) in the data header file and measuring diameter (D) and shadow length (S) inline with sun azimuth, we calculate the true depth (H T ) and slope (q) of the pit. Length not in shadow; S N ¼ D S Apparent height; H A ¼ S* tan a Slope angle; q ¼ arctanðh A =S N Þ Radius of crater; r ¼ 1 2D True depth; H T ¼ r* tan q [16] Given the formula for volume of a circular right cone, we get the following: Crater volume; V ¼ 1 3pr 2 * H T 5. Results 5.1. Distribution of Pit Chains on Mars [17] A total of 1242 pit chains were mapped with an additional 479 possible pit chains. Pit crater chains occur throughout the western hemisphere of Mars, but most are clustered in discrete regions (Figure 4). Alba Patera is a broad shield volcano with approximately 3 4 km relief, located at the northern end of the Tharsis Uplift region (summit area centered at 40 N, 110 W; [Mouginis-Mark et al., 1992]). Pit crater chains in this region trend NE on the eastern side (Figure 5a) and NW on the western side, tangential to the volcano in both cases. South of Alba Patera, pit chains trend north-south following the 6of20

7 Figure 3. (a) Schematic illustration of topographic profile parallel to sun illumination azimuth, through center of conical pit crater (shown in Figure 3b) for determining pit crater volume from shadow lengths. (b) Pit crater is part of a chain in the Tantalus Fossae, approximately 38.2 N, W. Image credit NASA/JPL/Arizona State University (ASU), THEMIS image a. general orientation of the graben system in the region (Figure 4). [18] The Valles Marineris is a 4000 km long canyon system located 10 to 20 south of the equator in the western hemisphere [Lucchitta et al., 1992] and radial to the Tharsis uplift region. Pit crater chains in this area trend N105E parallel to the Valles Marineris, and the pits typically coalesce into scallop-sided troughs (Figure 5b). West of the Valles Marineris is the Noctis Labyrinthus, a maze-like network of intersecting valleys [Lucchitta et al., 1992]. Pit chains in this region have numerous orientations, as do the troughs and grabens in the region. Pit chains commonly coalesce into scalloped troughs which form complex crosscutting relationships of pit chains, scalloped troughs and grabens (Figure 5c). [19] Pit chains were mapped around the Tharsis volcanoes of Ascraeus Mons, Pavonis Mons, and Arsia Mons. Pit craters in this region form concentric rings around each volcano and a smaller number have formed radial to the volcanoes. Pit chains cross multiple lava flows and are typically not perpendicular to present-day topographic contours (Figure 5d). Several long pit chains (several thousand kilometers) were mapped radial to the Tharsis Bulge, southward to the extent of image clarity (60 S). [20] Smaller scale pit chains are also observed on Mars. Although not within the hemisphere mapped for this study, pit chains in the southwest region of Utopia Planitia are an order of magnitude smaller than those mapped in this study (50 meters diameter in contrast to 1000 meters for pits in the Tantalus Fossae; Figure 5e). Because these smaller pit chains are only visible at the resolution of the THEMIS and MOC narrow angle images, complete mapping of this region is not feasible at this time Variations in Pit Chain Morphology and Their Evolution [21] Pit chains in the western hemisphere show associations with extensional features such as grabens. There are 7of20

8 Figure 4. Western hemisphere of Mars showing distribution of pit chains. MOC wide-angle image in simple cylindrical projection, which may exaggerate curvature. Yellow indicates pit crater chains and troughs. Blue indicates possible pit crater chains (used when resolution did not allow for certainty). Red boxes indicate areas in Figure 5. Image credit NASA/JPL/MSSS. many examples of changes in morphology along strike from pit chains to scalloped troughs to grabens. Pits are often found along a bounding fault of a graben. MOLA profiles show that the pits typically form along the footwall of a graben-bounding fault, rather than the hanging wall. Pit chains typically follow the same general orientation as the grabens, such as in the Alba Patera and Valles Marineris regions. Pit chains sometimes occur in en echelon arrays and a few display curved tips similar to interaction at the tips of en echelon faults [Ferrill et al., 1999a] and openingmode fractures [Olsen and Pollard, 1991]. [22] Mapping provided us with a large sampling of pit chain morphologies; these variations in observed pit chain morphology were used to interpret a sequence of pit crater chain evolution (Figure 6). Early pit chain development begins with individual conical pits in alignment. As the pits grow, they begin to coalesce into elliptical shapes with the long axis parallel to the chain s orientation. These elongate pits may have multiple low points, separated by narrow ridges as a result of interaction between adjacent pits. Further development results in a trough with scalloped edges. The scalloped troughs further evolve into troughs with straight edges. Some pit crater chains exhibit pits within scalloped troughs, implying further collapse. Although these developmental stages cannot be used for absolute or relative dating, they provide a basis with which to compare proposed mechanisms of pit chain formation Pit Volume Calculations [23] We measured diameters and calculated volumes for pit craters in five regions around Alba Patera, including Tantalus Fossae, Tractus Catena, Phlegethon Catena, Cyane Catena and Alba Catena (Figures 7a 7e). A total of 259 measurements of 156 individual pits were performed as detailed in the methodology section (see Table 2). Pit diameters range from 0.93 km to 4.53 km, with a mean value of km (Figures 7a and 7b). Pit volumes vary from km 3 up to km 3, with an average volume of 1.41 km 3. Most wall slopes range from 30 to 65, indicating some cohesive strength in the uppermost part of the Martian crust (Figures 7c and 7d); wall slopes of 30 indicate that material is at the angle of repose for loosely consolidated material [Burkalow, 1945]. [24] Regional variations were noted (Figures 7b and 7d). In the Phlegethon Catena, pits appear shallow and flatfloored; slopes of these pit walls are from 9 to 15, less than the angle of repose for loosely consolidated material [Burkalow, 1945], and average depths in the region were 8of20

9 Figure 5. See Figure 4 for locations of individual images. (a) Alba Patera showing pit chains associated with extensional grabens. Center of image approximately N, W. Image credit NASA/JPL/ MSSS, MOC image M (b) Valles Marineris showing pit chains in east-west trend of the Valles Marineris. Arrows indicate transition stages in pit evolution. Note en echelon pit chains with overlapping sections creating relay ramps, similar to normal faulting. Center of image approximately S, W. Image credit NASA/JPL/MSSS, MOC image M (c) Noctis Labyrinthus. Center of image approximately 4.66 S, W. Image credit NASA/JPL/MSSS, MOC image M (d) Ascraeus Mons showing pits concentric to the summit, not oriented downslope. Center of image approximately N, W. Image credit NASA/JPL/MSSS, MOC image FHA (e) Western Utopia Planitia showing small-scale pit chain formation in polygonally fractured terrain. Center of image approximately 44.9 N, W. Image credit NASA/JPL/MSSS, MOC image E of20

10 differences, or young pit crater ages. The Alba Catena pits yielded smaller average volumes; pits in this region lie predominately within troughs. The small pit volumes could be the result of most subsidence in the region having been accommodated by trough development. In this conceptual model, a pit chain evolves into a welldeveloped trough and subsequent displacement produces new pits within the trough. Both Tantalus and Tractus have wide ranges of pit volumes, with slope values falling between 30 and 65, indicating mechanically strong material in the near surface. These slope values are consistent with the observed occurrence of apparently resistant layers within pit crater walls in high resolution imagery (Figure 1b). In a plot of sun incidence angle against slope value (Figure 7e), there appears to be a positive correlation between low sun angle (where only upper parts of pit walls are illuminated and used in slope calculations) and low slope values and high sun angle (where upper and lower pit walls are illuminated and used in slope calculations) and high slope values. This suggests a pit morphology with convex upward pit walls, contrary to the impression conveyed by visual imagery. Figure 6. Stages of pit chain formation illustrated using examples from eastern Alba Patera. (a) Early: Individual conical shaped pits (37.5 N, W; image credit NASA/ JPL/MSSS, MOC image M ), (b) Middle: Pits coalesce into elliptical shapes with the long axis parallel to chain orientation (35.3 N, W; image credit NASA/ JPL/MSSS, MOC image M ), and (c) Late: Scalloped troughs develop (34.2 N, W; image credit NASA/JPL/MSSS, MOC image M ). much smaller (mean value km) than the other regions measured. Two possible explanations are that the pits have been subjected to erosion and infilling by sediments or infilling by lava flows. Measured pits in the Cyane Catena have markedly steeper slopes (up to 76 ) and a large average depth (mean value km), perhaps reflecting either more competent rock layers in the Cyane region, less slope degradation due to climatic 6. Discussion of Proposed Mechanisms 6.1. Lava Tubes [25] Most pit chains do not have the characteristics of collapsed lava tubes. Collapsed lava tubes are typically sinuous in plan view, whereas most of the mapped pit chains are linear. Many pit chains cross multiple lava flows and/or have linear trends that are oblique to the present-day topographic slope. Although thermal erosion of lava tubes through pre-existing bedrock has been interpreted [Greeley, 1987], it is a rare phenomenon and difficult to reconcile with the visual observation of stratification in the pit crater walls. The lava tube mechanism may have arisen from a misinterpretation of Carr et al. s [1977] description of sharp-crested ridges with collapse pits along the sinuous crest; these collapse pit chains clearly have a different morphology from the pit crater chains as defined and mapped in this study (see Pit Crater Morphology section) Dikes [26] The most common hypotheses regarding pit chains invoke dike injection as the mechanism behind graben formation, and hence pit formation. These hypotheses rely heavily upon the work of Mastin and Pollard [1988] and Rubin [1992], however the mechanisms described by these authors produce morphologies that we have not found to be associated with pit chains (e.g., thrust faults above the dike) and do not produce pits [Mastin and Pollard, 1988] or pit chains [Rubin, 1992]. The results of Mastin and Pollard s [1988] and Rubin s [1992] analog experiments and numerical modeling within narrow parameters have been extrapolated to explain the wide range of pit and trough formation. However, the alignment of pit chains with extensional features (normal faults and extensional fissures) in most regions argues that dike injection is not needed to form these structures. Evidence for dike 10 of 20

11 Figure 7. Pit crater volume calculations. Note that one pit (Cy10) was excluded from charts because of its exceptionally large size. See Table 2 for all pit crater data. (a) Pit volumes range from km 3 up to 6.52 km 3. (b) Pit volume sorted by region. (c) Comparison of volume and diameter data for 18 pits where 3 4 measurements were made; note slope values exceed angle of repose (30 ) for most pits. (d) Volume and diameter data (259 data points) measured from 156 pits, grouped by region. (e) Sun incidence angle and calculated slope values show a positive correlation. emplacement is lacking. Surface eruptions along a laterally propagating dike would be expected to occur at some locations along strike [Wilson and Head, 2002], however, in the regions mapped we found no direct evidence of dike eruption at the surface (e.g., cinder cones, lava flows). Although some positive relief features have been identified as volcanic constructs [Mège et al., 2003], their origins are ambiguous. An alternate interpretation is that these features are ejecta from nearby meteorite impacts. Plinian eruption fallout deposits were not found directly associated with pit crater chains and the argument that high winds would provide extreme dispersal [Scott and Wilson, 2002] cannot account for all the pit chains mapped in this study. It is difficult to reconcile the large volumes of the pit craters in this study with the accumulation of gas at the tip of a dike [Scott and Wilson, 2002]. MOC narrow angle images were used to determine whether dikes could be seen in canyon walls along the Valles Marineris, specifically chasma walls that are oblique to the canyon orientation [Mège, 1999b]; no direct evidence was noted. Although dike-induced graben and pit chain formation could have occurred on Mars, without sufficient evidence of related volcanic features, the dike injection model remains unsubstantiated Magma Chamber Collapse [27] In order to account for the large volumes of linear depressions, Mège et al. [2003] invoke collapsed magma chambers as a mechanism. However, their work relies on analogs to volcanic caldera collapse [e.g., Roche et al., 2001] and extrapolation to long linear chains of pits. Morphological differences in Martian pit craters and terrestrial caldera collapse (e.g., terraced crater walls are not associated with pit crater chains) and the lack of coeval volcanic features argue against this model of formation Karst [28] Spencer and Fanale s [1990] karst analog is confined to large basins and equatorial regions where carbonate 11 of 20

12 Figure 7. (continued) deposition may have occurred (e.g., Valles Marineris). In contrast, our mapping demonstrates that pit chains are not confined to equatorial regions and large basins, and are present throughout the western hemisphere of Mars over a latitude range of 60 N to 60 S. Although morphologically similar to terrestrial examples of tectonically controlled karst sinkholes (e.g., Dead Sea basin [Abelson et al., 2003]), spectroscopic evidence for soluble rock layers 12 of 20

13 Table 2. Pit Volume Data Pit Number a MOC Reference Image # Location, b Lat, Lon Sun Angle, a Diameter of Pit (D), Shadow Length (S), Slope of Pit (q), Degrees Depth (H T ), Volume of Pit (V), Kilometers 3 Al10.1 M N, W Al11.1 M N, W Al12.1 M N, W Al13.1 M N, W Al14.1 M N, W Al15.1 M N, W Al16.1 M N, W Al17.1 M N, W Al18.1 M N, W Al19.1 M N, W Al2.1 M N, W Al20.1 M N, W Al21.1 M N, W Al22.1 M N, W Al22.2 M N, W Al23.1 M N, W Al23.2 M N, W Al24.1 M N, W Al24.2 M N, W Al25.1 M N, W Al25.2 M N, W Al26.1 M N, W Al26.2 M N, W Al27.1 M N, W Al27.2 M N, W Al28.1 M N, W Al29.1 M N, W Al3.1 M N, W Al30.1 M N, W Al31.1 M N, W Al32.1 M N, W Al32.2 M N, W Al33.1 M N, W Al33.2 M N, W Al34.1 M N, W Al34.2 M N, W Al35.1 M N, W Al36.1 M N, W Al37.1 M N, W Al38.1 M N, W Al39.1 M N, W Al4.1 M N, W Al40.1 M N, W Al41.1 M N, W Al42.1 M N, W Al42.2 M N, W Al43.1 M N, W Al44.1 M N, W Al45.1 M N, W Al46.1 M N, W Al47.1 M N, W Al48.1 M N, W Al5.1 M N, W Al6.1 M N, W Al7.1 M N, W Al8.1 M N, W Al9.1 M N, W Cy1.1 M N, W Cy1.2 M N, W Cy1.3 M N, W Cy1.4 M N, W Cy10.1 M N, W Cy11.1 M N, W Cy12.1 M N, W Cy13.1 M N, W Cy14.1 M N, W Cy15.1 M N, W Cy15.2 M N, W Cy16.1 M N, W Cy16.2 M N, W Cy17.1 M N, W of 20

14 Table 2. (continued) Pit Number a MOC Reference Image # Location, b Lat, Lon Sun Angle, a Diameter of Pit (D), Shadow Length (S), Slope of Pit (q), Degrees Depth (H T ), Volume of Pit (V), Kilometers 3 Cy17.2 M N, W Cy18.1 M N, W Cy18.2 M N, W Cy2.1 M N, W Cy2.2 M N, W Cy2.3 M N, W Cy2.4 M N, W Cy3.1 M N, W Cy3.2 M N, W Cy3.3 M N, W Cy3.4 M N, W Cy4.1 M N, W Cy4.2 M N, W Cy4.3 M N, W Cy5.1 M N, W Cy5.2 M N, W Cy6.1 M N, W Cy6.2 M N, W Cy6.3 M N, W Cy7.1 M N, W Cy7.2 M N, W Cy8.1 M N, W Cy8.2 M N, W Cy9.1 M N, W Ph1.1 M N, W Ph2.1 M N, W Ph3.1 M N, W Ph3.2 M N, W Ph4.1 M N, W Ta1.1 M N, W Ta10.1 M N, W Ta11.1 M N, W Ta12.1 M N, W Ta13.1 M N, W Ta14.1 M N, W Ta15.1 M N, W Ta16.1 M N, W Ta16.2 M N, W Ta17.1 M N, W Ta17.2 M N, W Ta18.1 M N, W Ta18.2 M N, W Ta19.1 M N, W Ta19.2 M N, W Ta19.3 M N, W Ta2.1 M N W Ta20.1 M N, W Ta20.2 M N, W Ta20.3 M N, W Ta21.1 M N, W Ta21.2 M N, W Ta21.3 M N, W Ta21.4 M N, W Ta22.1 M N, W Ta22.2 M N, W Ta23.1 M N, W Ta23.2 M N, W Ta24.1 M N, W Ta25.1 M N, W Ta26.1 M N, W Ta27.1 M N, W Ta28.1 M N, W Ta29.1 M N, W Ta3.1 M N, W Ta30.1 M N, W Ta31.1 M N, W Ta31.2 M N, W Ta32.1 M N, W Ta32.2 M N, W Ta33.1 M N, W Ta33.2 M N, W of 20

15 Table 2. (continued) Pit Number a MOC Reference Image # Location, b Lat, Lon Sun Angle, a Diameter of Pit (D), Shadow Length (S), Slope of Pit (q), Degrees Depth (H T ), Volume of Pit (V), Kilometers 3 Ta34.1 M N, W Ta34.2 M N, W Ta35.1 M N, W Ta35.2 M N, W Ta36.1 M N, W Ta36.2 M N, W Ta37.1 M N, W Ta37.2 M N, W Ta37.3 M N, W Ta38.1 M N, W Ta38.2 M N, W Ta38.3 M N, W Ta39.1 M N, W Ta39.2 M N, W Ta4.1 M N, W Ta40.1 M N, W Ta40.2 M N, W Ta41.1 M N, W Ta41.2 M N, W Ta42.1 M N, W Ta43.1 M N, W Ta43.2 M N, W Ta44.1 M N, W Ta45.1 M N, W Ta46.1 M N, W Ta46.2 M N, W Ta47.1 M N, W Ta47.2 M N, W Ta47.3 M N, W Ta48.1 M N, W Ta48.2 M N, W Ta48.3 M N, W Ta49.1 M N, W Ta49.2 M N, W Ta5.1 M N, W Ta50.1 M N, W Ta50.2 M N, W Ta50.3 M N, W Ta51.1 M N, W Ta51.2 M N, W Ta51.3 M N, W Ta51.4 M N, W Ta52.1 M N, W Ta52.2 M N, W Ta52.3 M N, W Ta52.4 M N, W Ta53.1 M N, W Ta53.2 M N, W Ta53.3 M N, W Ta53.4 M N, W Ta54.1 M N, W Ta54.2 M N, W Ta54.3 M N, W Ta55.1 M N, W Ta55.2 M N, W Ta55.3 M N, W Ta6.1 M N, W Ta7.1 M N, W Ta8.1 M N, W Ta9.1 M N, W Tr1.1 M N, W Tr1.2 M N, W Tr10.1 M N, W Tr10.2 M N, W Tr11.1 M N, W Tr11.2 M N, W Tr12.1 M N, W Tr12.2 M N, W Tr13.1 M N, W Tr13.2 M N, W Tr14.1 M N, W of 20

16 Table 2. (continued) Pit Number a MOC Reference Image # Location, b Lat, Lon Sun Angle, a Diameter of Pit (D), Shadow Length (S), Slope of Pit (q), Degrees Depth (H T ), Volume of Pit (V), Kilometers 3 Tr14.2 M N, W Tr15.1 M N, W Tr15.2 M N, W Tr16.1 M N, W Tr16.2 M N, W Tr17.1 M N, W Tr17.2 M N, W Tr18.1 M N, W Tr18.2 M N, W Tr19.1 M N, W Tr19.2 M N, W Tr2.1 M N, W Tr2.2 M N, W Tr20.1 M N, W Tr20.2 M N, W Tr21.1 M N, W Tr21.2 M N, W Tr22.1 M N, W Tr22.2 M N, W Tr23.1 M N, W Tr23.2 M N, W Tr24.1 M N, W Tr24.2 M N, W Tr25.1 M N, W Tr25.2 M N, W Tr26.1 M N, W Tr26.2 M N, W Tr27.1 M N, W Tr27.2 M N, W Tr28.1 M N, W Tr29.1 M N, W Tr3.1 M N, W Tr3.2 M N, W Tr30.1 M N, W Tr31.1 M N, W Tr4.1 M N, W Tr4.2 M N, W Tr5.1 M N, W Tr5.2 M N, W Tr6.1 M N, W Tr6.2 M N, W Tr7.1 M N, W Tr7.2 M N, W Tr8.1 M N, W Tr8.2 M N, W Tr9.1 M N, W Tr9.2 M N, W a Al, Alba Catena; Cy, Cyane Catena; Ph, Phlegethon Catena; Ta, Tantalus Fossae; Tr, Tractus Catena; decimal numbers indicate multiple measurements. b MOC images in Mercator projection with latitude of true scale at image center. Coordinates are planetographic, longitude is positive-west. IAU/IAG 2000 definitions [see Seidelmann et al., 2002]. (e.g., carbonates) in the quantities called for by their model has not been found to date [Bandfield, 2002] Extension Fractures [29] Extension fractures (fissures) may be a viable interpretation for some pit chains, including the small scale polygonal pit chain patterns in southwest Utopia Planitia. Most of the mapped pit crater chains were found in regions that show features of local or regional extension, such as normal faults and graben systems (Figure 4). Pit chains associated with a significant vertical surface offset (throw) are more readily explained by faulting mechanisms Dilational Faults [30] Pit chains are generally observed in regions of Mars that are interpreted to have experienced crustal extension, such as the numerous grabens in the Alba Patera [Tanaka, 1990] and Valles Marineris [Wilkins and Schultz, 2003] regions. Pit chains also exhibit many features of normal faulting, such as en echelon arrays and curved tips. The strong correlation between grabens and pit chains (Figure 8) suggests a common tectonic mechanism for both features. Terrestrial examples of graben formation in regions of extension demonstrate that volcanic mechanisms are unnecessary [McGill and Stromquist, 1979; McClay, 1990; Dula, 1991; Trudgill and Cartwright, 1994; Diegel et al., 1995; Rouby et al., 1996; Rowan et al., 1999; Ferrill et al., 2000]. The regional stress field invoked for dike emplacement is also consistent with extensional faulting [Anderson, 1951]. The variations along strike from grabens to pit chains to trough formation are more readily explained by structural controls such as (1) changes in mechanical stratigraphy, 16 of 20

17 resulting in changes in near surface fault dip, (2) small variations in strike or dip that change the slip tendency [Morris et al., 1996] or dilation tendency [Ferrill et al., 1999b] along the fault, or (3) displacement variation along strike. Normal fault segments commonly occur in en echelon arrays in extensional settings [e.g., Ferrill et al., 1999a, and references therein], and as displacement increases, fault segments propagate laterally along curved traces to intersect with the next en echelon segment. This pattern is seen both in normal faults and in pit chains on Mars. As fault linkage continues, it produces a corrugated fault trace, with alongstrike variation in fault orientation. These varying orientations can produce different magnitudes and directions of slip along the fault [Morris et al., 1996]. Variation in the volumes of pit craters along strike may be either a function of this displacement variation along the fault or the depth of the dilational segment (Figure 9). The heterolithic stratigraphy observed in the Valles Marineris [McEwen et al., 1999] could provide the differential mechanical strength necessary to produce refracted fault geometries, including dilational faulting. 7. Non-Martian Examples [31] Long linear chains of collapse pits are rare on Earth, but have been documented in Hawaii, Israel/Jordan and Iceland. Okubo and Martel [1998] identified pit chains along Kilauea s east and southwest rift zone and concluded that collapse along a typical Hawaiian dike could not account for the large aperture of the pit craters. Large opening-mode fractures were likely caused by magma pressure and tensile stresses associated with the seaward sliding of Kilauea s south flank [Okubo and Martel, 1998]. Although magma flow through a pre-existing fracture might enlarge pit craters, it was not required to create them; regional extension alone produced open fractures [Okubo and Martel, 1998]. Figure 8. MOC narrow angle image showing pit craters superimposed on normal faults within graben in the Tractus Fossae region. Center of image approximately 24.1 N, W. Image credit NASA/JPL/MSSS, MOC image R Figure 9. Schematic drawing illustrating that pit volume is a function of (a) the depth of the dilational fault segment, (b) increased displacement along the fault or (c) the thickness of the overlying unconsolidated material. 17 of 20