ICARUS 71, 268-286 (1987)

Water or Ice in the Regolith?: Clues from Rampart Craters Seen at High Resolution r PETER J. MOUGINIS-MARK

Hawaii Institute of Geophysics, University of Hawaii, Honolulu , Hawaii 96822

Received October 23, 1986 ; revised March 3, 1987

Very high resolution Viking Orbiter images (8-17 m per pixel) have been used to investigate the morphology of Martian ejecta blankets and the crater interiors, with the objective of identifying the fluidizing medium for the ejecta and the physical properties of the target rock. The occurrence of well-preserved, small-scale pressure ridges and scour marks, evidence for subsidence around isolated buried blocks in partially eroded ejecta lobes, and the stability of crater walls and distal ramparts argue for ground ice being the dominant state for volatiles within the target rocks at the time of impact. Rare examples of channels (190-650 m wide) on the surfaces of ejecta blankets, and on the inner walls of the crater , indicate that in some instances liquid water was incorporated into the ejecta during its emplacement. No morphological evidence has been found to discount the idea that atmospheric effects were partially responsible for ejecta fluidization, but it is clear that these effects were not the sole reason for the characteristic lobate deposits surrounding at least some rampart craters on . © 1987 Academic Press, Inc.

INTRODUCTION laboratory experiments (Schultz and Gault 1981, 1984) have shown that rampart­ From the return of the first Viking bordered ejecta facies and ejecta flow lobes Orbiter images of Martian impact craters it can develop without the presence of water. was evident that the morphology of the Laboratory studies have also demonstrated ejecta blankets surrounding these craters that targets of varying viscosities can be was different from that of their lunar and used to investigate the fluidization of Mar­ Mercurian counterparts insofar as much of tian rampart craters (Gault and the ejecta appeared to have been emplaced 1978). Decreasing the viscosity of mud tar­ by surface flow (Carr et al. 1977). Over the gets promoted postdepositional flow of the last 10 years, considerable debate has ejecta and increased the radial extent of the focused on the possible cause(s) for this "continuous" ejecta deposit. In several ejecta fluidization and flow. Possible forms instances, these laboratory craters were that the fluidizing medium may have taken also seen to possess multiple flow units include water ice or liquid water within the similar in kind to the multilobed craters on target material (Carr et al. 1977, Boyce Mars (Carr et al. 1977). 1979, Johansen 1979) or atmospheric gases A groundwater system has been pro­ interacting with suitable particle sizes posed for Mars on a global scale (Carr 1979, within the ejecta curtain (Carr et al. 1977, Clifford 1986), and it is believed likely that Schultz and Gault 1979, Schultz 1986). such a system would influence the Indeed, Schultz (1986) draws attention to occurrence and physical state of impact the fact that the fluidized ejecta facies only crater ejecta. A major difficulty in correlat­ indicate fluid-like emplacement, and ing these inferences about ground volatile 268 0019-1035/87$3.00 Copyright © 1987 by Academic Press, In c. All rights of reproduction in any form reserved. MARTIAN CRATERS 269 content with the observed properties of cation (or lack thereof) of meltwater chan­ crater ejecta blankets lies in knowing the nels and small-scale topography on the extent to which the groundwater system surface and edges of the ejecta blankets and (located within the deep megaregolith) was the degree of stability of the ejecta lobes filled. Many of the large and parent crater walls (Mouginis-Mark seen on Mars may have formed through the 1986). Diagnostic features are likely to be eruption of groundwater under pressure physically small in size (probably less than from the megaregolith beneath the per­ a kilometer in extent) and so have not mafrost, which is estimated to have been previously been considered in either global about 1 km thick (Carr 1979). Carr (1986) studies of rampart craters (e.g., Johansen concluded that the megaregolith below this 1979, Horner and Greeley 1986) or permafrost layer, planetwide, might con­ cratering models (e.g., Schultz and Gault tain no less than the equivalent of a water 1984, Schultz 1986). This paper therefore layer 350 m thick. Clifford's (1986) cal­ describes morphologic features seen in culations of outgassed H 20 on Mars would very high resolution (8 to 17 m per pixel) suggest that if more than a few percent of Viking Orbiter images of crater ejecta blan­ the quantity of water required to saturate kets and interiors in order to constrain the the pore volume of the cryosphere were most likely target properties at the ti me of present, then a subpermafrost water system crater formation. of substantial proportions would result. After saturation of the pore volume of the PREVIOUS GLOBAL OBSERVATIONS cryosphere, the equivalent of an additional Johansen (1979) and Kargel (1986) have 100 m of H 20 would be sufficient to create suggested that the presence or absence of a an aquifer nearly 4.3 km deep. That this distal ridge on an ejecta flow may be an small volume of water is all that is required indicator of water (ridge present) or ice to produce the aquifer is due to the pre­ (ridge absent) within the target material. dicted low pore volume at depth (Clifford Although there is a predominance of ridged 1986). It is further predicted that much of craters at low and ridgeless craters the water originally at depths of less than 1 at high latitudes (Kargel 1986), no corrobo­ km would remain in situ for most of Mar­ rating models for ejecta emplacement, or tian history (Fanale et al. 1986), thus con­ theoretical or experimental data, were stituting a possible medium for fluidizing presented by these investigators to support ejecta. their ideas on ejecta fluidization. On the Because of the importance that the basis of finite-element studies of the stress presence of liquid water, liquid brines, or magnitudes, distributions, and directions in ground ice close to the surface would have hypothetical Martian ejecta blankets, for global models of volatile abundance and Woronow (1981) concluded that rampart physical state (e.g., Fanale and Jakosky craters had water contents between 16 and 1982, Fanale et al. 1986), more precise 72 volume percent soon after emplacement. observational information is needed to help However, Woronow hypothesized that the constrain these theoretical models of fluidized ejecta deposits were formed by volatile distribution and state. One the mechanical failure and subsequent approach to identifying both the former radial flow of lunarlike ejecta deposits; such presence (or absence) and physical state of a mechanism is not believed to be ap­ subsurface volatiles in the Martian past propriate for describing the ejecta comes from their inferred influence in con­ emplacement process for these Martian trolling the morphology of rampart craters craters (Gault and Greeley 1978, Schultz and crater ejecta blankets. Key testable and Singer 1980, Mouginis-Mark 1981, morphologic indicators are the identifi- Wohletz and Sheridan 1983). 270 PETER J. MOUGINIS-MARK

At issue is the relative importance of of atmospheric and target properties. atmospheric effects and target volatiles in However, Wohletz and Sheridan (1983) controlling the morphology of ejecta sur­ noted that since the formation of distal rounding Martian impact craters. Schultz rampart ridges around Martian craters is and Gault (1979) demonstrated quantita­ affected by local preexisting topographic tively that, during the emplacement of an obstacles, the formation of the ramparts ejecta cloud, ejecta deposition would be probably resulted from the deposition of controlled by the particle size distribution ejecta as the yield strength of the ejecta of the clasts and that, as a result of the size flow increased above a critical shear stress distribution of particles, a multiphase due to interparticle friction. This suggests emplacement sequence would probably that the morphology of the ejecta blanket result. Subsequent laboratory experiments was strongly related to the initial yield by Schultz and Gault (1981, 1984) have strength of the flow and, hence, the degree shown that under conditions scaled to of ejecta fluidization. simulate the Martian environment, craters A method for assessing the "average" with distal ridges can be produced without rheology, or degree of fluidization, of the any volatiles being present in the target. ejecta flows for crater populations found on These experiments suggest that craters uniform target materials has been the exam­ with contiguous ramparts may represent ination of the mechanical strength of the the products of relatively low-velocity ejecta lobes, based on their topographic impacts into target lithologies pqssessing relief and areal extent. Wohletz and Sheri­ lower volatile contents than the targets dan (1983) observed that the ejecta deposits associated with craters possessing highly appear to be composed of unconsolidated, fluidized ejecta facies. Increased volatile fine-grained particles that can be trans­ contents are predicted to decrease ejecta ported by Martian winds, and that the size by comminution and by aerodynamic thickness of the ejecta material is to­ breakup, thereby increasing the lateral pographically controlled. Ejecta lobe thick­ extent of the rampart. Ifthe impact velocity ness appears to be limited by some physical or volatile content were sufficiently high, aspect of the deposit, inferred to be the the energy trapped in the ejecta cloud maximum shear strength of the fluidized would result in long run-out flows or radial medium (Mutch and Woronow 1980). scouring. Thus, Schultz and Gault (1981 , Based on observations of ejecta flow 1984) concluded that volatiles within the around low topographic obstacles (Carr et Martian regolith were sufficient, but not al. 1977, Wohletz and Sheridan 1983), most essential, to produce the variety of lobate ejecta deposits are believed to be observed ejecta morphologies. Even under less than a few tens of meters thick, the current Martian climate and with rea­ although a few shadow length mea­ sonable ejecta size ranges, a variety of surements for the inner ejecta lobe of the emplacement styles may occur. The tran­ crater Arandas showed that some inner sition from lunar-type to rampart to mul­ ejecta deposits can be as much as 500 m tilobed to radial ejecta facies is believed to thick, and outer ejecta lobes can have ram­ represent part of a continuum which parts more than 50 m high (Mouginis-Mark depends at least in part on ejecta size. and Carey 1980). No global study has been While Schultz (1986) agrees that the made of this variation in the upper limit for atmospheric effects do not preclude a sup­ ejecta thickness, but for craters larger than porting (possibly even dominant) role for about 6 km in diameter, maximum ejecta target volatiles in the ejecta fluidization thickness appears to be uniform regardless process, no independent evidence has been of the crater size (W oronow 1981, presented to assess the competing roles Mouginis-Mark and Cloutis 1983) . This MARTIAN CRATERS 271 observation may imply that the rheology latitudes show evidence for quasi-viscous (and thus degree of ejecta fluidization) relaxation of the rim crests, and this remained uniform for cratering events over relaxation was attributed to creep defor­ a wide geographic area and an extended mation of near-surface materials that were period of Martian history. In turn, this ice-rich. would suggest that no appreciable change in volatile state or concentration existed at NEW OBSERVATIONS the time of crater formation for ejecta origi­ The Viking Orbiter image data set con­ nating at shallow depths (small crater tains approximately 400 frames at a spatial diameters) and greater depths (large crater resolution of better than 10m per pixel and diameters). However, these studies of 2200 frames at a resolution of better than 20 ejecta extent have not provided any unique m per pixel, for which the atmosphere was diagnostic features of the deposits which either clear or only slightly obscured (J. subsequently allow the identification of the Zimbelman, pers. comm. 1986). A search of fluidizing medium. all these images has revealed several exam­ With respect to the interior morphologies ples of both interior and exterior features of of Martian craters, early analysis (Wood et impact craters that bear on the nature of the al. 1978) drew attention to the numerous ejecta fluidizing medium. craters that possess central peaks with summit pits; however, these pits are not Crater Ejecta Blankets associated with all fresh rampart craters of While little positive information on the a given diameter. Wood et al. (1978) fluidizing medium can be gained from the hypothesized that the pits were formed by regional trends in the distribution of crater explosive decompression of strata con­ morphologies described above, previous taining subsurface volatiles. They sug­ studies of high-resolution images have gested that in addition to the volatiles that shown that certain small morphologic fea­ created the lobate ejecta deposits, volatiles tures may be used to distinguish between of another phase (possibly liquid water as water and ice existing within the target opposed to water ice) might have been material. For example, grooves on the present. Circumstantial evidence from the ejecta lobes of the crater Arandas analysis of crater morphometry (­ (Mouginis-Mark 1981) indicate that imme­ Erlich 1986) and volcanic landforms diately after lobe emplacement, but prior to (Mouginis-Mark 1985) in the Elysium cessation of ejecta curtain deposition, the Planitia region may support this interpre­ ejecta lobes had established sufficient phys­ tation, since here there is an unusually large ical strength to preserve these " scour number of craters that possess pitted cen­ marks" in their surface, which were tral peaks. This relationship may indicate created during the passage of the later that the liquid water responsible for the ejecta. A similar effect is also observed large Elysium outflow channels and where preexisting obstacles have created landforms associated with volcano/ground pressure ridges within the ejecta lobe on the ice interactions existed sufficiently close to craterward side of the obstacle. the surface to also affect the formation of Several examples of rampart crater impact crater ejecta blankets. ejecta lobes that were imaged at 9-17 m per Further corroborating evidence that ice, pixel are shown in Fig. 1. All of the craters or water, was included within the interiors are in the diameter range 16-26 km and are of impact craters lies in the " terrain soft­ located in the Southern Hemisphere ening" of landforms poleward of equatorward of 25°S. Prominent in these 30° north or (Squyres and Carr 1986). images is lack of modification of the distal It was observed that craters in these ramparts (Figs. 1A and IB) and the radial 272 PETER J. MOUGINIS-MARK ~ >:;o :;>-1 z n :;o > >-1 tTl :;o \/)

FIG. I. Examples of crater ejecta blankets seen at very high resolution. (A) Mosaic showing prominent rampart lobes on the ejecta blanket of a 26-km-diameter crater located at 9.47°S, 252.84°W. Maximum ejecta range from this crater is approximately 3 crater radii from the crater rim. Viking Orbiter frames 756A5 1-56, 17m/pixel resolution. Mosaic width equivalent to 47.2 km. (B) Ejecta lobes 1-1.5 crater radii from the rim of a 16-km-diameter crater located at 23.41°S, 240.31 °W (the same crater shown in Fig. 5C). Distal rampart is arrowed. Viking Orbiter frame 795A II , 9 m/pixel. Image height is equivalent to 12.4 km. (C) Near-rim continuous ejecta deposits of a crater 18 km in diameter located at 25.68°S, 162.33°W. Viking Orbiter frame 803A03, 9 m/pixel. Image height equivalent to 11.7 km. Illumination direction in all images is from the left.

N -....) w 274 PETER J. MOUGINIS-MARK scouring close to the crater rim (Fig. 1C). In not observed (Fig. 2B), implying that this particular, despite it being possible to iden­ water-charged ejecta flow mechanism did tify features as small as approximately not operate for the Schiaparelli craters. For 20-45 m, no postemplacement slumping of the "kettle hole model," the size of the the ejecta lobes can be seen, nor are there depressions within the ejecta blankets any superposed channels on these deposits. argues for solid material (rather than a Thus the relationship between a liquid liquid) being present (and buried) within the water fluidizing medium and the formation ejecta at the time of deposition and the of distal ejecta ramparts (Johansen 1979, subsequent postemplacement removal of Kargel 1986) does not appear to apply for this material to produce the collapse pits. these craters. The presence of such blocks transported The eroded ejecta blankets of two craters within the ejecta flow would strongly argue within the basin Schiaparelli (Fig. 2) are in favor of ice (rather than water) existing useful for examining the internal structure within the target material at the time of the of the ejecta lobes. Schiaparelli Basin is . partially infilled with eolian material The actual size of the ice blocks that (Mouginis-Mark et al. 1981) , and these two could be created and transported to the younger craters have been excavated in this appropriate radial distances (2-3 crater infill material. Subsequent subaerial ero­ radii, or 20-30 km from the rim crest) is sion has removed the topmost ejecta layers unknown for Martian cratering events. By to reveal some of the internal structures of analogy with the distribution of blocks the ejecta lobes. It is evident that there is a around lunar and terrestrial impact craters, much larger than normal number of irregu­ it is apparent that either the ice blocks on larly shaped depressions on both of these Mars were not derived from the parent ejecta blankets. These pits have long crater cavity or the comminution efficiency dimensions of up to 2 km and can be either of permafrost targets is different than that nearly circular or highly elongate. Irrespec­ for lunar and terrestrial craters. For exam­ tive of their shape, none of these pits ple, blocks around the 98-km-diameter appear to possess raised rims at this image lunar crater Tycho are less than 200 m in resolution. Thus it is believed that these diameter (Moore 1972), and therefore even numerous pits are unlikely to be primary in this case the ejecta blocks are much impact craters. smaller than the diameter of the Martian Two alternative interpretations for the collapse pits. Lacking the appropriate origin of these pits are proposed here. theoretical or observational data base to Either they can be secondary craters which distinguish the most likely mechanism, a were formed early in the cratering event method wherein the ice blocks were de­ and subsequently had their rim materials rived by plucking of the crater rim material removed by the ejecta flow, or the pits may during the radial component of the ejecta be collapse features akin to kettle holes flow appears more likely than changing the found in terrestrial periglacial environ­ comminution properties of ice-rich targets. ments. If the secondary crater model is Such erosion has been identified at the Ries valid , this would imply that the ejecta flow Crater in West Germany (Horz et al. 1983) , had considerable erosional potential, there­ where several tens of meters may have by removing the rim material from the been eroded from the local bedrock during secondary craters. Rapidly moving water­ the emplacement of the Bunte Breccia. rich ejecta flows are expected to have dis­ Locally derived "megaclasts" larger than tributary features (braided outwash chan­ 10 m in diameter occur within the con­ nels , sediment fans) located at the distal tinuous ejecta deposit of Ries, indicating ends of the ejecta lobes; such fe atures are that plucking of the preexisting terrain FIG. 2. Examples of eroded ejecta blankets for two craters within the basin Schiaparelli crs, 344°W) . These ejecta deposits have been eroded to a sufficient level to display some of the internal structure of the ejecta lobes, wh ile preserving sufficient form to identify flow lobes and ramparts. Of particular interest is the large number of irregularly shaped pits within the lobes that are suggestive of collapse of material within the ejecta after its emplacement, which is here hypothesized to be ice, in a manner similar to the formation of terrestrial kettle holes (see text for discussion). Note also that the eroded lobes still preserve some of the lineations also seen on fresh craters (Mouginis-Mark 1981) and that the ramparts erode into a series of isolated massifs, suggesting coherent blocks existed within the ejecta flows . Example (A) is a mosaic of Viking frames 748A0 1-09. The crater is approximately 16 km in diameter. Example (B) is frame 748A l4. Direction of ejecta flow was from the bottom left of this view. Width of image is equivalent to 18.5 km. In each case, image resolution is 17m/pixel. 275 276 PETER J. MOUGINIS-MARK occurred during the ejecta emplacement parent crater rim indicates that the radial process. However, in the case of the Mar­ velocity of the flow was sufficient to trans­ tian craters, the physical characteristics of port megaclasts to considerable distances, this plucking mechanism are poorly known and so is consistent with the idea that (although the process has been suggested to massive blocks of ground ice may have occur; Gault and Greeley 1978). Block size been initially incorporated within the ejecta may still be insufficient for producing the blanket. kettle holes by subsidence around an indi­ While most ejecta lobes evidently vidual ice-rich block, but it is also possible remained stable following their initial that clusters of smaller ice-rich blocks, emplacement, examples of postdeposition rather than single large pieces, could also fluid flow have been identified on two crater account for the large amount of localized ejecta blankets. In the first case, several subsidence. Due to the uncertainties channels not unlike the channel networks inherent in the observational data, the found in the Martian heavily cratered Sou­ quantitative description of the plucking thern Hemisphere have been identified on process is left here as a subject for future the ejecta blanket of a 9-km-diameter crater investigation. in Sinus Sabaeus (Fig. 3). These channels Further study of the Schiaparelli craters are approximately 190-340 m wide, can be also reveals two additional morphologic as much as 3.5 km in length, and extend up characteristics pertinent to the ejecta to 25 km from the rim of the parent crater. emplacement. First, the lobes shown in Individual examples may either branch or Fig. 2B evidently have retained many of the coalesce, depending presumably on small radial striations that appear to accompany variations in local topography. This rare lobe emplacement (Schultz and Singer occurrence of channels on the surface of a 1980, Mouginis-Mark 1981). With the small crater's ejecta blanket is hypothe­ inferred degree of erosion (evidently sized to be the consequence of the impact several meters, at least), these striations event occurring on the ejecta blanket of the can therefore be assumed to be part of the 150-km-diameter crater , the rim entire section through the lobe, rather than of which is located only 60 km from this simply being a surface phenomenon. This small crater. Carr (1983) and Brakenridge et in turn would argue in favor of a non­ al. (1985) have considered the possible turbulent emplacement process which mechanism(s) by which the morpholog­ would more likely be associated with ically similar Martian networks volatile lubrication than atmospheric could have been carved by low-volume effects, based on analogy with poorly fluid­ surface flow of water. Melting of ice pre­ ized terrestrial pyroclastic flows (Wilson cipitated onto the surface following 1980) and the cratering models developed injection of water into the atmosphere by by Schultz and Gauilt (1979). Second, the large impacts, and the subsequent incorpor­ distal ramparts of these craters are in the ation of this ice within the ejecta blanket, is process of being removed by eolian ero­ one possible source of the water. Alterna­ sion. The occurrence of isolated massifs tively, such channels may have been pro­ beyond the currently retreating edge of the duced by the interaction of possibly hot ejecta lobe suggests that there were large ejecta blankets and ground ice. In the par­ (300-500 m diameter) blocks entrained ticular case of the crater illustrated in Fig. within the ejecta that subsequently formed 3, we hypothesize that ice (or water) was the distal rampart. While it is not known brought sufficiently close to the surface by whether blocks of this size could be trans­ the formation of Bakhuysen that the ported more easily by atmospheric or ice second, smaller cratering event could fluidization, their occurrence far from the mobilize liquid water on the surface of its MARTIAN CRATERS 277

FIG. 3. The sole example found in this study of surface " drai nage" on top of the ejecta bl anket of a small (less than 50 km diameter) crater is seen for this 9-km-diameter crater at 25. 03°S, 345 .67°W. This crater is located on the ejecta blanket of a larger crater (Bakhuysen, 150 km diameter) that lies 60 km to the northeast of this scene (at the top of thi s image). This double impact event is hypothesized to be the cause of liquid water existing sufficiently close to the to be released from the ejecta blanket of the smaller crater. (A) Channels on ejecta lobes (Viking Orbiter frames 936A03-06, 15 m/pixel) . (B) Sketch map of same area as (A) , showing di stribution of channels (arrows indicate direction of flow) on the ejecta blanket. The primary crater is at the bottom left in thi s view, and other primary craters are denoted by inward-facing barbs. The limit of a partially buried/eroded outer ejecta lobe is shown by the dot-dashed line, while preimpact topographic highs are shaded. 278 PETER J . MOUGINIS-MARK

0 5 km

FIG . 3-Continued . ejecta blanket. In this manner, the small little of the original ejecta surface is still volumes of water implied by the dimen­ preserved. Due to their amphitheater head­ sions of the channels, and the inferred walls, undissected nature of the dividing atmospheric conditions that would have plateaus, and lack of slumping of these existed at this relatively recent time in upland remnants of the old ejecta surface, Martian history, need not be mutually these channels appear to be comparable to inconsistent. other landforms usually associated with On a larger scale, channeling on the sapping channels on Mars (Baker 1982) and ejecta blanket of the 400 x 460-km­ the Earth (Laity and Malin 1985 , Schumm diameter basin Schiaparelli (Fig. 4) was and Phillips 1986). previously recognized by Mouginis-Mark et a!. (1981) as a likely product of post­ Crater Interiors emplacement sapping of ground ice In a recent study, Squyres and Carr entrained within the ejecta. After their (1986) utilized the degree of relaxation, or initial formation, these channels have sub­ postformational flow , of numerous sequently been enlarged to the extent that landforms as an indicator of ground ice MARTIAN CRATERS 279

FIG. 4. Additional evidence for the release of meltwater around a few impact craters on Mars is seen in this view of the ejecta blanket of the basin Schiaparelli (6.64°S, 344.95°W). Despite the fact that the channels have evolved to the extent that intervalley divides have almost been removed, the existence of amphitheater headwall s and remnants of the original plateau-like ejecta surface suggest meltwater release by sapping from the emplaced ejecta bl anket. Direction of flow was from the top right in this view. Illumination is from the lower right. Viking Orbiter image 747A49, 15m/pixel. Image height equivalent to 11.9 km.

within the regolith. In the case of impact that craters equatorward of 30° contain craters, they found that, equatorward of little ground ice, all of the fresh craters that 30°, the crater morphology is crisp, with no Squyres and Carr (1986) presented as evidence of subsequent flow of the crater examples of this latitudinal variation in rim crests (which are sharp and have slopes ground ice possessed the characteristic that are generally concave upward in pro­ lobate ejecta deposits, and therefore file). Between latitudes 30° and 55°, crater experienced some degree of ejecta fluid­ rim crests are rounded, and slopes are ization. commonly convex upward. Poleward of 55° In this study, examples of the "ice-poor" the '','' as this subdued craters equatorward of 30° have been stud­ morphology is called, assumes a slightly ied at a resolution of 8-15 m per pixel to different character in that crater rims are further characterize the interior mor­ fairly sharp and more like those seen at low phology of rampart craters. Five craters are latitudes. Squyres and Carr (1986) attrib­ illustrated in Fig. 5. These craters range in uted this latitudinal variation in softening size from 4.5 to 15 km in diameter, and all to variations in rheology of the rim possess well-preserved lobate ejecta materials due to the percentage of ice-rich deposits. Evident from Fig. 5 are the sharp material contained within the rim deposits. rim crests observed by Squyres and Carr Although the implication of their work is (1986) and, with the exception of partial 280 PETER J. MOUGINIS-MARK

FIG. 5. Examples of crater interiors seen at very high resolution. (A) Pair of craters 4.5 and 8.0 km in diameter, located at 1.60°S, 198 .32°W (Viking Orbiter frame 725A29, 15m/pixel). (B) 6.6-km-diameter crater located at 22.66°$, 241.73°W (Viking Orbiter frame 795AOI , II m/pixel). (C) 12 .6-km-diameter crater located at 23.46°S, 240.58°W (Viking Orbiter frame 795AIO, 9 m/pixel). Height of image equivalent to 12.5 km. (D) 15.0-km-diameter crater located at 26.01 °$, 180.86°W (Viking Orbiter frame 801A74, 8 m/pixel). Image height equivalent to 10.5 km. lllumination direction in all images is from the left.

eolian infill, the almost lunarlike interior scalloping or subsidence on their walls morphology of these craters. The smallest (consistent with the conclusions of early craters (Figs. SA and 5B) have a simple studies of low-resolution Viking images; bowl-shaped form, continuous rims, and no Wood et al. 1978). For the larger craters MARTIAN CRATERS 281

(Figs. 5C and 5D), scallops and terraces subsequent flow and erosion of the channel­ have formed, but it is noticeable that even forming material (also unlike lunar melt for those craters where fluidized ejecta was deposits; Hawke and Head 1977), and the created (inferred from the occurrence of the inferred destruction of Martian impact lobate ejecta lobes), no unusual flows or a melts during the cratering process (Kieffer propensity for landslides and subsidence and Simonds 1980) all argue against the are seen on the inner crater walls. This Jack impact melt origin of these channels on of " unusual" interior features is believed to Cerulli. Thus the question remains as to be strong circumstantial evidence against why the inferred water discharge occurred the general occurrence of liquid water within Cerulli and not other Martian craters within the target rocks (and derived ejecta) of a similar size and latitude. A plausible at the time of crater formation. At a explanation would be that the water aquifer resolution of 8 m, it is expected that outflow was unusually close to the surface at the channels and slump features (due to the time of crater formation, as is also sug­ percolation of water from the rim material) gested by the unusually large amount of would be observable if they existed. slumping of the inner wall terraces and by Study of lower-resolution (42 m per the proximity of the source region for pixel) images of the crater Cerulli (32°N , Mamers Valles, which is located less than 338°W) does, however, show ample 50 km west of the rim of Cerulli. evidence of postcrater formation modifi­ In several instances, these channels cation by surface flow in the form of originate high on the inner walls of Cerulli numerous channels around the rim crest of and descend to the crater floor in a series of this 120-km-diameter crater. Figure 6 steps over the wall terraces. Evidence for shows both an example of these channels ponding of the flows can be found behind and a geomorphic sketch map illustrating these terraces, where flat, morphologically their distribution. Note that these channels bland areas can be seen in many local occur both on the inner wall and on the catchment basins prior to overflowing of exterior continuous ejecta deposits and the local obstacle and the formation of extend to a radial distance of 1.5 crater radii another channel on the downslope side of from the rim crest. Within Cerulli the chan­ the terrace. Only very subdued topography nels are generally 5-20 km in length, have and textural variations, possibly indicating widths of 150-650 m, are sinuous in plan, a deltaic plain, can be seen where these and appear similar to individual segments channels reach the crater floor. Because the of the channel networks identified in the channels appear to be deeply cut into the Martian heavily cratered terrain (cf. Baker wall material of Cerulli, this apparent lack 1982). The preferred interpretation for of erosional debris would imply that channel origin is that they are fluvial chan­ although the valleys were easily formed, nels carved by water percolating out of the the volume of material transported was emplaced wall material of Cerulli at an either quite small or the transported unspecified (but possibly quite short) time material was widely distributed over the after the cratering event. However, it crater floor. should be remembered that superficially While it is very unusual for such channels similar flows and channels have been seen and ponded material to be associated with on certain lunar impact craters, where an fresh impact craters on Mars, this type of impact melt origin was described (Howard flow and ponding has been seen elsewhere and Wil shire 1975). Lack of the crenulated on the planet and was also attributed to surface texture associated with lunar surface flow of small volumes of water. impact melt sheets (Howard and Wilshire Lucchitta and Ferguson (1983) have 1975), the occurrence of both ponding and recognized possible areas within Valles MARTIAN CRATERS 281

(Figs. 5C and 5D), scallops and terraces subsequent flow and erosion of the channel­ have formed, but it is noticeable that even forming material (also unlike lunar melt for those craters where fluidized ejecta was deposits; Hawke and Head 1977), and the created (inferred from the occurrence of the inferred destruction of Martian impact lobate ejecta lobes), no unusual flows or a melts during the cratering process (Kieffer propensity for landslides and subsidence and Simonds 1980) all argue against the are seen on the inner crater walls. This lack impact melt origin of these channels on of "unusual" interior features is believed to Cerulli. Thus the question remains as to be strong circumstantial evidence against why the inferred water discharge occurred the general occurrence of liquid water within Cerulli and not other Martian craters within the target rocks (and derived ejecta) of a similar size and latitude. A plausible at the time of crater formation. At a explanation would be that the water aquifer resolution of 8 m, it is expected that outflow was unusually close to the surface at the channels and slump features (due to the time of crater formation, as is also sug­ percolation of water from the rim material) gested by the unusually large amount of would be observable if they existed. slumping of the inner wall terraces and by Study of lower-resolution (42 m per the proximity of the source region for pixel) images of the crater Cerulli (32°N, Mamers Valles, which is located less than 338°W) does, however, show ample 50 km west of the rim of Cerulli. evidence of postcrater formation modifi­ In several instances, these channels cation by surface flow in the form of originate high on the inner walls of Cerulli numerous channels around the rim crest of and descend to the crater floor in a series of this 120-km-diameter crater. Figure 6 steps over the wall terraces. Evidence for shows both an example of these channels ponding of the flows can be found behind and a geomorphic sketch map illustrating these terraces, where flat, morphologically their distribution. Note that these channels bland areas can be seen in many local occur both on the inner wall and on the catchment basins prior to overflowing of exterior continuous ejecta deposits and the local obstacle and the formation of extend to a radial distance of 1.5 crater radii another channel on the downslope side of from the rim crest. Within Cerulli the chan­ the terrace. Only very subdued topography nels are generally 5-20 km in length, have and textural variations, possibly indicating widths of 150-650 m, are sinuous in plan, a deltaic plain, can be seen where these and appear similar to individual segments channels reach the crater floor. Because the of the channel networks identified in the channels appear to be deeply cut into the Martian heavily cratered terrain (cf. Baker wall material of Cerulli, this apparent lack 1982). The preferred interpretation for of erosional debris would imply that channel origin is that they are fluvial chan­ although the valleys were easily formed, nels carved by water percolating out of the the volume of material transported was emplaced wall material of Cerulli at an either quite small or the transported unspecified (but possibly quite short) time material was widely distributed over the after the cratering event. However, it crater floor. should be remembered that superficially While it is very unusual for such channels similar flows and channels have been seen and ponded material to be associated with on certain lunar impact craters, where an fresh impact craters on Mars, this type of impact melt origin was described (Howard flow and ponding has been seen elsewhere and Wilshire 1975) . Lack of the crenulated on the planet and was also attributed to surface texture associated with lunar surface flow of small volumes of water. impact melt sheets (Howard and Wilshire Lucchitta and Ferguson (1983) have 1975), the occurrence of both ponding and recognized possible areas within Valles " I •N • \

B

282 MARTIAN CRATERS 283

Marineris and the outflow channels Simud (I) The ejecta lobes retain much of the and , while Mouginis-Mark small-scale structure that was produced (1985) identified several small meltwater during ejecta deposition. Radial striations, channels and transient lakes to the north­ pressure ridges, and the sharp distal ram­ west of the Elysium volcanoes, where parts are all well preserved on craters rang­ similar surface flow and ponding evidently ing in size from 4.5 to 26 km in diameter. took place. Thus despite calculations by The general absence of remobilized Carr (1983) that surface flow of small materials as a consequence of water sap­ volumes of water is likely to be short-lived ping from the emplaced ejecta suggests that late in Martian history, there is sufficient the lobes possessed appreciable mechanical morphologic evidence from different geo­ strength at the time of their emplacement logic settings to indicate that surface flows and that the liquid water content of the at the scale described here for the ejecta flows was very low. No evidence to support blankets can occur under unusual cir­ the proposal by Woronow (1981) that the cumstances. ejecta contained 16 to 72% water has been found. SUMMARY AND SPECULATIONS (2) For two ejecta blankets within Shia­ Previous studies of Martian impact parelli Basin that have been partially craters have attempted to identify the mode eroded (Fig. 2), it is likely that large blocks of fluidization of the ejecta lobes from were transported within the ejecta flow and regional studies and from theoretical and that some of these blocks were most likely laboratory models. In this study, an ice-rich. Subsequent melting or ablation of alternative approach has been taken this ice has led to collapse of the ejecta wherein the highest-resolution Viking deposit in isolated places to form features Orbiter images have been used to search for similar to terrestrial kettle holes. A morphologic signs of water or ice existing plucking process, wherein material from in the ejecta blankets immediately follow­ beyond the crater cavity was incorporated ing ejecta deposition as a method by which within the ejecta flow , appears to be the to infer physical properties of the target most likely mechanism for the transport rocks. Images with spatial resolutions be­ and burial of these ice-rich blocks. tween 8 and 17 m per pixel have been used (3) The same eroded ejecta blankets to study both the ejecta deposits and the within Schiaparelli Basin reveal that flow crater interiors. The following observations striations extend through the thickness of and conclusions are made: the ejecta deposit, rather than simply being

FIG. 6. (A) Detail s of channels on inner wall of crater Cerulli (see Fig. 68 for location). Although this image is at a lower resolution than the other images used in this paper, many of these channels nevertheless show morphologic similarities to the channel networks in the Martian highland s, but are not di ssimilar to th e impact melt flow s on the lunar craters Copernicus and King (Howard and Wil shire 1975). The former mode of formation is preferred here, as di scussed in the text. Viking Orbiter frame 204S 19 . Picture height is equivalent to 54.4 km and image resolution is 42 m/pixel. (B) Sketch map of the 120-km-diameter crater Cerulli, located at 32.SON, 338.0°W, showing di stribution of channels (arrows show inferred direction of flow) within the interior of the crater and on the ejecta deposits. Note that these flow directions are almost random on the ejecta blanket, due to the undulating nature of the local topography. The rim crest of Cerulli is marked by triangular barbs and other crater rims by single barbs. Note that unusually large number of terraces for a crater of this diameter and the great width (in excess of 40 km in some places) of the inner wall material , suggestive of large-scale subsidence of the wall due to an unusually high concentration of entrained volatiles. Mapped from JPL photomosaic 211-5959, which includes Viking Orbiter images acquired on orbits 203S, 204S, and 205S. Location of Fig. 6A is outlined by the box. 284 PETER J. MOUGINIS-MARK

superficial features. This would imply that haps I 0 volume percent?) at the time of the ejecta flow was nonturbulent, akin to impact. The surface channels (both on the poorly fluidized terrestrial pyroclastic ejecta blankets and within craters) and flows. kettle holes argue against atmospheric (4) Most of the crater interiors studied effects dominating ejecta fluidization, while here reveal few signs of meltwater release the stability of small-scale topography and or the abnormal amounts of slumping that only the rare occurrence of surface chan­ would be expected if the rim material were nels indicate that liquid water was not com­ water-rich. monly associated with the ejecta. (5) Rare examples of surface flow can be As a topic for future investigation, nu­ found on both the ejecta deposits and crater merical models need to be developed to inner walls. In the case of a small crater determine the effects of different volume­ southwest of Bakhuysen, it is believed that tric amounts of ice that would have to be this surface flow was made possible by the entrained within the ejecta to produce the formation of this small crater on the ejecta observed morphologies. Topographic data blanket of Bakhuysen. For the crater Cer­ for the ejecta lobes are virtually absent, ulli, an unusual water-rich substrate was although photometric techniques (Davis evidently present. Extensive surface flow, and Soderblom 1984) offer exciting pros­ expressed by numerous small channels on pects for inferring the relief and, hence, the wall and continuous ejecta blanket, as the rheology of the flows when they were well as the creation of very wide terraces, emplaced. From the limited set of shadow took place following the formation of the measurements of topography that have so crater. The local water table may also have far been made, the several tens to hundreds been sufficiently disturbed by the formation of meters of relief on the lobes and ram­ of Cerulli crater to release the water parts would suggest a high mechanical responsible for the formation of Mamers strength for the ejecta lobes and that the Valles. volumetric amount of ice that was incorpo­ It should be remembered that only a very rated within the ejecta was quite low. small number of craters, all in a narrow Quantitative estimates of this amount of ice latitude band (predominately in the Sou­ are beyond the scope of this work, but it thern Hemisphere), have been included in appears likely that values for the volume of this investigation, primarily due to the ground ice are consistent with the regolith/ sparse coverage of the high-resolution atmosphere models proposed for Mars and Viking Orbiter images. Individually, the can explain the observed landform mor­ preceding observations do not argue con­ phology. vincingly for a single set of properties for Clearly, however, the observations the target material or the derived ejecta; presented here pertain only to a very small small volumes of liquid water, larger number of fresh craters that by chance volumes of ice, or purely atmospheric were imaged at very high resolution by the effects could all be used to explain ejecta Viking Orbiters. While these craters pro­ fluidization. Taken together, however, it is vide an insight into the physical state of the argued that the observational data support target material for these random areas, the proposal that ice, rather than water or image resolution is inadequate to draw firm atmospheric effects, dominated both the conclusions on volatile state for most of emplacement process and the mode of Mars. As a result, it is concluded that the ejecta fluidization for the majority of ram­ identification of small-scale features diag­ part craters studied here. By implication, nostic of water sapping and ejecta remobi­ ice must have been present in the target lization on crater ejecta blankets rocks in significant amounts (a few to per- constitutes one of the many interesting MARTIAN CRATERS 285 experiments that should be conducted with HALE-ERLI CH, W. S. 1986. Implications for substrate the submeter-resolution camera planned as volatile distributions on Mars from complex crater part of NASA's Mars Observer Mission. morphology and morphometry (Abstract). Lunar Planet. Sci. XVII, 303-304. ACKNOWLEDGMENTS HAWK E, B. R., AND J . W. H EAD 1977. Impact melt on lunar crater rims. In Impact and Explosion This research was supported by NASA under Grant Cratering (D. J. Roddy, R. 0. Pepin, and R. B. NAGW 437 from the and Geo­ Merrill , Eds.), pp. 815-841. Pergamon, New York. physics Program. We thank Don Gault, Lionel Wilson, HoRNER, V. M. , AND R. GREELEY 1986. Effects of Steve Clifford , and an anonymous reviewer for their elevation and plains thicknesses on martian crater comments on an earlier version of this manuscript. ejecta morphologies on the ridged plains (Abstract). This is Hawaii Institute of Geophysics Contribution In Rpts. Plan. Geol. and Geophys. Prog. 1985 , No. 1870. NASA TM-88383 , pp. 446-448. HoRZ, F. , R. OsTERTAG, AND D. A. RAINEY 1983. Bunte breccia of the Ries: Continuous deposits of REFERENCES large impact craters. R evs. Geophys. Space Phys. BAKER, V. C. 1982. The Channels of Mars. U niv. of 21, 1667-1725. Texas Press, Austin. HOWARD , K. A., AND H. G. WILSHIRE 1975. Flows of BoYCE, J. M. 1979. A method for measuring heat flow impact melt at lunar craters. J. Res. U.S. Geol. in the martian crust using impact crater morphology Survey 3, 237-251. (Abstract). In Rpts. Plan. Geol. Prog. 1978-1979 , JOHANSEN 1979. The latitude dependence of martian NASA TM-80339, pp. 114-118. splosh cratering and its relationship to water BRAKENRIDGE , G. R., H . E. NEWSOM , AND V. R. (Abstract). In Rpts. Plan. Geol. Prog. 1978-1979, BAKER 1985 . Ancient hot springs on Mars: Origins NASA TM-80339, pp. 123-125. and paleoenvironmental significance of small mar­ KARGEL, J . S. 1986. Morphologic variations of martian tian valleys. Geology 13, 859-862. rampart crater ejecta and their dependencies and CA RR , M. H. 1979. Formation of martian flood fea­ implications (Abstract). Lunar Planet. Sci. XVII, tures by release of water from confined aquifers. J . 410-411. Geophys. R es. 84, 2995-3007. KI EFFE R, S. W. , AND C. H . SIMONDS 1980. The role CARR , M. H . 1983 . Stability of streams a nd lakes on of volatiles a nd lithology in the impact cratering Mars. Icarus 56, 476-495. process. Revs. Geophys. Space Phys. 18, 143-1 81. CARR, M. H . 1986. Mars: A water-rich pl anet LAITY , J. E ., AND M. C. MALIN 1985. Sapping (Abstract). In Symposium on Mars: Evolution of Its processes and the development of the theater­ Climate and Atmosphere, LPI Contrib. No. 599, pp. headed valley networks on the Colorado Plateau. 9-11 . Geol. Soc. Amer. Bull. 96, 203-217. CARR, M. H ., L. s. CRUMPLER, J . A. CUTTS, R. LUCCHITTA , B. K., AND H. M. FERGUSON 1983. GREELEY , J. E . GuEST, AND H. MASURSKY 1977. Chryse Basin channels: Low-gradients and ponded Martian impact craters and emplacement of ejecta flows. Proc. Lunar Planet. Sci. Conf 13th , Part 2, by surface flow . J. Geophys. R es. 82, 4055-4065. J . Geophys. R es. 88, Suppl., A553-A568. CLIFFORD, S. M. 1986. Mars: Crustal pore volume, MooRE, H . J . 1972 . Ranger and other impact craters cryosphere depth, and the global occurrence of photographed by Apollo 16. In Apollo 16 Prelimi­ groundwater (Abstract). In Symposium on Mars: nary Science R eport, NASA SP-315 , pp. 29-45 to Evolution of Its Climate and Atmosphere, LPl 29-51. Contrib. No. 599, pp. 18-20. MouGINIS-MARK , P. J. 1981. Ejecta emplacement and DAVI S, P. A. , AND L.A. SODERBLOM 1984. Modeling modes of formation of martian fluidi zed ejecta crater topography and albedo from monoscopic craters. Icarus 45, 60-76. Viking Orbiter images. I. Methodology. J . Geophys. MouGINIS-MARK , P. J. 1985. Volcano/ground ice Res. 89, 9449-9457. interactions in , Mars. Icarus 64, FANALE, F. P., AND B. M . JAKOSKY 1982. Regolith­ 265-284. atmosphere exchange of water and carbon dioxide MOUGINIS-MARK , P. J . 1986. lee or liquid water in the on Mars: Effects on atmospheric history and climate martian regolith ? Morphologic indicators from ram­ change. Planet. Space Sci. 30, 819-831. part craters (Abstract). In Symposium on Mars: FANALE, F. P., J. R. SAL VAIL, A. P. ZENT, AND s. E. Evolution of Its Climate and Atmosphere, LPI Con­ PosTAWKO 1986. Global distribution and migration trib. No. 599, pp. 67-69. of subsurface ice on Mars. Icarus 61, 1-18. MOUGINIS-MARK , P. J., AND D. L. CAREY 1980. GAULT , D . E., AND R. GREELEY 1978. Exploratory Crater studies in the Northern Plains of Mars: experiments of impact craters formed in viscous­ Thickness estimates of fluidized ejecta deposi ts liquid targets: Analogs for martian rampart craters? (Abstract). In Rpts. Plan. Geol. Prog. 1979-1980, Icarus 34, 486-495. NASA TM-81776, pp. 105-107. 286 PETER J. MOUGINIS-MARK

MouGINIS-MARK, P. J. , AND E . A . CLOUTIS 1983. impact craters (Abstract). Lunar Planet. Sci. XV, Ejecta areas of impact craters on the martian ridged 732-733. plains (Abstract). Lunar Planet. Sci. X IV, 532-533. SCHULTZ , P. H. , AND J. SINGER 1980. A comparison MOUGINIS-MARK, P. J. , V. L. SHARPTON, AND B. R. of secondary craters on the Moon, Mercury and HAWKE 1981. Schiaparelli Basin, Mars: Mor­ Mars. Proc. Lunar Planet. Sci. Conf. lith, 2243- phology, tectonics and infilling hi story. Multi-ring 2259. Basins, Proc. Lunar Planet. Sci. UA, 155-172. Sc HUMM, S. A. , AND L. PHILLIPS 1986. Composite channels of the Canterbury Plain, New Zealand: A MUTCH, P. , AND A. WORONOW 1980. Martian rampart martian analog? Geology 14, 326-329. crater and pedestal craters' ejecta-emplacement: SQUYRES , S . W. , AND M. H. CARR 1986. Geomorphic . Icarus 41, 259-268. evidence for the distribution of ground ice on Mars. ScHULTZ, P. H. 1986. Crater ejecta morphology and Science 231, 249-252. the presence of (Abstract). Jn Sym­ WILSON , C. J. N. 1980. The role of fluidi zation in the posium on Mars: Evolution of Its Climate and emplacement of pyroclastic flows: An experimental Atmosphere, LPJ Contrib. No. 599, pp. 95-97. approach. J . Volcano/. Geotherm. Res . 8, 231-249. SCHULTZ, P. H. , AND D. E. GAULT 1979. WOHLETZ , K . H. , AND M. F. SHER ID AN 1983. Martian Atmospheric effects on martian ejecta em­ rampart crater ejecta: Experim ents and analysis of placement. J. Geophys. Res. 84, 7669-7687. melt-water interaction. Icarus 56, 15-37. ScHULTZ, P. H. , AND D. E. GAULT 1981. Ejecta WOOD , C. A. , J . w. H EAD, AND M. J. CINTALA 1978. emplacement and atmospheric pressure: Laboratory Interior morphology of fresh martian craters: The experiments (Abstract). In Third International effects of target characteristics. Proc. Lunar Planet. Colloq. on Mars, LPI Contrib. No. 441 , pp. Sci. Conf. 9th , 3691-3709. 226- 228. WoRONOW , A. 1981. Preflow stresses in martian ram­ SCHULTZ , P. H ., AND D. E . GAULT 1984. On the part ejecta blankets: A means of estimating the formation of contiguous ramparts around martian water content. Icarus 45, 320-330.