Meteoritics & Planetary Science 53, Nr 4, 672–686 (2018) doi: 10.1111/maps.13042
Martian cratering 12. Utilizing primary crater clusters to study crater populations and meteoroid properties William K. HARTMANN
1*
, Ingrid J. DAUBAR2, Olga POPOVA3, and Emily C. S. JOSEPH1
1
2
Planetary Science Institute, Tucson 85719–2395, Arizona, USA Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California, USA 3 Institute for Dynamics of Geospheres, Moscow, Russia * Corresponding author. E-mail:
[email protected] (Received 06 December 2016; revision accepted 30 November 2017)
Abstract–Images from Mars Global Surveyor and later images from Mars Reconnaissance Orbiter reveal that roughly half of the meteoroids striking Mars (at meter to few decameter crater diameters) fragment in the Martian atmosphere, producing small clusters of primary impact craters. Statistics of these “primary clusters” yield valuable information about important Martian phenomena and properties of interplanetary bodies, including meteoroid behavior in the Martian atmosphere, bulk strengths of bodies striking Mars, and the fraction of Martian “field secondary” craters, a datum that would improve crater count chronometry. Many Martian impactors fragment at altitudes significantly higher than 18 km above the mean surface of Mars, and we find that most bodies striking Mars and Earth have low bulk strengths, consistent with crumbly or highly fractured objects. Applying statistics of primary clusters at various elevations and independent diameter bins, we describe a technique to estimate the percentage of semirandomly scattered “field secondary” craters. Our provisional estimate of this percentage, in the diameter range ~250 m down to ~22 m, is ~40% to ~80% of the total impacts, with the higher percentages at smaller diameters. Our data argue against earlier suggestions of overwhelming dominance by either primaries or secondaries in this diameter range.
BACKGROUND Mars Global Surveyor MOC (Mars Orbiter Camera) imagery revealed that new primary impact craters of roughly decameter-scale are forming on Mars during few-year intervals. The resulting catalog of 19 new primary impact sites (Malin et al. 2006a) included the discovery that many small Martian impactors fragment during atmospheric entry; it listed 11 single craters (= 11–28 m) and 8 clusters (42% of total), with largest craters 10–22 m. Here we include one case where smaller craters were unresolved “dark spots” clustered around a 16 m new crater. (Note: In the original MOC catalog, a possible 20th new single crater at D = 148 m was later rejected.) Mars Reconnaissance Orbiter CTX (Context Camera) and HiRISE (High Resolution Imaging Science Experiment) imagery produced a larger, still growing, catalog of new primary impacts,
© The Meteoritical Society, 2018.
again with both single and clustered craters (Daubar et al. 2013). Their data indicate a higher percentage of clusters, ~53% at these small crater sizes. In summary, roughly half of incoming meteoroids in the relevant size range fragmented during passage through the Martian atmosphere (Malin et al. 2006a; Ivanov et al. 2009; Daubar et al. 2013, 2015). We refer to these as “primary clusters.” Examples of new clusters are shown in Fig. 1. Older primary clusters must, of course, exist. During various crater counting projects, we have noticed virtually identical-scale small clusters, many of them mixed among existing single craters, and many of them degraded in morphology, being presumably much older. Examples are shown in Fig. 2. As discussed in a later section, we infer that they are also primary clusters. In early studies of fragmentation in the Martian atmosphere, even before primary clusters were
672
Utilizing Martian crater clusters
(a)
(b)
673
(c)
Fig. 1. Images of newly formed clusters found by Mars Global Surveyor and by Mars Reconnaissance Orbiter (CTX and HiRISE imagers). a) 2006 MOC discovery image (MOC image S16-01426) of a newly formed cluster (Lat. 13.9 N, Long. 151.9 W, Elevation 3.6 km) formed in interval 2004–2006. Aureole with dark and bright features covers much of the middle and left part of frame. b) Same cluster photographed in 2007 at higher resolution by HiRISE (PSP_003925_1940). c) Cluster in eastern Amazonis: Lat. 18.7 N, Long. 151.8 W, Elev. 3.8 km (ESP_017770_1990). New craters are darkest spots at head of each dark aureole streak (possibly caused by strong prevailing winds during impact?). North approximately at the top in this figure and Fig. 2.
discovered on Mars, Popova et al. (2003) concluded that primary impactors should produce “small clusters” at roughly 50 m–200 m cluster diameter. Popova et al. (2007) also described two classes of clusters, “small clusters (crater diameter D ~ 10s m spread over few hundred m),” and large clusters (D ~ few hundred m, spread over 2–30 km). They discussed how secondary ejecta could produce the second category, which we will call “distant secondary clusters,” kilometers in diameter. They inferred that the small clusters were created by primary impacts and large clusters were created by secondary ejecta with long flight times. They noted that size distributions in large clusters were typically more bell shaped than the usual power-laws, possibly due to fragmentation and loss of the smaller ejecta blocks during flight. Here, we will discuss the formation and implications of clusters in additional detail. COMPARISON OF MARTIAN AND TERRESTRIAL ATMOSPHERIC INTERACTION WITH METEOROIDS Popova et al. (2003, table 1) showed 21 fragmentation events in Earth’s atmosphere. None are higher than 46 km; seven are at 30 km or below. For each event, they used entry fragmentation modeling theory to list equivalent breakup altitudes above Mars, from 24 km down to 0 km. In addition, they discussed 14 meteoroids photographed by Ceplecha, among which five had end heights and maximum luminosity (usually corresponding to major breakup) in the range of 90 km down to 60 km, corresponding to Martian elevations well
above the top of Olympus Mons. One of these terrestrial events, studied in detail, disrupted under load of about 0.2–1 bar. Popova et al. (2003) suggested a possible cometary origin for this weak object, and suggested that meteoroids of this strength, if smaller than 5000 kg, “would not reach the ground on present day Mars.” Ivanov et al. (2009, 2014) also examined fragmentation in the Martian atmosphere. They considered the separation of individual craters within the newly found primary clusters, inferring that a significant fraction of the current impactors may have a low density, given several assumptions about breakup altitude and the separation of fragments during descent. Similarly, Popova et al. (2011) showed that among 13 terrestrial fireballs with well-observed trajectories and recovered meteorites, in all cases the bulk strength estimated from first observed fragmentation at high altitude was much less than the bulk strength of the recovered stony meteorite samples on the ground, leading to the conclusion that most interplanetary meteoroids striking Earth, even at meter scales, are weaker than coherent rock meteorites. They inferred that most meter-scale meteoroids in space are laced with fractures from asteroid collision events, or may even have “rubble pile” structure. Figure 3a shows the breakup behavior of 22 wellobserved terrestrial fireballs (including observed initial breakups), translated to the Martian atmosphere. The graph indicates that ~41% of these objects would have fragmented above the highest volcanoes on Mars (~24 km), 50% of these objects above ~18 km altitude on Mars (the elevation of the cluster shown in Fig. 2b),
674
W. K. Hartmann et al.
(a)
(b)
(c)
(d)
Fig. 2. Probable primary clusters found during crater counts in different regions. a) A dune-covered cluster on Olympus Mons, at elevation +18 km (ESP_16042_1985_RED). This image illustrates how high crater density at these scales (and dunes) makes primary clusters difficult to distinguish from random groupings (hence we rejected data from Olympus Mons). b) Fresher cluster on Arsia Mons. Elevation +6.2 km (ESP_031008_1655_RED). c) Cluster of craters with similar, degraded morphology on Alba Mons. Elevation +5.8 km (PSP_004886_2200_RED). d) Fresher cluster in Hellas. Elevation 6.7 km (PSP_006779_1415_RED).
and 59% above 10 km. Figure 3b is a related plot, showing breakup altitudes versus entry velocity, derived from altitudes of maximum luminosity during entry of 141 fireballs observed from satellites, as discussed by Brown et al. (2016). Brown et al. assumed that maximum luminosity, which normally falls between the first breakup and final breakup shown in Fig. 3a, marks the major breakup event. Picking points on the vertical lines connecting first and final breakup events in Fig. 3a, we find plausible agreement with the data in Fig. 3b, and with the elevations noted in Figs. 1 and 2. In short, there is a reasonable consistency in the available terrestrial meteoritic breakup data and their translation to Mars. In a separate study of 521 new impact sites at various elevations observed by HiRISE on Mars,
Daubar et al. (2015) found that ~56% of the impact events involved fragmentation of the meteoroid and production of clusters, but their results indicate this result may vary with elevation, as discussed later in the Implications Regarding Meteoroid Fragmentation and Primary Cluster Production section. The availability of such statistics suggested to us that primary clusters may offer a direct method for measuring the percentage of secondary “field secondary” impacts among the total population of Martian craters in the decameter size range, as we will explain below. This would be a very valuable measurement for future Martian studies, because the percentage of field secondaries affects estimates of the uncertainty of crater retention ages, as measured by crater count procedures. The abundance of secondary
Utilizing Martian crater clusters
(a)
675
(b)
Fig. 3. Behavior of terrestrial fireballs if they had entered Mars atmosphere. Observations of terrestrial breakup events were used to calculate altitude of those fragmentations as they would have occurred in Martian atmosphere. a) First and last breakup events based on 22 well-observed fireballs with recovered meteorites. Light gray data points represent initial breakup events. Black data points represent last breakup events. Squares indicate three carbonaceous chondrites (based on recovered samples); they represent 3 out of the 10 highest initial breakups, above 20 km on Mars. Thirteen (59%) of these objects would have fragmented above 10 km on Mars. b) Brightest luminosity (most violent breakup?) events, based on 144 satellitebased terrestrial observations, plotted versus entry velocity (Brown et al. 2016). See text for further discussion of both data sets.
craters has been controversial. For example, Neukum and Ivanov (1994, p. 376), and Neukum (personal communication, 2007) proposed that the great majority of impacts at D < 1 km were primary (i.e., interplanetary bodies, favoring reliable ages), whereas McEwen et al. (2005) and McEwen and Bierhaus (2006) implied that the secondary craters at decameter scales overwhelmingly swamp the primaries and render crater count chronometry unreliable (even though field secondary abundances must correlate roughly with age over the long haul). Hartmann and Daubar (2017) and Bierhaus et al. (2018) offer further discussion. Our initial idea (Hartmann et al. 2013), expressed in very idealized terms using the above data, was as follows. Count the number of primary clusters in any area (in a given diameter bin), and convert it to what we call the “effective cluster diameter” (diameter of the crater that would have been formed had the impactor not fragmented). Then assume that about half the impacts make clusters and multiply the number of primary clusters by ~2 in each diameter bin. This should give roughly the total number of primary impacts in that bin. Thus, the rest of the randomly scattered craters would be field secondaries. We will evaluate this idea in much more detail below. In the following study, we define “elevation” as distance of a surface above or below the mean Martian surface, and “altitude” as distance in the atmosphere above the mean Martian surface. We then examine the fraction of impactors that fragmented before impact as a function of surface elevation and as a function of “effective cluster diameter.” The possible correlation of these fractions with surface elevation and with effective
crater size gives a clue about the altitudes at which meteoroids fragment in the Martian atmosphere, with corresponding implications for distribution of bulk strengths among those meteoroids. Applying all these ideas, we illustrate how such data may allow researchers to estimate percentages of field secondaries as a function of elevation and crater size at various Martian sites. The present paper provides a test of the concept; we point out various steps where further work would be valuable. CAN WE DISTINGUISH “PRIMARY CLUSTERS” FROM SECONDARY CLUSTERS ON MARS? We suggest that the major issue in the reliability of the present study is whether we can distinguish older “primary clusters” (formed by interplanetary impactors) from “distant secondary clusters.” A theoretical argument that this can be done starts with the observation that fragments of exploded impactors on Mars, traveling at cosmic speeds, have only a few seconds between the breakup event and impact on the ground. Popova et al. (2003) proposed Martian atmosphere entry velocities with a mean value of 10.2 km s1, based on work of Bland and Smith (2000), Davis (1993), and Flynn and McKay (1990). Lateral spreading velocities of fragments (during the few seconds available between breakup and impact), as estimated following Passey and Melosh (1980) and Artemieva and Shuvalov (1996), were used by Popova et al. (2003) to conclude that “the progressive fragmentation model predicts [primary] clusters of small craters with a diameter (D) of order 1–10 m, spread
676
W. K. Hartmann et al.
over a few hundred meters.” This appears to be roughly confirmed by the observations of the newly formed primary clusters (Fig. 1). For example, in the Mars Global Surveyor catalog (Malin et al. 2006b), six of the eight observed clusters have a cluster diameter of ≤70 m, and the other two have diameters of ~240– 250 m. In the work reported here, we defined “diameter” of the primary cluster itself as a rough circle that can enclose ~80% of the craters in the cluster, based on a visual estimate. Daubar et al. (2016) characterized cluster sizes more rigorously in a different way, according to “dispersion values,” defined as standard deviation between the distances between each combination of pairs of craters (believed to be part of the cluster). Their resulting cluster dispersion values ranged from a few to several hundred meters, but most were in the range of 20–80 m. (These should be somewhat less than the values used here, which are based on outer edges of clusters.) These concepts suggest an upper limit on the plausible sizes of primary clusters. Suppose we assume some limit on separation distance of fragments imposed by spreading velocities and the number of seconds before fragmenting primary material hits the ground. Let us call this hypothetical limit D. Ivanov et al. (2014) (their fig. 4) showed a most common value for separation of fragment pairs in the diameter bin 22– 32 m, and in models and observations, at roughly 20– 60 m. A great majority of their separations are at roughly 10–2000 m. (Thus, 2000 m may approximate our D concept.) If we now imagine larger and larger fragmenting meteoroids, we reach a point where the kinetic energy of the swarm of fragments is large enough to excavate a crater larger than our upper dispersion limit, D. Thus, at these large meteoroid sizes, even a fragmentation event creates only a single crater. Hence, the percentage of impact events forming primary clusters must drop to zero above some diameter. Our data and the discussions above suggest that the upper limit for primary cluster diameters (as we have defined them) may be in the range of 250–2000 m. Now let us consider secondary ejecta. A block of ejected debris must travel at less than escape velocity (5.0 km s1) in order to hit the ground. Quantin et al. (2016) reviewed literature on secondary fragment ejection velocities and considered ejection velocities of 0.3 to 1.5 km s1 for secondaries within the Gratteri ray system at the distances 50–450 km from primary. Thus, we suggest common velocities of ~2 to nearly 5 km s1 for typical distant “field secondaries,” which produce isolated, semirandomly spaced secondary clusters and individual secondary craters, beyond obvious ray systems (far enough that they might not be recognized as part of a local large primary’s field of
secondary ejecta). Here, we suspect that most of the boulder fragmentation events occur during the stresses and shock involved in explosive launch from the impact site (rather than during post-launch flight). If the ejection velocity is about 2 km s1 and angle is about 45°, then the flight time is ~750 s (no deceleration is assumed, implying a boulder size ~4 m or larger). The resulting field secondary impact is ~1000 km from the primary crater. If the lateral spreading velocity is ~5– 10 m s1 (plausible for secondary debris, see Popova et al. 2007; fig. 8), then the fragments would spread over ~4–8 km, similar in size to the proposed large (secondary) cluster shown by Popova et al. (2007). Even if some boulders fragment halfway through flight, the situation suggests common flight times >100 s for the fragments that produce the semi-randomly scattered distant secondary impacts. This in turn suggests that if the ejected fragments disrupt at spreading speeds comparable to those of primary meteoroids, then distant secondary clusters could have spreading diameters of more than an order of magnitude bigger than primary clusters. If we adopt 50–300 m as typical of primary clusters, and at least ~500 to ~4500 m as typical for distant secondary clusters, then small, tight clusters with diameters of ≲300 to 500 m, if far from obvious secondary crater field near larger primaries, should mostly be primary clusters. (In this study, we try to stay away from our estimated transition zone in diameters.) We tested these ideas by studying secondary clusters among ray systems and scattered secondaries from Martian craters Gratteri and Zunil, fresh craters with ray systems still prominent in infrared imagery. In the case of the young rayed crater Gratteri, 7 km in diameter, Quantin et al. (2016) studied the distribution of secondaries. Their fig. 3 shows a secondary cluster in a ray, 103 km from Gratteri, about 250 m wide and 900 m long, with the long axis oriented about 30° off the axis of the ray. (In these studies, we assume that the width is the result of the spreading velocity during flight; the length measured along the ray is a result of less important spreading of velocities along the flight path during the jetting phenomena and atmospheric drag, especially at low angles of launch and impact.) Quantin et al.’s fig. 4 shows a cluster 150 km from Gratteri, about 350 9 550 m, composed of several craters up to 130 and 180 m in diameter. This cluster is relatively isolated, as seen in the full HiRISE image. At that same distance, W. K. H. made crater measurements along a transect (perpendicular to the ray) across two Gratteri rays on HiRISE 008514_1630. The high secondary crater densities defining each ray extended over a width of ~5 km. One ray contained a cluster about 500 m wide stretching at least 2 km along the
Utilizing Martian crater clusters
ray; the other ray contained the cluster (shown in their fig. 4, mentioned just above) plus a larger cluster about 800–900 m wide and extending more than 2 km along the ray. Their fig. 5 shows part of a cluster 230 km from Gratteri, about 1.2 km wide and stretching at least 4.5 km in a direction radial to Gratteri. These results support our view that distant secondary clusters associated with rays, at least from this crater, reach larger sizes than primary clusters. We expected that the secondaries of young rayed crater Zunil (10 km diameter) would give an even better test, because (especially to the south of Zunil) they tend to stand out as fresh, sharp-rimmed craters, usually with a dark aureole and/or floor that is prominent on the sparsely cratered, lighter-toned plain (McEwen et al. 2005). (The dark features sometimes include more diffuse bright aureoles, lighter than the background.) The aureoles seem to suggest materials excavated from depth, since they are more prominent on larger, deeper craters. The aureoles appear to be tertiary ejecta veneers that disappear on geologically short time scales, but longer than the order-of-magnitude 0.06–1 Ma age estimated for Zunil (Hartmann et al. 2010). To test our ideas about cluster sizes at Zunil, in a less rigorous survey than the Gratteri work, W. K. H. and E. C. S. J. studied a set of HiRISE images, following a transect proceeding south (in a radial direction) from Zunil, seeking examples of Zunil secondary clusters. Here, however, the data on secondary clusters were more complex than at Gratteri. In an initial survey, we identified three classes of clustered features that appeared to be associated with Zunil ejecta. (1) “Ordinary clusters” of small craters with sharp, raised rims. These clusters can usually be circumscribed by a circular or mildly elliptical area containing less than one or two dozen resolved craters, typically lying within a few crater diameters of each other. The larger, deeper craters in the clusters are usually the ones distinguished by the dark aureoles. The smallest of these clusters would be most easily confused with primary clusters, except that they are associated with a relatively obvious secondary crater field surrounding a nearby primary crater (Zunil in this case). (2) Scattered “strings” of craters usually oriented within ~20° of radial to Zunil. As with class 1, these craters have sharp, raised rims and aureoles around the larger examples, but they stretched along ray-like streamers more than twice as long as wide (often three to five times longer than width), and typically contained several dozen craters. The larger numbers of craters, relative to class 1, are consistent with the ejecta being embedded in denser masses of (spinning?) jetted material that form ray-like distribution of impacting material. (3) At larger distances from Zunil, unusually
677
broad “patches” of hundreds of similar-appearing small craters (each crater generally 300–500 m, distinguishing primary clusters from secondary clusters is more difficult, and potential information content about field secondaries is more “noisy.” The clusters in ray-like strings that we labeled as class 2 seemed clearly to increase in size with greater distance from Zunil, and even the three class 2 clusters encountered at only 88–104 km from Zunil, 1.8–3.3 km long, mentioned above, seem too big (and with too many craters) to be mistaken for primary clusters. These clusters match the result we described above from the study of Gratteri secondaries in rays. The class 3 clusters, with hundreds of craters spread over many kilometers, appear too populated and too large to be confused with primary clusters. Their origin remains enigmatic, possibly being related to secondary debris being lofted at high velocities and high angles into space, with long flight times (rather than low angles, low velocities, and lower flight times that might reach the same distance from Zunil, then re-entering the atmosphere with further fragmentation, as discussed by Popova et al. 2007). To summarize the Zunil and Gratteri studies, we infer, on the grounds of the observations and theoretical arguments about spreading during flight times, that the smaller, randomly scattered clusters with cluster diameter ≲300 m are mostly primary clusters, and contain useful information about primary meteoroids and characteristics of primary versus secondary craters. We suggest that more systematic observations of the sort described above would be useful in understanding morphology of secondary crater clusters versus distance from primary craters of various ages, all in the cause of distinguishing primary clusters from secondary clusters. As we show below, this will allow
measurements of the fractions of field secondaries, needed to improve crater count chronometry. Moving to a different issue, our logic about fragment spreading time suggests that primary clusters forming at low surface elevations might have larger sizes than primary clusters at high elevations, because of longer time for lateral spreading of fragments during atmospheric passage. Moreover, one might argue that weak fragments continue to break up during descent, so that low-elevation clusters would be “clusters of clusters,” possibly larger than those for high-elevation clusters. We have not noticed the latter effect. Regarding the first effect, W. K. H. and E. C. S. J. examined clusters that E. C. S. J. had counted on Alba Mons, the highest elevation counts in our study, at +5.8 km (see next section). We examined about half the total sample of 110 clusters and recorded dimensions of 18 of the well-defined clusters. Focusing on the short axis (presumably most indicative of spreading time after breakup), we found that in our sample, 14 out of 18 had a short axis 1 m. In the CTX/HiRISE data set available for that paper, the production SFD of direct HiRISE detections of new, dated primary craters (singles + clusters listed by effective diameters) rolls over only at D ~ 6 m. The cause of these rollovers is of interest. Historically, a rollover in any set of crater counts at small sizes has been revealed by later higher resolution imagery merely to indicate incompleteness in crater detections at the smaller diameters in earlier imagery, rather than a truly reliable rollover in the actual numbers of impactors. During five decades of Mars orbiters, with cameras of increasing resolution, rollovers suggested by lower resolution cameras have repeatedly been disproven by higher resolution data. They were typically an effect of attempts to count craters with inadequate resolution, or in some cases observer fatigue from trying to count smallest
Utilizing Martian crater clusters
craters over too large an area, resulting in incomplete counts at smallest sizes. As an example, in the original MOC camera discovery of new, dated impacts on Mars, the SFD curve of the newly detected features rolls over at ~16 m, while the higher resolution HiRISE camera SFD for new features rolls over at ~6 m. A remaining question is whether some physical effect (atmospheric destruction of small meteoroids?) may reduce primary crater and cluster production at small sizes (e.g., 6 m?), since this is the size range in which theoretical models predict atmospheric entry breakup of weak impactors (Popova et al. 2003), with possible total loss of the weakest, most “crumbly” or ice-rich objects. In the Alba Mons data set, the diameter bins that we consider to give the best data, D = 22–31 m and D = 31–44 m (based on high total impact numbers and no rollover of cluster numbers), had total impact counts (single craters + clusters) of 281 and 113, respectively, of which 31 and 17 were clusters. These figures suggest fairly robust statistics for the percentage of clusters in those diameter bins. Intermediate Elevation: Daedalia Planum (+1.6 km) In this data set (Fig. 4b), initial counts by E. C. S. J. were combined with later counts in a smaller area by W. K. H., who attempted to extend the data set to smaller diameters in hilly and smooth terrain in HiRISE PSP_001960_1450 (resolution 25 cm/pixel), amounting to an area of 0.8 km2. The latter survey yielded only two clusters out of 89 P+fS impacts but is worth mentioning because of the paucity of clusters at smallest sizes. The most crater-populated bins at these sizes (5.6– 7.8 m, 7.8–11.0 m, 11–16 m) recorded 37, 34, and 17 total impacts, respectively, but only one cluster (presumed primary) was recorded in these three bins (the 11–16 m bin). The other, larger cluster (also presumed primary) was observed in the D = 44–62 m bin, and it was the only impact feature recorded in that bin. Among the counts of all craters, the rollover in crater number data occurred at D < 5.6 m. To present these data in the format of Fig. 4, we note that useful information is contained in the fact that various D bins with a dozen or more individual craters contained zero clusters of equivalent size, offering rough upper limits on cluster frequency. Thus, for example, in the diameter range 7.8–11 m, with zero clusters out of 34 impacts, we plot an upper limit for cluster abundance of
