Twenty confirmed impacts over a 7-year time period on Mars were qualitatively and statistically compared to 287 secondary craters believed to originate from Zunil, an ∼500 ka, 10-km diameter, primary crater. Our goal was to establish criteria to distinguish secondaries from primaries in the general crater population on the basis of their horizontal planforms. Recent primary impacts have extensive “air blast” zones, distal ray systems (>100 crater radii, R), and ephemeral ejecta. Recent primaries formed clusters of craters from atmospheric fragmentation of the meteoroid body. Secondary craters have ejecta blankets with shorter rays that are consistent with emplacement by low-impact velocities (near 1 km/s). The mean extent of the continuous ejecta blankets was less distal for secondaries (5.38 ± 1.57R) versus primaries (18.07 ± 7.01R), though primary ejecta were less fractal (Fractal Dimension Index (FDI) < 1.30) and more circular on average (Circularity Ratio (CR) = 0.55 ± 0.25 versus 0.27 ± 0.13 for secondaries). Crater rims were remarkably circular (primaries CR = 0.97 ± 0.02, secondaries at 0.94 ± 0.05), though secondaries have the lowest values (CR < 0.9). Secondary crater rims were elongated toward or orthogonal to their primary of origin. Uprange source directions for most secondaries, determined by ejecta planform and crater rim ellipticity, point toward Zunil, although contamination from other primaries is considered in some areas. Ejecta blanket discrepancies between recent primaries and Zunil secondaries are attributable to differences in impact velocity and retention age. After removal of the ejecta blanket, crater rims are generally not diagnostic for determining crater origin. Fragmentation of primaries may play some role in steepening the size-frequency distribution of crater diameters in the 5 m < D < 30 m range.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Understanding the geologic evolution of individual terrestrial planets in our solar system is limited by the absence of absolute chronological ages from their surfaces. This is due to a lack of samples returned from them, with the exception of the Earth and a spatially limited collection from the Moon. One widely used dating methodology is the measurement of the size-frequency distribution of impact craters. A planetary surface age is derived by comparing the crater distribution to an extrapolation of the impactor flux over time for that planet [e.g., Hartmann, 1999]. Our solar system meteoroid flux has decreased in object size and frequency over geologic time [Hartmann and Neukum, 2001; Bottke et al., 2005], though a continuous bombardment streams onto planetary surfaces daily. Over time, this geologic process builds up a record of impact history that can be used to date geologic surfaces in the solar system [Shoemaker et al., 1963; McGill, 1977]. Crater-counting measurements have established a production function and time-dependent cratering rate [Hartmann, 2002] on the Moon that combined with the absolute chronology of impact events from lunar samples [e.g., Turner, 1970; Schaeffer et al., 1970] can extrapolate absolute dating to derive surface ages on other planets [Neukum et al., 2001; Ivanov, 2001], including Mars [Hartmann and Neukum, 2001]. However, utilizing this dating technique with <1 km in diameter (D) craters has come under intense scrutiny on Mars [Bierhaus et al., 2005; McEwen et al., 2005], as well as Europa [Bierhaus et al., 2005; McEwen and Bierhaus, 2006]. The argument against using small (D < 1 km) craters for dating is that large (D = several kilometers) primary impacts may produce ∼109 distant small secondary craters of D ≥ 10 m [McEwen et al., 2005] that, over time, can be indistinguishable from primary impacts in the same size range, as was recognized during the first intensive studies of impacts on the Moon [Shoemaker, 1965]. Thus, the “steep branch” of the production function, where more craters are expected for smaller crater diameters [Hartmann, 2005], has been said to occur because of the numerous secondaries that increase the size-frequency distribution in this size range [McEwen et al., 2005; McEwen and Bierhaus, 2006]. The crossover diameter, when the number of primaries greater than diameter D exceeds the number of secondary craters, indicates most craters with D ≥ 12 m should be primaries, yet observations near the D = 10.1-km Martian primary Zunil show most craters in this size range to be obvious secondaries [McEwen et al., 2005]. Direct measurements of the impactor flux for small craters on Mars have been measured [Malin et al., 2006] that are approximately in agreement with crater-dating isochrons. However, the amount of dating error introduced by secondary impacts on crater-based dating continues to be debated [Bierhaus et al., 2005; McEwen and Bierhaus, 2006; Quantin et al., 2007; Hartmann, 2007; Hartmann et al., 2008, Werner et al., 2009]. It is clear that being able to distinguish secondary craters from primary craters in the small crater record would be useful in either affirming or adjusting dating on many planetary bodies that have not had samples returned for absolute dating. Furthermore, a better understanding of the primary to secondary ratio in the D < 1 km population can yield better insight into the current small crater impactor flux, the crossover diameter for secondary cratering dominance in the geologic record, and the validity of crater chronologies calculated from subkilometer diameter craters.
 The latest impacts are recognizable by their prominent rayed ejecta extending several crater radii beyond their continuous ejecta blankets (Figure 1). Malin et al.  have identified 20 rayed craters believed to be recent impacts, on the basis of their appearance after 1999 and before 2006 in the dusty regions of Amazonis, Tharsis, and Arabia. These small rayed craters (SRC) exhibit sharp crater rims and bright-rayed or dark-rayed ejecta extending tens of crater radii distal to the impact site. Many have an “air blast” region, assumed to be the result of local dust removed by a pressure wave during impact [Malin et al., 2006]. Other SRC, initially identified by Grier and Hartmann , show similar crater rim and ejecta morphology, except the rays tend to be shorter (Figure 2). These frequently bright-rayed SRC have been identified as secondaries from a larger primary, Zunil [McEwen et al., 2005]. Many fall in clusters within its large ray system, as observed in the thermal infrared [McEwen et al., 2005; Preblich et al., 2007]. Because of the quantity and asymmetric spatial distribution of these secondary SRC, it has been proposed that most SRC, and most craters less than 200 m in diameter, are secondaries that may skew dating based on crater counts in this size range [Bierhaus et al., 2005; McEwen and Bierhaus, 2006]. This “pollution” of the primary impact record is reinforced by nine other large-diameter (2 < D < 15 km) rayed primaries on the Martian surface [Tornabene et al., 2006; Tornabene and McEwen, 2008; A. S. McEwen, personal communication, 2009] that each may have contributed upwards of 106 to 108 or greater secondaries to the background crater production rate. On Mars, the crossover diameter for SRC on young surfaces (<10 Ma) is ∼60 m [McEwen et al., 2005].
 This research attempts to separate primary from secondary impacts in the SRC population on Mars on the basis of qualitative and quantitative observations of crater rim and ejecta planform. This is accomplished by creating a database of known primary and secondary SRC populations, digitizing their crater rim and continuous ejecta perimeters, and applying morphometric formulas to ascertain any process differences in simple crater formation and morphology.
1.1. SRC Primary and Secondary Populations
Malin et al.  discovered the first confirmed recent (≤6 Earth years) impacts onto Mars by comparing Mars Orbiter Camera (MOC) Wide-Angle (MOCWA) images at ∼230 m/pixel from mission start (1999) to new MOCWA imagery in 2006. By looking exclusively in the dusty regions of Amazonis Planitia, Tharsis Montes, and Arabia Terra, a “difference map” was created to identify dark or bright spots indicative of new impacts. Subsequent Roll-Only Targeted Observation (ROTO) and compensated Pitch and Roll Targeted Observations (cPROTO) MOC Narrow-Angle (MOCNA) images with 0.5 m/pixel along-track and 1.5 m/pixel cross-track confirmed initial assessments of 20 new SRC with crater diameters from 2 m to 148 m (Figure 3). Except for one 148 m diameter crater, the largest crater diameter observed for each primary impact was between 10 to 37 m, averaging ∼20 m. All new primary SRC, except one, exhibit extensive air blast regions surrounding their main crater (Figure 1). Several primaries also contain multiple craters in a clustered pattern indicating breakup during atmospheric entry before impact (Figure 4) .
 Zunil secondaries in the Elysium Planitia region occur over a circular area of approximately 1700 km in radius from Zunil [Preblich et al., 2007, Figure 10]. The most distant Zunil rays (and secondaries) lie west of Zunil due to its moderately shallow 30–60° ENE impact angle [McEwen et al., 2005]. Bedrock in this area is composed primarily of two Amazonian aged formations; volcanic assemblages of the Elysium Formation composed of lava flows of varying superposition and ridged plain units [Tanaka et al., 2005] and bounded to the south by the Medusae Fossae Formation (MFF), considered a young highly erodible [Preblich et al., 2007] pyroclastic deposit [Tanaka et al., 2005]. Though evidence strongly supports a pyroclastic origin for the MFF, the exact formation mechanism is still debated [Watters et al., 2007]. SRC in this study are concentrated in the young Amazonian lava plains west and north of Zunil, centered on Cerberus Fossae and Athabasca Valles (Figure 6). Athabasca Valles contains streamlined features indicative of catastrophic aqueous flooding [Burr et al., 2002], subsequently overlain by recent volcanic flood lavas from Cerberus Fossae [Berman and Hartmann, 2002; Jaeger et al., 2007] and other vents in the vicinity [Plescia, 1990], generating rootless volcanic cones indicative of lava and ground ice interaction [Mouginis-Mark, 1985; Lanagan et al., 2001; Fagents and Thordarson, 2007; Jaeger et al., 2007].
 The small primaries from Malin et al.  fall into two geologic regions: Arabia Terra (AR), a highly cratered and dissected Noachian plain and Amazonis Planitia/Tharsis Montes (AT), dominated by the smooth plains of Amazonis that are composed of the MFF and the numerous lava flows and wrinkle ridges from the Tharsis volcanoes to the east (Figure 3). The AT region appears to be most similar in geologic terms and potentially in physical properties to the target material impacted by secondaries in Elysium Planitia. Together, the AT and Elysium regions form a “corridor” [Dohm et al., 2008] hypothesized to be an area of recent volcanic and hydrologic activity anywhere from <1 Ga [Plescia, 2003] to ∼10 Ma (lava flows dated by Hartmann and Berman  and Hartmann and Neukum  with fluvial episodes dated by Burr et al.  and Berman and Hartmann ).
 Databases were constructed from MOCNA ∼6 m/pixel and Mars Reconnaissance Orbiter (MRO) High-Resolution Imaging Science Experiment (HiRISE) [McEwen et al., 2007] ∼0.25 m/pixel images for examining the primary and secondary populations. SRC images were selected along an approximate downrange transect extending west and at several azimuths relative to Zunil to assess any change in crater morphology with increasing distance or orientation from the launch origin. For each SRC, the crater rim and ejecta blanket planform were manually vectorized into a geographic information system (GIS). For SRC in MOCNA images, the crater rim was digitized on the basis of a hypothetical circular rim using the crater centroid and one point on the rim that defined an approximate average crater diameter. Using this two-point method is not expected to deviate significantly from a more accurate three-point method [Hopp, 1994]. In HiRISE images, the actual crater rim was digitized at a constant scale of 1:500. The perimeter around the primary ejecta blanket planform was also digitized at its distal extent, at a constant scale of 1:5000 for MOCNA and 1:2000 for HiRISE images.
 All primary SRC identified by Malin et al.  were examined as a “type” population for primary SRC on Mars. MOCNA ROTO images for each crater were radiometrically and geometrically processed in ISIS (version 3) to a Mercator projection with a latitude of origin equal to the image centroid latitude. Reference numbers, herein assigned as identification numbers (ID) in the auxiliary material (see Data Sets S1 and S2), as designated by Malin et al.  (e.g., “Impact Site 8”) in their supplementary catalog, are maintained for continuity. Primaries with multiple craters are labeled with the original ID number followed by a dashed reference number (e.g., 8-1, 8-2, etc.). HiRISE images were also examined for 16 of the primaries (see Data Set S1). Calibrated and projected HiRISE images were downloaded directly from the HiRISE instrument website (http://HiRISE.lpl.arizona.edu).
 Building on work from Preblich et al. , 49 SRC were examined from MOCNA images identified as crossing Zunil rays, including 237 SRC from HiRISE images we identified as near or crossing the same ray system, giving a high probability that these SRC are indeed Zunil secondary cratering events. Since most MOCNA images have 2 to 5 times the pixel size of ROTO images, secondary craters with diameters larger by a factor of 2 to 3 (tens of meters) were chosen to best resolve interior and exterior crater features, while remaining well within an order of magnitude diameter size range (<200 m) compared to the primary population. SRC identified in HiRISE images were resolution-limited only to craters of D < ∼2–5 m, though a lower limit of ∼D = 10 m was used for delineating ejecta. Sufficient image resolution allowed discrimination of the primary ejecta blanket distal edge, as well as crater rim and floor morphology for the secondary population. A secondary SRC was only included in our study when the crater rim and ejecta planform were readily resolved; therefore, our counts do not represent the total number of secondaries per image. SRC were also excluded when crater rims or ejecta planform were modified by subsequent or simultaneous impacts, again lowering the total number of SRC cataloged per image. Each secondary was assigned a unique number for its crater and corresponding ejecta, independent of the primary ID numbers (see Data Sets S3 and S4).
 Ejecta planforms were evaluated by measuring their radial extent relative to the crater center, planform morphology via three area-perimeter formulae, and downrange orientation (Figure 7). Minimum, maximum, and mean ejecta ranges, normalized by individual mean crater radii, were calculated from the digitized planform perimeter and crater rim centroid. Ejecta blanket symmetry for each of the impacts was evaluated with three equations:
the Circularity Ratio (CR) [Selkirk, 1982], Form Ratio (FR) [Selkirk, 1982], and Fractal Dimension Index (FDI) [McGarigal and Marks, 1994], respectively. In the formulas, A is the ejecta area (including the area inside the crater rim), P is the ejecta perimeter, and L is the maximum distance between any two points along the ejecta blanket perimeter. Both ratios attempt to measure “compactness,” an approximation of the efficiency of a feature's areal distribution [Selkirk, 1982]. Long sinuous areas will have values closer to 0 while more circular “compact” features will approach 1. CR is also reciprocal to ejecta lobateness, Γ:
used to measure the sinuosity of single, double, and multiple lobe Martian rampart ejecta deposits [Kargel, 1986; Barlow, 1994]. FDI measures the spatial complexity of a natural planar object [McGarigal and Marks, 1994] and will also be used as a proxy for asymmetries in the ejecta blanket; a symmetric ejecta blanket will have FDI = ∼1 (e.g., circular), while more chaotic, fractal ejecta would have FDI > 1. Since oblique impacts generate more ejecta downrange from the impact azimuth and uprange “forbidden zones” [Gault and Wedekind, 1978], and secondaries from Zunil are theoretically impacting at oblique angles on the basis of experimental data [Cintala et al., 1999, Anderson et al., 2003], one would expect secondaries to have similar ejecta planforms to other oblique impacts. We estimated the SRC uprange azimuth in degrees clockwise from north on the basis of the criteria of downrange ejecta and uprange “forbidden zones” to create rose diagrams that, in theory point toward the primary crater that the secondaries originated from. Effects of planetary rotation (i.e., Coriolis force) were ignored since modeling has shown that only ejecta with launch velocities of 3–5 km/s, at high latitude, and ballistic distances beyond a quarter hemisphere were greatly affected [Dobrovolskis, 1981, Wrobel and Schultz, 2004]; none of these criteria are met by this secondary population (to be discussed in section 5.1). Initial launch velocities calculated via distances from Zunil would represent overestimates because of the counterclockwise (relative to the north pole) rotation of Mars. Atmospheric effects were also disregarded as minor or beyond the scope of this study. We estimate that the uprange azimuth is accurate to within ± 15°.
 Crater rims were also investigated for circularity (similar to measurements of lunar crater rims by Murray and Guest ) using the same formulae as for the ejecta planform. Long-axis crater orientation (i.e., crater rim ellipticity) was estimated by calculating the standard deviation ellipse (SDE) from the rim planform. A standard deviation ellipse calculates the standard deviation of the x-y coordinates from the mean center of a set of points to define the two major axes of an ellipse oriented along a spatial trend [Environmental Systems Research Institute, 2006]. The vertices of the digitized crater rim were used as a point set to calculate the SDE to 1 standard deviation (covering ∼68% of the points), hence the long-axis orientation of the crater (Figure 8). We estimate that the orientation derived from this method is good to within ± 10° of the true crater long-axis azimuth. Similar to the ejecta, we plotted the long-axis orientation, from 0° ≤ θ ≤ 180°, in a rose diagram to evaluate secondary crater ellipticity in relation to the suspected source primary. Two MOCNA images, E1101849 and M401791, were visually estimated for secondary crater long-axis orientation to compare with the SDE results. To remove any directional bias, crater long-axis values were divided by 2 and placed in opposing bins (i.e., if three craters fell into the east (90° ± 22.5°) bin, the data were split evenly into both the due east and west (270° ± 22.5°) bins). Erring on the conservative side, each ejecta or rim azimuth was plotted into 45° bins to eliminate systematic error or bias in either the SDE or uprange/downrange estimate. Azimuths to six large primaries from Tornabene et al.  and Tornabene and McEwen  (Corinto, Dilly, Naryn, Tomini, Thila, and Zunil) were estimated from polar map projections centered on each individual crater and placed on the rose diagrams for evaluation of the likely primary of origin.
 The geologic target is a significant factor during the excavation stage [Wünnemann et al., 2006; Collins and Wünnemann, 2007], and its role is examined against the above measurements to make comparisons between the primary and secondary impacts. Several primary SRC were removed from our analysis because of ephemeral ejecta blankets that did not allow delineation from the larger air blast region. Some subjectivity is introduced by manual ejecta blanket digitization, so any results should be considered first-order measurements, especially in regard to the highly diffuse primary crater ejecta. The secondary SRC, with their sharp contrast between bright ejecta and darker background, allow more concise and reliable measurements.
3.1. Primary SRC
 The most distinguishing features of the primary SRC population are the air blast region, long distal rays, and evidence of impactor fragmentation. Each primary, except for primary 17 (identification numbers in Data Set S1), displays the results of a suspected downward/lateral atmospheric pressure shockwave, i.e., air blast [Malin et al., 2006], that disturbed dust-sized surface particles beyond the continuous ejecta blanket at distances >100 crater radii (e.g., Figure 9). The affected region ranges from somewhat circular about the main crater, as in Figure 9, to a distal buffered area around the continuous ejecta, as in Figure 1. In most cases, the boundary between air blast and ejecta is diffuse and not readily separated. Half of the primaries have continuous and discontinuous rays extending tens to >100 crater radii past the continuous ejecta blanket (Figure 10). These rayed primaries appear to have downrange ejecta and uprange forbidden zones, even in the air blast planform (e.g., Figure 10b), indicating nonvertical impact angles. Given that half of incoming meteoroids enter the atmosphere at less than 45° from horizontal [Shoemaker, 1962], this is expected. On the basis of MOCNA ROTO imagery, resultant craters average D = ∼28 m or even lower to ∼17 m if we exclude primary 17 at D = 148 m, the only D > 30 m crater in this population. In HiRISE images, eight primaries (40% of this population) show evidence of fragmentation with multiple sharp-rimmed, though not rocky, craters in the central area of continuous ejecta (see Data Set S1 and Figure 11). Primaries 3, 7, 8, 13, 15, and 19, all located in Arabia Terra (AR in Data Set S1), have little to no discernable ejecta or rays (except 13) at the image scale (Figure 12). A boulder-strewn crater rim is visible only in primary 17, the largest among them (Figure 13). This nested crater has perhaps 103 to 104 2–5-m boulders within 3 to 4 crater radii. Given that these primaries are only excavating a few meters below the surface, this depth may be insufficient to reach bedrock in many areas on Mars. However, at least ten of the primaries have bright ejecta over the darker “blast” scoured surface indicating interception of a brighter subsurface layer (e.g., Figure 10b). Six primaries (3, 6, 7, 8, 13, and 19) appear to be almost vertical impacts based on their circular air blast regions, though many of these have little to no ejecta to confirm this (Figure 11). All “no-ejecta” primaries cluster in the upper member of the Arcadia formation, a relatively young Amazonian deposit hypothesized to be made up of pyroclastic and flow material from the Tharsis region [Tanaka et al., 2005] with most craters well below 50 km in diameter based on measurements from gridded Mars Orbiter Laser Altimeter (MOLA) elevation data [Smith et al., 1999]. Primary 13 could also be considered part of this “no-ejecta” group, but a faint ejecta layer is discernable (Figure 12). Primaries 1, 6, and 9 (e.g., Figure 9) do have ejecta in the same Arcadia formation, but at a lower elevation to the north (see Figure 3), categorized as a volcanic unit [Tanaka et al., 2005]. Hypothetically, this unit may be an older emplaced lava flow that is more indurated than the upper units or perhaps the regolith is thinner here causing more spall from a “hard” target [Head et al., 2002].
 Of particular note is the “crater field” created by primary 20 (Figures 3 and 14). The two largest craters in the field have D = 16 m and D = 13 m (Figure 14a). If we look closely at the crater floor of the D = 13 m crater, rather than the larger crater because of insufficient lighting, the floor appears somewhat flat and irregular with several blocks or slumps of material, though less than five or so D < 1 m blocks reside within 1 crater radii outside the rim (Figure 14b). Given that this is a flat-floored crater, we use a simple equation to calculate the crater depth, d:
where L equals the horizontal shadow length and θ equals the solar incidence angle for the image [Chappelow and Sharpton, 2002]. With L = ∼3.75 m and θ = 58, the depth (d) is ∼2.3 m with a depth/diameter (d/D) ratio equal to ∼0.18; slightly lower than the expected value of ∼0.20 for small primaries on Mars [Pike, 1980; Pike and Davis, 1984]. Tens of craters in the D = 5–10-m range and perhaps hundreds more with D < 5 m can also be seen within the continuous ejecta blanket (Figure 14a). Within 200 m of the main ejecta body are tens of 1 m < D < 5 m dark-rayed craters. Dark rays from these meter-sized craters form a “V” in their planform, assumed to point uprange toward their impact entry azimuth (Figures 14b and 14c). The V-shaped ejecta all point approximately north, regardless of whether they are north or south of the main crater field. This indicates to us that these are not secondaries, but actually primaries from the same body, albeit more dispersed. On the basis of this observation, the original impactor appears to have fragmented and/or dispersed over a minimum of 1 km crossrange and 1.3 km downrange. One cluster of three dark V-shaped rayed craters with 1 m < D < 3 m occurs ∼1.7 km north and ∼1.1 km west of the main crater field (Figure 14d) indicating that the crater field may extend over 1.5 km crossrange and 2.5 km downrange.
3.2. Secondary SRC
 Secondaries in this population appear nearly uniform in their planform morphology. The predominate type consists of an optically bright, continuous primary ejecta blanket with several rays of varying length and width (Figure 15a). A dark annulus of ejecta is frequently found within one crater radii (Figure 15b), suggesting excavation of a two-layer target. Crater rims range from irregular (western crater, Figure 7) to nearly circular (eastern crater, Figure 7). Meter-sized boulders are visible near the rim, as might be expected for a secondary cratering event [Bart and Melosh, 2007], but their frequency varies from crater to crater (e.g., Figure 8). Crater floors are typically dark, though bright fill in the form of aeolian emplaced dunes is not unusual, often superimposed above the dark floor material that appears to represent the in situ regolith at depth (Figure 15c). Aeolian reworking of bright or dark crater floor material is universal. The continuous ejecta blanket is composed of several lobes or rays that in many cases appear to point downrange with an uprange forbidden zone in relation to their impact azimuth (e.g., Figures 7 and 15d). However, the ejecta do not appear as axial symmetric downrange and often display a chaotic border, “ramparting” as evident from cast shadows, and a sharp ejecta perimeter (e.g., Figures 15a and 15d). Signs of filamentary ray or ejecta structure beyond the primary ejecta blanket are nonexistent, though some discontinuous ejecta members do occur (white arrow, Figure 15d), but at a limited extent (one or two crater radii) past the continuous ejecta. Unlike the primaries, no atmospheric air blast region can be discerned down to the submeter level. Underlying topographic structures are still visible beneath the ejecta, signifying a thin ejecta layer, at least to the edge of submeter image resolution (Figure 15e). At meter scale, many secondaries look pristine, though submeter pixel sizes reveal wind scalloping and faceting of thicker ejecta material (Figure 15d, inset). This ventifact-like erosion gives the impression that the ejecta are indurated (e.g., Figures 15a and 15d). Further wind modification can be seen in some cases, but appears to be more depositional versus erosional (Figure 16).
 Another secondary type consists of a nearly “sans-ejecta” crater rim with only a hint of the past excavated material (usually the dark annulus), but sharing crater rim and floor morphology (Figure 15e). Since these two types are almost always mutually exclusive within one image, it would suggest differences in target material at the separate impact sites. Using the same methodology as in section 3.1, we measured the depth/diameter ratio of one secondary crater (ID 279; see Data Set S2) having a D = 106 m, L = ∼41.5 m, and θ = 57, yielding a depth (d) of ∼27 m with a d/D ratio equal to ∼0.25, well above the values for secondaries obtained by McEwen et al. . We are fairly certain these SRC are secondaries as they fall along a linear trend pointing back toward Zunil in image PSP_002806_1870_RED.
 We measured crater rims produced by 16 primary (see Data Set S1) and 286 secondary impacts (see Data Set S3 and Figure 17). To reiterate, these counts do not represent every secondary (or number of craters per primary) in every image, but a select population that allowed crater rim and ejecta planform to be measured; no size/frequency distribution is implied. Actual secondary densities for some MOCNA images used in this study can be found in the work of Preblich et al. . From these populations, 16 primary (see Data Set S2) and 197 secondary (see Data Set S4) continuous ejecta blankets were delineated for analysis. Some secondary craters had diffuse or little ejecta; thus the total number of secondary craters does not match the count of ejecta blankets. In cases where primaries have multiple impacts, the ejecta planform data are in Data Set S2, but not used in comparison with the secondaries to normalize the data to one crater per ejecta planform. Measurements were rounded to whole meters to be conservative in our results; therefore all numbers in the tables are ± 1 m. Crater diameter, calculated as twice the mean rim radius, may appear 1 m larger or smaller than expected because of the measurement rounding. Primary craters with overlapping or shared rims (e.g., ID 7; Figure 12b) were excluded from analysis, but listed in Data Set S1.
4.1. Crater Measurements
 The Circularity and Form Ratios calculated for crater rims were plotted versus crater diameter for the primary and secondary populations (Figure 18). Primary and secondary crater rims averaged CR = 0.97 ± 0.02 (1 standard deviation) and 0.94 ± 0.05, respectively. From these values alone, one could only say that a crater rim with a CR ≤ ∼0.90, hence more elliptical, is a candidate secondary, though 87% of secondary crater rims fell above this value. The Form Ratio yielded even less differentiation among the crater rims with primary averaging FR = 0.83 ± 0.10 and secondaries 0.84 ± 0.07. No obvious trend was noted in either ratio. The Fractal Dimension Index measured a distinct exponential trend from the primary to secondary population. While the FDI for primary rims was 1.70 ± 0.24 and 1.38 ± 0.06 for secondaries, an FDI versus crater diameter plot showed evidence of a power law relationship (Figure 19a). If the data are plotted in log-log space (Figure 19a), a trend line interpolated from the data gives us three relationships:
where D equals crater diameter in meters and r2 is the correlation coefficient. The strong correlation of the index from small to large craters is somewhat verified by the one D = 160 m crater (ID 17; see Data Set S1) that has a diameter larger than any secondary measured in our study (Figure 19). It is possible that the FDI values are only showing a scale effect with less “detail” of rim features being captured for smaller craters, yet in the 10 m < D < 25 m range there is strong agreement where the two populations overlap. Since these craters are well within the strength regime for Martian craters, it is reasonable that smaller diameter craters (i.e., less energy at impact) are more irregular (i.e., “fractal”) than larger diameter craters that required more energy to form and thus were able to overcome the inherent strength of the target. Target type did not appear to significantly affect the data trend as the primaries occurred over a heterogeneous range of possible targets compared to the more homogenous Elysium Planitia region of the secondaries. The seeming uniformity of target strength may indicate that all of these craters are regolith-based and never penetrated deep enough to reach the basement rock in these areas. The one exception may be the largest primary (ID 17) with its nested depression in the crater floor (Figure 13), though it too plotted on the secondary FDI trend.
4.2. Ejecta Measurements
 The Circularity and Form Ratios calculated for ejecta planforms were plotted versus crater diameter for the primary and secondary populations (Figure 20). Primary and secondary ejecta averaged CR = 0.55 ± 0.25 (1 standard deviation) and 0.27 ± 0.13, respectively. The considerable range for the primary ejecta CR values reflects the difficulty in separating the continuous ejecta versus air blast areas. If one counted the air blast region as part of the ejecta, the primaries would potentially have higher CR values, though even the air blast can be chaotic (e.g., Figure 1). One might conclude that SRC with CR > ∼0.60 were candidate primaries; however, there is too much variation in this data set to draw any firm conclusions. The Form Ratio was even more muddled with primaries averaging FR = 0.51 ± 0.14 and secondaries 0.39 ± 0.11. Again, qualifying the ejecta planform has reduced the usefulness of this statistic with the primary population. A clearer differentiation can again be found with the FDI (Figure 21). The FDI for primary ejecta was 1.29 ± 0.05 and 1.40 ± 0.06 for secondaries. Each population clusters above and below the 1.30 line for most size ranges. Secondary ejecta appear to have a weakly correlated negative sloping trend with no D < ∼40 m craters below 1.30, while primary ejecta cluster below this point for most of the craters. Ignoring scaling effects and some arbitrariness to selecting ejecta boundaries, FDI < ∼1.30 may represent candidate primaries to the first order (Figure 22). Even accounting for different geologic terrains, the six Amazonis/Tharsis (AT in Data Set S2) primaries with mean FDI = 1.29 ± 0.04 were nearly identical to the four in Arabia Planitia (AR) with a mean FDI = 1.29 ± 0.06. From these results, one may deduce that the secondary population has more “fractal,” irregular ejecta planforms than primaries in the same size range. In the case of a low angle (<15°) primary impact, ejecta asymmetries may be indistinguishable when compared to secondaries, as appears to be the case for some primaries in this study (e.g., primary ID 20 in Figure 22).
 A significant difference was found between minimum, mean, and maximum continuous ejecta planform range from the crater center (Figure 22). Normalized by crater radius (R), secondary SRC were a factor of 3 to 4 times lower in ejecta range. Secondary ejecta were nearly uniform in distribution from crater to crater with a mean value of 5.4 ± 1.6R (1 standard deviation). The primary ejecta distributed more broadly with a mean value of 18.0 ± 7.0R. The greater variance among the primary ejecta range data can be partially explained by the ambiguous continuous ejecta perimeter; however, the continuous ejecta range for the primaries was several factors more distal than the extent of any secondary (Figure 10) suggesting the average values are not off by a large factor. Impact theory does not offer a reasonable explanation for this disparity unless the target properties (e.g., granite versus pumice targets in the work of Gault and Wedekind ) or impact velocities [Hartmann, 1985] are dissimilar.
4.3. Directional Analysis
 We plotted both crater rim long-axis orientation and ejecta uprange azimuth to determine whether secondary crater or ejecta planform could be used as a proxy to “find” the primary they originated from and hence identify them as secondaries (Figures 23a and 23b). Our results indicate that the majority of ejecta blankets that we measured do point back toward Zunil, regardless of the distance or azimuth from this primary. We did have one image, HiRISE PSP_002820_1860_RED, where all secondary ejecta pointed uprange toward the primary Corinto (Figure 23a). Other images (e.g., 6801_1935 or 6788_1955 in Figure 23b) also appear to have non-Zunil secondaries “buried” in their secondary populations. Preblich et al.  noted that contamination of the Zunil secondary field with non-Zunil secondaries was possible; our research confirms this assertion.
 Crater long-axis data generally orient both toward (e.g., HiRISE PSP_006801_1935_RED) and orthogonal (e.g., secondaries in HiRISE PSP_006788_1935_RED) to Zunil (Figure 23b). In two images with “no-ejecta” craters, PSP_002806_1870_RED and PSP_003874_1815_RED, the directional analysis was either ambiguous (3874_1815) or contradictory when originating primaries occur parallel to the crater long-axis at opposite azimuths (2806_1870) (Figure 23a). On the basis of the north-south linear emplacement of all secondaries in image 3874_1815, it still remains unclear whether they originate from Zunil (29° clockwise from north) or Dilly (−23° clockwise), unlike other SRC that fall along a line directly to Zunil (Figure 24). In the absence of ejecta, the uprange azimuth based on crater planform alone may not be sufficient to determine origin.
5.1. Differences Between Primary and Secondary Ejecta Retention
 Ejecta preservation can be described in terms of formation, the unique properties of the target that form the ejecta, and retention, the depositional and erosive environment; the resulting preservation being a combination of both factors. Similar to the concept of optical maturity used on lunar crater rays [Lucey et al., 2000], some ejecta maintain a visible bright or dark tone because they have a different tone than the surface they overlay and retain their visible properties over time despite erosional processes (i.e., formation). Other ejecta are truly “young” because they have been recently emplaced and not weathered to the background tone (i.e., retention) [Hawke et al., 2004]. Unlike comminution and agglutinate formation from impact gardening on the lunar surface [Heiken et al., 1991], the current foremost mechanism of erosion on Mars is aeolian [Greeley et al., 2004], perhaps followed by periglacial activity (i.e., patterned ground, thermokarst, etc.) operating at latitudes poleward of ± 50° [Mangold et al., 2004] where abundant stable water ice exists near the surface in diffusive equilibrium with water vapor in the atmosphere [Mellon et al., 2004]. Given that the primaries in this study are equatorward of ± 30° latitude and the secondaries fall between 5°N and 15°N, the dominant retention process should be wind-related. However, during times of high obliquity, modeling has shown that ground ice would be stable near the equator [Mellon and Jakosky, 1995]. Evidence of polygonal ground [Page, 2007] and sorted stone circles [Balme et al., 2009] in the Elysium Planitia region offers the possibility that some additional erosion is derived from freeze-thaw action during the aforementioned periods. In regard to formation, the primaries impacted into a dusty desiccated regolith with nine in the Amazonis Planitia region dominated by volcanic features; an area geologically similar and near Elysium Planitia where the secondaries impacted. One would expect, given the low latitude and shallow depth of excavation for both primaries and secondaries, the formation properties should be roughly equivalent. This should leave only retention properties (i.e., erosion/deposition rates) to explain any differences in ejecta planform between the two populations. Experimental results have shown that a mere 1.5 × 10−4 g/cm2 deposition of ≤ 5 μm diameter dust is sufficient to reduce some visible wavelength reflectance by 70% [Wells et al., 1984]. Martian dust deposition, based on dust streak formation rates, has been estimated at ∼4 × 10−5 g/cm2/a [Aharonson et al., 2003], though this may be a maximum as the streaks fall in predominately low thermal inertia/high dust index areas. A dust layer 3 μm to 10 μm thick on both Mars Exploration Rovers (MER) caused a 37% to 42% reduction in reflectance on their dark solar panels over a mere 150 Mars sols [Kinch et al., 2007]. In contrast, aeolian erosion has been estimated as low as ∼0.03 nm/a on the Gusev plains (the closest site to the secondaries) to a high of 10 nm/a at Meridiani [Golombek et al., 2006]. Annual deflation rates (dust lifted by wind) are low in Elysium Planitia [Haberle et al., 2003], though dust deposition is also currently low [Ruff and Christensen, 2002] and perhaps stable for long time spans [Haberle et al., 2003]. Using the high end of the MER landing sites dust deposition and bedrock erosion estimates, a rough linear extrapolation equates to ∼10 m/Ma dust deposition to 0.01 m/Ma erosion. Given the temporal difference between the primaries and secondaries, one would expect dust deposition to exceed erosion in modifying ejecta on timescales <1 Ga. While we do not see visible evidence of meter-scale dust deposition, there are no signs that the ratio of deposition to erosion is <1, meaning net deposition in this section of Elysium dominates, even if only slightly.
 Our systematic and statistical examination has revealed several unique attributes that differentiate the ejecta planform of primary and secondary populations in our study. The majority of primaries observed have air blast regions surrounding their continuous ejecta blanket (Figure 9), are somewhat more circular (Figure 20a), with regular/less “fractal” ejecta planforms (Figure 21), have substantially longer ray systems (Figure 10) and more distal continuous ejecta relative to crater diameter (Figure 22). Primary continuous ejecta appear thin and in some cases ephemeral (Figure 12). To the contrary, secondary ejecta planforms have significantly shorter ray lengths (e.g., Figures 7 and 22) and more irregular/”fractal” planforms (Figure 21). Many secondary continuous ejecta appear to have thicker bright ejecta with ramparts at a sharp distal boundary (e.g., Figure 15a), and some display discontinuous ejecta “islands” (Figure 15). A subset of secondaries have little to no ejecta with only a brief 1–2 crater radii annulus of dark, perhaps blockier, ejecta, but thinner or nonexistent bright ejecta. The pitted and sometimes ventifact-like surface of some secondary ejecta retains the evidence of wind erosion (Figure 15d, inset), though it also appears indurated and able to resist current aeolian activity to some degree (Figure 16). While these qualitative and quantitative differences are potentially revealing as to the origin for these two populations, the time variable between the primaries and secondaries is significant. We know the primaries are currently < 10 years old [Malin et al., 2006] and the Zunil secondaries are ∼5 × 105 years on the basis of age estimates of Zunil [Kreslavsky, 2008]. The primaries also occur in regions of high dust cover [see Ruff and Christensen, 2002, Figure 14], indicating that the more ephemeral ejecta features are probably unique to impacts in dust-prone areas; further supported by ∼50 new impacts identified in similar dust-laden regions [Daubar and McEwen, 2009]. The original reflectance of ephemeral features such as the air blast and distal rays could be completely obscured by dust perhaps within a few centuries or readily in a few millennia, when they exist. Regardless, the presence of an air blast and/or distal rays (tens of crater radii) for craters with a diameter 5 m < D < 200 m is evidence of a ∼<102 year old primary, since there are no known secondaries in this diameter range produced within the last few centuries on Mars. Differences in the distal extent of continuous ejecta are diagnostic between the two populations; whether this is solely attributable to their age differences remains to be determined. The other indisputable difference is the notable thickness of some secondary ejecta that should, all things being equal except time, be thinner (i.e., more eroded) than ejecta from a primary of similar size. Hartmann  experimentally investigated the effects of impact velocity on several factors including ejecta extent and thickness. His results indicate that at impact velocities below 1100 m/s, less than ∼1% of impact energy goes into excavating ejecta material [Hartmann, 1985]. This has the concomitant result of reducing ejecta range (i.e., lower ejection velocities), thereby increasing ejecta thickness toward the crater rim. As impact velocities increase past 1100 m/s, energy input into excavation increases with more ejecta being launched distally (i.e., at higher velocities) resulting in a thinner, dispersed ejecta blanket overall. The majority of secondaries we measured landed between 300 km to 600 km away from Zunil, which can be achieved with an initial launch velocity of between ∼1050 m/s to ∼1450 m/s respectively, as calculated from a planetary ballistic range formula from Melosh  and ignoring atmospheric drag. Secondary crater range, and thus the launch velocity, toward the west of Zunil (where the majority of secondaries exist) is overestimated because of planetary rotation and perhaps impact-generated winds which would increase these values. Therefore, impact velocities are lower in the Athabasca Valles/Cerberus Fossae region than our estimates. Examining the radial distance of the continuous ejecta blanket, we find primaries with thin distal ejecta and secondaries with thick ejecta proximal to the crater rim (Figure 22) that appear to reflect the factor of ∼10 difference in impact velocities.
 Three images illustrate the effects of formation versus retention with SRC. The first shows a number of bright-rayed secondaries with one SRC with darker toned ejecta. We can quickly rule out formation as a factor, because the bright SRC are in close proximity around the darker SRC, they have roughly similar crater diameters, and all overlay the same terrain; a homogenous volcanic lava flow (Figure 25a). Since the target is horizontally uniform (and for all intents and purposes, vertically uniform as well) as far as can be observed, the difference in tonal quality of the ejecta must be due to retention. Looking more closely at one of the lava flow layers, we can see bright material deposited by erosion (mass wasting?) at their distal edge (Figure 25b). The deposits have a visually identical tone to the bright SRC, though would be younger if their deposition is ongoing. Upon closer examination, the dark SRC appears to be a thinner and less bright (though not truly “dark”) example of the bright secondaries with similar ejecta planform (though subdued) and irregular crater rim that appears to be structurally controlled by the linearity of adjacent target features (Figure 25c). The crater interior contains several wind-derived ripples of somewhat darker toned material than the background deposits interstitial to the knobby background (Figure 25c), while the bright secondary crater interiors hold bright material in equal tone to the ejecta. At the very least, we can conclude that the one “dark” SRC is either older or younger, but not contemporaneous, with the bright secondaries; otherwise we would expect a similar level of ejecta retention shared by the bright secondaries proximal to it. Given that the lava cliff face appears to yield bright material and the “dark” SRC has thinner ejecta as well as dark material within its crater rim, we posit this SRC must be older and likely a secondary based on its morphological similarity to the other secondaries.
 A second image, M0200581, gives a clear case of formation overriding retention properties. In this image (Figure 26), a lava front divides the image with older terrain (more craters) in the north half of the image and younger terrain (the lava flow with less craters) to the south. The terrain north of the young lava flow has several “no-ejecta” secondary craters with only a brief dark annulus (Figure 26a). At the transition from the older basement to the younger embaying lava flow, one “no-ejecta” crater can be seen on the stratigraphically lower unit (Figure 26b). However, secondaries on the young flow all have bright ejecta typical of many Zunil secondaries (Figure 26c). No observable bright secondaries appear in the northern half of this image. This leads us to conclude that the young lava flow unit is responsible for the bright ejecta, hence a product of formation. This leads us to consider that all SRC in this image are contemporaneous and likely a secondary cluster, though we cannot definitively rule out that the SRC in Figure 26a are from a different event.
 The final image, PSP_002661_1895_RED, displays several bright-rayed and dark-rayed craters (Figure 27). The dark SRC have a diffuse “star”-shaped ejecta planform with dark crater floors, while the bright-rayed craters have the “typical” sharp boundaries of Zunil secondaries with bright crater floors. Relative ray lengths for the bright and dark SRC are well within the maximums found for the secondary population (Figure 22). Another difference between them is the crater sizes; the dark SRC diameters range from ∼38 m < D < ∼60 m, while the bright SRC have D < ∼23. One exception occurs with SRC F that has a dark floor, but predominately bright continuous ejecta with D = ∼34 m. The disparity of ejecta brightness with crater diameter indicates a target with two distinct stratigraphic units. Since SRC F has bright ejecta, but a dark floor indicating it penetrated a lower dark layer yet it did not contribute the majority of the continuous ejecta, it must be near the boundary for the two units. On the basis of the standard depth-to-diameter (d/D) ratio of 0.2 for parabolic simple craters [Pike, 1980], the bright unit must transition between ∼4.6 m to ∼6.8 m below the surface to the dark unit. The bright ejecta from SRC E and F have a clear uprange forbidden zone while all the dark SRC ejecta have no such impact direction indicators; one would expect them to share a similar ejecta pattern if they represented a simultaneous impacting event. Some <10 m diameter bright SRC are coincident with some of the dark ejecta, and it appears that dark blocks override some of the bright ejecta (especially just north of SRC C). However, the image resolution does not entirely reveal if the blocks sit on top of the bright ejecta or are simply too big to be buried by it. In either case, the dark and bright SRC appear to be from separate events given their dissimilar ejecta planforms. We cannot determine whether the dark SRC are primaries or secondaries on the basis of their planforms alone.
5.2. Similarity of Primary and Secondary Crater Rims
 Our results show that, in terms of the crater rim, secondaries are remarkably similar in planform to primaries within 1 standard deviation (Figures 18 and 19). The majority of the secondary population displayed remarkably circular and “orderly” rims, making most of them indistinguishable from the primary population in this regard. This is in contrast to similar work done on the Moon where secondaries were significantly less circular (circularity = 0.54 versus 0.82 for primaries, [Pike and Wilhelms, 1978]). There is a definitive break at CR = 0.9; only secondaries had crater rims lower than this value. However, many secondary crater rims have similar long-axis orientations either parallel or orthogonal to the uprange azimuth of possible originating primaries (Figures 23a and 23b). In experimental data using single projectiles, crater rim circularity only becomes more elliptical downrange when the impact angle is ≤30° for “hard” targets and ≤10° in noncohesive sandy target [Gault and Wedekind, 1978]. For cohesive “dusty” targets, experiments with single projectiles at impact angles between 10–35° are elliptical in the cross-track dimension, oblique to the impact azimuth [Gault and Wedekind, 1978]. However, Schultz and Gault , using clustered projectiles, experimentally produced downrange elongated craters at angles as high as 60°. Experiments by Anderson et al.  showed a 30° angled primary impact produced secondary “blocks” ejected at (and assumed to land at) angles ≤40° during the first moments of high-velocity crater excavation flow. It remains unclear whether any inference to impact angles can be deduced from these crater planforms as we cannot ascertain whether each secondary crater formed from one impacting projectile or a dispersed cluster.
5.3. Identification of New Primary Cratering Events
 On the basis of our results, we have identified two rayed craters we believe are primary cratering events. The first is a dark-rayed crater in image MOCNA R0600296, located at approximately 150°E, 8.27°N (Figure 28a). It has a diameter D < ∼40 m, a dark ejecta signature close to the crater rim and several long, ephemeral dark rays with one extending ∼2.9 km or ∼72.5 crater radii, which fits well with our ejecta observations (Figure 10) and measurements (Figure 22) for primaries. There may be a faint air blast signature with a slight brighter intensity behind the rays and a darker signature just around the western edge of the feature that gives the appearance of a forbidden zone south/southwest of the crater, though it appears mostly faded at this time. Two bright-rayed secondaries may be overlain by rays ∼400 m and ∼620 m to the north/northwest of this crater, but the resolution of this image limits verification. This dark candidate primary has also been imaged in Thermal Emission Imaging System (THEMIS) visible (VIS) image V14186017 and V1603005, though no earlier image records this area other than this MOCNA image in 2003.
 A second candidate primary comes from image PSP_006801_1935 that has all of the diagnostic characteristics typical for definitive primaries in our study (Figure 28b). Four craters make up this cluster; the main crater at D = ∼6 m and three “child” craters at D = ∼3 m, making this the smallest primary yet reported for Mars. Since the meteoroid that formed these craters did fragment, it was likely a stony meteorite though the small crater diameters could also suggest an iron [Popova et al., 2003; Chappelow and Sharpton, 2005]. The cluster spreads over ∼280 m east-west and 120 m north-south. Bright air blast regions surround each crater with a darker inner annulus of assumed ejecta material. Discernable ejecta appear ephemeral, unable to bury meter-sized to submeter-sized blocks, and look well formed. The crater rim looks almost hexagonal with a possible floor modification (nested crater?) (Figure 28b, inset). Image resolution limits further quantification (e.g., measuring the FDI of the ejecta), despite the 0.25-m pixel size. No discernable forbidden zone is noted, indicating a probable impact angle ≥60°. The most distal ray from the D = ∼6 m crater is ∼150 m or 25 crater radii which puts it on the low end for a primary, but at the high end for secondaries (Figure 22), although a nearby secondary (ID 164; see Data Set S4) has a maximum ray length <7 crater radii. Compared to nearby secondaries, it is clear that this cluster is morphologically unique.
5.4. Crossover Between Primary and Secondary Cratering Events
 The problems inherent with using D < 1 km craters for dating can be either ignored because both primaries and distant, nonclustered secondaries are counted together for generating isochron-based dates [Hartmann, 2007; Hartmann et al., 2008; Werner et al., 2009] or these problems must be identified because secondaries dominate the crater counts in some regions causing orders of magnitude error in surface ages estimates [McEwen et al., 2005; Bierhaus et al., 2005]. When rayed ejecta are still retained around a crater, we have qualitatively and quantitatively shown that recent primaries have unique attributes that distinguish them from older secondaries. Over time (107 to 109 years), classification of an individual crater as a primary or secondary using only the crater rim and floor, as are the majority of cases on Mars once ejecta have been eroded or buried, becomes increasingly difficult. Except for some outliers, secondary crater rims are as round as primaries. Even some craters that fell in obvious linear trends were not readily associated with their primary (Figure 24), although modeling efforts of Popova et al.  show that only secondaries would fall in this manner, while primaries cluster within a few hundred meters of each other because of low lateral velocities during breakup. We did not examine depth/diameter (d/D) akin to Pike and Wilhelms  as our goal was to evaluate the horizontal planform, and most areas lack stereo coverage with sufficient image resolution (<1 m/pixel) to resolve 2-m or 3-m vertical relief. However, for illustrative purposes, we did measure the d/D of one “flat-floored” primary (Figure 14b) and one secondary (Figure 15f) whose shadow extended to the approximate center of the crater floor. Simple shadow measurements showed the primary with d/D = ∼0.18 and the secondary with d/D = ∼0.25; both values near the expected Martian d/D for primaries at ∼0.20 [McEwen et al., 2005]. While a statistic of one is not definitive, and we agree that many secondaries we examined were qualitatively “flat-floored,” it does show that “bowl-shaped” craters are not strictly primary forms. It is also unclear what role aeolian deposition/erosion or crater relaxation from thermokarst processes (at higher latitudes) have on SRC d/D considering they excavate predominately in the upper few meters of the surface where such activity dominates. Since the primary population we examined appeared to excavate only into regolith (except the floor of the one D > 100 m crater), over time their crater floors may also become shallow, though some argue that d/D is not affected by infilling in craters of this size range [McEwen et al., 2005].
 Not only can secondaries “masquerade” as primaries, but the reverse can also be true. As is the case for primary ID 20 (Figure 14), this one primary produced hundreds of craters D < 10 m, indicative of breakup of a weak stony meteorite according to Popova et al. , as are likely the other fragmented primaries we have measured [Hartmann, 2007]. Unique to this primary field are the hundred of “microprimaries” (1 m < D < 5 m, perhaps even more D < 1 m) that extend crossrange ∼1.5 km by 2.5 km downrange. Popova et al. [2003, 2007] predicted crater clustering to remain within a few hundred meters, yet here the ranges are an order of magnitude farther. To reiterate, what happens once the ejecta disappear? If this primary field is buried and exhumed, could you tell that these craters were all primaries? In the target surface are numerous craters in the same diameter range (D < 10 m) that, on the basis of the overlying dust layer, existed before the impact. Could you distinguish the new primary population from this one impact event versus the background craters after a few millennia? We would argue that even in the case of primaries, the total number of craters (without ejecta) with diameter D < ∼200 m on a surface is skewed positive because of breakup. During high obliquity cycles, which occur every ∼106 years on Mars starting at 3 × 106 years or earlier [Bills, 1990; Mellon and Jakosky, 1995], atmospheric temperature would increase [Bills, 1990; Mellon and Jakosky, 1995; Laskar et al., 2004], and pressure ranges anywhere from 30 mb [James et al., 1992] to approaching 80 mb or more at >30° obliquity in “thick” atmospheric models [Manning et al., 2006]. Fragmentation would increase with such atmospheric pressures causing even larger meteorites to fragment yielding larger crater diameters [Popova et al., 2003; Chappelow and Sharpton, 2005]. Thus, secondaries, as well as primaries, may “inflate” crater counts over a diameter range. This result complicates estimating the crossover diameter as both secondaries and primaries produce crater fields that are currently indistinguishable in many characteristics compared to a background population, especially once ejecta have been removed. As a thought experiment, let us consider the total number of confirmed primary craters produced during the past ∼10 years (an underestimate if more impacts occurred during the past 3 years) when nineteen primaries impacted ∼15% of the surface area of Mars (excluding the anomalous D > 100 m primary). These nineteen primaries produced 81 craters 5 m < D < 30 m over that time period. Multiplying these values by 6.9 to estimate the global cratering rate yields ∼127 primaries and ∼543 craters/10 years within the aforementioned diameter range and timespan. During the period it would take to produce a Zunil (∼500 ka) [Kreslavsky, 2008] with ∼108 secondaries D ≥ 10 m, ∼6.36 × 106 primaries could impact creating 2.72 × 107 craters D < 30 m. This ignores two possible “peaks” in the atmospheric pressure over the past 0.5 Ma [James et al., 1992; Manning et al., 2006] that should cause more fragmentation of larger meteorite bodies [Chappelow and Sharpton, 2005] increasing the primary count. Admittedly, there are also “troughs” where the atmosphere “collapses” into CO2 ice caps at both poles [Kreslavsky and Head, 2005], decreasing the fragmentation rate, while also reducing the current atmospheric filtering of smaller meteorites. It remains unclear whether the fragmentation-to-filtering ratio increases or decreases the net count of primaries permitted to the Martian surface during changes in Mars' atmospheric pressure.
 With these albeit simplistic calculations, if the number of primary craters produced was constant over 2 Ma, and no other >1-km diameter primaries occurred, the primary-to-secondary ratio of craters in the 5 m < D < 30 m diameter range across Mars could be close to parity. This would make sense as the secondary cratering size-frequency fits a Weibull distribution [Werner et al., 2009], resulting in a flattening of counts in this diameter range. Malin et al.  found the primaries to fall within an order of magnitude of their correct age on Hartmann  isochrons extrapolated to single digit years. Hartmann  also finds only an order of magnitude difference in isochron age when theoretical counts of primaries and secondaries over 10 and 100 million years ago are added together. While it does appear that the “steep branch” of the isochron graphs may well result from secondary contamination of crater counts, they do not appear to “crossover” beyond a small factor of the primary production rate [Hartmann et al., 2008; Werner et al., 2009], perhaps also because of primary fragmentation, even in the current thin (∼6 mb) Martian atmosphere. Not only are small secondaries “statistically clustered” in time [McEwen et al., 2005; Bierhaus et al., 2005; Hartmann, 2006], to some degree so are primaries. From these results, we estimate the primary to secondary ratio could approach parity (i.e., 1:1) for craters 5 m < D < 30 m.
 Understanding the ratio of primary-to-secondary crater formation on Mars, and all planetary bodies, is deemed critical for dating geologic surfaces and events, as well as defining the current primary crater production rate. Secondaries share many morphologic features with primary impacts, though some differences in the ejecta blanket planform and crater rim shape are quantifiable. We offer a matrix for resolving the similarities and differences between the primaries and secondaries in this study (Table 1). While these criteria are not absolute and better for a population-based estimate, they do provide some quantitative measure to evaluate SRC as to their origin; planetary or cosmic. Some caution must be taken as it is clear that target material and ejecta retention rates do play a factor in these results. It is also apparent that primary production rates based on crater counts alone are multifaceted not only by far-field secondaries [after Werner et al., 2009], but also by meteorite fragmentation.
Table 1. Criteria for Recognizing Primary Versus Secondary Craters on Mars
 We would like to thank Stephanie Werner, Alfred McEwen, and especially Nadine Barlow for their thorough and comprehensive reviews which brought the paper into focus. This work was partially supported by NASA MDAP grants to V.L.S. and R.R.H.