Radar Sounding of Open Basin Lakes on Mars
Abstract
Orbital observations of the surface of Mars indicate that ancient basins were once host to lakes that may have been depocenters of sedimentary materials, including hydrated minerals like phyllosilicates. Later volcanic and sedimentary resurfacing may have developed a stratigraphy within the basins detectable through radar sounding data from the Shallow Radar instrument. Our radar survey of 61 open basin lakes (OBLs) revealed only one plausible reflector in a single basin east of Hellas Planitia. We investigated possible factors contributing to lack of radar detections in the other basins through detailed analysis of radargrams and subsurface characteristics of OBLs. As possible hosts to previous habitable environments, OBLs on Mars are important sites for future robotic and human missions. A full understanding of the factors influencing the radar signal, as addressed here, is important for more fully characterizing the subsurface structure and geology at these locations.
Key Points
- A single reflector can be identified in four radar images in one of 61 open basin lakes
- A rough subsurface, volumetric scattering are likely more important attenuating factors to SHARAD than surface clutter, hydrated minerals
- Attenuation encountered must be addressed by future robotic and human missions investigating the subsurface in open basin lakes using radar
Plain Language Summary
Mars was once host to ancient lakes that can currently be observed as a topographic depression in the surface with channels where water is thought to have once flowed called an open basin lake. This flowing water transported sediments into the lake which was later covered by further sedimentary or volcanic deposits. The interface between layers of material in the subsurface can be detected from orbit using the Shallow Radar instrument. Our survey of 61 open basin lakes revealed only one detectable interface in a single basin lake. We investigate possible factors that could contribute to the lack of identified interfaces through detailed analyses of available radar images and surface characteristics of the open basin lakes. We develop a method to characterize the degree of noise (clutter) in the radar images created by large surface features (>100 m) up to 25 km away from the position of the ground track of the spacecraft. However, this alone cannot explain the lack of detected interfaces. A rough subsurface or interface may be more likely. Further modeling will be necessary to characterize the radar characteristics of these basins as many are landing site candidates for future missions searching for formerly habitable environments.
1 Introduction
Mars was once host to ancient lakes that may have been long lived during the Noachian and Early Hesperian. These “open basin lakes” (OBLs), classified by Fassett and Head (2008; FH2008), are currently observed as topographic depressions that have both inlet and outlet channels where surface water is thought to have once flowed. This fluvial activity transported sediment into the paleolakes, depositing the load as lake-bottom beds and deltaic features. A total of 224 OBLs were mapped throughout the southern highlands (Fassett & Head, 2008; Goudge, Head, Mustard, & Fassett, 2012). Many were resurfaced by Hesperian-aged volcanic material (Goudge, Head, Mustard, & Fassett, 2012) that could potentially overlay ancient lake deposits, creating a distinct subsurface stratigraphy that can be detected from orbit using radar sounding data from the Mars Reconnaissance Orbiter Shallow Radar (SHARAD) instrument. We conducted the first comprehensive survey of SHARAD radar observations over OBLs to assess their stratigraphic structure and their geologic properties affecting the radar signal. Many of these OBLs are of particular interest because of their candidacy as future landing sites for missions searching for formerly habitable environments.
SHARAD operates at a central frequency of 20 MHz with a vertical resolution of 15 m in free space that varies as the inverse square root of the real permittivity, ε, in different geologic materials. The horizontal footprint of SHARAD extends 3–6 km cross track and 0.3–1 km along track (Seu, Phillips, Biccari, et al., 2007). Empirical studies show that for radio frequencies, the real permittivity of dry material is primarily a function of its density (Campbell, 2002; Ulaby et al., 1988). Strong contrasts in permittivity within the subsurface will appear in the image (radargram) at a greater time delay than the initial, bright reflection off of the surface (Campbell et al., 2008; Seu, Phillips, Biccari, et al., 2007; Seu, Phillips, Alberti, et al., 2007). Reflectors can be obscured by cross-track surface echoes (clutter) that reach the receiver at a time delay comparable to that of the reflector. Such sources can be identified by comparing the radargram (Figure 1b) to simulations (Figure 1c) of surface clutter (cluttergrams) created from topography obtained by the Mars Orbiter Laser Altimeter (MOLA) instrument (Choudhary et al., 2016; Smith et al., 2001).
2 SHARAD Survey of OBLs
We conducted a survey of 937 SHARAD dayside and nightside radargrams for possible subsurface reflectors within 61 OBLs. We restricted our initial analysis to those OBLs with areas >1,000 km2, as smaller basins are more difficult to resolve due to the abundance of clutter from their steep walls within the SHARAD footprint. We used U.S. Team radargrams from the Planetary Data System (PDS) collected from December 2006 to September 2016. Each radargram was compared to its clutter simulation produced by the U.S. Team online processor, CO-SHARPS (Choudhary et al., 2016; Putzig et al., 2016), to confirm the presence of real subsurface reflectors. If a return was detected, the time delay of the reflector was determined in order to calculate depth to the interface and provide a constraint on the stratigraphy of the basin.
Despite the possibility of density variations at depth, we did not find convincing evidence of radar reflectors in all but four of the radargrams surveyed. This is surprising, given the dielectric contrasts that likely exist between surface volcanic plains and subsurface lake deposits in those OBLs that have been volcanically resurfaced (Goudge, Head, Mustard, & Fassett, 2012). However, most SHARAD observations in the southern highlands appear to lack subsurface reflections (Nunes et al., 2010; Stillman & Grimm, 2011) and our survey of OBLs is consistent with this trend.
A single basin (FH2008 Lake 218) located east of Hellas Planitia (102°E, −37°N) contains a single, plausible reflector. This reflector is observed in four radargrams at consistent time delays and is relatively low in power within a background of diffuse scattering (hereafter “rainfall effect”) that extends to high time delays. This scattering partially obscures the reflector, making accurate delineation of its extent difficult. A representative radargram (track number 1967301) and corresponding clutter simulation for this reflector is shown in Figure 2. The reflector is not present in the cluttergram (Figure 2b), which means that it is likely real. This reflector is observed at similar time delays in an additional three radargrams (see Figures S1–S4 in the supporting information). The track numbers for these radargrams are 663901, 592701, and 1679201. Previous mapping of the Hellas Planitia region by Leonard and Tanaka (2001) indicated that Hesperian lavas recently resurfaced Lake 218. Observations made by Rogers et al. (2018) suggest that throughout the highlands, bedrock plains are not sourced from lavas but are more commonly found to be of a pyroclastic or sedimentary origin. Goudge, Head, Mustard, and Fassett (2012) suggest that this resurfacing is glacial in origin, but further examination of several Mars Reconnaissance Orbiter Context Camera (CTX) images (Figure 3) shows that the glacial deposits are confined to small craters scattered throughout Lake 218. These images also show a distinct flow margin contacting the basin rim in addition to several wrinkle ridges located in the center of the basin supporting a volcanic origin for this resurfacing. Two channels crosscut this flow margin along the northeastern rim suggesting this volcanic resurfacing was followed by an additional, later period of fluvial activity that may have culminated in ponding after the initial formation of the lake by the valley networks in the Noachian. These channels appear similar to mapped drainage systems in the Navua Valles and Hadriacus Mons regions that empty into nearby Hellas Planitia (Hargitai et al., 2017). We interpret this reflector to be the contact between the resurfacing lavas at the near surface and some underlying sediments. If it is assumed that the volcanic materials extend to depth, and we assume a dielectric constant between 7 and 9 corresponding to Martian lava flows, the thickness of the deposit is ~93–105 m. If there are interlayered sediments or less dense materials present, lower dielectric constants between 4 and 6 can be assumed yielding a thickness of ~113–140 m.
A number of factors could contribute to lack of reflectors within the other OBLs. Campbell et al. (2013) describe clutter arising from topographic sources at two distinct scales in SHARAD radargrams. At larger scales (e.g., crater walls and ridges) clutter arising from facets facing the radar track is termed “deterministic clutter.” At smaller scales comparable to the SHARAD wavelength, clutter can be attributed to random roughness on the surface or rough subsurface interfaces that can attenuate the radar signal and is termed “statistical clutter.” Basins with high deterministic clutter, as modeled in the cluttergrams, can cause surface returns at time delays that obscure real reflectors if present at high enough density. Volumetric scattering within geologic units, which could arise from sources such as fractures or density heterogeneities at scales comparable to the SHARAD wavelength, can also attenuate the radar signal. Monolayers of water adhered to mineral grains, such as clays, have also been suggested to be a major source of radar attenuation in the Southern Highlands (Stillman & Grimm, 2011). It is also possible that no strong, continuous dielectric contrasts exist in the subsurface, resulting in the absence of radar reflections.
To better constrain the factors affecting detections of subsurface interfaces within OBLs, we conducted detailed studies of SHARAD radar data and surface characteristics within a subsample of OBLs (~10%, N = 26) chosen to be representative of the global population. The subsample incorporates basin sizes spanning the complete range of OBLs from ~50 km2 to the largest at 5 × 105 km2 (Eridania basin, FH2008 Lake 51), and over half of the basins (N = 15) have been resurfaced by volcanic material (Goudge, Head, Mustard, & Fassett, 2012). SHARAD data included 414 nightside radargrams from the PDS collected from March 2007 to February 2016.
For each basin, we developed methods for testing for the following factors: (1) deterministic and statistical clutter (roughness), (2) volumetric scattering, and (3) surface hydration. We applied these methods to each basin and discuss the methods and our observations in section 3. A full summary of our observations and calculated parameters used in the analysis are provided in Table S1 for the subsample of OBLs and in Table S2 for the global population of OBLs in the supporting information.
As examples, we also include detailed examinations of Antoniadi basin (FH2008 Lake 10) due to its diversity of radar characteristics and Gusev crater (FH2008 Lake 50) due to the availability of ground truth data from the Mars Exploration Rover Spirit that may help to interpret the radar observations. Antoniadi basin (61°E, 21°N) is located NW of Syrtis Major. This basin is large compared to the SHARAD footprint (D ~ 410 km) and exhibits a range of terrain types including smooth volcanic plains identified in previous mapping work from Greeley and Guest (1987) that suggested early Hesperian lavas from nearby Syrtis Major partially resurfaced Antoniadi along with other contributions from sources within the basin itself (Hiesinger & Head, 2004). Additionally, surface exposures of hydrated silica toward the eastern rim of Antoniadi in Figure 1a were identified by the Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) (Carter et al., 2013) and determined to have a stratigraphic relationship with the surrounding volcanic deposits indicating that formation occurred prior to the early Hesperian (Smith & Bandfield, 2012).
Gusev crater (175°E, 14°S; D ~ 160 km) is located near the edge of the southern highlands and is the landing site for the Mars Exploration Rover Spirit (Figure 4). Ma'adim Vallis, an 8- to 15-km-wide valley that traverses 900 km of the highlands (Irwin et al., 2002), joins Gusev on its southern rim. This inlet is interpreted to be the source for ponding water and sediments carried into the basin in the Noachian (Cabrol et al., 1996). Using the relationship between the depth and diameter for Martian complex craters of d = 0.2D0.53 (Garvin et al., 2000), Carter et al. (2001) calculated that Gusev has a depth of ~3 km unfilled compared to the current depth of 2.1 km in the center. They suggest that 900 m of sediment currently resides in the basin. Volcanic resurfacing in the Hesperian followed this period as evidenced by in situ observations from Spirit (Squyres et al., 2004). Detections by the Mössbauer spectrometer further confirm that rocks in the plains of Gusev are volcanic in origin (Morris et al., 2004). Although the source of the volcanic deposits is unknown, it is likely that they originated from within Gusev and are not from nearby Apollinaris Patera located downslope from the basin. The thickness of these volcanic plains remains very uncertain (Parker et al., 2010). Other notable observations made by Spirit include hydrated silica exposures also identified by CRISM (Carter et al., 2013; Figure 4) found at Home Plate in the Eastern Valley of Gusev interpreted to have formed in a hydrothermal environment within the basin (Squyres et al., 2008). The findings from the Spirit mission suggest that a subsurface stratigraphy exists within Gusev, making it an ideal target for radar sounding. However, as in our survey, Putzig et al. (2014) also found no subsurface reflectors in SHARAD radargrams. They suggest that the radar signal may be attenuated most effectively by older surfaces that have been subjected to significant alteration (water alteration or fracturing) as suggested by Stillman and Grimm (2011) or a lack of sufficient dielectric contrast between layers in the subsurface.
3 Geologic Controls on Detection of Subsurface Reflectors
3.1 Deterministic and Statistical Clutter
Since deterministic clutter is typically a visually identified characteristic of radargrams, we attempted to develop a new, simple quantitative proxy for predicting the degree of clutter in each radargram. Clutter is controlled by the presence of topographic facets larger than the radar wavelength (λ) that have slope aspects directed toward the spacecraft to produce off-nadir reflections (Choudhary et al., 2016; Holt et al., 2006). As a result, the degree of clutter will approximately increase with increasing topographic roughness measured at baselines >λ. As a clutter proxy, we used a directionally modified root-mean-square (RMS) height parameter. RMS height values (Shepard et al., 2001) were first calculated from MOLA gridded topography at 128 pixels per degree (PDS MOLA Mission Experiment Gridded Data Record product). The baseline used was one pixel, or ~463 m at the equator; clutter-contributing facets considered here are therefore approximately equal to this size. Facets whose slope aspect is perpendicular to the direction of the SHARAD track, or spacecraft, contribute most to reflections and clutter sources. SHARAD tracks analyzed here are oriented in an approximately north-south direction, with average bearings of ±6.8° from north. We therefore weighted the RMS height values by the absolute sine of slope aspect, determined from the MOLA MEGDR product. Doing so more greatly weights east or west facing slopes (slope aspects of 90° and 270° as measured clockwise from north); north and south facing slopes will result in parameter values of 0. It should be noted that the use of the 128 pixels per degree MOLA gridded topography may cause the RMS height parameter values to be biased low as the interpolation is mostly focused in the east and west directions (Som et al., 2008) resulting in an under sampling of these facets. However, the clutter simulations used in this study are derived using MOLA products (e.g., Choudhary et al., 2016) and so the RMS height parameter is a well-suited quantitative proxy for clutter in this case. Future investigations using this technique could improve this through the use of higher resolution topography (e.g., High Resolution Stereo Camera topography data).
For each portion of a SHARAD track crossing an OBL outline, we calculated an RMS height parameter value. The parameter was calculated as the average of weighted RMS heights falling within a buffer surrounding the SHARAD ground track (Figure 1a). To determine a suitable buffer size, we examined echo maps described by Choudhary et al. (2016) and available on CO-SHARPS (Putzig et al., 2016). These map first radar return locations and show contributions from clutter sources up to 45 km away from the ground track. Additional sources of cross-track surface echoes are also mapped that can contribute to clutter at much later time delays. Our qualitative assessment of these maps over OBLs show that clutter sources were often located close to the ground track at distances between 25 and 30 km. Based on these observations, a 25-km equidistant zone (buffer) was placed perpendicular to each SHARAD track for extraction of weighted RMS height values (Figure 1a). Further alteration of the buffer size or comparison between buffers of various sizes may yield a more refined result for future investigations of the OBLs or other regions of interest.
Statistical clutter was represented by calculating a SHARAD roughness parameter after the methods described in Campbell et al. (2013). The roughness parameter estimates the relative degree of topographic roughness at the 10 to 100 m scale and is therefore sensitive to smaller baselines than the RMS height parameter. Only nightside observations were used due to their higher signal-to-noise than dayside observations, which are affected by the ionosphere. While a roughness value was calculated for each column in the radargram, we report average values for each track segment within the OBL outline. SHARAD roughness parameter values typically range from 2 (smooth) to 8 (rough) (Campbell et al., 2013). Values anomalously higher or lower than this range are typically associated with calculations from radar observations with poor data quality (low signal-to-noise ratio) and are thus excluded from further consideration.
3.1.1 Verification of Clutter Proxy
For comparison and verification of use of the RMS height parameter as a quantitative clutter proxy, we also qualitatively assessed clutter for each radar observation included in our subsample of OBLs. We assigned a low, moderate, or high clutter rank to each radar observation based on the concentration of clutter features visually identified in the radargram and associated cluttergram. Radargrams with a low visual clutter rank had few cross-track surface echoes (N = 140), making detection of subsurface reflectors more likely (Figure 1b). High visual clutter ranks (N = 64) corresponded to radargrams where any subsurface reflector is likely to be entirely obscured (Figure 1e). Those with a moderate clutter rank (N = 219) had more cross-track echoes than those with a low rank. Identification of a reflector is still possible here, but there is an increased likelihood that a reflector will be obscured (Figure 1d). Obfuscation is not confined solely to near surface reflectors. Long, continuous returns from off-nadir sources can extend to considerably high time delays and so can also obscure any reflectors that are present at these later time delays.
We find that the RMS height parameter correlates well with our visual assessment of clutter. Figure 5 shows a statistically significant difference between the median and interquartile range of values between low, moderate, and high clutter radargrams. We find corresponding values of 4.5 ± 1.7 m, 7.4 ± 3.2 m, 13.5 ± 9.3 m for low, moderate, and high clutter by taking the median of each visual clutter rank with upper and lower bounds determined by the interquartile range. A Mann-Whitney U test between low to high, low to moderate, and moderate to high clutter ranks reveals p values <0.001 confirming a statistically significant difference between ranks.
3.1.2 Clutter Comparisons of Observations
Our observations show that OBLs possess abundant smooth terrain that result in many radargrams with low clutter that should be favorable for detection of subsurface interfaces. We predict that radargrams with the least clutter and that are most favorable for identifying subsurface interfaces will fall within a region of low RMS height parameter (0–6 m) and low to moderate SHARAD roughness (2–5; Figure 6, black box). Determining these values for the entire population of OBLs, we find that 249 radargrams (23.2%) covering 36 OBLs fall within this most favorable region. Radargrams for Lake 218 that show reflectors also fall within this range (Figure 6). The radargram shown in Figure 2 (track number 1967301) has an RMS height parameter and SHARAD roughness parameter of 5.8 and 4.2 m, respectively. Additional radargrams containing this reflector have track numbers 663901, 592701, and 1679201. The RMS height parameter for these tracks is 5.1, 3.4, and 4.7 m respectively. The calculated SHARAD roughness parameter for these tracks is 4.6, 4.0, and 4.5, respectively. These radargrams fall completely within the region of most probable detection and is therefore consistent with our predictions.
In contrast, radargrams least favorable to detections are predicted to fall within a region of high RMS height parameter (14–30 m) and SHARAD roughness spanning the full range of 2–8 (Figure 6, red box). By definition (section 3.1), radargrams that rank as “high” both visually and by the calculated RMS height parameter are considered far too cluttered to detect subsurface reflections regardless of their value for the SHARAD roughness parameter. A total of 171 radargrams (15.9%) from 56 OBLs fall within this least favorable region. The remaining radargrams (N = 278) covering 70 OBLs have moderate RMS height values (6–9 m) that are predicted to have intermediate clutter characteristics.
3.2 Volumetric Scattering
Internal physical characteristics that cause volumetric scattering can also contribute to attenuation of the radar signal. Stillman and Grimm (2011) suggested that internal fracturing at scales near the wavelength of SHARAD could help to explain the necessary attenuation over many geologic units on Mars. Other sources of volumetric scattering include dielectric heterogeneities, such as large, discontinuous density variations within the unit.
Although it is difficult to directly assess the internal characteristics of layering from surface observations, billions of years of erosion and tectonic modification of most OBL surfaces have created windows into the subsurface that may be examined from high resolution orbital images. We examined available CTX images (Malin et al., 2007) and several High Resolution Imaging Science Experiment images (McEwen et al., 2007) in order to assess possible sources of volumetric scattering. In particular, we search for evidence of exposed fracturing to test the hypothesis of Stillman and Grimm (2011).
We found that only two basins in our subsample, FH2008 Lake 115 and Antoniadi (FH2008 Lake 10), had observable evidence of fracturing on the surface. Some smooth volcanic plains near the center of Antoniadi exhibit a number of intersecting linear ridges in CTX images that we interpret as filled fractures exposed by differential erosion (Figure 7a.iii). Other pockets of fractures appear concentrated near the southeast rim of the basin as well. Their isolated occurrence suggests that such fractures have not been preserved at the surface but could be pervading the subsurface. It is also possible that subsurface fracturing may pervade the other basins as well, but this is difficult to test without direct evidence.
Further examination of a CTX image (Figure 7) along the western rim of Antoniadi reveals an exposed contact between the Hesperian lavas and underlying rough, Noachian-aged material. This rougher surface beneath the lavas may be ejecta from a nearby crater that is capped by lavas sourced from Syrtis Major and subsurface conduits within Antoniadi (Greeley & Guest, 1987; Hiesinger & Head, 2004). While a dielectric contrast may be present, the rough terrain may efficiently scatter the radar signal from SHARAD and create the widespread lack of reflectors we have encountered.
3.3 Hydrated Minerals
A recent global compilation of hydrated mineral detections by CRISM (Carter et al., 2013) indicates that these minerals occur in large regions throughout the southern highlands. If present within the near surface of OBLs, they could act to attenuate the radar signal and prevent detection of subsurface reflectors due to their increased absorption losses relative to dry materials (e.g., Stillman & Grimm, 2011).
Based on the CRISM detections (Carter et al., 2013), only four basins in the subsample contain exposed hydrated minerals. Additionally, three CRISM observations are present in Lake 218: one over each of the two inlet channels on the eastern rim and one in the center. These spectral parameter maps indicate an absence of exposed hydrated minerals, which is consistent with the observations made by Carter et al. (2013). While it is possible that hydrated minerals may be obscured by dust or could occur below the surface, many volcanically resurfaced lakes have clear basaltic signatures suggesting little alteration by water after emplacement of the volcanic material (Goudge, Mustard, Head, & Fassett, 2012). This suggests that hydrated minerals exposed at the surface or in the near subsurface play a minor role in the lack of observed OBL subsurface reflectors. Further, at least some of the CRISM detections of hydrated minerals appear to represent surface exposures of material stratigraphically below resurfacing lava plains. For example, in Antoniadi basin, two outcrops of hydrated silica occur along the basin's southeastern rim (Smith & Bandfield, 2012). They form high-standing knobs of Noachian material that are surrounded by Hesperian volcanic plains (Figure 7b.iii). Similarly, in Gusev crater there is a lack of hydrated minerals at the surface (Figure 4a) and detections are confined to isolated regions such as Home Plate toward the center of Gusev. While their presence at the surface may attenuate the radar signal locally, the relatively low number of detections and the detection of reflectors in Lake 218 despite later fluvial activity suggests that hydration is likely not a major attenuating factor within Antoniadi or the OBL population.
4 Discussion
While surface clutter on multiple scales can account for lack of observed subsurface detections in 16% of the radargrams traversing OBLs, 23% of the radargrams have favorable, low-clutter characteristics that make them ideal for investigating potential reflectors. However, only four radargrams in Lake 218 possess a plausible subsurface detection. Our investigation for possible factors contributing to these lack of detections suggest that contrary to previous hypotheses (Stillman & Grimm, 2011), adsorptive losses due to increased surface hydration is unlikely a major factor.
Despite some evidence of fracturing resolvable with CTX images (~18 m) within Antoniadi and one other OBL, sources of volumetric scattering and other internal characteristics were difficult to investigate from orbital images alone. Due to this difficulty, we cannot rule out the possibility that features contributing to volumetric scattering are still present within OBL units despite the lack of evidence at the surface. Alternatively, observations from the radargrams themselves may reveal the influence of volumetric scattering. Indeed, we find that 91% of the radargrams in our subsample had a “rainfall” effect that we suggest may be related to internal scattering (Figure 1b). This effect is discontinuous along track and typically appears comparable to or lower in power than clutter in the radargram. It decreases in power with increasing time delay; its maximum time delay is also highly variable. A similar effect described by Campbell et al. (2013) suggests that this rainfall effect may be a result of surface roughness at SHARAD wavelength scales. However, we find that this rainfall effect does not appear to be well correlated with surface clutter, as it occurs in many tracks with low RMS height parameter and low SHARAD roughness parameter values. For example, in Antoniadi, radargrams with the rainfall effect and low clutter are located over smooth areas of volcanic plains (Figures 1a and 1b). We also find it within the smooth interior of Gusev crater (Figure 4). Possible sources of the rainfall effect are internal scattering from facets within the subsurface or small-scale fractures that lower the bulk density of the layer as well as patchy or pinching out of layering that could create diffuse scattering at SHARAD scales. Additionally, multiple overlapping lava flows often have complex internal structures that could prevent SHARAD from detecting a uniform interface, and tuff deposits can have inhomogeneities such as welding contacts and density variations that lead to large radar attenuation values (Grimm et al., 2006).
Subsurface clutter could also be a contributing factor. Like surface clutter, a rough subsurface interface can cause diffuse scattering if the facets along the interface are randomly oriented at the 10- to 100-m scale. Due to the ancient nature of OBLs and the possibly large lag time between sedimentary infill and later episodes of resurfacing, early erosional processes likely modified and roughened the tops of the sedimentary fill. In volcanically resurfaced OBLs, the rough contact between sediments and volcanic material can be preserved while creating smooth topography at the surface. Such is the case in Antoniadi, where more recent erosion of surface volcanic unit has exposed roughened terrain at depth (Figure 8). It is possible that this type of interface extends across the interior of Antoniadi and may also occur in other OBLs, explaining the prevalence of diffuse scattering that we observe in SHARAD radargrams.
5 Conclusions
OBLs on Mars have been important sites for understanding the habitability of the planet and will continue to be targets for future landed missions. Our current ability to actively probe the shallow subsurface of OBLs is limited to radar sounding; therefore, understanding the radar response at these locations is important to more completely characterize basin geology. Despite evidence of stratigraphic layering from orbital observations, only one OBL showed evidence of a subsurface radar reflection. Detailed examination of SHARAD radar observations and surface geology at the remaining OBLs suggest that surface clutter and hydration alone are unlikely explanations for lack of detections in the smooth and unaltered interiors of many OBLs. Although surface expressions of fracturing or layer characteristics are limited, more likely candidates for radar attenuation are volumetric scattering or scattering off rough subsurface interfaces. A lack of a dielectric contrast or a deep reflector are also possibilities, but the presence of reflectors in other areas of known contacts between sediments and lava flows (e.g., Campbell et al., 2008; Morgan et al., 2015) suggests that such interfaces should be readily sensed by SHARAD. Further investigation and modeling will be necessary to provide more insight into the complex Martian subsurface of these basins and the southern highlands.
Acknowledgments
SHARAD, MOLA, CTX, and HiRISE data used in this work are publicly available on NASA's Planetary Data System (PDS). NASA Mars Reconnaissance Orbiter SHARAD team funding to L. M. Carter and funding from the Maryland Space Grant Consortium (grant NNX15AJ21H) to E. S. Shoemaker Thackston supported this work. D. M. H. Baker acknowledges support from the NASA Postdoctoral Program administered by Universities Space Research Association. We thank the Associate Editor Wenzhe Fa and two anonymous reviewers for their insightful comments and suggestions for improvement.
