A fragile record of fleeting water on Mars

Decameter-scale, rhythmically bedded layers are exposed within craters throughout Arabia Terra, Mars, and map a rich chronology of depositional surface environments that date back to the Noachian (Lewis and Aharonson, 2014; Hynek and Di Achille, 2017). Because deposition appears to have occurred in recurring episodes over millions of years (Lewis and Aharonson, 2014), these units likely record key tracers of a Martian climate in flux. Orbital observations consistently show hydrated minerals, including polyhydrated sulfates (Wiseman et al., 2010), which hint at past interactions with water. Nonetheless, without in situ measurements of these sulfate-bearing layered deposits (SBLDs), their physical properties and origins have remained obscure.

Elevated apparent thermal inertia (TI) in the SBLDs relative to surrounding units was previously thought to reflect enhanced induration (e.g., Fergason and Christensen, 2008; Andrews-Hanna et al., 2010) because TI increases with cementation (Piqueux and Christensen, 2009). However, these sediments enigmatically tend to lack embedded craters and show widespread erosional landforms suggestive of relatively high erosion rates (Lewis and Aharonson, 2014). Because dust cover often masks the TI of underlying rock (Edwards et al., 2011), one explanation is that dust is heterogeneously accumulating in the region (Rogers et al., 2018). If materials surrounding the SBLDs are mantled by thicker dust deposits than the SBLDs, the relative difference in apparent TIs may not actually reflect greater SBLD induration.

We expanded on previous efforts to determine the style and extent of depositional processes in Arabia Terra by examining controls on the contrasting remote sensing observations. We first quantified the relationship between dust cover and TI for different surface types within large craters using orbital data sets. We then investigated the physical conditions leading to apparent TI heterogeneity in dusty regions by considering predicted wind patterns, small crater densities, failing slope geometries, and ultimately by drawing analogies to materials measured in situ on Mars’ surface.

While hydration-related spectral absorptions accompany the SBLDs (Wiseman et al., 2010), the extent to which water contributed to their deposition is debated. Hydrated minerals, elevated TIs, and basin-filling morphologies have led some studies to propose origins in playa, lakebed, or groundwater spring settings under persistent, water-stable conditions (Wiseman et al., 2010; Zabrusky et al., 2012; Pondrelli et al., 2019). Indeed, models indicate that a deep groundwater supply could have lasted in the region through the end of the Noachian (Andrews-Hanna et al., 2010). Nonetheless, mineral hydration and basin filling are not unique to aqueous deposition (Michalski and Niles, 2012), and groundwater activity does not necessitate persistent nor high water-to-rock ratios (Kite, 2019). Several observations seem instead to challenge a fully aqueous deposition model. Consistent layer thicknesses and exposures of layered deposits outside topographic lows suggest these deposits stem from depositional processes with regional extent (Lewis and Aharonson, 2014; Hynek and Di Achille, 2017; Annex and Lewis, 2020) contrary to expectations for individual paleolake basins. Shallow, topography-conforming layer inclinations (Annex and Lewis, 2020) and central-mound landforms comprising many SBLDs are also inconsistent with the horizontal layers or spring-like heaps expected in a water-fed system (Kite et al., 2013; Pondrelli et al., 2019).

As an alternative, several airfall sources have been proposed, including impact processes (Knauth et al., 2005), pyroclastic ejecta (Hynek and Di Achille, 2017), and eolian deposition (Day and Catling, 2019). Without an aqueous depositional setting, hydration and weak cementation may occur from vapor diffusion (Bridges and Muhs, 2012), diagenesis (Fraeman et al., 2016), and/or ice sublimation (Michalski and Niles, 2012), similar to conditions in the sulfate-bearing layers at Mars’ poles. Generally, processes involving submergence in pooled water induce higher degrees of cementation than those involving sublimated ice or vapor (Bridges and Muhs, 2012), while diagenetic cementation varies with the water-to-rock ratio (Kite, 2019). Since pore-filling cements have a strong influence on TI (Piqueux and Christensen, 2009), improving interpretations of TI in these deposits adds an integral constraint on their formation and alteration.

Spectrally derived dust maps of Arabia Terra (e.g., Ruff and Christensen, 2002) show widespread dust cover except in isolated patches (Fig. 1). At >2 km/pixel, however, such maps lack the spatial resolution required to differentiate units within the region's crater basins. We identified 10 craters at the southwestern margins of the dustier central region of Arabia Terra, where SBLDs appear in 18–36 m/pixel Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) data (Fig. 1). At each location, we paired 100 m/pixel Thermal Emission Imaging System (THEMIS) TI (Edwards et al., 2011) with CRISM band parameter maps (Viviano-Beck et al., 2014) as inputs to a clustering model (see the Supplemental Material¹) to classify surface units (Fig. 2) and quantify how TI varies between regions of different spectral character. The units classified reflect a similar set of units identified previously in the region (Edgett, 2002; Wiseman et al., 2010; Zabrusky et al., 2012), which include SBLDs, clay-bearing crater floors and rims, and mafic capping materials (Figs. 1 and 2).

While SBLDs are characterized by sharp 1.9 µm and modest 2.4 µm absorption features, the majority of surrounding units, regolith, and apparently dusty layers display a broad pyroxene spectral absorption spanning 1.4–2.4 µm (Fig. 1). SBLDs showing distinct sulfate absorptions averaged TIs of ∼250–400 Jm^–2 K^–1 s^–1/2 (hereafter SI [SI units of thermal inertia]). We note that because image-to-image uncertainty (up to 50 SI) is much greater than pixel-to-pixel uncertainty (∼10–20 SI) (Fergason et al., 2006; Edwards et al., 2011), the differences in TI between neighboring units can be more useful than absolute values alone for exposing regional trends (Fig. S1 in the Supplemental Material). At each sample site, SBLDs with distinct sulfate absorptions averaged 66.49 ± 0.34 SI higher than SBLDs missing those features and 28–89 SI higher than crater floor and wall units. Lower TI regions also showed fine-scale, smooth surface textures that conform to topography (Fig. 2), which we infer to reflect a dust mantle.

To place the potential effect of dust on TI in context, models show that dust layers 40–200 µm thick can reduce bedrock surface TIs by as much as 200 SI (Edwards et al., 2011). At our easternmost site, Kaporo Crater, TIs reach as low as ∼50 SI, which is consistent with a dust layer of at least a few centimeters thick (Piqueux and Christensen, 2009). When comparing dust-covered SBLDs to nearby low-albedo sand dunes, TIs averaged 33.63 ± 0.51 SI higher for the sand, reinforcing that even thin cover from dust-sized sediment lowers TI to values beneath that of sand-sized sediment (Golombek et al., 2008; Piqueux and Christensen, 2009).

Two questions remain: (1) why do some SBLDs retain less dust than surrounding units?, and (2) what dictates where dust is being removed? One possibility is that mechanical differences between each unit are leading to differential erosion, and erosion in some SBLDs is outpacing the rate of dust accumulation.

Modern erosion of Mars’ layered sedimentary rocks is enhanced by wind stress and saltating particles (Bridges et al., 2012). Using prevailing wind directions from atmospheric models (Toigo et al., 2012) and observational indicators of eolian direction (e.g., Edgett, 2002), we estimated the spatial distribution of erosional stress within each crater (Fig. 3; Fig. S2). Fiercer erosion is anticipated for wind-facing slopes (Day et al., 2016), areas impacted by slope winds (Kite et al., 2013), and downwind of saltating sand deposits (Edgett, 2002; Bridges et al., 2012). Although this approach to identifying higher-stress surfaces is qualitative, layered terrain in those zones was consistently 50–200 SI greater than at other intracrater areas. In Sera Crater, a wind-facing slope of a mounded portion of the crater floor unit (Fig. 3) shows notably high TIs (>500 SI) and few craters, which indicates that this may be one of the few areas where the erosional force is strong enough to resurface that material.

Another useful indicator of erosion history is the divergence of observed crater distributions from expected crater production functions (Kite and Mayer, 2017). While crater densities can differ based on age and non-erosional processes, the similarity in expected unit ages (Hynek and Di Achille, 2017) and stark differences in crater densities (Fig. 3) point to exhumation as a likely control.

Assuming steady exhumation, we use a model (Kite and Mayer, 2017; see the Supplemental Material) for deriving SBLD erosion rates from crater densities in seven crater basins (Fig. 3). Determining a precise erosion rate for the crater floors is difficult since crater superposition suggests prolonged burial (Fig. 2). However, average >20-m-diameter crater density is similar to what is observed for bedrock at nearby Meridiani Planum (Golombek et al., 2014), implying that the long-term erosion rates of 1–10 nm/yr derived there may be similar. The resulting mean estimate for SBLDs was 131–1088 nm/yr, which is in line with prior upper estimates of ∼1000 nm/yr for other layered deposits in Arabia Terra (Smith et al., 2008) and weakly indurated materials in Meridiani Planum (Golombek et al., 2014). Given Mars’ low atmospheric pressure, such high rates in the SBLDs must reflect a combination of locally heightened erosional forces and/or notable friability (Bridges et al., 2012).

To understand if enhanced erosion of the SBLDs relates to rock strength, we applied material failure criteria, as outlined in Wyllie and Mah (2004), to derive rock mass cohesion. We measured the geometries of failed slopes at High Resolution Imaging Science Experiment (HiRISE) image scales (Fig. 4; see the Supplemental Material), limiting the approach to outcrops showing evidence of failure at (50 m scales. For circular-style failures, we derived an estimate of ∼35 ± 10 kPa for the upper limits of SBLD cohesion. Using more precisely measurable wedge failure surfaces, we derived cohesions ranging from 0.13 kPa to 17.9 kPa. These values are consistent with various forms of surface-crusted regolith measured in situ at Meridiani Planum and other mission landing sites (Golombek et al., 2008). For comparison, analog tensile strength measurements of sedimentary rocks in Gale Crater (Kronyak et al., 2020), which complement prior in situ unconfined compressive strength measurements (Peters et al., 2018), yielded estimates of 30–100 kPa for Murray mudstone drill sites and 5000–20,000 kPa for Stimson sandstone drill sites, both of which are considered very soft compared to most sedimentary rock on Earth. Bulk cohesion is generally ∼2× the tensile strength in sedimentary rock (Sivakugan et al., 2014), which means that the SBLDs are at most as strong as the weakest Murray mudstones but likely much weaker. TI of bedrock in Murray and Stimson outcrops was modeled from Mars Science Laboratory (MSL) temperature observations to be ∼650 SI and ∼700 SI, respectively (Vasavada et al., 2017), which is >200 SI above the values we derive for SBLDs. The crusted materials in Meridiani Planum appear to be a better analog, and TI estimates of 300–400 SI for those surfaces, derived from ground-based measurements (Golombek et al., 2008), are also consistent with our observations of the highest TIs in the SBLDs.

Our estimates of low cohesion and low TI, combined with context from prior bedding measurements (Annex and Lewis, 2020), indicate that episodic weak induration similar to crusting observed in Meridiani Planum (Golombek et al., 2008) would have been enough to sustain the observed SBLD structures. Even small amounts of intergranular cement dramatically increase TI (Piqueux and Christensen, 2009), and if the sulfates within SBLDs comprise a major cementing secondary mineral, akin to gypsum-bound rocks on Earth, we should expect much higher cohesion and TI (Wyllie and Mah, 2004; Piqueux and Christensen, 2009; Peters et al., 2018). Instead, these deposits could have formed without any direct water infiltration (e.g., Bridges and Muhs, 2012; Michalski and Niles, 2012) or following only brief interactions with water (Kite, 2019). While it is possible that the SBLDs were once more cohesive and have since been degraded, post-cementation decomposition (such as from decompression, thermal heaving, mineral substitution, and impact deformation) of this magnitude covering such a large region seems unlikely given the dependency of such processes on localized conditions (Melosh, 2011). Although an exact mineral-forming process remains uncertain, we conclude that Arabia Terra's SBLDs most likely formed via eolian deposition and only mild subsequent induration.

Our investigation of how dust controls thermal inertia and derivation of the physical properties of layered deposits within Arabia Terra craters provides new information about their depositional history. SBLDs in Arabia Terra are among the fastest-eroding and least cohesive rocks observed on Mars, which implies both local eolian stresses above previous regionally resolved estimates and relatively weak induration. When paired with observations of low bedding inclinations, regionally extensive sequences, and exposures of SBLDs outside of topographic lows, low cohesion in SBLDs is interpreted to result from airfall processes and intermittent wetting. If the SBLDs record conditions on Mars prior to atmospheric loss, then investigations of similar deposits by rovers in Gale Crater and near Jezero Crater should provide significant clues to habitability at the end of the Noachian as well as to the form and cadence of Mars’ dehydration.

We thank Y. Itoh for processed CRISM data products, and three anonymous reviewers for comments that substantially improved this work. This work was funded by the NASA Mars Data Analysis Program (grant 80NSSC17K0672).