Introduction

Layered sedimentary rocks record planetary evolution over billions of years1. On Earth, many attributes of the sedimentary rock record, including layer thickness, depocenter location, and rock volume (accumulation rates) are sensitive to environmental conditions2. The same is true of these attributes for Mars’ sedimentary rocks2,3. Despite this potential, there have been no studies of the global-scale variations in these parameters in Mars’ sedimentary rock record or their variations in time.

(Previous work involved regional studies Annex and Lewis4, Schmidt et al.5,6, Koeppel et al.7, Di Pietro et al.8, Annex et al.9). Today, high resolution (25 cm/pixel) orbital images10 and digital elevation models (DEMs) (1 m/pixel)11 of different deposits allow us to form the most comprehensive and highest resolution dataset to date11,12,13, which permits accurate measurements of layer thicknesses at different outcrops within the global sedimentary rock record. This record consists primarily of, in stratigraphic order, the clay, sulfate and rhythmite units, which are distinguished by composition, facies and stratigraphic position12. These rock units also demonstrate Mars’ environmental transition from a planet with relatively more abundant liquid water (globally distributed clays) to present day Mars, which is cold and dry (globally distributed anhydrous oxides)2.

As they comprise our best record of planet-wide environmental change, understanding the formation, growth and evolution of these stratigraphic units is critical14. Clays on Mars have been hypothesized to record abundant aqueous activity3,12,15, and many large deltas and channels visible from orbit formed around the same time as clay formation. In contrast, orbital facies of the rhythmites and underlying sulfates consist of laterally extensive, parallel layers of uncertain origin. These units make up the youngest, and most voluminous, portion of the Martian sedimentary rock record. Together, these units apparently accumulated over a time span of  ~3.5 billion years (beginning ~3.9 Ga)16, accounting for the majority of sedimentary rocks on Mars and recording a transition (sulfates to rhythmites) hypothesized to record global climate change17. Global mechanisms such as ashfall from large caldera-forming eruptions and climate-controlled dust circulation and induration have been proposed for rhythmite accumulation, (though alternative sedimentation models exist2). While ashfall would result in sedimentary rock volume and layer thicknesses thinning away from eruptive centers18,19, climate-modulated dust circulation would result in these parameters being concentrated at low latitudes and thinning to the north and south20. Here we test and reject these hypotheses using global datasets, while providing an additional integrated synthesis of the layer thickness, deposit thickness and volume, and age of Mars’ young layered sedimentary rocks. We obtain a surprising trend where the decrease in rate of sedimentary rock production post-dates an initial drop in layer thickness that occurred 3.5 Ga. Based on this result, we suggest a planetary evolution scenario where an initial drop in atmospheric pressure results in thinner layers and a later, gradual decrease in liquid water availability leads to a slow decline in planet-integrated sedimentary rock volume.

Results and discussion

Regional trends in layer thickness, global trends in outcrop age

We obtained 972 layer measurements from 49 HiRISE (High Resolution Imaging Science Experiment) Digital Elevation Models (DEMs) and orthomosaics of sedimentary deposits on Mars (including new DEMs made for this study) and also analyzed previously collected sedimentary mound volume estimates, along with crater chronology data. Data selection was largely controlled by data availability (Fig. 1; Table S1), and image pairs were chosen to sample each of Mars’ major rhythmite and sulfate deposits12. We require accurate layer thickness measurements throughout an outcrop, so mosaics with continuous, laterally consistent and relatively undeformed layering were prioritized.

Fig. 1: Methods demonstration.
figure 1

Workflow for measuring layer thicknesses. (Top) Images and DEMs are obtained from the locations marked in the map (black markers) (Fig. S1), with black polygons outlining the extent of the targeted sedimentary deposits53. (Middle) Sections of continuous layering are identified and traced directly onto the images. Bedding plane orientations are obtained using the layer traces and the corresponding topography. Dip corrected layer thicknesses are obtained for each layer section using the mean layer orientation for the different layer packages (N:3-11). (Bottom) Layer thicknesses are reported as the mean thickness for each package, with the uncertainty quantified using the thinnest and thickest of the individual layer thicknesses.

Full size image

We obtained data from globally representative sedimentary mounds at Valles Marineris, Arabia Terra, Meridiani Planum, the Medussae Fossae Formation (MFF), in addition to Galle, Terby, and Gale crater (Figs. 1, S1).

Using this global approach, we report large differences in layer thickness between regions: Valles Marineris layers are on average 2 × thinner than the Arabia Terra sites with a maximum layer thickness that is 23m smaller. Furthermore, layer thicknesses are more uniform within Valles Marineris: mean fractional variance of these sites is 0.8 ± 0.4 versus 2.1 ± 2.1 for Arabia sites (Figs. S12, S14, S16). Thus, accumulation in these different geographic areas was likely governed by regional depositional processes (Figs. 2–4, S3, S6, S12) (although in principle global dust storms could account for this type of variability with a single sediment source21). Valles Marineris sites show thinning away from the Tharsis volcanoes, as expected for a volcanic origin18,19, although this trend is slight. Conversely, the Arabia and Medusae Fossae Formation sites lack clear intra-regional trends in latitude or distance from volcanic sources (Fig. S10). The differences in trends and in fractional variance between deposits suggests that sediment deposition in these two regions was governed by independent regional controls/independent sediment sources. Additionally, we do not observe any global trends of thicker layers at lower latitudes20, nor thinning layers away from the major volcanic centers18,19 (Figs. S4, S6). This is inconsistent with previously proposed simple models.

Fig. 2: Layer thickness vs. Time.
figure 2

Layer thicknesses of measured packages decrease over time. Regions in the 3.5–4.5 Ga bin include sulfates and rhythmites from Meridiani Planum24and sulfates from Gale crater22. Valles Marineris sulfates and deposits at Terby crater23 are in the 3.5–2.5 Ga bin, and Gale crater rhythmites and the Eastern MFF are in the 2.5–1.5 Ga bin16,54. The youngest bin contains layer thicknesses from Zephyria Planum and the Valles Marineris rhythmites16. The size of each box plot is determined by the 25th and 75th percentile value, while the median is plotted with the black line. Whiskers are drawn to show the minimum and maximum values when excluding all outliers. Outliers are defined using the interquartile range (IQR), where a value is defined as an outlier if it is below Q1 - 1.5*IQR or above Q3 + 1.5*IQR. The actual data (black points) is plotted to the right of each bin.

Full size image

To assess trends with time, crater counting statistics were obtained from the Arabia and Valles Marineris deposits (Materials and Methods, Fig. S2) and used together with existing statistics from Gale crater16,22, the MFF16, Terby crater23, and Meridiani Planum24. These data place both upper limits on the age of sedimentary rocks (from the crater retention age of underlying materials) and lower limits on the age of sedimentary rocks (from the crater retention age of the sedimentary rocks themselves), allowing for our data to be ordinated in time. We found that the oldest deposits (intercrater mounds in Arabia Terra, and Terby crater (Fig. 1)), have thicker layers than younger regions (Fig. 2). Moreover, deposits with the same orbital facies based on globally representative morphology, mineralogy and stratigraphic order have vastly different ages (Fig. 3), ruling out global climate as the pacemaker of the shift from sulfate orbital facies to rhythmite orbital facies17.

Fig. 3: Time-integrated sedimentary rock volume.
figure 3

Estimate of the minimum rate of sedimentary rock formation through time. Each curve and the underlying shaded area (blue for sulfates, yellow for rhythmites) marks a different sedimentary deposit mapped on the Martian surface, shown in map view in Fig. 4. A detailed breakdown of all mounds represented in this plot is shown in Fig. S18, and the uncertainty in MFF mound volumes (Fig. S15) is represented by the dashed lines. In most cases, ages are obtained using the crater age of the underlying surface as lower bound and the exhumation age, the age of the current exposure, as an upper bound. All crater retention ages were obtained using a crater counting software, either within this study or previous work16,22 -- 24,45,46,55 (Fig. S2, Table S2). Additionally, ages for the Medusae Fossae Formation (MFF) mounds and rhythmite deposits in Valles Marineris are formation age estimates corrected for erosion16.

Full size image

Using age estimates and mound boundaries (Materials and Methods), we estimated the total accumulation of sedimentary rock volume as a function of time (i.e., km3/yr) (Fig. 3). While previous work assumes a gradual drying out of the Martian surface, with a drop-off in sedimentary rock formation with time12, Figs. 2–3 show a disconnect between the drop in layer thickness, which occurs early in Mars’ history, compared to that in sedimentary rock accumulation, which occurs much later ( ~0.4 Ga16). Early on, there is a strong pulse of accumulation and preservation at Meridiani Planum. However, other older deposits (Gale sulfates, Terby crater, Arabia Terra rhythmites) correspond to a global low in sedimentary rock volume. Sediment volume rebounds with the accumulation of the Valles Marineris mounds, and later, the MFF deposits, before finally dropping off to no sedimentary rock accumulation in the present day (excluding polar deposits). The cause of the apparent secondary rebound in rock accumulation after the low preservation levels immediately following Meridiani Planum is unknown: a similar pattern in terrestrial data corresponds to Earth’s Great Unconformity1. This result for the Martian rock record highlights a complex history of feedbacks between the factors influencing sedimentary rock formation, namely water availability, erosion, and sediment source, through time.

Rhythmite formation on Mars

Our layer thickness data exclude both the simple volcanic19 and snowmelt-controlled deposition20 models as a global depositional mechanism. The snowmelt model predicts layer thinning at higher latitudes20, inconsistent with these data (Fig. S4). A global volcanic ashfall depositional model is also not consistent with these data because the global minimum in layer thickness in Valles Marineris is closest to Tharsis, Mars’ primary magmatic center (Figs. S1, S6). Instead of global control, semivariogram results indicate a regional (103 km scale) control of layer thickness (Fig. S12).

The spatial variation of layer thickness within the Valles Marineris deposits could be consistent with volcanic ashfall, given the subtle drop-off in layer thickness toward the southeast, away from the Tharsis volcanoes (Fig. S9). The layers in these deposits also thicken toward the equator, as predicted by the snowmelt model. However, other outcrops deposited during the same time period lack equatorial thickening. Because of this, we feel that a regional volcanic-source scenario is the more likely of these options. The thicknesses of the Arabia Terra deposits lack intra-regional trends (Figs. S4, S10), meaning that neither of the simple models explain the deposits in this region. This is consistent with prior studies of this region that favor groundwater upwelling as a control on sedimentary rock accumulation5,8,25, though alternative models exist6,7.

Over time, layers thinned (Fig. 2) and the amount of sedimentary rock of a given age that has been preserved (km3/yr) decreased (Fig. 3). In principle, these trends could be caused by a decrease over time in sediment availability (sediment starvation), an increase in erosion, or a decrease in cementation potential (for conversion of sediment into rock). We now consider each in turn.

The sediment starvation hypothesis predicts that the sediment-rich times correspond to older, thicker layers. However, the older, thickly layered rhythmites (yellow wedges at 3.7 Ga in Fig. 3, corresponding to Arabia Terra) formed during a period with a lower volume of rock preservation. Most of Mars’ sedimentary rocks are younger (i.e. Valles Marineris, MFF). The presence of abundant young sedimentary deposits implies that sediment must have been available much later in time (Figs. 3, 4). Alternatively, thinning layers might correspond to an increase in erosion over time. However, this contradicts the expected decrease in erosion potential over time as Mars’ atmosphere was lost26. Therefore, we propose that the decrease in layer thickness and rock volume most likely corresponds to a decrease in cementation potential with time. This is based on the assumption that water is the primary cementing agent, therefore more liquid water availability would correspond to a higher cementation potential and more sedimentary rock formation. Alternatively, a drier environment would result in a general decrease in trapped, deposited sediment, corresponding to thinner layers.

Fig. 4: Post-3.5 Ga Martian stratigraphy.
figure 4

Global distribution of the sedimentary rock units analyzed in this study. The representative columns from each region show the relative regional volume (column width), maximum outcrop thickness (column height), and layer thickness (hatching: quantified to the right of each column) of the sulfate and rhythmite units. Mound volumes range over three orders of magnitude. Tick marks at the base of each column correspond to log10 units.

Full size image

Though a decrease in cementation potential can explain decrease in rock formation rate and layer thickness over time, it does not explain the decoupling in the timing of these declines: layer thickness by a factor of  ~ 2–3 Gyr (Figs. 3, 4) before sedimentary rock formation rate declines. We propose that this can be explained by an initial drop in atmospheric pressure, which would inhibit erosion26, prior to the start of decreasing cementation potential caused by loss of liquid surface water (Fig. 5). The oldest thickest layers (Arabia Terra rhythmites, Gale and Terby sulfates) would be deposited before the drop in atmospheric pressure, leading to a low preservation potential due to erosion by a thicker atmosphere. The persistence of sedimentary rock volume late into the Amazonian signals a drop in atmospheric pressure allowing for sedimentary rock to be more easily preserved. The combined presence of a lower erosional potential with a longer-lived, but decreasing, liquid surface water reservoir would then explain the large volume of young, thinly layered sedimentary rocks. The large preserved rock volume of Meridiani Planum sulfates (Fig. 3) is an outlier relative to the framework described above. We suggest this is explained by greater erosional resistence, implying that these deposits are more indurated. Meridiani Planum sulfates have a relatively large density of craters24, some of which are well-preserved despite being  > 3.6 Ga (e.g. Fig. S19). This indicates greater erosional resistance because erosion effaces craters.

Fig. 5: Macrostratigraphy of Mars 3.5 Ga to present.
figure 5

Deposit columns and interpreted stratigraphic correlations are organized by relative age and longitude. Stratigraphic correlations are marked by arrows and lines and by the colored masks. We find that Valles Marineris and Arabia Terra sedimentary rocks evolved asynchronously despite the similarities in stratigraphy. In contrast, deposits of the Eastern and Central Medusae Fossae Formation, Aeolis-Zephyria Planum, and Gale crater are hypothesized to be correlative54, which would span roughly 120∘ in longitude around the planet. Alternatively, Gale crater could have evolved asynchronously from the rest of the Medusae Fossae Formation, perhaps via atmospheric teleconnections to Arabia Terra or Valles Marineris. The bars on the right side of the figure display the interpreted changing controls (cementation potential and atmosphere) and the resulting observable characteristic (layer thickness). Layer thickness peaks during the deposition of rhythmites in Arabia Terra, and sulfates at the Mid-Latitudes and Gale- in our interpretation, this is after the atmosphere has thinned and before significant loss of liquid water has occurred.

Full size image

Spatial trends in sedimentary rock deposition through time

Surprisingly, given the complete lack of evidence for continental drift on Mars, we find that late stage sedimentation was unsteady and regularly shifted locations (Figs. 4, 5). Sedimentation stops early near 0∘ longitude while deposition of rhythmites near 60∘ W and 150∘ E persists to  ~0.4 Ga16. This order-of-magnitude age difference corresponds to the same orbital facies: these deposits all display the characteristic, light-toned, regular layering of the rhythmite facies and always superpose sulfate facies when both units co-occur12,13,24. Thus, orbital facies are not a reliable time-marker in the Mars geologic record as has been proposed in the past17, but are more consistent with alternate hypotheses of deposition2. Why did these shifts occur? Possible controls include depositional basin size and age, groundwater table fluctuations or a shift from groundwater to surface water as the water source for aqueous cementation20,25, true polar wander and corresponding shifts in topography27,28, or the timing of volcanic eruptions19. We now address each in turn.

Most mounds analyzed in this study sit within natural geologic containers. Examples include crater basins and canyon interiors, for which the craters’ rims and canyons’ walls limit the extent of the interior sediment. The inter-crater deposits in Arabia Terra (which have thicker layers) have much smaller geologic containers than the mounds within the canyons of Valles Marineris (which have thinner layers). This is consistent with a simple model that predicts that the deposits in smaller containers have thicker layers (Fig. S13).

When sedimentation started, some containers did not yet exist. Estimates of the age of the Valles Marineris rift system converge around the Noachian-Hesperian boundary ( ~3.7 Ga)29, so lack of basins could have limited early deposition in Valles Marineris. This alone does not explain why sedimentation stopped in Arabia Terra, and does not account for the Medusae Fossae Formation, which has no container.

Groundwater hydrology models predict upwelling in most of the locations where sedimentary mounds occur25. If groundwater were the primary source of aqueous ions for minerals in these sedimentary rocks, the spatial shifts could be explained by fluctuations in the groundwater table through time. One model does predict that groundwater flow at Arabia Terra would dry up first (Fig. S6)25, consistent with our data. Another potential mechanism could be a shift in the source of aqueous fluid needed for cementing minerals (cementation source) from bottom-up (groundwater controlled), to top-down, snowmelt-supplied20,30. For this scenario, snowmelt would periodically permit the cementation of wind-blown deposits, forming most of the sedimentary mounds we observe today.

Another way to explain the spatial shifts in depocenters with time is true polar wander. As lava piles up at Tharsis, the planet’s axial tilt changes, moving the whole mass equatorward27. This shift would also result in changing topography relative to latitude (and wind patterns)28, which affects sedimentation. However, published models predicting different depocenters in these scenarios do not predict surface water accumulation in Arabia Terra, in tension with the data27. Furthermore, because Arabia Terra lies  ~90∘ longitudinally from the Tharsis load, the region’s latitude would not change much from true polar wander.

The spatial shifts in depocenter could (in principle) be explained by shifting eruptive centers, or multiple eruptive centers contributing to a single location31. Predicted ashfall from calderas overlaps with mapped sedimentary deposits18,19,32. However, if deposits correspond to direct deposition of ash, then their distribution should match the timing of volcanic eruptions. This does not appear to be the case for deposits in Arabia Terra, which are far from large volcanoes and receive relatively little ash based on atmospheric models18,19,32. These sites are predicted to experience south- and eastward thinning from hypothesized volcanoes to the north31,32,33, and thinning westward away from Nili Patera18,19, highlighting the potential for complex layer thickness predictions from multiple regional sources31. However, despite this potential added complexity, the chronology of sedimentary deposition relative to eruptions and regional coherence of sites in close proximity should remain consistent with this hypothesis. Age dating of the largest caldera upwind of Arabia Terra (Nili Patera) shows that eruptions continued long after sedimentation in Arabia Terra stopped34. Furthermore, layer thicknesses in Arabia Terra lack both regional coherence (Figs. S10, S14), and a thinning trend related to the closest volcanic centers32 (Fig. S10). Because of this mismatch in timing and lack of regional coherence in layer thicknesses, shifting eruptive centers cannot straightforwardly explain these data.

Apart from the variable spatial trends within the same unit discussed above, we also find that the spatial variability of the rhythmite unit is completely independent of trends in the sulfate unit. Wherever they co-occur, rhythmite outcrop overlies the sulfates. However, despite this global consistency, there are many differences between regions (Figs. 4, 5). The regions with the thickest rhythmites are not the same as the regions with the thickest sulfates. This applies both to absolute thickness of layers, and relative layer thickness compared with other deposits of the same facies. For example, the thickest rhythmite layers in this dataset (Arabia Terra) are 8000 km away from the thickest sulfate layers (Gale crater). Furthermore, previous studies record even thicker sulfate layers in Juventae Chasma within Valles Marineris13. This decoupling of spatial patterns between the two facies suggests that the rhythmites and sulfates had independent sources and depositional settings21.

An irregular record of layer thickness and sedimentation for Mars

These results show that models of Mars’ sedimentary rock formation need to be updated, and are also a benchmark for tests of current and future hypotheses. Our global survey results are the largest, most detailed, and systematically consistent dataset of layer thickness measurements on Mars. We find several new constraints for Mars’ sedimentary rock record: 1) Layer thickness results support provincial, rather than global, mechanisms of transport, accumulation and preservation (most notable in Valles Marineris and Arabia Terra) (Figs. 4, 5); 2) we confirm a diachronous boundary between sulfates and rhythmites on Mars despite consistent facies, composition and stratigraphic ordering (Figs. 3, 5); and 3) these constraints persist all while layering notably thins post-3.5 Ga (Fig. 2).

Although layers thin over Gyr timescales, we do not observe any systematic thinning within the stratigraphy of a single DEM (10s of Myrs35) (Fig. S7).

The spatial variations in depocenter and layer thickness are robust, and reject simple global models that were built upon observations of global consistency in morphology and composition from orbit17. These trends, taken together with diachroneity (Fig. 5), imply a paleoenvironment that facilitated a dynamic and migratory sedimentary rock cycle despite the absence of plate tectonics.

Methods

Though HiRISE stereopair coverage is sparse, it includes many layered sedimentary mounds that are observed near the equator of Mars. We only study mounds previously identified: sedimentary mounds within the Valles Marineris canyon system36,37,38, the layered inter-crater mounds at Arabia Terra4,5,6,8,9 and Meridiani Planum24, Mount Sharp within Gale crater12,39,40, the Medusae Fossae Formation12, the terraced outcrop in Terby crater23, and the intercrater mound in Galle13. We obtained layer thicknesses with known stratigraphic ordering in at least one HiRISE mosaic and DEM from each of these locations. Each mosaic was assigned to either the rhythmite or sulfate orbital facies to obtain the averages shown in Fig. 4.

We used 49 HiRISE mosaics and DEMs from these outcrops. We only use HiRISE mosaics due to their high spatial resolution (up to 25 cm/pixel imagery and 1 m/pixel DEM). This resolution is necessary given the layer thickness for these deposits; which ranges from sub-1 meter to greater than 10 meters. Because this study uses layering presumably accumulated over orbital timescales13,35, as a proxy for each orbitally defined sedimentation event, we measure the thinnest, laterally continuous layers that can be resolved at each location. Data from Curiosity suggests that layering observable from orbit is frequently composed of finer scale, sub-meter, and millimeter-scale laminae. We cannot resolve these smaller features, and purposefully do not measure finer scale sub-layering within orbital outcrops that is not continuously visible. The selection of this subset of DEMs is controlled by data availability and layer exposure. Our dataset includes previously published and publicly available DEMs from existing stereopairs and 7 new DEMs produced from existing stereopairs using the Integrated Software for Imagers and Spectrometers and Ames Stereo Pipeline (ASP) software41. The vertical resolution of the resulting DEMs depends on the horizontal resolution and the convergence angle of the two images in the stereopair11, and is typically 10 to 30 centimeters. This vertical precision is therefore well below the minimal thickness of the individual layers and layer packages that we measured, as demonstrated in Fig. 1. New DEMs were selected based on layer exposure and perceived data gaps, assessed in part based on previous literature4,5,6,8,9,13,42. Layer exposure also influenced our choice of DEMs. For this work we need accurate, dip corrected thicknesses at different intervals within an orbitally-resolved outcrop. This requires that multiple exposures of laterally continuous and coherent packages of layering be visible. We prioritize layer exposures that lack intense faulting. We also sought to have at least two mosaics for each sedimentary mound to test the null hypothesis that layer thicknesses are spatially uncorrelated.

Outcrops were assigned to the rhythmite or sulfate facies as follows. If outcrops were mentioned in pre-existing, geologic mapping-based literature as being part of one of the facies, we adopted that assignment. This applied to much of the Valles Marineris canyon system37,38, Meridiani Planum24, the Arabia Terra mounds4,5,6,8,9, Gale crater39,40, and Terby crater23. We made new assignments for three sedimentary mounds in Valles Marineris, some of which had not been extensively mapped since the Viking era, and the MFF, which is mostly dust covered. In these cases, we classified the different units using the orbital facies visible from HiRISE and CTX (Context Camera), underlying and overlying contacts, and spectroscopic indices from the Compact Reconnaissance Imaging Spectrometer for Mars (CRISM)43, using criteria outlined in previous literature12.

Boundaries between the rhythmite and sulfate units were mapped in Ophir Chasma and Candor Chasma within Valles Marineris (Fig. S11). These boundaries were determined based on previous work mapping sulfates36, along with detailed geologic maps of individual mounds that isolate different facies consistent with the rhythmite unit37,44. In West Candor and Ophir Chasma, the rhythmite occurs above a large angular unconformity 37,44. For East Candor we mapped the boundaries based on compositional signatures from CRISM and the overlap with the characteristic orbital facies in HiRISE12. The sulfate unit orbital facies consists of laterally consistent (10s–100s kms) bedding of variable thicknesses, ranging from  ~1 to  ~100 meters, with sulfate signatures identifiable from orbit. By contrast, the rhythmite orbital facies consists of light-toned, highly regular and rhythmically layered, fine grained strata with no spectroscopic evidence for aqueous minerals12.

Layer thicknesses at each site are determined by first measuring a series of layers throughout each DEM. Within the images, continuous packages of at least 3 layers are identified. The relative stratigraphic ordering of packages is determined using a combination of layer dips, topography, and existing geologic maps38. Within each package, the layers are traced manually on the HiRISE image, and the underlying topography is extracted in order to calculate layer orientation. A plane is fit to each three-dimensional topographic trace using Ordinary Least Squares (OLS) Regression, assuming that all of the uncertainty is confined to the z-direction13. The dip and azimuth is calculated from each planar fit, and the associated error is defined as the angular error on the normal to each fitted plane. Planes with an error greater than 2∘ are considered to not be good fits, and are not considered in subsequent steps when calculating mean orientations for each package. This 2∘ is an empirical threshold used in many previous studies measuring orbital layering from HiRISE4,9,13,35. For this study, 96% of all measurements satisfy this error threshold.

Layer fits with large error can be caused by a lack of natural outcrop curvature, manifesting on the surface as a linear exposure without a large enough profile in the xy-plane, or by layer curvature, which can be caused by deformation, block displacement, erosion and slumping. For each continuous layer package, fits with better than 2∘ error were used to calculate a mean dip and azimuth for the entire package of layering. If the layers measured within the exposure had more than one layer with an error exceeding this threshold, the package was not used, and an alternate location was found instead.

Layer thicknesses are calculated by transforming each layer into the new coordinate space defined by the plane of the mean dip, where the z’ direction, corresponding to height within stratigraphy, is parallel to the normal of mean layer fit for each package. Within this rotated coordinate space, the stratigraphic height of each layer is determined by the mean height of the entire rotated trace, with the uncertainty of each fit shown as the amplitude of the residuals on each layer fit, as shown in Fig. 1. Uncertainty in the stratigraphic elevation of the fit has similar causes to the error on the fit itself. We use the mean layer thickness of each package as opposed to individual layer thicknesses for our analyses. This minimizes the error associated with measuring a single layer, because the residuals of a single fit are unlikely to extend the span of an entire layer package given the error constraints on the planar fit for the layers. In addition to reporting the mean, we also show the thinnest and thickest individual layer thicknesses (Fig. 1).

Between 3 and 11 separate packages of layers are measured within each mosaic. This depends on the availability of measurable, continuous exposed layers. For analyses of spatial trends in layer thickness (Figs. 4, S6), the mean layer thickness of all package averages is used as the representative value for a DEM. For analyses of layer thickness with age (Fig. 2), individual package averages are used.

We also obtain estimates of age and volume for the different sedimentary mounds in this study (Fig. S2, Table S2). For age estimates, we combined existing age estimates from the literature and our own crater counts. Crater counting and statistics were performed on the mounds in Valles Marineris, the mounds in Arabia Terra (excluding the Meridiani Planum plateau), and Galle crater. In each of these regions, craters with diameters larger than 2 km, 5 km, and 16 km were identified using the Robbins crater database45. The diameter thresholds were chosen for consistency with existing crater counts in Meridiani Planum reported in previous literature24. Age estimates were created from the mapped craters using CraterStats46. We used previous age estimates for Gale crater16,22, the Medusae Fossae Formation16, Meridiani Planum24, East Candor16, and Terby crater23.

For most mounds, the age range was determined by the age of the underlying surface (i.e. crater floor, canyon rifting), and the exhumation age of the mound. The likelihood of sedimentation was assigned a uniform probability of occurring between the age limits defined by these two bounding ages, and the error on each end was taken directly from the associated age distributions. In some cases, this method assumes that the underlying basin surface has been exposed for the majority of its geologic history, implying that the overlying deposits were not part of larger continuous deposits that have been eroded back. At Arabia Terra, we use the same Noachian aged basement as the underlying strata for the sulfate and rhythmites in Arabia Terra, even though it is speculated that the rhythmites may lie unconformably above the sulfate units of Meridiani Planum24. For Gale crater, we use the exhumation age of the entire mound as the top-age estimate, and the age of Gale itself, believed to be 3.6 Gyrs old22, for the underlying bounding age. Where erosion was accounted for in the age estimates16, these values serve as a constraint for the entire mound accumulation as opposed to exhumation and basin age. For these locations, we implemented the the age and error distribution directly from the model used for the source study.

Volume estimates were calculated in ArcGIS by interpolating basal surfaces from the mapped unit boundaries and largely follow published values47. For each boundary, the MOLA topography (MEGDR 128 pixels per degree)48 was extracted and used to interpolate a basal surface for that unit using Inverse Distance Weighted (IDW) interpolation. The volume, and maximum height of the units was then determined by subtracting the modern MOLA topography from the interpolated basal surface. For locations with rhythmite and sulfate unit exposures, such as Gale, the rhythmite portion was determined using the method above, and the sulfate portion was determined by subtracting the rhythmite volume from the volume of the entire mound.

Volume estimates are uncertain, especially for the Medusae Fossae Formation. This uncertainty is rooted in differences in mapping approaches between investigators, small variations in mapped boundaries, and, because this method requires interpolation of a basal surface, the interpolation method used. These uncertainties are biggest for the MFF for multiple reasons, including dust cover which hinders mapping by obscuring spectral signatures and primary orbital morphology. This uncertainty is demonstrated in Fig. S15. Because there are few ground-truth-based constraints on rock volume in the MFF, and the error-sensitivity of the value, we estimate the volumes for the MFF mounds have an uncertainty of roughly 50%, which is represented in Fig. 3 by the dashed lines. For similar reasons (e.g. dust cover49), smaller uncertainties also exist for the Arabia Terra deposits, particularly when accounting for the added volume of deposits in northeast Arabia Terra. This uncertainty amounts to  ~ 6% of the volume for Arabia Terra50.