Sustained fluvial deposition recorded in Mars’ Noachian stratigraphic record

Orbital observation has revealed a rich record of fluvial landforms on Mars, with much of this record dating 3.6–3.0 Ga. Despite widespread geomorphic evidence, few analyses of Mars’ alluvial sedimentary-stratigraphic record exist, with detailed studies of alluvium largely limited to smaller sand-bodies amenable to study in-situ by rovers. These typically metre-scale outcrop dimensions have prevented interpretation of larger scale channel-morphology and long-term basin evolution, vital for understanding the past Martian climate. Here we give an interpretation of a large sedimentary succession at Izola mensa within the NW Hellas Basin rim. The succession comprises channel and barform packages which together demonstrate that river deposition was already well established >3.7 Ga. The deposits mirror terrestrial analogues subject to low-peak discharge variation, implying that river deposition at Izola was subject to sustained, potentially perennial, fluvial flow. Such conditions would require an environment capable of maintaining large volumes of water for extensive time-periods, necessitating a precipitation-driven hydrological cycle. Using high-resolution orbital imagery of the Martian surface, the authors Salese et al. here describe the first discovered stratigraphic product of multiple extensive fluvial-channel belts in an exposed vertical section at Izola Mensa in the northwestern rim of the Hellas Basin.

W hile the present-day Martian surface is generally dry and cold, its geomorphic record contains compelling evidence for the former presence of liquid water [1][2][3][4][5][6][7] . In addition to this rich geomorphic archive, Mars' increasingly accessible sedimentary rock record provides a repository of information from which to study how planet-wide patterns in deposition have changed over time. From orbit, Martian sedimentary rocks have been observed for more than 20 years [8][9][10][11][12][13] , but detailed descriptions of large Noachian alluvial successions have so far been lacking. In fact, unequivocal sedimentarystratigraphic evidence of alluvium has only been identified in-situ by rover-led studies 14 , and whilst rovers provide unprecedented direct access to extraterrestrial strata 15 , methodologies are typically limited by accessible outcrop dimensions. For example, the Shaler outcrop, an interpreted Hesperian fluvial deposit identified by Curiosity 16 , comprises a single 70-cm-thick, 20-m-wide sandstone body. This is a crucial scale difference when compared with orbital investigation: the physical dimensions of most rock outcrops studied on the ground to date are smaller than the dimensions of many geomorphic components of even moderatesized extant river systems 17,18 .
In this study, we use high-resolution imaging science experiment (HiRISE) image (25 cm/pixel) and topographic (1 m/pixel) data to describe the sedimentary-stratigraphic architecture of a far-larger, 1500-m-wide, 190-m-thick sedimentary succession. The recently identified Izola outcrop is located in the northwestern rim of the Hellas basin (Fig. 1a), a~2000 km diameter impact structure containing a variety of 3.7 Ga Noachian Fe/Mg phyllosilicate-rich sedimentary intercrater plains, overlain by Hesperian-aged (~3.3 Ga) lava flows 19 .
These intercrater plains offer erosional windows which expose stratigraphic sections with well-preserved channel forms, and which must be older than the~3.7 Ga overlying plains 19 . The channel forms and associated sedimentary packages are interpreted as the product of an actively depositing fluvial system, with the final sedimentary architecture suggesting that these Noachian-aged rivers were not typified by high-energy episodic floods, but rather perennial or semi-perennial fluvial flow. The scale and completeness of the sedimentary succession offers a so far unique opportunity to assess the larger scale morphology of an evolving Noachian-aged fluvial system.

Results and discussion
The sedimentary succession. The outcrop exposes layered sedimentary strata, which display a variety of large-scale stratal architectures consistent with an alluvial interpretation. Alluvial sedimentary strata can be subdivided into genetically related three-dimensional packages [20][21][22] . The aggregate of these packages is referred to as the succession's sedimentary architecture, and is the product of the scale and behaviour of the fluvial system over time. The outcrop appears to have undergone little postdepositional deformation, has a gentle dip and large-dimensions, so is suitable for the analysis of sedimentary architecture ( Supplementary Fig. 5). A hierarchy of bounding surfaces is applied to divide stratigraphy, hierarchically ordered to reflect river processes at varying scales (Table 1, Supplementary Fig. 1). Two distinct sedimentary packages were recognised in the studied outcrop: (1) channelised packages; and (2) inclined accretionary surfaces. As details of sedimentary facies (usually discriminated by grain size and centimetre to decimetre-scale bedding and sedimentary structure) are unattainable without in-situ investigation on the ground, only the three-dimensional geometry of the sedimentary packages are described, with no details of internal structure attempted.
Channelised packages. In the observed stratigraphy, packages bound by lower erosional, channel-shaped (fourth order) surfaces and truncated by flat, erosional (fourth and fifth order) surfaces are attributed to channel-fill deposition (Figs. 2, 3b, Supplementary Fig. 2A) . They are 5-15 m thick, with observable lateral  extents of up to 210 m (although outcrop limitations restrict  observation of the full lateral extent of many channel forms:  Table 2, Supplementary Fig. 2). Internally, packages appear succession-dominated 23 , comprising multiple aggrading (third order) surfaces (Fig. 3b). Final channel banks and former channel margins coalesce, indicating that the original channels laterally migrated (Fig. 3b, d). Some channelised packages have a distinct channel wing (yellow asterisks in Fig. 2b; Supplementary Figs. 3,4), which may archive a genetically associated levee or crevasse and thus strengthen the alluvial interpretation. Particular areas of the outcrop show high concentrations of discrete, but partly amalgamated, channelised packages (Fig. 3b). In these areas, smaller channelised packages may be nested inside larger examples (Fig. 3b), implying that periods of net erosion locally occurred within a dominantly depositional regime. Areas containing clusters of channels are bottomed and topped by laterally extensive (up to 640 m), low relief, fifth-order surfaces, which suggest a change in the type or location of the dominant depositional process. These surfaces probably reflect channel avulsion, in which the location of the active channel changes abruptly. The sediments enclosed between these fifth-order surfaces were therefore laid down between avulsion events and are thus defined as channel belts 24 .
Inclined accretionary surfaces. In the observed stratigraphy, wedge-shaped packages comprise gently inclined depositional surfaces (third and fourth order; Fig. 3d). They are topped and bottomed by flat, erosional surfaces (both fourth order), except on occasions where they can be traced laterally into an associated channelised package (Fig. 3b). Most frequently only erosional remnants are preserved, with deposits often passing laterally into areas of non-exposure (Fig. 3d). Thicknesses range from 1.5 to 14 m, similar to the associated channels. Geometry and internal stacking patterns are suggestive of shifting fluvial barforms. For example, in some instances, bottomsets of fourth-order surfaces are downlapped by third-order foresets (Fig. 3d), demonstrating discrete intervals of bar-building. The outcrop orientation with respect to paleoflow is not well known, meaning it is not possible to build any unequivocal consensus on the direction of barform accretion with respect to the flow of the original channels (e.g., downstream accretion, lateral accretion). However, the identification of channel forms in conjunction with the outcrops alignment with the regional slope direction suggests a broadly depositional-strike succession.
Alternative explanations. Without in-situ validation of fluvial deposition other alternative possibilities must additionally be considered. In this section, we examine two alternative origins for the studied sedimentary succession: aeolian deposition, and deposition within submarine channels.
Aeolian: if these deposits have an aeolian origin, then the working hypothesis is that the channel forms are in fact localised scour fills formed by migrating aeolian strata, and that this succession records periods of aeolian accumulation punctuated by erosion or stagnation. However, there are disconnects between the stratigraphic architecture described here, and that of largescale cross-bedding typical of aeolian bedforms described on both Earth 25,26 and Mars 27 . No aeolian interpretation can explain the observable lateral migration surfaces associated with the discrete channelised packages (Fig. 3b). Multiple generations of aeolian dunes would partially erode preceding examples, perceived at outcrop by marked downlapping surfaces not apparent anywhere across the succession. Also, the fill geometry of aeolian dune deposits is often symmetrical and concordant with the trough base, unless the final outcrop is cut obliquely to the original flow direction, in which case planes are perceived to fill asymmetrically and downlap onto the trough base 28   across the outcrop (Fig. 2b, Supplementary Figs. 3, 4). Such elements can be variably interpreted as the topmost story of an individual channel or a genetically associated levee or crevasse 23 , but do not concur with any aeolian model. Submarine channels: certain depositional structures are found in both submarine and alluvial environments (e.g., channel wings), but several architectural characteristics in these deposits favour alluvial interpretation and suggest against a submarine origin. Evidence of lateral migration is widespread across the outcrop (i.e., inclined accretionary surfaces). Such an architectural style is more typical of alluvial settings, with submarine slope architectures more regularly dominated by vertical accretion 29 . Some inclined accretion surfaces additionally have distinct foreset and bottomset elements (Fig. 3d), consistent with those of fluvial barforms which scale to bankfull water-depth. Furthermore, most described submarine slope channels on Earth are at least an order of magnitude greater in size than the channels identified in this study 30 , which have dimensions consistent with   Table 2. The well-exposed channels are labelled with numbers and their parameters reported in Table 2. ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15622-0 many other geomorphic fluvial channels identified across the Martian surface 31 . Finally, regional observations of paleolakes occupying the Hellas basin at various stages in its evolution 32 are compatible with an alluvial interpretation for these deposits. Although a degree of ambiguity exists in any interpretation of ancient sedimentary strata, whether on Earth or Mars, and disputes between subaqueous and subaerial interpretations are commonplace on both planets 10,33,34 , considering the discussion herein, we argue that outcrop evidence strongly favours a fluvial origin for the described Izola architecture. demonstrate that preserved sand-body thickness is on average 30% of the original channel depth 35 , so this approximation is almost certainly an underestimate. These hypothesised waterdepths would be comparable with those of modern, mature fluvial systems on Earth draining extensive basins 36 , suggesting these Noachian-aged rivers had similar drainage capacities. Channels and barforms account for the entire observable stratigraphy. The evidence for channels being successiondominated (Fig. 3b) implies that entrenchment was quickly followed by bed aggradation 23 . The occurrence of distinct channel clusters is consistent with channel belts with relatively stable banks and limited lateral mobility 37 . The internal structure of the identified barforms comprises remnants of numerous phases of bar-building activity (Fig. 3d). On Earth, deposits of rivers with low-peak discharge variability preserve macroform structure, enabling the reconstruction of barform morphology 35,38 . The similar preserved architectural style seen here, of relatively intact barforms and channels, implies that the depositing Izola rivers were also characterised by low-peak discharge variability. Episodic flooding events may not have been responsible for deposition, but rather long-term, potentially perennial fluvial flow. This conclusion also indicates that the Izola rivers were not the product of meltwater flowing from a glacial front, as these environments are more widely associated with fluctuating, high discharges, and so would be unlikely to leave a stratigraphic record comprising relatively intact channels and bars.
Siliciclastic deposition on an unvegetated planet. Depositional models are largely based on sedimentary environments on Earth, where physical form and process is near ubiquitously influenced by biology 39 . Comparisons between the terrestrial and Martian sedimentary record therefore require careful consideration before application. For instance, copious observations demonstrate the various ways extant vegetation modifies fluvial processes and landforms 40,41 . Most important from an astrobiological standpoint is the likelihood of alluvial mud becoming preserved on an unvegetated planet. In addition to increasing mud production through chemical weathering 42 , terrestrial vegetation promotes mud retention on the continents through above-ground baffling and below-ground stabilisation 41,43,44 . Through these combined processes, an upsurge in mudstone abundance within alluvial stratigraphy is observed in stratigraphic alignment with evolving land plants 45 . Before this time, terrestrial alluvium is predominantly sand-grade or coarser with few preserved muddy floodplain facies 45,46 . Direct comparison between the prevegetation Earth and unvegetated Martian record might therefore imply that the studied Izola alluvium might also be lacking in preserved mudstone. Such tangible characteristics of the terrestrial pre-vegetation alluvial record have been used previously as supportive evidence for environmental interpretations elsewhere on Mars: for example, the scarcity of pre-vegetation terrestrial floodplain mudstones was used to recommend a lacustrine, not floodplain, origin of the Sheepbed mudstone at Gale crater 16 . The abundance of mudstone in the studied outcrop here is not known or speculated upon, as their intrinsic fine-grained components require rover-based observations. However, the preserved channel-belt architecture, comprising relatively intact channels and barforms, suggests some degree of original channel-belt stability, with naturally shear-resistant sediment such as mud a possible candidate 47,48 . With regard to channel and barform facies, evidence of relatively stable, deep-channelled drainage on pre-vegetation Earth is being increasingly reported 37 . Such findings are helping to dispel notions that pre-vegetation rivers were ubiquitously wide and shallow 49 , an observation that can now be extended to Mars (Fig. 2).

Preservation of time and implications for early Mars climate.
Disentangling the total duration recorded in any sedimentary outcrop is difficult as stratigraphic records are highly fragmentary, incomplete chronicles of time 50,51 . Despite uncertainty, broad approximations from the studied outcrop can be made which can inform ongoing debates about the early Martian climate 52 . The outcrop comprises at least four possible channel belts (Fig. 2), discrete packages of strata bound by laterally extensive (fifth order) surfaces and, on Earth, channel belts of this scale can require up to~10 4 -10 5 terrestrial years to deposit 52 (though we recognise that this approximation may differ under Mars' distinct boundary conditions). However, the amount of additional time hidden within the bounding fifth-order surfaces between individual channel belts is unknown (and unknowable) 53 . If a channelbelt interpretation is correct, the laterally extensive surfaces developed after the active channel-belt was transposed through avulsion. During this time, deposition was likely occurring elsewhere in the basin, and the region covered by our sedimentary outcrop was undergoing intervals of net erosion or stasis 54 . In other words, hiatuses in deposition in the studied outcrop were accompanied by deposition elsewhere in the basin. This implies that the time recorded in the 190-m-thick succession represents only a fraction of the total time fluvial deposition was ongoing in this region. The majority of strata and time will have either been lost to erosion, or preserved in outcrops as of yet undiscovered, or currently buried and not amenable to study. While we are only beginning to understand the chronostratigraphic exactness of sedimentary rock outcrops on Earth 53,55,56 , let alone Mars, it appears likely that the period of deposition in the northwestern rim of the Hellas basin exceeded 10 5 terrestrial years. Furthermore, the preservation of relatively intact channel margins and barforms advocates that throughout this protracted period fluvial deposition was a relatively constant phenomena 38 .
The architectural interpretation of this so far unique sedimentary succession feeds into ongoing debates about the early Martian climate. Our interpretation of long-lived, deep, perennial or semi-perennial rivers necessitates a climate in which active water-conduits were maintained for 10 5 years or longer. For the first time, orbital data has allowed us to examine, through detailed high-resolution architectural analysis, a large (1500 m by 190 m) pre-late Noachian outcrop, and draw reliable paleoenvironmental interpretations based on sedimentary-stratigraphic evidence. Our observations and analysis favour steady water discharges that are most consistent with a precipitation-driven hydrological cycle. This conclusion aligns with previous arguments for the prolonged presence of water on the early Martian  60 . Tie points were automatically populated in SOCET SET between the two images. We ran a series of bundle adjustments, removing erroneous tie points until the remaining points had an RMS pixel matching error of ≤0.6 pixels. The resultant DTM was then tied to Mars Orbital Laser Altimeter 61 topography and exported with a horizontal post spacing of 1 m/pixel and a vertical precision of~1 m.
Calculation of sedimentary package dimensions. The acquired HiRISE DTM of the Hellas outcrop was of sufficient resolution to enable accurate tracing of beds and plotting of architectural elements (Fig. 2). Channels thickness were obtained measuring the exact elevation of the channel top and the exact elevation of the channel base using ArcMap 10.6 elevation tools. Channel widths were measured by tracing an edge-to-edge topographic channel profile using the HiRISE DTM. This allowed calculation of true thickness, given that they are bound by laterally extensive fifth-order surfaces, which are almost flat lying ( Supplementary Fig. 5).
Architectural analysis. Line drawings were only attempted at areas where stratigraphy is clearly visible. Outcrop orientation with respect to paleoflow is not known, so no architectural elements with distinct directional components (e.g., downstream accretion, lateral accretion) were assigned. Sediment grain size is also unknown, thereby prohibiting the distinction between active and abandoned channel-fill deposits. The completed architectural panel enabled the various sediment stacking patterns and lateral relationships assessed.
Bounding surface hierarchy. A hierarchy of bounding surfaces was used to describe partitions of fluvial strata at outcrop, with different order surfaces reflecting river processes at varying scales 62 (Table 1, Supplementary Fig. 1). The principles of the hierarchical division applied here follows that detailed in previous papers 22,62 . Succinctly, zeroth, first and second order surfaces relate to foreset, set and coset boundaries, respectively, and are not observable from HiRISE imagery. Third-and fourth-order surfaces indicate the presence of macroforms (e.g., a barform deposit) or a channel. Fourth-order surfaces represent the upper and lower boundaries of the macroform or channel, whereas third-order surfaces relate to internal growth increments (indicating flow fluctuation, but no significant changes in predominant fluvial style). Fifth-order surfaces are the highest order observed at the studied outcrop and bind major depositional packages (e.g., channel belts). More fifth-order surfaces are likely present in the studied outcrop than highlighted on Fig. 2. This is simply because their confident recognition depends on an understanding of their relationship with lower order surfaces, and in some instances the vagaries of outcrop exposure (particularly towards the bottom of the section) prevent this.
Estimation of outcrop age. The age of the intercrater plains, which form the planform cover of the studied sedimentary-stratigraphic succession, has been estimated in a previous work 19 using various crater count techniques. They date 3.70 + 0.03/−0.04 Ga (Noachian). The latter represents the age of the surface at the top of the outcrop studied in this work, which lies in the S1 unit (Fig. 18 19 ). Crater counts were performed on Context Camera (CTX; 5-6 m/pixel) data using Crater Tools. Crater statistics and crater model ages were analysed with Craterstats2 software. For more details and references see a previous study 19 .

Data availability
The HiRISE data that support the findings of this study were obtained freely from the Planetary Data System (PDS) and are publicly available online at https://pds.nasa.gov/index. shtml. Satellite imagery and the Extended Data were generated with ISIS 3 (Integrated Software for Imagers and Spectrometers) available online at https://isis.astrogeology.usgs. gov. All these data were integrated into ArcMap 10.6 project. The DTM was produced using the USGS Integrated Software for Imagers and Spectrometers (ISIS 3) software and the BAE photogrammetric package SOCET SET with a post spacing of 1 m/pixel.