Rivers and Lakes in Western Arabia Terra: The Fluvial Catchment of the ExoMars 2022 Rover Landing Site

Oxia Planum, the landing site for the ExoMars rover mission, is a shallow basin on the southern margin of Chryse Planitia that hosts remnants of fan‐shaped sedimentary deposits associated with the ancient channel system Coogoon Vallis. This indicates runoff from a catchment in Arabia Terra has transported sediment into the landing site. To explore this fluvial system we created a model catchment for Oxia Planum and, using 6 m/pixel ConTeXt camera orbital remote sensing image data, we digitized the fluvial and lacustrine landforms in Western Arabia Terra in and around this catchment. We find: (a) The catchment has a minimum area of ∼2.1 × 105 km2 and has been episodically deformed by tectonic activity; (b) There were at least two phases of fluvial activity. The first created a mature landscape associated with Coogoon Vallis, which may have deposited alluvial or deltaic deposits in the Oxia Basin. After a substantial hiatus, a second phase of activity incised U‐section channels into the pre‐existing landscape and channel systems; and (c) Evidence for numerous possible paleolake deposits within the catchment. These are not well connected to the fluvial system and were probably sustained by groundwater activity contemporaneous with both phases of fluvial activity. This groundwater might have modified the mineralogy of Oxia Planum. Oxia Planum probably experienced an alluvial or distal deltaic/lacustrine depositional environment during the mid Noachian, which was later overprinted by a younger phase of fluvial activity.


of 27
Oxia Planum (Figure 1) is located at the northern edge of Arabia Terra at ∼20°W & 18°N where the ancient cratered terrains transition into the low-lying plains of Chryse Planitia. The landing ellipse for Rosalind Franklin is within a small (80 × 80 km) topographic basin (hereafter referred to as the Oxia Basin) within Late Noachian Highlands (Figure 1; lNh; Tanaka et al., 2014). To the south and west of the Oxia Basin are extensive clay-bearing units identified in Observatoire pour la Minéralogie, l'Eau, les Glaces et l'Activité (OMEGA; Bibring et al., 2006) and Compact Reconnaissance Imaging Spectrometer for Mars (CRISM; Murchie et al., 2007) infrared hyperspectral data (Figure 1, Quantin-Nataf et al., 2021). The north of the basin is open to Chryse Planitia, where the clay-bearing terrains are overlain by Chryse Basin fill materials (Greeley et al., 1977;Tanaka, 1997), and toward the southeast margin of the basin are several inlet channels associated with the remnants of several fan-shaped sedimentary deposits Molina et al., 2017;Quantin-Nataf et al., 2021). We make no interpretation about the genesis of these fans as alluvial or deltas and are hereafter simply referred to as "sediment fans." These sediment fan remnants appear to be associated with Coogoon Vallis and, although the exact relationship is obscured by Kilkhampton crater (IAU, 2019; Figure 1c), the channel systems that enter the Oxia Basin from the east, along with a group of unnamed channels from the south, are likely to have supplied water and sediment into the basin from Arabia Terra.
Valley networks on Mars, first observed by Mariner 9 (Masursky, 1973;Schultz & Ingerson, 1973) have long been considered evidence of overland flow and erosion by water in the ancient past (Ansan et al., 2008;Head, 2008, 2011;Grant, 2000;Howard et al., 2005;Hynek et al., 2010;Irwin & Howard, 2002;Mangold & Ansan, 2006;Pieri David, 1980). As our understanding of the variety of channel forms and their morphology has increased with improvements in data coverage and resolution Baker et al., 1983), it is now possible to pick out increasingly subtle landforms (Davis et al., 2019) and sedimentary structures (Salese et al., 2020;Skinner et al., 2021) that record fluvial systems throughout martian history. Recent observations of a regional network of exhumed channel-belt depositional systems demonstrate evidence for aggradational river systems in Arabia Terra, and their detailed geomorphology suggests they formed over substantial time periods as a result of regional precipitation Davis et al., 2016;Williams et al., 2009). Some of these inverted fluvial systems transition into erosional valley networks south of Arabia Terra, but it remains unclear how these ancient fluvial systems relate to Chryse Planitia, the natural sink for regional fluvial activity, but into which few channels seem to debouch.
The aim of this work is to better constrain the timing of the fluvial activity in the region, and the provenance of sedimentary material that may be present in the landing site. To do this we investigate the morphology and extent of paleo-fluvial landforms associated with Oxia Planum. Using the continuous coverage of ConTeXt camera (CTX; Malin et al., 2007) data to pick out increasingly subtle channel forms, we explore the density and distribution of fluvial geomorphology.
We also use a hydrological model to reveal the catchment of the contemporary Oxia Basin and compare it with ancient fluvial activity in the paleo-catchment to estimate the size of the paleo-catchment which drained into the Oxia Basin and the Rosalind Franklin landing site. We characterize the fluvial and lacustrine landforms observed within the model catchment and determine the relative timing and evolution of aqueous inputs into the Oxia Basin. Finally, we consider the evidence for lacustrine activity and consider the implications for biosignature formation, transportation and preservation, the prime targets for the Rosalind Franklin rover, from the fluvial history of the catchment.

Data
Geomorphological observations of fluvial features were made using CTX 6 m/pixel data at a scale of 1:50,000, georeferenced to High Resolution Stereo Camera MC11 quadrangle mosaic basemap (HRSC; Gwinner et al., 2016;Neukum et al., 2004). THermal EMission Imaging System (THEMIS; Christensen et al., 2013) day and night IR controlled mosaic of Oxia Palus (00N, 315E 100mpp) were used to inform identification of features observed in the CTX data, and Color and Stereo Surface Imaging System (CaSSIS; Thomas et al., 2017) images were used where available for color interpretation. Mars Orbital Laser Altimeter (MOLA; Zuber et al., 1992) data were used for topographic information. Using these data, a fluvial (valley/channel and sinuous ridges) and 10.1029/2021JE007045 4 of 27 lacustrine features was identified. After the initial survey, a topographic flow accumulation model was used to identify areas to revise where the model suggested channels might be present, and these were then searched more closely for any subtle morphological evidence of fluvial landforms. This iterative, multi-data process enabled many more fluvial systems to be identified than using one data set alone. The result from this survey can be accessed in Fawdon et al. (2021).

Watershed Calculation
To determine the watershed area for Oxia Planum (Figure 1), the ArcMap 10.5 Spatial Analyst "ArcHydro" toolset (Esri, 2016) was used to calculate a model of flow accumulation grids and a drainage network map using topographic data from the MOLA DEM (Smith et al., 2001). Areas in the DEM that created sinks or basins were filled prior to calculating flow direction and accumulation. It is important to note that these processing steps "fill in" areas of low-lying terrain and impact craters, as well as unwanted error and noise in the DEM. These "filled in" areas create model flow pathways stretched across basins that were retained to identify where ponding may have occurred and where the model flow is likely to deviate from the geomorphic observations. The watershed and contributory areas were calculated using the D8 flow direction and flow accumulation model upslope of two pour points located in the Oxia Basin. The location of both pour points (see Figure 2) was based on the correspondence of preliminary model flow accumulation paths calculated for the whole MC-11 Quad and geomorphological features resolved in the MOLA DEM. The eastern pour point (the lowest point in the "fan" watershed) was located where the channel of Coogoon Valles opens out into the Oxia Basin at the highest elevation of the sediment fan remnants. The northern pour point (the lowest point in the basin watershed) was located at the lowest point of the Oxia Basin leading northwards to Chryse Planitia. The watershed is defined where the flow accumulation is 0 (i.e., there are no cells from which water would flow). Both D-infinity and D8 flow direction and accumulation models were tested. We found no points where D-infinity flow accumulation pathways crossed the watershed delineated from the D8 flow accumulation model no difference in model watershed area produced between the two methods. The pour points, their watersheds, and the flow accumulation pathways were converted to Strahler stream order (Strahler, 1957) and are shown in Figure 2.

Reconstruction of the Palaeo-Catchment
We used surface flow accumulation model (Esri, 2016) to estimate the catchment area of the Oxia Basin and to constrain the extent of our geomorphological study. The hydrological flow accumulation calculations show a total model catchment area of 2.1 × 10 5 km 2 ( Figure 2). This represents the total upslope area (including the Oxia Basin) from which water would flow over the contemporary topography though the Oxia Basin and northwards to Chryse Planitia. From the catchment area 71% (∼1.5 × 10 5 km 2 ) accumulates into the Oxia Basin through the pour point labeled "fans" (Figures 2 and 3) which is coincident with remnant sedimentary fans (Molina et al., 2017;Quantin-Nataf et al., 2021). 26% (∼0.6 × 10 5 km 2 ) of the catchment contributes to the Oxia Basin from the south ( Figure 2). This area is at lower elevation with low-relief relative to the eastern part of the catchment flow, where accumulation is diverted by tectonic ridges (e.g., Figure 1b). The remaining 3% of the catchment represents the Oxia Basin itself (∼80 × 80 km). This is included because it has not yet been established if the Oxia Basin hosted a standing body of water at any time, and whether this postulated water body was isolated from-or contiguous with-a hypothesized northern ocean (Carr & Head, 2003;Di Achille & Hynek, 2010;Dickeson & Davis, 2020;Parker et al., 1993).
The margins of the catchments are defined by three types of geomorphological structure (Figure 2): (a) Tectonic ridges influence model flow accumulation pathways in low-lying plains regions, particularly south of Oxia In context with regional geological units (Tanaka et al., 2014), previously reported "crater lakes" and "valley networks" Hynek et al., 2010). (c) The local topographic context of the 2022 landing ellipse area (3 sigma confidence landing probability ellipse shown for the opening of the launch window) in Oxia Planum including; "The Oxia Basin",  detections of phyllosilicate minerals made by the CRISM and OMEGA instruments shown in red , Coogoon Vallis , the ejecta of Kilkhampton crater (gray) the −3,000 m MOLA elevation contour and sediment fan remnants reported in (Molina et al., 2017 Planum and in the northwest of the catchment (Figures 2b and 2c). In the north of the catchment (Figures 2b  and 2c), a tectonic ridge that forms the boundary of the catchment is crosscut by a narrow branch of Coogoon Vallis. (b) Throughout the catchment, impact craters form local sinks that have been filled by the watershed model (Figuer 2a). For some craters, modeled flow accumulation is concordant with the fluvial morphological features present on the surface, but for others it is discordant with fluvial features. (c) The model flow accumulation also follows a number of early Noachian highland massifs (Figures 1b, 2c, and 2d;Tanaka et al., 2014) with valleys on their flanks located in the south and far west of the catchment (Figures 1 and 2).
Our preliminary interpretation of the model catchment is that it is generally representative of the Oxia Planum paleo-catchment, as indicated by the general concordance between the flow accumulation model, and the previously reported fluvial features in ancient highland terrains ( Figure 1). However, the catchment area must be considered as a minimum estimate because there has been tectonic activity and "modern" impact structures' which have strongly influenced the topography, and therefore affected the modeled flow accumulation. A good example is the observation showing a branch of Coogoon Vallis (as recorded in Hynek et al., 2010) traversing a modeled watershed boundary (Figure 2b).

Geomorphic Evidence for Ancient Fluvial Activity
We used orbital images to explore how extensive the fluvial network that fed Oxia Planum was, and to test the veracity of the reconstructed paleo-catchment. We find evidence for 4 scales of valley/channel features in the Oxia catchment. In addition to the trunk valley Coogoon Vallis, we identify three types of incised paleochannels within the paleo-catchment area which transported water towards the Oxia Basin: (a) Wide flat-floored channels which are set within broad valleys in the landscape; (b) U-section channels which crosscut the older flat floor trunk channel of Coogoon Vallis; (c) Low-relief channels/valleys which are the smallest features resolvable in the CTX data, and that are also often set within broad valleys in the landscape. These features are described below.

The Relationship Between Oxia Planum and the Trunk Valley of Coogoon Vallis
The main contributory input to the Oxia Basin is from the east through the Coogoon Vallis fluvial system (Figure 3a). However, the relationship between the channels/valleys, sediment fans, the flat floored part of Coogoon Vallis and Kilkhampton crater has considerable stratigraphic complexity (Molina et al., 2017).
East of Kilkhampton crater, Coogoon Vallis is a ∼6 km wide, filled or flat-floored structure (which we refer to as "Paleo-Coogoon" Vallis; Figures; 2 and 3a, c) but this Paleo-Coogoon valley is not observed west of Kilkhampton crater ejecta. West of Kilkhampton crater are remnant sediment fan systems (Molina et al., 2017;Quantin-Nataf et al., 2021). These are associated with three narrow channels >1 km wide to the North-and South west of Kilkhampton crater ( Figure 3d). These narrow channels (along with Paleo-Coogoon) are collectively named "Coogoon Vallis" in Hynek et al. (2010) but to differentiate them we name them "North", "central" and "south Neo-Coogoon" (Figure 3a).
An important stratigraphic observation is that North Neo-Coogoon underlies the ejecta of Kilkhampton crater and is incised into the floor of "Paleo-Coogoon" Vallis ( Figure 3d), therefore postdating it. This demonstrates that the system supplied water and sediment to Oxia Planum at different times and that the sediment fan remnants in Oxia Planum are not associated with the development of Paleo-Coogoon which represent a substantially larger, and older, phase of fluvial activity.  The widest flattest part of Coogoon Vallis referred to as "Paleo-Coogoon" (d) the three Neo-Coogoon Valles are obscured by the ejecta of Kilkhampton Crater in CTX DTM data, however (e) these channels, which are "U" shaped in profile, incise (and thus postdate) the paleo-Coogoon Vallis as seen in CTX image G01_018662_1981_XN_18N022W.

Wide Flat Floored Channels/Valleys
Wide, Flat-Floored channels/valleys occur within the catchment (Figure 4a). The largest example trends east from the trunk valley Palaeo-Cogoon Vallis ( Figure 3c) where this channel is ∼5 km in width and up to 150 m deep. (Figures 3 and 4). To the east the channel narrows up slope to ∼1 km in width and ∼50 m in depth where it is intersected by subsidiary channels at high junction angles. Down slope, towards Oxia Planum, this channel opens out, and the margins become less distinct. Here, it is >10 km in width with a smooth floor that appears to be infilled by a younger dark-toned unit. Other wide (>1 km) channels with shallow banks elsewhere in the catchment are much less extensive. Wide flat floored channels lie in valleys which conform to the long wavelength (10's km) topography of the catchment (Figure 4c).

U-Section Channels
U-section channels are found in the northwest of the model catchment ( Figure 4a). They are generally ∼50-100 m in depth ( Figure 4) and ∼500-1,000 m wide. Their topography is significant enough to be captured in the flow accumulation model results. U-section channels incise the valleys of wide flat floored channels and their upslope terminations are often abrupt, forming "amphitheater" shapes . Unless crosscutting the floor of an older channel, U-section channels are not situated within their own valleys ( Figure 4). U-section terminations often trend into Low-relief Channels/valleys (Section 3.2.4) upslope. In some cases, U-section have been deformed by later tectonic activity ( Figure 2c).
The cross-cutting relationships show that U-section channels were active after landscape modification associated with the Palaeo-Coogoon channel. U-section channels appear to be superimposed on this older landscape, reactivating the older flow pathways; their relatively straight courses and sharp terminations (Figures 4g-4i) suggest that they were not active for long enough to equilibrate with the pre-existing topography. U-section channels predate the most recent stages of regional tectonic deformation (Figure 2b), so the elevation of their terminations, and their present-day topographic long-profiles, might not reflect their original elevations when they were active.

Low-Relief Channels
Low-relief channels Figure 5 incise plains materials ( Figure 1b; unit mNh) throughout the model catchment. They range in width from ∼500 m to less than ∼100 m and their depths are estimated to be <10 m based on the limited CTX DTM coverage. At length-scales smaller than this, the expression of low-relief channels becomes muted and it is no longer possible to clearly observe them. Low-relief channel generally occur as upslope continuations of U-section and wide flat floored channels, or as lower order branches that feed into U-section or wide flat floored channels. Low-relief channel are also often observed upslope of amphitheater shaped onset points of U-section channels ( Figure 4). Low-relief channel have a low sinuosity often branch up slope into smaller channels (Figure 5d) and also trend into (and from) low-relief sinuous ridges ( Figure 6).
Low-relief channels represent the smallest, narrowest channels in the network that formed contemporaneously with wide flat-floored valleys. Low-relief type channels were probably subsequently reactivated in association with the development of the U-section channels as they are observed to merge into channels of this type. We interpret that the minimum observable size is related to the effects of erosion (e.g., by wind and small scale impacts [Golombek et al., 2014]) and burial or regolith development since they were last active. Where low-relief channels trend into sinuous ridges, this suggests that that these channel were part of an aggregational alluvial system, so they have deposited sediment in places and did not just incise a pre-existing surface.

Sinuous Ridges
In addition to the network of incised channels, a network of sinuous ridges are found in the west of the catchment (Figures 5a and 6). Sinuous ridges are up to ∼500 m in width, are estimated to be up to 10 m in height and often form as continuations of low-relief channel channel forms. Sinuous ridges are found in areas with "inverted" crater fills and eroded remnant mounds (McNeil et al., 2021). This suggests that substantial burial and erosion has occurred. Sinuous ridges are also associated with light toned fractured terrains. Sinuous ridges exhibit branching patterns (Figure 6e), as well as a variety of surface textures, and are overlain by the regionally significant rounded buttes (Figure 6e) part of a significant regional population (McNeil et al., 2021).
We interpret the sinuous ridges as inverted fluvial channel deposits (e.g.,Figures; 5a, 6; Pain & Ollier, 1995) comparable to similar features seen throughout western Arabia Terra (e.g., Balme et al., 2020;Davis et al., 2016, Davis et al., 2019Day et al., 2019;Edgett & Malin, 2002;Newsom et al., 2010;Williams et al., 2009). We attribute the variation in occurrence to regional variations in exhumation processes and degree of erosion. To the northwest of the catchment where there has been the most erosion, sinuous ridges predate the formation of the rounded buttes ( Figure 6e) which are likely to have been part of a regionally extensive layer (McNeil et al., 2021). This means the material of the ridges (i.e., fossilized stream beds, assuming the widely accepted interpretation) are at a similar stratigraphic level to the clays (lNh, mNh; Figure 1) that comprise Oxia Planum Mandon et al., 2021;Quantin-Nataf et al., 2021) consistent with other exhumed channel systems in the region (e.g., Balme et al., 2020;Marzo et al., 2009).

Geomorphic Evidence for Paleolakes in the Region of the Model Catchment Area
Within the model paleo-catchment, 180 locations show possible evidence of liquid water having ponded in the past as "paleolake basins". The strongest evidence to support a paleolake basin interpretation is the presence of a topographic basin associated with an overspill outlet channel (e.g., Figure 8) which suggests that water once filled the basin to at least the level of the overspill (e.g., . A variety of other landforms are often observed in these paleolake basins (Cabrol & Grin, 1999;Fassett and Head, 2008a;Salese et al., 2019;Wharton et al., 1995) in addition to overspill channels, so we used the following evidence observed in the catchment to infer the presence of paleolakes: (a) smooth layered crater floor deposits not associated with evidence for volcanic architecture (Cabrol et al., 2003;Hauber et al., 2011;Salese et al., 2019); (b) sediment fans (e.g.,; Goudge, Aureli, et al., 2015;Hauber et al., 2009;Palucis et al., 2014); (c) presence of interior crater wall channels with circumferential interior crater wall slumps (e.g., Salese et al., 2019) and (d) possible strandlines (e.g.; Cabrol & Grin, 1999;Komatsu et al., 2001;Le Deit et al., 2013;Ori et al., 2000;Salese et al., 2019).
We compiled an initial catalog of possible paleolake locations from impact structures identified by (Robbins & Hynek, 2012a, 2012b, our survey of the THEMIS and CTX data and locations where the slope of a "filled" (Esri, 2017;Planchon & Darboux, 2002) MOLA DEM was 0°. We consider possible paleolakes as any location that exhibited at least one of these pieces of morphological evidence for ponded water when found in association with a plausible topographic context. We also consider CRISM/OMEGA observations associated with geomorphic evidence (Carter et al., 2015;Wray et al., 2011) as supporting evidence for lake activity when identified in context with an assemblage of the other landforms. Based on the presence of these landforms, the degree of evidence for ponded water and their size, we have divided the possible paleolake basins in the model catchment (  Table 1.

Large Crater Lakes
Large Crater lakes Figure 8 are the class of larger impact crater which show the most compelling evidence for being ancient lakes. They have a mean area of ∼1,100 km 2 and mean diameter of ∼33 km and retain an upstanding crater rim. Of the 12 examples observed in this class, eight have identifiable inlet valleys and four of these also have outlet channels exiting the crater. Large crater lakes also often (8 of 12 examples) have channels on the inside of their crater walls (Figure 8b). Four examples of large crater lakes form in a chain of craters ( Figure 8). In this chain, the western-most crater has an outflow channel (Figure 8c), the eastern most crater has inlets from highland to the south-west and outlets to one of the central craters (Figure 8d), although this is not directly connected   The presence of Inlet and outlet channels show that many of these craters hosted at least one instance of a lake. The layered deposits on the floor of one crater, in which Carter et al. (2015) identified the spectral signatures of phyllosilicate minerals, give us confidence that similar floor morphologies (i.e., all large crater lakes and rimless crater lakes) are also likely to constitute aqueously altered sedimentary deposits. Where large crater lakes do not have outflow channels but do have interior channels predominantly below the elevation of the surrounding plains. We interpret this morphology to have formed by flow of water through the crater walls, because the groundwater table was likely above the crater floor. Consequently, several instances of progressively shallower high stands and lake depths could form with fluctuating groundwater levels without breaching the crater rim. No evidence of volcanic architecture was seen to suggest the craters were flooded by lava.

Rimless Crater Lakes
Rimless Crater lakes are a class of smaller basins that are flat-floored, quasi-circular depressions with little or no topographic rim (Figure 9). Some also have inflow and or outflow channels (only 17/52 examples; Figure 9a).  We created 24 crater rim-floor profiles, extracted from 100 m/pixel HRSC DEM (Gwinner et al., 2016) data, of six rimless crater lakes basins and six other proximal (within <100 km) "reference" craters that do not appear to have hosted lakes. All examples are ∼10 km in diameter (Figure 9d). The profiles show that rimless crater lakes craters have lost ∼200 m of rim height and 600 m from the rim to floor depth compared to the reference crater population, which did not host any paleolake criteria morphologies. Accounting for the change in rim height the rimless crater lake basins also have been infilled by ∼300-400 m of material not present in other craters in the same geographic area.
We interpret rimless crater lakes to be the remnants of substantially degraded impact craters where the majority or all of the impact crater rim has been removed and the floor infilled by hundreds of meters of material. The removal of the crater rim is consistent with them being part of a mature (Craddock et al., 2018) mid-Noachian landscape (Craddock & Maxwell, 1990).
We interpret the smooth flat floors and interior topographic steps/breaks in slope as "strand lines" indicating deposition conforming to an approximately equipotential surface. Where these features are found in association with inlet channels and sediment fans, we are confident that the rimless crater lakes have hosted instances of paleolakes at some point in their history. The lack of outlets associated with rimless crater lakes means they did not overflow and are not connected by overland flow into the regional fluvial systems (Section 3.2). This means lake depth and longevity is likely to have been controlled by the depth of the crater and height of the groundwater table.

Irregular Dark Depressions
The most numerous class of potential paleolake landforms are irregular dark depressions (IDD; Figure 10). In CTX images, these terrains are predominantly dark in tone and occur either on the floor of impact craters associated with rimless crater lakes or large crater lakes (Figures 4a and 4b) or in topographic lows in intercrater plains. Where IDD occur in inter-crater plains they are often associated with channels or sinuous ridges. The best example of this (Figures 7f, 10a, and 10b) has one very clear outlet which leads to Coogoon Vallis, and several possible inlet channels or sinuous ridges. The margins of IDD appear to follow specific contours unless they have been crosscut by tectonic. They often show concentric albedo bands at their margins and have a strong association with periodic bedrock ridges (e.g., Favaro et al., 2021;Silvestro et al., 2021). These bands comprise a bright fringe at the edge of the feature, and a darker band inside, which transitions to a tone comparable to that of the surrounding plains in the center of the feature. There is often a topographic step at the edge of each of these bands, which appears bright in night time THEMIS data. This is most noticeable where IDD are also found in association with sinuous ridges (e.g., Figure 6). Where HiRISE image data and CTX topographic data are available (Figures 10c  and 10d) the darker irregular dark depression material has distinct layers ( Figure 10d) and forms a sharp boundary with the underlying light-toned, clay-bearing crater floor material (see Figure 8).
We tentatively interpret the IDD recorded in this study as paleolake or playa deposits based on their location in the topography, and their association with the regional network of channels and ridges. We interpret the dark deposit to be dark previously unconsolidated transported material cemented in local topographic lows at times of highstand of the groundwater table. We attribute the concentric bright and dark rims to be this darker capping material and a brighter, less erosion resistant underlying material that is undergoing a terrain inversion process. However, there are many processes that could produce these features, and many examples of dark materials in different contexts that are not associated with topographic lows for which we do not explore an interpretation. For example, similar features have previously been interpreted as remnants of mega-tsunami runout deposits (Rodriguez et al., 2016) or, because of their pyroxene-dominated minerology, as volcanic deposits (Ivanov et al., 2020;Quantin-Nataf et al., 2021).

Distribution of Channels and Lakes
The distribution of channels and possible paleolakes (Figures 4a, 5a, and 7) approximately divides the region into zones based on the pattern of fluvial and lacustrine features in the landscape. In the Eastern Zone channels form dendritic patterns branching out from Coogoon Vallis. When traced upslope, the channels transition in morphology from wide flat-floored to low-relief-channels, or from wide flat-floored to U-section and to low-relief channels with the U-section channels superposing the wide flat-floored channels (Figure 4). The topography of the east quadrant slopes down towards the channels but the paleolakes generally occur at higher elevations. In the Southern Zone the channels are predominantly low-relief channels and form a discontinuous pattern. The trends of this pattern are associated with the slopes of the mid Noachian (Figure 1) highland massifs to the south of the quadrant. In the Western Zone the topography forms a low-lying and low-relief plain that is crosscut by the regional north-south tectonic fabric, and here the craters are often infilled or "ghost" craters picked out by the ridges. To the north of the West Zone, a network of contributory low-relief channel and sinuous ridges extend out of the model catchment south of Oxia Planum, whilst in the south, numerous IDD appear to be constrained by the tectonic ridges. In the center of the catchment (the east of the West Zone) is a large irregular dark depression associated with outflow from the central chain of craters ( Figure 8) and U-section channels in the south of the Oxia Zone. The Oxia Zone, containing the ExoMars landing site is a shallow basin (Figures 1 and 7) is located at the confluence of drainage from the East and West Zones. The Oxia Basin is fed from the East zone by Coogoon Vallis, U-section channels which superpose Coogoon Vallis ( Figure 3) and from the western zone by U-section channels which appear to superpose an westwards trending the pattern of sinuous ridges and low-relief channels.

The Catchment Area of Oxia Planum
The catchment represents the source area from which the fluvially transported sedimentary deposits found in Oxia Planum are likely to be derived. The modeled catchment area of 2.1 × 10 5 km 2 is larger than that of other basins investigated by rovers such as Jezero crater (∼0.3 × 10 5 km 2 ;  or Gale (Ehlmann & Buz, 2015), but similar in scale to the watershed of Hypanis Valles (∼4.6 × 10 5 km 2 ; Fawdon et al., 2018) and Gale crater (∼1.2 × 10 5 km 2 ; Thomson et al., 2019). To consider the validity of the model catchment we compared the model flow accumulation pathways and watershed boundary to geomorphological observations on orbital images. The low-relief channels and wide flat floored channels match the model flow accumulation pathways across hundreds of kilometres showing that these ancient channels are in good agreement with the current topography ( Figure 2).
The ArcHydro-derived flow accumulation network deviates from the network defined by mapping where U-section channels incise older valley landscapes or are deformed by tectonic features. Channels cross into the model catchment from the north, and exit to the west (Figures 2, 3e, 7, and 10). This also suggests the topography must have changed since the formation of the channels. In both instances, tectonic "wrinkle ridges" deform the pre-existing flow pathways and have created tens to hundreds of meters of relief. In the western region of the model catchment ( Figure 6) degraded or inverted impact structures, sinuous ridges and remnant rounded buttes ( Figure 6) indicate tens of meters of erosion have altered the topography. This mean the contemporary topography does not represent the paleo-surface into which the channels were eroded and negates the finer-scale validity of

The Timing of Fluvial Activity
Within the model catchment, there is a partially exposed channel network containing three different styles of erosional channels, and a set of sinuous ridges interpreted to be inverted channels or small exhumed channel belts. Both the largest wide flat-floored channels, and smallest (low-relief channels) examples of channels, fit within the shape of the present-day regional topography over length scales of hundreds of kilometres but show local deviations from the present day landscape over length scales of tens of kilometres, where wrinkle ridges modify these channels. In contrast, the intermediate sized U-section channels commonly appear incised into the landscape with well-defined breaks in slope along their banks (e.g., Figure 4). U-section channels superimpose both the floor and banks of wide flat floored channels, as well as occurring along the path of low-relief channels, but the opposite relationship is not observed. This means that U-section style channel formation generally postdates both the formation of the wide flat floored channels and the initial formation of low-relief channels. The similarities between their flow paths suggests that many U-section channels are reactivations of the low-relief channel flow paths and overprint the medial section of ancient river profiles now represented only by the wide flat floored channels downstream, and low-relief channels upstream.
The interpretation that the flow paths of the low-relief channels were formed in an ancient landscape is further supported by their along-track transition into sinuous ridge inverted channels. Sinuous ridges are observed in the western portion of the catchment, which has experienced the most erosion ( Figure 6). Such channel inversion of fluvial channels can occur by a variety of processes, including cementation of fluvial sediments to create differential erosion resistance between coarser grained channel deposits and finer grained overbank deposits, or later infill of channels and armoring by some subsequent volcanic or impact-sourced materials (e.g.; Pain & Ollier, 1995). All models require later erosion to leave the resistant channel body upstanding in the landscape. Sinuous ridges in the model catchment are not seen to be associated with any volcanic features or impact ejecta, suggesting that the relief inversion has occurred over a wide region in a similar style. This further implies that  Note. The frequency of observed geomorphological evidence for paleolake activity. Note, these features are not weighted by importance. The full data table is included in the supplementary material as text and GIS-ready formats .

Table 1
Summary of Landforms Observed in the Possible Paleolakes the differential erosional properties of the channel fill are likely related to endogenic factors (e.g., grain size or cementation) in the channel fill and are an integral part of the ancient landscape (Williams et al., 2009) rather than exogenic factors such as volcaniclastics or impact ejecta. A possible exception is an example of a sinuous ridge trending into a irregular dark depression (Figure 3e).
The implication of these observations for fluvial activity in western Arabia Terra is as follows: first, that there was a period of fluvial activity during which Coogoon Valles developed within the ancient topography of the region (Figure 5a). Subsequently, and likely after a hiatus when impact craters accumulated (Quantin-Nataf et al., 2021), a second phase of fluvial erosion led to U-section channel morphologies being superimposed on the pre-existing network (Figure 5c). This later period of incision was likely short-lived because the U-section channels cut into a more mature landscape (Burbank & Anderson, 2009;Howard et al., 2005) but have not reached a sufficient stage of landscape maturity to erase the landscape associated with Coogoon Vallis. Sinuous ridges occur at the same stratigraphic level as low-relief channels so the fluvial activity they represent occurred around the same time as the first stage of development of Coogoon Vallis. To date, no regional stratigraphic markers have been identified that correlate this period to the formation of the many other inverted or exhumed channels systems observed in Arabia Terra (Davis et al., 2016(Davis et al., , 2019, but it would be parsimonious for both sets of channel system to be co-evil in the mid-to late-Noachian Davis et al., 2019).
These relationships predict that sediment fans associated with U-section channels should postdate the Noachian landscape (which is primarily associated with wide flat floored and low-relief channels). This is validated by observations in the Oxia Basin where sediment fans associated with U-section channels superpose both wide flat floored channels and impact craters associated with older surfaces . The stratigraphy also supports a relationship between this initial stage of widespread fluvial activity, associated with the wide flat floored/low-relief channels that developed with the regional relief during the mid/late Noachian, and the regional spectral evidence of hydrated minerology and weathering profiles found in the region (Carter et al., 2015;Loizeau et al., 2007;Michalski & Noe Dobrea, 2007;Noe Dobrea et al., 2010;Poulet et al., 2020). This allows for phyllosilicates in the Oxia Basin to be either authigenic or allogenic depending on rates of alteration and sediment transportation.

Possible Lacustrine Activity in Craters
Within the modeled catchment area we identified two types of potential paleolake basins associated with impact craters and inter-crater topographic lows. Each of these types (Figures 8-10) have distinct rim morphologies and show a suite of landforms indicative of lacustrine activity (Table 1) similar to criteria from the global study of Salese et al. (2019). The landforms we can observe are dependent on the preservation, exposure and context of the lake. Consequently not all possible palaeolakes exhibit every diagnostic landform.
Inlet and outlet channels are not common in large crater lakes or rimless crater lakes, despite this, the evidence from other features that supports a lake interpretation reinforcing the conclusion that groundwater has been an important driver of lake stability in the region (Michalski et al., 2013;e.g.,Salese et al., 2019;Schmidt et al., 2021). Inlet channels are seen where craters are located on a topographic rise with a large enough upslope area catchment. Few rimless crater lakes have outlet channels, and only one large crater lake has an outlet channel ( Figure 8). This suggests that the formation of outlet channels depends on the physiographic setting of these lakes (i.e., if the groundwater-supplied lake could be episodically supplemented by overland flow; Figure 11). If the large crater lakes and rimless craters with no channels did host lakes at the same time as those in adjacent terrain, then were predominantly supplied by groundwater.
An important case is a chain of four craters (Figure 8c). Here, four interconnected ( Figure 8f) large crater lake craters were identified that posses a suite of consilient evidence supporting lacustrine activity (Figure 8d), including a channel breaching a crater rim (Figure 8e). The terrain slopes away from the western end of the channel in all directions (and there is some evidence for a sediment fan) so it is not topographically plausible for this channel to be an inflow channel. This interpretation is supported by the linear elongate streamlined nature of this channel, which we interpret as a single outflow that was not followed by any subsequent persistent or episodic channel erosion. The importance of this feature and these craters is that the water level throughout the crater system must have reached the level of the outlet spillway at least once. A suite of other observations (strand lines, hyperspectral observation of hydrated mineralogy in the layered floor, and a sediment fan) further support the conclusions Figure 11. Sequence of events. Schematic showing the relationship between the timing of channel incision, putative lake activity, groundwater level, tectonic subsidence and erosion in western Arabia Terra. that these craters host paleolakes comparable the interpretation for the nearby McLaughlin crater (Michalski et al., 2013).
From this, we interpret that large crater lakes and rimless craters hosted predominantly groundwater-controlled lakes at times when climatic conditions were conducive to a combination of high groundwater levels (Andrews-Hanna et al., 2010;Pondrelli et al., 2019;Schmidt et al., 2021) and supplementary overland water supplied through the regional fluvial network. The younger phase of fluvial activity is likely more closely associated with craters hosting well-preserved feature associated with this more intense valley network formation (e.g., Howard et al., 2005;Irwin et al., 2005) but without detailed study and modeling of individual lakes and local groundwater (e.g., Andrews-Hanna et al., 2010) we cannot say when the contribution maintained this lake activity.

Inter-Crater "Lakes"
Irregular Dark Depressions ( Figure 10) often overlie or infill channel features, form parts of the floor in other larger crater paleolakes, or are found on plains between craters. On inter-crater plains, IDD are often associated with sinuous ridges where the dark material of the IDD trends into the inverted channel's topmost resistant layer ( Figure 6). This indicates that the infilling or armoring of the channels occurred before the differential erosion that created the sinuous ridges. This erosion has most affected the western part of the catchment, especially in and south of Oxia Planum. Irregular dark depressions have a dark surface which often retains many small craters, indicating that this material type is less susceptible to the processes that erase the smallest part of the crater population (Golombek et al., 2014;Warner et al., 2015) than the surrounding terrains.
Where irregular dark depression like material is identified in Coogoon Vallis and in the north-west of the study area it has been interpreted as tsunami runout deposits (Molina et al., 2017;Rodriguez et al., 2016) that have ponded in topographic lows. Previously, similar surfaces have been attributed to a variety of depositional processes, including as extrusive volcanic materials due to their commonly observed mafic spectral signatures in Jezero crater  and Oxia Planum . However, a volcanic interpretation is not necessarily appropriate (e.g.; Edgett, 2005;Edgett & Sarkar, 2021) and in this case is inconsistent with the distribution and variety of contexts seen for IDD seen in the Oxia Basin catchment: none of the 109 examples within the study area show evidence for volcanic architecture that might be expected from Amazonian volcanism (Brož et al., 2017).
We propose that these dark deposits, when located in crater floors and associated with the channel network, represent allochthonous mafic volcaniclastic/aeolian materials (Edgett & Sarkar, 2021) that have been cemented by highstand of groundwater activity (Andrews-Hanna et al., 2007;Schmidt et al., 2021). The deposits accumulated in topographic lows associated with the ancient low-relief/wide flat floored channel network, which could have provided hydrological corridors within the regional groundwater flow (Andrews-Hanna et al., 2007, or mafic enrichment in a desert pavement-like process (Irwin et al., 2018;Rogers et al., 2018). We do not presume that all deposits exhibiting these properties formed through the same process, or even concurrently, and understanding the context of such deposits is key to understanding their origin; many relief-filling, low albedo deposits occur on Mars. Furthermore, the process we propose may have occurred at several instances in time and is probably inconsistently preserved within the study area.

Implications for Mars's Geological History
Our study explores fluvial and lacustrine systems in Western Arabia Terra. From this we have established a sequence of events which constrain the hydrological development of Oxia Planum (Figure 11), spanning events from the mid Noachian to Amazonian in Mars's history.
This sequence began with an early stage of landscape evolution when the main channel of Coogoon Vallis (Figure 3c) was active (Figure 11e). During this time, sediment transported though Paleo-Coogoon Vallis was deposited in an undetermined location to the west of Kilkhampton crater coinciding with the burial of now exhumed channels ( Figure 6). The rimless craters associated with evidence of lakes (RCL; Figure 9) and the low-relief channels as part of a mature fluvial landscape is comparable to other mid Noachian terrains (e.g.; Irwin et al., 2013).
The implication is that the landscape of the catchment and the formation of Paleo-Coogoon Vallis preserved in the Oxia Basin are of at least Mid Noachian age, contemporaneous with best estimates for the age of clay-bearing terrains  at the ExoMars rover landing site Quantin-Nataf et al., 2021). expect alluvial or riparian facies in the Oxia Basin associated with a regional alluvial landscape.
If the deposition was subaqueous then all three of these depositional environments (lacustrine, deltaic, alluvial) can have occurred through different transgressive or regressive events. Inverted or exhumed channels in the west of the catchment support an alluvial landscape ( Figure 6), but this may only represent what post Noachian denudation (McNeil et al., 2021) has exposed and a grater range of environments contributing to infilling the Oxia Basin are represented in the stratigraphic record. Additionally the groundwater table would have been high and thus actively modifying (secondary) mineralogy.
Following a hiatus (Figure 11d) there was a second phase of fluvial activity and an associated groundwater highstand ( Figure 11c) in which classic "valley network" style channels were incised into the landscape and sediment fans developed at lake high stands. This phase is associated with the peak in valley network formation at the Noachian/Hesperian boundary Irwin et al., 2005) and sediment fan formation around the dichotomy boundary (Di Achille Fawdon et al., 2018) and is also likely to be a time of diagenetic overprinting of the mineralogy by groundwater activity. The modern elevation of Oxia Planum does not match the elevation of other sediment fans, but the presence of tectonic ridges throughout the catchment (e.g., Figures 2, 9, and 10; Molina et al., 2017) suggest substantial changes in elevation since their formation. This observation is also supported by models of the development of Tharsis which suggest topographic changes of up to 1 km for the region (Chan et al., 2018;Citron et al., 2018;Sholes et al., 2021).
The global influence of Tharsis (Citron et al., 2018) causing long wavelength flexure means that the complex of sediment fan remnants west of Kilkhampton Crater may have been deposited at a substantially different elevation than they are at now. The interspersed episodes of tectonic and fluvial/lacustrine activity suggest that this global scale flexure developed alongside evolving climatic conditions and explains mismatches between the model catchment and geomorphic channel network (Figure 2). Since the Hesperian, further tectonic deformation (Figure 11b) and erosion, especially in the west of the catchment, have altered the landscape and has exhumed older fluvial channel deposits from within surrounding deposits, forming the sinuous ridges seen today (Figure 11a).
This sequence shows that an environment supporting fluvial activity occurred several times, but that later phases were insufficiently intense to totally erase evidence of earlier activity. There was progressively less surface water available over time, but during interfluvial periods sufficient water was still stored globally to permit hydrological activity to recommence.

Conclusions
From our observations of ancient fluvial and lacustrine processes within the hydrological catchment for Oxia Planum, the landing location of the ExoMars 2022 rover, we established a sequence of events for hydrological activity in the catchment ( Figure 11) and draw the following conclusions: 1. The Oxia Basin was fed by an extensive fluvial system with a minimum catchment area of ∼2.1 × 10 5 km 2 .
2. Within the model catchment we identified three type of fluvial channel (in addition to the main trunk channel of Coogoon Vallis); "Wide flat floored", "Narrow U-section" and "low-relief", and a network of sinuous ridges. We also identified three classes of possible Paleolake deposit; "large crater lakes", "rimless crater lakes", and "IDD" which are likely to have been concurrent with the fluvial network. 3. We find that there were two major phases of fluvial activity and associated groundwater activity. The first, during the mid Noachian, is associated with the development of the main trunk valley of Coogoon Vallis (Paleo-Coogon Vallis). During this phase the majority of the geomorphic work to form the present day landscape took place and probable geochemical alteration of basement rocks. After a hiatus the second phase of activity reactivated relict channels and incised a system of U-section channels (the key sections of which we term Neo-Coogoon Vallis) into the older valley floors. This is likely to coincide with diagenetic overprinting in the Oxia Basin. 4. We find that large craters with evidence of paleolake deposits are not well integrated into the fluvial system.
These craters show evidence for having been sustained by groundwater, and appear to have had contained lakes with multiple different water levels over time comparable other regional paleolake systems (Michalski et al., 2013;Schmidt et al., 2021). 5. Tectonic activity, probably associated with Tharsis formation, occurred after the formation of Paleo-Coogoon Vallis, and again after the reactivation of the fluvial network and formation of the U-section channels. Consequently the contemporary elevation of pour points out of the catchment do not represent the elevation when channels formed.
These conclusions about the development of the ancient environment in the catchment have the following implications for the formation, transportation, and preservation of biomarkers in the Oxia Basin and for the ExoMars Rosalind Franklin Rover mission: 1. Sedimentary rock samples obtained by the ExoMars rover are likely to record regionally diverse examples of the Noachian environment, as they might have been transported from anywhere in the extensive catchment. 2. Deposits located to the west of Kilkhamption crater, and at the lowest points in the stratigraphy, are likely to be associated with deposition from Paleo-Coogoon Vallis. These could be either deltaic or alluvial sediments, depending on where Paleo-Coogoon Vallis terminated, but a record of this should be preserved within the Oxia Basin and this determination could be made from rover observations. 3. The second phase of activity associated with U-section channel formation was also responsible for the formation of the prominent sediment fan at the eastern margin of the proposed landing site. This phase of fluvial and associated groundwater activity is likely to have imparted a diagenetic overprint on the more ancient clay bearing terrains (Quantin-Nataf et al., 2021). 4. The numerous lakes in the catchment appear to have been well-connected into the regional groundwater system. Given that lakes provide excellent potential habitats, biomarkers from these lacustrine systems could have been transported to and concentrated within the Oxia Basin during the multiple episodes of regional groundwater activity but it is unlikely they where transported though the fluvial system.

Data Availability Statement
The data used in this study is freely available from NASA, (CTX; Malin, 2007;MOLA;Neumann et al., 2003)