The Evolution of Ancient Fluvial Systems in Memnonia Sulci, Mars: Impact Crater Damming, Aggradation, and a Large Water Body on the Dichotomy?

There is conflicting evidence for an ancient ocean which occupied the northern hemispheric basin on Mars. Along different regions of the dichotomy boundary, sediment fans have been interpreted as either forming into a large water body or a series of smaller paleolake basins. Here, we investigate fluvial systems in the Memnonia Sulci region of Mars, set along the dichotomy, which comprise erosional valley networks, paleolake basins, inverted channel systems, and sediment fans. We focus our analysis on the evolution of the upslope catchment and characterizing the ancient environment of a large, downslope basin, bound by the topographic dichotomy and the Medusae Fosse Formation. The catchment fluvial systems comprise highly degraded valley networks and show a complex history of incision and filling, influenced by paleolake basin overflow, impact crater damming, aggradation, and possibly a downstream water body. The morphology of the sediment fans is consistent with either fluvial fans or deltas and they form at discrete elevations, rather than a common elevation plane. Our analysis is consistent with the sediment fans forming into a series of paleolake basins set along the dichotomy, rather than into a large inner sea or ocean‐sized water body. The fluvial systems were likely active between the mid Noachian and early Hesperian periods. Our results demonstrate the complex, multi‐phase evolution of fluvial systems on ancient Mars and highlight the importance of regional and local studies when characterizing ancient regions of the dichotomy.

Here, we investigate a series of fluvial systems along the dichotomy boundary in the Memnonia Sulci region of Mars using high-resolution image and topographic datasets. The fluvial systems in this region have previously been investigated Ori et al., 2000), however, these studies were prior to the widespread coverage of high-resolution datasets, which could give new insight into their evolution. The fluvial systems comprise valley networks, paleolake basins, inverted channel systems, and sediment fans, which debouch into a large basin bound by the Medusae Fosse Formation (MFF) to the north and the dichotomy to the south (Figure 1). We investigate the upslope catchment and the influences on its evolution, and the relationship to regional stratigraphic units and use impact crater size-frequency distributions to constrain the age of the fluvial systems. Finally, we use the sediment fans to characterize the nature of the MFF-dichotomy bound basin and consider the implications for a northern ocean on early Mars.

Regional Geology of Northwest Memnonia Sulci
Our study region is situated in the northwest section of the Memnonia quadrangle ( Figure 1). Here, the highland terrain south of the dichotomy is mid-Noachian in age (3.9-3.8 Ga; Tanaka et al., 2014) and contains numerous, large, degraded impact craters. Erosional valley networks in the region converge on and debouch at the dichotomy (Alemanno et al., 2018;Hynek et al., 2010). Previous authors have identified multiple open paleolake basins situated within the erosional valley networks here Ori et al., 2000).
Northwest Memnonia is adjacent to Lucus Planum, a central exposure of the MFF (Figure 1b). Stratigraphic relationships indicate that the Lucus Planum section of the MFF was likely emplaced and modified from the Hesperian (3.7-3.0 Ga) to Amazonian (3.0-0 Ga) periods (Kerber & Head, 2010;Tanaka et al., 2014). Lucus Planum and the hemispheric dichotomy together form an WSW-ENE orientated basin, which runs for 500 km along its longest axis (Figure 1b). Although much of this basin is topographically enclosed, at its NE margin it intersects the northern lowlands. Our main study area is situated at the SW margin of this basin.

Datasets
We investigate the fluvial systems in the Memnonia Sulci region primarily using Context Camera (CTX; 5-6 m/pixel; Malin et al., 2007) and High Resolution Imaging Science Experiment (HiRISE; 0.25 m/pixel; McEwen et al., 2007) datasets. We combined eight CTX images to produce a mosaic basemap for the main 10.1029/2021JE007021 3 of 20 study area. We produced HiRISE and CTX digital elevation models (DEMs; Table S1 in the Supporting Information S1) of the fluvial catchment and sediment fans using the USGS Integrated Software for Imagers and Spectrometers (ISIS) software and the BAE photogrammetric package SOCET SET (Kirk et al., 2008). The resultant DEMs were tied to Mars Orbital Laser Altimeter (Zuber et al., 1992) topography and exported with a post spacing of 1 and 20 m/pixel for HiRISE and CTX, respectively. These DEMs were supplemented by MOLA topography (450 m/pixel). We used CTX and MOLA topography to measure the along and cross-section profiles of the catchment fluvial systems. We also used available Color and Stereo Science Imaging System (CaSSIS; 4.5 m/pixel; Thomas et al., 2017) images to investigate the sediment fans. We used images with the combined NIR-BLU-PAN filters (NBP) and synthetic RGB images created from the BLU and PAN filters. Our investigation

Measurements of the Fluvial Catchment
Within our main study area (Figure 1), we used the valley networks mapped by Alemanno et al. (2018) to identify the largest and most continuous of the catchment fluvial systems, which we refer to as Unnamed Valley 1 (UV1). We used the standard ArcGIS hydrology toolkit to measure the possible drainage catchment of UV1 using the MOLA DEM clipped to the western Memnonia region. We filled the DEM to a depth of 750 m to account for local hydrological sinks (value determined based on deepest sink; B1) and to ensure continuous flow. Our fill value was defined based on the maximum depth of a 50 km diameter open basin paleolake, midway along the course of UV1. To define a watershed, we placed a pour point near the outlet of UV1. Finally, we compared our watershed to the morphological expression of valley networks. As the ArcGIS hydrology toolkit is designed for active river systems on Earth and not ancient topography on Mars, we consider our hydrology measurements as a rough guide to the potential size of the catchment.

Age Estimates From Geologic Maps and Impact Crater Size-Frequency Distributions
Estimating when small or linear features such as valley networks or sediment fans formed using impact crater size-frequency distributions ("crater counting") is challenging and can produce misleading results (e.g., Warner et al., 2015). Instead, we mostly rely on the stratigraphy of published geological maps (Tanaka et al., 2014) to estimate when fluvial processes were occurring. We do perform one crater count of an 18 km diameter impact crater (Crater A) which overlies the fluvial catchment and has well-defined impact ejecta. We mapped out the impact ejecta, before counting all overlying craters visible in our CTX mosaic using the ArcMap extension CraterTools (Kneissl et al., 2011) which were greater than 50 m in diameter. We then used CraterStats2 (Michael et al., 2012) software to produce a model formation age for Crater A, using the production function of Ivanov (2001) and the chronology function of Hartmann and Neukum (2001).

Observations and Results
To constrain the ancient environment in this region of the dichotomy, we investigate the fluvial catchment (4.1), its minimum age (4.2), and the sediment fans fed by the catchment (4.3).

Catchment of the Fluvial Systems
In this section, we describe the catchment fluvial systems within the main study area, which fed the downslope sediment fans and the MFF-dichotomy bound basin. Erosional valley networks in this region converge on the dichotomy (Allemanno et al., 2018;Hynek et al., 2010, Figure 1). The longest of the systems within our main study area, Unnamed Valley 1 (UV1), was fed by a watershed potentially 58,000 km 2 in area and extended 450 km south of the dichotomy (Figures 2 and 3). In the lower reaches of UV1, our reconstructed catchment area mostly coincides with the morphological expression of valley networks (Figure 2). To contrast, valley networks are largely absent from the upper reaches of the reconstructed catchment, although there are some examples of degraded basins with outlet valleys (Figure 2). Approximately midway along the course of UV1 is a degraded, flat-floored 50 km impact crater, with multiple sets of terraced structures around its interior rim ( Figure 2). This crater is breached by UV1 at both its northern and southern margins and has previously been interpreted as an open paleolake basin Ori et al., 2000). We refer to this open paleolake basin as Basin 1 (B1). The morphology of UV1 is consistent with an erosional valley network formed by fluvial erosion (e.g., Allemanno et al., 2018;Hynek et al., 2010).
We focus our observations on the 190 km long section of UV1 north of B1, which continues toward the dichotomy ( Figure 3). The walls of UV1 here are poorly defined but appear controlled by ancient crater rims and ejecta blankets ( Figure 2). Paired erosional terrace structures which run parallel to the valley walls are sometimes present in some of this northern section (Figures 3b and S1 in the Supporting Information S1). The valley floor of UV1 appears infilled (Figure 3), although in two locations, this valley infill is incised. The first of these incised sections begins at the paleolake outlet (I1) and continues for approximately 60 km (Figure 3b). The incision depth of I1 shallows as it increases with distance from B1 (from ∼300 m depth at B1 to 0; Figure 3). I1 eventually intersects an 18 km diameter impact crater (Crater A) which superposes both UV1 and I1 (Figure 3). The ejecta from Crater A appears mostly unmodified and does not appear to have been subsequently incised. The second incised section (I2) of UV1 begins at a knickpoint approximately 70 km south of the valley mouth (north of Crater A), where there is a significant increase in gradient ( Figure 2b). I2 initially deepens to a maximum depth of 600 m before shallowing (Figure 3a). At its northern end, I2 is superposed by ejecta from a 12 km diameter impact crater (Crater B; Figure 3a). However, the ejecta which superposes I2 is not continuous and appears to have been incised since its formation (Figure 4a). A 5 km long ridge is situated within I2, which also superposes the Crater B ejecta (Figure 4a). The ridge is 50-150 m wide and some discontinuous ridge segments are observed within I2 further upslope. Little to no structure is exposed in the top surface or margins of the ridges. As the ridges are situated within I2, we interpret them as inverted channel deposits (deposits of indurated fluvial . Crater A (red outline) superposes UV1 and has not been incised. Crater B (blue lines) also superposes UV1 but has subsequently been incised. Cross-sectional profiles show sections where the fill of UV1 has been incised (I1 and I2). Profiles were extracted from two CTX DEMs. Yellow circles show sediment fan locations. (b) CTX mosaic of I1 which begins at the outlet for the terraced paleolake (B 1), shallowing in the downslope direction (Profiles I1A-I1A′ and I1B-I1B′). The original walls are highly degraded, and terrace structures are visible. (c) CTX image showing the southern section of I2, which begins at a knickpoint north of Crater A. I2 initially deepens (Profiles I2A-I2A′ and I2B-I2B′) before shallowing. sediment exhumed by differential erosion) and their morphology is consistent with other examples of Mars (e.g., Burr et al., 2010;Davis et al., 2016;Hayden et al., 2019;Williams et al., 2009).
North of Crater B, I2 debouches onto a plain bound by the MFF and the dichotomy scarp. South of Crater B is an additional valley network, which we refer to as Unnamed Valley 2 (UV2). UV2 is approximately 80 km in length and runs approximately parallel to UV1 (Figures 2 and 3a). Like UV1, UV2 has also been heavily modified since its formation and attempts to reconstruct its watershed were unsuccessful. Near its uppermost reach, parts of UV2 have been completely buried by Crater A (Figure 2). As UV2 approaches Crater B, its abruptly changes direction to the SW, where it debouches into a 13 km diameter basin, Basin 2 (B2; Figure 4b). B2 appears to be a highly degraded impact structure. At the eastern margins of B2 are branching and re-joining ridge segments, approximately 10 km long, situated within the fill of UV2 ( Figure 4c). As with the previous example in UV1, we interpret these ridges as inverted channel deposits. At the SW edge of B2, there is a 5 km long, poorly exposed, fan-like form, comprised of channel and lobate structures ( Figure 4c).

Minimum Age of the Fluvial Catchment
In this section, we use crater counting to constrain the minimum age of the catchment. Crater A superposes and infills UV1 and UV2, and its impact ejecta does not appear incised. Crater A and its ejecta cover an area of 1.79 × 10 3 km 2 . Figure 5 shows the count area and the binned cumulative crater frequency histograms for Crater A. In total, we counted 1,755 craters (D ≥ 50 m) which were overlying Crater A. Of these, the largest craters (n = 23) fall along the 3.5 ± 0.1 Ga isochron, defined by Hartmann and Neukum (2001) and using the chronology of Michael (2013).

Description and Interpretation of the Sediment Fans
In this section, we describe and interpret the major sediment fans found along the dichotomy in this region. Figure 6a shows the locations of fans in the main study area.

Fan 1
Observations: Fan 1 begins 10 km north of UV1 outlet ( Figure 6a). Fan 1 comprises a series of linear to quasi-sinuous ridges, orientated in the downslope direction ( Figure 6b). The ridges are approximately 50-200 m wide, ∼5 m high, and form 2-5 km long continuous segments. The ridges progressively bifurcate in the downslope direction ( Figure 6c) and multiple ridge systems superpose one another (Figure 6b). Approximately a third of the way downslope, Fan 1 divides into two main branches, Branch A and B (Figure 6b). Branch A appears stratigraphically higher than Branch B. The distal section of Branch B is ∼20 m lower in elevation than Branch A. There are also examples of ridges amalgamated together laterally (Figure 6c). At the downslope termination of Branch A, some ridges bifurcate into narrower, digitate ridges ( Figure 6d). In the distal region of Branch B, the  Hartmann and Neukum (2001) and using the chronology of Michael (2013). 23 of the largest craters show a best fit to the 3.5 ± 0.1 Ga isochron. ridges show a higher degree of sinuosity (Figure 6b). Adjacent and parallel to some of the ridges are corridors of darker material (Figures 6c and 6d). The ridges are set within low relief plains and appear to being exhumed from a mid-toned material, with a distinct mottled and pitted surface texture (Figure 6b). At its longest extent, Fan 1 extends 16 km from its southern to northern most point, over an average slope of 0.3° (1:190).
Interpretation: Due to their sinuous and branching nature and their proximity to the upslope fluvial systems, we interpret the ridges as inverted channel systems (e.g., Burr et al., 2010;Davis et al., 2016;Williams et al., 2009). We further interpret the ridges as the exhumed remnants of a distributary channel and sediment fan system. The occurrence of Fan 1, 10 km north of the UV1 outlet, set within low relief plains, could be consistent with either an aggrading fluvial fan (e.g., Ventra & Clarke, 2018) or a delta . In this work, we consider a fluvial fan to comprise multiple, extensive, distributary fluvial systems (i.e., migrating and aggrading channel-belts) with low gradient surfaces and often fed by an extensive catchment (e.g., Ventra & Clarke, 2018). A delta is broadly similar to a fluvial fan, with the addition of forming into a standing body of water, resulting in some sediment stored below base level (e.g., Ganti et al., 2014;Ventra & Clarke, 2018). To contrast, alluvial fans often form adjacent to short, steep catchments and can have high, proximal surface gradients (Blair & McPherson, 1994).

Fan 2
Observations: Fan 2 begins at the end of a 25 km long NW trending channel, which originates from the outlet of UV1 (Figure 7a). The feeder channel appears to have been deflected around Crater C (Figure 7a), which likely pre-dates Fan 2. The main body of Fan 2 comprises a plethora of curvilinear, discontinuous ridges which significantly vary in orientation, although commonly WNW-ESW or NNW-SSE (Figures 7b and 7c). Ridges are up to 5 km long, ∼50-300 m wide, and 5-10 m high (Figures 7b and 7c). The ridges occur on an expansive, planar surface, from which some ridges are being exhumed ( Figure 7a). As in Fan 1, ridge systems divide in the downslope direction and commonly superpose one another (Figures 7c and 7d). In some examples, distinct caprock is visible on the ridges (Figure 7c). Multiple ridge systems appear to have amalgamated together laterally (Figures 7c and 7d). In total, the ridges cover an area of 18 km by 12 km, over an average slope of 0.005° (1:1.1 × 10 4 ).
Interpretation: Of all the sediment fans in the study, we are least confident in our interpretation of Fan 2. The gradational boundary between the impact ejecta from Crater C and the ridges suggests that the ridges maybe eroded, distal ejecta deposits (e.g., Barlow et al., 2000). However, the orientation and planimetric pattern of some of the ridges are inconsistent with impact ejecta originating from Crater C. The curvilinear and bifurcating nature of some ridges, as well as the examples of multiple ridges amalgamated together are instead more consistent with the exhumed remnants of a distributary channel and sediment fan system (i.e., a fluvial fan). It is possible that Fan 2 may represent a sediment fan overlying ejecta deposits in places. Unlike Fan 1, Fan 2 forms northwest of UV1, away from the main MFF-dichotomy bound basin to the northeast (Figure 6a).

Fan 3
Observations: Fan 3 is found at the southwest interior margin of a highly degraded, 10 km impact crater approximately overlying the dichotomy, where an erosional valley network debouches (Figure 7e). The catchment for Fan 3 extends at least 20 km southwest. Fan 3 forms an arcuate shaped deposit, 3 km long by 6 km, filling the width of the crater (Figure 7e). Distal erosional remnants suggest that Fan 3 was previously more extensive and may have filled the diameter of the crater entirely. Erosion has exposed 50-100 m wide, curvilinear ridges which are generally orientated NNE (Figure 7f). At the margins of these ridges, sub-horizontal layering is exposed (Figure 7f). The northern wall of the crater has been eroded away and some of the ridges extended beyond the proximal fan deposits and outside the confines of the crater (Figure 7e).
Interpretation: The curvilinear ridges exposed in Fan 3 are consistent with inverted channel deposits. The morphology of Fan 3 is consistent with a fluvial fan or a delta body. As the distal deposits of Fan 3 steps out of the hosting crater, if it formed into a water body, such a body may not have been confined to the crater.

Fan 4
Observations: Fan 4 is found 70 km ENE of Fan 1 (Figure 6). The apex of Fan 4 is found ∼100 m lower in elevation than the most distal sections of Fan 1 and at the end of a 35 km long feeder channel, the pathway of which suggests it has been diverted around at least some of the MFF (Figure 8a). The origin of the feeder channel is unclear as it appears buried at its upslope margin. Fan 4 comprises multiple, discrete, flat-topped sediment bodies, typically ∼100-1,500 m wide and ∼1-5 km long, orientated in the downslope direction (Figure 8b). Sub-horizontal layering is exposed in the margins of the sediment bodies, which is continuous for hundreds of meters ( Figure 8b). The sediment bodies appear to be grouped into three branches (C-E), with Branch E being the stratigraphically youngest. On the top surfaces of the sediment bodies, erosion has exposed low relief, linear ridges, typically 50 m wide and <1 km long, which are orientated parallel to the long axis of the sediment bodies (Figure 8c). In some examples, the sediment bodies superpose one another. The sediment bodies broaden in the downslope direction (Figure 8d). These sediment bodies are being exhumed from within the surrounding terrain. Dividing the sediment bodies are negative relief channels, which are of generally similar dimensions. In total, from its apex to its most distal section, Fan 4 is ∼20 km long and up to ∼15 km wide. Overlying Fan 4 are erosional outliers of the MFF (Figure 8b). Interpretation: Fan 4 is morphologically distinct from Fans 1-3. We interpret the low relief ridges on the sediment bodies as channel deposits which have amalgamated together. Although the sediment bodies broaden downslope, a clear channel-lobe transition is not observed, as seen in other martian sediment fans, interpreted as deltaic (e.g., Fawdon et al., 2018). Like previous examples, Fan 4 could plausibly be a fluvial or deltaic fan.

Fan 5
Observations: Fan 5 is outside the main study area but is relevant to understanding the evolution of MFF-dichotomy bound basin, as it occurs near its northeast margin (Figure 1b). Fan 5 begins at the dichotomy boundary at the end of a catchment at least 300 km long (Figure 1b). Fan 5 strongly differs from Fans 1-4 in its morphology: it comprises a series (at least three) of arcuate to lobate-shaped sedimentary deposits, which occur at discrete elevations and are interconnected by channels (Figure 9a). We term the three recognizable lobes, Lobes A-C. All three lobes have branching, distributary channels visible on their top surfaces. Lobe A, the most proximal lobe, is ∼5 km long and forms at the dichotomy scarp, where the source bedrock valley network cross-cuts a wrinkle ridge and debouches into a shallow, degraded 4 km impact crater (Figure 9b). Lobe A is incised by a NNW orientated channel, which connects to Lobe B. Lobe B begins 10 km from the valley outlet and is 10 km long (Figure 9c). Like Lobe A, Lobe B is set within a local topographic low, possibly a degraded impact crater (Figure 9d). At its northern margin, Lobe B has left the confines of this crater. Lobe B has subsequently been incised through by a channel, which has exposed sub-horizontal layering in its margins. The channel incising Lobe B connects to the most distal lobe, Lobe C. Lobe C (Figure 9e), begins 40 km NNW of the valley outlet and almost 200 m lower in elevation (Figure 9a). Lobe C forms into a topographically closed region of the MFF-dichotomy bound basin (rather than a local topographic low). Lobe C is approximately 12 km long and is not incised by subsequent channels. Interpretation: Fan 5 occurs over a much wider area than Fans 1-4, with distinct lobes separated by 200 m in elevation. The successive channels incising and connecting Lobes A-C suggest that each lobe is progressively younger (e.g., the channel cutting through and postdating Lobe A can be followed downslope to source Lobe B). The basinward-step of Fan 5 is consistent with both fluvial fans and delta bodies (e.g., Blum & Törnqvist, 2000).

Physiographic Setting of Downslope, MFF-Dichotomy Bound Basin
In this section, we describe the physiographic setting of the downslope, MFF-dichotomy bound basin and its relationship to the sediment fans ( Figure 10). Above the −2,000 m contour, the basin is open to the northern lowlands, and encompasses the outlet of UV1. Fan 3 forms near this elevation ( Figure 10). However, both Fans 1 and 2 form into a topographically closed basin defined by the −2,100 m contour, which we term Basin 3 ( Figure 10). Similarly, Fan 4 forms into a topographically closed region of the main MFF-dichotomy bound basin, defined by the −2,300 m contour ( Figure 10). To the east, Lobe A of Fan 5 forms close to the −2,400 m contour, which at this elevation is open to the lowlands (Figure 10). To contrast, Lobe C of Fan 5 forms into a topographically closed region defined by the −2,700 m contour (Figure 10).

Incision, Filling, and Damming of the Catchment
In this section, we discuss the evolution of the catchment fluvial systems (UV1 and UV2). The presence of an inlet valley, an outlet, and terraces structures in Basin 1 supports an interpretation as an open paleolake basin (Figure 2; Fassett & Head, 2008;Goudge et al., 2012Goudge et al., , 2016Irwin et al., 2005;Ori et al., 2000). Additional open basin paleolakes may have existed further upslope, although these are poorly preserved (Figure 2). The evolution of the northern section of UV1 appears complex, with multiple stages of bedrock incision and sediment aggradation within the valley, in agreement with the multi-stage evolution interpreted by Irwin et al. (2005). Firstly, fluvial downcutting into bedrock must have occurred to form the initial valley. We estimate from CTX topography that incision was to a depth of up to 600-800 m, although the original valley walls are so poorly preserved that this is difficult to determine. Bedrock strath terraces along UV1 and in Basin 1 indicates that multiple periods of downcutting may have occurred (Figure 3). The course of UV1 appears to have been controlled by the topography generated by impact craters and their ejecta along the valley margins (Figure 3a). UV1 is not graded to an equilibrium profile (Figure 2b), indicating that it was an immature fluvial system, like many other examples across Mars (e.g., Fassett & Head, 2008;Goudge & Fassett, 2018;Hynek et al., 2010).
The valley floor of UV1 has been filled by sediment to an unknown depth (we estimate at least several hundred meters; Figure 3). The fill may be due to impact cratering, input from the MFF, or channel-fill associated with the fluvial system itself. Regardless of its origin, incision into the fill in two distinct regions (I1 and I2) indicates that the fill was later modified by subsequent fluvial processes (Figures 3b and 3c). I1 (Figure 3b) begins at the outlet of Basin 1, so the incision may have been caused by an overflow flooding event from the lake which occupied Basin 1 (Goudge & Fassett, 2018). We note that overflow flooding cannot be responsible for the downcutting which originally formed UV1 as the original valley walls exceed the maximum elevation of the walls of B1. I1 shallows in the downslope direction (from 300 m to 0) and it eventually becomes buried by Crater A.
The reason for the formation of I2 (Figure 3c), at a prominent knickpoint north of Crater A, is unclear. Knickpoints may be caused by lithologic variation in the underlying bedrock of the incised valley, when a material that is resistant to erosion is encountered (e.g., Gilbert, 1896). As the resistant layer is eroded, the knickpoint migrates upstream. Knickpoint formation may also have been initiated by base level fall within a downstream body of water, concentrating incision in the downslope region of UV1 (e.g., Goudge & Fassett, 2018;Howard, 1994). Near the dichotomy boundary, I2 has been buried by the ejecta from Crater B (Figure 4a). However, the ejecta from Crater B has subsequently been incised, indicating that the formation of Crater B was syn-fluvial, temporarily damming UV1, or that Crater B formed in between periods of fluvial activity. In either scenario, UV1/I2 later incised through the ejecta from Crater B. South of Crater B, the inferred flow directions of UV2 indicate that it may have originally been a tributary of UV1. Following the formation of Crater B, its ejecta formed a dam and UV2 was likely diverted SW into Basin 2 (Figure 4b), where water may have pooled, forming a paleolake.
The inverted channel deposits in the downslope regions of UV1 and UV2 indicates the presence of channel-fill and net deposition within the valleys (Figures 4a and 4c). Inverted channel deposits bound within the downslope region of erosional valleys are observed elsewhere on Mars (Dickson et al., 2021), notably Arabia Terra, where they are often found in proximity to paleolake basins (Davis et al., 2016. In filled valley sequences on Earth, these processes can be driven by changes in base level, or water or sediment supply (e.g., Blum & Aslan, 2006;Singh et al., 2017). The proximity to the downslope MFF-dichotomy bound basin (if it contained water) opens the possibility that these regions of UV1 and UV2 were subject to the effects of water level fluctuations (Blum et al., 2013). Stratigraphic relationships indicate that the formation of these inverted channel deposits likely postdates the formation of Crater B. Thus, this association indicates that the formation of Crater B may be responsible for the aggradation within both UV1 and UV2, potentially by supplying ample unconsolidated sediment to the system or causing a local rise in base level. The formation of Crater A, which superposes both UV1 and UV2, and whose ejecta does not appear incised, seemingly occurred after the cessation of fluvial activity within the catchment.

Formation of the Sediment Fans
In this section, we interpret sediment fans and consider their formation environment. Many fans appear dominated by ridges (Figures 6 and 7) or sediment bodies which erode into ridges (Figure 8), which we have interpreted as inverted channel systems. The stacked nature of the ridges (Figures 6, 7c, and 7d) and the amalgamation of multiple ridges (Figures 6, 7d, and 8) supports an interpretation as exhumed channel-belt deposits, rather than the inversion of single-fill channels (DiBiase et al., 2013;Hayden et al., 2019;Martinez et al., 2010). Ridges within Fans 1, 2 and 4 occur in stratigraphically and topographically separated branches (Figures 6-8), suggesting that these fans built up multiple, laterally stacked sedimentary deposits, formed by the avulsion and aggradation of river channels (e.g., Balme et al., 2020;Davis et al., 2019). Within the stratigraphically distinct branches of Fan 1, bifurcation nodes are exposed at different elevations (Figure 6), suggesting that Fan 1 may have grown out basinward over time, possibly in response to changing downslope water levels (e.g., Fawdon et al., 2018).
To contrast, Fans 3 (Figures 7e and 7f) and 5 are defined by lobate-shaped deposits (Figure 9). Although Fan 3 forms near an elevation contour open to the lowlands (Figure 10), the majority of Fan 3 is confined to its hosting impact crater (Figures 7e and 7f), suggesting it formed into a lake confined to the crater (e.g., Fassett & Head, 2008), rather than a more expansive basin. The distal inverted channel deposits on Fan 3 and the degraded state of the northern crater margin, point to deposition eventually leaving the confines of the crater.
Similarly, although Lobes A and B of Fan 5 occurs near an elevation open to the northern lowlands (Figure 10), we note these deposits form into local topographic lows (e.g., craters; Figure 9) not apparent in MOLA topography, suggesting that sediment deposition infilled local accommodation (i.e., impact craters), rather than a wider basin. These observations are generally more consistent with the behavior of a fluvial fan than a delta. The 40 km progression in downslope lobe formation indicates a downstream migration in sediment deposition over time, which was concentrated in local topographic lows. Lobes A and B are successively incised by channels, with the new lobe deposition occurring downslope of the incised channel, possibly driven by reduction in sediment supply or lowering of base level (e.g., Blum & Törnqvist, 2000). The terminal lobe, Lobe C, forms into a wider basin (Figure 10), and it is conceivable that a larger body of water may have pooled here (although not filling the main MFF-dichotomy bound basin). In summary, all five fans are generally consistent with distributary channel and sediment fan systems, and there is tentative evidence at least some of them may have formed in water bodies.

Chronology of the Fluvial Systems
In this section, we consider when the fluvial systems were active to place them into broader context. UV1, UV2, and other nearby valley networks incise into mid Noachian age terrain (∼3.9-3.8 Ga; Tanaka et al., 2014), which places an upper bracket on their formation. We note that the original valley walls are highly degraded (Figures 2  and 3), so it is possible that the valley fill, inverted channel systems and sediment fans are significantly younger. Fans 1, 2, 4 and 5 all form onto terrain mapped as early Hesperian (∼3.7-3.5 Ga; Tanaka et al., 2014) in age. The age of the adjacent MFF is difficult to determine as its extreme friability limits conventional crater counting techniques. Kerber and Head (2010) analyzed the stratigraphic relationship of the MFF to lava flows and impact craters to determine its emplacement age. Kerber and Head (2010) suggest that in this region, the MFF (Lucus Planum) was emplaced during the Hesperian. The feeder channel for Fan 4 is clearly diverted around, and infilled by, the MFF, and Fan 4 itself is overlain by erosional outliers of the MFF (Figure 8). These observations suggest that the MFF was at least partially emplaced before the formation of Fan 4 and that its formation continued afterward. Significant volumes of sediment may have been supplied to the fluvial systems if they were coeval with the emplacement of the MFF (i.e., volcaniclastic).
The feeder channel for Fan 4 may have been sourced from overspill from Basin 3 (which hosted Fans 1 and 2; Figure 10), although we cannot exclude the possibility that these systems were active at different times. As Crater A superposes UV1 and it is not incised, its formation at 3.5 Ga ( Figure 5) likely represents a lower age bracket for the cessation of fluvial processes. Fans 1, 2 and 3 were likely fed by UV1 and its associated valley networks, also putting a bracket on their activity. It is possible that some fluvial systems were locally reactivated by late-stage flows, as seen elsewhere on Mars (e.g., Bouley et al., 2009). In summary, we conclude that fluvial processes were likely active between the mid-Noachian and early Hesperian (3.9-3.5 Ga), although most landforms described here probably represent the younger part of this age bracket.

Nature and Evolution of the Downslope Basin
In this section, we consider the environment of the downslope basin (Figure 1), bound by the MFF and the dichotomy scarp, in particular relation to the existence of an ancient ocean. We consider five possible scenarios for the evolution of this basin: 1. The fluvial systems predate the MFF and thus the basin did not exist at the time they were forming 2. The fluvial systems formed into an entirely dry, sub-aerial basin 3. The fluvial systems formed into a large, sub-aqueous basin, which was disconnected or 4. Connected to a water body in the northern lowlands 5. The MFF-dichotomy basin was mostly dry, but the fluvial systems formed in localized paleolake basins The stratigraphic relationship of the MFF to Fan 4 ( Figure 8) rules out scenario 1. There is tentative evidence that at least some of the fans may have formed into water bodies, making scenario 2 less likely, although we cannot rule it out entirely. Some of the fans do form near the dichotomy equipotential level defined by Di Achille and : ∼−2,500 m. However, we note that the fans all form at distinct elevations (separated by up to 700 m), either into closed sub-basins within the main MFF-dichotomy bound basin (Figure 10), or into local topographic lows (i.e., impact craters). Unlike other sediment fans identified along the dichotomy (Di Achille , the fans in this region do not form an equipotential surface. It therefore seems unlikely that the fluvial systems, as they are currently expressed, formed into a large, sub-aqueous basin, that was either disconnected or connected to the northern lowlands (scenarios 3 and 4).
Instead, the morphology and physiography of the fluvial systems suggests that at least some of the sediment fans formed into local paleolake basins, set within the wider MFF-dichotomy bound basin (scenario 5). It is conceivable that the lake basin hosting Fans 1 and 2 (Basin 3) was connected to the basin hosting Fan 4 ( Figure 10). The paleolake basins which the sediment fans form into could represent the last stages of a large, drying body of water. The basinward growth of Fan 1 ( Figure 6) may record this drying, however, the changes to Fan 1 are much smaller and more localized than those observed elsewhere along the dichotomy, such as Aeolis Dorsa (e.g., Cardenas et al., 2017;Hughes et al., 2019) or Hypanis Valles (e.g., Fawdon et al., 2018). Fan 5 also shows evidence for basinward growth, however, Lobes A and B form into successive local topographic lows, suggesting that deposition shifted basinwards once the local accommodation was filled, rather than being driven by a larger water body that receded.
Our results agree with previous studies of the dichotomy boundary in the Gale crater region (García-Arnay & Gutiérrez, 2020; Rivera-Hernández & Palucis, 2019), in which sediment fans were characterized as forming into a series of paleolake basins, rather than a hemisphere-spanning ocean. We note that the Memnonia Sulci systems were likely active at a similar time to the Hypanis Valles and Gale crater region systems (Noachian to early Hesperian; García-Arnay & Gutiérrez, 2020; Fawdon et al., 2018). Additionally, our results generally agree with a recent global-scale investigation which demonstrated that the sediment fans along the dichotomy mostly do not have the spatial-temporal relationships to indicate they formed into a hemisphere-spanning ocean where the water levels were changing (De Toffoli et al., 2021).

Conclusions
We used high-resolution image and topographic datasets to investigate valley networks, paleolake basins, inverted channel systems, and sediment fans in the NW Memnonia Sulci region of Mars. The extensive upslope catchment comprises highly degraded valley networks, which reveal a complex evolutionary history including multiple periods of incision and filling. The fluvial catchment has been influenced by overflow floods from upslope basins and impact crater damming, leading to the diversion of flows and aggradation within the catchment. The catchment also shows some evidence for the influence of a downslope body of water. Downslope, a series of sediment form into a large basin, bound by the topographic dichotomy and Medusae Fosse Formation. The morphology of the sediment fans is generally consistent with a formation as fluvial fans or deltas. The physiographic setting of the fans suggests they formed into a series of paleolake basins, situated on the margins of the dichotomy, rather than a large regional water body or ocean, although it is unclear if a more extensive water body may have existed earlier. Stratigraphic relationships and crater counts suggest that fluvial activity likely occurred between the mid Noachian and early Hesperian. Our results (a) demonstrate the complex, multi-phase evolution of fluvial systems on ancient Mars (e.g., Edgett et al., 2020) and (b) highlight the importance of regional studies when assessing the nature of the northern lowlands basin on early Mars, which can differ from the global picture.