Satellite data for the field site in the Californian Mojave Desert was examined on Google Earth (Google Earth, 2024). In addition, we used aerial high-resolution orthoimagery from the USGS EarthExplorer including the Digitial Orthophoto Quadrangles (DOQs) originally collected on 01 October 1995: Harvard Hill (O3411603.NWS.839561) and Alvord Mountain West (O3511659.SWS.839536). Regional meter-scale topographic data are from airborne LiDAR (Light Detection And Ranging) acquired in August 2017 (accessible at USGS Lidar Explorer https://www.usgs.gov/tools/lidarexplorer).

Field studies, measurements, and observations at the paleochannel ridges were conducted between the years 2014 and 2016. Photo documentation of paleochannel ridges included landscape views and nadir surface images for grain size statistics following methods of Bunte and Abt (2001). Topographic surveys of paleochannel ridges used the Trimble Pro XRS Differential Global Positioning System (DGPS), with 2–4 cm horizontal and vertical precision in the acquired elevation data. Elevation measurements were used to characterize ridge shape with spot points acquired along longitudinal and cross-sectional ridge transects.

Images used in this study are from two cameras on board the Mars Reconnaissance Orbiter: the High Resolution Imaging Science Experiment (HiRISE; 0.25 m/pix; McEwen et al., 2007) and the ConTeXt camera (CTX, ∼6 m/pix; Malin et al., 2007) rendered into a global mosaic (Dickson et al., 2024; available online through the Caltech Murray Lab; https://murray-lab.caltech.edu/CTX).

We also examined multipsectral Thermal Emission Imaging System (THEMIS) infrared images (Christensen et al., 2001). Decorrelation stretched (DCS) images are produced from THEMIS band 9 (center wavelength 12.57 μm), band 6 (10.21 μm) and band 3 (7.93 μm) mapped to red, green and blue, respectively. These data are advantageous because they illustrate the maximum spectral variability within THEMIS data in a 3-band set. THEMIS DCS-963 quadrangles are generated according to the procedures described in Hill (2022) to have globally consistent colors. These data are currently available to the THEMIS science team for validation prior to archival for the planetary science community.

Topographic data are from the Mars Global Surveyor Mars Orbiter Laser Altimeter (MOLA, Zuber et al., 1992) blended with High-Resolution Stereo Camera (HRSC; Neukum et al., 2004) digital elevation models (DEM, 200 m/pix resolution, ∼10 m vertical precision; Fergason et al., 2018; Gwinner, 2010; Kirk et al., 2021). Where available, we generated high-resolution CTX-DEMs (24 m/pix resolution, ∼5 m vertical precision Kirk et al., 2021) using the Ames Stereo Pipeline (Beyer et al., 2018; Mayer and Kite, 2016; Mayer, 2018). Data products produced during this study are archived with the USGS Science Base (Williams et al., 2024).

Geomorphic features mapped in previous studies were also incorporated into the JMARS (Christensen et al., 2009) GIS database. We used the global map of valley networks (Hynek, 2016). Davis et al. (2019) mapped sinuous ridges, interpreted as inverted channels, in Arabia Terra. In a regional study of the northern Hellas basin, Gullikson et al. (2023) mapped fan deposits and catchments, valley networks, ridges, fan deposits and catchments.

We selected two case study locations on Mars with paleochannel ridges within closed topographic basins: northern Arabia Terra and Harris crater. These sites were selected because they have not previously received detailed consideration as lacustrine sites, and they have mapped ridges interpreted as raised channels formed through topographic inversion. Our approach is described below and the workflow is illustrated in Figure 1.

The former course of rivers is preserved in a number of ways. Paleochannels are the relict fluvial path identified by sedimentary deposit and/or landform attributes (e.g., planimetric network pattern; Bridge, 1984; 2003; Miall, 1985; Church, 2006; Fielding, 2007; Hooke and Yorke, 2011; Ashmore, 2013). These deposits can be single channel-fill, channel belt or alluvial-valley-fill deposits (Hayden and Lamb, 2020). Paleochannel ridges are positive-relief features whose origin may be primary (constructional deposit) or secondary (inverted-relief landform) fluvial deposits (Figure 2).

Topographic inversion in fluvial systems occurs where differential erosion removes the intervening erodible material and the relic fluvial channel fills, channel belts or valley fill material are preferentially preserved in an elevated state relative to the surrounding terrain (Maizels, 1990; Pain and Ollier, 1995; Pain et al., 2007; Hayden et al., 2019; Zaki et al., 2020; Zhao et al., 2021). Such landforms are termed raised channels, inverted valleys, the colloquial ‘inverted channels’ (a term that is pervasive in the martian literature), or exhumed channels when formerly buried prior to topographic inversion (Williams et al., 2007; Cuevas Martínez et al., 2010; Foix et al., 2012; Cardenas et al., 2020; Korus and Joeckel, 2023; Clarke et al., 2020; Williams et al., 2009; Williams et al., 2011). Various processes can cause fluvial deposits to become indurated, whether from cementation, armoring by coarse sediment or infilling by lava flows (Pain and Ollier, 1995; Pain et al., 2007). Subsequent erosion of the channel banks and/or floodplain results in topographic inversion, transforming the original low-lying channel deposits into an elevated ridge. A recent global inventory of over 100 raised channels on Earth documents a range of erosion-resistant capping materials (Zaki et al., 2021), but all of these landforms have a destructive origin.

Primary paleochannel ridges are well documented in the literature. Alluvial ridges are near-channel positive-relief regions associated with changing river courses due to avulsion (Fisk, 1944; Allen, 1965). As a river channel migrates over time, channel belt dynamics generate alluvial ridges through overbank deposition (Allen, 1965; Bridge and Leeder, 1979; Bridge, 1984; Nicholas et al., 2018; Slingerland and Smith, 1998). This sediment deposition is the river’s store of potential energy (Gearon and Edmonds, 2025). Aggradation of the channel beds and banks of rivers form alluvial ridges. The most detailed study of alluvial ridges is from the modern Mississippi River and more broadly along the coastal plain of the northern Gulf of Mexico (Fisk, 1944; 1952; Aslan and Blum, 1999; Swartz et al., 2022), and thus the term is generally associated with fine-grained sedimentary systems. Alluvial ridges are discontinuous landforms that record abandoned river channels generated over timescales of centuries to millennia (Fisk, 1944; Blum and Price, 1988; Kiss et al., 2022).

In contrast, and of relevance to this work, a newly recognized mechanism of primary paleochannel ridge formation is subaqueous generation at the interface between a river and a rising lake (Miller et al., 2018). This is a constructional scenario, wherein the fluvial deposits are emplaced in a ridge-form at the time of deposition. Paleochannel ridge relief may be enhanced due to secondary erosional processes as the lake level falls. In the next section, we describe field observations from such a location as an illustration.

Lake Coyote is a playa (∼5 km × 7 km), one of four subbasins of Pleistocene Lake Manix in southern California (Figure 3; Supplementary Figure S1). Lacustrine phases at Lake Coyote occurred at times during the last 25 ka when water in the Mojave River breached the 543 -meter-above-sea-level (masl) threshold, crossing the Minneola fluvial plain and entering the Coyote basin (Meek, 1994). During this time interval, the basin was tectonically stable (Miller et al., 2018).

Distributary paleochannel ridges above the Minneola fluvial plain were first identified by Hagar (1966). Trenches dug into the ridges revealed channel-fill geometry with thicker coarse sand deposits at ridge center tapering laterally to the edge and sediment composition confirmed provenance (Hagar, 1966; Dudsah, 2006). Mojave River deposits contain arkosic sand sourced from the California Transverse Ranges, and gravels of plutonic rocks and metasedimentary rocks, including distinct rounded quartzite (Miller et al., 2018). At the ridge margins, the Mojave River deposits interfinger with lacustrine silt deposits. Further, stratigraphic sections along the Lake Coyote margin show Mojave River sediments (identified by the presence of diagnostic quartzite pebbles, the bed load of the Mojave River) buried by thin lacustrine deposits (Miller et al., 2018).

The spatial configuration of the ridges is best recognized in remotely sensed data. The ridges at Lake Coyote are evident in aerial photographs by dark tone (varnished gravel) of ridges in contrast to lighter-toned plains, and minor vegetation patterns (sparser on crest). Dudsah (2006) mapped two broad paleochannel ridge systems (eastern and western). Miller et al. (2018) conducted a detailed map of individual ridges in sub-meter topographic data, documenting a direct connection to the Mojave River. The eastern set of paleochannel ridges have bulbous terminations, whereas the western set end as individual radial ridges (Figure 3B). The terminal elevation of the western paleochannel ridges merge with an arcuate barrier beach.

Mapped ridges are widespread across the Minneola fluvial plain, covering an area of ∼115 km². These landforms vary in size (Supplementary Figure S2), with ridges ranging from 10–30 m wide, and segments extending in length from a few meters to over a kilometer. The ridges exhibit low relief, generally less than 1 m, with a very shallow northerly slope (∼0.11°) along their crest (Supplementary Figure S2).

Due to their subtle topographic form, the ridges are difficult to recognize from surface expression in the field. A monolayer of clasts blankets the surface of the Minneola fluvial plain, with grain size varying with topography (Figure 4). Gravel is concentrated on the ridge crest, while finer-grained sediment is present in the lowlands. At the ridge apex, the maximum grain size reaches coarse gravel (2 cm), in contrast to the topographic lows where the coarsest fraction is fine gravel (0.5 cm). Correspondingly, median grain size also differed: D₅₀ = 0.9 cm at the crest and D₅₀ = 0.2 cm in the trough. These gravelly fluvial deposits are thin (a veneer) and superpose earlier lacustrine deposits composed of silt-to sand-size grains. In general, ridge sediments are unconsolidated sediments and weakly cemented by halite salt, making them easily excavated by hand. Locally, ridges have soil present beneath the gravel pavement that can reach 5 cm thickness and is well indurated with calcium carbonate.

In a preliminary United States Geological Survey (USGS) geologic map, Dudsah (2006) outlined three working hypotheses for the Lake Coyote paleochannel ridge origin with reference to analog sites: 1) differential erosion of Mojave River channel deposits (raised channels similar to the gravel-capped exhumed river channel systems in Oman; Maizels, 1990), 2) fluvial deposition in a low gradient aggrading system (potentially akin to the Australian Lake Eyre; Lang et al., 2004), or 3) a mixture of both destructive and constructive processes (e.g., a combination of hypotheses 1 and 2).

Observations support both destructive and constructional processes (favoring model 3), which highlights the complex history preserved in the Lake Coyote paleochannel ridges. Importantly, these two paleochannel ridge types are gravel-capped ridges that are indistinguishable in undisturbed surface observations. Clast armoring by river-transported gravels protected finer-grained sediment in this deflated landscape. Thus, the gravel-capped ridges are interpreted as raised channels, differentially eroded stream channel deposits, with the low-lying finer grain deposits between ridge crests interpreted as marsh and floodplain deposits (Hagar, 1966; Dudsah, 2006; Miller et al., 2018). However, at the lake margin the geomorphic and sedimentologic signature of streams interacting is consistent with constructional channel deposition (e.g., gravel-capped subaqueous ridge).

The Lake Coyote site is noteworthy for the positive-relief geomorphic expression of the stream-to-lake margin with two varieties of constructional subaqueous ridges. The eastern paleochannel ridges end very close to the maximum Lake Manix highstand (543 masl at 24.5 ka; Figure 3B), potential evidence that the river flowed into standing water (Dudsah, 2006; Miller et al., 2018). Miller et al. (2018) proposed that rivers entering the Coyote basin encountered a rapidly rising lake such that river deposits were buried by lacustrine sediments as water levels rose, a situation they suggested inhibited delta formation.

In contrast, the western paleochannel ridges are interpreted as constructional channels whose sediment was reworked to form an arcuate barrier beach. Paleochannel ridge elevation matches the barrier beach at an altitude of ∼541 masl. In this area, Miller et al. (2018) documented river channel position moved progressively eastward from stratigraphic relationships of the radial ridges and confirmed by radiometric dating (Figure 3B). They propose that wave action and lake currents along the beach transported north-flowing bed load away (along shore), explaining the source of gravel sediment and overall shape of the arcuate barrier beach.

Lake cycles were initially constrained by geomorphic mapping to establish a relative sequence. In situ stratigraphic sections and radiometric dating of lacustrine shells were then used to establish an absolute chronology (Oviatt, 1997; Miller et al., 2018). These data allowed Miller et al. (2018) to identify six lake cycles spanning ∼5 ka, with an associated fluvial sequence as the paleochannel ridges progressively stepped basinward (Figure 3B). Radiometric ages differentiate depositional phases for individual radial ridges (‘spokes’) at the end of the western paleochannel ridges, and the apparent distributary configuration of ridges is a time-integrated pattern. The eastern paleochannel ridges formed first, followed by sequential radial ridge accretion along the western lake margin (Figure 3B). The paleochannel ridges record a time sequence of river channel switching.

Lake Coyote had a minimum fill time of 30 years based on estimated river discharge rates, but the integrated regional climate-hydrological record indicates sustained lake stands with duration of several hundred years (Wells et al., 2003; Miller et al., 2018). This highstand longevity argues against paleochannel deposits occasionally drowned by the lake, and supports the river-lake confluence model.

Lake hydrographs are often compared to other paleoclimate records to infer information on regional or global trends. Reheis et al. (2015) compared the Lake Manix hydrograph with ice core, marine and spleothem paleoclimate records. They documented three highstands for Lake Manix that coincide with known climate events (abbreviated here as Henrich events: H2, H3 and H4). Supplementary Figure S3 shows the reconstructed lake level curve for Late Pleistocene Lake Manix. The subaqueous paleochannel sequence at Lake Coyote corresponds to the most recent warming period, occurring after the H2 event.

Arabia Terra is the northern extent of the martian ancient cratered highlands and is noteworthy for the relative scarcity of valley networks compared to other Noachian-era terrains (Hynek et al., 2010). Fluvial systems in Arabia Terra were active between the mid-Noachian to Early Hesperian (Davis et al., 2018). With the recognition that some martain fluvial systems are preserved as topographic highs due to differential erosion (e.g., Dickson et al., 2021), paleochannel ridges help bridge the gaps in some valley networks and expand the record of surface runoff on Mars.

In Arabia Terra, Davis et al. (2016) mapped a contiguous pattern of fluvial landforms showing negative-relief valleys transitioning to positive-relief ridges (‘inverted channels’). This morphological transition was interpreted as crossing an erosional boundary, reflecting a complex history of burial and exhumation (Davis et al., 2016; 2018).

Northeastern Arabia Terra has long been recognized as the site of an extensive mantle up to several hundred meters thick that covers the ancient Noachian cratered terrain (Moore, 1990; Fassett and Head, 2007; Zabrusky et al., 2012). The horizontally layered mantle is thought to be a regional aeolian, dust or pyroclastic deposit that has been extensively denuded to form scabby or ‘etched’ terrain (Edgett, 2005; Hynek and Di Achille, 2017), with an estimated original etched unit thickness of ∼800 m at Phison Patera (Zabrusky et al., 2012). Davis et al. (2019) demonstrated the high correlation (∼80%) of paleochannel ridges and the former extent of etched terrain in Arabia Terra. They suggested that the blanketing etched terrain may have protected the paleochannels from erosion such that the low-relief ridges visible today are the final stages of the significant deflation experienced across Arabia Terra. With this history of burial and erosion, the Phison Patera paleochannel ridges are exhumed channels.

Phison Patera is a topographic basin located in northeastern Arabia Terra, centered near 30°N, 48°E (Figure 6). [The International Astronomical Union name is for an albedo feature originally mapped by Antoniadi (1930)]. Identified as a quasi-circular depression in MOLA data by Frey et al. (2002) and dated to the Early Noachian, it is an eroded composite impact basin. As a large regional sink, valley networks incise the perimeter but lack triangular deposits at their termini. Davis et al. (2019) first proposed this site as a candidate paleolake based on transitions from valley to exhumed channels terminating within a ∼400 km-diameter closed basin (∼10⁵ km²). This location is not included in martian crater lake databases (Fassett and Head, 2008; Goudge et al., 2012).

To evaluate a possible paleolake at the Phison Patera basin, we considered morphological evidence for consistency with standing water at this site. First, we assessed candidate lake levels based on the transition zone between valleys and exhumed channels. Second, we examined the spatial distribution of indurated crater fill that are likely remnant paleolake deposits, as described in Section 3.3.2.

The plains northeast of Hellas Planitia have drawn a lot of research interest due to the diversity of interior deposits within crater basins. Martian alluvial fans are geographically restricted, with one regional cluster of megafans (10–40 km radial length) present within craters in Southwesten Tyrrhena Terra, including Harris crater (Moore and Howard, 2005; Wilson et al., 2021). Very few (N = 2) of the intracrater fans in this region are interpreted to be deltaic (e.g., Wilson et al., 2021). Nevertheless, multiple closed basin lakes are identified in this region, all of which lack a triangular fan deposit (Goudge et al., 2015). Some craters have layered sedimentary deposits (Malin and Edgett, 2000) containing hydrated minerals that are thought to be lacustrine in origin (Ansan and Mangold, 2004; Wilson et al., 2007).

Harris crater has a diameter of 82 km and is located at ∼ 21.9°S, 66.8°E (Figure 10). The Harris impact event occurred between Early Hesperian and Early Amazonian (∼3.7 to ∼3.3 Ga) and is part of a fresher class of craters thought to postdate the main era of valley network activity (Mangold et al., 2012). Gullikson et al. (2023) mapped water-related landforms in the northern region of Hellas Planitia, including at Harris crater (Figure 10). Five intracrater fans at Harris crater were previously interpreted as alluvial fans (Figure 11; Williams et al., 2011; Anderson et al., 2023; Wilson et al., 2021). In this study, we evaluate if there was lacustrine overprinting of the original alluvial deposits.

Drawing upon a terrestrial analog, we demonstrate that some depositional ridges on Mars may have formed subaqueously in lakes. This new insight paves the way for future studies to use this landform in detailed lake hydrograph reconstructions. On Mars, consistency tests were passed in evaluating two case studies, substantiating the plausibility of subaqueously-formed ridges. We examined commonalities between paleochannel ridge onset elevation and termination points of valley networks to assess possible paleolake levels within topographic basins. Additional morphological evidence of paleolake deposits was present at both sites: indurated crater fill at Phison Patera, evaporite salts and desiccation polygons at Harris crater. Our two examined sites differ in physical scale and energy setting, raising the prospect for considering constructional paleochannel ridges in detailed geologic studies.

We find the case for a paleolake at Phison Patera stronger than the one at Harris crater and acknowledge that the available data does not support a unique interpretation (Table 1). Critically, both sites retain a valid alternative explanation, namely the original raised channel hypothesis. A subaqueous ridge origin for some ridges at the studied locations is a reasonable interpretation with a comparable level of confidence as previous findings with the available observational evidence. Future work could further evaluate paleolake feasibility through three-dimensional modeling of fluvial and groundwater inputs. Mineralogical signatures of salts in spectral data, especially if localized to the basin, would further corroborate a paleolake interpretation.

This proof-of-concept study provides a basis for examining the formation setting of other martian paleochannel ridges, especially in cases where they converge within a topographic basin as would be consistent with a subaqueous origin. Geomorphic mapping is a way to synthesize the geologic context. Factors to consider include the relationship of paleochannel ridges to modern topography, the regional drainage pattern, the spatial distribution of water-formed minerals, and the presence of lacustrine landforms (e.g., deltas, barrier beaches, lake terraces, layered sedimentary deposits, shrinkage polygons, etc.).

Recognition of paleochannel ridges as lake level indicators would critically inform 1) identification of additional paleolake sites, especially in degraded terrains, and 2) future hydrograph reconstruction through constraints on lake extent and water level sequence. The significance of this finding is that it introduces a new shoreline-associated landform for reconstructing lake-level fluctuations, a tool for deepening our understanding of hydrology and habitability on Mars. By assessing paleochannel ridges as shoreline indicators, new insights into the regional and/or global martian climate oscillations can be probed.