Deposits from giant floods in Gale crater and their implications for the climate of early Mars

This study reports in-situ sedimentologic evidence of giant floods in Gale crater, Mars, during the Noachian Period. Features indicative of floods are a series of symmetrical, 10 m-high gravel ridges that occur in the Hummocky Plains Unit (HPU). Their regular spacing, internal sedimentary structures, and bedload transport of fragments as large as 20 cm suggest that these ridges are antidunes: a type of sedimentary structure that forms under very strong flows. Their 150 m wavelength indicates that the north-flowing water that deposited them was at least 24 m deep and had a minimum velocity of 10 m/s. Floods waned rapidly, eroding antidune crests, and re-deposited removed sediments as patches on the up-flow limbs and trough areas between these ridges forming the Striated Unit (SU). Each patch of the SU is 50–200 m wide and long and consists of 5–10 m of south-dipping layers. The strike and dip of the SU layers mimic the attitude of the flank of the antidune on which they were deposited. The most likely mechanism that generated flood waters of this magnitude on a planet whose present-day average temperature is − 60 °C was the sudden heat produced by a large impact. The event vaporized frozen reservoirs of water and injected large amounts of CO2 and CH4 from their solid phases into the atmosphere. It temporarily interrupted a cold and dry climate and generated a warm and wet period. Torrential rainfall occurred planetwide some of which entered Gale crater and combined with water roaring down from Mt. Sharp to cause gigantic flash floods that deposited the SU and the HPU on Aeolis Palus. The warm and wet climate persisted even after the flooding ended, but its duration cannot be determined by our study.

www.nature.com/scientificreports/ This geological history led to the extensive exposures of sedimentary rocks in Gale crater, making it an ideal place to study the evolution of early Mars [21][22][23] . Gale crater has two physiographic provinces: Aeolis Mons, also known as Mt. Sharp, and the crater floor (Fig. 1A). Mt. Sharp is a crescent shaped mound near the center of the crater and consists of 5 km of sedimentary rocks (Fig. 1A). The crater floor is a ring-like depression that surrounds Mt. Sharp (Fig. 1A). Its northern portion is named Aeolis Palus (Fig. 1A). Many channels cut the rim of Gale crater (Fig. 1A). The largest one is Farah Vallis that is 3 km wide and 700 m deep (Fig. 1A). In addition, huge canyons are also carved into Mt. Sharp (Fig. 1A). One of them , called the Grand Canyon of Gale crater, is about 2 km wide and is 200-300 m deep (Fig. 1A).
The Curiosity rover landed on the Bradbury Landing Site on the northern margin of Aeolis Palus in August of 2012 (Figs. 1B, S1). It has been conducting detailed investigations along its path from the landing site toward Mt. Sharp (Figs. 1B, S1). Six rock units were identified before and after the landing of the rover (Fig. 1C). They are the Bedded and Fractured Unit (BFU), the Hummocky Plains Unit (HPU), the Striated Unit (SU), the Rugged Terrain Unit (RTU), the Murray formation (MF), and the Stimson formation (SF). Each has specific orbital attributes and distinct in-situ lithological and sedimentological characteristics 11,12 . Our goal is to present sedimentological characteristics and depositional environments of two of these rock units: the SU and the HPU (Fig. 1C-E). To conduct this investigation we have used images from High Resolution Imaging Science Experiment (HiRISE), from the Mast-mounted cameras (Mastcams), and from Mars Hand Lens Imager (MAHLI) camera mounted on the arm of the Curiosity rover. Our study presents a documentation of large floods in Gale crater based on in-situ, rover-based observations.

Results
The HPU and the SU are exposed in an 8 km by 8 km area on the northern areas of Aeolis Palus (Fig. 1C). They occur in well-defined stratigraphic order: The SU consistently overlies the HPU. In addition, both are overlain by the RTU throughout their exposures (Fig. 1C).
The HPU is distinguished by its dark gray tone, smooth surface, and sparse impact craters on HiRISE images (Fig. 1D). Lithologically, the HPU is composed of poorly sorted conglomerate with particles ranging in size from silt to boulders throughout its exposures along the rover's path (Figs. 2A, S2). In few localities where its internal structure is exposed, the HPU displays large cross beds that are 2-6 m high and up to 60 m long (Figs. 2, S2, S3). The flow direction is directed toward north as inferred from the cross beds ( Figs. 2A, S2). Most importantly, the HPU displays linear ridges which occur in variable states of preservations (Fig. 1D). These ridges are relatively straight, parallel, and about 500-800 m long in their exposed outcrops (Fig. 1D). Their crests trend at N60° E (Fig. 1D), are nearly flat, and appear to be eroded (Fig. 3). Ridges are symmetrical with each limb sloping about 4° to 9° (Fig. 3). The distance between these ridges is about 150 m (Figs. 1D, 3). Their present amplitude ranges from 5 to 10 m (Fig. 3). The base of the HPU is not exposed but the presence of 10 m-high ridges suggest it is a least 10 m thick (Fig. 3). These ridges are best preserved in the Kimberley region (Fig. 1D). They gradually disappear northward (Fig. 3).
The SU occurs in patches rather than as continuous layers (Fig. 1C). In contrast to other strata in Gale crater which are horizontal to nearly horizontal, patches of the SU consists of dipping layers that strike N60° E and have an estimated dip of about 10° SE (Figs. 4A,B, S4, S5). However, the dipping layers become conformable to the surface on which they deposited at about 50-100 m in a down dip direction (Figs. 4C, S6). Each patch is concave upward and extends 100-200 m along the strike and 50-100 m along the dip (Figs. 4A,B, S4, S5). The thickness of the SU in each patch is about 5-10 m.
Most importantly, the SU layers deposited on the south limbs of ridges of the HPU and extend to troughs between them (Fig. 3A). The strike of the SU is parallel to the trend of crests of symmetrical ridges, both at N60° E (Fig. 1D). The SU layers consistently dip at about 10ºSE southeast (Figs. 4A,B, S4, S5) which is similar to the measured 4°-9° southerly slope of limbs of the ridges on which it was deposited (Fig. 3A).
The elevation of the surface on which the SU deposited is about − 4500 m near its northernmost exposures and − 4460 m at its southernmost exposure (Fig. 1C). This represents a rise of 40 m over 3 km of horizontal distance, or a slope of about 1º northward (Fig. 5). That is, the surface on which each patch of the SU deposited rises in elevation from north to south or from the landing site toward Mt. Sharp (Fig. .1B,C). This is also clearly demonstrated by the cross section at the Kimberley area (Fig. 1D). Here, the base of the SU patches rises 6 m in elevation over 300 m of horizontal distance, also indicating that the surface on which the SU patches deposited had a slope 1.2º northward (Fig. 3A).
Two layers of the SU were examined closely by the rover at two different locations. In one area investigated on Sol 586, the SU layer begins with about 60 cm of conglomerate (the b interval). It overlies the conglomerates of the HPU with a sharp but conformable contact (Figs. 4B, 6A,B, S5, S7). The conglomerate of the b interval is thin bedded, grain-supported, and consists of grains ranging in size from 2 to 5 mm (Figs. 6A, S7).
The b interval is overlain by a distinct, 13 cm-thick, normally graded bed with a sharp decrease in grain size (Figs. 6A, S7). The base of the graded bed consists of 5 cm of pebbly sandstone (the d interval) that is grain supported and poorly sorted (Figs. 6A,B, S7).
The d interval grades upward to 7-8 cm of laminated, medium grain sandstone (the e interval) that shows irregular laminations (1 mm to 2 mm thick), and displays soft sediment deformation (Figs. 6A, S7). Laminations are disrupted at two locations on the left side of the image (Figs. 6A, S7). A MAHLI image of the graded bedded interval shows that it is highly unsorted but consists of grains that are well-rounded and dark in color (Fig. 7 www.nature.com/scientificreports/

Discussion
The previous sedimentological study suggested a fluvial-deltaic-lacustrine depositional system for strata examined by the Curiosity rover between the landing site and the Pahrump Hills location 12 . The model advocated that melting of ice and snow on the northern rim of Gale crater generated liquid water that flowed southward and drained into a shallow lake when Gale crater was nearly empty 12 . The proposed system resulted in deposition of four lithologies each representing a specific environment: lacustrine, clinoform (foreset beds) of a delta, fluvial, and alluvial fan (Fig. S9). The model clearly equates the mudstones of the MF to deposition in a lake or a lacustrine environment (Fig. S9). It also presents the SU as the clinoform (foreset beds) of a delta. Authors of this model 12 do not use terms such as the RTU or the HPU in their proposed depositional environments. However, they describe sedimentological characteristics of sandstone outcrops mapped as the RTU and interpret them as river deposits 12 . Similarly, they describe conglomerates mapped as the HPU and interpret them as deposition in an alluvial fan environment 12 . Therefore, the four lithologies of the study area namely the MF, the SU, the RTU, and the HPU www.nature.com/scientificreports/ are equated to four depositional environments such as the lucustrine, delta, fluvial, and alluvial, respectively 12 (Fig. S9). The proposers of the fluvial-deltaic-lacustrine model considered these four rock units to be time-equivalent 12 ; that is, their deposition evolved side-by-side through time. After the lake dried up, these four rock units were buried by about 5 km of sediment that filled Gale crater. They were then exposed along an erosion surface (Fig. S9) when margins of Gale crater were excavated creating the modern morphology of the Gale crater 12 . According to this model, deposition of all four rock units predated the formation of Mt. Sharp 12 .
We have found four major discrepancies between our observations and the proposed fluvial-deltaic-lacustrine depositional environment that was applied to Gale crater 12 . First, this depositional system is very common on Earth 35 , as was also pointed out by the proposers of the model 12 . As a result, we have a clear understanding of its sedimentary record 35 . Basin-ward migration of time equivalent environments of a fluvial-deltaic-lacustrine system on Earth produces a coarsening upward stacking pattern 35 . It consists of lacustrine mudstone at the base which transitions upward to deltaic strata, then to fluvial deposits, and is finally caps by alluvial layers 35 . This stacking pattern is shown schematically to exist in Gale crater (Fig. S9), but it has not been seen anywhere along the rover's path. www.nature.com/scientificreports/ Second, the key component of the fluvial-deltaic-lacustrine interpretation is the SU: the rock unit that was interpreted as a clinoform component (foreset beds) of a delta and is the main focus of our investigation. According to the model, the deltaic clinoform (i.e. the SU) transitioned into the lacustrine mudstone of the MF (Fig. S9). Therefore, the SU layers should overlie the lacustrine mudstones of the MF as the system migrated basinward (Fig. S9). But the SU neither transitions nor overlies the MF anywhere along its exposures imaged by the rover (Fig. 1C). In fact, the SU has never been seen to transition into any rock unit. It consistently overlies the conglomerates of the HPU and underlies the sandstones of the RTU both with sharp contacts (Figs. 3, 4).
Third, patches of the SU occur at successively higher elevations from north to south, that is, from the landing site to Mt. Sharp (Figs. 1C, 5). This characteristic was also noted by previous investigators 12 . They suggested that each exposure of the SU is a small delta that deposited successively at high elevations due to the continued rises in lake level 12 . That is each patch of the SU becomes younger southward (Fig. S9). We demonstrated that the occurrence of the SU at higher elevation southward was because it deposited on a surface that was rising in elevation in that direction (Fig. 5). It was not because the SU became younger southward due to rises in lake level. Such an interpretation would have required the SU and the MF to deposit up-section together as shown schematically to have occurred by these authors (Fig. S9). But the SU has never been seen in contact with the MF along its exposures. In fact, dipping layers of the SU at each patch become tangential to the surface on which they deposited (Figs. 4C, S6); but they neither overlie nor transition into lacustrine mudstone of the MF (Fig. 4C, S6). In fact, the SU does not transition into any other unit along its exposures. As such, each occurrence of the SU could not have been a separate delta that was deposited at higher elevation during rises in lake level.
Lastly, the giant cross beds of the HPU (Fig. 2) suggest that the flow that deposited this rock unit was towards north. Combined with the north-sloping surface on which the SU was deposited, these indicate that the SU and the HPU formed on a north sloping surface. The only north sloping surface in this area is the northern flank of Mt. Sharp. That means that both the HPU and the SU were deposited after the emergence of Mt. Sharp. Another investigation has already documented that the RTU (the rock unit that overlies both the HPU and the SU) was also deposited after the emergence of Mt. Sharp 9 . This means that three out of four components of the fluvialdeltaic-lacustrine interpretation (the SU, the HPU, and the RTU) formed after the emergence of Mt. Sharp, not before it as previous investigators suggested 12 . Therefore, the reason the SU has not been seen in contact with the MF anywhere along its exposures is because the two rock units formed in two different times. One (the MF) was deposited when the crater was empty, then buried by 5 km of strata, and was subsequently exposed when margins of Gale crater were excavated resulting in the emergence of Mt. Sharp. The other three (the SU, the RTU, and the HPU) formed after the emergence of Mt. Sharp when Gale crater had acquired its modern morphology.
These major differences between our observations and the existing interpretation have persuaded us to reexamine the sedimentation history of strata examined by the Curiosity rover in Gale crater in general and the depositional environments of the individual rock units in particular. This contribution primarily addresses the depositional environments of the SU and, secondarily, that of the HPU, as it relates to the main goal of this study.
Proposed Interpretation. Our alternative interpretation for the depositional environments of the SU is based on three characteristics of this unit. They are: (1) the sedimentological features of its layer, (2) the geometry of its occurrences and the southerly dips of its strata, and (3) its relationship with the rock unit that underlies it.
Sedimentologically, the two layers of the SU that were investigated by the rover show systematic upward changes in lithology, grain size, and sedimentary structures (Fig. 6). Each layer begins with a thin-bedded www.nature.com/scientificreports/ conglomerate (the b interval) consisting of granule to coarse pebble size fragments (Fig. 6). Thin layering and the grain-supported texture suggest deposition as bedload. Their sharp plane bedding indicates that they were deposited under a condition commonly referred to as the upper flow regime 36 . The b interval is overlain by a cross bedded granule conglomerate, the c interval (Fig. 6C,D). The decrease in grain size and the change in sedimentary structures from sharp plane layering to cross bedding indicate a decrease in flow strength or a transition from upper to lower flow regime conditions 36 .
The cross bedded c or the thin-bedded b intervals are followed upward by a normally graded capping layer (the d and the e intervals) across a sharp transitional boundary (Figs. 6A, S7). A rapid reduction in grain size, co-deposition of pebble and sand grains, the absence of cross beds, and irregular laminations indicate that the transition from the c (or the b) interval to the d interval involved a rapid and sudden decrease in flow strength 36 . It resulted in co-occurrence of well-rounded granule, small pebble, and sand-size grains (Fig. 7), suggesting that the flow was so weak that deposition was most likely occurred from suspension. Upward grading of pebbly sandstone (the d interval) to a sandstone interval (the e interval) with irregular laminations suggests suspension www.nature.com/scientificreports/ sedimentation continued until the end of the deposition of the normally graded bed (Fig. 6). The presence of soft sediment deformation and water escape structures (Figs. 6A, S7) indicate deposition under subaqueous conditions. Morphologically, the SU occurs as patches rather than as continuous layers (Figs. 1C, 3). In contrast to other strata in Gale crater which are horizontal, layers within patches of the SU have a strike of about N60° E and a dip of about 10° SE (Figs. 4A,B, S4, S5). However, the dipping layers become tangential to the surface on which they were deposited within about 50-100 m in a down-dip direction (Figs. 4C, S6). These patches are 5-10 m thick, concave upward, and are 100-200 m in the strike direction and 5-100 m in the dip direction (Figs. 4A,B, S4, S5).
Patches of the SU occur exclusively between ridges of the HPU conglomerate (Figs. 1D, 3). Most importantly, the strike of the SU layers and the trend of the ridge crests in the HPU are both at about N60° E (Fig. 1D). In addition, the dip of the SU strata is similar to the slope of the limb of the HPU ridge (4°-9°) on which they were deposited (Fig. 3). These relationships strongly suggest that the deposition of the SU is related to the formation of ridges of the HPU. Furthermore, at least in one location along the rover's path in the Kimberley region, the SU sharply truncates and overlies a 60 m-long, 2-3 m thick giant cross bed within the HPU (Figs. 2D, S3).
Systematic changes in lithology, the upward decrease in grain size, and the sequence of sedimentary structures suggest that each layer of the SU is a single sedimentation unit that was deposited by a waning flow. Deposition began with a flow that was strong enough to transport grains as large as coarse pebbles as bedload (the interval b) and terminated by a rapid decrease in flow strength so that sand and granules deposited from suspension (intervals d and e). These characteristics are traditionally attributed to deposition as turbidites: a type of sediment gravity flow deposits in which sediments are kept in suspension by flow turbulence 37,38 . Turbidites are very common in deep water environments of modern and ancient oceans and lakes on Earth 37,38 .
In addition, layers with features similar to those found in the SU have also been documented by waning catastrophic glacial floods of the Pleistocene Epoch on Earth and commonly referred to as rhythmites [39][40][41] . This is particularly apparent when floods drain into lakes, best recorded in Lake Lewis that formed along the path of the Missoula Lake floods 40 .
The absence of strata indicative of deepwater lacustrine sedimentation within the SU layers challenges the turbidite and rhythmite interpretations. In addition, neither of the two hypotheses satisfies occurrences of the SU as concave upward patches of dipping layers. Furthermore, the turbidite and rhythmite hypotheses do not account for the deposition of the SU between ridges of the HPU (see Figs. 1D,E, 3A). Nevertheless, it is clear that layers of the SU were deposited by waning flows; but they are neither true turbidites nor true rhythmites.
Three lines of evidence suggest that the deposition of the SU patches was related to the formation of 10 m-high ridges of the HPU. First, the SU occurs exclusively between these ridges and specifically on their south limbs. This is best expressed in areas where these ridges are well preserved such as the Kimberley region (Figs. 1D,E, 3A). Second, the strike of the SU layers is identical to the trend of ridge crests of the HPU both at N60°E (Figs. 1D,E); and third, the value and the direction of the dip of the SU layers mimics those of the slope of the southern limb of the ridge on which they deposited (Fig. 3A).
These relationships suggest that the deposition of the SU is directly related to the origin of the ridges of the HPU. These ridges are parallel and symmetrical (Figs. 1D, 3A). They have a consistent wavelength of about 150 m and amplitudes that range from 5 to 10 m (Figs. 1D, 3). They consist of unsorted conglomerate with fragments as large as 20 cm across (Fig. 2). In few locations where their internal structures were preserved, they display 2-5 m high, 60 m long cross beds, particularly in their basal part (Figs. 2D, S3). The topmost part to the ridge consists of thin-bedded conglomerate, however. Formation of these ridges cannot be attributed to wind erosion after the deposition of the HPU. This is because wind could not have removed cobbles and boulders of the HPU to carve perfectly symmetrical ridges at identical spacing and at the same time leave behind sandstones of the www.nature.com/scientificreports/ SU layers. In contrast, lithology, morphology, giant cross beds, and thin bedding suggest that ridges of the HPU are giant bedforms produced by moving water. Bedforms of this size and their lithology are characteristics of deposition by huge floods similar to those formed by Pleistocene floods on Earth [42][43][44] . They are referred to as giant current ripples, but are actually giant gravel dunes [42][43][44] . However, giant bedforms (ridges) of the HPU are symmetrical (Figs. 1D, 3); and bedforms with these characteristics are antidunes not dunes 36,[45][46][47][48] . They are also known as stationary waves or standing waves 36,[45][46][47][48] . Antidunes form when flow is strong enough to remove previously deposited sedimentary structures such as dunes and high energy plane laminations 36,[45][46][47][48] . Sedimentation, erosion, and migration of dunes are very different from those of antidunes. In dunes, erosion occurs on the stoss side, deposition takes place on the lee side, and migration is in the same direction as the flow. In antidunes, erosion occurs on the lee sides, deposition on the stoss side. Antidunes migrate down-flow, up-flow, or remain stationary 36,[45][46][47][48] . In addition, the sediment surface and water surface are out of phase in dunes but in phase in antidunes (Fig. 8C). When flow depth decreases, water breaks on antidune crests and partly to completely erodes them, depositing removed sediments on flanks and troughs of antidunes as backset beds 36,[45][46][47][48] . As a result, antidunes are rarely preserved in the rock record except when flow wanes rapidly 36,[45][46][47][48] . Flow that breaks over the crests of antidunes is sediment-saturated and loses its strength quickly, leading to the deposition of fining upward layers which are identical to those found in turbidites 37,38 or rhythmites [39][40][41] . The main difference is that backset beds occur as dipping patches because of deposition on sloping flanks of anitdunes; whereas true turbidites and rhythmites occur as continuous, horizontal layers [37][38][39][40][41] . In essence, we have discovered a new environment and a new mechanism in Gale crater, Mars, where layers with sedimentological characteristics identical to turbidite deposited by processes other than sediment gravity flows.
Indeed, the 10 m-high antidunes of the HPU that consist of cobble-and boulder-size fragments indicate fast-moving flows such as floods. However, the presence of 2-5 m high cross beds in few locations suggests these floods initially deposited as giant gravel dunes (Figs. 2D, 8B). As flooding continued and flow gained additional strengths, gravel dunes were partly to completely eroded and were replaced by 10 m-high antidunes (Figs. 3A, 8C). The remnants of eroded giant dune cross beds were seen in several locations along the path of the rover (Fig. 2D). When flooding waned, water broke over the crests of antidunes ridges of the HPU and began to erode them (Fig. 8D). Eroded sediments were re-depositing on the up-flow limbs (southern flanks) of antidune ridges as south-dipping patches of backset beds (Figs. 3A, 8E). The erosion of antidune ridges was variable along the flow path which was from south to north. In the southern up-flow regions (high elevations), antidunes suffered minor erosion and were well preserved possibly because the rapid waning exposed these ridges preventing their erosion (Figs. 3A, 8C). In the northern down-flow areas (low elevations), however, ridges experienced major erosion to the point that they are not easily identifiable (Figs. 2B, 4A, 8E). They were totally eroded north of the Dingo Gap region (Fig. 1C).
The southerly dips of the backset beds gave the false impression that the main flow direction was from the north to the south delivering sediments from the northern rim of the crater to form a clinoform delta in a lake near the center of Gale crater 12 . However, this study demonstrated that southerly dips of the SU layers was due to their deposition on south-facing limb (up-flow flanks) of antidunes as backset beds when the main flow was northward delivering sediments from upland regions (Mt. Sharp), to lowland areas northward (Fig. 8A). Initially, backset flows were strong enough to carry coarse pebble fragments as bedload and leave behind secondary bedforms which were up to 1.5 m tall (Fig. 4C). However, the rapid transition to a graded bed layer suggests that backset flow velocity decreased quickly so that granule, small pebbles, and sands deposited from suspension in the capping graded beds of each layer (Figs. 6,7). This interpretation suggests that the SU and the UPU formed during the same event, and that event was a gigantic flash flood on the northern flank of Mt. Sharp (Fig. 8).
Backset beds of the size and magnitude seen in Gale crater, Mars, have not yet been discovered in Earth's fluvial depositional system. However, centimeter-size backset beds identical to those of Gale crater have been reported in ancient rocks and modern rivers on Earth 49,50 , and have been extensively documented in flume studies using fine sand grains [45][46][47][48] .
Wave length (L) of antidune ridges in the HPU can be used to estimate flow depth (h) and mean flow velocity (U) via Eqs. (1) and (2), respectively 45-48 : where g is the Martian acceleration due to gravity (3.72 m/s 2 ) 51 . The wave length of antidunes in the Kimberley region is about 150 m suggesting that flow depth at this location was about 24 m and that flow velocity was about 9.5 m/s. A 24 m deep flood at the Kimberley region would have extended for 11 km from the crater wall to the west and Mt. Sharp to the east. This is consistent with the distribution of the HPU in this area of Gale crater (see Fig. 1C). As such, the discharge of flood would have been about 2.75 10 6 m 3 /s. We consider our calculate discharge as a minimum. However, its exact value will not affect our results because our conclusions for the presence of floods in Gale crater are based on physical sedimentology. The discharge would have depended on the amount of water vapor in the Martian atmosphere at time of flooding (see below). It is important to note that our calculated discharge for floods of Gale crater is 7-10 times lower than the discharge for the Pleistocene floods on Earth such as Missoula Lake and Altai Mountain floods 2 calculated to be about 10 7 m 3 /s, and 100-1000 times lower than discharge through major outflow channels on Mars 2 estimated to range from 10 8 to 10 9 m 3 /s. www.nature.com/scientificreports/

Source of flood waters
Mars is presently so cold (average temperature of -60 °C) that water freezes and ice subsequently sublimates 52 . In fact, low solar luminosity would have kept Mars below freezing temperatures during its entire history regardless of how much green houses gases such CO 2 were present [53][54][55][56] . Despite the cold climatic condition, abundant geological, morphological, and sedimentological features indicate that liquid water flowed on Mars and in some cases accumulated in lakes 1,12,34,57,58 . Therefore, the question of how liquid water formed has been debated extensively 59,60 . This means that we either need a mechanism to warm Mars to above freezing temperature or prevent water from freezing under a cold planetary condition. Suggestions include geothermal heating 1,61 , CO 2 clouds 62 , asteroid impact 63 , high water salinity 64 , volcanism 65 , orbital changes 66,67 , and high methane concentration [68][69][70][71] .
At the present state of knowledge all of these mechanisms are plausible. Additional in-situ observations and data are needed for a definitive solution to how liquid water formed and flowed on a cold planet like Mars. Nevertheless, two processes appear to be more likely than others to have produced the giant floods that caused deposition of the SU and the HPU in Gale crater. The first is the conceptual model proposed to explain floods through outflow channels of Mars 1 . According to this scenario, the cold and dry Mars was punctuated by intervals of geothermal heating associated with volcanic activities. The event melted glaciers, similar to the generation of Icelandic floods 72 , and resulted in a rapid release of liquid water which flowed to Gale crater via a series of ancient drainages on the Southern Highlands. Water vapor combined with abundant CO 2 and CH 4 released from their solid reservoirs produced a warm and wet climate 1 .
The second possibility is the warming of Mars by a giant impact that vaporized glacial ice, melted solid CO 2 reservoirs, and released a large amount CH 4 from gas hydrate sources 63 . Subsequent cooling would have resulted in condensation of water vapors to produce planet-wide torrential rain 63 .
The warming by an impact is a more likely to have triggered floods of Gale crater than heating by a geothermal event. This is because over seven years of investigation, no evidence of ice-covered lakes such as dropstones or glacial deposits have been found in lacustrine strata of Gale crater [8][9][10][11][12][13][14][15][16][17][18][19][20] . This suggests that not only the source region was warm enough to generate liquid water but the sink area where water accumulated was also warm preventing water from freezing. This requires an instantaneous planet-wide heating event that would have resulted from a large impact 63 .
A giant impact may have punctuated the cold and dry climate of Mars and created an interval of a warm and wet one. A large volume of water vapor was generated from frozen reservoirs. Subsequent cooling resulted in a period of planetwide torrential rain. Runoff would have flown downhill from vast areas of the Southern Highlands and poured into a bowl-shaped, tilted depression such as Gale crater basically from all sides. This water would have combined with that flowed downhill from Mt. Sharp producing gigantic flash floods that deposited the SU and the HPU on Aeolis Palus (Fig. 8).
As such, the flood discharge was not limited by the catchment area in Gale crater; rather, it would have been controlled by the volume of water vapor in the atmosphere which itself depended on the total amount of water available on Mars at that time. Estimated water budget of Mar ranges from 34 to 550 GELs (global equivalent layer) of water 73 . We consider a value of about 150 GELs of water, estimated by hydrogen isotope analysis 74,75 , as a reasonable minimum approximation. This indicates that the Late Noachian had at least 5 times more water than the present-day Mars, which is estimated to be 34 GELs 52,76 . If all of the 150 GELs of water was put into the atmosphere as water vapor, then gigantic flash floods of Gale crater were the most likely outcome. In fact, sufficiently large liquid water would have been available to accumulate on Northern Lowlands forming Oceanus Borealis [77][78][79] .
The flooding event may have lasted only a few days. But, the warm and wet period persisted for much longer. This is because flood deposits of Gale crater (the SU and the HPU) are overlain by the RTU throughout their exposures (Fig. 1C). The basal strata of the RTU consist of thin-bedded fine grain sandstone 9 and have been interpreted as a lacustrine sedimentation 80 . This suggests that the warm and wet conditions continued even after the flooding ended. However, its duration cannot be determined by our study. It may have lasted from few weeks to few centuries depending on the size of the impactor and the amount CO 2 and CH 4 in the atmosphere 63 ; but it was certainly not long enough to cause the complete removal of unstable minerals by chemical weathering, leading to their preservation in Gale crater strata [81][82][83] . It will take longer than a few decades or several centuries of warm and wet condition to remove all unstable minerals 84 .
Since this warming event occurred when Gale crater had acquired its modern morphology, torrential rain may have also carved several huge canyons on Mt. Sharp (Fig. 1A). One of them, called the Grand Canyon of Gale crater, is 2 km wide and 200-300 m deep (Fig. 1A). It is also very likely that flooding of Gale crater was a part of the proposed intense fluvial activity that affected the entire Mars at the end of the Noachian period 32 .

The age of the flooding and the potential impact that triggered it
Neither the SU nor the HPU have any features that can be used to determine the timing of their deposition. But, both rock units consistently underlie the RTU throughout their exposures (Fig. 1C). The top surface of the RTU is so heavily cratered that it was initially mapped as the cratered surface 11 . This indicates that deposition of the RTU occurred when the heavy bombardment of Mars had not ended. This event is attributed to the boundary between the Noachian and the Hesperian periods of the Martian history and considered to have occurred around 3.6 billion years ago 85 . Therefore, this relative dating indicates that flooding had already occurred prior to the end of Noachian Period. That is, floods are older than 3.6 billion years.
Our relative dating of floods is supported by a crater count study 33 . The HPU, the SU, and the RTU occur on the northern part of the crater floor or on Aeolis Palus (Fig. 1C). With all likelihood, they are the same strata that were named the crater floor layer whose top surface yielded a crater count date of 3.6 billion years 33  www.nature.com/scientificreports/ Therefore, floods occurred after the formation Gale crater but before 3.6 billion years ago. Unfortunately, there is a considerable uncertainty about the age of impact craters on Mars depending on methods used 34 . For example, the proposed age for Gale crater ranges from the Early Noachian to Early Hesperian, a difference of over 400 million years 34 . A recent study suggested that the most likely age of Gale crater is the Middle Noachian 34 . On the basis of these constraints, we tentatively suggest that flooding is not younger than 3.6 billion years and not older than 3.85 billion years. Therefore, we searched for impact craters within this age range. In addition, the triggering impactor must have been large enough to initiate a planet wide warming to melt frozen reservoirs of H 2 O, CO 2 , and CH 4 . As such, we limited our search to impactors which produced craters with diameters ranging from 200 to 400 km which occurred between 3.85 to 3.6 billion years. Of course the size of these impact craters is partly related to the size of the impactors. Other factors such as density, velocity, and the angle of the impactor, as wells as characteristics of the target, will also play roles 86,87 .
With these considerations, four impactors are likely to have triggered the warming episode that initiated the floods discussed in our investigation; the one that caused the Antoniadi crater (3.83, 4.01, 21.4, 60.8), the one that formed the Vinogradov crater (3.76, 210, − 19.8, − 37.7), the one that produced the Lowell crater (3.72, 199, − 50.0, − 81.4), and the one that created the Galle crater (3.61, 223, − 50.6, − 30.9). Data in the parentheses are from 34 and include (from left to right): the age of the crater in billions of years before present based on Hartman's isochron, crater diameter in kilometer, latitude, and longitude, respectively. Future refinement in age dating should resolve this issue.

Conclusions
Our study suggests that the SU and the HPU were deposited by gigantic flash floods after the emergence of Mt. Sharp when Gale crater had acquired its modern morphology. The flood formed giant gravel antidunes of the HPU which were at least 10 m high and 150 m apart. The SU deposited when floods waned eroding ridge crests and depositing sediments on the up-flow sides of antidunes of the underlying the HPU. Each patch is concave upward, consists of 5-10 m of south dipping layers, and is 50-200 m long and wide. At the Kimberley region, flood waters were at least 24 m deep and had a velocity of 10 m/s. The most likely cause of flooding was melting of ice by heat generated by a large impact. The event also mobilized CO 2 and CH 4 from their solid reservoirs that combined with H 2 O vapor to form a short period of warm and wet climate. Condensation of water vapor produced torrential rain planetwide some of which entered Gale crater and combined with water coming down from Mt. Sharp to produce gigantic flash floods that deposited the HPU and the SU.

Materials and methods
The study was conducted by analysis of images taken by Mars Hand Lens Imager (MAHLI) and Mastcam cameras on board the Curiosity rover. MAHLI is a 2-megapixel color camera with a focusable macro lens that is mounted on Curiosity's robotic arm to investigate rocks and minerals in Gale crater 88 . It acquires in-focus images at working distances ranging from 2.1 cm to infinity. It is capable of resolving fine sand grains 89 . Mastcam cameras are a multispectral imaging system, which consist of two digital cameras mounted on the rover's mast (1.97 m above the ground). The left and right cameras have 34 mm (M-34) and 100 mm (M-100) focal lengths respectively, yielding pixel scales of 0.22 and 0.074 mrad/pixel, respectively. Mastcams are capable of full color panoramic and stereoscopic measurements 89 . Mastcams images were used to delineate sedimentary structures, sedimentary facies, stratal bounding surfaces and sedimentary architecture and determine dip directions of bedding. MAHLI and Mastcam images were used for sedimentological and textural characterization of targets. Criteria for identification of targets as sedimentary rocks include the presence of rounded grains, sedimentary structures, and layering. Then, dominant grain size was used to classify targets as mudstones or sandstones.