Contents lists available at ScienceDirect

Icarus

journal homepage: www.elsevier.com/locate/icarus

Mars: Formation and fate of a frozen Hesperian ocean

Michael Carra,⁎, James Headb

aUS Geological Survey, 345 Middlefield Rd, USAb Brown University, Providence, RI 02912, USA

A B S T R A C T

Late Hesperian-aged, circum-Tharsis floods interpreted to have formed by catastrophic release of groundwater cut large channels and debouched significantquantities of water into the northern lowlands of Mars. The floods are thought by many to have formed an ocean of significant volume and depth, encircled bycontacts that have been interpreted as shorelines. Models of catastrophic groundwater release require a thick and continuous cryosphere with mean annual tem-peratures well below freezing much like those today. In this environment, the bodies of liquid formed by individual outflow events would have been very short lived,undergoing rapid freezing. We investigate the case where floods repeatedly flowed into the northern lowlands under climatic conditions that resemble those of thepresent day; the water from each flood froze in a geologically short period of time to form an ice layer. Successive ice layers accumulated to form an ocean-sized bodyof ice that filled the basin up to the -3650 contour, thereby enclosing a 110 m global equivalent layer (GEL) of water. Subsequent to the filling of the basin the iceslowly sublimated into the atmosphere to be lost to space or to accumulate in various surface and near-surface cold traps, such as the polar layered deposits. Where isthis excess water today? The presence of the thick global cryosphere meant that only minor amounts of water were lost from the surface back into the globalgroundwater system. Approximately 20-30 m GEL of water is estimated to be at or near the surface today and exchanging with the atmosphere on geologic time scales(this includes the polar layered deposits and deposits elsewhere at depths less than approximately 80 m). Below the exchangeable reservoir is a non-exchangeablereservoir of unknown capacity. Present day loss rates to space fall far short of those needed to eliminate the mid-Hesperian ice ocean and those needed to cause theobserved doubling of the D/H of the exchangeable reservoir since the mid-Hesperian. The discrepancy implies earlier loss rates must have been higher. Assuming alinear increase in loss rates because of the higher early EUV output of the Sun, we estimate that for a present inventory of 30 m GEL, 42 m GEL remains to beaccounted for. Possibilities include greater dependence of losses of EUV than assumed, and enhanced losses during periods of high obliquity. In addition, largevolumes of ice may be present near the surface at high latitudes outside the polar layered deposits as indicated by recent discoveries of ice layers in cliffs.

1. Introduction

The possibility that the northern plains of Mars could have oncecontained ocean-sized bodies of water has long been recognized (e.g.Baker et al., 1991; Lucchitta et al., 1986; Parker et al., 1989, 1993).Large outflow channels, widely interpreted as cut by floods, drain intothe plains; the plains have low slopes and roughness (Kreslavsky andHead, 2000) comparable to terrestrial ocean basins; many features onthe plains resemble sub-glacial terrestrial landforms and could haveformed under a frozen ocean (Kargel et al., 1995); long, continuousbreaks in slope surrounding the plains could be shorelines (Parker et al.,1993); widespread fluvial dissection of the highlands may imply anextended water source (see summary in Carr, 2006). More recently,Costard et al. (2017) and Rodriguez et al. (2016) have found geo-morphic features that they interpret as evidence of impact-generatedtsunamis.

The former presence of oceans has been suggested by two main linesof reasoning. First, high erosion rates and widespread valleys and lakesin the cratered highlands suggest that late Noachian Mars could havehad a hydrologic cycle with evaporation from an ocean, precipitation

on the ancient highlands to form the valley networks and drainage backinto the northern lowlands basin (e.g. Baker et al., 1991; Craddock andMaxwell, 2002). Second, a precipitous drop in rates of erosion andvalley network formation at the end of the Noachian has been inter-preted to indicate that the climate changed from one that was warmand wet, or warm and arid (Craddock and Howard, 2002; Ramirez andCraddock, 2018), to an extremely cold and dry climate more like thepresent, and that a thick cryosphere developed during the transition(Clifford, 1993; Carr, 1996). Large outflow channels then formed, in-terpreted to be the result of massive eruption of groundwater trappedbelow the cryosphere (Carr, 1979; Andrews-Hanna and Phillips, 2007).The channels drained into the northern plains to form large sea orocean-scale bodies of water that would then have frozen in the globalcold and arid climate (Kreslavsky and Head, 2002a; Turbet et al.,2017). The size and fate of this later Hesperian-aged ice-ocean thatformed contemporaneously with the outflow channels, according to thishypothesis, is the main concern of this paper. A possible earlier Noa-chian-aged ocean is mentioned only peripherally.

Almost everything about the proposed Hesperian ocean is con-troversial (see summary in Head et al. (2018)) including the cause, the

https://doi.org/10.1016/j.icarus.2018.08.021Received 16 March 2018; Received in revised form 9 July 2018; Accepted 23 August 2018

⁎ Corresponding author.E-mail address: mhcarr265@gmail.com (M. Carr).

Icarus 319 (2019) 433–443

Available online 22 September 20180019-1035/ Published by Elsevier Inc.

T

number, magnitude and timing of the contributing outflow events,whether the water would have quickly frozen so that any standing bodyof liquid water would have been short-lived, how compelling the evi-dence for tsunamis is, and where all the water subsequently went.Climatic considerations are a major source of skepticism that therecould have been a long-lived liquid ocean in the Hesperian, one thatsurvived long enough to erode topographically identifiable shorelinesalong its shore and to have experienced impact generated tsunamis. Along-lived, liquid, high-latitude ocean would require near-equatorialtemperatures close to 273 K. Partly because of the less-luminous earlySun, greenhouse warming of a thick CO2eH2O atmosphere falls farshort of that required to raise the effective surface temperature from thepresent 215 K to in excess of 273 K (e.g. Kasting, 1991; Haberle et al.,2017). The supplemental effects of other greenhouse gases, such as SO2,CH4, NH3, H2 and N2O have also been studied, but for a variety ofreasons all, either independently or in combination, fail to give therequired long-lived boost in temperature (summarized in Haberle et al.,2017). Temporary warming as a consequence of large impacts has alsobeen examined (Segura et al., 2008; Palumbo and Head, 2017; Turbetet al., 2017) but such effects are more likely in the Noachian, whenimpact rates were much higher, than in the Hesperian. Complete dis-cussion of all the controversies is beyond the scope of this paper. Weacknowledge that large amounts of water flowed into the northernplains during large floods but are skeptical that there was a long-livedliquid ocean. Alternatively, we propose that the water from each suc-cessive flood rapidly froze to form a layer of ice (Kreslavsky andHead, 2002a; Turbet et al., 2017) and that multiple floods resulted inaccumulation of a long-lived, ocean-sized body of ice, but not an ocean-sized body of liquid water. We assume that the ice slowly sublimated aswater was lost to space and as changes in obliquity caused redistribu-tion of the water. The formation and fate of this body of ice is the mainconcern of this paper.

Parker et al., (1989, 1993) and Clifford and Parker (2001) identifiedseveral features encircling the northern basin that they interpreted asshorelines of an ocean. The youngest and most easily identifiable ofthese proposed shorelines (the Utopia shoreline, or contact 2) is closestto an equipotential surface (Head et al., 1999). In order to explore thevalidity of the Hesperian ocean hypothesis, we interpret this distinctivecontact as marking the edge of a frozen ocean, and we assess the im-plications of this hypothesis. We re-estimate the volume enclosed bythis shoreline in light of recent geologic mapping and THEMIS data. Wethen try to reconcile this ice volume with the present near surface in-ventory of water. We re-examine the amounts of water that could havebeen lost to sinks (such as to space and to groundwater), and theamounts gained from sources other than the interpreted ice-ocean.Again, our concern here is exclusively with the youngest and bestpreserved of the proposed shorelines (Contact 2 or the Deuteronilusshoreline). We recognize that there could have been additional, oldershorelines of Noachian age that imply much larger volumes(Clifford and Parker, 2001), but these are not discussed. We are con-cerned here exclusively with the volume of the supposed Hesperian ice-ocean and where the water went.

We use the contrast between the present atmospheric D/H(Villanueva et al., 2015) and the D/H in sediments from Gale Crater(Mahaffy et al., 2015) as one constraint on how the near surface in-ventory of water may have evolved over time. The age of Gale crater ispoorly constrained but crater counts on, and superposition relations of,the sediments in the central mound suggest that they are 3.6–3.8 b.y.old and straddle the Noachian –Hesperian boundary (Thomson et al.,2011). In contrast, the outflow channels and the sediments of thenorthern plains that were deposited by them are upper Hesperian in age(3 to 3.5 b.y. old) (Tanaka et al., 2005). The sediments at Yellowknifebay in Gale Crater unambiguously accumulate in a lake that may havebeen long-lived (Grotzinger et al., 2014). Our assumption that thewaters from the later outflow channels would rapidly freeze may imply,therefore, a change in surface conditions between the early and late

Hesperian, and this may not survive further scrutiny.

2. Present inventory of near-surface water

There is abundant evidence of unbound water near the martiansurface, particularly at high latitudes. MOLA topography (Smith et al1999) and MARSIS and SHARAD data (Plaut et al., 2007; Byrne, 2009)indicate that the layered deposits at the two poles combined contain17–22m Global Equivalent Layer (GEL) (Lasue et al., 2013; Carr andHead, 2015). Ground ice has also been detected at latitudes above 60°outside the layered terrains by gamma-ray and neutron spectrometers(Boynton et al., 2002), by ground-penetrating radar (Mouginot et al.,2010), at the Phoenix landing site (Smith et al., 2009), in images offresh impact craters (Byrne et al., 2009) and in a latitude-dependentmantle derived from mobilization of polar ice during cycles of increasedobliquity amplitude (Head et al., 2003) and thought to contain a7–70 cm thick water global equivalent layer (GEL) today(Kreslavsky and Head 2002b). Ice layers extending from near the sur-face to depths of 100m have also been observed in cliffs at latitudes aslow as 55° (Dundas et al., 2018). Mouginot et al., (2010) interpret thesteep decrease in the 3–5MHz reflectivity observed by MARSIS at la-titudes greater than 50° as due to the presence of ground ice, which atthese latitudes is stable a few meters below the surface under presentconditions (e.g. Mellon and Jakosky, 1993). Mouginot et al., (2010)estimate that the radar is effectively probing to depths of 60–80m andthat at least 106 km3, or 7m GEL, of ground ice is present outside thepolar layered deposits to these depths at latitudes greater than 50°.

Near surface ice may be present at even lower latitudes. SHARADsounding has detected interfaces interpreted as the bases of extensiveice sheets in Utopia and Arcadia Planitiae that extend as far south as 38°(Bramson et al., 2017). Ice may also be present in the 30°–50° latitudebands in a meters-thick, dust-ice-rich mantle, interpreted to have beendeposited during recent obliquity excursions (Head et al., 2003), and inlineated valley fill Head et al., 2010), concentric crater fill (Levy et al,2010), pedestal craters (Kadish and Head 2014), debris-covered glaciers(Plaut et al., 2009) and tropical mountain glaciers (Head andWeiss, 2014). Levy et al. (2014a,2014b) estimate that as much as 2.6mGEL may be present near the surface in these bands. In summary, weestimate that 20–30m GEL of water is present in the polar layereddeposits and at depths shallower than 80m in the regions surroundingthe polar layered deposits extending down to 30° latitude. We recognizethat in addition to the 20–30m GEL of near-surface water, abundantunbound water (possibly hundreds of m GEL) may exist as ice cementand secondary ice within the cryosphere at depths greater than 80mand as groundwater below the cryosphere (Clifford, 1993).

3. Ocean volume

Parker et al. (1989; 1993) were the first to suggest that variouslinear features around the northern basin were shorelines of a formerocean. They identified two contacts, an outer contact 1 and an innercontact 2 around the northern plains that they interpreted as shorelines.Subsequent acquisition of MOLA elevation data (Smith et al., 1998)enabled the ocean hypothesis to be tested more rigorously than hadpreviously been possible. Head et al. (1999) used MOLA altimetry datato test whether the Parker et al. shorelines represented equipotentialsurfaces, and found that Contact 1 (−1680m) ranged over 11 km andwas, therefore, not likely to be a shoreline. Contact 2 (−3760m)ranged over 4.7 km, but had extensive stretches that had much lessvariation and could be consistent with an equipotential surface re-presenting a shoreline.

Carr and Head (2002) examined in detail the geologic relations,morphology and elevations of contacts 1 and 2 in light of the MOLAaltimetry data and the more precise information on location of thecontacts given in Clifford and Parker (2001). They found that somecontacts are clearly of volcanic origin and that contact 1, at the

M. Carr, J. Head Icarus 319 (2019) 433–443

434

boundary between the southern uplands and the northern plains, has awide spread of elevations and is unlikely to be a shoreline. In agreementwith previous work, however, they concluded that large expanses ofcontact 2 are close to a constant elevation of around −3760m, andcould potentially be consistent with a shoreline. There was, howeverconsiderable uncertainty in the exact location of the shorelines and itwas not clear how much the spread in elevation was due to the contactsnot being ancient shorelines and how much was due to mis-location ofthe shorelines. The latter problem was particularly acute where thecontact was on or close to a steep slope such as along the plains-uplandboundary. Carr and Head (2002) concluded that better support for theformer presence of water over large parts of the northern plains isprovided by the Vastitas Borealis Formation (VBF), an Upper Hesperianthin veneer of material in the central northern lowlands overlying theLower Hesperian volcanic ridged plains (Tanaka et al., 2005). Theyfound that support for this interpretation came from: (1) the similarityin age between the outflow channels and the VBF, (2) the presence ofthe VBF at the lower ends of the outflow channels, and (3) identificationof numerous features in the outcrop areas of the VBF that are suggestiveof basal melting of an ice sheet (Kargel et al., 1995).

Publication of the global geologic map of Mars (Tanaka et al., 2014)and the availability of global THEMIS daytime images (http://mars.asu.edu/data/thm_dir_100m/) enables a new assessment of the exact lo-cation of the proposed shorelines and, in conjunction with the MOLAdata, their elevation. The availability of the geologic map is particularlyuseful. Identification of geologic units and the depiction of their dis-tribution are based solely on the observed surface characteristics,having been compiled independently of any model for the northernplains or intent to prove or disprove the presence of shorelines. TheTHEMIS daytime images have proven particularly useful in providing anew source of data that enables better portrayal of contrasts in thetextural and thermal properties of the surface, distinction betweenunits, and sharper definition of their contacts, thereby reducing thedependence solely on topography for recognizing possible shorelines.

We find that one the most compelling cases for a shoreline is pro-vided by the outer boundary of the Vastitas Borealis Formation (VBF)around the southern and western rim of the Utopia basin, as portrayedin the Tanaka et al. (2014) global geologic map (Fig. 1) and inIvanov et al. (2017). The VBF occurs widely throughout the northernplains. It is characterized by numerous closely spaced low hills, somepitted, and some arrayed in closely spaced curved ridges to form a‘thumb-print’ texture. Areas of polygonal fractures are common, atdifferent scales and in different locations. Locally in the interior ofUtopia, graben-like linear intersecting troughs form complex networks(Heisinger and Head, 2000). A distinctive variety of surface units andtextures can be distinguished, assessed stratigraphically and datedusing superposed craters (Ivanov et al., 2014, 2015, 2017). Super-imposed impact craters commonly have bright halos, thereby giving theunit a mottled appearance (Tanaka et al., 2005). Many of the textureshave been attributed to the former presence of a frozen ocean (Kargelet al., 1995; Ivanov et al., 2014, 2015, 2017). The unit is labelled asLate Hesperian on the Tanaka et al., (2014) map but the accompanyingcorrelation chart indicates that the unit may extend down into the EarlyHesperian. Kreslavsky and Head (2002a) suggest that the unit is ef-fluent deposited from an ocean (or multiple lakes) that formed con-temporaneously with the outflow channels in the Late Hesperian. Fromthe presence of buried ridges and ‘ghost’ craters, they estimated that theunit is approximately 100m thick. Around the south and west marginsof the Utopia basin the outer boundary of the VBF is easily recognizablein the THEMIS daytime mosaics and readily distinguishable from thesurrounding units by contrasts in albedo and surface texture (Fig. 2).The boundary here has little topographic expression. It closely followsthe −3650m contour for 2800 km around the south and west marginsof the Utopia basin (Fig. 1). It is difficult to imagine any geologic fea-ture or process, other than a shoreline, that would so closely follow acontour over such large distances, and thus we choose to pursue this

interpretation. The precise origin of such a shoreline in southern Utopiais uncertain, however. One possibility is that it marks the furthest reachof the last major flood that entered the northern basin and marks thelimit of deposition of effluent from that flood. At the time of the lastflood most of the water from the previous floods may have alreadyfrozen to form a solidified ice-ocean.

Away from southern Utopia the −3650m contour as a shoreline isnot so compelling (Carr and Head 2002) although its characteristics arenot inconsistent with this interpretation. To the east of Utopia Planitiathe contour crosses Amazonian volcanics that extend from Elysiumdown into the northern basin, so any Hesperian or older shoreline thatmight have been formerly present would now be buried. Similarly, theVBF abuts against younger units in northern Arcadia Planitia and northof Alba Patera. It is not clear how far south the VBF, and hence anypotential shoreline, extended into Arcadia Planitia because of the pre-sence of these younger deposits. The −3650m contour follows thesteep outer boundary of the Olympus Mons aureole deposits. Thisboundary is almost certainly of volcano-tectonic, not marine origin(Carr and Head, 2002), yet would have been a shoreline if present at thesame time as the proposed ocean marked by the −3650m contour. Ifthe outer aureole deposits formed subsequent to the ocean, as is likely(Tanaka et al., 2014), then the deposits would have buried the shorelineof the southern extension of the ocean into Arcadia Planitia. To the westof Utopia Planitia, the contour follows the plains/upland boundary,which is much older than the Hesperian, but which would neverthelesshave been a shoreline in the Hesperian if an ocean was present up to the−3650m elevation.

In Chryse Planitia, into which most of the large outflow channelsempty, the −3650m contour outlines numerous streamlined islandsand other breaks in slope such as the plains-upland boundary (Fig. 3).Ivanov and Head, 2001 used MOLA altimetry data to show that thelargest and most prominent outflow channels (Kasei, Maja, Simud, Tiu,Ares and Mawrth) debouch into Chryse Planitia and disappear into thenorthern lowlands at average elevations that are all within less than∼170m of a mean elevation of −3742m, all within 190m of theelevation of contact 2, over a lateral distance of over 2500 km. Theyproposed that the distinctive change of channel morphology could beconsistent with a rapid loss of stream energy encountered at a base levelsuch as a subaerial/submarine boundary. Contact 2 does not follow theVBF boundary specifically, which is to the north further into the basin.The −3650m contour here is an erosional boundary rather than whatappears to be a depositional boundary in Utopia Planitia. Possible ex-planations in the context of this interpretation are that late stage floodsremoved the VBF in Chryse or that the rapidly moving floods enteringChryse Planitia eroded the ocean floor and did not deposit their sedi-ment load until well below the −3650m contour.

The fate of the water released by the outflow channels and anydeposited in an ocean would have depended on the climatic conditionsunder which it formed. Turbet et al. (2017), using 3-D Global ClimateModel simulations, explored the release of water in individual outflowchannel events, its fate, and its effect on the climate. They modeled theshort and long term climatic impacts of a wide range of outflow channelformation events under cold Mars conditions interpreted to exist in theLate Hesperian, and found that even the most intense (largest, highestflux) of these events cannot trigger long-term greenhouse globalwarming. In a typical event, the outflow channel water reaches thebottom of the basin in a few days, and freezes over within just a fewhundred days. It then sublimates to cold traps at the poles possiblywithin a few hundred thousand years.

An ocean would clearly have to have accumulated by multipleevents as indicated by the number of outflow channels entering thebasin. If, as appears likely and as described above, global climateconditions were similar to present conditions and a thick cryospherewas present (required for the overpressurized aquifer origin for theoutflow channels), then the body of water that formed after each eventwould have frozen within a geologically short period of time (103 to 104

M. Carr, J. Head Icarus 319 (2019) 433–443

435

years depending on the size of the event; Kreslavsky and Head, 2002a;Turbet et al., 2017). The ice could potentially stabilize and remainbetween outflow channel events if a sublimation lag was formed rapidlyafter each event (Kreslavky and Head, 2002a, 2002b; Bramson et al.,2017). Mellon and Jakosky (1995) show, for example, that underconditions expected at an obliquity of 30°, ice is stable at all latitudes atdepths greater than a 10–20 cm. In this specific scenario, repetitivefloods could build an ocean-sized body of ice, layer upon layer as eachwater outflow event built on the frozen, sublimation lag protected,earlier layers. Once the era of flooding ended the ice would slowlysublimate, possibly over hundreds of thousands to millions of years,with the water being both lost to space and redistributed across themartian surface (Fig. 4). According to this scenario, the −3650mshoreline never enclosed an ocean-sized body of liquid water; rather, itmarked the edge of the last outflow channel flood that flowed across the

northern basin, which at the time would be almost completely filledwith ice from the previous floods. For the rest of this paper, we willassume this specific scenario, that in the mid-late Hesperian, there wasa solidified ice-ocean in the northern plains composed of residual icelayers from the individual outflow channel events. The −3650m con-tour, enclosing ∼110m GEL marks the limit of the final flooding event.We will attempt to determine where the 110m GEL of water went. Wesaw above that we can account for approximately 20–30m GEL incurrently existing deposits, so that approximately 80–90m is un-accounted for.

4. Outgassing

In order to know what fraction of the present 30m near-surfaceinventory of water has been inherited from the Hesperian we need to

Fig. 1. Section of the Tanaka et al., (2014) geologic map of the northern hemisphere with the −3650m contour superimposed (sinuous black line). The contourmostly follows the outer boundary of the Vastitas Borealis Formation (VBF). Where the contour is colored red, as northwest of Elysium and in Arcadia Planitia, theboundary is buried by younger units.

M. Carr, J. Head Icarus 319 (2019) 433–443

436

know how much has volcanically outgassed since that time. To de-termine this we used a procedure somewhat similar to that used byGreeley (1987) and Greeley and Schneid (1991). The volume of ex-trusives was derived from the areas of volcanic units, as depicted on theavailable geologic maps, coupled with thicknesses derived from theburial of craters; then an assumption was made about the water contentof the lavas. Our procedure was as follows. The areas of different vol-canic plains units were taken from Table 6 in the brochure accom-panying the global geologic map of Tanaka et al., (2014). The thicknesswas estimated by finding the thickness needed to bury all craters on theunderlying unit within the mapping resolution, which was assumed tobe 105 km2, on the assumption that if there were more craters than thedesignated age, then that 105 km2 would have been mapped as theolder unit. The number of craters on the different aged buried surfaceswas determined from equation 13 and Table III in Ivanov (2001), takinginto account protection from cratering by the overlying unit. The sur-face underlying Amazonian units was assumed to have ND(3.8)–N(D)

(3.0) craters upon it, where ND(3.8) is the number of craters of diameter(D) that has accumulated on a 3.8 b.y. old surface. Similarly, a surfaceunderlying a Hesperian unit was assumed to have ND(4.0) – N(D) (3.8)craters upon it. Rim heights as a function of diameter were taken from

Garvin et al. (2002). The procedure suggests a minimum thickness of800m for Hesperian plains units (very similar to the 800–1000m es-timates for the Hesperian ridge plains in the northern lowlands derivedby different techniques; Head et al., 2002) and 500m for Amazonianplains units. In addition, edifice volumes from Plescia (2004) were as-sumed to have accumulated at a constant rate through the Amazonianand Hesperian. Following this procedure we derive extrusive volumesof 1.21× 107 km3 for the Amazonian and 1.24× 107 km3 for the He-sperian. These numbers are smaller than those of Greeley and Schneid(2.6× 107 and 3.3×107 km3) mainly because of the smaller areasdepicted as volcanic on the Tanaka et al., (2014), map as comparedwith earlier maps. Intrusives (probably mostly in the form of dikes) arenot included in the estimates. For estimating the effect of outgassing onD/H (see below) we assume, consistent with these figures, that vol-canism has declined exponentially over the last 3.5 b.y and that thecumulative volume, V, over the last t billion years (t=0 to 3.8) is givenby V=2.56e7*(1−exp(−0.83*(3.8−t))). The D/H was assumed to be1x SMOW (Standard Mean Ocean Water).

A significant problem in estimating the amount of water outgassedis in knowing the amount of water originally present in the magmas andthe fraction of this amount that was outgassed to the atmosphere.Estimates of the water contents of the Shergottite parent liquids rangefrom 0.0075% to 2.8% (McCubbin et al., 2012; Usui et al., 2012).Following Greeley (1987) and Greeley and Schneid (1991), we assumedthat one wt. percent is outgassed, which gives 5m GEL for the totalamount of water outgassed in the Amazonian and Hesperian. A laterestimate used 0.5 wt% water on the basis of terrestrial data(Craddock and Greeley, 2009). These assumptions, if valid, imply ap-proximately 5m GEL of the present 30m inventory was added byvolcanism after the end of the Noachian..

5. Infiltration into the ground

Clifford (1987, 1993) suggested that, despite the presence of a near-global cryosphere during much of Mars’ history, the planet may havemaintained a slow global hydrological cycle as meltwater at the base ofthe polar layered deposits infiltrated into the groundwater system tocreate a groundwater mound and cause percolation equatorward underthe cryosphere. At low latitudes, the groundwater was episodically re-turned to the surface as a result of impacts, seepages and eruptions ofgroundwater, thereby completing the cycle. The proposed circulation isdriven by the infiltration of meltwater at the poles. Here we examinethe plausibility of melting at the base of the polar deposits in light ofrecent estimates of the global heat flow as a function of time.

Present estimates of global heat flow are significantly less thanwhen the Clifford model was formulated, and these lower heat flowvalues make it unlikely that basal melting could occur at the polesduring the Amazonian and Hesperian. A model byStevenson et al. (1983), for example, predicted a present heat flow of30mWm−2 for present day Mars and approximately 100mW m−2 atthe Amazonian-Hesperian boundary 3 b.y. ago. In contrast, from thegeodynamical response of the lithosphere to the presence of the polardeposits, Phillips et al. (2008) estimate that the present heat flow at thenorth pole is in the 8–17mWm−2 range. In addition,McGovern et al. (2002) and Ruiz et al. (2011) have inferred lithospherethicknesses and heat flows for a number of regions of Mars fromgravity/topography admittance spectra. Estimated heat flow values,while in excess of 40mW/m2 for most Noachian features, are in the10–40mW/m2 range for Amazonian and Hesperian features except forthe large Tharsis volcanoes which have larger local values(Cassanelli et al., 2015). The lower heat flow estimates have also beencompared and reconciled with geochemical models (Ruiz, 2014). Theupper limit for heat flow (mW m2) as a function of time given inRuiz (2014) was approximated by

= + − +H t t t14.4 5.6 1.61 0.82 3

Fig. 2. The outer contact of the Vastitas Borealis Formation at 20N, 120E. Thesharp, easily recognizable contact can be traced at a constant elevation for wellover 2000 km. The VBF is in the upper half of the image. Its southern boundaryis marked by small convex outward lobes, interpreted here as a depositionalboundary formed by the last major flood that entered the basin. The prominentcrater in the lower left is 17 km across (THEMIS day).

Fig. 3. The Chryse basin between 300–360°. E and 10–50°. N. The −3650mcontour around the basin traces numerous sharp breaks in slope between theuplands and the plains and around islands in the flooded areas. The contourhere follows erosional features cut by the large floods entering the basin fromthe south and west. The VBF boundary is further to the northeast (MOLA).

M. Carr, J. Head Icarus 319 (2019) 433–443

437

where t is the time before present in billions of years. According to thisformulation the heat flow is 14.4 mW/m2 at present and 38mW/m2 at3 Ga. For a surface temperature of 150 K, a basal temperature of 273 Kand a thermal conductivity of 2W m−1 K−1 (Clifford, 1987), the polardeposits would have to be over 6 km thick for basal melting 3 b.y. ago,over 9 km thick 2 b.y. ago and over 12 km thick 1 b.y. ago. The presentdeposits are approximately 3 km thick. We conclude, therefore, thatpolar basal melting has been unlikely since the Hesperian ocean waspresent and loss of water though the poles into the groundwater systemcan be ignored. Basal melting may, however, have been significantduring the Noachian when heat flows and possibly surface temperatureswere higher. Following earlier work (Harrison and Grimm, 2004;Russell and Head, 2007), Cassanelli et al. (2015) assessed the possibilitythat snow and ice accumulation on Tharsis could be sufficient to causebasal melting and recharge of the groundwater system. They found,however, that even the heat flow at Tharsis would be insufficient tocause significant basal melting and recharge, except under the verylocalized and volumetrically insignificant volcanic edifice heatpipe/drainpipe situation.

Another possibility is that the Hesperian ocean formed under warmclimatic conditions before a global cryosphere developed and that somesignificant fraction of the ocean evaporated, thereby causing pre-cipitation and infiltration into the groundwater system in the sur-rounding regions. There is, however, little evidence to support such ascenario (Turbet et al., 2017). The outflow channels that are interpretedto have caused the ocean are largely undissected by valley networks as

are many plains that clearly pre-date the outflow channels, such asLunae Planum in the case of Kasei Vallis.

6. Loss of equatorial ice and groundwater to the atmosphere

One of the more puzzling aspects of the evolution of near-surfacewater is the role of the loss of water from the equatorial cryosphere.Under present climatic conditions, ice present at the near-surface at lowlatitudes will tend to sublimate and be transferred to the atmosphere,thereby causing depletion of the near-surface ground ice (Fanale et al.,1986; Clifford and Hillel, 1983). The depth of depletion will increase intime at a rate dependent on the properties of the near-surface materialsand conditions at the surface, with a particular sensitivity to conditionsat different obliquities. This could result in substantial transfer ofequatorial ground ice to the polar regions over geologic time.Grimm and Painter (2009), and Grimm (2017) estimate, for example,that approximately 60m GEL would have been removed from theequatorial regions after the transition to modern cold, dry surfaceconditions in the early Hesperian, most of the losses having taken placein the first billion years. They assume an abrupt transition at 3 b.y. agofrom an earlier climatic regime that allowed retention of equatorial iceto the “modern” regime under which such ice is unstable. The losses areconsistent with the scarcity of near-surface ice at latitudes less than50°–60° as indicated by neutron spectrometer (Feldman et al., 2004)and MARSIS (Mouginot et al., 2010) data. As we saw above, recyclingof water through polar basal melting is unlikely, so the ice removed

Fig. 4. Schematic diagram showing the proposed scenario for formation and elimination of the Hesperian ice ocean. Top: Floods that cut the large outflow channelsenter the northern basin and form temporary lakes that rapidly freeze. Successive floods build a sequence of ice layers up to the−3650 contour which marks the edgeof the last large flood. Bottom: After the flooding era the ice slowly sublimates, some being lost to space and some being redistributed to form the polar layereddeposits and other near-surface ice deposits.

M. Carr, J. Head Icarus 319 (2019) 433–443

438

from the equatorial regions would remain near the surface at high la-titudes or be lost to space. The water lost would have had a D/H of thenear-surface reservoir at the time of transition, which is assumed to bethat of that of the Hesperian aged clays at Gale crater (3x SMOW)(Mahaffy et al., 2015).

If this reasoning is even approximately valid, it would imply that asignificant fraction of the present near-surface inventory of water isderived from equatorial ground ice, not directly from a Hesperian oceanand that we would have to account for the fate not only of the 110mGEL estimated to have been in the ice-ocean, but also the additionalapproximately 65m GEL derived from the equatorial cryosphere.However, initiation of loss of groundwater and ground ice from theequatorial regions probably occurred much earlier than the 3 Ga as-sumed by Grimm (2017). The transition more likely occurred at the endof the Noachian, 3.8 b.y. go, as indicated by the rapid drop off of therate of formation of valley networks at that time. Indeed, loss ofequatorial groundwater may have been a contributing factor in thesubsequent filling of the northern basin both by eruption of ground-water to form the outflow channels and by direct transfer from theequatorial cryosphere to the polar regions through the atmosphere. As aresult of these uncertainties we do not include additions from theequatorial cryosphere in our assessment of the fate of the water thatformerly filled the northern basin. We assume that the equatorial losses,even if they were substantial, occurred earlier than the final phase offilling the northern basin indicated by the Deuteronilus shoreline).

7. Exchange between near-surface ice and the atmosphere andother reservoirs

The evolution of the D/H of the near-surface water reservoir pro-vides clues concerning the fate of water that formed the ice-ocean.Water in the present atmosphere has a value of 6–7x SMOW (Owenet al., 1988; Aoki et al.,2015; Villanueva et al., 2015). The value for themid-Hesperian is approximately 3x SMOW (Mahaffy et al., 2015). Themain cause of the long-term doubling of the D/H of near-surface watersince the mid-Hesperian is losses to space from the upper atmosphere(e.g. Yung et al., 1988). To understand the implications of the D/Hmeasurements, we need to know the size of the water reservoir that isbeing affected by such losses. We saw above that the bulk of the near-surface inventory of water (17–22m GEL) is in the polar deposits.Observational evidence and modeling suggest that the polar layereddeposits are geologically young and dissipate and re-accumulate ongeologically short time scales. Crater counts indicate that the northpolar layered deposits (NPLD) are only ∼105 years old (Herkenhoff andPlaut, 2002). The south polar layered deposits (SPLD) are much older,ranging from 107 to 108 years (Herkenhoff and Plaut, 2002; Koutniket al., 2002), but are still late Amazonian in age. Levrard et al., (2007)modelled the exchange of water between the poles and low latitudesusing a 3-dimensional global climate model. Their results indicate thatthe NPLD completely dissipated during a period of high (> 30°) ob-liquities 5–10Ma ago and that they have been re-accumulating duringthe last 4 Ma as mean obliquity decreased to the present value of 25°.The NPLD are now in an accumulation phase, hence their young age.Superimposed on this long term trend are smaller oscillations. Mostrecently, geologically, polar water is thought to have been transferredfrom the poles to lower latitudes where it formed latitude-dependent,ice-rich mantles (Head et al., 2003); the most recent north polar de-posits are likely to be derived from the sublimation of the southern partof the latitude dependent mantle. On longer time scales, but still in thelate Amazonian, when mean obliquity exceeded 30°, more widespreadequatorward transfer of polar ice is interpreted to have resulted in re-gional ice deposits in the mid-latitudes, including glaciers, lobate debrisaprons (LDA), lineated valley fill (LVF) and concentric crater fill (CCF)(Head et al., 2010; Levy et al., 2014a, 2014b; Kadish and Head, 2014;Madeleine et al., 2009; Fastook and Head, 2014). Tropical mountainglaciers are interpreted to have accumulated when the mean obliquity

was ∼45° (Head and Marchant, 2003; Fastook et al., 2008). The southpolar deposits are more stable than those in the north but still appear tohave exchanged with the atmosphere on geologically short time scales(100Ma). There is little reason to think that the spin-axis/orbital factorsthat govern the transfer of ice in the present epoch are different fromthose operating at other times during the last 3.5 Ga, although theiramplitudes will differ (Laskar et al., 2004). The polar layered depositsare likely to have been dissipating and re-accumulating throughout thisperiod, with their D/H increasing as a consequence of losses from theupper atmosphere. It is less clear how well the 10m GEL of ice depositsperipheral to the pole have exchanged with the atmosphere, but thepresence of mid-latitude ice sheets and tropical mountain glaciersduring this period means that at least some of the ice in the upper partof the regolith is being mobilized as cold traps migrate.Laskar et al. (2004) estimate that there is a 90% probability that ob-liquity reached 47° in the last 20Ma, which would have led to morewidespread loss of polar ice than occurred during the recent episodewhen obliquities were mostly in the 30°–40° range. In summary, it islikely that on a time scale of ∼100Ma, all of the 17–22m GEL polar iceundergoes exchange with the atmosphere as does much of the 10m GELnear the surface in the circum-polar regions.

Based in part on D/H measurements, Kurokawa et al., (2014; 2016)proposed a three-reservoir model for the distribution of water (the at-mosphere, a near-surface reservoir exchangeable with the atmosphereand a deeper ground-ice reservoir). The D/H in the atmosphere andnear-surface reservoir co-evolve, differing only as a result of the frac-tionation factor between ice and vapor. The D/H of the deeper groundice retained the value it had at the time of transition from a hypothe-sized warm and wet/arid (Ramirez and Craddock, 2018), verticallyintegrated hydrologic regime around 3.8–4.0 b.y. ago, to the presenthorizontally stratified regime with a global cryosphere. Kurokawa et al.(2014, 2016) suggest that there could have been slow exchange be-tween the deep ground ice and the near-surface reservoir as a result ofinfiltration, hydrothermal activity, impacts and diffusion. The ex-changeable reservoir is held constant in the model, with losses to spacefrom the exchangeable reservoir being offset by net additions from theunderlying reservoir. From meteorite data (Usui et al., 2012), they as-sume that the deep ground ice has a D/H of 1–2 x SMOW. With theseassumptions they estimate that a few tens of meters GEL of water havebeen lost to space for D/H fractionation factors of 0.016 to 0.11, and afew tens to hundreds of meters GEL for a fractionation factor of 0.33.

We assume here a similar model, in which a near-surface reservoirthat exchanges with the atmosphere overlies a larger ground-ice re-servoir that interacts episodically with the overlying exchangeable re-servoir. As water is lost from the exchangeable reservoir, its D/H in-creases and the losses are replaced by incorporating ice from theunderlying reservoir, which retains mid-Hesperian D/H of 3x SMOW(Mahaffy et al, 2015) . The size of the exchangeable reservoir dependson the time scale of interest. In the present epoch, on a time scale of 103

years, the deeper layers of the NPLD are sealed from the atmosphereand so do not interact with it, but on a time scale of 105 years all theNPLD interact with the atmosphere. On time scales of 108 years, be-cause of excursions to higher obliquities, a larger volume exchangeswith the atmosphere. The SPLD are erased and the areas peripheral tothe poles exchange with the atmosphere (Levrard et al., 2007). We areinterested here in the long term evolution (3.5 b.y.) and the volume thatis exchanging and equilibrating with the atmosphere on time scales of108 years. We will assume that all the NPLD, all the SPLD, and theground-ice peripheral to the poles to a depth of 80m, a total of 20–30mGEL, all interact with the atmosphere on times scales of 108 years, andthat this volume has been approximately constant for the last 3.5 b.y.

8. Losses to space

If the atmosphere has been episodically exchanging with a reservoirof near-surface ground ice, as just discussed, then the high D/H ratio of

M. Carr, J. Head Icarus 319 (2019) 433–443

439

the atmosphere (and by implication the exchangeable ice reservoir)indicates a substantial loss of hydrogen to space. At present day lossrates the equivalent of only 3.6–25m GEL of water would have beenlost over the last 4.2 b.y. (Jakosky et al., 2018). Such a low loss rate isincompatible with the evolution of D/H and the size of exchangeablereservoir as just described. If there had been no loss or gain of waterbetween the near-surface exchangeable reservoir and the deeper sur-face, then the Hesperian and present inventories would be related bythe expression

=−I I R R( / ) f

H P p H(1/1 )

where IH and IP are the Hesperian and present near-surface inventories,RH and RP are the Hesperian and present D/H ratios, and f is the frac-tionation factor (Kurokawa et al., 2016). Even in the extremely unlikelycase of no deuterium lost to space (f=0), a present inventory of20–30m and a doubling of the D/H (see above), would imply that atleast 20–30m has been lost.

As noted by almost all who have addressed the issue of losses tospace from the upper atmosphere, the loss rates depend on the longterm history of the output of the Sun, particularly in the EUV.Comparisons with nearby stars indicate that in the first few hundredmillion years the Sun's EUV output declined from 100s to 10 times thepresent value. Gudel (2007) and Gudel and Kasting (2011) estimatethat by 3.5 b.y. ago the output had declined to 6 times the presentvalue. Their data on the decline can be approximated by the relation-ship log E=0.0647e0.726t where E is the enhancement over present rateand t is the time before the present in billions of years. There are un-certainties concerning the dependence of escape rates for oxygen andhydrogen on the intensity of EUV. We assume here a linear dependence.There is also some uncertainty as to whether the solar ages are exactlyequivalent to the martian crater ages. The solar ages assume the Sun'searly history is typical of that of nearby stars and derivation of craterages depend on crater scaling laws and knowledge on the mix of objectsin the early Solar System. Such discrepancies could be significant inview of the rapid fall off in the EUV between 3 and 4 billion years ago.

Higher obliquities may also have caused higher hydrogen loss ratesin the past. Mellon and Jakosky (1995) suggested that higher ob-liquities would result in higher column abundances of water in theatmosphere as more water is transferred from pole to pole with theseasons. Their estimated increases are substantial, with an over twoorders of magnitude increase from the present 25° to the average ob-liquity of 42° over the last 4 b.y. Such increases in the water content ofthe atmosphere could result in significantly enhanced losses from theupper atmosphere.

Early estimates of the losses of oxygen to space vary widely from theequivalent of 5–15m GEL of water (Lammer et al., 2003) to 50–80mGEL (Kass and Yung, 1995; Luhmann et al., 1992). More recently,Jakosky et al. (2017) estimated oxygen losses from MAVEN measure-ments of 38Ar/ 36Ar in the upper atmosphere. They conclude that asmuch as a half a bar of CO2 could have been removed by sputteringthrough time. They note that if the O lost came from H2O then waterlosses could have been “substantial”. However, even if most of the Olost was from water, it remains unclear when the water was lost,whether before or after formation of the Deuteronilus shoreline. We areinterested here in losses in the last ∼3.5 billion years, that is sinceformation of the Hesperian ice-ocean and much of the “substantial” losscould have taken place before that time.

9. Evolution of D/H

For the present and Hesperian inventories of 30m and 110m GEL,and D/H values of 6 and 3, equation 1 implies a fractionation factor of0.46 which is a higher value than most published values (e.g.Krasnapolsky, 2000; Yung et al., 1988). But as we saw above, the near-surface reservoir has not been isolated from the rest of the planet and

volcanism has added to the surface inventory. Moreover, the loss of80m GEL over the last 3.5 b.y. implies a substantially larger present-day loss rate of hydrogen than the early MAVEN estimates(Jakosky et al., 2018), even taking into account the higher, earlier EUVenhancements.

In order to determine how the mid-Hesperian ice inventory of 110mGEL with a D/H of 3 x SMOW might have evolved to the present near-surface exchangeable inventory of 20–30m GEL with a D/H of 5–7xSMOW, a spread sheet was constructed which started with the presentsituation and stepped back in time in 100m.y. increments. At each stepthe losses to space were calculated starting with different nominal va-lues for present day loss rates and scaling with time to account for thepast higher EUV as described above under Losses to Space. Additions asa result of outgassing were made as also described above, using 1xSMOW for D/H. The exchangeable volume, nominally 30m GEL waskept constant by adding from a deeper non-exchangeable reservoir witha D/H of 3x SMOW in order to replace losses from above. The value of3x SMOW is based on the assumption that the non-exchangeable re-servoir has been protected from interaction with the atmosphere sincethe ice ocean was present. With these constraints the fractionationfactor that results in evolution of D/H of the exchangeable reservoirfrom 3x SMOW at 3.5 b.y. ago to the present value of 5–7 was de-termined by trial and error. Typical results are shown in Fig. 5.

Assuming a present inventory of 30m GEL of which 4m is derivedfrom outgassing, then 84m GEL must be lost to space to eliminate the110m GEL estimated to have been in the Hesperian ocean. Present lossrates of 4–25m GEL in 4.2 b.y. (Jakosky, 2016) fall well short of therequired rates even correcting for the earlier higher EUV fluxes . Apresent day loss of 25 m GEL in 4.2 b.y. implies a loss of 42m GEL in 3.5b.y., or half the required loss, even taking into account the higherearlier EUV fluxes. In addition, doubling of the D/H since the midHesperian requires present day loss rate of at least 21m GEL in 4.2 b.y.,that required in the unlikely event of no losses of deuterium since themid-Hesperian (Fig. 6). In order to lose the 84m GEL to space estimatedto be required to eliminate the 110m GEL ocean, we estimate that apresent day loss rate of 49m GEL in 4.2 b.y. is required, almost doublethe measured rate.

One possibility for the shortfall is that the losses to space have beenunderestimated. This could occur because the measured rates are not

Fig. 5. Typical result of modeling the evolution of D/H as described in the text.Here the exchangeable inventory is 30m GEL and the present day loss rate is30m. GEL in 4 b.y. Curves for two different values of the present polar D/Hshow how the D/H of the exchangeable reservoir may have evolved over time.

M. Carr, J. Head Icarus 319 (2019) 433–443

440

typical of the present epoch, because the effects of the early EUV en-hancements were underestimated or because possible higher loss ratesduring periods of high obliquity were not taken into account. Anotherpossibility is that the present near-surface inventory has been under-estimated. Mouginot et al. (2010) acknowledge that their estimate fromMARSIS of 7m GEL outside the polar caps at depths less than 60–80mis a lower limit. We argued above that significant amounts of watersublimated from the ice-ocean could not have re-entered the ground-water system because of the presence of a thick cryosphere. If true, themissing volume, if not lost to space, must be near the surface. It mustalso be outside the −3650m contour that defines the 110m volume ofthe basin and at latitudes greater than 30° because of stability con-straints. Since the basin occupies most of the northern hemisphere atlatitudes greater than 30°, the missing volume, if still present, is mostlikely at high southern latitudes. Consistent with this conclusion, sevenof the eight cliffs in which layers of ice have been observed are in the55°–60° S. latitude band (Dundas et al., 2018). The southern latitudes inexcess of 30° constitute only one quarter of the planet's surface area soto accommodate the missing 30–55m the southern high latitudes wouldhave to typically be underlain by 120–220m of ice. While this seems animprobably large amount, the ice layers observed byDundas et al. (2018) are tens of meters to over 100m thick so that thehigh southern latitudes could contain a significant fraction of themissing volume.

10. Conclusions

• In the mid to late Hesperian, floodwaters repeatedly flowed into thenorthern basin under climatic conditions that resemble those of thepresent day. We assume that each flood froze in a geologically shortperiod of time to form an ice layer that remained there until the nextflood. Successive ice layers accumulated to form an ocean-sizedbody of ice that filled the basin up to the −3650m contour therebyenclosing 110m GEL of water.

• Subsequent to the filling of the basin the ice slowly sublimated intothe atmosphere to be lost to space or to accumulate in various near-

surface cold traps such as the polar layered deposits.

• Only minor amounts of water were lost from the surface to theglobal groundwater system because of the thick cryosphere and lowheat flow values. The minor amounts lost were in places whereanomalously high heat flow values prevailed as in active volcanicedifice regions.

• 20–30m GEL of water are estimated to be near the surface todayand exchanging with the atmosphere on geologic time scales. Thisincludes the polar layered deposits and deposits elsewhere at depthsless than approximately 80m. A non-exchangeable reservoir of un-known capacity underlies the exchangeable reservoir.

• Approximately 5m of water has been outgassed to the surface sincethe mid Hesperian.

• Present day loss rates to space fall far short of those needed toeliminate the volume of the mid Hesperian ice ocean and thoseneeded to cause a doubling of the D/H of the exchangeable reservoirsince the mid Hesperian. Earlier loss rates must have been higher.Assuming a linear increase in loss rates because of the higher EUVoutput of the Sun, a 30m present inventory, and the upper limit onpresent loss rates from MAVEN, 42m GEL of water remain to beaccounted for. Possibilities include greater dependence of losses ofEUV than assumed and enhanced losses during periods of high ob-liquity. In addition, large volumes of ice may be present near thesurface at high southern latitudes as indicated by recent discoveriesof ice layers in cliffs.

Acknowledgements

The paper was improved substantially following comments fromattendees at the 4th Early Mars Conference, Flagstaff, Oct 1–6, 2017and by thoughtful reviews by R. Irwin and an anonymous reviewer.JWH gratefully acknowledges support for participation in the MarsExpress High-Resolution Stereo Camera (HRSC) team (JPL 1237163).

Supplementary materials

Supplementary material associated with this article can be found, inthe online version, at doi:10.1016/j.icarus.2018.08.021.

References

Andrews-Hanna, J.C., Phillips, R.J., 2007. Hydrological modeling of outflow channels andchaos regions on Mars. J. Geophys. Res. Planet. 112, E08001. https://doi.org/10.1029/2006JE002881.

Aoki, S, Nakagawa, H, Sagawa, H, Gierann, M, Sindoni, G, Aronica, A, Kasaba, A, 2015.Seasonal variation of the HDO/H2O ration in the atmosphere of Mars at the middle ofnorthern spring and beginning of northern summer. Icarus 7–22 260.

Baker, V. R., Strom, R. G., Gulick, V. C., Kargel, J. S., Komatsu, G. and Kale, V. S., 1991.Ancient oceans, ice sheets and the hydrologic cycle on Mars. Nature, 352, 589–59.

Boynton, W.V., et al., 2002. Distribution of hydrogen in the near-surface of Mars: evi-dence for subsurface ice deposits. Science 296, 81–85.

Bramson, A.M., Byrne, S., Bapst, J., 2017. Preservation of mid-latitude ice sheets on Mars.J. Geophys. Res. 122, 2250–2260. https://doi.org/10.1002/2017JE005357.

Byrne, S., et al., 2009. Distribution of mid-latitude ground ice on Mars from impactcraters. Science 325, 1674–1676. https://doi.org/10.1126/science.1175307.

Byrne, S., 2009. The polar deposits of Mars. Ann. Rev. Earth Planet. Sci. 37, 535–560.https://doi.org/10.1146/annurev.earth.031208.100101.

Carr, M.H., 1979. Formation of martian flood features by release of water from confinedaquifers. J. Geophys. Res. 84, 2995–3007.

Carr, M.H., 1996. Water On Mars. Oxford Press, New York.Carr, M.H., 2006. The Surface of Mars. Cambridge University Press, Cambridge.Carr, M.H., Head, J.W., 2002. Oceans on Mars: an assessment of the observational evi-

dence and possible fate. J. Geophys. Res. 108https://doi.org/10.1029/2002JE001963. E5.

Carr, M.H., Head, J.W., 2015. Martian surface/near-surface water inventory: sources,sinks, and changes with time. Geophys. Res. Lett. 42, 726–732.

Cassanelli, J.P., Head, J.W., Fastook, J.L., 2015. Sources of water for the outflow channelson Mars: implications of the Late Noachian "icy highlands" model for melting andgroundwater recharge on the Tharsis rise. Planet. Space Sci. 108, 54–65. https://doi.org/10.1016/j.pss.2015.01.002.

Clifford, S.M., 1987. Polar basal melting on Mars. J. Geophys. Res. 92, 9135–9152.Clifford, S.M., 1993. A model for the hydrologic and climatic behavior of water on Mars. .

Geophys. Res. 98 (E6), 10973–11016. https://doi.org/10.1029/93JE0022.

Figure 6. Estimated losses to space for different values of the present loss rates.The losses are corrected for earlier EUV enhancements. The colored areas showthe measured range of 3.6–25m GEL (Jakosky et al., 2018). The darker areaindicates where loss rates fall short of those required to explain the doubling ofD/H since the mid-Hesperian. The upper curve is for a present loss rate of 49mGEL in 4.2 b.y., that required to eliminate a 110m GEL ocean assuming a 30mGEL present inventory and earlier EUV enhancements.

M. Carr, J. Head Icarus 319 (2019) 433–443

441

Clifford, S.M., Hillel, D., 1983. The stability of ground ice in the equatorial region of Mars.J. Geophys. Res. 88, 2456–2474. https://doi.org/10.1029/LB088iB03p02456.

Clifford, S.M., Parker, T.J., 2001. The evolution of the martian hydrosphere: implicationsfor the fate of a primordial ocean and the current state of the northern plains. Icarus154, 40–79.

Costard, F., Séjourné, A., Kelfoun, K., Clifford, S., Lavigne, F., Di Pietro, I., Bouley, S.,2017. Modeling tsunami propagation and the emplacement of thumbprint terrain inan early Mars ocean. J. Geophys. Res. 122, 633–649.

Craddock, R.A., Greeley, R., 2009. Minimum estimates of the amount and timing of gasesreleased into the martian atmosphere from volcanic eruptions. Icarus 204, 512–526.

Dundas, C.M., et al., 2018. Exposed surface ice sheets in the Martian mid-latitudes.Science 359, 199–201. https://doi.org/10.1126/science.aao1619.

Fanale, F.P., Salvail, J.R., Zent, A.P., Postawko, S.E., 1986. Global distribution and mi-gration of sub-surface ice on Mars. Icarus 67, 1–18.

Fastook, J.L., Head, J.W., 2014. Amazonian mid- to high-latitude glaciation on Mars:supply-limited ice sources, ice accumulation patterns, and concentric crater fill gla-cial flow and ice sequestration. Planet. Space Sci. 91, 60–76. https://doi.org/10.1016/j.pss.2013.12.002.

Fastook, J.L., Head, J.W., Forget, F., 2008. Tropical mountain glaciers on Mars: altitude-dependence of ice accumulation, accumulation conditions, formation times, glacierdynamics, and implications for planetary spin-axis/orbital history. Icarus 198,305–317. https://doi.org/10.1016/j.icarus.2008.08.008.

Feldman, W.C., et al., 2004. The global distribution of near surface hydrogen on Mars. J.Geophys. Res 109 (E9). https://doi.org/10.1029/2003LE02160.

Garvin, J.B., Sakimoto, S.E.H., Frawley, J.J., Schnetzler, S., 2002. Global geometricproperties of martian impact craters. In: Proceedings of the Lunar and PlanetaryScience XXXIII, Abstract 1255.

Godel, M., Kasting, J.F, 2011. Origin and Evolution of Life. In: Gargaud, M., Lopez-Garcia,P., Martin, H. (Eds.), Cambridge, pp. 167–182.

Greeley, R., 1987. Release of juvenile water on Mars: Estimated amounts and timingassociated with volcanism. Science 236, 1653–1654.

Greeley, R., Schneid, B.D., 1991. Magma generation on Mars: amounts, rates and com-parisons with Earth, Moon and Venus. Science 254, 996–998.

Grimm, R.E., Painter, S.L., 2009. On the secular evolution of groundwater on Mars.Geophys. Res. Lett. 36, L24803. https://doi.org/10.1029/2009GL041018.

Grimm, R.E., 2017. On the secular retention of groundwater and ice on Mars. J. Geophys.Res. 122, 94–109. https://doi.org/10.1002/2016JE005132.

Grotzinger, J.P., et al., 2014. A habitable fluvio-lacustrine environment at YallowknifeBay, Gale Crater. Mars Sci. 343, 1242777.

Gudel, M., 2007. The sun in time: activity and environment. Living reviews in SolarPhysics 4 (3).

Harrison, K.P., Grimm, R.E., 2004. Tharsis recharge: a source of groundwater for Martianoutflow channels. Geophys. Res. Lett. 31, L14703. https://doi.org/10.1029/2004GL020502.

Herkenhoff, K.E., Plaut, J.J., 2002. Surface ages and the resurface rates of the polar de-posits on Mars. Icarus 144, 243–253.

Haberle, R.M., Catling, D.C., Carr, M.H., Zahnle, K.J., 2017. The early Mars climatesystem,. In: Haberle, R.M., Clancy, R.T, Forge, F., Smith, M.D., Zurek, R.W. (Eds.),The Atmosphere and Climate of Mars. Cambridge Press, pp. 526–568.

Head, J.W., Hiesinger, H., Ivanov, M.A., Kreslavsky, M.A., Pratt, S., Thomson, B.J., 1999.Possible ancient oceans on Mars: evidence from Mars Orbiter Laser Altimeter data.Science 286, 2134–2137.

Head, J.W., Kreslavsky, M.A., Pratt, S., 2002. Northern lowlands of Mars: evidence forwidespread volcanic flooding and tectonic deformation in the Hesperian period. J.Geophys. Res. 107, 5003. https://doi.org/10.1029/2000JE001445.

Head, J.W., Marchant, 2003. Cold-based mountain glaciers on Mars: Western Arsia Mons.Geology 31 (7), 641–644.

Head, J.W., Weiss, D.K., 2014. Preservation of ancient ice at Pavonis and Arsia Mons:tropical mountain glacier deposits on Mars. Planet. Space Sci. 103, 331–338. https://doi.org/10.1016/j.pss.2014.09.004.

Head, J.W., Mustard, J.F., Kreslavsky, M.A., Milliken, R.E., Marchant, D.R., 2003. Recentice ages on Mars. Nature 426, 797–802.

Head, J.W., Marchant, D.R., Dickson, J.J., Kress, A.M., Baker, D.M., 2010. Northern mid-latitude glaciation in the Late Amazonian period of Mars: criteria for the recognitionof debris-covered glacier and valley glacier land system deposits. Earth Planet. Sci.Lett. 294, 306–320. https://doi.org/10.1016/j.epsl.2009.06.041.

Head, J.W., Forget, F., Turbet, M., Cassanelli, J., Palumbo, A., 2018. Two oceans on Mars?History, problems and prospects. In: Proceedings of the 49th Lunar and PlanetaryScience Conference. 49 Abstract #2194.

Heisinger, H., Head, J.W., 2000. Characteristics and origin of polygonal terrain insouthern Utopia Planitia, Mars: results from Mars Orbiter Laser Altimeter and MarsOrbiter Camera data. J. Geophys. Res 105 (E5), 11999–12022.

Ivanov, A.B., 2001. Mars/Moon cratering rate ratio estimates. In: Kallenbach, R., Geiss, J.,Hartmann (Eds.), Chronology and Evolution of Mars. Kluwer, Dordrecht, pp. 97–104.

Ivanov, M.A., Heisinger, H., Erkeling, H., Reiss, R., 2015. Evidence for large reservoirs ofwater/mud in Utopia and Acidalia Planitiae on Mars. Icarus 248, 383–391.

Ivanov, M.A., Heisinger, H., Erkeling, H., Reiss, R., 2014. Mud volcanism and morphologyof impact craters in Utopia Planitia on Mars: evidence for the ancient ocean. Icarus228, 121–140.

Ivanov, M.A., Erkeling, G., Hiesinger, H., Bernhardt, H., Reiss, D., 2017. Topography ofthe Deuteronilus contact on Mars: evidence for an ancient water/mud ocean andlong-wavelength topography readjustments. Planet. Space Sci. 144, 49–70.

Ivanov, M.A., Head, J.W., 2001. Chryse Planitia, Mars. Topographic configuration, out-flow channel continuity and sequence, and tests for hypothetical ancient bodies ofwater using Mars Orbiter Laser Altimeter (MOLA). J. Geophys. Res. 106, 3275–3295.

Jakosky, B.M., 2016. MAVEN Observations of Mars atmospheric Loss and Implications

For Long Term Evolution. Fall AGU meeting, San Francisco December 2016.Jakosky, B.M., Slipski, M., Benna, M., Mahaffy, P., Yelle, R., Stone, S., and Alsaeed, M.,

2017. Mars’ atmospheric history derived from upper-atmosphere measurements of38Ar/ 36Ar, Science 355, 1408–1410.

Jakosky, B.M., et al., 2018. Loss of the Martian atmosphere to space: present-day lossrates determined from MAVEN observations and integrated loss through time. Icarus315 (In press). www.sciencedirect.com/science/article/pii/S0019103517306917?via%3Dihub.

Kadish, S.H., Head, J.W., 2014. The ages of pedestal craters on Mars: evidence for a late-Amazonian extended period of episodic emplacement of decameters-thick mid-lati-tude ice deposits. Planet. Space Sci. 91, 91–100. https://doi.org/10.1016/j.pss.2013.12.003.

Kargel, J.S., Baker, V.R., Beget, J.E., Lockwood, J.F., Pewe, T.L., Shaw, J.S., Strom, R.G.,1995. Evidence of ancient continental glaciation in the martian northern plains. J.Geophys. Res. 100, 5351–5368.

Kass, D.M., Yung, Y.L., 1995. Loss of atmosphere from Mars due to solar wind inducedsputtering. Science 268, 697–699.

Kasting, J.F., 1991. CO2 condensation and the climate of early Mars. Icarus 94, 1–13.Koutnik, M., Byrne, S., Murray, B., 2002. South polar layered deposits of Mars. J.

Geophys. Res 107, 5100. https://doi.org/10.1029/2001JE001805. E11.Krasnapolsky, V., 2000. NOTE on the deuterium abundance on Mars and some related

problems. Icarus 148, 597–602.Kreslavsky, M.A., Head, J.W, 2000. Kilometer-scale roughness of Mars: results from

MOLA data analysis. J. Geophys. Res 105 (E11), 26695–26711.Kreslavsky, M.A., Head, J.W., 2002a. Fate of outflow channel effluents in the northern

lowlands of Mars: the Vastitas Borealis Formation as a sublimation residue fromfrozen, ponded bodies of water. J. Geophys. Res 4 (1), 4-1–4-25. https://doi.org/10.1029/2001JE001831.

Kreslavsky, M.A., Head, J.W., 2002b. Mars: nature and evolution of young latitude-de-pendent water-ice-rich mantle. Geophys. Res. Lett. 29 (15), 1719. https://doi.org/10.1029/2002GL015392.

Kurokawa, H, Sato, M, Ushioda, M, Matsuyama, T, Morikawa, R, Dohm, J, 2014.Evolution of water reservoirs on Mars: Constraints from hydrogen isotopes in martianmeteorites. Earth Planet Sci. Lett 394, 179–0185.

Kurokawa, H., Usui, T., Sato, M, 2016. Interactive evolution of multiple water-ice re-servoirs on Mars: insights from hydrogen isotope compositions. Geochem. J. 50,67–79.

Laskar, J., Correia, A.C.M., Gastineau, M., Joutel, F., Levrard, B., Robutel, P., 2004. Longterm evolution and chaotic diffusion of the insolation quantities of Mars. Icarus 170,343–364.

Lammer, H., Lichtenegger, H.Kolb, C., Ribas, J., GuinanE., Anart, R., Bauer, S., 2003. Lossof water from Mars: Implications or the oxidation of the soil. Icarus 165, 9–25.

Lasue, J., Mangold, N., Hauber, E., Clifford, S., Feldman, W., Gasnault, O, Grima, C,Sylvestre, M, Mousis, O, 2013. Quantitative assessments of the Martian hydrosphere.Space Sci. Rev. 174, 155–212.

Levrard, B., Forget, F., Montmessin, F, Laskar, J., 2007. Recent formation and evolution ofthe martian polar layered deposits as inferred from a global climate model. J.Geophys. Res 112, 343–364. https://doi.org/10.1029/2006JE002772.

Levy, J., Head, J.W., Marchant, D.R., 2014a. Concentric crater fill in the northern mid-latitudes of Mars: Formation processes and relationships to similar landforms ofglacial origin. Icarus 209, 390–404. https://doi.org/10.1016/j.icarus.2010.03.036.

Levy, J., Fassett, C.I., Head, J.W., Schwartz, C., Watters, J.L., 2014b. Sequestered glacialice contribution to the global martian water budget: geometric constraints on thevolume of remnant, midlatitude debris-covered glaciers. J. Geophys. Res. https://doi.org/10.1002/2014JE004685.

Lucchitta, B.K., Ferguson, H.M., Summers, C., 1986. Sedimentary deposits in the northernlowland plains, Mars. J. Geophys. Res. 91, E166–E174.

Luhmann, J.G., Johnson, R.E., Zhang, M.G., 1992. Evolutionary impact of sputtering ofthe martian atmosphere by O+ pick up ions. Geophys. Res. Lett. 19, 2151–2154.

Mahaffy, P.R., et al., 2015. Mars atmosphere. The imprint of atmospheric evolution in theD/H of Hesperian clay minerals on Mars. Science 347, 412–414. https://doi.org/10.1126/science.1260291.

Madeleine, J.B., Forget, F, Head, J.W, Levrard, B, Montmessin, F, Millour, E, 2009.Amazonian northern mid-latitude glaciation on Mars: a proposed climate scenario.Icarus 203 (2009), 390–405. https://doi.org/10.1016/j.icarus.2009.04.037.

McCubbin, F.M., Hauri, E.H., Elardo, S.M., Kaaded, K.V., Wang, J., Shearer, C.K., 2012.Hydrous melting of the martian mantle produced both depleted and enriched sher-gottites. Geology 40, 683–686. https://doi.org/10.1130/G33242.1.

McGovern, P.J., Solomon, S.C., Smith, D.E., Zuber, M.T., Simons, M., Wieczorek, M.A.,et al., 2002. Localized gravity/topography admittance and correlation spectra onMars: implications for regional and global evolution. J. Geophys. Res 107, E12.https://doi.org/10.1029/2002JE001854.

Mellon, M., Jakosky, B.M., 1995. The distribution and behavior of martian ground iceduring past and present epochs. J. Geophys. Res. 111, 11781–11799.

Mouginot, J., Pommerol, A., Kofman, W., Beck, P., Schmitt, B., Herique, A., Grima, C.,Safaeinili, A., Plaut, J., 2010. The 3–5MHz global reflectivity map of Mars byMARSIS/Mars Express: implications for the current inventory of subsurface H2O.Icarus 210, 612–625. https://doi.org/10.1016/j.icarus 2010.07.003.

Owen, T., Maillard, J.P., de Berg, C, Lutz, B.L., 1988. Deuterium on Mars: the abundanceof HDO and the value of D/H. Science 240, 1767–1770.

Palumbo, A.M., Head, J.W., 2017. Impact cratering as a cause of climate change, surfacealteration, and resurfacing during the early history of Mars. Meteor. Planet. Sci. 1–39.https://doi.org/10.1111/maps.13001.

Parker, T.J., Saunders, R.S., Schneeberger, D.M., 1989. Transitional morphology in thewest Deuteronilus Mensae region of Mars: implications for modification of the low-land/upland boundary. Icarus 82, 111–145.

M. Carr, J. Head Icarus 319 (2019) 433–443

442

Parker, T.J., Gorsline, D.S., Saunders, R.S., Pieri, D., Schneeberger, D.M., 1993. Coastalgeomorphology of the martian northern plains. J. Geophys. Res 98, 11061–11078.

Phillips, R.J., et al., 2008. Mars north polar deposits: stratigraphy, age, and geodynamicalresponse. Science 320 (5880), 1182–1185. https://doi.org/10.1126/science.1157546.

Plaut, J.J., et al., 2007. Subsurface radar sounding of the south polar layered deposits ofMars. Science 316, 92–99. https://doi.org/10.1126/science.1139672.

Plaut, J.J., Safaeinili, A., Holt, J.W., Phillips, R.J., Head, J.W., Seu, R., Putzig, N.E.,Frigeri, A., 2009. Radar evidence for ice in lobate debris aprons in the mid-northernlatitudes of Mars. Geophys. Res. Lett 36, L2203. https://doi.org/10.1029/2008GL036379.

Plescia, J.B., 2004. Morphometric properties of martian volcanoes. J. Geophys. Res. 109,E03003.

Ramirez, R.M., Craddock, R,A, 2018. The geological and climatological case for a warmerand wetter early Mars. Nature Geosci. 11, 230–237. https://doi.org/10.1038/s41561-018-0093-9.

Rodriguez, J.A., Fairen, A.G., Tanaka, K.L., Zarroca, M., Linares, R., Platz, T., Komatsu,G., Miyamoto, H., Kargel, J.S., Yan, J., Gulick, V., Higuchi, K., Baker, V.R., Glines, N.,2016. Tsunami waves extensively resurfaced the shorelines of an early Martianocean. Scientific Reports, 25106.

Ruiz, J., 2014. The early heat loss evolution of Mars and their implicaitons for internaland environmental history. Scientific Reports 4, Article 4338. doi:10.1038/srep04338.

Ruiz, J., McGovern, P, Jimenez-Diaz, A., Lopez, V., Williams, J., Hahn, B., Tejero, R.,2011. The thermal evolution of Mars as constrained by paleo-heat flows. Icarus 215,508–517.

Russell, P.S., Head, J.W., 2007. The Martian hydrologic system: multiple recharge centersat large volcanic provinces and the contribution of snowmelt to outflow channelactivity. Planet. Space Sci. Planet Mars II 55, 315–332. https://doi.org/10.1016/j.pss.2006.03.010.

Segura, T.L., Toon, O.B., Colaprete, A., 2008. Modeling the environmental effects ofmoderate-sized impacts on Mars. J. Geophys. Res 113, 1953–1954. https://doi.org/

10.1029.2008/JE003147.Smith, D.E., Zuber, M.T., Frey, H.V., Garvin, J.B., Head, J.W., Muhleman, D.O., Pettengill,

G.H., Phillips, R.J., Solomon, S.C., Zwally, H.J., Banerdt, W.B., Duxbury, T.C., 1998.Topography of the northern hemisphere of Mars from the Mars Orbiter LaserAltimeter. Science 279, 1686–1692.

Smith, D.E., et al., 1999. The global topography of Mars and implications for surfaceevolution. Science 286, 94–97. https://doi.org/10.1126/science.284.5419.1495.

Smith, P.H., et al., 2009. H2O at the Phoenix landing site. Science 325, 58–61. https://doi.org/10.1126/science.1172339.

Stevenson, D.J., Spohn, T, Schubert, G, 1983. Magnetism and thermal evolution of theterrestrial planets. Icarus 54, 466–489.

Tanaka, K.L., Skinner, J.A., Dohm, J.M., Rossman, P.I., Kolb, E.J., Fortezzo, C.M., Platz,T., Gregory, G.M. and Hare, T.M., 2014. Geologic Map of Mars. U. S. Geol. Survey,Scientific Inv. Map 3292.

Tanaka, K.L., Skinner, J.A., and Hare, T.N., 2005. Geologic map of the northern plains ofMars. U. S. Geol. Survey, Scientific Inv. Map 2888.

Thompson, B.J., et al., 2011. Constraints on the origin and evolution of the central moundin Gale Crater using Mars Reconnaisance Orbiter data. Icarus 214, 413–432.

Turbet, M., Forget, F., Head, J.W., Wordsworth, R., 2017. 3D modelling of the climaticimpact of outflow channel formation events on early Mars. Icarus 288, 10–36.https://doi.org/10.1016/j.icarus.2017.01.024.

Usui, T., Alexander, C., Wang, J., Simon, J.I., Jones, J.H., 2012. Origin of water andmantle-crust interactions on Mars inferred from the hydrogen isotopes and volatileelements of olivine-hosted melt inclusions of primitive shergottites. Earth Planet. Sci.Lett. 357–358, 119–129. https://doi.org/10.1016/j.epsl.2012.09.008.

Villanueva, G.I., Mumma, M.J., Novak, R.E., Kaufl, H.U., hartogh, P., Emcrenaz, T.,Tokunaga, A., Khayat, A., Smith, A.D., 2015. Strong wter isotopic anomalies in themartian atmosphere: probing current and ancient reservoirs. Science 348, 218–221.

Yung, Y.L., Wen, J.S., Pinto, J.P., Pierce, K.K., Allen, M., Paulson, S., 1988. HDO in themartian atmosphere: Implications for the abundance of crustal water. Icarus 76,146–159.

M. Carr, J. Head Icarus 319 (2019) 433–443

443