JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 115, E09006, doi:10.1029/2009JE003546, 2010

Explosive lava‐water interactions in Elysium Planitia, Mars: Geologic and thermodynamic constraints on the formation of the Tartarus Colles cone groups Christopher W. Hamilton,1 Sarah A. Fagents,1 and Lionel Wilson2 Received 16 November 2009; revised 11 May 2010; accepted 3 June 2010; published 16 September 2010.

[1] Volcanic rootless constructs (VRCs) are the products of explosive lava‐water interactions. VRCs are significant because they imply the presence of active lava and an underlying aqueous phase (e.g., groundwater or ice) at the time of their formation. Combined mapping of VRC locations, age‐dating of their host lava surfaces, and thermodynamic modeling of lava‐substrate interactions can therefore constrain where and when water has been present in volcanic regions. This information is valuable for identifying fossil hydrothermal systems and determining relationships between climate, near‐surface water abundance, and the potential development of habitable niches on Mars. We examined the western Tartarus Colles region (25–27°N, 170–171°E) in northeastern Elysium Planitia, Mars, and identified 167 VRC groups with a total area of ∼2000 km2. These VRCs preferentially occur where lava is ∼60 m thick. Crater size‐frequency relationships suggest the VRCs formed during the late to middle Amazonian. Modeling results suggest that at the time of VRC formation, near‐surface substrate was partially desiccated, but that the depth to the midlatitude ice table was ]42 m. This ground ice stability zone is consistent with climate models that predict intermediate obliquity (∼35°) between 75 and 250 Ma, with obliquity excursions descending to ∼25–32°. For lava thicknesses ranging from 30 to 60 m and ground ice fractions ranging from 0.1 to 0.3, an ice volume of ∼4–23 km3 couldhavebeenmelted and/or vaporized by the time the lava solidified, and the associated hydrothermal systems could have retained temperatures >273 K for up to ∼1300 years. Citation: Hamilton, C. W., S. A. Fagents, and L. Wilson (2010), Explosive lava‐water interactions in Elysium Planitia, Mars: Geologic and thermodynamic constraints on the formation of the Tartarus Colles cone groups, J. Geophys. Res., 115, E09006, doi:10.1029/2009JE003546.

1. Introduction Fagents, 2001; Lanagan et al., 2001; Fagents et al., 2002; Head and Wilson, 2002; Fagents and Thordarson, 2007; [2] Volcanic rootless constructs (VRCs) are generated by Hamilton et al., 2010a, 2010b]. Integrated geological map- lava‐water interactions through repeated cycles of frag- ping of VRCs, age‐dating of VRC‐hosting lava surfaces, and mentation and pyroclastic dispersal [Thorarinsson, 1951, thermodynamic modeling of lava‐substrate interactions can 1953]. Terrestrial VRCs typically range from 1 to 35 m in establish where and when near surface groundwater (or ice) height, 2–450 m in basal diameter, and occur in groups with has been present in volcanic regions. This information is total areas of up to ∼150 km2 [Fagents and Thordarson, valuable for constraining models of global climate change on 2007]. VRC analogs on Mars are typically larger than ^ Mars [Head et al., 2003; Laskar et al.,2004;Head et al., those on Earth, with heights of 25 m, basal diameters of 2009] and for understanding the relationships between cli- 30–1000 m [Fagents et al., 2002], and cone group areas of mate, volcanism, volatile stability, and hydrothermal systems up to ∼1300 km2 [Hamilton and Fagents, 2009]. VRCs are [Carr, 1996; Head et al., 2009]. significant because they imply the presence of active lava [3] The Tartarus Colles are located in eastern Elysium flows and an underlying volatile phase (e.g., groundwater or Planitia, ∼750 km northeast of Grjótá Vallis. Lanagan et al. ice) at the time of their formation [Thorarinsson, 1951, 1953; [2001], Greeley and Fagents [2001], Fagents et al. [2002], Frey et al., 1979; Frey and Jarosewich, 1982; Greeley and Bruno et al. [2004, 2006], Baloga et al. [2007], and Hamilton et al. [2010b] used morphological and geospatial evidence to demonstrate that this region contains landforms that are 1Hawaii Institute of Geophysics and Planetology, University of Hawaii, analogous to VRCs in Iceland and, therefore, the Martian Honolulu, Hawaii, USA. cone groups provide evidence of near surface groundwater 2Department of Earth and Planetary Sciences, Lancaster University, Lancaster, UK. (or ice). [4] In this study, we map the western Tartarus Colles Copyright 2010 by the American Geophysical Union. region and present a thermodynamic model of lava‐substrate 0148‐0227/10/2009JE003546

E09006 1of24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006 interactions to: (1) better understand the geological history [8] Fuller and Head [2002] proposed that ascending dikes of Elysium Planitia, (2) constrain the depth to the ground ice in Elysium Planitia may have cracked the cryosphere and table when the VRC groups formed, (3) estimate the volume released groundwater to form floods that were overlain by of ground ice that could have been vaporized and/or melted lava once the dikes reached the surface. In this model, at the time of the cone‐forming eruptions, (4) infer the rootless eruptions initiated when lava flows interacted with obliquity‐driven climate conditions required for ground ice surface water and/or groundwater that had infiltrated into the stability during the emplacement of the Tartarus Colles lava substrate during the antecedent aqueous floods. In other flow, and (5) explore the potential for lava to generate models, dikes were not the cause of aqueous floods and lava hydrothermal systems that could have provided habitable flows, but rather groundwater and magma traveled toward niches for life on Mars. the surface through zones of weakness in the cryosphere that were generated by regional tectonic stresses [Burr et al., 2002; Berman and Hartmann, 2002; Plescia, 2003]. 2. Geologic Setting [9] The youngest of the landforms on Mars to have been [5] The western Tartarus Colles cone groups are located interpreted as VRCs are located in eastern Cerberus Palus, on Mars between approximately 25–27° North and 170– near the mouth of Marte Vallis [Lanagan et al., 2001]. In 171° East (Figure 1). This region is situated between this region, lava surface ages are <1–10 Ma [Hartmann and northeastern Elysium Planitia and southern Arcadia Planitia, Berman, 2000; Berman and Hartmann, 2002]. Strati- and includes the Nepenthes Mensae and Elysium rise units graphically, the Tartarus Colles cone groups are older than [Tanaka et al., 2005] as well as a younger VRC‐hosting the VRCs in Marte Vallis, but if the VRCs in the Tartarus volcanic unit. Colles region volatilized ground ice associated with a [6] The Nepenthes Mensae unit includes knobs and mesas northeast component of an aqueous flood from Grjótá with intervening slope‐ and plains‐forming materials that Vallis, then the Tartarus Colles cone groups may be on the outcrop in the Tartarus Colles and were emplaced during the order of several tens of millions of years old. If rootless early Hesperian to late Noachian [Tanaka et al., 2005]. The eruptions occurred in the Tartarus Colles region ∼10–40 Ma Tartarus Colles are surrounded by the Elysium rise unit ago then ice and fossil hydrothermal systems may be pre- (Figure 1b), which consists of lava flows that were erupted served to this day. These structures could potentially pro- from Elysium Mons, Hecates and Albor Tholi, and local vide information about the development of habitable niches sources during the early Amazonian to late Hesperian on Mars. However, if the volatile source for the rootless [Tanaka et al., 2005]. eruptions in the Tartarus Colles region was part of the global [7] The Cerberus Fossae are located in Elysium Planitia cryosphere, then the timing of these rootless eruptions and are associated with Amazonian age lava flows inter- would not depend on the age of floodwater from Grjótá spersed with aqueous flood deposits, tectonic features (e.g., Vallis. In this case, the Tartarus Colles cone groups could be fissures), and mantling deposits [Keszthelyi et al., 2004]. significantly older than the Cerberus Fossae 2 unit, which Lava flows and aqueous floods associated with the Cerberus would reduce the likelihood that fossil ice and hydrothermal Palus originated from the Cerberus Fossae fissure system structures would be preserved. The geological evidence in [Burr et al., 2002; Fuller and Head, 2002; Plescia, 2003; the Tartarus Colles region is examined to distinguish Head et al., 2003]. Deposits related to Cerberus Fossae between these two hypotheses. divide into three units. The oldest of these units is the early Amazonian age Cerberus Fossae 1 unit [Tanaka et al., 3. Methodology 2005], located in southern Elysium Planitia. This unit is overlain to the north by the late to middle Amazonian age 3.1. Integrating Geologic and Thermodynamic Cerberus Fossae 2 unit. Source regions for the Cerberus Constraints Fossae 2 unit are primarily located in Grjótá Vallis and, [10] Constraints on the formation of the Tartarus Colles in this region, the youngest aqueous flood deposits are cone groups are established using a threefold methodology 10–40 Ma old [Burr et al., 2002]. Mapping of scoured involving: (1) geologic mapping, (2) surface age estimations channels, streamlined forms, and longitudinal lineations based on crater size‐frequency distributions, and (3) ther- suggests that floodwater was released from the northernmost modynamic modeling of lava‐substrate heat transfer. Geo- fissures of Cerberus Fossae to form north and south branches logic mapping is used to identify geologic units, measure (Figure 1b), which flowed southeast toward Cerberus Pla- their areas, and constrain their stratigraphic relationships. nitia [Burr et al., 2002; Plescia, 2003; Burr and Parker, This information is important for calculating VRC group 2006]. Burr and Parker [2006] also noted small stream- areas, estimating lava thicknesses, and identifying the lined forms that suggest a portion of the floodwater may have properties of the substrate beneath the VRC‐hosting Tartarus traveled northeast (Figure 1), toward the Tartarus Colles. The Colles lava flow. Crater size‐frequency distributions are Cerberus Fossae 2 unit is embayed by aqueous flood chan- used to constrain the age of the cone groups and determine nels [Burr et al., 2002; Plescia, 2003; Head et al., 2003] and the timing of their emplacement relative to other geologic overlain by lava flows that have platy‐ridged textures events in Elysium Planitia. Thermodynamic modeling in- resembling those of historic flood basalts in Iceland volves preliminary exploration of our model’s sensitivity to [Keszthelyi et al., 2000, 2004]. The youngest of these variations in initial conditions and validation of its results. We aqueous flood deposits and lava flows belong to the late then apply the model to the Tartarus Colles region using Amazonian age Cerberus Fossae 3 unit [Tanaka et al., 2005; geologic constraints (e.g., lava flow thicknesses and VRC Jaeger et al., 2010]. distributions) to calculate isotherm depths in the substrate,

2of24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006

Figure 1. (a) Context map. (b) Geological map of eastern Elysium Planitia. Black inset depicts the loca- tion of Figures 2 and 3, and white inset shows the extent of the geological map presented in Figure 3. Lobate boundaries, lithostratigraphic units, and emplacement chronology are based on Tanaka et al. [2005]. Lava flow directions (white arrows within the Elysium rise unit) are inferred from the axis of elongation of lobate structures identified by Tanaka et al. [2005]. Aqueous flow directions (white arrows within the Cerberus Fossae 2 unit) are based on Burr and Parker [2006]. The orange arrow shows a possible flow path of floodwater from Grjótá Vallis toward the Tartarus Colles region [Burr and Parker, 2006].

3of24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006 estimate ice table depth, and constrain the maximum lon- biases that affect piecemeal crater counting methods. By gevity of hydrothermal systems. By combining total VRC assuming circular crater geometry, the areas of digitized group area with estimated ground ice fractions and modeled polygons were used to calculate equivalent crater dia- isotherm depths, we quantify the maximum volume of ground meters. The diameters were binned following the method ice that could have been vaporized and/or melted at the time of Hartmann [1999, 2005] and plotted using the Martian of VRC formation. Last, we combine constraints on ground crater‐count isochron diagram of Hartmann [2007], which ice stability with surface age estimates to infer obliquity employs a revised estimate of the Mars:Moon ratio of 2 conditions on Mars during the late Amazonian. bolide impacts per km (Rbolide =2.6).Berman and Hartmann [2002] and Hartmann [2005, 2007] discuss 3.2. Geologic Mapping sources of error associated with deriving absolute ages [11] Photogeological mapping of the Tartarus Colles from crater size‐frequency relationships and estimate that region was performed in ArcGIS using a base map of twelve for Martian surfaces < 2 Ga old, absolute ages could vary Mars Reconnaissance Orbiter (MRO) High Resolution from model ages by a factor of approximately ±2. Even with Imaging Science Experiment (HiRISE) images, eleven these large uncertainties, model ages can be used to assess if MRO Context Camera (CTX) images, nineteen Mars Global the Tartarus Colles lava flow, and/or the smooth terrain it Surveyor (MGS) Mars Orbiter Camera Narrow Angle rests upon, were contemporaneous with the ∼10–40 Ma old (MOC‐NA) images, the MGS Mars Orbiter Laser Altimeter aqueous flood deposits in Grjótá Vallis [Burr et al., 2002], or (MOLA) Mission Experiment Gridded Data Record if they are significantly older and thus do not share a genetic (MEGDR) at a resolution of 128 pixel/degree (0–44°N, 90– association with the Cerberus Fossae 2 unit. 180°E), the MOLA Precision Experiment Data Record (PEDR), thirteen Mars Odyssey Thermal EMission Imaging 3.4. Thermodynamic Modeling System (THEMIS) Visible (V) images, the THEMIS global [14] Subaerial lava flows transfer heat into the substrate daytime infrared (IR) mosaic (0–30°N, 150–180°E), and and surrounding atmosphere. In this study, we quantify heat additional THEMIS daytime and nighttime IR images to fill transfer rates between lava flows and permafrost substrates gaps in the global mosaic and qualitatively examine the containing ground ice. Wilson and Head [2007] modeled thermal inertia properties of surface materials. Appendix A heat transfer during volcano‐ice interactions on Earth for provides details regarding the HiRISE, CTX, MOC‐NA, supraglacial lava flows, subglacial flows, and dyke intru- and THEMIS‐V images used within this study. sions into glaciers. Supraglacial flows may appear to be [12] We mapped the Nepenthes Mensae unit using the analogous to subaerial lava emplaced over an ice‐bearing criteria of Tanaka et al. [2005] and subdivided it into high‐ lithic substrate, but the supraglacial model of Wilson and standing knobs and mesas, and slope‐ and plains‐forming Head [2007] assumes that meltwater can effectively deposits. We mapped the Elysium rise unit, following the escape from underneath the flow, which would buffer the conventions of Tanaka et al. [2005] and added subdivisions basal lava temperature TB at the ice melting point for high and low elevation smooth terrains and anomalous (273.15 K). This assumption cannot be satisfied if the lava flows. We also identified a new volcanic unit, herein substrate contains a significant lithic component because termed the Tartarus Colles volcanic unit, which hosts the the heat conducted into the underlying material will raise Tartarus Colles cone groups. To constrain the interpretation the substrate temperature TS and reduce the downward that the cone groups are composed of volcanic rootless thermal gradient. Since a constant TB cannot be assumed constructs (VRCs), we draw upon previous studies [Lanagan when lava is emplaced over a lithic‐rich substrate, we treat et al., 2001; Fagents et al., 2002; Bruno et al., 2004, 2006; this scenario using the analytical solution of Carslaw and Baloga et al., 2007; Fagents and Thordarson, 2007; Jaeger [1986, article 2.4, equation (14)]. Hamilton et al., 2010b]. These studies used morphological [15] To solve for temperatures in the lava and underlying and geospatial evidence to demonstrate that Tartarus Colles substrate, we consider a solid bounded by the plane x =0 cone groups are analogous to VRC groups in Iceland and (i.e., the top of a lava flow) that extends downward to therefore are the products of explosive lava‐water interac- infinity in the direction of x positive (i.e., through a lava tions. In this study, we mapped the cone group boundaries flow and into the underlying substrate; see Section 4.3.1 for and used down‐flow elongated tephra deposits [Jaeger et al., an example of the model’s geometry). The initial tempera- 2007, 2008; Hamilton et al., 2010a, 2010b] to identify paleo‐ ture distribution is specified by a lava layer of thickness d flow directions in the Tartarus Colles lava flow. and constant initial temperature TM that overlies an infinitely thick substrate at constant initial temperature T . The model ‐ A 3.3. Estimating Model Surface Ages Using Crater Size assumes that heat lost through the edges of the flow is neg- Frequency Distributions ligible, the emplacement time of the flow is short relative to [13] Hartmann [1966, 1999, 2005, 2007] discussed how the total cooling time of the lava, the substrate and atmo- crater size‐frequency relationships can be used to date the sphere are initially at thermal equilibrium at ambient tem- formation age of ideal surfaces, such as broad lava flows perature TA, and that relative to the total cooling history of the formed during a single event. Using this approach, we flow, the temperature at the top of the lava TT rapidly cools to estimated the ages of the Tartarus Colles lava flow and the TA. For the Earth, we consider TA to be 270 K, following the underlying smooth terrain. In ArcGIS, all discernable craters conventions of Wilson and Head [2007], whereas for Mars, were digitized using a base map of CTX images with a we assume TA is equal to its blackbody temperature of 210 K resolution of ∼6 m/pixel. We employed a seamless crater [Barlow, 2008]. At any time, t, after local flow emplacement, digitizing approach because it eliminates the local sampling

4of24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006

[16] Our model differs from equation (3) of Wilson and Head [2007] in that we treat the effects of variable ground ice fractions in the substrate by accounting for the heat absorbed in melting and vaporizing H2O. To accomplish this, is obtained by dividing the ice‐free thermal diffu- −7 2 −1 sivity 0 (assumed to be 7 × 10 m s for basaltic lava flows) by a correction term kAVG. Appendix B shows how we calculated kAVG for a range of ambient temperatures and ground ice conditions. [17] Wilson and Head [2007] modeled d = 1, 3, 10, and 30 m, TM = 1450 K, TA = 270 K, and they considered substrates composed of only pure ice and rock. In contrast, we apply our model over a wider range of initial condi- tions that includes d = 1 to 75 m, TM = 1450 and 1617 K, TA = 180 to 270 K, and FI = 0 to 0.3, where FI describes the initial ground ice fraction in the substrate. Our broader range of initial conditions encompasses approximate permafrost conditions on Mars (∼210 K) and the Earth (∼270 K), which allows us to improve upon the study of Wilson and Head [2007] by exploring how different planetary environments affect lava‐substrate heat transfer.

4. Results

[18] This section presents photogeologic maps, surface age determinations, and lava‐substrate heat transfer models for permafrost environments on Mars and the Earth. The latter part discusses general thermodynamic modeling results and limitations, uncertainties associated with ambient tempera- ture conditions, effects of the latent heat of crystallization, Figure 2. MOLA MEGDR 128 pixel/degree topography of isotherm significance, model validation, and hydrothermal the Tartarus Colles region. Circled letters show locations of system longevity associated with lava‐substrate interactions. interest discussed in Section 4.1. The boundary of the Tar- In Section 5, thicknesses of the Tartarus Colles lava flow are tarus Colles lava flow is delimited by the diagonally hatched estimated from MOLA data and the results are used to region. White outlines depict boundaries of other major constrain lava‐substrate heat transfer models and infer paleo‐ lithostratigraphic units. White dashed line shows the location environmental conditions. of a transitional boundary between a smooth sedimentary 4.1. Geological Mapping mantle and a lava surface in the smooth terrain (Figure 3). Rectangular insets identified by the labels g, h, and j corre- 4.1.1. Nepenthes Mensae and Elysium Rise Units spond to Figures 4a, 4c, and 5a, respectively. The thick black [19] The oldest deposits in our study area belong to the outline shows the extent of the geological map presented in Nepenthes Mensae unit, which is exposed in the Tartarus Figure 3. Colles (e.g., location a in Figure 2; HiRISE PSP_008818_2055, 25 cm/pixel). This unit includes high relief knobs and mesas with intervening slope‐ and plains‐forming materials the temperature, T, at any depth, x, beneath the lava surface (purple in Figure 3). The Tartarus Colles are surrounded by is given by volcanic plains belonging to the Elysium rise unit (shades of red in Figure 3). The volcanic plains mainly consist of lava 1 x x d x þ d T ¼ T þ ðÞT T 2erf pffiffiffiffiffi erf pffiffiffiffiffi pffiffiffiffiffi ð1Þ flows with east to northeast flow directions (Figure 1b). B 2 L B 2 t 2 t 2 t The lava flows exhibit lobate margins and irregular surface depressions (e.g., location c in Figure 2; HiRISE where TB is the temperature at the base of the flow (initially PSP_007526_2075, 25 cm/pixel) that are typical of pahoehoe equal to TM), TL is the lava temperature (initially equal to lava that has inflated to produce lava‐rise plateaus with ∼ ∼ TM, which for basaltic magma is 1450 or 1617 K if an lava‐rise pits [Walker, 1991, 2009]. adjustment is made to account for the latent heat of crys- [20] Along the southeastern margin of our geological map tallization, see Section 4.3.3), is thermal diffusivity, and (location d in Figure 2), there are two surfaces that appear to erf is the error function. The general form of equation (1) is be part of a lava flow that traveled northward and embayed the same as equation (3) of Wilson and Head [2007]; the topographic lows along the eastern side of the Tartarus however, we note that there was a typographical error in Colles. We assign this lava flow to the Elysium rise unit—in their equation (3), whereby the leading factor of 1/2 was accord with Tanaka et al. [2005]—but caution that the omitted. This error was not present in their numerical general northward flow direction of this lava is inconsistent simulations and does not affect the validity of their calcula- with other lava flows in this sector of Elysium Planitia tions or results. (Figure 1b) and may have originated from a different source.

5of24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006

Figure 3. Geological map showing major unit boundaries, volcanic rootless cone (VRC) groups (yellow), and lava flow directions (black arrows) within the Tartarus Colles region. The black inset shows the location of Figure 11a.

[21] Smooth terrains belonging to the Elysium rise unit ejecta with a high thermal inertia (based on qualitative underlie the Tartarus Colles lava flow (Figure 4a) and, examination of THEMIS nighttime IR images, e.g., therefore, it is important to determine the nature of this I14080005, 99.24 m/pixel) that is similar to the nearby material to constrain potential volatile sources for the rough surface material (Figure 4c). This suggests that the rootless eruptions in the Tartarus Colles region. The smooth impact crater excavated through a thin mantle of smooth terrain includes two components: a low‐relief surface at low light‐toned material and into the underlying darker and elevation and an elevated component. The elevated com- rougher material. Craters on the light tone, low‐relief ponent covers a prominent south to north trending topo- surface tend to exhibit greater sedimentary infilling than graphic ridge that is located in the western part of the study the craters on the darker rough surface (Figure 4c). The area (Figures 2 and 3). The elevated component mantles eastern boundary between the smooth terrain and the lava parts of the Nepenthes Mensae and Elysium rise units of the Elysium rise unit is indistinct, but may occur along (location e in Figure 2, and Figure 4a), but the contacts are the western margin of a south‐north trending wrinkle partly obscured by sedimentary deposits (Figure 4b). Along ridge (location g in Figure 2). In this vicinity, the smooth the topographic ridge, the elevated component of the smooth terrain exhibits a dark, rough surface with irregular de- terrain reaches a height of ∼175–225 m above the sur- pressions resembling a lava‐rise plateau (e.g., location c rounding plain. There is a transition within the smooth terrain in Figure 2). Within the smooth terrain (e.g., CTX image (dashed white line in Figure 2) where a light tone and low‐ P03_002344_2067_XN_26N189W, 5.73 m/pixel) there are relief surface grades into a darker and rougher surface bright streaks trending northeast‐southwest. These streaks (location f in Figure 2, and Figure 4c). In the light‐toned appear preferentially on the southwest side of craters and region, a 1.01 km diameter impact crater generated blocky

6of24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006

Figure 4. (a) See location g in Figure 2. Black triangles identify the contact between the elevated component of the smooth terrain and older surfaces (CTX image P07_003900_2050_XI_25N189W, 5.70 m/pixel, sub‐solar azimuth (i.e., illumination direction measured clockwise from 3 o’clock) = 144.6°). The smooth terrain underlies the VRC‐hosting Tartarus Colles lava flow. (b) Magnified view of the contact between the elevated component of the smooth terrain and older surfaces. Recent transverse aeolian dunes mantle parts of the Elysium rise unit and suggest a recent southwestward wind direction. (c) See location h in Figure 2 for spatial context. Transition between smooth and rough surface materials (CTX image P03_002344_2067_XN_26N189W, 5.73 m/pixel, sub‐solar azimuth = 171.8°). Blocky ejecta from the large crater resembles the rough surface material, and impact craters on the smooth surface show greater infilling than craters on the rougher surface. may have been caused by sediment transport under a pre- but we have found no evidence to suggest that these sedi- vailing northeast wind. ments were deposited by aqueous floods. [22] Based on the observations presented in this section, 4.1.2. Tartarus Colles Volcanic Unit we conclude that the smooth terrain consists of one or more [23] The Tartarus Colles volcanic unit includes the lava flows belonging to the Elysium rise unit, but that its youngest volcanic features in the western Tartarus Colles surface has been partially mantled by a westward‐thickening region. In the northern portion of the study area, there is a low veneer of aeolian sediments. We infer that the substrate (∼60–80 m high) shield‐like edifice (dark gray in Figure 3) beneath the Tartarus Colles lava flow consists primarily of with a well‐developed network of distributary channels lava and that there may be an intervening layer of sediments, and/or collapsed lava tubes (e.g., location h in Figure 2;

7of24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006

Figure 5. Path of the Tartarus Colles lava through the Tartarus Colles (see location j in Figure 2 for spatial context). (a) The Tartarus Colles lava flow overtopped a low in Nepenthes Mensae unit and flowed north (mosaic of CTX images: P12_005601_2061_XN_26N189W, 5.69 m/pixel, sub‐solar azimuth = 138.5° and P14_006669_2050_XN_25N188W, 5.71 m/pixel, sub‐solar azimuth = 161.7°). (b) Interac- tions between the Tartarus Colles lava flow and cone groups provide evidence of contemporaneous emplacement (see Section 4.1.3). The white circle highlights a VRC with radial symmetry, and the white ellipse bounds an elongate VRC with lava‐rafted rootless tephra deposits that can be used to infer paleo‐ flow directions (white arrows). (c) Magnified view of the overtopped region of the Nepenthes Mensae unit showing lava highstands and late‐stage lava channel.

HiRISE PSP_008027_2070, 25 cm/pixel). The lava shield given the flow geometry, local slope directions, and orien- has lobate margins and, where the shield surrounds high‐ tation of the flow margins, the mantling lava appears to standing outcrops of the Nepenthes Mensae unit, the lava associated with the same episode of volcanism that con- appears to have developed inflation clefts that are typical structed the rest of the shield. of inflated pahoeoe [Walker, 1991; Self et al., 1996, 1998]. [25] The shield‐like edifice is overlain by the Tartarus [24] The surface of the lava shield generally has a scal- Colles lava flow (light gray in Figure 3) and its associated loped appearance, pervasive dust mantling, and exhibits cone groups (yellow in Figure 3). The Tartarus Colles lava extensive degradation. Topographic lows in the northern flow traveled downslope toward the north, but encountered part of the volcanic shield have been inundated by lava a topographic barrier formed by the Nepenthes Mensae unit (location i in Figure 2 and medium gray in Figure 3), but (e.g., location j in Figure 3). The constricted flow of lava

8of24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006 through this narrow (∼1.2 km‐wide) breach allowed the lava pronounced down‐flow elongation; however, this branch to over‐thicken on the south side of topographic obstacle extends beyond the extent of the continuous CTX image (Figure 5). Gradual drainage of the lava through the coverage and cannot be mapped entirely. constriction caused the lava level to decrease from its [29] Mapping of the Tartarus Colles region reveals that highstand elevation—marked by continuous lava deposits there are 167 cone groups, but additional VRCs are likely to along the margins of the Nepenthes Mensae unit—to form a be located within portions of the northeastern lava branch lower elevation surface that includes a late‐stage channel that extend beyond the geological map. Mapped VRC (Figure 5c). Lava that flowed through the constriction then groups range in area from ∼0.01 to 1258 km2 and cover a spread out within the topographic lows formed between the total area of 2014 km2. knobs and mesas of the Nepenthes Mensae unit (e.g., location k in Figure 2). 4.2. Age Estimates for the Smooth Terrain and [26] Locally, the Tartarus Colles lava flow developed Tartarus Colles Lava Flow platy‐ridged surfaces (e.g., Figures 4a and 4b), which form [30] To constrain the ages of the smooth terrain and the as disrupted sections of lava crust are transported down‐ Tartarus Colles lava flow, we digitized all recognizable flow [Keszthelyi et al., 2000, 2004]. Where cone groups craters on these two surfaces. The resulting geospatial occur in association with platy‐ridged lava surfaces (e.g., database includes 4018 craters on the smooth terrain and Figure 5b), there are chains of raised‐rim deposits, which 8721 craters on the Tartarus Colles lava flow. Figure 6 Jaeger et al. [2007, 2008] and Hamilton et al. [2010a, presents results for craters ≥44.2 m in diameter. The turn- 2010b] interpret to be the products of rootless explosions ing point in the frequency distribution occurs at the next that deposit tephra onto moving lava surfaces. The lava then largest bin size (62.5 m), which implies that smaller crater rafts the tephra down‐flow to form elongated VRCs. The frequencies are incomplete due to data resolution limita- Tartarus Colles cone groups also include radially symmetric tions. The crater‐size frequency relationships also show that conical landforms with central craters (e.g., Figure 5b), craters > 1.0 km are rare within both units (Figure 6). For which Hamilton et al. [2010a, 2010b] interpret to be the craters < 1.0 km in diameter, the smooth terrain exhibits a products of rootless eruptions that deposit tephra onto sta- gentler slope on the crater‐count isochron diagram than the tionary surfaces. If rootless eruption sites are active before Tartarus Colles lava flow (Figure 6). We attribute this and after a lava surface comes to rest, then rootless eruptions gentler slope to infilling and burial of small craters in the can generate transitional landforms (e.g., Figure 5b), which smooth terrain due to aeolian mantling. Consequently, when include a chains of low‐relief tephra deposits (formed while using crater size‐frequency relationships to estimate the the surface of the flow was moving) that can be traced model age of the smooth terrain, we weight the larger craters up‐flow to radially symmetric VRCs (formed when the more heavily. surface of the flow came to rest). Interactions between the [31] Within the southern part of the study area, small Tartarus Colles lava flow and the cone groups provide (<176 m in diameter) craters in the Tartarus Colles lava flow evidence of their concurrent emplacement. tend to concentrate within elongate regions with unusually [27] Mapping of >250 lava‐rafted tephra deposits shows dense crater populations relative to their surroundings. We that the Tartarus Colles lava flow moved through the interpret the dense concentrations of small craters to be of topographic lows between the high‐standing outcrops of the secondary origin and therefore weigh the significance of Nepenthes Mensae unit to form several radiating branches larger craters more heavily when estimating the age of the (Figure 3). The flow directions were dominantly to the Tartarus Colles lava flow. north, but there was also a western branch that curved to the [32] Based on the crater size‐frequency relationships south (Figure 3). Within the southern portion of this lava presented in Figure 6, and the above mentioned caveats, we branch (location l in Figure 2) there is a region with HiRISE estimate that the smooth terrain and the Tartarus Colles lava image coverage (PSP_003900_2055, 25 cm/pixel) that in- flow are late Hesperian and middle Amazonian age, cludes excellent examples of radially symmetric VRCs respectively. Given that uncertainties in model ages derived ranging ∼10–400 m in basal diameter with most cones from Martian crater size‐frequency relationships can vary by being 50–150 m in diameter [Hamilton and Fagents, 2009]. a factor of ±2 [Berman and Hartmann, 2002], we infer that To the northeast there is a type locality for Martian VRCs the smooth terrain could be of early Amazonian to early (location k in Figure 2), which has been studied by Lanagan Hesperian age. However, given that this terrain overlies et al. [2001], Fagents et al. [2002], Bruno et al. [2004, materials from the Elysium rise unit and that Tanaka et al. 2006], Baloga et al. [2007], and Hamilton et al. [2010b]. [2005] estimate the Elysium rise unit to be early Amazo- Hamilton et al. [2010b] interpret the smooth regions sur- nian to late Hesperian, we can exclude the possibility that rounding the conical landforms to be tephra deposits that the smooth terrain was emplaced during the early Hesperian. grade from a sheet‐like distal facies into a platform facies This constraint allows us to refine our age estimate for the with a proximal cone facies that includes welded spatter smooth terrain to the early Amazonian to late Hesperian. deposits. This age estimate is in accord with the age of the Elysium [28] The shield‐like edifice (location h in Figure 2) Rise unit [Tanaka et al., 2005], and is consistent with the bifurcated the northward flowing lava into northwest and geologic interpretation that the smooth terrain belongs to the northeast branches. The northwestern branch directly over- Elysium Rise unit. With a factor of ±2 uncertainty in the age lies the smooth terrain and includes the largest cone group in of the Tartarus Colles lava flow, we estimate that it formed the study area (1258 km2; Figure 3). VRCs in the north- ∼125 Ma ago with a range of uncertainty from 75 to 250 Ma eastern flow branch (location m in Figure 2; CTX ago (i.e., late to middle Amazonian). This crater retention P14_006669_2050_XN_25N188W, 5.71 m/pixel) exhibit age is significantly older than the 10–40 Ma age of the

9of24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006

Figure 6. Crater size‐frequency relationships on the surface of the smooth terrain and Tartarus Colles lava based on the crater‐count isochron diagram of Hartmann [2007]. The upper black line marks saturation equilibrium. Heavy short lines mark the divisions between the Amazonian, Hesperian, and Noachian periods, whereas the nearby solid gray lines mark the era subdivisions. Thin dashed lines represent isochrons and uncertainties in the isochron positions are estimated to within a factor of approximately ±2 [Berman and Hartmann, 2002]. Model ages suggest that the smooth terrain formed during the early Amazonian to late Hesperian, whereas the Tartarus Colles lava flow formed during the late to middle Amazonian. aqueous flood deposits associated with the Cerberus Fossae Earth‐like conditions—estimated lava solidification times 2 unit in Grjótá Valles [Burr et al., 2002]. show good agreement with the emplacement duration of equivalent thickness terrestrial lava flows (see Section 4.3.6). 4.3. Thermodynamic Modeling Nevertheless, the model suffers from the following limita- 4.3.1. General Lava‐Substrate Heat Transfer Results tions. It is a one‐dimensional model that cannot directly and Model Limitations account for the flux of lava though internal pathways (e.g., [33] Figure 7 presents the components of our thermody- lava tubes), which would supply heat after the initial namic model. In this example, we consider a 1 m thick lava emplacement of the flow. By increasing TM to account for the flow with an initial lava emplacement temperature TM = latent heat of crystallization, the heat transfer rate at the base 1617 K (elevated from 1450 K by 167 K to account for the of the flow is artificially increased. It is assumed that the latent heat of crystallization, see Section 4.3.3), ambient emplacement duration is short relative to the total cooling temperature TA = 210 K (to simulate permafrost conditions history of the flow and thus the flow is treated as an on Mars), and an initial ground ice fraction FI of 0.2. Figure 7 instantaneously emplaced unit. The temperature at the top of shows the 1273, 373, and 273 K isotherm depths as a the flow TT is assumed to rapidly decrease to TA, but in a function of time. These isotherms approximately correspond natural system heat flux through the top surface of the lava to the temperature of the basaltic solidus, water boiling point, flow would maintain TT above TA for a significant period of and ice melting point, respectively. In this model, depths are time. Last, the analytical solution of Carslaw and Jaeger calculated from the lava flow surface (x = 0), but when [1986, article 2.4, equation (14)] assumes uniform thermal discussing isotherm depths in the substrate, the lava flow diffusivity throughout the semi‐infinite half‐space of the thickness d is subtracted from x. model and thus does not treat the thermal properties of the [34] The model reproduces the well‐documented square lava flow and substrate separately. In our model, the ice‐free root of time dependence on the growth of chilled lava crusts thermal diffusivity is divided by a correction term kAVG [Hon et al., 1994; Keszthelyi and Denlinger, 1996] and—for to account for the presence of water in the substrate

10 of 24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006

Figure 7. Lava‐substrate interactions shown at one‐tenth of the model resolution. The initial tempera- ture of the substrate is assumed to be in equilibrium with the atmosphere at ambient temperature, TA. The temperature at the top of the lava flow, TT, is set to the boundary condition TT = TA. Results are shown for a 1 m thick lava flow given TA = 210 K, TM = 1617 K, and an initial ground ice fraction of 0.2. The initial lava temperature, TM, is elevated by 167 K to account for the latent heat of crystallization that would be generated as the lava flow cools (see Section 4.3.3 for details).

(Appendix B), but improved models would treat heat transfer conditions on Earth causes the depths of the 373 K isotherm through a system containing multiple layers, each with at TL = 1273 K, 273 K isotherm at TL = 1273 K, maximum unique thermophysical properties. 373 K isotherm, and maximum 272 K isotherm, to be 4.3.2. Uncertainties in Ambient Temperature overestimated by factors of 1.10, 1.48, 1.24, and 3.96, Conditions respectively. This sensitivity analysis demonstrates that [35] The thermal effects of lava‐permafrost interactions although isotherm depths are sensitive to TA, this model is are calculated given a range of flow thicknesses (d = 1, 10, more robust for Mars than for the Earth given equivalent 30, 60, and 75 m), initial lava temperatures (TM = 1450 uncertainties in initial ambient temperature. and 1617 K), initial ground ice fractions (FI = 0, 0.1, [37] Figures 9a and 9b show the linear relationship 0.2, and 0.3), and ambient temperature conditions (TA = between lava flow thickness and isotherm depth for ambient 180 to 270 K). This range of TA includes approximate temperature conditions on Mars (210 K) and the Earth permafrost conditions on Mars (210 K) and the Earth (270 K), respectively, given no ice in the substrate and (270 K). Figures 8a–8d show how isotherm depths vary with TM = 1617 K. These results show that each of the four TA and d, given TM = 1617 K and FI = 0. The results show isotherms achieve greater depths on Earth than on Mars that isotherm depths have a nonlinear dependence on TA by the following factors: 1.21 for the 373 K isotherm at and that uncertainties in TA become increasingly significant TL = 1273 K, 1.73 for the 273 K isotherm at TL = 1273 K, as TA approaches 273 K. 1.48 for the maximum 373 K isotherm, and 6.09 for the [36] For approximate ambient temperature conditions on maximum 273 K isotherm. The 273 K isotherms are influ- Mars and the Earth, we examined the effects of incorrectly enced the most by variations in TA because as the ambient estimating TA by 10, 20, and 30 K. The results are sum- temperature increases, less thermal energy is required to marized in Appendix B and they show that for assumed elevate the substrate to the ice melting point. ambient conditions on Mars, an overestimate (underesti- 4.3.3. Latent Heat of Crystallization mate) of 30 K modifies the corresponding isotherm depths [38] The thermodynamic model does not directly calculate by factors of 1.09 (0.90), 1.14 (0.85), 1.16 (0.83), and 1.33 the latent heat of crystallization generated within cooling (0.65). In contrast, overestimating TA by 30 K for ambient lava flows, but the latent heat of basaltic magma LM (∼2×

11 of 24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006

Figure 8. Thermodynamic modeling results for isotherm depths as a function of lava flow thickness and ambient temperature, TA (see Section 4.3.2 for details). Ground ice fractions affect cooling rates, but not maximum isotherm depths. Simulations use an initial lava temperature, TM, of 1617 K to account for the latent heat of crystallization. The 373 K and 273 K isotherms are assumed to represent the water vaporization and ice melting points, respectively, and the lava flow is assumed to stagnate when its temperature TL reaches ∼1273 K.

5 −1 10 Jkg ) can be estimated by increasing TM by an amount TL = 1273 K; 1.14 ± 0.01 for the maximum 373 K isotherm DT such that LM is approximated by the sensible heat depth; and 1.10 ± 0.01 for the maximum 273 K isotherm cMDT, where cM corresponds to the specific heat of the depth. magma (∼1200 J kg−1 K−1 at magmatic temperatures). For 4.3.4. Interpretations of Isotherm Significance in basaltic magma, this correction amounts to an increase of TM Terms of H2O Phases by ∼167 K [Wilson and Head, 2007]. Results shown in [40] The isotherm depths presented in Figures 8 and 9 are Figures 8 and 9 were generated with TM = 1617 K. How- significant because the maximum depths of the 373 K and ever, to assess how this assumption affects our results we 273 K isotherms describe the maximum depths to which have also performed simulations with TM = 1450 K, where ground ice could be vaporized and melted, respectively. In 1450 K is assumed to be the eruption temperature of the lava. contrast, the depths of the 373 K and the 273 K isotherms at [39] For the same ambient temperature conditions, iso- TL = 1273 K describe the thermal conditions that would therm depth ratios for TM = 1617 K to TM = 1450 K should exist in the substrate when the last parcel of lava cools to a be constant because these isotherm depth relationships are temperature that is too low to generate dynamic rootless erup- linear with respect to changes in lava thickness and ground tions. This limit occurs because cooling‐induced increases in ice fraction (e.g., Figures 9a and 9b), but small uncertainties viscosity would reduce the efficiency of dynamic lava‐water were obtained due to the resolution of the time steps in the mixing and inhibit explosive molten fuel‐coolant interac- simulations. For TA = 210 K (i.e., ambient thermal condi- tions (MFCIs) [Wohletz, 1986, 2002; Zimanowski, 1998]. tions on Mars), the ratio of isotherm depths obtained using Although MFCIs may be suppressed before TL = 1273 K, TM = 1617 K to TM = 1450 K, averaged over all lava flow the time required for the lava to reach its solidus defines the thicknesses and ground ice fractions (±1 standard deviation), maximum duration of substrate heating that could occur are as follows: 1.31 ± 0.02 for the 373 K isotherm depth at prior to complete lava solidification and, after this time, it TL = 1273 K; 1.26 ± 0.02 for the 273 K isotherm depth at would be impossible to generate rootless eruptions through

12 of 24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006

Figure 9. (a and b) Thermodynamic modeling results for isotherm depths, (c and d) lava solidifica- tion times, and (e and f) hydrothermal system longevity, given ambient thermal conditions on Mars (TA = 210 K) and the Earth (TA = 270 K) with ground ice fractions ranging from 0 to 0.3 (isotherm depths presented in Figures 9a and 9b are invariant to initial ground ice fraction; see Section 4.3.2 for details).

13 of 24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006

Figure 10. Application of the thermodynamic model to the Tartarus Colles lava flow. (a) THEMIS‐V image (V26915031, sub‐solar azimuth = 174.4°, 18.58 m/pixel) showing the boundary between the smooth terrain and overlying Tartarus Colles lava flow (white line intersecting the MOLA track at “B“), the boundaries between lava plateaus 1 and 2, and outlines of the volcanic rootless construct (VRC) groups. Filled black circles show the location of MOLA shots along a segment of track #18940 between X and Y. Features labeled A–I are described in Section 5.1.1. (b) Vertical exaggeration × 50. Isotherms beneath the Tartarus Colles lava flow are based on MOLA derived lava flow thicknesses, TA = 210 K, and TM = 1617 K. Note that VRCs concentrate in Plateau 2. explosive MFCIs. Nevertheless, variations in either water thermal conditions because the local mean atmospheric pres- purity or pore pressure could affect interpretations of the sure is approximately 740 to 760 Pa. significance of these isotherms in terms of water phase [41] In the pressure range from 612 to 101,325 Pa (i.e., stability. For instance, the triple point of H2O is at 273.16 K from the H2O triple point to standard atmospheric pressure at 612 Pa, and below this pressure water would only be at sea level on Earth), the ice melting point changes by only present in a gas phase [Carr, 2006]. Fortunately, the ele- 0.01 K (i.e., from 273.16 to 273.15 K). Consequently, for vation of the Tartarus Colles region ranges from approxi- expected atmospheric conditions on Mars and the Earth, it is mately −3000 m to −3300 m (Figure 2) and thus even under inferred that the 273 K isotherm will approximately equal current environmental conditions (i.e., global mean atmo- the ice melting point. However, over the same pressure spheric pressure of 560 Pa and mean atmospheric scale range, the H2O boiling point changes from 273.16 to 373.15 K. height of 10.8 km), liquid water could exist under the right The pressure‐dependence of the H2O boiling point therefore

14 of 24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006 affects interpretations of the 373 K isotherm in terms of lava‐substrate interactions during the emplacement of the water phase stability, but it does not affect estimates of the Tartarus Colles lava flow. Second, associations between combined total volume of ground ice that was melted or VRC groups, lava thicknesses, and modeled isotherms are vaporized. used to infer ice table depths. Third, mapped VRC group 4.3.5. Validation of the Model Results areas are combined with lava thickness information and [42] For TM = 1617 K, lava flows ranging from 1 to 75 m modeled isotherm depths to estimate the maximum volume in thickness, and ambient temperature conditions on Mars of melted and/or vaporized ground ice that could have and the Earth, respectively, Figures 9c and 9d present the contributed to cone‐forming eruptions. Fourth, constraints time required for lava core temperatures TL to reach 1273 K. on the thickness of the Tartarus Colles lava flow are used to This cooling time is approximates the lava emplacement estimate hydrothermal system longevity. duration, but to assess whether the modeling results are reasonable, we compared our modeled lava solidification 5.1. Lava Thickness Constraints times—given FI = 0 and TA = 270 K—to the emplacement 5.1.1. Lava Thicknesses Variations in the Tartarus durations of equivalent thickness pahoehoe sheet lobes on Colles Lava Flow Earth [Thordarson and Self, 1998]. For the fifteen flows in [45] Figure 3 depicts the location of THEMIS‐V image Table 4 of Thordarson and Self [1998]—excluding Alviðr- V26915031 (18.58 m/pixel), which is shown in Figures 10a uhamrar because the total lobe thickness was not measured— and 11. Overlapping HiRISE and CTX image coverage the ratio of calculated stagnation times to their estimated enables MOLA PEDR elevations to be precisely correlated emplacement times is 0.91 ± 0.33. Thus, our results tend to with geologic features. The region shown in Figures 10a and slightly underestimate the time for the lava emplacement, 11a includes 24 MOLA tracks. All of these tracks show but we infer that this is partly due to the fact that the model similar associations between elevations and morphological does not account for the supply of heat that would be units, but we focus on a segment of track #18940 because it introduced into the system by lava fluxing through internal crosses all of the major surface types in the region. The track pathways (e.g., lava tubes) and partly due to excess heat loss extends north‐northwest to south‐southeast from the smooth through the top surface of the lava flows because of the TT = terrain (X in Figure 10), across a portion of the Tartarus TA boundary condition. Colles lava flow, and into part of the largest cone group (Y in 4.3.6. Longevity of Hydrothermal Systems Generated Figure 10). Lava thicknesses were obtained by subtracting a by Lava‐Substrate Interactions local datum—extrapolated from the surface elevation of the [43] The vast majority of all life on Earth draws its energy smooth terrain—from MOLA PEDRs of the lava surface. either directly or indirectly from photosynthesis, but there [46] Estimating lava thicknesses as the difference between are also organisms that derive their energy from chemical lava surface elevations and a local datum is reasonable for disequilibria generated in hydrothermal systems, and it has this portion of the Tartarus Colles lava flow because the been proposed that chemosynthesis may be the most plau- underlying smooth terrain has a surface slope of only 0.03 to sible source of energy for putative life on Mars [Shock, 0.09°. These slopes correspond to an elevation difference of 1997; Jakosky and Shock, 1998; Varnes et al., 2003]. To less than ∼1.5 m/km and are comparable to Tharis‐related assess the potential for establishing near surface hydro- volcanic units, which slope gently at 0.08° for a distance of thermal systems on Mars and the Earth, given a range of 2500 km [Aharonson et al., 1998]. lava flow thickness from 1 to 75 m and ground ice fractions [47] Along the MOLA transect, the smooth terrain (fea- from 0.1 to 0.3, we calculated the maximum duration for ture A in Figure 10) is overlain by lava with a sharply which temperatures in the substrate could remain above defined flow front (feature B in Figure 10) that consists of 273 K. The results are shown in Figures 9e and 9f for 100–250 m wide lobes. The lobes coalesce into a lava‐rise ambient conditions on Mars and the Earth with TM = 1617 K plateau [Walker, 1991] with a mean thickness of 30.3 ± 6.9 m and FI = 0.2, but given that our model does not take into (at 1 standard deviation, se = ± 0.8 m). The surface of the account convective circulation, our calculations will over- lava‐rise plateau is interrupted by an impact crater (feature C estimate the longevity of hydrothermal systems. For ambient in Figure 10) with a rim‐to‐rim diameter of 1.53 km. The temperature conditions in permafrost environments on Mars thickness of the lava‐rise plateau continues to increase (TA = 210 K) and an initial ground ice fraction of 0.2, liquid gradually to the south‐southwest until it intersects a higher water could only remain for <0.3 years beneath a 1 m thick relief plateau (feature D in Figure 10). 3 flow, and <1.6 × 10 years beneath a 75 m thick flow. Given [48] This second lava plateau has indistinct margins, the same ground ice fraction and ambient temperature which vary from 1 to 3 km in width, and its surface is pitted conditions on Earth (TA = 270 K), liquid water would be and blocky. Local flow directions, defined by the orientation stable from <6.5 years for a 1 m thick flow to <3.8 × 104 of elongated VRCs (Figure 3), show that the supply of lava years for a 75 m thick flow. into this region was from the east‐southeast. Along the northern slope of the second lava plateau there is a VRC 5. Lava‐Substrate Interactions in the Tartarus group located between features E and F in Figure 10. Between features F and G (Figure 10), the second lava Colles Region plateau attains a thickness of 60.2 ± 12 m (at 1 standard [44] The thermodynamic model (equation (1)) is applied deviation, se = ± 2.6 m) and extends to the east to form a to the Tartarus Colles region in four ways. First, lava flow broad continuous platform that underlies the largest VRC thicknesses are estimated using geologic evidence and group in the study area. Between features H and I (Figure 10) MOLA PEDR and MEGDR data to constrain models of there is a topographic saddle between two branches of the

15 of 24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006

Figure 11. Lava thickness distributions. (a) THEMIS‐V image (V26915031) with overlaid segment of MOLA PEDR track #18940 (X to Y). Estimated lava thicknesses obtained by subtracting a local datum from the 128 pixel/degree MOLA MEGDR. Cone groups (diagonally hatched regions) are overlaid above the lava thickness raster and show that VRCs preferentially concentrate within the thicker parts of the flow. (b) Lava thickness distribution for the Tartarus Colles flow and within the cone group regions. Thicknesses of the Tartarus Colles lava flow are bimodal (with peaks at 25–30 m and 55–60 m), whereas lava thicknesses associated with VRCs are unimodal (with a peak at 60–65 m).

∼60 m thick lava plateau. At the southern end of the or slabs [Keszthelyi et al., 2004; Guilbaud et al., 2005]. The MOLA profile (south of feature I in Figure 10), the lava second plateau is inferred to be bounded by rubbly pahoe- flow appears to reach a thickness of ∼100 m; however, this hoe that inflated behind a peripheral blockage—perhaps region coincides with a VRC group and thus the thickness caused by cooling‐induced stagnation along the thinner flow of the Tartarus Colles volcanic unit at this locality would margins of the first plateau. VRCs concentrate in the second include a thickness contribution from both the Tartarus lava plateau, where lava is > 30–60 m thick, but tend not to Colles lava flow and overlying rootless tephra deposits. form in the first lava plateau, which is < 30 thick (Figure 10). 5.1.2. Interpretation of Lava Thickness Variations [50] To demonstrate that the MOLA PEDR profile shown [49] We attribute the bimodal lava thickness variations in in Figure 10 is representative of regional lava thickness the study area to two plateaus of inflation within the Tar- distributions, we have applied a similar method of sub- tarus Colles lava flow, rather than the superposition of two tracting a local datum from the MOLA MEGDR 128 pixel/ separate flows. The sharp lobate flow margins, thickness degree data for a region between 26 and 27°N and 170.2– profile, and smooth surface morphology of the lower lava 171.2°E. This region was chosen because CTX and plateau are characteristic of inflated pahoehoe [Thordarson THEMIS‐V imagery shows that the Tartarus Colles lava and Self, 1998]. In contrast, the second plateau has indis- flow directly overlies the low elevation component of the tinct margins and a rough, pitted, and blocky surface. These smooth terrain, which permits us to extrapolate the ∼0.05° characteristics are consistent with rubbly pahoehoe, which slope of the smooth terrain beneath the Tartarus Colles forms due to pressurization of a molten lava core—either by volcanic unit to estimate the overlying lava thickness. In pulsed lava supply or interactions with obstacles to flow— Figure 11a, estimated lava thicknesses and cone group followed by disruption of the overlying lava into blocks and/ boundaries show that the VRCs preferentially concentrate in

16 of 24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006 the thicker parts of the lava. Figure 11b shows the lava and 273 K isotherms would have reached depths of 26.4 m thicknesses frequency distributions. (+2.4 m and −2.6 m given TA ± 30 K) and 37.2 m (+5.2 m [51] The lava thickness distribution is bimodal, with peaks and −5.6 m given TA ± 30 K), respectively. at 25–30 m and 55–60 m, whereas the lava thicknesses [55] These isotherm depths represent maximum estimates within the cone groups are unimodal, with a mean of 59.5 m because the rootless eruptions must have initiated before the (s = ±8.3 m, se < ±0.1 m). The bimodal lava thickness lava reached its solidus. For instance, within the Cerberus peaks correspond to the two lava plateaus identified in Fossae 3 unit on Mars, Jaeger et al. [2010] measured the Figure 10 and thus we infer that results based on MOLA thickness of tilted lava slabs in the Athabasca Valles flood PEDR track #18940 are representative of general lava lava that were upended in response to loading by rootless thickness distributions within the Tartarus Colles lava flow. tephra. Jaeger et al. [2010] applied a lava cooling model— 5.1.3. Relevance of Lava Thicknesses to adapted from Keszthelyi and Denlinger [1996]—to estimate Thermodynamic Modeling that the ∼1 m thick tilted slabs would have formed ∼11 days [52] If the bimodal variations in the surface elevation of after the lava flow surface came to rest, but it is uncertain the Tartarus Colles lava flow correspond to two plateaus of how long the flow surface was in motion before it stagnated inflation within a single flow, then the cooling lava layer of and began to form a competent surface crust. In the Tartarus thickness d (equation (1)) corresponds to the combined Colles region, lava surfaces have not been tilted by the thickness of both lava plateaus. Figure 10b shows the iso- loading of rootless tephra, which suggests that lava crusts therm depths obtained by applying our thermal model to the were able to support the rootless tephra because they had a estimated lava thicknesses along the MOLA profile in longer cooling duration and thus were thicker and had Figure 10. The simulation assumes TM = 1617 K and TA = greater mechanical strength. This implies that prior to the 210 K. In this case, isotherms are diachronous and ground initiation of rootless eruptions, the Tartarus Colles lava flow ice fractions are irrelevant because they affect modeled generally had a longer cooling duration than the Amazonian cooling rates, but not the maximum isotherm depths. The Athabasca Valles lava flow. However, the presence of lava‐ isotherms in Figure 10b exhibit bimodal depth distributions rafted rootless tephra deposits in the study area suggest that that reflect the bimodal variation in lava thicknesses, but the some rootless eruptions initiated soon after the local isotherm depths are expected to be overestimated beneath emplacement of the Tartarus Colles lava flow because the the VRC groups because rootless tephra would contribute to flow surface was still undergoing a period of autobrecciation the apparent lava thickness. Given that the ∼30 and ∼60 m due to flow motion [Jaeger et al., 2007, 2010]. Thus, while thick lava plateaus identified in Figure 10 occur in a region the time required for lava core temperatures to reach their that contains the majority of the Tartarus Colles VRCs solidus represents a maximum upper bound on the time (Figure 3), we assume that the combined thickness of these available for heating the substrate prior to the initiation of two lava plateaus is representative of the total lava thick- rootless eruptions, it is inferred that rootless eruptions ini- nesses in other VRC‐hosting sections of the Tartarus Colles tiated considerably before TL = 1273 K and that results lava flow. overestimate isotherm depths in the substrate at the time of cone formation. 5.2. Ground Ice Depths at the Time of Cone Formation 5.3. Calculations of Mobilized Ground Ice Volumes [53] Isotherms at TL = ∼1273 K describe the thermal conditions that exist when the core temperature of a lava [56] Given that VRCs are indicative of an underlying flow reaches its approximate solidus temperature. After this groundwater (or ice) reservoir at the time of their formation, time, rootless eruptions cannot be generated by explosive the mapped areal extent of rootless cone groups (i.e., MFCIs because the lava will be in a solid state and can no 2014 km2) can be combined with the thermodynamic model longer participate in the MFCI processes of hydrodynamic to estimate the maximum volume of ground ice that could premixing, triggering, fine‐fragmentation, and explosive have been vaporized and/or melted at the time of the cone water vaporization [Wohletz, 1986, 2002; Zimanowski, forming eruptions. These calculations only consider ground 1998]. Where VRCs are present, the ground ice table must ice volumes directly below the Tartarus Colles cone groups therefore have been located above the depth of the 273 K and, therefore, lateral migration of volatiles in the substrate isotherm when TL = 1273 K. is ignored. It is assumed that TM = 1617 K and TA = 210 K, but [54] On Mars, a 30 m thick lava flow would reach 1273 K to account for uncertainties in estimates of FI and d, the model after ∼3 to 9 years, for a range of ground ice fractions from 0 is applied over a range of conditions that include FI = 0.1 to to 0.3. When TL = 1273 K, the depth of the 373 K isotherm 0.3 and d = 30 to 60 m. The H2O melting point is generally would be 12.6 m (with an uncertainty of +1.1 m and −1.3 m insensitive to pressure and, therefore, the 273 K isotherm given potential errors in the estimate of TA by up to ±30 K), will approximately correspond to the ice melting tempera- whereas the 273 K isotherm would reach a depth of 18.3 m ture in the Tartarus Colles region. Thus, 3.7 to 22.5 km3 of (+2.6 m and −2.7 m given TA ± 30 K). The general absence ground ice could have been melted by the time the Tartarus of VRCs on sections of the Tartarus Colles lava flow that are Colles lava flow solidified. <30 m thick suggests that the uppermost substrate (i.e., to [57] In contrast to the ice melting point, the boiling point depths of up to ∼16–21 m) contained insufficient ground ice of H2O is highly sensitive to pressure and therefore given to generate rootless eruptions. In contrast, rootless cones the same temperature distribution, the partitioning of melted tend to occur where lava is >30–60 m thick. For ground ice ground ice into liquid or gas phases will depend on substrate fractions ranging from 0 to 0.3, a 60 m thick flow would pore pressure. If the pore pressure were ∼105 Pa (i.e., cool to its solidus within ∼11 to 34 years and the 373 K standard atmospheric pressure at sea level on Earth) then the

17 of 24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006 boiling point of water would be 373 K. Given that Martian ∼220 ± 5 K at the equator, and the global frost point tem- surface gravity g is 3.71 m/s2, the pressure at the base of a perature of water is ∼200 K [Squyres et al., 1992]. Conse- basaltic lava flow, with a density r = 2500 kg/m3, would quently, ground ice in equilibrium with the water vapor reach 101,325 Pa when the lava reached a thickness h = content in the atmosphere is restricted to latitudes that are 9.93 m (lithostatic pressure = rgh). For pore pressure con- poleward of approximately ±40° [Farmer and Doms, 1979; ditions of ∼105 Pa, the volume of ground ice that could have Squyres et al., 1992]. At higher obliquities, the polar regions been vaporized at TL = 1273 K would therefore have been receive more insolation, which causes the mean annual polar 2.5–16.0 km3, and the total volume of ground ice in a liquid temperatures to rise and the annual equatorial temperatures state would have been 1.2–6.5 km3. If pore pressure were to decrease. Increased polar temperatures cause the polar ice <105 Pa, there would be more vapor and less liquid water, caps to melt/sublime, thereby increasing the amount of whereas if the pore pressure was >105 Pa—due to lithostatic water in the atmosphere, increasing the frost point temper- loading by the overlying substrate and Tartarus Colles lava ature, and widening the near equatorial latitude belt over flow—then there would have been a larger proportion of which ground ice is stable. liquid water relative to vapor. Given that rootless eruptions [61] The obliquity of Mars is currently 25.19°, but over in the Tartarus Colles region tend to occur where lava is 30– the past 5 Ma, Mars’ obliquity has oscillated between 15 and 60 m thick, we expect lithostatic pressure at the base of the 40°, with a mean of 27 ± 6° [Laskar et al., 2004]. There was flow to be 2.78–5.56 × 105 Pa, which corresponds to a water a significant increase in obliquity 5 Ma ago and, from 5 to boiling point of ∼403–427 K. Given that explosive MFCIs 20 Ma ago, the mean obliquity was 36 ± 5°, with a range of require coolant (i.e., water) to initially be in the liquid phase, 24–47° [Laskar et al., 2004]. VRCs near the mouth of Marte lithostatic pressure controls on the water boiling point may Vallis are estimated to be <10 Ma old [Lanagan et al., 2001] therefore be an important factor in determining where and therefore these low‐latitude rootless eruptions likely rootless eruptions can occur on Mars. formed when obliquity was higher than it is at present. If the Marte Vallis and Tartarus Colles cone groups formed under 5.4. Hydrothermal Systems in the Tartarus Colles similar environmental conditions, then it is probable that the Region Tartarus Colles lava flow was emplaced under intermediate [58] Within the Tartarus Colles region, lava thicknesses (∼35°) obliquity conditions as well. Prior to 10–20 Ma ago, typically range from 30 to 60 m, which would generate precise obliquity solutions cannot be derived due to chaotic hydrothermal systems with substrate temperatures > 273 K variations [Laskar et al., 2004], but probabilistic obliquity for less than 182 to 1333 years given ground ice fractions of scenarios can be constrained using geological evidence that 0.1 to 0.3. Link et al. [2005] demonstrated that chemical documents changes in ground ice stability through time weathering of minerals in aqueous environments on Mars [Head et al., 2009] (see Section 6.2). can provide enough geochemical energy to support the [62] At and below the present obliquity of 25.2°, ground construction of potential Martian organisms at temperatures ice is unstable at all latitudes less than 40–50°, whereas at as low as 273 K. However, given that even very thick lava obliquities ≥32.3° ground ice would be stable at all latitudes, flows on Mars are unlikely to generate long‐lived hydro- with the ice table being <5–10 cm below the surface at all thermal systems, the rates of these geochemical weathering latitudes [Mellon and Jakosky, 1995; Carr, 2006]. The processes could pose a significant barrier to the habitability presence of VRCs within the Tartarus Colles region implies of ephemeral hydrothermal systems associated with lava that at the time of the rootless eruptions there was ice in the flows in Martian environments. substrate and, therefore, that obliquity was >25.2°. Never- theless, the general absence of VRCs in sections of the flow ∼ – 6. Discussion that are <30 m thick suggests that the uppermost 16 21 m of substrate was at least partially desiccated, which implies that 6.1. Potential Sources of Ground Ice the paleo‐obliquity was <32.3°. The heterogeneous VRC [59] The critical lava thickness required to generate root- distributions in the Tartarus Colles region may have resulted less eruptions in the Tartarus Colles lava flow (∼30–60 m) from variations in lava flow thicknesses. Rootless eruptions is greater than in other VRC‐hosting regions of Mars, such may have preferentially initiated where the flow was >30 m as Marte Vallis, which have estimated lava thicknesses of thick because only these portions of the flow would have ∼10 m [Lanagan et al., 2001]. Relative to Marte Vallis, the contained sufficient thermal energy to enable the 273 K greater lava thickness required to generate VRCs in the isotherm to reach the depth of the global ice table. If so, the Tartarus Colles region may be due to a lower ground ice ice table must have been located at a depth <42.4 m based abundance per unit volume of substrate, which would on the maximum depth of the 273 K isotherm beneath a 60 m require thicker flows to supply sufficient thermal energy to thick lava flow when TL = 1273 K. mobilize an equivalent volume of ice. A lower abundance of [63] Alternatively, the Tartarus Colles flow could have ground ice in the near surface substrate of the Tartarus volatilized water associated with a localized ground ice Colles region could have been due to lower substrate porosity body that was perched above the global cryosphere. In this and/or a greater depth to a buried ground ice reservoir. scenario, the distribution of rootless cone groups in the [60] The depth to the top of the global cryosphere is Tartarus Colles region could have been controlled by strongly dependant on latitude and time‐dependent varia- groundwater that infiltrated into the substrate during one or tions in obliquity [Squyres et al., 1992; Mellon and Jakosky, more aqueous floods [Lanagan et al., 2001; Fuller and 1995]. The current mean annual surface temperature on Head, 2002]. This process would have preferentially gen- Mars varies with latitude from ∼160 ± 5 K at the poles to erated rootless eruptions over former topographic depres-

18 of 24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006 sions because these sites would have been the most favor- trending valley that runs between Grjótá Vallis and the able locations for concentrating the flow of floodwater and Tartarus Colles. Although aqueous flood deposits from lava. However, based on observed crater size‐frequency Grjótá Vallis may obscure the source region of the Tartarus relationships, we estimate that the smooth terrain was em- Colles lava flow, age estimates suggest that this lava flow placed during the early Amazonian to late Hesperian. The preceded the most recent episode of aqueous flooding in smooth terrain directly underlies the Tartarus Colles lava Grjótá Vallis. The cone‐forming rootless eruptions must flow and, therefore, age constraints suggest that the timing therefore have volatilized an older water source, such as the of a potential aqueous flood in the Tartarus Colles region is global cyrosphere, and not groundwater that infiltrated into incompatible with the late to middle Amazonian age the substrate during any known aqueous flood. If the Tar- floodwaters associated with the Cerberus Fossae 2 unit tarus Colles lava flow is associated with an event that [Tanaka et al., 2005]. Consequently, if one or more aqueous occurred before the emplacement of the Cerberus Fossae 2 floods did pass through the Tartarus Colles region, then unit, then it may have resulted from an precursory phase of they were not affiliated with any of the previous described volcano‐tectonic activity that was associated with the units associated with the fossae in Grjótá Vallis [Burr et migration of active volcanism away from central Elysium al., 2002; Plescia, 2003; Burr and Parker, 2006]. Conse- Planitia and toward Cerberus Fossae during the early to quently, the hypothesis that the volatile source for rootless middle Amazonian—perhaps in accord with changes in the eruptions in the Tartarus Colles region was the global regional state of stress that led to the formation of the fossae cryosphere is favored over that of a perched ground ice in Grjótá Vallis during the late Amazonian. reservoir that formed by infiltration of floodwater into the substrate. This suggests that rootless eruptions on Mars are not limited to regions that have been inundated by aqueous 7. Conclusions floods. [66] The Tartarus Colles region includes outcrops of the 6.2. Paleoclimate Implications early Hesperian to late Noachian Nepenthes Mensae unit, [64] The preferential occurrence of VRCs on ∼30–60 m which is exposed in the Tartarus Colles, and volcanic thick portions of the Tartarus Colles lava flow suggests that landforms that have been attributed to the early Amazonian obliquity was between 25 and 32° at the time of the cone to late Hesperian Elysium rise unit [Tanaka et al., 2005]. formation. If this obliquity range is compared to the fifteen Within the Elysium rise unit we have identified a series of ‐ probabilistic obliquity scenarios of Laskar et al. [2004] for lava rise plateaus, which include the sediment mantled lava the period over which the Tartarus Colles lava flow could flow(s) that comprise the smooth terrain. The Tartarus have been emplaced (i.e., 75–250 Ma ago based on an Colles lava flow overlies the smooth terrain and was em- estimated age of ∼125 Ma and a factor of ±2 uncertainty), placed as a northward traveling flow that overtopped a geological evidence excludes models that predict extended topographic low in the Nepenthes Mensae unit and inun- periods of unusually high or low obliquity during the past dated the valleys within the Tartarus Colles. Photogeologi- 250 Ma. This conclusion is in accord with that of Head et al. cal mapping shows that the Tartarus Colles lava flow hosts a minimum of 167 VRC groups with a total area of at least [2009] who suggested that the intermediate obliquity sce- 2 nario 301003BIN_A.N001 [Laskar et al., 2004] provides 2014 km . These VRC groups include radially symmetric the best agreement with the geological evidence for mid- conical landforms with central pits, rootless tephra com- ‐ latitude glaciations during the Amazonian Period in other plexes, and VRCs with down flow elongated tephra de- regions of Mars. This scenario shows that on average obliq- posits. Interactions between the Tartarus Colles lava flow uity was >32.3° during the past 250 Ma, but that ∼125 Ma ago and cone groups demonstrate that the lava and VRCs were ‐ there were frequent obliquity excursions descending into the emplaced concurrently and crater size frequency relation- range of ∼25–32° [Laskar et al., 2004]. Head et al.’s [2009] ships suggest that the Tartarus Colles lava flow is late to preferred scenario agrees with obliquity estimates in this middle Amazonian age. This age estimate is younger than study and supports our conclusion that ∼125 Ma ago, a that of the Elysium rise unit, but older than the most recent shallow (<42.4 m depth) midlatitude (25–27°N) ice table aqueous flows in Grjótá Vallis [Burr et al., 2002]. Conse- was present below a thin (<16–21 m) zone of near surface quently, the Tartarus Colles lava flow may have been desiccation. associated with changes in the regional state of stress that led to a migration of volcanic activity away from Elysium 6.3. Regional Tectonic Implications Planitia and toward the Cerberus Fossae. [65] The crater retention age for the Tartarus Colles lava [67] The Tartarus Colles lava flow exhibits two lava ∼ ∼ flow (i.e., late to middle Amazonian) is younger than the plateaus, which are 30 m and 60 m thick. VRCs con- – early Amazonian to late Hesperian lava flows that have been centrate where lava is >30 60 m, but tend not to form where attributed to the Elysium rise unit [Tanaka et al., 2005]. The the lava is <30 m thick. This suggests that the underlying Tartarus Colles lava flowed north, whereas in this sector of ground ice was located at <42 m depth, but that VRCs did Elysium Planitia other lava flow directions generally range not form where the lava was <30 m thick because the – from east to northeast (Figure 1b). This anomalous flow uppermost 16 21 m of the substrate was partially desic- direction could suggest that the Tartarus Colles lava flow cated. This zone of ground ice stability suggests the Tartarus originates from a local source to the south, or that an east- Colles cone groups formed during a period of extended ∼ ward moving lava flow was deflected to the north after intermediate obliquity ( 35°), with excursions descending to ∼ – being captured by the south‐southwest to north‐northeast 25 32°.

19 of 24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006

Table A1. Images Used for Mapping the Western Tartarus Colles Region Image Center Center Pixel Identification Latitude Longitude Width Original Image Instrument Number (°N) (°E) (m) Description VRCs HiRISE ESP_012431_2050 24.9 169.7 0.50 Small cones near the Tartarus Montes No HiRISE ESP_017244_2050 24.9 169.4 0.50 Cluster of secondary impact craters No HiRISE PSP_002344_2065 26.1 170.5 0.25 Cratered knobs in Amazonis region Yes HiRISE PSP_003900_2055 25.4 170.4 0.25 Ring/cone structures in the Yes Tartarus Montes/Phlegra Dorsa region HiRISE PSP_008818_2055 25.2 169.9 0.25 Cones among the Tartarus Montes No (Stereo pair: 1 of 2) HiRISE PSP_009029_2055 25.2 169.9 0.50 Cones among the Tartarus Montes No (Stereo pair: 2 of 2) HiRISE PSP_008528_2060 25.8 170.6 0.25 Field of cones in the Tartarus Montes Yes (Stereo pair: 1 of 2) HiRISE PSP_009675_2060 25.8 170.6 0.25 Field of cones in the Tartarus Montes Yes (Stereo pair: 2 of 2) HiRISE PSP_007671_2065 26.0 170.9 0.25 Cones among the Tartarus Montes Yes (Stereo pair: 1 of 2) HiRISE PSP_007882_2065 26.0 170.9 0.25 Cones among the Tartarus Montes Yes (Stereo pair: 2 of 2) HiRISE PSP_008027_2070 26.5 171.2 0.25 Distributary channel system in the Phlegra Dorsa Yes HiRISE PSP_007526_2075 27.3 171.3 0.25 Boundary between lava flow and No potential flood deposits CTX P01_001579_2072_XN_27N188W 27.3 171.1 5.71 Textured flow material in Phlegra region Yes CTX P03_002344_2067_XN_26N189W 26.8 170.5 5.73 Knobs and flow material in Phlegra region Yes CTX P04_002555_2055_XN_25N189W 25.5 170.1 5.71 Knobs and flow material in Phlegra region Yes CTX P07_003900_2050_XI_25N189W 25.0 170.5 5.70 Tartarus Colles region Yes CTX P07_003900_2050_XI_25N189W 25.0 170.5 5.70 Tartarus Colles region Yes CTX P12_005601_2061_XN_26N189W 26.2 170.8 5.69 Knobs and flow material in Phlegra region Yes CTX P14_006669_2050_XN_25N188W 25.0 171.5 5.71 Knobs and flow material in Phlegra region Yes CTX P17_007526_2073_XN_27N188W 27.4 171.3 5.78 Tartarus Colles region Yes CTX P18_008027_2071_XN_27N188W 27.2 171.1 5.77 Ride‐along with HiRISE Yes CTX P20_008818_2054_XN_25N190W 25.5 169.9 5.78 Ride‐along with HiRISE Yes CTX P20_009029_2054_XN_25N190W 25.5 169.9 5.93 Ride‐along with HiRISE Yes MOC‐NA M03/06223 25.7 169.9 3.01 Sample No MOC‐NA M08/01962 25.7 170.6 4.52 Traverse among knobs southeast of Yes Phlegra Montes region MOC‐NA M09/02788 25.5 171.1 3.02 Sample knobby terrain of Tartarus Montes Yes MOC‐NA M09/02788 25.8 170.8 3.02 Sample knobby terrain of Tartarus Montes Yes MOC‐NA M22/02497 26.2 171.8 4.54 Massifs of Tartarus Montes Yes MOC‐NA R09/03347 25.9 170.0 6.35 Phlegra Montes knobs No MOC‐NA R12/03400 25.9 171.5 3.17 Massifs in Tartarus Montes No MOC‐NA R15/01132 26.2 169.8 4.76 Knobs in the Phlegra region No MOC‐NA S12/02407 25.8 170.6 1.57 Mounds with summit pits in Yes M08–01962 (cPROTO) MOC‐NA S13/00179 25.7 170.7 3.18 Sample Yes MOC‐NA S14/00989 25.2 169.9 4.78 Landforms in the Phlegra Dorsa region No MOC‐NA S15/00197 24.5 170.6 4.79 Landforms in the Phlegra Dorsa region No MOC‐NA S15/02698 25.5 171.5 4.76 Landforms in the Phlegra Dorsa region Yes MOC‐NA S17/00366 25.0 171.1 4.79 Landforms in the Phlegra Dorsa region Yes MOC‐NA S18/00428 25.0 171.2 3.16 Knobs and flow features in Phlegra region No MOC‐NA S18/00977 24.6 170.3 4.74 Landforms in the Phlegra Dorsa region No MOC‐NA S19/00402 26.6 171.1 4.80 Textured flow features in Phlegra region Yes MOC‐NA S20/00396 26.8 171.2 4.77 Textured flow features in Phlegra region No MOC‐NA S21/00745 25.9 170.5 4.79 Knobs and flow features in Phlegra region Yes THEMIS‐V V26029013 26.0 170.9 18.56 No description given Yes THEMIS‐V V26603028 26.1 171.6 18.56 No description given Yes THEMIS‐V V26890021 26.2 171.7 18.58 No description given Yes THEMIS‐V V26915031 26.6 170.6 18.57 No description given Yes THEMIS‐V V27177027 27.7 172.1 18.44 No description given No THEMIS‐V V27177028 25.1 171.7 18.36 No description given No THEMIS‐V V27776022 25.2 170.8 18.54 No description given Yes THEMIS‐V V28063013 25.2 170.9 18.36 No description given Yes THEMIS‐V V28088015 27.5 170.0 18.43 No description given No THEMIS‐V V28662043 25.5 169.9 18.55 No description given Yes THEMIS‐V V29211014 25.2 171.0 18.51 No description given Yes THEMIS‐V V29236016 26.3 169.9 18.56 No description given No THEMIS‐V V31232008 27.0 170.9 37.31 No description given Yes

20 of 24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006

Table B1. Thermal Diffusivity Corrections kT for TA = 210 K Table B3. Average Thermal Diffusivity Corrections kAVG Ground Ice Fraction Ground Ice Fraction

kT 0.001 0.01 0.05 0.10 0.15 0.20 0.25 0.30 TA (K) 0.001 0.01 0.05 0.10 0.15 0.20 0.25 0.30 k273 1.00 1.01 1.06 1.11 1.17 1.22 1.28 1.33 180 1.01 1.06 1.32 1.64 1.96 2.28 2.60 2.92 k373 1.00 1.05 1.25 1.50 1.74 1.99 2.24 2.49 190 1.01 1.07 1.33 1.66 1.98 2.31 2.64 2.97 k450 1.01 1.12 1.60 2.21 3.81 3.42 4.02 4.63 200 1.01 1.07 1.34 1.67 2.01 2.35 2.68 3.02 k500 1.01 1.10 1.51 2.02 2.53 3.04 3.55 4.06 210 1.01 1.07 1.35 1.69 2.04 2.38 2.73 3.08 k600 1.01 1.08 1.40 1.79 2.19 2.58 2.98 3.37 220 1.01 1.07 1.36 1.71 2.07 2.43 2.78 3.14 k700 1.01 1.07 1.33 1.65 1.98 2.31 2.64 3.96 230 1.01 1.07 1.37 1.73 2.10 2.47 2.84 3.20 k800 1.01 1.06 1.28 1.56 1.85 2.13 2.41 2.69 240 1.01 1.08 1.38 1.76 2.14 2.52 2.90 3.28 kAVG 1.01 1.07 1.35 1.69 2.38 2.38 2.73 3.08 250 1.01 1.08 1.39 1.79 2.18 2.57 2.96 3.36 260 1.01 1.08 1.41 1.81 2.22 2.63 3.04 3.44 270 1.01 1.08 1.42 1.85 2.27 2.69 3.12 3.54

[68] For measured cone group areas, a range of estimated lava thicknesses from 30 to 60 m, and ground ice fractions ranging from 0.1 to 0.3, it is inferred that ∼4–23 km3 of we have identified volcanic rootless constructs (VRCs) in ground ice could have been vaporized and/or melted by the each image. time the Tartarus Colles lava flow solidified and that beneath 60 m thick lava flows, the temperature of the re- Appendix B: Effects of H2O and Ambient sulting hydrothermal system could have remained >273 K ‐ for a maximum of ∼1300 years. The longevity of hydro- Temperature on Lava Substrate Heat Transfer thermal systems generated by lava flows on Mars are more [70] To adjust thermal diffusivity to account for the heat than an order of magnitude smaller than those calculated for absorbed in melting and vaporizing H2O in a dominantly equivalent thickness lava flows on Earth, which places a lithic substrate, we divide the ice‐free thermal diffusivity 0 significant constraint on the time available for achieving the by a correction term kAVG. For ambient thermal conditions chemical disequilibria required for biological chemosyn- (TA) on Mars and the Earth, kAVG is obtained by averaging thesis through aqueous weathering of minerals in ephemeral kT, solved at temperatures (T) of 273, 373, 450, 500, 600, hydrothermal systems on Mars. 700, and 800 K, where kT is the ratio of the H2O‐adjusted thermal diffusivity to 0. For Mars, it is assumed that regions containing near surface ground ice will have T = 210 K, Appendix A: Image Database Used for Mapping A whereas on Earth, TA = 270 K. At 273 K, kT is calculated as the Western Tartarus Colles Region follows, [69] Table A1 provides an inventory of the satellite ima- ges used to map the western Tartarus Colles region. These FRcRðÞþT273 TA FW cI ðÞT273 TA images were acquired by cameras onboard the Mars k273 ¼ ; ðB1Þ Reconnaissance Orbiter (MRO), Mars Global Surveyor cRðÞT273 TA (MGS), and Mars Odyssey spacecrafts. The instruments include the MRO High Resolution Imaging Science Experiment (HiRISE) camera, MRO Context (CTX) cam- FW cI k273 ¼ FR þ ; ðB2Þ era, MGS Narrow Angle (NA) Mars Orbiter Camera cR (MOC), and Mars Odyssey Thermal EMission Imaging System (THEMIS) Visible (V) camera. Table 1 also presents the scaled pixel widths and metadata image where FR is the fraction of rock in the substrate, cR is the −1 −1 descriptions. These original descriptions demonstrate the specific heat of the rock (900 J kg K ), FW is the fraction widely varying nomenclature used to describe the image of H2O in the substrate, and cI is the specific heat of the − − content. In the final column, we summarize whether or not water ice (1900 J kg 1 K 1). Equation (B2) describes condi- tions associated with the heating of the substrate up to a

Table B2. Thermal Diffusivity Corrections kT for TA = 270 K

Ground Ice Fraction Table B4. Ratio of Isotherm Depths for Assumed TA (210 K) to

kT 0.001 0.01 0.05 0.10 0.15 0.20 0.25 0.30 True TA for Mars True T /Assumed T – True T k273 1.00 1.01 1.06 1.11 1.17 1.22 1.28 1.33 A A A k373 1.01 1.07 1.36 1.72 2.08 2.44 2.80 3.16 200 K/ 190 K/ 180 K/ 220 K/ 230 K/ 240 K/ k450 1.02 1.16 1.79 2.57 3.36 4.15 4.94 5.72 Isotherm 10 K 20 K 30 K −10 K −20 K −30 K k500 1.01 1.13 1.63 2.26 2.89 3.52 4.15 4.78 k600 1.01 1.09 1.46 1.91 2.38 2.83 3.29 3.74 373 K (TL = 1273 K) 1.02 1.05 1.09 0.97 0.93 0.90 k700 1.01 1.07 1.37 1.73 2.10 2.46 2.83 3.19 273 K (TL = 1273 K) 1.03 1.10 1.14 0.96 0.90 0.85 k800 1.01 1.06 1.31 1.62 1.92 2.23 2.54 2.85 373 K (Maximum depth) 1.05 1.12 1.16 0.94 0.88 0.83 kAVG 1.01 1.08 1.42 1.85 2.27 2.69 3.12 3.54 272 K (Maximum depth) 1.08 1.21 1.33 0.89 0.77 0.65

21 of 24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006

Table B5. Ratio of Isotherm Depths for Assumed TA (270 K) to uncertainties in ambient temperature conditions, predicted True TA for the Earth isotherm depths for a range of assumed TA versus true TA were examined given T = 1617 K and F = 0 (i.e., k = 1 for all True T /Assumed T – True T M I AVG A A A temperatures). Tables B4 and B5 show the results for Mars Isotherm 260 K/10 K 250 K/20 K 240 K/30 K and the Earth, respectively. For Mars, overestimates and 373 K (TL = 1273 K) 1.03 1.06 1.10 underestimates of TA by 10, 20, and 30 K were considered. 273 K (T = 1273 K) 1.23 1.36 1.48 L For the Earth, the effects of overestimating TA by 10, 20, and 373 K (Maximum depth) 1.08 1.16 1.24 30 K were examined. Underestimates of T are not consid- 272 K (Maximum depth) 2.26 3.14 3.96 A ered because permafrost on Earth would not form above ∼270 K. Simulated isotherm depth ratios for different TA values are invariant to lava flow thickness because of the linear rela- temperature of 273 K, which is assumed to be just below tionship between lava flow thickness and isotherm depth. the ice melting point. At 373 K, kT becomes, F c ðÞþT T F ½c ðÞþT T L þ c ðÞT T Notation k ¼ R R 373 A W I 273 A I L 373 273 ; 373 c ðÞT T −1 −1 R 373 A cM, cR, cI, cL, cV Specific heat of magma (1200 J kg K ðB3Þ for basaltic magma), rock (900 J kg−1 K−1 for solidified basalt), water ice (1900 J kg−1 K−1), liquid water (4186 J kg−1 K−1), F ½c ðÞþT T L þ c ðÞT T −1 −1 −1 k ¼ F þ W I 273 A I L 373 273 ; ðB4Þ and water vapor (2000 J kg K )(Jkg 373 R − cRðÞT373 TA K 1).

5 −1 d Lava flow thickness (m). where LI is the latent heat of water ice (3.34 × 10 Jkg ) and −1 −1 FR, FI, FW Fraction of rock, ice, and water in the sub- cL is the specific heat of liquid water (4186 J kg K ). strate, respectively (dimensionless). Equation (B4) accounts for heat absorbed in melting H O ice 2 Corrected thermal diffusivity ( = 0/kAVG) and warming the substrate to 373 K, where 373 K is assumed (m2 s−1). to be just below the water vaporization temperature. Over the −7 2 −1 0 Ice‐free thermal diffusivity (7 × 10 m s range of pressure conditions that are relevant to lava‐substrate for basaltic lava) (m2 s−1). heat transfer processes, ice melting temperatures are generally kAVG Average thermal diffusivity correction term insensitive to pressure, but the vaporization temperature can (dimensionless). change dramatically. In our calculation of k , we assume 373 kT Thermal diffusivity correction factor at that significant volumes of water vapor will not be generated temperature T (K) (dimensionless). on Earth and Mars below 373 K. This assumption is generally 5 LM, LI, LV Latent heat of fusion of magma (∼2×10 J valid beneath lava flows on Mars because the pressure at the −1 5 kg ), latent heat of fusion of water (3.34 × base of a 10 m thick lava flow would be ∼1.11 × 10 Pa (lava 105 Jkg−1), and latent heat of water vapor- thickness × lava density × gravitational acceleration on Mars ization (1.80 × 105 Jkg−1)(Jkg−1). = 10 m × 3000 kg/m3 × 3.7 m/s2), which is greater than 5 t Time since local lava flow emplacement standard sea level pressure on Earth (1.01 × 10 Pa), where (s). water vaporizes at 373.15 K. However, more sophisticated TA Ambient temperature of the air and sub- applications of our model should take into account pressure strate (K). controls on phase changes when calculating corrections to TB Temperature of lava flow base (initial TB = ice‐free thermal diffusivity. For T above 373 K, k becomes, T TM) (K).

FRcRðÞþT TA FW ½cI ðÞþT273 TA LI þ cLðÞþT373 T273 LV þ cV ðÞT T373 kT ¼ ; ðB5Þ cRðÞT TA

FW ½cI ðÞþT273 TA LI þ cLðÞþT373 T273 LV þ cV ðÞT T373 kT ¼ FR þ ; ðB6Þ cRðÞT TA

TL Temperature of the lava (initial TL = TM) where LV is the latent heat of vaporization of water (1.80 × (K). 6 −1 10 Jkg ) and cV is the specific heat of water vapor (2000 J T Initial (i.e., magmatic) temperature of the − − M kg 1 K 1). Tables B1 and B2 show adjusted thermal diffu- lava, which we assume is 1450 K for sivity results for conditions on Mars (TA = 210 K) and basaltic lava flows +167 K to account Earth (TA = 270 K), respectively. Table B3 presents kAVG for the latent heat of crystallization (K). corrections that would apply where TA varies between 180 K TS Temperature of the substrate (initial TS = and 270 K. To assess the sensitivity of the model to TA) (K).

22 of 24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006

TT Temperature at the top surface of the lava Hamilton, C. W., T. Thordarson, and S. A. Fagents (2010a), Explosive flow (initial T = T ) (K). lava‐water interactions I: Architecture and emplacement chronology of T A volcanic rootless cone groups in the 1783–1784 Laki lava flow, Iceland, x Distance below the top surface of a lava Bull. Volcanol., 72, 449–467. flow (m). Hamilton, C. W., S. A. Fagents, and T. Thordarson (2010b), Explosive lava‐water interactions II: Self‐organization processes among volcanic rootless eruption sites in the 1783–1784 Laki lava flow, Iceland, Bull. [71] Acknowledgments. We thank Thorvaldur Thordarson, Scott Volcanol., 72, 469‐485. Rowland, Bruce Houghton, and Susanne Still for their insightful comments Hartmann, W. K. (1966), Early lunar cratering, Icarus, 5,406–418, during the preparation of this manuscript. We also thank Lori Glaze and an doi:10.1016/0019-1035(66)90054-6. anonymous reviewer for their excellent recommendations during the review Hartmann, W. K. (1999), Martian cratering VI. Crater count isochrons and process. Financial support was provided from the NASA Mars Fundamen- evidence for recent volcanism from Mars Global Surveyor, Meteorit. tal Research Program (MFRP) grant NNG05GM08G, and NASA Mars Planet. Sci., 34, 167–177, doi:10.1111/j.1945-5100.1999.tb01742.x. Data Analysis Program (MDAP) grant NNG05GQ39G. SOEST publication Hartmann, W. K. (2005), Martian cratering 8: Isochron refinement and the number 7979. HIGP publication number 1866. history of Martian geologic activity, Icarus, 174, 294–320, doi:10.1016/j. icarus.2004.11.023. Hartmann, W. K. (2007), Martian cratering 9: Toward resolution of the References controversy about small craters, Icarus, 189, 274–278, doi:10.1016/j. Aharonson, O., M. T. Zuber, G. A. Neumann, and J. W. Head III (1998), icarus.2007.02.011. Mars: Northern hemisphere slopes and slope distributions, Geophys. Res. Hartmann, W. K., and D. C. Berman (2000), Elysium Planitia lava flows: Lett., 25(24), 4413–4416, doi:10.1029/1998GL900057. Cratercount chronology and geological implications, J. Geophys. Res., Baloga, S. M., L. S. Glaze, and B. C. Bruno (2007), Nearest‐neighbor 105, 15,011–15,025, doi:10.1029/1999JE001189. analysis of small features on Mars: Applications to tumuli and rootless Head, J. W., and L. Wilson (2002), Mars: A review and synthesis of gen- cones, J. Geophys. Res., 112, E03002, doi:10.1029/2005JE002652. eral environments and geological settings of magma/H2O interactions, in Barlow, N. G. (2008), Mars: An Introduction to its Interior, Surface, and Volcano‐Ice Interaction on Earth and Mars, edited by J. L. Smellie and Atmosphere, 264 pp., Cambridge Univ. Press, New York, doi:10.1017/ M. G. Chapman, Geol. Soc. Spec. Publ., 202, 27–57. CBO9780511536069. Head, J. W., L. Wilson, and K. L. Mitchell (2003), Generation of recent Berman, D. C., and W. K. Hartmann (2002), Recent fluvial, volcanic, and massive water floods at Cerberus Fossae, Mars by dike emplacement, tectonic activity on the Cerberus plains of Mars, Icarus, 159,1–17, cryospheric cracking, and confined aquifer groundwater release, Geo- doi:10.1006/icar.2002.6920. phys. Res. Lett., 30(11), 1577, doi:10.1029/2003GL017135. Bruno, B. C., S. A. Fagents, T. Thordarson, S. M. Baloga, and E. Pilger Head, J. W., D. R. Marchant, F. Forget, J. Laskar, J.‐B. Madeleine, and J. L. (2004), Clustering within rootless cone groups on Iceland and Mars: Fastook (2009), Deciphering the late Amazonian climate history of Mars: Effect of nonrandom processes, J. Geophys. Res., 109, E07009, Assessing obliquity predictions with geological observations and atmo- doi:10.1029/2004JE002273. spheric general circulation models, Lunar Planet. Sci., XL, Abstract 1349. Bruno, B. C., S. A. Fagents, C. W. Hamilton, D. M. Burr, and S. M. Baloga Hon, K., J. Kauahikaua, R. Denlinger, and R. MacKay (1994), Emplace- (2006), Identification of volcanic rootless cones, ice mounds, and impact ment and inflation of pahehoe sheet flows: Observations and measure- craters on Earth and Mars: Using spatial distribution as a remote sensing ments of active lava flows on Kilauea Volcano, Hawaii, Geol. Soc. tool, J. Geophys. Res., 111, E06017, doi:10.1029/2005JE002510. Am. Bull., 106,351–370, doi:10.1130/0016-7606(1994)106<0351: Burr, D. M., and A. Parker (2006), Grjotá Valles and implications for flood EAIOPS>2.3.CO;2. sediment deposition on Mars, Geophys. Res. Lett., 33, L22201, Jaeger, W. L., L. P. Keszthelyi, A. S. McEwen, C. M. Dundas, and P. S. doi:10.1029/2006GL028011. Russell (2007), Athabasca Valles, Mars: A lava‐draped channel system, Burr, D. M., J. A. Grier, A. S. McEwen, and L. P. Keszthelyi (2002), Science, 317, 1709–1711, doi:10.1126/science.1143315. Repeated aqueous flooding from the Cerberus Fossae: Evidence for very Jaeger, W. L., L. P. Keszthelyi, A. S. McEwen, C. M. Dundas, and P. S. recently extant, deep groundwater on Mars, Icarus, 159,53–73, Russell (2008), Response to comment on “Athabasca Valles, Mars: A lava‐ doi:10.1006/icar.2002.6921. draped channel system”, Science, 320, 1588, doi:10.1126/science.1155124. Carr, M. (1996), Water on Mars, 248 pp., Oxford Univ. Press, Oxford, U. K. Jaeger, W. L., et al. (2010), Emplacement of the youngest flood lava on Mars: Carr, M. (2006), The Surface of Mars, 307 pp., Cambridge Univ. Press, A short, turbulent story, Icarus, 205, 230–243, doi:10.1016j.icarus. New York, doi:10.1017/CBO9780511536007. 2009.09.011. Carslaw, H. S., and J. C. Jaeger (1986), Conduction of Heat in Solids, Jakosky, B. M., and E. L. Shock (1998), The biological potential of Mars, 2nd ed., 510 pp., Oxford Univ. Press, New York. the early Earth and Europa, J. Geophys. Res., 103, 19,359–19,364, Fagents, S. A., and T. Thordarson (2007), Rootless volcanic cones in Ice- doi:10.1029/98JE01892. land and on Mars, in The Geology of Mars: Evidence From Earth‐Based Keszthelyi, L., and R. Denlinger (1996), The initial cooling of pahoehoe Analogs, edited by M. G. Chapman, pp. 151–177, Cambridge Univ. flow lobes, Bull. Volcanol., 58,5–18, doi:10.1007/s004450050121. Press, New York. Keszthelyi, L., A. S. McEwen, and T. Thordarson (2000), Terrestrial ana- Fagents, S. A., P. Lanagan, and R. Greeley (2002), Rootless cones on Mars: logs and thermal models for Martian flood lavas, J. Geophys. Res., A consequence of lava‐ground ice interaction, in Volcano‐Ice Interaction 105(E6), 15,027–15,049, doi:10.1029/1999JE001191. on Earth and Mars, edited by J. L. Smellie and M. G. Chapman, Geol. Keszthelyi, L., T. Thordarson, A. McEwen, H. Haack, M.‐N. Guilbaud, Soc. Spec. Publ., 202, 295–317. S. Self, and M. J. Rossi (2004), Icelandic analogs to Martian flood lavas, Farmer, C. B., and P. E. Doms (1979), Global seasonal variation of water Geochem. Geophys. Geosyst., 5, Q11014, doi:10.1029/2004GC000758. vapor on Mars and the implications for permafrost, J. Geophys. Res., 84, Lanagan, P. D., A. S. McEwen, L. P. Keszthelyi, and T. Thordarson (2001), 2881–2888, doi:10.1029/JB084iB06p02881. Rootless cones on Mars indicating the presence of shallow equatorial Frey, H., and M. Jarosewich (1982), Subkilometer Martian volcanoes: ground ice in recent times, Geophys. Res. Lett., 28,2365–2367, Properties and possible terrestrial analogs, J. Geophys. Res., 87,9867– doi:10.1029/2001GL012932. 9879, doi:10.1029/JB087iB12p09867. Laskar, J., A. C. M. Correia, M. Gastineau, F. Joutel, B. Levrard, and Frey, H., B. L. Lowry, and S. A. Chase (1979), Pseudocraters on Mars, P. Robutel (2004), Long term evolution and chaotic diffusion of the J. Geophys. Res., 84, 8075–8086, doi:10.1029/JB084iB14p08075. insolation quantities of Mars, Icarus, 170,343–364, doi:10.1016/j. Fuller, E. R., and J. W. Head III (2002), Amazonis Planitia: The role of icarus.2004.04.005. geologically recent volcanism and sedimentation in the formation of Link, L. S., B. Jakosky, and G. D. Thyne (2005), Biological potential of the smoothest plains on Mars, J. Geophys. Res., 107(E10), 5081, low‐temperature aqueous environments on Mars, Int. J. Astrobiol., 4(2), doi:10.1029/2002JE001842. 155–164, doi:10.1017/S1473550405002648. Greeley, R., and S. A. Fagents (2001), Icelandic pseudocraters as analogs to Mellon, M. T., and B. M. Jakosky (1995), The distribution and behavior of some volcanic cones on Mars, J. Geophys. Res., 106, 20,527–20,546, Martian ground ice during past and present epochs, J. Geophys. Res., doi:10.1029/2000JE001378. 100, 11,781–11,799, doi:10.1029/95JE01027. Guilbaud, M.‐N., S. Self, T. Thordarson, and S. Blake (2005), Flow forma- Plescia, J. B. (2003), Cerberus Fossae, Elysium, Mars: A source for lava tion, surface morphology, and emplacement mechanism of the AD 1783– and water, Icarus, 164,79–95, doi:10.1016/S0019-1035(03)00139-8. 4 Laki lava, in Kinematics and Dynamics of Lava Flows, edited by Self, S., T. Thordarson, L. Keszthelyi, G. P. L. Walker, K. Hon, M. T. Murphy, M. Manga and G. Ventura, Spec. Pap. Geol. Soc. Am., 396, 81–102. P. Long, and S. Finnemore (1996), A new model for the emplacement Hamilton, C. W., and S. A. Fagents (2009), The Tartarus Colles cone group of Columbia River Basalts as large, inflated pahoehoe lava flow fields, and its implications for explosive lava‐water interactions in the Grjotá Geophys. Res. Lett., 23, 2689–2692, doi:10.1029/96GL02450. Valles region of Mars, Lunar Planet. Sci., XL, Abstract 1924.

23 of 24 E09006 HAMILTON ET AL.: LAVA‐WATER INTERACTIONS ON MARS E09006

Self, S., L. Keszthelyi, and T. Thordarson (1998), The importance of pahoe- Walker, G. P. L. (2009), The endogenous growth of pahoehoe lava lobes hoe, Annu. Rev. Earth Planet. Sci., 26,81–110, doi:10.1146/annurev. and morphology of lava‐rise edges, in Studies in Volcanology: The Legacy earth.26.1.81. of George Walker, edited by T. Thordarson et al., Spec. Publ. IAVCEI,2, Shock, E. L. (1997), High temperature life without photosynthesis as a 17–32. model for Mars, J. Geophys. Res., 102, 23,687–23,694, doi:10.1029/ Wilson, L., and J. W. Head III (2007), Heat transfer in volcano‐ice interac- 97JE01087. tions on Earth, Ann. Glaciol., 45(1), 83–86, doi:10.3189/ Squyres, S. W., S. M. Clifford, R. O. Kuzmin, J. R. Zimbelman, and F. M. 172756407782282507. Costard (1992), Ice in the Martian regolith, in Mars,editedbyH.H. Wohletz, K. (1986), Explosive magma‐water interactions: Thermodynam- Kieffer et al., pp. 523–554, Univ. of Ariz. Press, Tucson. ics, explosion mechanisms, and field studies, Bull. Volcanol., 48, 245– Tanaka, K. L., J. A. Skinner, and T. M. Hare (2005), Geologic map of the 264, doi:10.1007/BF01081754. northern plains of Mars, U.S. Geol. Surv. Sci. Invest. Map, 2888. Wohletz, K. (2002), Water/magma interaction: Some theory and experi- Thorarinsson, S. (1951), Laxargljufur and Laxarhraun: A tephrachronologi- ments on peperite formation, J. Volcanol. Geotherm. Res., 114,19–35, cal study, Geogr. Ann., H1–2,1–89. doi:10.1016/S0377-0273(01)00280-3. Thorarinsson, S. (1953), The crater groups in Iceland, Bull. Volcanol., 14, Zimanowski, B. (1998), Phreatomagmatic explosions, in From Magma 3–44, doi:10.1007/BF02596003. to Tephra, edited by A. Freundt and M. Rosi, pp. 25–554, Elsevier, Thordarson, T., and S. Self (1998), The Roza Member, Columbia River Amsterdam. Basalt Group: A gigantic pahoehoe lava flow formed by endogenous pro- cesses, J. Geophys. Res., 103, 27,411–27,445, doi:10.1029/98JB01355. S. A. Fagents and C. W. Hamilton, Hawaii Institute of Geophysics and Varnes, E. S., B. M. Jakosky, and T. M. McCollom (2003), Biological ‐ – Planetology, University of Hawaii, 1680 East West Road, Honolulu, HI potential of Martian hydrothermal systems, Astrobiology, 3(2), 407 96822, USA. ([email protected]) 414, doi:10.1089/153110703769016479. L. Wilson, Department of Earth and Planetary Sciences, Lancaster Walker, G. P. L. (1991), Structure, and origin by injection of lava under University, Lancaster, UK. surface crust, of tumuli, “lava rises,”“lava‐rise pits,” and “lava‐inflation clefts” in Hawaii, Bull. Volcanol., 53(7), 546–558, doi:10.1007/ BF00298155.

24 of 24