Climate optimum on Mars initiated by atmospheric collapse

Edwin S. Kite1, Michael A. Mischna2, Peter Gao3,4, Yuk L. Yung5

1. University of Chicago (kite@uchicago.edu), Chicago, IL 60637. 2. Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109. 3. NASA Ames Research Center, Mountain View, CA 94035. 4. University of California, Berkeley, CA 94720. 5. California Institute of Technology, Pasadena, CA 91125.

Abstract.The progressive drying-out of Mars’ surface was punctuated by a dramatic transient increase in fluvial erosion around the Noachian-Hesperian boundary (~3.7 Ga). Standard explanations of this climate optimum appeal to volcano- or impact-triggered climates and imply that individual runoff episodes were brief, apparently inconsistent with evidence for persistent runoff. We examine a scenario in which the duration, intensity and uniqueness of the Noachian-Hesperian climate optimum result from degassing of CH4-clathrate consequent to atmospheric collapse. Atmospheric collapse causes low-latitude surface water ice to sublimate away, depressurizing and thus destabilizing CH4 clathrate in subglacial pore space. Subsequent atmospheric re-inflation leads to warming that further destabilizes CH4 clathrate. CH4-induced warming is efficient, permitting strong positive feedbacks, and possibly raising Mars into a climate optimum. The optimum is brought to a close by photochemical destruction of CH4 or by a new atmospheric collapse, and drawdown of the CH4-clathrate reservoir prevents recurrence. This scenario predicts a 105-106 yr climate optimum, transient connections between the deep hydrosphere and the surface, mud volcanism, and strong surface weathering, all of which are consistent with recent observations. Crustal hydrothermal circulation very early in Mars history could yield CH4 that would be incorporated into clathrate on approach to the cold surface. The scenario explains why regional watershed integration on Mars occurred relatively late and apparently only once, and suggests that the contrasts between Noachian versus Hesperian climate-sensitive deposits on Mars correspond to a transition from a never-collapsed atmosphere to a collapse-prone climate, ultimately driven by slow loss of CO2 to space.

1. The problem: an anomalous wet-climate episode on Mars.When Mariner 9 reached Mars orbit in 1971, it was soon realized that ancient Mars terrain is crisscrossed by networks of dry valleys. Today, valley-network formation around the Noachian/Hesperian boundary (~3.7 Ga) is recognized as the biggest anomaly in Mars’ geomorphic record of global drying [1]. The precipitation-fed valleys flowed into lakes that covered 2-20% of the ancient terrain [2-3]. Lakes spilled over to form regionally-integrated watersheds [4]. This widespread regional

1

integration of watersheds is unique in the geomorphic record, and in sharp contrast to smaller-scale fluvial integration both before and subsequently [4-7] (Fig. 1). Valley depths suggest that, during the climate optimum, column runoff totaled >5 km (Supplementary Information). This water could not have been delivered catastrophically because some, but not all lakes overflowed. Instead, a hydrological cycle lasting >105 yr [8] or 105-107 yr [9] is inferred. Because more of the planet’s surface was covered by liquid water than at any other time in Mars’ discernable history, we refer to this time interval as a climate optimum.

Given that Mars’ valley-network-forming climate is the most habitable known extraterrestrial surface environment, and widespread interest in the physical basis of sustained surface habitability [e.g., 10-11], it is perhaps surprising that there is currently no physical model that accounts for the distinctiveness, late start, and relative brevity of the Mars climate optimum [5,12]. The climate optimum must be connected in some way to the evolution of Mars atmospheric CO2, because today Mars’ atmosphere is too thin (atmospheric pressure ≈ partial pressure of CO2, PCO2 = 6 mbar) to permit extensive liquid water, and past PCO2 was much greater, due to gradual and irreversible atmospheric loss by carbonate formation and escape-to-space [13-14]. Because models suggest that the rate of atmospheric escape-to-space fell off more quickly than the rate of volcanic outgassing, some models yield a broad maximum in PCO2 around the Noachian-Hesperian boundary [15]. However, CO2–warming alone is insufficient to explain the climate optimum (even with H2O-vapor feedback) [16-17]. What was the source of additional warming? Why was it limited to the observed interval around the Noachian/Hesperian boundary? Why does Mars’ surface show relatively little evidence for intense aqueous weathering, despite the evidence for lakes and rivers? Why was the uptick in warming not repeated?

Previously proposed triggers for the climate optimum include bolide impacts, volcanic eruptions, true polar wander, and climate cycles [16]. However, almost all large bolide impacts on Mars predate the climate optimum [18], and the direct warming due to bolide impact is too brief to match lake-overspill constraints [8]. For bolide impact to kick-start a hydrological cycle, the impact must access a metastable climate state for which there is no uncontested evidence [16]. Volcanic SO2 warming [19] is weak and ephemeral [20]. Rapid polar wander (inertial-interchange) is unlikely [21]. Climate cycles involving rapid silicate weathering during wet intervals [22] predict >3 × 104 kg/m2 carbonate in Noachian/Hesperian boundary sediments, which is not observed [23]. For Early Mars climate more broadly, the gap between proxy data and models remains wide [24-25].

2

Fig. 1. Geomorphic history of Mars’ river-forming climates. The valley network timing is from Ref. 26, using the Neukum chronology. Data and models for magma flux and crater flux are consistent with monotonic decrease, unlike the geomorphic record. (Modified after Ref. 27).

Here we propose a new scenario for the climate optimum, that makes use of recent advances in the understanding of Mars’ CO2 and H2O cycles [e.g., 28-29]. The scenario tracks the transient release of CH4 gas, released by atmospheric-collapse-triggered depressurization and warming of CH4 clathrate. CH4 clathrates are a candidate driver of climate change on Titan and on Phanerozoic Earth [30-31]. Yet CH4 from clathrates is rarely invoked as a greenhouse agent on Mars, in part because CH4 has a <106 yr photochemical lifetime [32-33]. However, recent work shows that <106 yr pulses of liquid water are consistent with data, and shows that the warming potential of O(1%) CH4 in Early Mars’ atmosphere has been underestimated [34-35], indicating that this dismissal may have been premature. The collapse-trigger mechanism explains why valley incision occurred around the Noachian/Hesperian boundary but not thereafter, explains the duration of the event, seems consistent with other orbiter and rover data, and makes several testable predictions.

2. Recent advances in the understanding of Early Mars set the stage for the scenario.The collapse trigger scenario is motivated by evidence for the following:- (i) Noachian water-rock reactions in the deep subsurface; (ii) a 0.5 bar past ≳atmosphere that (iii) drove surface H2O ice to high ground; and (iv) chaotic obliquity variations. Understanding of all four has improved in recent years.

(i) Noachian water-rock reactions are recorded by hydrothermal minerals [e.g. 36-39]. Hydrothermal reactions between Mars’ mafic/ultramafic crust and waters charged with magmatic and/or atmospheric C should yield H2 and CH4 [e.g. 40]. We emphasize CH4 here and assume abiotic production [41]. CH4 production faces kinetic barriers that can be overcome by high temperatures, recycling of fluid by hydrothermal circulation, and olivine-rich rocks that host catalysts (e.g., Ni). The reaction-rate would slacken as the Noachian progressed and the geotherm cooled:

3

Fig. 2. Overview of the collapse-triggered climate optimum scenario. Inset at top: Showing effect of atmospheric collapse and reinflation on the phase diagram of methane clathrate. Upper panel: Depth-versus-time schematic of the long-term evolution of a column of the low-latitude highlands of Mars. Figs. 3-4 show details. Lower panel: Latitude-versus-time schematic, showing key volatile reservoirs: CO2, H2O, and CH4. Below φc, 90-99.9% of the atmosphere will condense in ~1 kyr (step 1). This unloads high ground (step 2), releasing some CH4 from sub-ice clathrate. Later re-inflation of the atmosphere triggered by φ rise (step 3) leads to massive CH4 release and climate optimum.

4

most evidence for hydrothermal circulation predates the Noachian-Hesperian boundary [42-43]. Methane produced by water-rock reactions in the deep subsurface would be transported by hydrothermal circulation to the cooling near-surface (e.g. beneath ice sheets or primordial seas), and trapped into clathrates (Fig. 2) [44-46]. Methane clathrate is stabilized by increasing pressure (stable for P ≥ 2.1 MPa at 273.15K), and by decreasing temperature (stable for P ≥ 15 kPa at 200 K). Once formed, clathrate could be destabilized by orbital forcing [34,47] or by other mechanisms (ref. 33; this paper). Methane clathrate destabilization on Mars has been proposed as an explanation for chaos terrain and mounds interpreted as mud volcanoes (e.g., ref. 48). (ii) Most estimates of PCO2 at the time of the valley networks are in the range 0.2 – 2 bar [e.g., 49-50]. For such thick atmospheres, surface temperatures track the atmospheric thermal lapse rate [29]. (iii) As a result, H2O ice would be located at high ground [29], including most of the Southern hemisphere. Average ice sheet thickness, assuming ice above +1 km elevation, was ≥300 m [e.g., 51-52]. This thickness is sufficient to stabilize CH4 clathrate in the regolith pore-space beneath the ice sheet. (iv) Unlike Earth, Mars undergoes chaotic large-amplitude shifts in obliquity, φ ( = φ 0-70°; ref. 53) on which are superposed quasi-periodic φ variations of period 105-106 yr and amplitude 0-20° (10× Earth).

3. Climate optimum scenario.The proposed scenario (Fig. 2) is motivated by a separation of timescales: CO2-atmosphere collapse and re-inflation (~103 yr) is lagged by H2O ice migration (105-107 yr). Furthermore, the thermal-conduction timescale linking surface warming to subsurface clathrate breakdown, and the time lag between clathrate breakdown and release of CH4 to the atmosphere following, are both much shorter ( 10≲ 3 yr) than photochemical destruction of CH4 (~105 yr). Thus, CH4-induced warming can cause further CH4 release.

Step 1. Collapse of an initially-thick CO 2 atmosphere. As the atmosphere is thinned by escape-to-space and carbonate formation, the remaining atmospheric CO2

becomes increasingly vulnerable to collapse (Fig. 3). Collapse is triggered if polar CO2 ice caps grow from year to year. Cap growth occurs below a critical polar temperature. Polar temperature is set by insolation, which increases with obliquity, and by heating from the CO2 atmosphere (both the greenhouse effect, and equator-to-pole heat transport, increase with PCO2). Given secular atmospheric-pressure decline, obliquity variations will eventually lower insolation below the threshold for perennial CO2-ice caps. Once nucleated, caps grow quickly (103-104 yrs; ref. 28) because ice-cap sequestration of CO2 further reduces polar temperatures. Collapse typically sequesters 90-99.9% of the planet’s CO2 . Post-collapse PCO2 is in equilibrium with polar CO2-ice cap surface temperature - i.e., very low (Fig. 3) [55]. Assuming caps poleward of 80°, CO2 ice cap thickness is ~1200 m/bar CO2 (lower if caps flow; ref. 56). Atmospheric collapse involves a hysteresis (Fig. 3, Fig. S2), but is

5

ultimately reversible. CO2 trapped as CO2 ice is only sequestered temporarily. Eventually, obliquity rise will trigger rapid re-sublimation of CO2 condensed at low obliquity. Many collapse/re-inflation cycles should have occurred in Mars history [29], and there is strong geologic evidence for recent collapse (e.g., ref. 57). However, Mars’ small total CO2 inventory at present precludes the 102-103-fold swings in pressure that should have occurred on Early Mars. There is sedimentological evidence for PCO2 < 10 mbar at ~3.7 Ga [58], in which case the first collapse occurred >3.7 Ga. There is also indirect noble-gas evidence that collapse did not occur prior to ~4 Ga, because collapse in the face of strong XUV ~4 Ga would lead to virtually complete loss of noncondensible noble gases, in contradiction with data [50].

Fig. 3. Phase portrait for atmospheric collapse on Mars, showing how atmospheric collapse drives H2O-ice distribution (this calculation uses the GCM of ref. 17). Thin black lines show annual-mean polar temperature as a function of atmospheric pressure assuming Faint Young Sun luminosity. Thick black line is the condensation curve for CO2; atmospheres below this line are collapsing onto polar CO2 ice caps (e.g., AB). Blue dashes outline the approximate pressures and obliquities below which H2O ice is stable only at Mars’ poles [e.g. 17, 29, 54]. Collapse leads to relocation of surface H2O ice from highlands to poles. In this GCM, for an initial CO2 inventory of 8×1018 kg (= 2 bar), the atmosphere is stable until φ ≤ 15 deg (at A). Rapid collapse (~103 yr) moves the system to point B. Increasing obliquity (over 105-107 yr) moves the (ice cap)/atmosphere system along the condensation curve to C, (the highest consistent with permanent COφ 2 ice caps). Further rise leads to φsublimation of the CO2 ice cap (~103 yr) and the system returns to A.

6

Cooling of the surface due to loss of the greenhouse effect helps to stabilize CH4 -clathrate, far outweighing the destabilizing effect of loss of the weight of the atmosphere. Therefore, no CH4 release occurs at this stage. Cold post-collapse conditions are unfavorable for liquid H2O [59]. Correspondingly, we are not aware of any previous suggestion that atmospheric loss and atmospheric collapse could favor surface habitability.

Step 2: H2O-ice unloading of low latitudes triggers some CH 4 release. Following Mars’ first-ever atmospheric collapse, the <100 mb atmosphere can no longer provide much heat to the poles. The cold poles are now the stable location for H2O ice. Water ice condenses at the poles, and sublimates away from low-latitude highlands for the first time in Mars’ history. The sublimation rate is low, due to the Faint Young Sun: ~0.1 mm/yr (according to our GCM), rising to ~1 mm/yr for a dust-like ice albsedo. Thus, the timescale of complete H2O-ice unloading depends on the initial H2O thickness. For a thickness of 300 m, unloading time is 3×105-3×106 yr. On this timescale H2O-ice glacial flow is small [51].

This slow latitudinal shift in H2O-ice overburden pressure depressurizes and thus destabilizes CH4-clathrate in subglacial pore space. Unloading-induced destabilization eventually exceeds cooling-induced stabilization. The top of the clathrate hydrate stability zone (HSZ) will deepen. Clathrate that is stranded above the deepening HSZ boundary will dissociate in <1 kyr [60-61]. We assume that the CH4 released will be swiftly discharged to the atmosphere, in part because of fracturing induced by the large volume change involved in decomposition.

The corresponding CH4(g) release is proportional to g, the fraction of surface area initially shrouded by ice, and to f, the fraction of HSZ volume occupied by clathrate. g ≈ 50% in published models [e.g., 29]. Although the growing polar (H2O+CO2)-ice cap overburden stabilizes polar clathrate, this represents <10% of the planet’s surface area and is ignored. The clathrate is charged up with CH4 produced by water-rock reactions over >108 yr (Fig. 2a). It is released <103 y after destabilization [60-61], possibly via explosive blow-outs and mud volcanism [62]. How closely outgassing approaches this maximum will depend on the prior history of unloading. Because in our model, the clathrate reservoir is recharged slowly if at all, zero CH4 will be released from regolith volume elements that have formerly been destabilized. We track these histories using realistic orbital forcing (Figure S8).

Methane is ineffective at warming Mars at this stage, mainly because (as on today’s Mars) the outgassing source is swamped by photochemical sinks [63]. Moreover, what little CH4(g) exists is radiatively ineffective, because CH4-CO2 collision-induced absorption (CIA) is weak when the collapsed-atmosphere PCO2 is low [35].

7

(a) (b)

Fig. 4. (a) Map of initial depth (within regolith), in meters, to the CH4-clathrate hydrate stability zone (HSZ). This depth is zero where H2O-ice (located at topographic elevations >+1 km, by assumption) is thick enough to stabilize CH4 clathrate throughout the regolith. Vertical line at 27°E is shown in cross-section (b). (b) Blue line corresponds to initial HSZ boundary. Red line corresponds to cold, post-collapse conditions, assuming complete H2O movement to the poles. Yellow line corresponds to re-warmed, post-reinflation conditions, assuming H2O has not yet returned from the poles, but not including CH4 warming. Green line shows effect of an additional 10 K of warming (e.g., due to CH4-CO2 CIA).

Step 3. CO 2 -atmosphere re-inflation and further CH 4 release. Mars atmospheric collapse occurs at low obliquity. As obliquity rises, the collapsed atmosphere is still too thin to warm the poles – a hysteresis effect (Fig. S2). Therefore, PCO2 rises slowly in vapor-pressure equilibrium with polar temperature, and with most CO2 sequestered in ice. The value of PCO2 rises in equilibrium with CO2-ice temperature because, although stiff H2O-ice may enshroud the CO2 -ice, flow of thick, soft CO2–ice [56] opens crevasses. When obliquity rises to the point where no value of pressure yields a polar temperature below the condensation point, the remaining polar CO2-ice caps rapidly (103 yr) and completely sublimate (Fig. 5).

Loss of the stabilizing H2O-ice overburden exposes clathrate to the CO2-greenhouse warming associated with re-inflation, and clathrate irreversibly decomposes, releasing CH4. The wait time for CH4 clathrate decomposition is, for 102 m depth-to-HSZ, equal to (re-inflation time + subsurface conductive-warming time + decomposition time) = (103 yr + 102 yr + <10 yr) ≈ 103 yr. Because this is much quicker than the >106 yr needed for all H2O ice to return to the low latitudes [29], CH4 is outgassed before H2O ice can re-load (and re-stabilize) the highlands. This is also much faster than the CH4 photolysis timescale. In the simulations, the atmospheric concentration of CH4 is much larger after Step 3 than in Step 2, because of rapid release (Fig. S4).

8

In our scenario, the climate optimum occurs for the first atmospheric collapse that lasts for 1 Myr (i.e., a collapse so deep that the peaks of individual 10≳ 5 yr obliquity cycles are insufficient to trigger atmospheric re-inflation). This deep collapse gives time for substantial unloading of the highlands, such that much of the remaining CH4-clathrate reservoir is destabilized upon re-inflation. The first prolonged collapse happens geologically soon ( 10≪ 9 yr) after Mars’ first-ever atmospheric collapse (Supplementary Information). (By contrast, brief collapses do not cause a large CH4-burst, because, according to our GCMs, unloading is so incomplete that most of the highlands clathrate is still stabilized by the H2O(i) remaining at the time of re-inflation). We track CH4-clathrate through multiple collapse/re-inflation cycles and ensure that the same subsurface volume is not degassed twice. Once degassed, pore space is not recharged with CH4, because diffusion of CH4 through cold clathrate is slow.

For f > 3%, the atmosphere now contains 1-5% CH4 with abundant CO2 – ideal conditions for strong CH4-CO2 CIA warming [35].

Fig. 5. Example climate evolution model output. Top: Partial pressure of CO2 (blue) and CH4

(red). Pressure evolution. Typical collapse/re-inflation follows on from >>100 Myr of continuously high PCO2. Bottom: Temperature evolution. After re-inflation, a ~100 kyr-long, >10 K warming occurs. Green highlights bringing times when annual mean temperatures in the ±40° latitude, -2 to +3 km elevation zone (valley network zone; ref. 69) exceed those for lakeshore weather stations in Taylor Valley, Antarctica (Doran et al. 2002).

Step 4. CH 4 + CO 2 = Mars climate optimum. The cocktail of circumstances enabled by Mars’ first atmospheric collapse (a massive CH4 pulse, a thick CO2 atmosphere, low-latitude H2O ice) now permits low-latitude rivers and lakes (Fig. 5). H2O snow falling

9

on the equator (plus any H2O ice that did not have time to sublimate) encounters high insolation and high greenhouse forcing. Seasonal snowmelt runoff forms valleys that drain into perennial ice-covered lakes [e.g. 64]. For warming exceeding 10 K, temperatures exceed measured lake-shore temperatures in Taylor Valley, Antarctica [65] over most of the Southern Highlands (Fig. 6). Lakes overspill to form regionally-integrated valley networks. Lake-bottom warming is a further positive feedback on greenhouse warming. High lake-bottom temperatures destabilize sub-lake CH4-clathrates (Fig. 4). Methane-induced warming is strong to destabilize additional CH4 in the regolith (and under some circumstances this positive feedback can produce runaway CH4 outgassing; Fig. S6). CH4 can remain in the atmosphere at radiatively important levels and thus maintain wet conditions for 105-106 yr, consistent with geomorphic estimates of the time needed to form the VNs. This duration allows sublake taliks to connect with deep aquifers. The required valley incision rate is 0.1-1 mm/yr, similar to the mountainous Western United States [66]. This duration also satisfies mineralogic upper limits on past liquid H2O availability [67-68].

Fig. 6. Expected geographic distribution of rivers and lakes. Before the climate optimum, conditions warmer than lakeshore weather stations in Taylor Valley, Antarctica only occur in the deepest part of Mars (thick black line). However, with 17 K of uniform CH4-induced warming (consistent with our outgassing model), the area warmer than lakeshore weather stations in Taylor Valley Antarctica expands to cover (thick red line) most of the planet. Dark gray area is obscured by young lavas. Light gray dots are mapped valleys [69], some of which postdate the climate optimum. Colored contours show topography in meters (present-day topography is used).

Step 5. CH 4 loss shuts down wet climate in 10 5 -10 6 yr. The wet climate ends because atmospheric CH4 is destroyed by photochemical processes (Fig. S7). As CH4 is destroyed without resupply, eventually the rivers dry up, and atmospheric collapse

10

is again possible. At this time, many mbar of CH4 remain in the atmosphere, so the newly-collapsed atmosphere will have a CH4/CO2 ratio that transiently exceeds the threshold for Titan-like photochemistry, haze, and soot [70]. This phase is brief, because CH4 destruction is rapid when CO2 shielding is ineffective.

4. Parameter sensitivities and the possibility of rainfall.The most important parameter in our scenario is f, the fraction of the never-degassed portion of the HSZ that is initially occupied by clathrate (f can also be interpreted as a measure of degassing efficiency, if some CH4(g) is trapped by permafrost). For f < 0.01, little warming occurs. For f ~ 0.01-0.05, rivers and ice-covered lakes are fed by snowmelt. For f > 0.05, rivers and lakes are fed by rainwater.

Methane is released by the warming triggered by CH4 release. This feedback CH4 release increases with f. For large f, runaway clathrate breakdown is possible (Fig. S6). Runaway outgassing can produce a Tave > 273 K (temperate) Mars climate (Fig. S6) [71]. A temperate Mars climate lasting ~1 Myr might explain the surface leaching event apparently recorded by Al-clays globally [72].

Runaway clathrate breakdown depends on f, PCO2 and Tinit. The minimum f for runaway clathrate breakdown increases with decreasing PCO2 (because the CIA warming per unit CH4 increases with PCO2), and increases for cooler pre-release temperatures (due to nonlinearity in the clathrate phase diagram) (Figs. S5-S6).Even when the system does not runaway, the positive feedback on CH4-induced warming can be large (Figs. S5). Because warming has a nonlinear dependence on pCH4 [35], the warming-outgassing-warming cycle may self-arrest before the clathrate reservoir is fully depleted (Fig. S6).

5. Discussion and predictions.The collapse trigger scenario makes testable predictions for Mars geology: (1) absence of Noachian large wind ripples [58], because these form only at low atmospheric density; (2) absence of D < 50 m craters that predate the valley networks [49], because if PCO2 was high before the valley networks as required by the model, then small impactors would be screened by the atmosphere; and (3) extreme rarity of impact craters (of any size) that formed during the main interval of valley network integration, because interbedded craters would indicate an extended interval of valley network integration.

Serpentine outcrops are widespread on Mars [37] but uncommon. This is not surprising, because serpentinization may be mostly restricted to rocks too deep for impact exhumation [36, 39]. Regardless of production depth and process, CH4-charged waters can be swept to the near-surface by hydrothermal circulation [38].

11

On Earth, mud volcanism is usually powered by CH4 gas. There is strong and widespread evidence for mud volcanism on Mars [e.g., 48].

The collapse trigger scenario suggests new interpretations of Mars geochemistry. Predictions include:- (1) Widespread charge-up of the clathrate reservoir; clathrate remnants could persist today and diffusively release CH4 that could be detected by spacecraft; (2) A major shift in Mars’ Cl cycle around the time of valley-network incision, because ClO4

- production is thought to require galactic cosmic radiation (GCR) [73] that is absorbed by the thick atmosphere that exists before the first atmospheric collapse; (3) A significant change in Mars’ S cycle around the time of valley network incision [74]. In a <200 mbar collapsed Mars atmosphere, but not in a ≥2 bar atmosphere before the first collapse, S species dominate volcanic degassing [75], and atmospherically-derived S can build up to high concentrations in sediments, because reworking by erosion and weathering under a thin atmosphere is slow. Therefore, geologic mapping should show that S-rich layered deposits postdate the climate optimum; (4) Enrichment in photochemical soot (<103-104 kg/m2) just after the climate optimum. Such soots can indicate a reducing (CH4/CO2 ≳ 0.1) past atmosphere (Supplementary Information). CH4/CO2 ~ O(1) is predicted to occur immediately following climate optimum. Complex abiotic organic matter stratigraphically associated with river and lake deposits is a potential life detection false positive.

Our model suggests (Supplementary Information) a physical basis for the late timing of valley network formation.

The collapse trigger scenario has implications for Mars astrobiology. For astrobiology rovers, climate-optimum fluviolacustrine deposits represent candidate landing sites with good life detection potential [76]. However, in the collapse trigger scenario, sterilizing surface conditions (temperatures ~200 K and >1 MRad surface radiation doses; ref. 77) immediately precede the climate optimum. Near-surface refugia might occur near hot springs. The sub-lake taliks predicted by the model would allow deep-subsurface biota, if any, to inoculate the lakes.

Planets with Mars’ size and temperature are likely common in the universe (e.g., TRAPPIST-1h). Our scenario suggests two false positives for exoplanet habitability. (1) Chemical disequilibrium is considered a biosignature [e.g. 78]. However, in our scenario, chemical disequilibrium is caused by (abiotic) CH4 release. Exoplanet analogs to our scenario would have a duration that would depend on solid-planet parameters, such as serpentinization efficiency, that are hard to determine for exoplanets. (2) A habitable, >500 Myr old planet suggests prolonged habitability; but in our scenario, the surface is sterilized just before the climate optimum. Therefore, our results underline the importance of planetary history for evaluating planetary habitability [11].

12

6. Conclusions.Mars’ climate optimum “appears to be fundamentally different than the fluvial environment characteristic of [still earlier times]” [5]. Since 1971, there has been extensive exploration of connections between mineralogy, atmospheric evolution, redox, and hydrology associated with the climate optimum on Mars. We examined a scenario in which the optimum is initiated by atmospheric collapse. The scenario anchors these connections and makes them precise (and predictive). Specifically, we infer that the valley networks of Mars are the consequence of an irreversible loss of reducing power from the shallow-subsurface of Mars, permitting a temporary (but widespread) integration of the surface and deep-subsurface hydrospheres, and explaining why Mars’ climate optimum also marks a global shift in sedimentation patterns and mineralogy. Although hypothetical and speculative, the scenario makes several novel testable predictions. The scenario cannot explain all wet climates in Mars history (Fig. 1), and previously proposed triggers for Mars’ climate optimum may supplement the scenario discussed here and are not inconsistent with it [e.g., 79].

For exoplanets as for Mars, both the loss of a secondary (outgassed) atmosphere, and atmospheric collapse, are usually considered to detract from habitability [e.g., 11,80-81]. However, in the collapse trigger scenario, these processes have the opposite effect. First, a >3 bar CO2 atmosphere, which represents a plausible initial condition for Mars, is invulnerable to runaway collapse and so would never see a climate optimum. Second, the collapse triggers the optimum. Therefore, in our scenario, both atmospheric loss and atmospheric collapse are prerequisites for the most habitable known extraterrestrial surface environment.

Materials and Methods.

In order to model collapse-triggered climate optima, we use simulated obliquity forcing to drive a model of surface temperature evolution (Fig. S1). The model resolves latitude and longitude and uses a GCM-derived parameterization of atmospheric collapse and re-inflation (Fig. S2). Changes in surface temperature and overburden pressure cause subsurface CH4-clathrate to decompose and outgas. Once outgassed, CH4 warms the surface and is destroyed by photochemical processes. Different assumptions about volatile escape-to-space can be incorporated by varying initial conditions, but once initialized, the total surface-exchangeable H2O and CO2 inventories are held constant for the duration of a model run (107-108 yr).

The total CH4 outgassed due to Mars atmospheric collapse and re-inflation can be decomposed into:

Prompt CH4 release from atmospheric collapse (reduction in PCO2 pressure, combined with planetwide cooling). This is always zero.

13

Minor CH4 release from H2O-ice unloading following atmospheric collapse (Step 4 in Fig. 2).

Major, re-inflation CH4 release: from warming during re-inflation (Steps 3-4 in Fig. 2).

CH4 release due to feedbacks (Steps 3-4 in Fig. 2).

Direct CH4 release is proportional to f, and feedback CH4 release increases nonlinearly with f.

Atmospheric Collapse and Surface Temperature Modeling.To calculate atmospheric collapse as a function of and φ PCO2 (Fig. S2), we used the Mars Weather Research and Forecasting General Circulation Model (MarsWRF GCM; refs. 17,82-83) to build a look-up table for Mars annual-mean surface temperatures as a function of , log of φ PCO2, and latitude, for a solar luminosity 75% of modern. H2O radiative effects are neglected. We used a polynomial fit to these results to determine the boundary of atmospheric collapse (coldest temperatures at the CO2 condensation point). We also investigated a -φ PCO2 phase portrait (Fig. S2) based on the results of the GCM of Ref. 84, finding generally more prolonged and intense climate optima (not shown).

Collapse rate is set to ~1 mbar/yr [28]. This pace is set by the ratio of the surface longwave radiation at the CO2-condensation temperature, to the product of density and latent heat for CO2 ice. We use the same rate for re-inflation [28]. We force the model with (t) from realistic obliquity histories, because is the dominant controlφ φ on polar insolation. Collapse takes ~1 kyr and is followed after a collapsed interval of 104-107 yr duration by ~1 kyr of re-inflation (Fig. 5, Fig. S8). We adopt φc = 15°, for which the (model-dependent) collapse-triggering PCO2 is 1.7 bar. The sequence of events is not sensitive to the choice of φc.

Clathrate modeling.Our scenario assumes the existence of shallow CH4-clathrate on Mars, with CH4

produced abiotically in association with deep water-rock reactions [34,41,46]. The P-T pathway forming CH4-clathrate is shown in Fig. S3.

Guided by the GCM output, we calculated T and P as a function of depth within the regolith, latitude, and longitude, using Tsurf from the GCMs, and assumptions about H2O ice distribution motivated by the water-cycle output of previously published GCMs [e.g., 17,29]. Spatial resolution is (2° latitude) × (2° longitude) × (1 m depth). Tsurf decreases with elevation [51], more so as PCO2 increases (Wordsworth 2016). The initial H2O ice distribution is set by allocating 1.4×107 km3 of H2O ice [52] uniformly across land with elevations >1 km (using modern topography). Where H2O ice rests on top of the regolith, we model the corresponding overpressure and thermal insulation (assuming thermal conductivity, kice = 2 W m-1 K-1). Destabilization of clathrate to depths >300 m requires annual-mean Tsurf > 273 K, for which Mars is

14

already in a climate optimum. Therefore, the grid is truncated at 300 m depth-within-regolith. For 300 m, the vertical thermal conduction timescale is 1 kyr – much shorter than the other timescales affecting atmospheric PCH4. In particular, the CH4 destruction timescale is 104-106 yr. This separation of timescales holds even when the low thermal diffusivity of CH4 clathrate and the latent heat of clathrate decomposition are considered. Therefore, we approximate geotherm adjustment as rapid.

Assuming rapid geotherm adjustment, we calculate the depth-to-HSZ (in meters below the top of the regolith, so a depth-to-HSZ of 100 m below an ice sheet of 200 m thickness means that clathrate is stable ≥300 m below the surface). We use the phase diagram of Ref. 85 (their Table 4.1), as shown in Fig. S3. We set porosity = 0.3, assume 120 kg CH4/(m3 clathrate), lithospheric heat flow Q = 0.03 W m-2, regolith thermal conductivity kreg = 2.5 W m-1 K-1; and regolith density ρreg = 2000 kg m-3. We consider both a case with H2O ice on high ground and a case without H2O ice on high ground. The resulting depth-to-HSZ look-up tables are passed to the main driver.

The main driver takes as input the depth-to-HSZ look-up tables, the phase portraits for atmospheric collapse from the GCMs (Fig. S2), a CH4 destruction look-up table (described below), a parameterization of greenhouse warming due to CH4 -CO2 Collision-Induced Absorption (CIA) [35], and a simulated obliquity time series obtained using the N-body code of Ref. 86 and an obliquity wrapper script [87]. For computational efficiency, we shift the obliquity time series up and down as needed to trigger collapse relatively early in the run (the Supplementary Information and Fig. S9 describe likely true wait times for atmospheric collapse). For a given atmospheric collapse phase portrait (Fig. S2), the selected value of φc defines the initial, pre-collapse, PCO2.

Fed by these inputs, the main driver first calculates the depth-to-HSZ for four key states: pre-collapse (H2O ice on highlands, inflated atmosphere – before Step 1 in Fig. 2); immediately post-collapse (H2O ice on highlands, collapsed atmosphere – end of Step 1 in Fig. 2); fully unloaded but collapsed (negligible H2O ice on highlands, collapsed atmosphere – end of Step 2 in Fig. 2); and immediately after re-inflation (negligible H2O ice on highlands, re-inflated atmosphere – Step 3 in Fig. 2). We then compute depth-to-HSZ following re-inflation for a range of intervals that are too short for complete unloading of H2O ice. We assume that unloading is steady and uniform (we explored nonuniform unloading specifications, but found only trivial differences). Next, based on a HSZ occupancy fraction, f, we compute the volume of CH4-clathrate that decomposes on CO2-atmosphere re-inflation for complete H2O-ice unloading of the highlands. Methane decomposition is rapid [60-61]. CH4-clathrate breakdown involves a >14% reduction in solid volume, and we assume fractures allow methane gas released at ≤100m depth to reach the surface in 10≪ 5 yr. The

15

greenhouse warming corresponding to this CH4 outgassing [35] is calculated. The corresponding maximum CH4 release, including CH4-warming-induced CH4 release, is obtained by iteration. For incomplete H2O ice unloading of the highlands, we linearize the feedback CH4 release. Linearization is invalid for extreme warming (see Fig. S6), but is reasonable for determining whether or not a climate optimum can occur.

The main time-stepping loop in the main driver uses the phase portraits in Fig. S2, as well as warming due to CH4 (if any), to track atmospheric collapse and re-inflation. Water ice net sublimation rate, i, is set to i = 0.15 mm/yr for migration from high ground to poles. This low value (based on MarsWRF GCM results) is due to the low vapor pressure of H2O in the ~200 K collapsed atmosphere. Movement of ice from poles back to high ground under the warm re-inflated atmosphere should be faster: we use i = -1.5 mm/yr. Our results are qualitatively unaffected by reasonable changes in i. The overpressure due to polar ice caps is not explicitly modeled. This is acceptable because polar cap area is small. For collapse duration tΔ ≥ (~300 m)/i, the H2O ice is completely removed from the highlands. If tΔ < (~300 m)/i, we use a look-up table to find the partial CH4 release. We also track release of CH4 from previous re-inflations and collapses to ensure that the same CH4 is not released twice (Fig. S4). Finally, we track inhibition of orbitally paced atmospheric collapses by CH4 warming (Fig. 5). (This effect is analogous to the delay in the onset of the next terrestrial ice age due to industrial release of CO2).

Methane destruction parameterisation. We used the Caltech/JPL 1-D Mars photochemistry code, modified to include CH4 (and C2H6, and C2H2, etc.) [88-90], to build a look-up table for CH4 destruction by photochemical processes. The CH4 destruction model is the same as that in ref. 34, except for the use of a UV flux appropriate for the Sun 3.8 Ga [91]. Boundary conditions include surface burial of O2, O3, H2O2, and CO. H2O is set to saturation at the surface; results are insensitive to H2O concentration. As in ref. 34, we found that the main control on CH4 destruction rate is the CH4/CO2 ratio, so we fit a curve to the CH4 destruction rate as a function of the CH4/CO2 ratio for use in the climate-evolution model (Fig. S7).

Acknowledgements. We thank R. Wordsworth for sharing model output. We thank A. Howard, C. Goldblatt, R.P. Irwin, R. Craddock, A. Soto, J.C. Armstrong, F. Tian, I. Halevy, T. McCollom, and C. Oze for discussions. Grants: NASA (NNX16AG55G).

16

References.

1. Irwin, R. P., III; Howard, A. D.; and Craddock, R. A. (2008), Fluvial valley networks on Mars, in River Confluences, Tributaries, and the Fluvial Network, edited by S. Rice, A. Roy, and B. Rhoads, pp. 409–430, John Wiley, West Sussex, U.K

2. Fassett, C. I. and J. W. Head (2008) Valley network-fed, open-basin lakes on Mars: Distribution and implications for Noachian surface and subsurface hydrology. Icarus, 198, 37-56.

3. Matsubara, Y.; Howard, A.D.; Drummond, S.A. (2011) Hydrology of early Mars: Lake basins, J. Geophys. Res., 116, E4, CiteID E04001.

4. Goudge T.A.; Fassett, C.I.; Head, J.W.; Mustard, J.F.; Aureli, K.L. (2016) Insights into surface runoff on early Mars from paleolake basin morphology and stratigraphy, Geology 44, 419-422

5. Howard, A. D.; Moore, J. M.; and Irwin, R. P. (2005) An intense terminal epoch of widespread fluvial activity on early Mars: 1. Valley network incision and associated deposits. J. Geophys. Res. (Planets), 110:E12S14, 2005. doi: 10.1029/2005JE002459.

6. Irwin, R. P.; Tanaka, K. L.; and Robbins, S. J. (2013). Distribution of Early, Middle, and LateNoachian cratered surfaces in the Martian highlands: Implications for resurfacing events and processes. J. Geophys. Res. (Planets) 118, 278-291.

7. Zabrusky, K.; Andrews-Hanna, J.C.; Wiseman, S. (2012), Reconstructing the distribution and depositional history of the sedimentary deposits of Arabia Terra, Mars, Icarus 220, 311-330

6. Irwin, R. P.; Tanaka, K. L.; Robbins, S. J. (2013). Distribution of Early, Middle, and LateNoachian cratered surfaces in the Martian highlands: Implications for resurfacing events and processes. J. Geophys. Res. (Planets) 118, 278-291.

8. Barnhart, C.J.; ;Howard, A.D.; Moore, J.M (2009) Long-term precipitation and late-stage valley network formation: Landform simulations of Parana Basin, Mars. J. Geophys. Res. (Planets) 114, E01003.

9. Hoke, M.R.T.; Hynek, B.M.; Tucker, G.E. (2011) Formation timescales of large Martian valley networks, Earth Planet Sci. Lett. 312, 1-12

10. Kasting, JF, & D Catling (2003) Evolution of a habitable planet, Annual Review of Astronomy and Astrophysics 41 (1), 429-463

11. Catling, D.C. & Kasting, J.F. (2017) Atmospheric evolution on inhabited and lifeless worlds, Cambridge University Press.

12. Irwin, R. P., A. D. Howard, R. A. Craddock, and J. M. Moore 2005. An intense terminal epoch of widespread fluvial activity on early Mars: 2. Increased runoff and paleolake development. J. Geophys. Res. (Planets) 110, E12S15.

13. Jakosky, B. M. et al. (2017), Mars’ atmospheric history derived from upper-atmosphere measurements of 38Ar/36Ar, Science 355(6332), 1408-1410.

17

14. Lammer, H., E. et al. (2013) Outgassing History and Escape of the Martian Atmosphere and Water Inventory, Space Sci. Rev., 174, 113–154

15. Tian, F., Kasting, J.F., & Solomon, S.C. (2009) Thermal escape of carbon from the early Martian atmosphere, Geophys. Res. Lett., DOI: 10.1029/2008GL036513

16. Wordsworth, R. (2016), The Climate of Early Mars, Annual Reviews of Earth and Planetary Sciences, 44, 381-408, https://doi.org/10.1146/annurev-earth-060115-012355

17. Mischna, M.A., Baker, V., Milliken, R., Richardson, M., Lee, C. (2013) Effects of obliquity and water vapor/trace gas greenhouses in the early martian climate. J. Geophys. Res. (Planets) 118, 560-576. doi:10.1002/jgre.20054.

18. Robbins, S. J., B. M. Hynek, R. J. Lillis, and W. F. Bottke (2013) Large impact crater histories of Mars: The effect of different model crater age techniques. Icarus 225, 173-184.

19. Halevy, I., Head, J.W., III (2014) Episodic warming of early Mars by punctuated volcanism. Nat. Geosci. 7, 865-868.

20. Kerber, L., Forget, F., Wordsworth, R (2015). Sulfur in the early martian atmosphere revisited: experiments with a 3-D Global Climate Model. Icarus 261, 133-148.

21. Matsuyama, I., & M. Manga (2010), Mars without the equilibrium rotational figure, Tharsis, and the remnant rotational figure, J. Geophys. Res.: Planets 115 (E12)

22. Batalha, N.E.; Kopparapu, R.K.; Haqq-Misra, J.; Kasting, J.F. (2016), Climate cycling on early Mars caused by the carbonate-silicate cycle, Earth Planet Sci. Lett., 455, 7-13

23. Edwards, C.S.; Ehlmann, B.L. (2015) Carbon sequestration on Mars. Geology 43, 863-866

24. Hynek, B., The great climate paradox of ancient Mars, Geology, 44, 879-880 (2016).

25. Luo, W., Cang, X. & A.D. Howard, (2017) New Martian valley network volume estimate consistent with ancient ocean and warm and wet climate, Nat. Communications 8, 15766

26. Fassett, C.I., Head, J.W. (2008). The timing of martian valley network activity: Constraints from buffered crater counting. Icarus 195, 61-89.

27. Kite, E.S., Sneed, J., Mayer, D.P. & Wilson, S.A. (2017), Persistent or repeated surface habitability on Mars during the Late Hesperian - Amazonian, Geophys. Res. Lett., doi:10.1002/2017GL072660.

28. Soto, A.; Mischna, M.; Schneider, T.; Lee, C.; Richardson, M.I. (2015), Martian atmospheric collapse: Idealized GCM studies, Icarus, 250, 553-569.

29. Wordsworth, R.; Kerber, L.; Pierrehumbert, R.; Forget, F.; Head, J.W. (2015), Comparison of warm and wet and cold and icy scenarios for early Mars in a 3-D climate model, JGR: Planets, 120, 1201-1219.

30. Tobie, G., Lunine, J.I. and Sotin, C. (2006), Episodic outgassing as the origin of atmospheric methane on Titan, Nature 440, 61-64.

31. Bowen, GJ, BJ Maibauer, MJ Kraus, U Röhl, T Westerhold, A Steimke (2016) Two massive, rapid releases of carbon during the onset of the Palaeocene-Eocene thermal maximum, Nat. Geosci. 8, 44-47.

18

32. Sagan, C. (1977), Reducing greenhouses and the temperature history of Earth and Mars, Nature 269, 224 - 226; doi:10.1038/269224a0

33. Chassefière, E., Lasue, J., Langlais, B., Quesnel, Y. (2016) Early Mars Serpentinization Derived CH4 Reservoirs and H2 Induced Warming. Meteoritics & Planet. Sci., 51, 2234–2245.

34. Kite, E.S., Gao P., Goldblatt, C., Mischna M., Mayer D., and Yung Y.L., Methane bursts as a trigger for intermittent lake-forming climates on post-Noachian Mars, accepted by Nat. Geosci..

35. Wordsworth, R., et al (2017) Transient reducing greenhouse warming on early Mars, Geophys. Res. Lett. 44, 665-671.

36. Carter, J.; Poulet, F.; Bibring, J.-P.; Mangold, N.; Murchie, S. (2013) Hydrous minerals on Mars as seen by the CRISM and OMEGA imaging spectrometers: Updated global view, J. Geophys. Res.: Planets, 118, 831-858.

37. Ehlmann, B.L., J. F. Mustard, and S. L. Murchie (2010) Geologic setting of serpentine deposits on Mars. Geophys. Res. Lett. , 370:L06201, 2010. doi: 10.1029/2010GL042596.

38. Parmentier, E.M., Zuber, M.T. (2007) Early evolution of Mars with mantle compositional stratification or hydrothermal crustal cooling. JGR (Planets) 112, E02007.

39. Sun, V.Z.; Milliken, R.E. (2015) Ancient and recent clay formation on Mars as revealed from a global survey of hydrous minerals in crater central peaks, J. Geophys. Res. Planets, 120, 2293-2332.

40. Lyons, J.R.; Manning, C.; Nimmo, F. (2005) Formation of methane on Mars by fluid-rock interaction in the crust, Geophys. Res. Lett. 32, CiteID L13201.

41. Etiope, G., Sherwood Lollar, B. (2013) Abiotic Methane on Earth. Rev. Geophys. 51, 276-299.

42. Ehlmann, B. L., et al. (2011) Subsurface water and clay mineral formation during the early history of Mars, Nature, 479(7371), 53–60.

43. Fassett, C.I., and J. W. Head (2011) Sequence and timing of conditions on early Mars. Icarus 211, 1204-1214, doi: 10.1016/j.icarus.2010.11.014.

44. Chastain, B.K.; Chevrier, V. (2007) Methane clathrate hydrates as a potential source for martian atmospheric methane, Planetary and Space Science 55, 1246-1256.

45. Chassefière, E., & Leblanc, F. (2011) Methane release and the carbon cycle on Mars, Planetary & Space Sci. 59, 207-217

46. Mousis, O., et al. (2013) Volatile Trapping in Martian Clathrates. Space Sci. Rev. 174, 213-250.

47. Prieto-Ballesteros, O., et al. (2006) Interglacial clathrate destabilization on Mars: Possible contributing source of its atmospheric methane. Geology 34, 149.

48. Skinner, J.A., & Tanaka, K.L. (2007) Evidence for and implications of sedimentary diapirism and mud volcanism in the southern Utopia highland–lowland boundary plain, Mars, Icarus, 186, 41-59.

19

49. Kite, E.S., Williams, J.-P. , Lucas, A. , Aharonson, O. (2014) Low palaeopressure of the martian atmosphere estimated from the size distribution of ancient craters. Nat. Geosci. 7, 335–339.

50. Kurokawa, H., Kurosawa, K., and Usui, T. (2017) A lower limit of atmospheric pressure on Early Mars inferred from nitrogen and argon isotopic compositions, Icarus, doi:10.106/j.icarus.2017.08.020

51. Fastook, J. L., and J. W. Head III (2015) Glaciation in the Late Noachian Icy Highlands: Ice accumulation, distribution, flow rates, basal melting, and top-down melting rates and patterns, Planet. Space Sci., 106, 82-98, doi: 10.1016/j.pss.2014.11.028.

52. Mahaffy, P.R., et al. (2015) The imprint of atmospheric evolution in the D/H of Hesperian clay minerals on Mars. Science 347, 412-414.

53. Laskar, J., et al. (2004), Long term evolution and chaotic diffusion of the insolation quantities of Mars, Icarus, 170, 343–364.

54. Kite, E. S., I. Halevy, M. A. Kahre, M. Manga, and M. Wolff (2013) Seasonal melting and the formation of sedimentary rocks on Mars, Icarus, 223, 181–210.

55. Sagan, C.; Toon, O. B., Gierasch, P. (1973) Climatic Change on Mars, Science, 181, 1045-1049.

56. Mellon, M.T. (1996) Limits on the CO2 Content of the Martian Polar Deposits, Icarus 124, 268-279.

57. Kreslavsky, M. A.; Head, J. W. (2005) Mars at very low obliquity: Atmospheric collapse and the fate of volatiles, Geophys. Res. Lett., 32, 12, CiteID L12202

58. Lapôtre, M. G. A., et al. (2016), Large wind ripples on Mars: A record of atmospheric evolution, Science, 353, 55–58.

59. Hecht, M.H. (2002) Metastability of Liquid Water on Mars. Icarus, 156, 373-386, doi: 10.1006/icar.2001.6794.

60. Stern, L.A., Circone, S., Kirby, S.H., Durham, W.B. (2003) Temperature, pressure, and compositional effects on anomalous or ''self'' preservation of gas hydrates. Can. J. Phys. 81, 271-283.

61. Gainey, S. R.; Elwood Madden, M. E. (2012) Kinetics of methane clathrate formation and dissociation under Mars relevant conditions, Icarus 218, 513-524.

62. Andreassen, K., et al. (2017), Massive blow-out craters formed by hydrate-controlled methane expulsion from the Arctic seafloor, Science, 356, 6341, 948-953, DOI: 10.1126/science.aal4500

63. Krasnopolsky, V.A., Maillard, J.P., Owen, T.C. (2004) Detection of methane in the martian atmosphere: evidence for life? Icarus 172, 537-547.

64. McKay, C. P.; Wharton, R. A., Jr.; Squyres, S. W.; Clow, G. D. (1985) Thickness of ice on perennially frozen lakes, Nature 313, Feb. 14, 1985, p. 561, 562.

65. Doran, P.T., et al. (2002) Valley floor climate observations from the McMurdo dry valleys, Antarctica, 1986-2000, J. Geophys. Res. Atmospheres, 107, D24, ACL 13-1, CiteID 4772, DOI 10.1029/2001JD002045.

20

66. Dethier, D.P. (2001) Pleistocene incision rates in the western United States calibrated using Lava Creek B tephra, Geology, 29, 783-786.

67. Olsen, A.A., & Rimstidt, J.D. (2007) Using a mineral lifetime diagram to evaluate the persistence of olivine on Mars, Am. Mineral., 92, 4, 598–602.

68. Tosca, N.J., Knoll, A.H. (2009) Juvenile chemical sediments and the long term persistence of water at the surface of Mars. Earth Planet. Sci. Lett. 286, 379-386.

69. Hynek, B.M.; Beach, M.; Hoke, M.R. T. (2010) Updated global map of Martian valley networks and implications for climate and hydrologic processes, J. Geophys. Res., 115(E9), CiteID E09008.

70. McKay, C.P.; Pollack, J.B.; Courtin, R. (1991) The greenhouse and antigreenhouse effects on Titan, Science 253, 1118-1121.

71. Halevy, I.; Fischer, W. W.; Eiler, J. M (2011) Carbonates in the Martian meteorite Allan Hills 84001 formed at 18±4°C in a near-surface aqueous environment, Proc. Natl. Acad. Sci., 108, 16895-16899

72. Carter, J., D. Loizeau, N. Mangold, F. Poulet, J.-P. Bibring (2015) Widespread surface weathering on early Mars: A case for a warmer and wetter climate, Icarus, 248, 373–382.

73. Wilson, E.H., S.K. Atreya, R.I. Kaiser, P.R. Mahaffy (2016) Perchlorate formation on Mars through surface radiolysis-initiated atmospheric chemistry: A potential mechanism, J. Geophys. Res. – Planets, DOI: 10.1002/2016JE005078

74. Bibring, J.-P., et al. (2006), Global mineralogical and aqueous Mars history derived from OMEGA/Mars express data, Science, 312(5772), 400–404.

75. Gaillard, F.; Michalski, J.; Berger, G.; McLennan, S.M.; Scaillet, B. (2013) Geochemical Reservoirs and Timing of Sulfur Cycling on Mars, Space Sci. Rev., 174, 1-4, 251-300.

76. Summons, R.E., et al. (2011) Preservation of martian organic and environmental records: Final report of the Mars biosignature working group, Astrobiology, 11(2): doi:10.1089/ast.2010.0506.

77. Hassler, D.M., et al. (2014) Mars’ surface radiation environment measured with the Mars science laboratory’s curiosity rover. Science 343, 1244797 .

78. Krissansen-Totton, J., D. Bergsman, D. C. Catling (2016), On detecting biospheres from chemical thermodynamic disequilibrium in planetary atmospheres, Astrobiology 16, 39-67.

79. Maclennan, J., M. Jull, D. McKenzie, L. Slater, K. Grönvold (2002), The link between volcanism and deglaciation in Iceland, Geochem. Geophys., Geosystems 3, 1–25.

80. Heng, K., & P. Kopparla (2012) On the stability of Super-Earth atmospheres, Astrophys. J. 754, Number 1.

81. Zahnle, K. & Catling, D. (2017) The Cosmic Shoreline: The Evidence that Escape Determines which Planets Have Atmospheres, and what this May Mean for Proxima Centauri B, Astrophys. J., 843:122 (23pp).

82. Richardson, M.I., Toigo, A.D., Newman, C.E. (2007) PlanetWRF: a general purpose, localto global numerical model for planetary atmospheric and climate dynamics. J.Geophys. Res. 112, E09001. http://dx.doi.org/10.1020/2006JE002825.

21

83. Toigo, A.D., Lee, C., Newman, C.E., Richardson, M.I. (2012) The impact of resolution onthe dynamics of the martian global atmosphere: varying resolution studies with theMarsWRF GCM. Icarus 221, 276–288.

84. Forget, F.; et al. (2013), 3D modelling of the early martian climate under a denser CO2 atmosphere: Temperatures and CO2 ice clouds, Icarus 222(1), p. 81-99.

85. Sloan, E.D., & Koh, C.A. (2008) Clathrate Hydrates of Natural Gases (3rd Edition), CRC Press.

86. Chambers, J.E. (1999), A hybrid symplectic integrator that permits close encounters between massive bodies. Monthly Notices Royal Astron. Soc. 304, 793-799.

87. Armstrong, J.C., Leovy, C.B., Quinn, T. (2004), A 1 Gyr climate model for Mars: new orbital statistics and the importance of seasonally resolved polar processes. Icarus 171, 255-271.

88. Summers, M.E.; Lieb, B. J.; Chapman, E.; Yung, Y.L. (2002) Atmospheric biomarkers of subsurface life on Mars, Geophys. Res. Lett. 29, 24-1, CiteID 2171.

89. Nair, H.; Allen, M.; Anbar, A.D.; Yung, Y.L.; Clancy, R.T. (1994) A photochemical model of the martian atmosphere, Icarus 111, 124-150.

90. Nair, H.; Summers, M.E.; Miller, C.E.; Yung, Y.L. (2005) Isotopic fractionation of methane in the martian atmosphere, Icarus 175, 32-35.

91. Claire, M.W, et al. (2012) The Evolution of Solar Flux from 0.1 nm to 160 m, μ Astrophys. J. 757, article id. 95, 12 pp.

22

Supplementary Information.

Geologic data.Valley incision represents 0.4 m – 4 m of area-averaged erosion: only 104-105 yr of erosion at Earth rates (Hoke et al. 2011, Luo et al. 2017). Modest erosion is also suggested by incomplete grading of valley long-profiles (Aharonson et al. 2002, Som et al. 2009). The rarity of delta entrenchment suggests abrupt shutdown of the climate optimum (Irwin et al. 2005). Column runoff likely totaled >5-100 km (Luo et al. 2017, but see Rosenberg & Head 2015). Climate physics predicts global synchroneity of wet conditions, broadly consistent with valley-network crater counts (Fassett & Head 2008b; see Hoke & Hynek 2009 for another view). Overall, the challenge to modelers is to explain a ~3.7 Ga wet climate that persisted for >104 yr but was geologically brief.

The Mars climate optimum appears to be unique in the geomorphic record (Fig. 1), consistent with the collapse trigger scenario (Fig. S9). We cannot definitively exclude the possibility that valleys formed earlier, but were buried. However, there is no evidence for this, and there is clear evidence for a major change in hydrology at the climate optimum (Irwin et al. 2013, Goudge et al. 2016). Summary of previous studies of Mars atmospheric collapse.Previous work on Mars atmospheric collapse includes proposals for geologically recent or near-future climate change driven by atmospheric re-inflation (Sagan 1973, Gierasch & Toon 1973, McKay et al. 1991). This is ruled out by new upper limits on the present-day inventory of CO2 ice (Bierson et al. 2016). Nakamura & Tajika (2001, 2002) show that the extent and seasonality of CO2 ice-sheets can be important for collapse. Nakamura & Tajika (2003) study the atmospheric collapse hysteresis loop as a function of CO2 inventory and solar luminosity. Using a 1D model, they find that the maximum obliquity for a collapsed atmosphere is currently ~32°, but due to reduced solar luminosity this would have been higher for the ancient climate. They conclude that the ancient climate “might have been rather unstable due to the repeated climate jumps” (Nakamura & Tajika, 2003). Kahre et al. (2013) assess the importance of carbon dioxide ice cap albedo and emissivity in setting the boundaries of the atmospheric-collapse zone. Manning et al. (2006) considered atmospheric collapse as part of a study of the long-term evolution of the Mars C inventory. The most recent analysis of atmospheric collapse is by Soto et al. (2015). In their study, collapse onto Olympus Mons plays a key role, but this volcano was probably not very tall at 3.7 Ga (Isherwood et al. 2013). Recently, radar data have provided direct evidence of atmospheric collapse <<10 Myr ago on Mars (Bierson et al. 2016).

23

Parameter sensitivities.Consistent with the results of all published Mars ancient-water-cycle GCMs, the scenario assumes that collapse occurs only for <φ 40°. Otherwise, no major H2O-ice shift will occur at collapse (Fig. S8). The collapse trigger atmospheric pressure also has to be 1 bar, otherwise surface warming from ≳ CH4 is weak or net-negative (Fig. S8; Wordsworth et al. 2017).

CH4 release increases strongly with initial T, because of the nonlinearity of the CH4-clathrate decomposition curve (Fig. S3). The scenario is qualitatively insensitive to reasonable variations of the total H2O ice inventory, to the value of the obliquity at collapse, to possible changes in Mars’ precession constant over time, or to obliquity-oblateness feedbacks.

Calculation of runaways.For the calculations underlying Fig. S5, we find the runaway clathrate breakdown threshold by comparing direct CH4-induced warming to the feedback warming (induced by the CH4 release driven by direct warming). If the indirect, feedback warming exceeds the direct warming, then the system will run away. Otherwise the asymptotic warming gain (G) is given approximately by

G = 1/(1-R), (1)

where R is the feedback factor (Roe 2009). These calculations linearize the feedback. For the calculations underlying Fig. S6, nonlinear feedbacks are also included.

Long-term evolution.The size of CH4 burst is affected by HSZ outgassing prior to the climate optimum. Pre-optimum outgassing can occur due to obliquity variations (Kite et al. 2017), or as the result of ice unloading during pre-optimum atmospheric collapses. We used an ensemble of long-termobliquity simulations to explore these effects. In these calculations, we made the simplifying assumption that CH4 release was proportional to the volume of H2O ice removed, but with CH4 release suppressed from depths that had previously been degassed. This simplifying assumption allowed us to carry out >30 long-term simulations. We determined the fraction of runs for which at least 100 m of H2O-ice first-time unloading occurs during a single atmospheric collapse interval, and the dependence of this fraction on φr and the ice sublimation rate. The 100 m first-time ice-unloading value was chosen as one for which a climate optimum is likely provided that the sub-ice regolith is charged with CH4-clathrate. The "first-time" qualifier refers to the high probability that the CH4-clathrate reservoir is not recharged, meaning that repeated ice unloading will only yield CH4 gas from a given latitude-longitude-depth voxel the first time that voxel is unloaded. We found that the fraction of “successful” runs increases steeply with the size of the hysteresis loop, but is insensitive to the ice unloading rate.

24

Timing of the optimum. The collapse trigger hypothesis suggests a physical basis for the timing of valley network formation (Fig. S9). Valley network formation occurred >0.5 Gyr after solar system formation, based on crater counts (Hartmann & Neukum 2001). This does not match impact nor volcanism forcings, but there is indirect evidence that atmospheric collapse began around this time (Lapôtre et al. 2016, Kurosawa et al. 2017). In the collapse trigger scenario, valley network formation is triggered <108 yr after Mars’ first atmospheric collapse, which in turn is caused by the interplay between atmospheric loss and chaotic obliquity (Fig. S9). Using >100 realistic obliquity trajectories, constant solar luminosity and CO2 inventory, and assuming the hysteresis loops from Fig. S2, we find that CO2 inventories >0.8 bars can match the data for the Forget et al. (2013) GCM output; for the Mischna MarsWRF GCM output, ≥2 bar CO2 inventory is favored (Fig. S9). Our calculations assume that the basic orbital architecture of the Solar System at ~3.7 Ga was similar to today (see Brasser & Walsh 2011 for an alternative view). Folding in estimates of Mars atmospheric pressure over time (Lammer et al. 2013), a wait of 108-109 yr is expected, consistent with data, although there are large uncertainties.

Supplementary-only references

Aharonson, O.; Zuber, M.T.; Rothman, D.H.; Schorghofer, N.; Whipple, K.X. (2002) Drainage basins and channel incision on Mars, Proc. Natl. Acad. Sci. 99, 1780-1783.

Bierson, C. J.; Phillips, R. J.; Smith, I. B.; Wood, S. E.; Putzig, N. E.; Nunes, D.; Byrne, S. (2016) Stratigraphy and evolution of the buried CO2 deposit in the Martian south polar cap, Geophys. Res. Lett., 43, 9, 4172-4179

Brasser, R.; Walsh, K.J. (2011) Stability analysis of the martian obliquity during the Noachian era, Icarus 213, 423-427.

Gierasch, P. J.; Toon, O. B. (1973) Atmospheric pressure variation and the climate of Mars. J. Atmos. Sci., 30, 1502 – 1508.

Brasser, R.; Walsh, K.J. (2011) Stability analysis of the martian obliquity during the Noachian era, Icarus 213, 423-427.

Hartmann, W.K.; Neukum, G. (2001) Cratering Chronology and the Evolution of Mars, Space Sci. Rev. 96, 1/4, p. 165-194.

Hoke, M.R.T. & Hynek, B.M. (2009) Roaming zones of precipitation on ancient Mars as recorded in valley networks, J. Geophys. Res. 114(E8), CiteID E08002.

Isherwood, R.J.; Jozwiak, L.M.; Jansen, J.C.; Andrews-Hanna, J.C. (2013), The volcanic history of Olympus Mons from paleo-topography and flexural modeling, Earth Planet Sci. Lett., 363, 88-96.

25

Kahre, M.A., Vines, S.K., Haberle, R.M., Hollingsworth, J.L. (2013), The early Martian atmosphere: Investigating the role of the dust cycle in the possible maintenance of two stable climate states, J. Geophys. Res. – Planets, DOI: 10.1002/jgre.20099

Manning, C.V., McKay, C.P., Zahnle, K.J. (2006) Thick and thin models of the evolutionof carbon dioxide on Mars. Icarus 180, 38–59.

McKay, C.P.; Toon, O.B.; Kasting, J.F. (1991) Making Mars habitable, Nature 352, 489-496.

Nakamura, T.; Tajika, E. (2001) Stability and evolution of the climate system of Mars, Earth, Planets and Space, 53, 851-859.

Nakamura, T.; Tajika, E., (2002) Stability of the Martian climate system under the seasonal change condition of solar radiation, J. Geophys. Res. (Planets), 107(E11), 4-1, CiteID 5094, DOI 10.1029/2001JE001561

Nakamura, T.; Tajika, E. (2003) Climate change of Mars-like planets due to obliquity variations: implications for Mars, Geophys. Res. Lett., 30, 18-1, CiteID 1685, DOI 10.1029/2002GL016725.

Roe, G. (2009) Feedbacks, timescales, and seeing red, Ann. Rev. Earth Planet. Sci. 37, 93-115

Rosenberg, E.N.; Head, J.W., III (2015) Late Noachian fluvial erosion on Mars: Cumulative water volumes required to carve the valley networks and grain size of bed-sediment, Planet. Space Sci. 117, 429-435.

Som, Sanjoy M.; Montgomery, David R.; Greenberg, Harvey M. (2009) Scaling relations for large Martian valleys, J. Geophys. Res. - Planets 114(E2), CiteID E02005. Supplementary figures.

Fig. S1. Schematic of the model.

26

Fig. S2. Atmospheric-collapse phase portraits. The hysteresis loops that are highlighted by thick lines are examples of physically realizable climate system trajectories. The thin lines correspond to unstable parts of the polar-CO2-condensation curve. MarsWRF GCM results shown in orange. Loop fit to output of the GCM of Forget et al. (2013) shown in yellow.

27

Fig. S3. CH4 clathrate phase diagram. Phase boundaries shown in dark blue. Mars geotherms shown in red (early, steep geotherm) and cyan (later, shallow geotherm). Early in Mars history, cooling of the geotherm locks-in CH4 as clathrate in regolith beneath ice sheets (gray arrow). Further geotherm cooling and escape of ice-sheet H2O to space has little effect on CH4-clathrate stability. Atmospheric collapse and consequent unloading cause minor CH4-clathrate breakdown (Fig. 2). Warming of the surface upon re-inflation, plus feedback warming, cause major CH4-clathrate breakdown.

Fig. S4. Methane inventory history for one run with f = 0.03.

28

Fig. S5. Gain map for a single-column model, assuming initial surface temperature of 240 K. Gain is defined in Eqn. 1.

29

Fig. S6. Showing potential for runaway clathrate outgassing on Mars due to CH4 release from an f = 0.045 clathrate reservoir. Colored dashed lines correspond to CH4-induced warming (Wordsworth et al. 2017). Solid lines show the corresponding CH4 release. The arrows labeled 1-3 give an example of how to read the diagram. Supposing (1) an initial collapsed-triggered release of 15 mbar CH4 in a 1500 mbar pCO2 atmosphere, the warming (2) is ~10K. For an initial surface temperature Tinit = 220K, this is sufficient to release (3) a further 9.5 mbar of CH4, which will lead to further warming and thus more CH4 release. When the warming for a given PCO2 plots above a CH4 release line, runaway outgassing occurs. When the warming for a given PCO2 plots below a CH4 release line, there is no runaway, but positive feedback still occurs. Circles highlight selected stable equilibria that terminate the runaway. These results are for a single column of clathrate-charged regolith, but are representative of the full-model behavior.

30

Fig. S7. CH4 destruction rate.

Fig. S8. Sketch to show that simulations that incorporate CO2-ice clouds predict the obliquity at first-collapse is less than 40°, and the pressure at first-collapse is >500 mbar, consistent with a collapse-triggered climate optimum.

31

Fig. S9. Effect of chaotic diffusion of obliquity on timing of valley network formation. Gray bar corresponds to timing of VNs. The lines show the spread of outcomes for >10 randomly initialized obliquity simulations. Symbols correspond to the median wait time for initial atmospheric collapse for initial CO2 inventories of (from left to right) {200, 500, 1000, 2000} mbar, for different models of polar temperature (asterisks: Forget et al. 2013 GCM; circles, Mischna / MarsWRF GCM; symbols are vertically offset for clarity).

32