Oxygen as a control over 2.4 billion years of Earth's atmospheric evolution
Gregory Cooke, Dan Marsh, Catherine Walsh, Benjamin Black, Jean-François Lamarque
aa r X i v : . [ a s t r o - ph . E P ] F e b Oxygen as a control over 2.4 billion years of Earth’s atmosphericevolution
G. J. Cooke ∗ , D. R. Marsh , † , C. Walsh ‡ , B. A. Black , J.-F. Lamarque School of Physics and Astronomy, University of Leeds, Leeds LS2 9JT, UK National Center for Atmospheric Research, Boulder, CO 80301, USA Department of Earth and Atmospheric Sciences, The City College of New York, NewYork, NY, USAFebruary 23 rd , 2021 Since the formation of the terrestrial planets, atmospheric loss has irreversibly altered their at-mospheres, leading to remarkably different surface environments - Earth has remained habitablewhile Venus and Mars are apparently desolate . The concept of habitability centres around theavailability of liquid water which depends greatly on the composition of the atmosphere . Whilethe history of molecular oxygen (O ) in Earth’s atmosphere is debated , geological evidence sup-ports at least two major episodes of increasing oxygenation: the Great Oxidation Event and theNeoproterozoic Oxidation Event. Both are thought to have been pivotal for the development andevolution of life . We demonstrate through three-dimensional simulations that atmospheric O concentrations on Earth directly control the evolution and distribution of greenhouse gases (suchas O , H O, CH and CO ) and the atmospheric temperature structure. In particular, at ≤ %the present atmospheric level (PAL) of O , the stratosphere collapses. Our simulations show that abiologically ineffective ozone shield, lower than previously thought, existed during the Proterozoic, ∗ E-mail: [email protected] † E-mail: [email protected] ‡ E-mail: [email protected] acts as a valve for the loss rate of atmospheric hydrogen through the exosphere. Estimatedlevels of hydrogen escape for the Proterozoic eon are all lower than present day, enabling us to es-tablish Earth’s water loss timeline. Furthermore, we demonstrate how O on terrestrial exoplanetsdetermines their theoretical transmission spectra, challenging signal-to-nose ratio assumptions con-tributing to the design of next generation telescopes that will facilitate the characterisation ofEarth-like worlds. As most atmospheric O is produced from life, and Phanerozoic O is critical tosustaining stratospheric O and thus UV-shielding terrestrial habitats , O is a key mechanismby which biology can reshape planetary habitability. Earth’s atmospheric evolution
It is thought that Earth began with a primordial atmosphere with a composition that was repre-sentative of the protoplanetary disk around the Sun . This primordial atmosphere escaped tospace and was replaced by a secondary anoxic atmosphere with nitrogen, carbon dioxide and water,as well as other gases in trace amounts .As molecular-oxygen-producing life evolved from the first cyanobacteria ∼ , theEarth’s atmosphere and ocean became progressively more oxygenated, a transition with at leasttwo major shifts towards higher atmospheric oxygen: the Great Oxidation Event ∼ . Gyr ago and a second rise in atmospheric oxygen to near-modern levels at the close of the Precambrian,known as the Neoproterozoic Oxidation Event . During the Proterozoic eon, which continued forover a billion years, O values are estimated to have been between 0.001% and 1% the presentatmospheric level (PAL) ; O now comprises 21% of the modern atmosphere.In this work, we show how Earth’s atmospheric loss, temperature structure, greenhouse gasquantities, ozone shield, and Earth-like transmission spectra, are regulated through O concentra-tions, by simulating Earth’s atmosphere up to the lower thermosphere with a three-dimensional(3D) model employing fully coupled chemistry and physics.This study uses version 6 of the Whole Atmosphere Community Climate Model (WACCM6) to produce five different simulations - see Extended Data Table 1. We simulate an 1850 pre-2ndustrial atmosphere (hereafter PI) that does not have present day pollutants and greenhouse gasconcentrations and then we vary the mixing ratio for O to potential Proterozoic levels of 1% PAL,0.5% PAL and 0.1% PAL. The fifth simulation (hereafter YS) assumes O levels of 1% PAL and usesa theoretical spectrum of the Sun 2 billion years ago to investigate the impact of a less luminousyounger Sun. This young Sun’s modelled total energy output was 13% less than the present Sun,with a stronger extreme ultraviolet flux by a factor of 2.97. We will show how this young Sunreduces hydrogen escape by cooling the tropopause. Vanishing ozone during the rise of Eukaryotes
The stratospheric ozone (O ) layer strongly absorbs ultraviolet wavelengths of incoming solar ra-diation, thereby warming Earth’s stratosphere and providing surface biota with a critical shieldagainst ultraviolet radiation. Our simulations show that a reduction in O to Neoproterozoic levelsfundamentally alters the atmospheric structure. The primary mechanism behind these changes is adramatic reduction in O and consequent radiative heating in the middle atmosphere.The maximum O volume mixing ratio in the 0.1% PAL simulation is 38.6 times lower thanin the PI simulation. These reduced levels of O cause less heating in the middle atmosphere (seeFig. 1 and Fig. 2), with the modern stratosphere effectively disappearing. In the PI simulation, −90 −60 −30 0 30 60 90Latitude [ ∘ ]10 −1 −2 P r e ss u r e [ h P a ] . K . K . K . K . K . K . K . K PI (100∘ PAL) −90 −60 −30 0 30 60 90La i ude [ ∘ ] . K . K . K . K . K . K
1∘ PAL −90 −60 −30 0 30 60 90La i ude [ ∘ ] . K . K . K . K . K . K . K . K T e m p e r a u r e [ K ] Fig. 1: Atmospheric temperature structure changes resulting from varied O concen-trations. The time-averaged mean atmospheric temperature structure across the Earth’s latitudesis plotted for the PI simulation ( left ), the 1% PAL simulation ( centre ) and the 0.1% PAL simula-tion ( right ). The warm (red) stratosphere in the PI simulation has been replaced by a cool (blue)middle atmosphere in the lower oxygen simulations.3he time-averaged mean stratospheric temperature increase is ≈ K between 100 hPa and 1 hPa.In the same pressure range for the lower O simulations, there is no global temperature increasegreater than 3 K. The position of the mesopause, which marks the start of the thermosphere, hasbeen moved up in pressure (down in altitude) from 0.02 hPa ( ∼ km) in the PI case to 0.2 hPa( ∼ km) in the 0.1% PAL case.In addition to changes in the thermal structure of the atmosphere, Fig. 2 shows how ProterozoicO levels also lead to striking changes in the chemical structure of the atmosphere. A decrease in O concentrations and O column density results in increased ultraviolet flux penetrating the lower at-mosphere and increasing photolysis rates. Panels a, b and h in Fig. 2 show that enhanced photolysishas reduced the mixing ratios of H O, CH and CO : at − hPa, the time-averaged mean volume −7 −6 −5 H O mi−ing ratio10 −3 −1 P r ss u r [ h P a ] PI (100% PAL)1% PAL0.5%PAL0.1% PAL a 10 .8 .7 .6 CH %i−ing )atio b 10 .7 .6 .5 H %i−ing )atio c 10 .21 .15 .9 H %i−ing )atiod250 500 750T %p )atu) [K]10 .3 .1 P ) ss u ) [ h P a ] T)oposph ) St)atosph ) M sosph ) Th )%osph ) 10 .12 .9 .6 O %i−ing )atio f 10 .12 .10 .8 OH %i−ing )atio g 10 .6 .5 .4 CO %i−ing )atioh Fig. 2: Chemical mixing ratio profiles resulting from various O concentrations. Thetime-averaged global mean atmospheric profiles output from the WACCM6 simulations are plotted.The PI (black), 1% PAL (dark blue), 0.5% PAL (blue) and 0.1% PAL (light blue) simulationsare shown, where O relative to its present atmospheric level, is at 100%, 1%, 0.5% and 0.1%,respectively. The top row shows the volume mixing ratios for H O (a) , CH (b) , H (c) and H (d) . The bottom row shows the temperature profile of the atmosphere (e) with the volume mixingratio profiles for O (f ) , OH (g) and CO (h) . The PI atmospheric layers are indicated by greydashed lines - see panel e . 4
90 −75 −60 −45 −30 −15 0 15 30 45 60 75 90Latitude [ ∘ ]050100150200250300350400 O z o n e ( O ) c o l u m n [ D U ] PI (100% PAL)1% PAL 0.5% PAL0.1% PALFull UV screen lower estimate Full UV screenupper estimatea 10 −4 O concentration [PAL] StandardmodelW+ 2017K & D 1980L+ 1979
Fig. 3: O column varying with latitude . a, The O column in Dobson Units is plottedagainst latitude for the PI (black), 1% PAL (dark blue), 0.5% PAL (blue) and 0.1% PAL (lightblue) simulations. Biologically effective full UV screens for life are given, with upper and lower estimates indicated by magenta dotted and magenta dashed lines, respectively. b, O columnsvarying with O concentrations are compared with different models used in previous studies .The three circles for each WACCM6 simulation, from top to bottom, represent the maximum, mean,and minimum O columns as seen in panel a .mixing ratios for these species from the PI to the 0.1% PAL case have been reduced by factors of . × , ∼ and . respectively. Excluding the YS simulation, all global mean surface temper-atures are within 4 K of each other (see Extended Data Fig. 1b). The lower surface temperaturescompared to the PI atmosphere occur due to stratospheric reductions in O , CO , H O and CH ; allare greenhouse gases. OH, a key driver of atmospheric chemistry, is increased at the surface from avolume mixing ratio of . × − to . × − between the PI case and the 0.1% PAL case dueto production from photolysis of several chemical species. As a result of oxidation by OH, the lossof CH from the surface is increased.The Earth’s present-day O column varies geographically depending on incident sunlight andthe Brewer-Dobson circulation . Fig. 3 shows this variation across Earth’s different latitudes. Thesimulated global mean total O column for the PI atmosphere case is 289 Dobson Units (DU = . × molecules m − ), decreasing to O columns of 69 DU and 19 DU for the 1% PAL and0.1% PAL simulations, respectively. A full biologically effective UV shield of & DU is no longerpresent in the reduced O cases . Additional simulations show that an ozone shield of & DU is reached in the midlatitudes at 10% PAL of O and sustained across all latitudes at 50% PAL5f O (both are Phanerozoic levels of O ).An order-of-magnitude reduction in O column density from the pre-industrial atmosphere tothe Proterozoic implies that much stronger UV fluxes reached Earth’s surface, with wide-rangingconsequences for habitable niches on the Proterozoic Earth during the rise of Eukaryotes. For O levels between 0.5% PAL and 1% PAL, the total mean O column levels in our 3D simulations are afactor of 1.5-3 times lower than in prior 1D and 3D simulations . How the O shield evolvedthrough time is critical to explaining the development of life from the beginning of the Proterozoicthrough to the Cambrian explosion. Barring another shielding mechanism such as hydrocarbon orsulphur haze, a viable O shield was a likely prerequisite for the emergence of life on land . Atmospheric loss through the O valve Hydrogen can be lost from planetary atmospheres through a variety of mechanisms . Photolyticloss of CH and H O, and subsequent irreversible loss of hydrogen, can abiotically increase theoxidation state of the Earth .The current hydrogen escape rate is ≈ × mol of H yr − , corresponding to the productionof 0.1 bar of O and 1 metre of water loss per 1 billion years . From isotopic evidence andcomputational modelling, since its formation, the Earth may have lost between 0.26 and 2 timesthe present ocean volume , although such estimates depend on the transfer between the Earth’soceans and its interior .The hydrogen escape rate is given by a diffusion limiting rate , Φ esc , such that Φ esc ∝ f T ( H ) , (1)where f T ( H ) is the total mixing ratio of hydrogen components at the homopause (the atmosphericpoint above which molecular diffusion dominates and below which turbulent diffusion dominates),such that f T ( H ) = f ( H ) + 2 f ( H ) + 2 f ( H O ) + 4 f ( CH ) ..., (2)6here f ( H ) is the volume mixing ratio of H, and so forth for the other species.In the modern atmosphere, the escape rate of total hydrogen is fixed by a diffusion limited rate( Φ esc ) because the hydrogen that is lost from the exosphere is populated by Φ esc slower than itcan escape . This is the upper hydrogen bottleneck. There is another bottleneck in the loweratmosphere, known as the ‘cold trap’: the coldest parts of the tropical tropopause layer (TTL)freeze-out water as it propagates upwards . Consequently, the water vapour mixing ratio enteringthe stratosphere is reduced from tropospheric levels ( f (H O) of ∼ . to ∼ . This limits the entry value of f T ( H ) into the lower stratosphere. CH isnot halted by the ‘cold trap’ due to its low boiling point, but it is oxidised to produce H O in thelower stratosphere primarily through reaction with OH.Fig. 4 presents the total hydrogen mixing ratio profile in each simulated atmosphere. For thePI simulation, the mixing ratio of total hydrogen entering the stratosphere is f T ( H ) = 12 ppmv.Above the cold trap, the H O mixing ratio increases until it reaches a maximum mesospheric value( ≈ ppmv) due to CH reacting with OH. Turbulent mixing causes f T ( H ) to remain roughlyconstant between the lower stratosphere and homopause. Above the homopause, through diffusiveseparation, the lighter atmospheric constituents increase in relative abundance with decreasingpressure (increasing altitude).The low oxygen cases present rather different scenarios because the cold trap mechanism issensitive to O concentrations, thus affecting hydrogen escape. As the TTL cools when O andsimultaneously O decrease, a lower amount of water is able to enter the stratosphere. Due tothe ≈ K difference in time-averaged zonal mean TTL temperatures between the PI and 0.1%PAL simulations, more water is frozen out in the form of ice and ice clouds before the temperatureinversion in the PI atmosphere at ≈ hPa (see Extended Data Table 1 and Extended DataFig. 2). For example, there is ≈ ppmv less H O at 100 hPa (lower stratosphere) for the 0.1% PALcase. Consequently, a reduced amount of water propagates into the stratosphere in the low oxygencases compared to the PI atmosphere, contributing less to f T ( H ) .Resulting from the increase in OH, there is tropospheric production of H O from CH oxidation;methane is now indirectly affected by the ‘cold trap’ (because the water produced can freeze-out)7 .5 5.0 7.5 10.0 12.5 15.0 17.5f T (H) [ppm)]10 −1 −2 −3 −4 P r e ss ( r e [ h P a ] PI (100% PAL)1%PAL0.5%PAL0.1% PAL YS (1% PAL)DecreasingH escape IncreasingH escapeHomopausea Time before present I n c r e a s i n g H e s c a p e Archean Proterozoic Phanerozoic 4 Gyr ~2.4 Gyr 0.54 Gyr 0 GyrPI0.1% PAL0.5% PAL1% PAL YSMuch higherH escapeb 4 Gyr ~2.4 Gyr 0.54 Gyr 0 Gyr
Fig. 4: Levels of diffusion-limited hydrogen escape varying with O concentrations. a, The total hydrogen mixing ratio f T (H) profile is plotted for the simulations of PI (black), 1% PAL(dark blue), 0.5% PAL (blue), 0.1% PAL (light blue) and YS (red), in units of parts per millionby volume (ppmv). Total hydrogen is mainly composed of water vapour (H O), methane (CH ),atomic hydrogen (H) and molecular hydrogen (H ). The approximate pressure of the homopauseis indicated by the grey dotted line. b, A qualitative diagram showing the relative levels (not toscale) of hydrogen escape through Earth’s history.in the 0.1% PAL simulation. Some oxidation of CH occurs above the tropopause in the 0.5% PALand 0.1% PAL simulations, however, there is not enough production of water from this reactionto reverse the decrease in contribution to total hydrogen (see Extended Data Fig. 3 for all f T ( H ) contributions).The f T ( H ) profile shows that for the lower oxygen cases the rate of hydrogen escape is lower thanthat of present day escape ( Φ esc,PI ). The rates for the lower oxygen simulations are: . esc,PI , . esc,PI , and . esc,PI , for the 0.1% PAL, 0.5% PAL, and 1% PAL simulations, respectively.In the YS simulation, the 2 Gyr younger Sun decreases the escape rate from the 1% PAL case bya further 10% because the TTL is cooler and more water freezes out lower in the atmosphere - seeExtended Data Fig. 2. Therefore, both the young Sun and O levels during the Proterozoic willhave driven the limiting flux for hydrogen escape.Our estimates of f T ( H ) during the Proterozoic compared with the pre-industrial atmosphereillustrate how evolving O levels would have driven fluctuating but generally low rates of hydrogenescape through the Proterozoic eon, and demonstrate that most of Earth’s water loss throughhydrogen escape must have occurred in the preceding Archean and Hadean eons .8 he Proterozoic faint young Sun problem The chemical lifetime of a molecule is its mean lifetime before it is destroyed. The effect of reducingO vastly reduces atmospheric chemical lifetimes, such as those of H O ( τ H O ) and CH ( τ CH ) -see Extended Data Fig. 4. τ H O at the surface in the PI simulation is 60 thousand years, whereasit is ≈ years for the 0.1% PAL case. τ CH is 3.8 years at the surface for the PI case and just 13days for the 0.1% PAL case. Each simulation has a constant surface CH mixing ratio. However, τ CH at the surface in each case is different such that the ratios between the surface lifetimes isthe inverse of the ratio between the surface fluxes. These CH flux ratios are 1:15:30:104 for the(PI):(1% PAL):(0.5% PAL):(0.1% PAL) cases.The Faint Young Sun Paradox is the problem associated with the early Sun outputting less totalenergy, yet the surface temperatures of Earth remaining high enough for liquid water to exist .While the Faint Young Sun Paradox may have been solved for the Archean climate , the questionof how the Earth maintained an ice-free surface throughout most of the Proterozoic remains tobe answered . 3D simulations have shown that an ice-free surface can be sustained during theMesoproterozoic if CO is at 10 times its pre-industrial level and there is between 28 and 140 ppmvof CH . Surface f (CH ) in our simulations is 0.8 ppmv. Consequently, given the surface τ CH values for the low O cases, our results show that a CH flux increase (compared to present day) ofa factor between 525 and 3640 is needed to reach levels of 28 ppmv during the Proterozoic, and 5times these values to reach 140 ppmv. With a 2 Gyr younger Sun, τ CH is further reduced since theLyman- α radiation intensity has doubled compared to the modern Sun. For these reasons CH isunlikely to be one of the most important greenhouse gases during the Proterozoic, in agreement withprevious work . Instead, other mechanisms are required to explain a mostly ice-free Proterozoic,such as elevated levels of N O (also unlikely due to high rates of photolysis), alterations in thecontinental coverage , and large-scale mantle thermal mixing variations .9 xoplanet analogs Our simulations represent not only the Earth through geological time, but possible terrestrial exo-planets (rocky planets orbiting other stars). The Archean, Proterozoic and Phanerozoic atmospheresrepresent key analogs for observations of potentially habitable exoplanet atmospheres. The transmis-sion, emission and reflectance spectra for exoplanets with O levels similar to the Proterozoic Earthwill differ markedly from expectations for higher O atmospheres resulting from the chemical andtemperature variations that we demonstrate. These potential differences should be accounted for inthe design of next generation telescopes such as the LUVOIR and HabEx concepts, which arebeing designed around simulations that present mostly different volume mixing ratios for severalgreenhouse gases compared to the simulations presented here . It is crucial that such missionsare developed alongside the forefront of exoplanet simulations, where inter-comparison projects cangive insights into model disparities .During an exoplanet transit of its host star, light passes through the atmosphere. It is scattered,absorbed, and re-emitted along the line of sight to the observer. Such transmission spectra probe T r a n s i t d p t h [ pp m ] O O O E ff c t i v h i g h t [ k m ] CO CO CH CO O H OO -X CH O CO PI (100% PAL) 1% PAL 0.1% PAL
Fig. 5: Theoretical transmission spectra of possible Earth-like exoplanets.
Three theo-retical transmission spectra of Earth are produced using the Planetary Spectrum Generator , anonline radiative transfer suite. Transmission spectra probe the signal from the Earth’s terminator(day-night dividing line) from afar during a transit of its host star. Three simulations are included,with O at present day levels (100% PAL - black) and reduced levels (1% PAL - dark blue; 0.1%PAL - light blue). The signal is given in terms of: transit depth - the relative dip in starlightcaused by absorption; effective altitude - the effective additional thickness added to the planet’sopaque disk blocking light in a specific wavelength bin. The spectra have been binned to a spectralresolution of 100, between 0.1 and µ m . Atmospheric features which are due to the absorption oflight by specific chemicals are indicated in grey. 10arth-sized exoplanets around Sun-like stars above the tropopause . Comparable models (1Dand 3D) that have explored the impact of low O concentrations neglect the upper layers of theatmosphere . Including the upper atmosphere as we do here results in substantially reducedH O and CH concentrations, and contrasting O concentrations, above the troposphere. CO isalso reduced by a lesser extent but its higher concentrations still impacts the propagation of light.In Fig. 5 we show the theoretical transmission spectra for a pre-industrial Earth and Earths withO levels of 1% and 0.1% PAL. Absorption differs considerably for molecules such as O , O , CH ,CO , H O and the collision induced absorption for O (known as O – X, of which O – O andO – N contribute the most ). O and O have been proposed as potential biosignatures, alongwith CH when observed in combination with O . We find that although O , O and CH arestill present in an atmosphere with Proterozoic O levels, their signal is greatly reduced, despite theknown habitation of the Proterozoic Earth . If biosignature presence is confirmed on an exoplanet,before conclusions about extraterrestrial life can be drawn, ascertaining whether the biosignature isproduced biologically or abiotically is essential .Oxygen is central to life. It is a vital ingredient of cellular respiration and a key product ofphotosynthesis. As we have demonstrated, oxygen exerts a first-order control on the structureand composition of Earth’s atmosphere, modulating hydrogen escape, surface habitability, and theabsorption of light during planetary transits. Our results show that the atmospheric evolution ofnon-reducing terrestrial planets are affected by varied O concentrations, and future simulationsshould include the fact that upward diffusion of CH can be limited by the tropopause ‘cold trap’.The low-O Proterozoic motivates the need to investigate biosignature detection for exoplanetswhere oxygen richness is not a prerequisite for life.In the era of exoplanet discovery, there has been a renewed interest in the Earth’s history. Tounderstand the habitability of newly discovered worlds, it is becoming increasingly clear that themost robust terrestrial exoplanet models, combined with reliable retrieval models, are required toprepare for the forthcoming age of exoplanet characterisation.11 ethods
The climate model: WACCM6
This work uses the most recent model of the Whole Atmosphere Community Climate Model -WACCM6 . Previous WACCM models have been used for a variety of purposes, such as simulatingclimate change between the industrial revolution and the 21 st century , as well as investigating theeffects of solar flares on the middle atmosphere .WACCM6 is a 3-dimensional (3D) global climate model (GCM). It has 70 atmospheric layers,from a surface pressure of 1000 hPa down until a pressure of . × − hPa (an approximate altitudeof 140 km, which corresponds to the thermosphere) . We used a horizontal grid of . ◦ × . ◦ (longitude × latitude), and a 30 minute time step for the atmosphere, in all simulations.The standard downloadable WACCM6 pre-industrial baseline simulation source code was alteredto produce the 1% PAL, 0.5% PAL, 0.1% PAL and YS simulations. For the 1% PAL (and YS),0.5% PAL and 0.1% PAL simulations, the lower boundary condition for the O mixing ratio wasscaled by − , × − and − , respectively - see Extended Data Fig. 5 for the O mixing ratioprofiles. The 0.1% PAL case was perturbed from the 0.5% PAL case at 29 model years and ran untila total of 77 model years. In each simulation other than the PI case, new minimum mixing ratios forO and CH , both set at − , were set to − and − , respectively. A constant mixing ratiocondition for O at the lower boundary was imposed for the 0.5% and 0.1% PAL simulations becausesurface O rapidly decreases below these scaled values without the imposed boundary condition.The length of the simulations, given in terms of model years, is given in Extended Data Table1. These durations allowed the effects from various chemical changes to fully mix. The simulationswere halted once 4 years of inter-annual equilibrium for the total hydrogen mixing ratio at all levelswas observed. A time-averaged mean was taken over the last 4 years of each simulation (for example:the beginning of year 28 to the end of year 31 for the 1% PAL case). Data was output in termsof monthly and 5 day means, as well as snapshots every 5 days (instantaneous values for modelvariables).Although WACCM6 simulates up to . × − hPa, the figures in this study have a minimum12ressure of × − hPa. The reason for this is because we do not know what the upper boundaryconditions would be when oxygen is reduced and did not want to specify boundary conditionsthat bias the upper atmosphere. From running many perturbation experiments, we found that theupper boundary condition does not affect the atmosphere below × − hPa. Despite this, weselected from these experiments what we decided would best represent the upper boundary in eachsimulation for each particular chemical constituent. The upper boundary was changed for H , H,H O, CH , O, O , and N depending on these perturbation experiments. It is important to notethat the final set of upper boundary conditions did not affect the total hydrogen mixing ratio at thehomopause, the temperature structure below × − hPa nor the chemical mixing ratios below × − hPa. Therefore, the choice of upper boundary conditions do not affect the conclusions ofthis work.With increased photolysis of CH , there is a question surrounding the fact that an organic hazelayer may form, cool the surface and reduce photolysis below this layer. WACCM6 does not supportthe formation of organic haze, although a haze layer is unlikely to exist in our simulated atmospheresbecause the CH /CO ratio ≪ . Furthermore, it has been found experimentally that hazeparticle production decreases as O levels increase above − PAL . The effects of a haze layerwere not investigated for these reasons. Data processing
Total hydrogen
The hydrogen bearing species in WACCM6 used for the total hydrogen mixing ratio in this workwere: H O, H , H, CH , HNO , OH, H O , HO , H SO , CH O , and CH O, liquid clouds, iceparticles and ice clouds. They were weighted as in Eq. (2). Note that above ≈
100 hPa, H O, H ,H and CH contributed more that 99.5% towards the total hydrogen mixing ratio ( f T ( H ) ). Tropopause temperatures
Time-averaged values for the tropopause temperature in terms of global mean ( T Tp ), global min-imum ( T Tp,min ) and global maximum values ( T Tp,max ) were calculated from taking the TROP_Toutput variable (model calculated tropopause temperature) and then averaging over time for thefinal 4-year period. The mean, minimum and maximum of all the time-averaged TROP_T gridcells were then taken.
Chemical lifetimes
Chemical lifetimes were calculated for H O and CH to produce Extended Data Fig. 4. Thechemical loss rate of H O, L H O , and the loss rate of CH , L CH are calculated within WACCM6.The H2O_CHML and CH4_CHML output variables are the chemical loss rate in units of moleculesm − s − , of H O and CH , respectively.To calculate the molecular lifetime, the number density of the constituent is divided by it’s lossrate in each grid cell and then the mean lifetime is computed globally. Therefore, the lifetime ofH O ( τ H O ) is given by τ H O = n H O L H O , (3)where n H O is the number density of H O, with a similar expression for CH olar evolution model We used an existing solar evolution model to produce the solar spectrum at 2 Gyr before present.The model can produce theoretical spectra for the Sun between 4.4 Gyr in the past and 3.6 Gyr inthe future. The model is validated from 0.1 nm to µ m . We extend the model further into the farinfrared, modelling the Sun in this region as a blackbody. The spectrum from the solar evolutionmodel was re-binned (using a Python tool called SpectRes ) whilst conserving flux, to ensure thatthe new spectrum was interpolated onto the WACCM6 solar spectrum grid.Changes in the total solar insolation and UV flux are important as they effect the overalltemperature structure of the atmosphere. Greater solar insolation increases the surface temperature,whereas more ultraviolet (UV) and extreme ultraviolet (EUV) flux can change the temperature ofdifferent atmospheric layers, depending on the location of the dominant UV and EUV absorbers. Inthe case of the modelled 2 Gyr younger Sun, the EUV increases in flux by a factor of 2.97 but theUV decreases to 0.84 times the present day flux. This increases the total O column (reducing thephotolysis of H O and CH ) but increases the flux of Lyman- α radiation (increasing the photolysisof H O and CH ). The UV wavelength range was assumed to be between 10 nm and 400 nm andthe EUV wavelength range was assumed to be between 10 nm and 91 nm . This younger Sunreplaced the standard solar spectrum used in WACCM6 to produce the YS simulation. Planetary Spectrum Generator transmission spectra
The transmission spectra in Fig. 5 were produced using the Planetary Spectrum Generator (PSG).PSG is a diverse radiative-transfer suite that can be accessed online (https://psg.gsfc.nasa.gov/). Weused PSG’s Global Exoplanet Spectra application to directly incorporate the 3D data (temperature,pressure, and chemical mixing ratios) produced from the WACCM6 simulations. To enable anupload that did not exceed the maximum allowed file size, the snapshot data was averaged overlongitude, in chunks of 4 longitude grid spaces. This reduced the total number of longitudinal pointsfrom 144 to 36. Snapshot data on January st , on the last year of each simulation, was used as thedate for each transmission spectrum. PSG contains orbital information for solar system bodies andexoplanets. In this case, we used the orbital properties for Earth on the st of January. There is no15articular reason for selecting this date; although spectra differences exist between seasons, seasonalvariations are small compared with the variation between simulations. The 0.5% simulation wasnot included in Fig. 5 for clarity, but has absorption between the 1% PAL and 0.1% PAL cases.Due to refraction, the atmosphere below ∼ . km for Earth-like planets around Sun-like starscannot be probed . PSG accounts for this, such that refraction creates a roughly flat spectrum at ∼ were used for molec-ular absorption (of O , O , H O, CO , CH , N , OH and HNO ) radiative transfer calculations.Collision induced absorption (of O – O , O – N , O – CO , N – N ) were also included, making useof data from the HITRAN database .The radius of the Earth was assumed to be 6371 km, with an atmospheric average molecularweight of 28.97 atomic mass units for the pre-industrial atmosphere (PI), and 28.00 atomic massunits for the low O atmospheres, which are primarily composed of N . Code availability
Author information
D.R.M., J.-F.L. and B.A.B. initiated the preliminary research. B.A.B. performed preliminary sim-ulations. G.J.C., D.R.M. and J.-F.L. performed the final simulations. G.J.C. produced the figures.All authors analysed and interpreted the simulation output data. C.W. and D.R.M advised on theconsequences of early earth simulations on exoplanet observations. G.J.C. wrote the manuscriptwith input and comments on the final manuscript preparation from all authors.16 orrespondence and requests for materials should be addressed to G.J.C.
Ethics declarations
Competing interests
The authors declare no competing interests.
Acknowledgements
We thank Dr. Geronimo Villanueva (NASA-Goddard Space Flight Center) for helpful advice withthe use of the Planetary Spectrum Generator.G.J.C. acknowledges the studentship funded by the Science and Technology Facilities Coun-cil of the United Kingdom (STFC). C.W. acknowledges financial support from the University ofLeeds and from the Science and Technology Facilities Council (grant numbers ST/R000549/1 andST/T000287/1).This work was undertaken on ARC4, part of the High Performance Computing facilities at theUniversity of Leeds, UK. We would like to acknowledge high-performance computing support fromCheyenne (doi:10.5065/D6RX99HX) provided by NCAR’s Computational and Information SystemsLaboratory, sponsored by the National Science Foundation. The CESM project is supported pri-marily by the National Science Foundation (NSF). This material is based upon work supported bythe National Center for Atmospheric Research (NCAR), which is a major facility sponsored by theNSF under Cooperative Agreement 1852977.
References [1] Lammer, H. et al. Origin and evolution of the atmospheres of early Venus, Earth and Mars.The Astronomy and Astrophysics Review , 2 (2018).172] Kasting, J. F. & Catling, D. Evolution of a Habitable Planet.Annual Review of Astronomy and Astrophysics , 429–463 (2003).[3] Wolf, E. T., Shields, A. L., Kopparapu, R. K., Haqq-Misra, J. & Toon, O. B. Constraintson Climate and Habitability for Earth-like Exoplanets Determined from a General CirculationModel. The Astrophysical Journal , 107 (2017). 1702.03315.[4] Lammer, H. et al. What makes a planet habitable? Astron Astrophys Rev , 181–249 (2009).[5] Lyons, T. W., Reinhard, C. T. & Planavsky, N. J. The rise of oxygen in Earth’s early oceanand atmosphere. Nature , 307–315 (2014).[6] Planavsky, N. J. et al. Low Mid-Proterozoic atmospheric oxygen levels and the delayed rise ofanimals. Science , 635–638 (2014).[7] Lenton, T. M., Boyle, R. A., Poulton, S. W., Shields-Zhou, G. A. & Butterfield, N. J. Co-evolution of eukaryotes and ocean oxygenation in the Neoproterozoic era. Nature Geoscience , 257–265 (2014).[8] Warke, M. R. et al. The Great Oxidation Event preceded a Paleoproterozoic “snowball Earth”.Proceedings of the National Academy of Science , 13314–13320 (2020).[9] Gaudi, B. S. et al. The Habitable Exoplanet Observatory (HabEx) Mission Concept StudyFinal Report. arXiv e-prints arXiv:2001.06683 (2020). 2001.06683.[10] The LUVOIR Team. The LUVOIR Mission Concept Study Final Report. arXiv e-printsarXiv:1912.06219 (2019). 1912.06219.[11] Ratner, M. I. & Walker, J. C. G. Atmospheric Ozone and the History of Life.Journal of Atmospheric Sciences , 803–808 (1972).[12] Berkner, L. V. & Marshall, L. C. On the Origin and Rise of Oxygen Concentration in theEarth’s Atmosphere. Journal of Atmospheric Sciences , 225–261 (1965).[13] Zahnle, K. J. Earth’s earliest atmosphere. Elements , 217–222 (2006).1814] Stüeken, E. E. et al. Mission to Planet Earth: The First Two Billion Years.Space Science Reviews , 31 (2020).[15] Godfrey, L. V. & Falkowski, P. G. The cycling and redox state of nitrogen in the Archaeanocean. Nature Geoscience , 725–729 (2009).[16] Gettelman, A. et al. The Whole Atmosphere Community Climate Model Version 6 (WACCM6).Journal of Geophysical Research (Atmospheres) , 12,380–12,403 (2019).[17] Claire, M. W. et al. The Evolution of Solar Flux from 0.1 nm to 160 µ m: Quantitative Estimatesfor Planetary Studies. The Astrophysical Journal , 95 (2012).[18] Way, M. J. et al. Resolving Orbital and Climate Keys of Earth and Extraterrestrial En-vironments with Dynamics (ROCKE-3D) 1.0: A General Circulation Model for Simulatingthe Climates of Rocky Planets. The Astrophysical Journal Supplement Series , 12 (2017).1701.02360.[19] Levine, J. S., Hays, P. B. & Walker, J. C. G. The evolution and variability of atmosphericozone over geological time. Icarus , 295–309 (1979).[20] Kasting, J. F. & Donahue, T. M. The evolution of atmospheric ozone.Journal of Geophysical Research: Oceans , 3255–3263 (1980).[21] Butchart, N. The Brewer-Dobson circulation. Reviews of Geophysics , 157–184 (2014).[22] Cockell, C. S. & Raven, J. A. Ozone and life on the Archaean Earth.Philosophical Transactions of the Royal Society of London Series A , 1889–1901 (2007).[23] Gronoff, G. et al. Atmospheric Escape Processes and Planetary Atmospheric Evolution.Journal of Geophysical Research (Space Physics) , e27639 (2020). 2003.03231.[24] Catling, D. C., Zahnle, K. J. & McKay, C. P. Biogenic Methane, Hydrogen Escape, and theIrreversible Oxidation of Early Earth. Science , 839–843 (2001).[25] Zahnle, K. J., Catling, D. C. & Claire, M. W. The rise of oxygen and the hydrogen hourglass.Chemical Geology , 26–34 (2013). 1926] Pope, E. C., Bird, D. K. & Rosing, M. T. Isotope composition and volume of Earth’s earlyoceans. Proceedings of the National Academy of Science , 4371–4376 (2012).[27] Kurokawa, H., Foriel, J., Laneuville, M., Houser, C. & Usui, T. Subduc-tion and atmospheric escape of Earth’s seawater constrained by hydrogen isotopes.Earth and Planetary Science Letters , 149–160 (2018). 1806.03792.[28] Korenaga, J., Planavsky, N. J. & Evans, D. A. D. Global water cy-cle and the coevolution of the Earth’s interior and surface environment.Philosophical Transactions of the Royal Society of London Series A , 20150393 (2017).[29] Hunten, D. M. The Escape of Light Gases from Planetary Atmospheres.Journal of Atmospheric Sciences , 1481–1494 (1973).[30] Fueglistaler, S. et al. Tropical tropopause layer. Reviews of Geophysics , RG1004 (2009).[31] Dessler, A. E. et al. Variations of stratospheric water vapor over the past three decades.Journal of Geophysical Research (Atmospheres) , 12,588–12,598 (2014).[32] Wang, T. & Dessler, A. E. Analysis of cirrus in the tropical tropopause layer from CALIPSOand MLS data: A water perspective. Journal of Geophysical Research (Atmospheres) ,D04211 (2012).[33] Feulner, G. The faint young Sun problem. Reviews of Geophysics , RG2006 (2012).1204.4449.[34] Charnay, B., Wolf, E. T., Marty, B. & Forget, F. Is the Faint Young Sun Problem for EarthSolved? Space Science Reviews , 90 (2020). 2006.06265.[35] Olson, S. L., Reinhard, C. T. & Lyons, T. W. Limited role for methane in the mid-Proterozoicgreenhouse. Proceedings of the National Academy of Science , 11447–11452 (2016).[36] Fiorella, R. P. & Sheldon, N. D. Equable end Mesoproterozoic climate in the absence of highCO2. Geology , 231–234 (2017). 2037] Bradley, D. C. Secular trends in the geologic record and the supercontinent cycle.Earth Science Reviews , 16–33 (2011).[38] Jellinek, A. M., Lenardic, A. & Pierrehumbert, R. T. Ice, Fire, or Fizzle: The Climate Footprintof Earth’s Supercontinental Cycles. Geochemistry, Geophysics, Geosystems , e08464 (2020).[39] Rugheimer, S. & Kaltenegger, L. Spectra of Earth-like Planets through Geological Evolutionaround FGKM Stars. The Astrophysical Journal , 19 (2018). 1712.10027.[40] Bolcar, M. R. et al. The Large UV/Optical/Infrared Sur-veyor (LUVOIR): Decadal Mission concept design update. InSociety of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, vol. 10398of Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, 1039809(2017).[41] Mennesson, B. et al. The Habitable Exoplanet (HabEx) Imaging Mission: prelim-inary science drivers and technical requirements. In MacEwen, H. A. et al. (eds.)Space Telescopes and Instrumentation 2016: Optical, Infrared, and Millimeter Wave, vol.9904 of Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series,99040L (2016).[42] Fauchez, T. J. et al. TRAPPIST-1 Habitable Atmosphere Intercomparison (THAI): motivationsand protocol version 1.0. Geoscientific Model Development , 707–716 (2020). 2002.10950.[43] Villanueva, G. L., Smith, M. D., Protopapa, S., Faggi, S. & Mandell, A. M. Planetary Spec-trum Generator: An accurate online radiative transfer suite for atmospheres, comets, smallbodies and exoplanets. Journal of Quantitative Spectroscopy and Radiative Transfer , 86–104 (2018). 1803.02008.[44] Bétrémieux, Y. & Kaltenegger, L. Impact of Atmospheric Refraction: HowDeeply can We Probe Exo-Earth’s Atmospheres during Primary Eclipse Observations?The Astrophysical Journal , 7 (2014). 1312.6625.2145] Gebauer, S. et al. Evolution of Earth-like Extrasolar Planetary Atmospheres: Assessing theAtmospheres and Biospheres of Early Earth Analog Planets with a Coupled Atmosphere Bio-geochemical Model. Astrobiology , 27–54 (2017). 1807.06844.[46] Fauchez, T. J. et al. Sensitive probing of exoplanetary oxygen via mid-infrared collisionalabsorption. Nature Astronomy , 372–376 (2020). 2001.01361.[47] Meadows, V. S. et al. Exoplanet Biosignatures: Understanding Oxygen as a Biosignature inthe Context of Its Environment. Astrobiology , 630–662 (2018). 1705.07560.[48] Marsh, D. R. et al. Climate Change from 1850 to 2005 Simulated in CESM1(WACCM).Journal of Climate , 7372–7391 (2013).[49] Pettit, J. et al. Effects of the September 2005 Solar Flares and Solar Proton Events on theMiddle Atmosphere in WACCM. Journal of Geophysical Research (Space Physics) , 5747–5763 (2018).[50] Pavlov, A. A., Brown, L. L. & Kasting, J. F. UV shielding of NH and O by organic hazes inthe Archean atmosphere. Journal of Geophysical Research: Planets , 23267–23288 (2001).[51] Trainer, M. G. et al. Inaugural Article: Organic haze on Titan and the early Earth.Proceedings of the National Academy of Science , 18035–18042 (2006).[52] Hörst, S. M. et al. Exploring the Atmosphere of Neoproterozoic Earth: The Effect of O onHaze Formation and Composition. The Astrophysical Journal , 119 (2018). 1804.02775.[53] Carnall, A. C. SpectRes: A Fast Spectral Resampling Tool in Python. arXiv e-printsarXiv:1705.05165 (2017). 1705.05165.[54] France, K. et al. The LUVOIR Ultraviolet Multi-Object Spec-trograph (LUMOS): instrument definition and design. InSociety of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, vol. 10397of Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, 1039713(2017). 1709.06141. 2255] Peacock, S., Barman, T., Shkolnik, E. L., Hauschildt, P. H. & Baron, E. Predicting the ExtremeUltraviolet Radiation Environment of Exoplanets around Low-mass Stars: The TRAPPIST-1System. The Astrophysical Journal , 235 (2019). 1812.06159.[56] Gordon, I. E. et al. The HITRAN2016 molecular spectroscopic database.Journal of Quantitative Spectroscopy and Radiative Transfer , 3–69 (2017).23 xtended Data Table 1: Five different simulations are presented for this work. There is apre-industrial (PI) case, three cases with reduced O levels (1%, 0.5% and 0.1% PAL cases) anda young Sun case, also with lower O (YS). The amount of model years simulated is given. Thevolume mixing ratio for O , f (O ), is given in terms of present atmospheric level (PAL). The volumemixing ratio for N , f (N ), is listed. Time-averaged values for the tropopause temperature in termsof global mean ( T Tp ), global minimum ( T Tp,min ) and global maximum values ( T Tp,max ) are alsolisted, alongside the time-averaged global mean O column, C O , in Dobson Units (DU). Simulation name Length [model years] f ( O ) [PAL] f ( N ) T Tp [K] T Tp,min [K] T Tp,max [K] C O [DU]PI 17 1 0.78 206.8 193.3 216.0 289.21% PAL 31 0.01 0.98 204.2 191.6 213.5 69.30.5% PAL 48 0.005 0.98 203.0 189.2 212.6 49.30.1% PAL 77 0.001 0.98 199.4 184.5 209.5 19.0YS 41 0.01 0.98 201.7 188.7 212.8 87.4 −90 −75 560 545 530 515 0 15 30 45 60 75 90La0i01de [ ∘ ]220240260280300 T i m e - a v e r a g e d s u r f a c e t e m p e r a t u r e [ K ] aJan Feb Mar Apr May Jun Jul Aug Sep Oct Nov DecMonth284286288290 G l o b a l m e a n s u r f a c e t e m p e r a t u r e [ K ] PI (100% PAL) 1% PAL0.5% PAL0.1% PAL b
Extended Data Fig. 1: Surface temperature of Earth affected by O concentrations. a, The surface temperature of Earth varying with latitude averaged over longitude and time is plottedfor the pre-industrial simulation and lower oxygen simulations. The largest difference between thesimulations occurs at ◦ latitude (the North pole) and is ≈ K. b, The time-averaged monthlyglobal mean surface temperature is plotted for the pre-industrial simulation and lower oxygen sim-ulations. All global mean surface temperatures are within ≈ K. The PI (black), 1% PAL (darkblue), 0.5% PAL (blue) and 0.1% PAL (light blue) simulations are displayed in a and b .24
75 180 185 190 195 200 205 210Temperature [K]50100150200250300 P r e ss u r e [ h P a ] PI (100% PAL)1% PALYS(1% PAL)0.5% PAL0.1% PAL a
Temperature P r e ss u r e [ h P a ] b H O + liquid cloudsIce particles + ice clouds
Extended Data Fig. 2: O concentrations altering temperature and water freeze-out inthe tropical tropopause layer (TTL) . a, A time-averaged zonal mean ± ◦ from the equator ispresented for temperature in the tropical tropopause layer (TTL) for the simulations of PI (black),1% PAL (dark blue), 0.5% PAL (blue), 0.1% PAL (light blue) and YS (red). b, Mixing ratios areplotted for water (the summation of water vapour and water clouds - dotted line) in the TTL andice (the summation of ice particles and ice clouds - dashed line) for the same simulations as in a .Volume mixing ratios are given in parts per million by volume (ppmv).25
25 50 75 100H O contribution to f T (H) [%]10 −1 −3 P r e ss % r e [ h P a ] a 0 25 50 75 100CH co trib%tio to f T (H) [%]b0 25 50 75 100H co trib%tio to f T (H) [%]10 −1 −3 P r e ss % r e [ h P a ] c 0 25 50 75 100H co trib%tio to f T (H) [%]d PI (100% PAL)1% PALYS (1% PAL)0.5% PAL0.1% PAL
Extended Data Fig. 3: Percentage contributions to total hydrogen.
Percentage contribu-tions to total hydrogen throughout the atmosphere are plotted for: a, H O; b, CH ; c, H ; d, H.In each panel, the PI (black), 1% PAL (dark blue), 0.5% PAL (blue), 0.1% PAL (light blue) andYS (red) simulations are displayed. −1 H O lifetime [ ears]10 −1 −3 P r e ss u r e [ h P a ] a PI (100% PAL)1% PAL0.5% PAL0.1% PAL −1 CH lifetime [ ears]b Extended Data Fig. 4: Chemical lifetimes of H O and CH reduced by lower O con-centrations. The atmospheric chemical lifetimes of H O (a) and CH (b) are plotted for the PI(black), 1% PAL (dark blue), 0.5% PAL (blue) and 0.1% PAL (light blue) simulations. Lifetimesfor both molecules depend on the number density of the molecule and its loss rate.26 −7 −6 −5 −4 −3 −2 −1 O mixing ra io10 −4 −3 −2 −1 P r e ss u r e [ h P a ] PI1% PAL0.5% PAL0.1% PAL 1% PAL(YS)
Extended Data Fig. 5: Mixing ratio profiles of O . The mixing ratio of O2