EDP Sciences logo
Subscriber Authentication Point
Search
Menu
Home All issues Volume 680 (December 2023) A&A, 680 (2023) A103 Full HTML
Press Release
Open Access
Issue
A&A
Volume 680, December 2023
Article Number A103
Number of page(s) 15
Section Planets and planetary systems
DOI https://doi.org/10.1051/0004-6361/202346936
Published online 18 December 2023

© The Authors 2023

Licence Creative CommonsOpen Access article, published by EDP Sciences, under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

This article is published in open access under the Subscribe to Open model. Subscribe to A&A to support open access publication.

1 Introduction

The continuous improvement of telescopes, instruments, and techniques allow us to detect and characterize a growing number of exoplanets. The new generation of instruments will enable the characterization of the atmospheres of small, rocky, and temperate planets, which are now within reach thanks to the James Webb Space Telescope (JWST, e.g. Morley et al. 2017; Lustig-Yaeger et al. 2019; Fauchez et al. 2019; Wunderlich et al. 2019). The first results of the JWST are likely to open up a new era in the characterization of such targets. Future astronomical ground-based instruments in development such as RISTRETTO@VLT (Lovis et al. 2022) and ANDES@E-ELT (Marconi et al. 2022) will allow the community to go even further and study Earth-like planets. The possibility of detecting objects similar to our planet brings up the ancestral question of life (and “habitability”) in the Universe. Astrobiology is the field of science that aims to answer this question, while exoplanetology is now a big piece of that gigantic puzzle. In particular, an essential condition for the emergence and the maintenance of life as we know it is the presence of liquid water. Therefore, modeling the potential climate of planets we aim to characterize is essential to determining the conditions under which water can exist in liquid phase at their surface or in the atmosphere as clouds. This modeling effort, coupled with advanced spectroscopy techniques, will allow us to make progress in the search for life in the Universe.

For an Earth-like planet with a given amount of water, there is a range of distances to the host star for which this water can be liquid at the surface. Too close and the water evaporates and too far, then it freezes. This ideal zone is the Habitable Zone (HZ; Kasting et al. 1993; Kopparapu et al. 2013). This concept is extremely useful in the exoplanet community to statistically study large samples of targets under the spectrum of habitability and search for life. The HZ is function of the brightness of the star, the size of the planet, and the composition of its atmosphere Kopparapu et al. (2013, 2014, 2017). If the planet is close enough to the star, the temperature rises and oceans start to evaporate which may add enough vapor in the bottom part of the atmosphere to completely absorb the thermal emission because of the strong greenhouse effect of water vapor. Once thermal emission reaches the Simpson-Nakajima limit (Simpson 1929; Nakajima et al. 1992), the planet is not able to cool down anymore and the positive feedback of the runaway greenhouse arises (Komabayasi 1967; Ingersoll 1969; Kasting 1988; Goldblatt & Watson 2012). This leads to the evaporation of the entire surface ocean and a dramatic increase in global surface temperature of several hundreds of Kelvin. This climatically unstable transition separates two population of planets: temperate planets and hot post-runaway planets (e.g., Hamano et al. 2013; Turbet et al. 2021). This is one of several scenarios that are meant to explain the difference between the Earth and early-Venus (Way et al. 2016). Temperate planets are similar to the Earth with surface condensed water and a thin atmosphere, while hot planets have an extended water-dominated atmosphere with high surface temperature, up to a few thousand Kelvins (Turbet et al. 2019). The inner edge of the HZ is determined in part by the runaway greenhouse process. The key question of the existence of surface liquid water on terrestrial planets is thus inherently connected to the runaway greenhouse with strong implications for our understanding of the possible evolution of exoplanets. As shown in Turbet et al. (2020), this process may also strongly influence the mass-radius relationships of terrestrial planets by clearly separating temperate planets hosting a water ocean and hot steam planets.

Regarding the importance of this process, many papers studied it using simple 1D climate models. However, as shown by some 3D General Circulation Models (GCMs) studies (e.g., Leconte et al. 2013; Kopparapu et al. 2017), approximations assumed in 1D models lead to a wrong estimation of the climate. For example, Hadley cells on Earth-like planets tend to reduce the relative humidity around the tropics, allowing for a greater Outgoing Longwave Radiation (OLR) than the one estimated by 1D models (Leconte et al. 2013). Moreover, specific clouds pattern may appear for slow rotating planets planets (Yang et al. 2013) or for post-runaway states (Turbet et al. 2021), which (by definition) cannot be accounted for in a 1D model.

Even if 1D models are useful in many cases because of their adaptability and fast computation time, more complex 3D models should be preferred to capture intrinsically 3D physical processes, such as clouds and global dynamics. Several studies explored the runaway greenhouse effect on Earth-like planets with different GCMs (e.g., Leconte et al. 2013; Yang et al. 2013; Wolf & Toon 2014, 2015; Popp et al. 2016; Kopparapu et al. 2017) in order to determine under which insolation this positive feedback arises. Unfortunately, highly unstable climates and high water vapor content may cause numerical instabilities or reveal model limitations. As explained by Kopparapu et al. (2016) and Kane et al. (2018), usually, existing studies assume a runaway greenhouse when the simulation fails due to high water content and large differences between the OLR and the Incident Stellar Radiation (ISR). Some other studies found moist stables states (e.g., Popp et al. 2016), for which the planet can be in a stable regime with a wet troposphere and a global temperature lower than 350 K. This prevents any dramatic huge increase in the global temperature that would lead to a post-runaway state, thereby challenging the definition of the inner edge of the HZ, as well as the implications for habitability.

The original aspects of our work are rooted in our efforts to push the study beyond the tipping point and to describe the evolution of the climate across the runaway greenhouse transition in order to link the temperate and post-runaway states. We also explore the reversibility of the runaway greenhouse and we discuss the possible existence of an intermediate moist greenhouse state. We describe inherently 3D processes in detail to have a complete and accurate overview of the evolution of the atmosphere, mainly through its enrichment in water vapor and the consequences of the modification of the cloud coverage. In Sect. 2, we describe the method followed to explore the runaway greenhouse transition, as well as the GCM we use and the considered numerical setups. We present our results in Sect. 3 and we discuss the limits and the prospects of this work in Sect. 4.

2 Modeling the climate

2.1 Modeling the runaway greenhouse

Our method is similar the one used by previous works that studied the onset of the runaway greenhouse with GCMs (Yang et al. 2013; Leconte et al. 2013; Wolf & Toon 2014, 2015; Popp et al. 2016; Kopparapu et al. 2017). First, we modeled a temperate planet similar to present-day Earth, then we increased (step by step) the insolation (ISR) to find the tipping point at which the well-known positive feedback arises. We kept the shape of the stellar spectrum unchanged when we modified the insolation. The novelty of this work is in our modeling of the climate beyond the onset of the runaway greenhouse, that is, for a higher temperature and water vapor mixing ratio. When the insolation is high enough to initiate a runaway greenhouse, we fixed it to let the ocean evaporate in order to study the evolution of the climate during this critical transition phase. The insolation and the global surface temperature for which this threshold arises depend on the configuration of the simulation (e.g., presence of continents, CO2 content). Our aim is not here to find the exact value of the insolation corresponding to the tipping point, but to study the differences among the climate patterns. It could be interesting to find the exact tipping point, as done by Leconte et al. (2013) or Kopparapu et al. (2017), for example, for various atmospheric compositions; however, this is not the aim of this study. We performed simulations following this method for different atmospheric configurations: with and without Earth’s continents, as well as with and without CO2 and for different N2 pressures (see Table 1). In this work, our reference simulation is a waterworld, including 1 bar of N2, without CO2 (W1), as a representative example of the evolution of the climate during the runaway transition. Other simulations (presented in Table 1) are discussed in the context of a sensitivity study.

The key process allowing a runaway greenhouse is the evaporation of the surface ocean. To model the ocean, we used a two-layer slab ocean without heat transport, coupled to the atmosphere (see Sect. 2.2). Evaporating the totality of the Earth’s oceans will induce a vapor surface pressure of 273 bar. Regarding the large heat capacity of the water vapor, any temperature modification will be extremely slow in this situation, making such a simulation unachievable from a numerical point of view. Moreover, the timescale required to evaporate enough water to reach this pressure are also hugely long. To circumvent that difficulty, we choose to stop the evaporation process when a given quantity of water is evaporated (see the evaporation phase described in Sect. 3.1). For numerical reasons, we only tested two different evaporated water vapor pressures: 1 and 1.5 bar, which correspond to 10 and 15 m of a global layer equivalent (GEL), respectively; namely, the depth of the equivalent liquid ocean covering the entire planet. Then we removed the ocean from the simulation (i.e., we turned off the surface evaporation and consistently adjusted the surface albedo and the thermal inertia) and we set a new simulation by using the final state of the evaporation phase as the initial state (this is the dry transition phase described in Sect. 3.2). This method allowed us to study the impact of the size of the water reservoir on the obtained post-runaway state. In addition, based on the conclusions of Chaverot et al. (2022) on the radiative transfer for non-diluted water atmospheres, we used opacity data and took into account the transition of broadening species from nitrogen-dominated to water-dominated.

Table 1

Parameters of the simulations.

2.2 Numerical setup

The Generic-PCM (G-PCM) (e.g., Wordsworth et al. 2011; Leconte et al. 2013; Turbet et al. 2016, Turbet et al. 2018, 2021; Fauchez et al. 2019), previously known as the LMD Generic GCM, has been adapted to model water-rich atmospheres (Leconte et al. 2013). This makes it particularly appropriate for studying the runaway greenhouse. The horizontal resolution is 64×48 (longitude × latitude) with 30 vertical levels from 0.1, 1, or 10 bar (depending on the simulation setup; see Table 1) to 10 Pa. We consider the Earth’s obliquity and all the simulations are initialized assuming summer in the northern hemisphere. The dynamical time step is equal to 90 seconds. The physics (evaporation, convection, etc.) is computed every 15 min while the radiative transfer is computed every hour. The radiative transfer, assuming a Sun-like star, is done using a correlated-k table from Chaverot et al. (2022) for the mixture H2O+N2 and from Leconte et al. (2013) for the mixture including 376 ppm of CO2. The absorption line-list are taken from HITRAN 2012 (Rothman et al. 2013) and HITRAN 2016 (Gordon et al. 2017), respectively. We include also the N2-N2 collision-induced absorption (CIA) from the HITRAN CIA database (Karman et al. 2019). The H2O-H2O and H2O-air continua are from version 3.3 of the MT_CKD database (Mlawer et al. 2012). In our simulation, water is a variable gas which may condense or evaporate from the surface ocean, but this is also the case in the atmosphere, and the relative humidity is computed consistently. The subgrid-scale dynamical processes (turbulent mixing and convection) as well as the moist convection scheme are described in Leconte et al. (2013). The number of cloud condensation nuclei (CCNs) per unit mass of moist air used is fixed: 4 × 106 kg−1 for liquid water clouds and 2 × 104 kg−1 for ice clouds, while their radius is variable (see Leconte et al. 2013 for more details).

The model includes also a two-layer slab ocean without heat transport (Codron 2012; Charnay et al. 2013). The thermal inertia is equal to 18 000 Js−1/2 m−2 K−1 for oceans and 2000 J s−1/2 m−2 K−1 for continents. The albedo of the ocean is 0.07 and varies for continents up to 0.35, following Leconte et al. (2013) The heat capacity (cp) and the mean molecular weight are consistently initialized for every setup, but they stay constant during the simulation, which means that we do not account for the modification induced by the addition of water vapor.

3 The runaway greenhouse transition

In this section, we present evolutions of the climate from temperate stable states to stable post-runaway states, following the method described in Sect. 2. First, we describe the onset of the runaway greenhouse (Sect. 3.1), called the “evaporation phase” here, in reference to the enrichment of the atmosphere in water vapor. Second, we describe the “dry transition phase” (Sect. 3.2) for which the ocean is considered entirely evaporated.

3.1 Onset of the runaway greenhouse

3.1.1 Evolution of the OLR during the onset

We performed simulations for the different configurations presented in Table 1 to quantify the influence of continents and strong infrared (IR) absorbers such as CO2 on the onset of the runaway greenhouse.

Figure 1 provides an overview of the different simulations carried out in this work. The bottom panel is the OLR as a function of the mean surface temperature while the top panel indicates the different values of the insolation considered for every setup. In Fig. 1, we indicate the stable states with black dots; in other words, these are simulations for which an equilibrium is possible. The surface temperature of stable states is function of the insolation; therefore, we found different stables states by increasing (step by step) the insolation (colored dots on top panel of Fig. 1) as described in Sect. 2.1. When the insolation is high enough, a tipping point is reached and the atmosphere enters an unstable state (colored arrows on top panel of Fig. 1). Due to a large amount of water vapor, the bottom part of the atmosphere becomes optically thick to the thermal emission and the OLR is limited. Therefore, there is no longer any equivalence between the OLR and the Absorbed Stellar Radiation (ASR), as shown in Fig. 2, and the global temperature increases continuously due to this radiative imbalance (solid colored lines on the bottom panel of Fig. 1).

For every configuration presented in Fig. 1, the relation between the OLR and the mean surface temperature for stable states is roughly linear, which is consistent with previous studies (e.g., Leconte et al. 2013; Wolf & Toon 2015). For a given insolation, the presence of continents contributes to the reduction of the global surface temperature of few Kelvins because of their high albedo (higher than the deep ocean), as shown on the top panel of Fig. 1 (W1 versus E1 and WCO2 versus ECO2). For atmospheres with 1 bar of nitrogen (W1, E1, ECO2, and WCO2), the evolution of the OLR during the runaway transition (i.e., when the surface temperature reaches ~340 K) is not sensitive to the presence of continents and/or the presence of low concentration of CO2 for a given total surface pressure; that is, every simulation converges roughly on the same OLR curve. This is because when the surface temperatures are high enough, background gases become negligible compared to the radiative effect of high vapor pressures (Chaverot et al. 2022). Moreover, the radiatively active part of the atmosphere is decoupled from the surface conditions; therefore, the surface albedo (i.e., continents) does not influence the radiative balance any longer. For higher background gas pressures (W10 in Fig. 1), the onset of the runaway greenhouse arises at higher temperature and for a higher insolation (450 W m−2). For this extreme case with 10 bar of nitrogen, the temperature profile is closer to a purely dry adi-abatic profile, which is “less steep” than a wet adiabatic profile. Consequently, for a given global surface temperature, the atmosphere is colder and, thus, it is dryer. Therefore, even if there is less water in the atmosphere (vapor and clouds), the averaged OLR is lower, as shown in Fig. 1. This effect of high nitrogen pressures on the steepness of the temperature profile is also discussed in Chaverot et al. (2022), but they did not find smaller OLR values for high nitrogen pressure. This is probably linked to the pseudo-adiabatic assumption fixing the water vapor partial pressure as a function of the temperature. This assumption usually done in 1D models fixes the relative humidity to unity and underestimates the OLR (Ishiwatari et al. 2002; Leconte et al. 2013). On the other side, lower background gas pressures (0.1 bar of nitrogen; see W01 in Fig. 1) lead to a runaway greenhouse transition at a lower temperature, OLR value, and insolation (350 W m−2). A lower OLR bump for a low nitrogen pressure is consistent with results obtained by Chaverot et al. (2022) with a 1D model. The strong pressure broadening of H2O-H2O collisions makes the absorption more efficient than in the diluted case assuming 1 bar of nitrogen. The temperature profile is also closer to a wet adiabat which warms the stratosphere (compare to the dry adiabat case) and trigger the onset of the runaway greenhouse.

Finally, adding 376 ppm of CO2 shifts the onset of the runaway greenhouse to lower the insolation (375 W m−2 for ECO2 and WCO2 instead of 400 W m2 for E1 and W1) because of the strong greenhouse power of CO2. This drop is discussed using simpler 1D models as well (e.g., Nakajima et al. 1992; Goldblatt & Watson 2012; Goldblatt et al. 2013; Koll & Cronin 2019; Chaverot et al. 2022). For a given insolation value, the global surface temperature is also about 20 K higher in presence of CO2 (top panel in Fig. 1), which is consistent with respect to the greenhouse power of this gas.

During the runaway greenhouse, the insolation is kept unchanged to let the system evolve. The OLR reaches a maximum then decreases strongly. This is due to a combination of radiative effects and a quick modification of the cloud pattern. Spectroscopic processes are described by using a 1D radiative-convective model in Chaverot et al. (2022) and the modification of the cloud coverage is described in Sect. 3.1.3. In Fig. 1, the colored envelope delimiting the one-sigma variability of the OLR increases largely during the first steps of the runaway greenhouse. The continuous evaporation characteristic of this process densifies the cloud coverage. For medium temperatures (between 320 K and 360 K), the global dynamics induces variable patterns in this cloud coverage, thus strong variations of the spatial distribution of the OLR. When the global surface temperature is high enough to maintain a large amount of water vapor in the atmosphere (beyond 360 K), the cloud coverage becomes more homogeneous, thereby reducing the thermal emission windows as well as the OLR and its variability.

thumbnail Fig. 1

Evolution of the thermal emission during the onset of the runaway greenhouse. Bottom panel: values of the OLR function of the global surface temperature for each simulation setup described in Table 1. Colored dots are the stables states corresponding to various insolations, while the full curves are the instable state of the runaway greenhouse. Top panel: insolation values of each stable state. The colored arrows represent the insolation for which the runaway greenhouse arises for each setup. The OLR and temperature values are averaged over two years (for both the stable states and the runaway greenhouse) and the colored fills are the 1-sigma uncertainties due to variability on the OLR calculation.
thumbnail Fig. 2

Thermal emission and absorbed flux as a function of the mean surface temperature. Blue and purple solid lines are OLR and ASR, respectively, while the blue and purple dashed lines are their clear-sky equivalent (csOLR and csASR). Black dots are the stable states and the blue solid line is the runaway greenhouse for an insolation equal to 400 W m−2. The simulation setup is a waterworld with 1 bar of nitrogen without CO2 (W1). For comparison, the red line is the OLR computed with the 1D reverse version of the Generic-PCM which uses the same radiative transfer module and assume a fully saturated atmosphere (named the 1D reverse version of LMD-Generic in Chaverot et al. 2022). The green line is the OLR computed using a 1D reverse model assuming a subsaturated atmosphere (saturation mixing ratio less than 5%) from Goldblatt et al. (2013). The flux and temperature values are averaged over a two-year period and the colored areas are the 1 sigma uncertainties due to variability on the fluxes calculation.

3.1.2 Radiative imbalance

To study the evolution of the cloud coverage and its radiative effects, we compare in Fig. 2 the OLR and the ASR to their clear sky equivalent (csOLR and csASR). To compute the csOLR and csASR, we ran a simulation, including clouds but we neglect their radiative effect to estimate the emitted and absorbed fluxes. On the same figure, the red curve correspond to OLR values obtained using the 1D cloud-free version of the Generic-PCM1 described in Chaverot et al. (2022), with an atmosphere including H2O and 1 bar of N2 and a relative humidity value fixed to unity. The aim is to compare 1D and 3D models sharing the same radiative transfer calculation.

At low and high temperatures, respectively, when the atmosphere is nitrogen- or water-dominated, csOLR values from 1D and 3D simulations tend to converge. As there are no clouds in the 1D simulation, the only remaining difference is the absence of dynamics in the 1D model. Therefore, at low temperatures, the atmosphere is very thin and the csOLR is not much affected by the global dynamics. In other words, it is close to the Planck function. On the other hand, at high temperatures, the water vapor is well distributed around the planet and the csOLR is not affected by the dynamics. For intermediate temperatures and especially for the onset of the runaway greenhouse when the evaporation begins, the distribution of the vapor is not homogeneous and the 1D version of the Generic PCM largely underestimates the global OLR. As explained in Ishiwatari et al. (2002) and Leconte et al. (2013), this underestimation of the OLR of 1D models assuming a saturated atmosphere is due to the downward branches of the Hadley cells which reduce the relative humidity around the tropics, creating radiative windows by reducing the quantity of water vapor. To circumvent this issue, 1D models with variable humidity ought to be preferred (e.g., Ramirez et al. 2014). In Fig. 2, the green curve which is from Goldblatt et al. (2013) is obtained by using a cloud free 1D model assuming a saturation mixing ratio of less than 5%. This allows us to fit accurately csOLR from GCM simulations for temperate states. Even if some differences remain for the runaway greenhouse, the shape of the csOLR (and particularly the bump) is quite well represented. Thus, the major difference between results from 1D and 3D simulations is induced by the evolution of the cloud coverage. The differences remaining between Goldblatt et al. (2013, green curve) and the csOLR could come from dynamics, the different opacity database used or a mix of the two. Indeed, Goldblatt et al. (2013) used HITEMP, whereas we use HITRAN in this work.

The difference between the OLR and csOLR curves in Fig. 2 is only due to the radiative effect of the clouds. While this difference is rather constant for stable states, it expands during the runaway greenhouse. This is due to the strong evaporation, which allows for a densification and a drift of the cloud coverage toward the upper layers of the atmosphere (see Sect. 3.1.3), thereby increasing their radiative effect (see also Fig. 11, which shows the usual cloud forcing metrics). This is visible through the sharp increase in the Bond albedo and the drop in the ASR (pink solid line in Fig. 2). Nevertheless, as the OLR decreases as well, the effective flux (Seff=ASR/OLR) is greater than unity and the atmosphere warms-up. The energy imbalance is constant beyond 390 K (≈12 W m−2) and never exceed ≈25 W m−2, while it reaches ≈150 W m−2 in Kopparapu et al. (2017) for a global surface temperature equal to 350 K around a 3700 K star. We found that this energy imbalance may reach higher values for higher insolations (e.g., ≈80 W m−2 for ISR=475 W m−2 instead of ISR=400 W m−2 for the reference simulation W1).

During the runaway greenhouse, the difference between the csOLR and the csASR increases, highlighting the increase of the water vapor content in the atmosphere and, thus, of its absorption. This means that the OLR drops due to a combined absorption induced by the clouds and by a efficient horizontal homogenization of the increasing content of water vapor in the atmosphere. As the csASR is roughly constant during the runaway greenhouse, the drop in the ASR is mainly due to the increase of the Bond albedo due to the evolution of the cloud coverage.

3.1.3 Evolution the cloud coverage

As explained in Sect. 3.1.2, during the runaway greenhouse, the thickness of the cloud coverage grows (due to a strong evaporation) and the altitude of the cloud deck drifts toward the upper layers of the atmosphere. This effect is visible through the evolution of the mean vertical profile of the cloud coverage as a function of the temperature (given in Fig. 3). The increase in the surface pressure through the evaporation of the ocean is also visible on the same figure. The same evolution of the clouds has been observed by Wolf & Toon (2015) using CAM4. The water cloud content is constant in the bottom part of the atmosphere while the top-of-atmosphere layers are enriched from ≈10−7 to ≈10−1 kg m−3. In the same way, the stratosphere becomes wetter with the augmentation of temperature. At 400 K the vapor profile is quasi-constant and tends toward 1 kg kg−1 (see Appendix A for more details). This tendency is consistent with the results obtained by Turbet et al. (2021) studying the runaway greenhouse with hot and water-dominated initial conditions. As the quantity of clouds at the top of the atmosphere increases, the Bond albedo increases (from 0.2 to 0.45), inducing a drop in the ASR, as discussed in Sect. 3.1.2.

The day-to-night distribution of the clouds is not homogeneous during the runaway greenhouse, as shown on the latitude-longitude maps averaged over a year and centered at the substellar point presented in Fig. 4. The pattern observed at 400 K (right panel on the figure) is similar to that observed and discussed in Turbet et al. (2021) for hot and steam post-runaway atmospheres – except that in this case, a liquid surface ocean is still present and, thus, the albedo on the day side is not strictly zero. The ocean allows cloud formation on the day-side, even if they are mainly located on the night-side. The cloudy night-side provides an efficient greenhouse effect, contributing to the warming of the atmosphere. While it is non-existent for temperate stable states, the contrast of this specific pattern is amplified during the evaporation of the ocean. In this work, we assume a rapidly rotating Earth-like planet which induces an uniform cloud coverage of temperate stables states. Previous studies (Yang et al. 2013; Kopparapu et al. 2017) have shown that slow-rotating planets preferentially form cloud at the sub-stellar point for temperate stables states, preventing an intense warming and delaying the runaway greenhouse thanks to a high albedo on the day-side. These two different scenarios are discussed in Turbet et al. (2021). The evaporation of the clouds at the sub-stellar point during the onset of the runaway greenhouse described in detail in this section was also observed by Kopparapu et al. (2017) for planets orbiting various stellar types.

thumbnail Fig. 3

Map of the evolution of the mean vertical profile of the total water cloud (liquid + ice) content as a function of the global surface temperature during the evaporation phase. The simulation setup is the waterworld with 1 bar of nitrogen without CO2 (W1). The profiles are averaged over one year.

3.1.4 Atmospheric circulation

The global circulation of the atmosphere changes during the runaway greenhouse. The usual Hadley cells are shown in Fig. 5 for temperate states (bottom left panel), with an upward branch at the equator and downward branches at the tropics. This terrestrial circulation disappears during the runaway greenhouse while a new circulation arises at the top of the atmosphere (bottom right). As the water vapor pressure increases in the bottom part of the atmosphere, the opacity increases and there is less and less stellar flux reaching the surface. Because of this, the temperature is more homogeneously distributed around the planet, which has a decreasing effect on the surface wind jets. On the other side, as the absorption in the upper layers rises, the radiative heating strongly increases, inducing strong meridional jets at the top of the atmosphere. This stratospheric circulation characteristic of the runaway greenhouse and decorrelated from the surface conditions is visible in every simulation we performed. The same circulation was found by Turbet et al. (2021) for similar conditions and by Fujii et al. (2017), using ROCKE-3D, for water-dominated tidally locked planets.

As described in Sect. 2, we used a slab ocean without heat transport. For simulations without continents, we observe a hemispheric difference of the average annual temperature; more precisely, the northern hemisphere is warmer. This hemispheric difference can also be seen in Fig. 4, where the cloud coverage is slightly more important for the two left panels. This induces an atmospheric circulation around the equator along the north-south direction that is visible on the two first panels of Fig. 5. In our simulations, the atmosphere is initially water vapor-free and the surface temperature is homogeneous. Due to the non-zero obliquity of the planet (see Table 1), heating and intense evaporation happen alternatively in both hemispheres. For the waterworld setup, the evaporation of the ocean is continuous over time, unlike the Earth’s continent setup for which the evaporation rate over time is perturbed by the presence of continents during the rotation of the planet. Consequently, for a given insolation, the atmosphere of the waterworld hosts more water vapor. In addition to this, the non-symmetric evaporation due to the obliquity of the planet induces a difference of the opacity between the two hemispheres. Therefore, this creates a north-south forcing of the heat transport and one Hadley cell overrides the other. The downward branch of the Hadley cell reduces the relative humidity, as described in Leconte et al. (2013), which allows locally a larger thermal emission, thus preventing the evaporation. This induces a feedback which maintain this specific circulation. This modification of the Hadley cells is not function of the seasons and seems stable along the entire year. When the runaway greenhouse occurs, as the physics is de-coupled from the surface this circulation weakens to converge on a unique stratospheric circulation pattern (last two panels in Fig. 5, see also Fujii et al. 2017)

This non-symmetric temperate circulation seems stable in the simulation W1 as the zonal averaged temperature is constant over time as well as the water vapor pressure. However, as the numerical integration time is limited by our resources, a very long equilibrium time (longer than a hundred years) should not be excluded. Even if this state seems stable, it is function of the initial conditions of the simulation. The north to south hemispheric circulation shown in Fig. 5 (top left panel) corresponds to a simulation initialized assuming summer in the northern hemisphere. By initializing the simulation assuming summer in the southern hemisphere, the obtained circulation is reversed, meaning that the dominant wind is from south to north. For simulations with continents (E1 and ECO2), the evaporation rate and the cloud distribution are less homogeneous along the longitude than for their waterworld equivalents (W1, WCO2). This disturbs the feedback described above, thus allowing more symmetrical circulation around the equator that is similar to Hadley cells we see on Earth. Moreover, a symmetrical circulation is observed for a planet without obliquity as well as for low insolations (i.e., global surface temperatures below 280 K). This confirms that the modified circulation can happen preferentially on a waterworld, and it is induced by combining obliquity and high evaporation rates due to an insolation higher than the one of the Earth. Nevertheless, the sensitivity of this particular circulation needs to be deeply investigated to understand the contribution of each processes on it, as well as its stability criteria. Moreover, the oceanic circulation could largely affect it.

thumbnail Fig. 4

Latitude-longitude map of the clouds (liquid + ice) distribution during the evaporation phase. The maps were calculated in the heliocentric frame, that is keeping the subsolar point at 0° longitude and 0° latitude, and using an annual average. The simulation setup is the waterworld with 1 bar of nitrogen without CO2 (W1).
thumbnail Fig. 5

Evolution of the wind circulation during the runaway greenhouse. The color map represent the zonal wind (positive values represent eastward winds) and the arrows the vertical and meridional winds averaged over 1 year. The first row is the evolution of the winds for the waterworld setup (W1) and the bottom row is considering the Earth’s continents (E1), both without CO2. The first panel of each row is the warmer stable state obtained for the two considered setups at I = 375 W m2. Others show the evolution during the runaway greenhouse for different temperatures. Wind speeds corresponding to the length of the arrows vary with temperature.

3.1.5 Reversibility of the runaway greenhouse

In each simulation presented above, we studied the evolution of the atmosphere during the warming phase of the runaway greenhouse transition. We performed a set of simulations, presented in Fig. 6, where we started from a planet in runaway and reduce the insolation it receives in order to condense back the vapor into the surface ocean. First, a simulation following the method described in Sect. 2 was performed in order to partially evaporate the ocean (blue curve in Fig. 6; also see W1 in Table 1). This induced an increase of the global surface temperature, symbolized by the blue arrow in Fig. 6. Then we selected a runaway greenhouse state with a global temperature equal to 369 K as initial state for a second set of simulation (shown as a black star in Fig. 6). For this second step, we decrease the insolation compared to the one required to trigger a runaway greenhouse transition (ISR=400 W m−2) in order to find the insolation required to condense back the vapor into the surface ocean (i.e., to cool down the planet). The bottom panel of Fig. 6 presents the obtained OLR curves for different insolations (yellow: ISR=300 W m−2, orange: ISR=325 W m−2, brown: ISR=350 W m−2, red: ISR=375 W m−2). Colored arrows on the two panel represent the tendency evolution of the global surface temperature for each corresponding insolation (cooling or warming).

The non-zero difference between the OLR and the ASR during the runaway (see Fig. 2) suggests that a small decrease of insolation is not enough to condense back the water vapor and cool the planet. While ISR=400 W m−2 is required to initiate the runaway greenhouse for the W1 setup, ISR=375 W m−2 is not low enough to condense back the water vapor onto the surface starting at a global surface temperature of 369 K. An insolation equal to 350 W m−2 is able to cool down the planet, converging on a temperate stable state. For every cooling simulation, we recovered the same stable stables as those found by increasing step-by-step the insolation (see Sect. 2.1). We show in Fig. 6 that during this cooling phase, the OLR increases following the same pathway than the warming W1 simulation. This is also true for the evolution of the albedo and the cloud coverage. This resilience of the runaway greenhouse is visible on the top panel of Fig. 6 through the hysteresis loop on the insolation.

thumbnail Fig. 6

Evolution of the thermal emission when water vapor condense back into the surface ocean. The bottom panel is the OLR function of the mean surface temperature. The top panel is the insolation of each simulation. The blue curve is the warming simulation presented in Fig. 1 with an insolation equal to 400 W m−2. Black dots are stable states assuming a two-year average. Other colored curves (yellow, orange, brown, and red) are the evolution of the OLR for lower insolations and assuming a initial temperature equal to 369 K (the black star). Arrows represent the tendency of the evolution of the global surface temperature (warming or cooling) for different insolations. Colored crosses represent the stable stable found for cooling simulations (ISR=300 W m−2, 325 W m−2, and 350 W m−2, respectively). The top panel highlight the hysteresis loop induced by the clouds feedback during the runaway greenhouse. However, the bottom panel shows that there is no dependence of the OLR on the insolation for a given surface temperature. The simulation setup is a waterworld with 1 bar of nitrogen without CO2 (WCO2).

3.2 Runaway greenhouse dry transition

As explained in Sect. 2.1, to study the transition of the runaway, we arbitrarily remove the surface ocean when the surface water vapor pressure reaches some threshold values (1 and 1.5 bar). In the evaporation phase, we evaporate up to 2.6 bar of water but we were not able to model the dry transition phase for vapor pressure higher than 1.5 bar because of numerical instabilities. High vapor pressures induce a huge radiative imbalance leading to numerical failures similar to those discussed by Kopparapu et al. (2017).

We did two simulations for two different vapor surface pressures (1 bar and 1.5 bar corresponding to 10 and 15 m of GEL) in order to explore the impact of the water reservoir on the obtained post-runaway states. We used the corresponding runaway greenhouse states as initial conditions then we remove the surface ocean from the simulation and we let the system evolve by keeping the atmospheric composition unchanged, as well as the insolation. As there is no longer any humidity source, the temperature profile of the atmosphere changes. More precisely, there is a transition from a previously wet to a dry adiabatic profile in the bottom part of the atmosphere (as seen in Boukrouche et al. 2021 with a 1D model) as visible in Fig. 7, thus we choose to name this second simulation phase the “dry transition.” As the pressure-temperature profile of a dry adiabat is steeper than a wet one, the global surface temperature rises quickly. During the evaporation phase, the large thermal inertia of the ocean produces a cooling in the bottom part of the atmosphere preventing a quick increase of the surface temperature. In the dry transition, such forcing does not exist allowing a strong and fast increase of the global surface temperature. For the evaporation phase (W1 simulation, ISR=400 W m−2), more than a hundred years are needed to increase the temperature by about 80 K (i.e., evaporating 1.5 bar of water); whereas for the “dry” phase, only few tens of years are needed to increase the temperature to ≈400 K.

Figure 8 shows the evolution of the total water cloud content (liquid + ice) during the dry transition. We see re-evaporation taking place in the bottom part of the cloud layer. This contributes to the enrichment of the upper part of the atmosphere in water vapor. Moreover, due to the high temperatures at the surface, the water vapor migrates towards the upper layers and the vapor mass mixing ratio profile tends toward a constant value equal to 0.5 kg kg−1 (see Appendix A for more details). We notice also a slightly increase of the cloud coverage at the very top of the atmosphere.

Figure 9 gives the evolution of the OLR as a function of the global mean surface temperature. The evaporation phase (below 420 K) corresponds to the simulation W1 presented in Fig. 1. The blue dashed and dashed-dotted lines are respectively the dry transitions when 1 bar and 1.5 bar of water vapor is evaporated. The red curve is taken from Turbet et al. (2021) assuming 10 bar of vapor, 1 bar of nitrogen and no CO2. Black dots are the different stables states obtained in each cases, and for different insolations. Blue and red curves are computed using different methods. Blue curves are obtained by increasing the insolation from temperate stable states (warming simulation) while for the red curve, the insolation is reduced step-by-step to find the tipping point for which condensation occurs (cooling simulation). We show that the size of the water reservoir determines the equilibrium temperature of the post-runaway state (black dots beyond 800 K in Fig. 9), as suggested by 1D models (Turbet et al. 2019, Turbet et al. 2021; Boukrouche et al. 2021). Dotted lines are reference OLR curves computed using different 1D models and for different atmospheric compositions. Cyan, orange, and green dotted lines correspond to a pure water atmosphere (Goldblatt et al. 2013), a pre-industrial Earth (Kopparapu et al. 2013), and an atmosphere containing 1 bar of nitrogen without CO2 (Chaverot et al. 2022), respectively. All these models consider a fully saturated atmosphere (see Sect. 3.1.2 for more details about differences between 3D and 1D saturated and sub-saturated models).

The reason for which we consider small amounts of water (1 and 1.5 bar) come from the fact that the addition of vapor in the atmosphere increases its thermal inertia, thereby decreasing the evaporation rate. Due to numerical constraints, we were not able to evaporate more than 2.6 bar of water vapor in our simulations. Moreover, we were not able to model the dry transition phase itself for vapor pressures beyond 1.5 bar because of numerical instabilities. Indeed, for such pressures, the radiative unbalance skyrockets inducing numerical failures similar to those described by Kopparapu et al. (2017) for the onset of the runaway greenhouse. There is no strong evolution of the wind pattern during the dry transition. The day-to-night cloud dichotomy discussed in Sect. 3.1.3 remains during this phase (as shown in Fig. 10) and becomes even more contrasted. It is accentuated by the re-evaporation of the clouds to converge on the exact same pattern, as described in Turbet et al. (2021), with a purely cloud-free day-side. OLR values obtained in this work are also consistent with those obtained for different insolations in Turbet et al. (2021). Moreover, even with a quantity as high as 10 bar of water vapor in the atmosphere, they obtain global circulation, temperature, and cloud distributions strongly similar to those presented here. Therefore, we can reasonably expect the same climate pattern and evolution for post-runaway states with higher water pressures (few tens of bar).

Figure 11 shows the evolution of the cloud forcing during both phases of the runaway greenhouse. For temperate habitable states, the cooling effect of the clouds in the shortwave (blue curve) is more efficient than the warming due to their greenhouse effect in the longwave (red curve). During the evaporation phase, as the cloud deck densities, the Bond albedo increases (Fig. 12) and the colling effect of the clouds is even more efficient (blue curve in Fig. 11). However, when the surface ocean is considered entirely evaporated (vertical grey dotted lines at 383 K and 393 K corresponding, respectively, to 1 bar and 1.5 bar), after a strong increase of the global temperature, the net radiative effect of the clouds switches from cooling to warming (green curve) around 700 K. We show that during the dry transition phase, as the Bond albedo decreases (see Fig. 12 and blue curve in Fig. 11), the greenhouse effect of clouds becomes the dominant process and the net effect of the clouds tends to warm the planet (Fig. 11). In other words, the thick layer of clouds at the top of the atmosphere acts as a shield that retains the thermal emission of the planet. The cloud forcing of the obtained post-runaway states is very similar to the one from Turbet et al. (2021), even though those authors considered ten times more water vapor. This suggest that 1 bar of vapor is enough to stabilize the climate in a post-runaway configuration, with a thick cloud deck at the top of the atmosphere, preferentially on the night side of the planet. The temperature for which the net effect of the clouds switch from cooling to warming is concomitant with the increase of OLR and function of the water vapor content. More vapor and clouds will require a higher temperature to evaporate and modify the cloud albedo. The 50 K gap between the evaporation phase and the dry transition in Figs. 9, 11, and 12 is due to the very fast evolution of the system during the first stages of the simulation.

thumbnail Fig. 7

Temperature profile for different surface temperature during the dry transition with 1 bar of water vapor. Arrows indicate the transition between a dry adiabat in the lower part of the atmosphere (high pressures) and a wet one in the bottom part (low pressures) for the coldest (blue) and hottest (red) cases. Dry region is below the transition while moist one is above (Abe & Matsui 1988; Kasting 1988). The simulation setup is the waterworld with 1 bar of nitrogen without CO2 (W1).
thumbnail Fig. 8

Map of the evolution of the mean vertical profile of the total water cloud (liquid + ice) content as a function of the global surface temperature during the dry transition phase. The simulation setup is the waterworld with 1 bar of nitrogen without CO2 (W1) assuming 1 bar of water vapor. The profiles are averaged over one year.
thumbnail Fig. 9

Evolution of the OLR as a function of the mean surface temperature from temperate stable states to post-runaway stable states. For blue curves, the simulation setup is the waterworld, 1 bar of nitrogen without CO2 (W1). The red curve is taken from Turbet et al. (2021) assuming 10 bar of water vapor, 1 bar of nitrogen and no CO2. The insolation of the evaporation phase and the dry transition is equal to 400 W m−2. The vertical grey lines are the temperatures for which 1 bar and 1.5 bar of water vapor are evaporated. Stable post-runaway states from Turbet et al. (2021) assume different insolations. The OLR values are averaged over 2 years which explain the gap between the evaporation and the dry transition phases. Cyan, orange and green dotted lines are OLR curves from reference works using 1D models and assuming 273 bar of water (pure water from Goldblatt et al. 2013, Earth’s case from Kopparapu et al. 2013 and with 1 bar of N2 from Chaverot et al. 2022, respectively).
thumbnail Fig. 10

Latitude-longitude map of the cloud (liquid + ice) distribution during the dry transition phase with 1 bar of vapor. The maps were calculated in the heliocentric frame that is keeping the subsolar point at 0° longitude and 0° latitude, and using an annual average. The simulation setup is the waterworld assuming 1 bar of nitrogen without CO2 (W1).
thumbnail Fig. 11

Evolution of the cloud forcing during the runaway greenhouse. Solid lines are the evaporation and dry transition phases while black dots are stable states assuming a two-year average. Negatives values indicate a cooling effect of the clouds and positive ones a warming effect. In the longwave (blue curves), clouds have a cooling effect due to their high albedo reflecting the incoming stellar flux back to space. The red curve correspond to the warming contribution due to the greenhouse effect of the clouds (i.e., thermal mission absorbed by the clouds). The green curve is the net effect of the clouds, including shortwave and longwave contributions. The vertical grey lines are the temperatures for which 1 bar and 1.5 bar of water vapor are evaporated. These are used as initial conditions for the dry transition phase (colored dotted lines). Colored triangles are the values obtained in Turbet et al. (2021) for the same insolation (400 W m−2), but including 10 bar of water vapor. The fluxes values are averaged over two years, which explain the gap between the evaporation and the dry transition phases. The simulation setup is the waterworld with 1 bar of nitrogen without CO2 (W1).
thumbnail Fig. 12

Evolution of the bond albedo as a function of the mean surface temperature from temperate stable states to post-runaway stable states. The simulation setup is the waterworld with 1 bar of nitrogen without CO2 (W1). The insolation of the evaporation phase and the dry transition is equal to 400 W m−2. The vertical grey lines are the temperatures for which 1 bar and 1.5 bar of water vapor are evaporated. The albedo values are averaged over 2 years which explain the gap between the evaporation and the dry transition phases. Pink squares and green triangles are steady states of the Earth’s case from Leconte et al. (2013) and Wolf & Toon (2015) respectively. Dotted lines are albedo curves from reference works using 1D models (1 bar of N2 from Goldblatt et al. 2013 and Earth’s case from Kopparapu et al. 2013).

4 Discussion

4.1 Literature comparison

We show in Sect. 3.1.2 that the radiative imbalance is constant during all the evaporation phase. Similarly to Leconte et al. (2013), we did not find any moist stable states; whereas Wolf & Toon (2015) did, using CAM4, or Popp et al. (2016), using ECHAM6. There are multiple discussions in the literature around this open question (e.g., Kopparapu et al. 2017; Yang et al. 2019), some pointing out that the specific humidity of the stratosphere is a key factor in explaining this difference. Indeed, the GCM inter-comparison done by Yang et al. (2019) highlights that the Generic-PCM2 tends to be wetter than CAM4 for planets around G-type stars. In the same time, the cloud fraction and the cloud thickness is higher in CAM4 inducing a cooling of the planet. This was discussed in Kopparapu et al. (2017), showing that Wolf & Toon (2015) found higher stratospheric temperatures than Leconte et al. (2013) for a same surface temperature. High relative humidity in the stratosphere reduces the thermal emission from the upper layers, allowing a larger difference between the OLR and the ASR. This is the commonly accepted answer to explain the existence (or not) of a moist stable state. More inter-comparisons based on the method used in Yang et al. (2019), as well as high-resolution cloud resolving simulations (e.g., Lefèvre et al. 2021), should be done to solve this question. By using a 1D climate model, Popp et al. (2015) suggests that the convection scheme could be the answer to explain potential moist runaway greenhouse. The number of Cloud Condensation Nucleii (CCN) present may also impact the cloud formation thus the possibility of a moist state.

We compare our results for the Earth case with pre-industrial CO2 quantities (ECO2), with major results from the literature (Leconte et al. 2013; Wolf & Toon 2015) in Fig. 13. We do not show a direct comparison of the ASR, but we compare the Bond albedo from simulation W1 with Leconte et al. (2013) and Wolf & Toon (2015) in Fig. 12. As shown by Fig. 13, we found slightly different global surface temperature and OLR values compared to Leconte et al. (2013) and Wolf & Toon (2015), even for a same insolation. This is not surprising regarding numerous differences between the models highlighted by Yang et al. (2019). In Leconte et al. (2013), they used the Generic-PCM as this work, but they mimic the ocean thank to a large thermal inertia of the ground; whereas, we used a more complex slab ocean (without heat transport). These aspects, combined with the development of the model, may explain the differences between our results – especially on the relation between insolation and the global surface temperature. Even if Wolf & Toon (2015) found a moist stable state, the evolution of the OLR as a function of the global surface temperature is qualitatively consistent with our simulations.

The red dashed line is taken from Goldblatt et al. (2013), corresponding to a pre-industrial composition with a subsatured atmosphere. This curve is comparable with our ECO2 case to estimate the impact of 3D processes. The constant offset of OLR values for temperate states between 1D and 3D simulations is due to the greenhouse effect of the clouds. In the same way, the large difference of OLR during the runaway greenhouse is due to the contribution of the thick cloud coverage. As discussed in several works (e.g., Ishiwatari et al. 2002; Abe et al. 2011; Leconte et al. 2013), reducing the relative humidity of 1D models can produce results in a good agreement with 3D modelling for temperate stable states. However, substantial differences remain for the runaway greenhouse as the key process is the evolution of the cloud coverage that is not estimable using 1D models. From a qualitative point of view, the shape of the OLR curve we show is similar to historical estimations of the thermal emission of an atmosphere composed of H2O and N2 (e.g., Nakajima et al. 1992), but the quantitative values are different because of the addition of more complex physical processes (non-grey atmosphere, clouds, global dynamics).

thumbnail Fig. 13

Literature comparison of the OLR and the insolation as a function of the temperature. The black dots are the stables states corresponding to various insolations, while the full curves are the unstable state of the runaway greenhouse. The top panel indicates the insolation values of each stable state. The colored arrows represent the insolation for which the runaway greenhouse arises for each setup. The OLR and temperature values are averaged over two years and the colored fills are the 1 sigma uncertainties due to variability on the OLR calculation. Pink squares and green triangles are steady states of the Earth’s case from Leconte et al. (2013) and Wolf & Toon (2015), respectively. The red dashed curve is a reference result from Goldblatt et al. (2013) for a pre-industrial atmosphere and using a 1D model, assuming saturation mixing ratio below 5%.

4.2 Limitations and prospects

The H2O+N2 correlated-k table used in this work was made by Chaverot et al. (2022) using the HITRAN database (Gordon et al. 2017). Goldblatt et al. (2013) showed that the HITEMP database is more accurate and complete than HITRAN for high temperatures and this impacts the estimation of OLRs and ASRs (see also Kopparapu et al. 2013 and Ramirez 2014). More precisely, they highlight using a 1D climate model, which HITRAN overestimates the albedo and thus underestimates the ASR for temperatures beyond ≈360 K. As in 3D GCM simulations, the relative humidity (thus the vapor partial pressure) is lower than in 1D models (Leconte et al. 2013), we can assume no significant difference in our simulations up to ≈400 K. Therefore, this does not affect much the evaporation phase but may change the dry transition phase and the final equilibrium temperature by slightly modifying the Bond albedo of the planet the heating rates. This could be investigated but we are confident on the global evolution described in this paper and on the reliability of general mechanisms which determine the climate pattern transition between pre- and post-runaway. We describe general processes and tendencies and our results are in accordance with previous works using different GCMs.

In every simulation, the heat capacity (cp) and the mean molecular weight are assumed to be constant. This is inaccurate when the temperature increases strongly (because of the dependence of cp on the temperature) and when the dominant gas changes (from nitrogen-dominated to water -dominated during the evaporation phase, see Fig. A.1 and Chaverot et al. 2022). Our version of the Generic-PCM does not include variable heat capacity or molecular weight, but ongoing developments will allow for studies of the impact of variations of these quantities in the near future. For reasons of consistency, we did not update the value of cp when we initialized the simulation of the dry transition phase.

In this work, we describe in detail the simulation with 1 bar of nitrogen without CO2 (W1) to explore the effect of the water without additional absorption sources (nitrogen is assumed to be optically thin, except for the CIA). Therefore, the minimal absorption given by the correlated-k table is for a water volume mixing ratio equal to 10−6. When the mixing ratio is lower (e.g., in the stratosphere at low temperature), the water vapor absorption is given by this threshold value. This approximation may induce an error for very low water content, but it becomes negligible when water is the dominant gas (i.e., during the runaway greenhouse, which we focus on in this work).

In Sect. 3.2, we describe a transition between a wet to a dry adiabat. This is true for low partial water pressures, but as shown by Selsis et al. (2023), it becomes wrong for pressures higher than roughly 20 bar (function of the considered host star). In this case, the temperature profile follows a radiative profile.

Finally, as we did not search for the exact tipping point triggering the runaway greenhouse, we probably overestimated the radiative imbalance by a few W m−2. This does not change the physics described in this work, however, a study focused on this could be interesting in the context of estimating the evaporation timescale of different water reservoirs.

5 Conclusions

We studied the runaway greenhouse with the 3D Generic-PCM in order to link temperate stables states to post-runaway stable states. The positive feedback of the runaway greenhouse is triggered by the strong absorption of the water vapor and driven by a large water inventory (i.e., surface ocean) available through evaporation. From terrestrial climate conditions, we increase the insolation step-by-step in order to find the onset of the runaway greenhouse, following the method used by previous works (Leconte et al. 2013; Wolf & Toon 2014; Popp et al. 2016). The originality of this work is to go beyond the onset and to study the transition toward post-runaway states3. We show that there are two phases:

  1. An “evaporation phase” in which the water vapor content and the cloud thickness increase rapidly.

  2. When the ocean is considered to be entirely evaporated, a “dry transition phase” begins, characterized by a switch from a wet adiabatic profile to a dry adiabatic profile. During this second step, the temperature increases quickly, then the climate converges onto a new stable state at high temperature, namely, a post-runaway state.

We describe the evolution of the composition of the atmosphere as well as the changes of the cloud pattern in Sect. 3.1. We show that even if the tipping point of the runaway greenhouse depends on the simulation setup (e.g., continents, with CO2 as a minor constituent), the evolution path during the transition does not, for a given background gas pressure. As the bottom part of the atmosphere is water-dominated, it becomes optically thick and the physics is decoupled from the surface conditions. In the same way, a low content of CO2 (376 ppm) does not influence the climate beyond the onset (Fig. 1).

The main characteristic of the first phase of the runaway greenhouse (evaporation phase) is the evolution of the cloud coverage. It grows because of the intense evaporation and it migrates toward the upper layers (Fig. A.1). This migration increases the Bond albedo creating a drop in the ASR (Fig. 2). The day to night distribution of the clouds is also affected. As shown by Fig. 4, the night-side is more cloudy than the day-side, which tends to increase the greenhouse effect of the clouds. This competition between the warming and cooling effect of the clouds is shown in Fig. 12. We highlight that during the entire evaporation phase, the net effect of the clouds tends to cool down the planet. However, the addition of water vapor due to evaporation creates an extra absorption able to overshoot this cooling, leading to a roughly constant radiative imbalance (Fig. 2) that continuously warms the planet. The general atmospheric circulation is strongly affected by the evaporation and the evolution of the cloud coverage. As the top part of the atmosphere becomes wetter, it absorbs more stellar flux and the heating rate rises creating an intense global circulation (Fujii et al. 2017). As visible in Fig. 5, strong stratospheric jets appear substituting the usual Hadley cells of the temperate Earth. We explore also the potential reversibility of the runaway greenhouse. The idea is to use a runaway state as a initial state, then to reduce the insolation in order to condense back the water in the surface ocean. We show in Fig. 6 that the radiative unbalance characteristic of the runaway greenhouse creates a hysteresis loop on the insolation making it resilient to any small variations of incoming flux. We also show that if a re-condensation happens, the OLR increases when the surface temperature decreases, following the exact same evolution pathway than during the evaporation phase.

In a second step, when a given quantity of water is evaporated (1 bar or 1.5 bar), we consider the ocean to have entirely evaporated and we remove it from the simulation (see Sect. 3.2). The absence of humidity sources at the bottom of the atmosphere induced a transition from a wet to a dry adiabatic profile (Fig. 7). Water vapor migrates towards the upper layers of the atmosphere and lower clouds evaporate (Fig. A.2). We also observed a more contrasted dichotomy between a cloudy night-side and a cloud free day-side (see Fig. 10,) which is consistent with the results from Turbet et al. (2021) on post-runaway states for similar atmospheres. Because of the re-evaporation of the clouds, the albedo decreases (Fig. 12) and the net effect amounts to a tendency to warm the planet (Fig. 11, see also Turbet et al. 2021). This reduces also the radiative unbalance, allowing the simulation to converge on a post-runaway state as shown in Fig. 9. This final state strongly depends on the quantity of water available in the atmosphere, in other words: on the size of the initial water reservoir.

Finally, by comparing our results to historical studies using 1D models, we show that 3D processes (clouds and dynamics) strongly impact the evolution of the climate during the first stages of the runaway greenhouse, as well as during the dry transition phase. This is of major importance for discussions of the potential observability of this kind of planet with reflected or emitted light.

Acknowledgements

This work has been carried out within the framework of the NCCR PlanetS supported by the Swiss National Science Foundation under grants 51NF40_182901 and 51NF40_205606. G.C. and E.B. acknowledge the financial support of the SNSF (grant number: 200021_197176 and 200020_215760). M.T. thanks the Gruber Foundation for its generous support to this research. M.T. acknowledges support from the Tremplin 2022 program of the Faculty of Science and Engineering of Sorbonne University. The authors thank the Generic-PCM team for the teamwork development and improvement of the model. The Generic-PCM (and documentation on how to use the model) can be downloaded from the SVN repository https://svn.lmd.jussieu.fr/Planeto/trunk/LMDZ.GENERIC/ (version 2521). More information and documentation are available on http://www-planets.lmd.jussieu.fr. The computations were performed at University of Geneva on the Baobab and Yggdrasil clusters. This research has made use of NASA’s Astrophysics Data System. We thank the two referees Ravi K. Kopparapu and Colin Goldblatt for their helpful comments.

Appendix A Evolution of water vapor and clouds during the runaway greenhouse

As discussed in Sect. 3, the two phases of the runaway greenhouse strongly modify the distribution of water vapor as well as the cloud coverage. In this section, we provide additional information concerning these quantities. During the evaporation phase, the evolution of the average water vapor profile (left panel of Fig. A.1) shows a strong enrichment of the upper layers of the atmosphere in water vapor from 1×10−13 kg kg−1 to 0.1 kg kg−1. This is coherent with the densification and the drift of the cloud coverage towards the stratosphere shown on the right panel of Fig. A.1 (this panel corresponds to the results presented in Fig. 3). This drift of the clouds is also discussed in Kopparapu et al. (2017).

thumbnail Fig. A.1

Evolution of the water vapor mass mixing ratio and of the mean vertical profile of the water cloud (liquid + ice) content during the evaporation phase. The simulation setup is the waterworld with 1 bar of nitrogen without CO2 (W1). The profiles are averaged over one year. The dashed black line corresponds to the temperate stable state with the highest insolation (ISR=375 W m−2).

During the dry transition, as there is no longer any moisture source from the surface and the surface temperature rises, leading to a migration of the water vapor and of the water clouds toward the upper layers of the atmosphere. The left panel of Fig. A.2 shows the evolution of the water vapor mass mixing ratio profile. We see that the bottom part of the atmosphere become dryer (i.e., the water vapor content decreases) and the mass mixing ratio profile tends toward a constant value up to 1×103 Pa. This is the pressure above which the pressure temperature conditions are favorable to the formation of clouds (see left panel of Fig. A.2 and Fig. 8). At the same time, the bottom part of the cloud layer evaporate due to the increase of surface temperature. This is clearly visible on the right panel in Fig. A.2 (this panel corresponds to the results presented in Fig. 8).

The latitudinal distribution of the clouds coverage during the runaway greenhouse is presented in Fig. A.3 for global surface temperatures of 320 K, 340 K and 370 K (panels A, B, and C, respectively). The setup is the Earth without CO2 (E1). While the usual cloud pattern we see on the Earth remains during the onset of the runaway greenhouse (panel A), it is modified when the water vapor content becomes sufficient to drift the cloud toward the upper atmosphere (panel B). When the stratospheric circulation described in Sect. 3.1.4 appears (panel C, which corresponds to a global surface temperature equal to 370 K), the cloud deck is located at the top of the atmosphere with a maximum density around the equator where the upward winds are the strongest. Cloud minima are around the tropics because of the downward winds (see Fig. 5 for the wind circulation).

thumbnail Fig. A.2

Evolution of the water vapor mass mixing ratio and of the mean vertical profile of the water cloud (liquid + ice) content during the dry transition phase. The simulation setup is the waterworld with 1 bar of nitrogen without CO2 (W1) assuming 1 bar of water vapor. The dashed black line corresponds to the post-runaway stable state obtained after the runway greenhouse transition.
thumbnail Fig. A.3

Evolution of the latitude-altitude map of the water cloud (liquid + ice) distribution during the evaporation phase. The simulation setup is the Earth with 1 bar of nitrogen without CO2 (E1). The maps are averaged over two years. The A, B and C panels correspond to global surface temperatures of 320 K, 340 K, and 370 K, respectively.

Appendix B Effect of the out-of-equilibrium evolution

By definition, the runaway greenhouse is an out-of-balance process making its analysis quite complex. In order to study the impact of the insolation on the evolution of the climate, we performed four different simulations with the W1 setup and various insolations, with all of them sharing the same initial conditions. Results are shown in Fig. B.1, where the left panel is the vertical water cloud profile, and the right panel is the temperature profile (both annually averaged). For every curve, the global surface temperature is equal to 370 K. For high insolations, the cloud coverage is less dense at equivalent global surface temperature (left panel) and where the temperature is lower in the upper layers (right panel). Due to highly imbalanced radiative conditions at high insolations, the evolution is so fast that there is a delay of the growth of the cloud coverage compared to the increase of global surface temperature. This affects the global evolution of the atmosphere by reducing the cooling effect of the clouds at a given surface condition. The enrichment in water vapor of the top part of the atmosphere is also delayed. However, as the cloud coverage is less dense at high insolations, the difference between cloud maxima at the equator and cloud minima around the tropics is less contrasted, smoothing the latitudinal distribution of both the water vapor and the temperature.

thumbnail Fig. B.1

Cloud (liquid + ice) profiles and temperature profiles during the runaway greenhouse when the global surface temperature is equal to 370 K, from simulations assuming different insolations and a common initial condition. The simulation setup is the waterworld with 1 bar of nitrogen without CO2 (W1). The profiles are averaged over one year.

References

  1. Abe, Y., & Matsui, T. 1988, J. Atmos. Sci., 45, 3081 [NASA ADS] [CrossRef] [Google Scholar]
  2. Abe, Y., Abe-Ouchi, A., Sleep, N. H., & Zahnle, K. J. 2011, Astrobiology, 11, 443 [CrossRef] [Google Scholar]
  3. Boukrouche, R., Lichtenberg, T., & Pierrehumbert, R. T. 2021, ApJ, 919, 130 [NASA ADS] [CrossRef] [Google Scholar]
  4. Charnay, B., Forget, F., Wordsworth, R., et al. 2013, J. Geophys. Res. Atmos., 118, 10414 [NASA ADS] [CrossRef] [Google Scholar]
  5. Chaverot, G., Turbet, M., Bolmont, E., & Leconte, J. 2022, A & A, 658, A40 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  6. Codron, F. 2012, Climate Dyn., 38, 379 [NASA ADS] [CrossRef] [Google Scholar]
  7. Fauchez, T. J., Turbet, M., Villanueva, G. L., et al. 2019, ApJ, 887, 194 [Google Scholar]
  8. Fujii, Y., Del Genio, A. D., & Amundsen, D. S. 2017, ApJ, 848, 100 [NASA ADS] [CrossRef] [Google Scholar]
  9. Goldblatt, C., & Watson, A. J. 2012, Philos. Trans. Roy. Soc. A: Math. Phys. Eng. Sci., 370, 4197 [NASA ADS] [CrossRef] [Google Scholar]
  10. Goldblatt, C., Robinson, T. D., Zahnle, K. J., & Crisp, D. 2013, Nat. Geosci., 6, 661 [NASA ADS] [CrossRef] [Google Scholar]
  11. Gordon, I. E., Rothman, L. S., Hill, C., et al. 2017, J. Quant. Spectrosc. Rad. Transf., 203, 3 [NASA ADS] [CrossRef] [Google Scholar]
  12. Hamano, K., Abe, Y., & Genda, H. 2013, Nature, 497, 607 [NASA ADS] [CrossRef] [Google Scholar]
  13. Ingersoll, A. P. 1969, J. Atmos. Sci., 26, 1191 [NASA ADS] [CrossRef] [Google Scholar]
  14. Ishiwatari, M., Takehiro, S.-I., Nakajima, K., & Hayashi, Y.-Y. 2002, J. Atmos. Sci., 59, 3223 [NASA ADS] [CrossRef] [Google Scholar]
  15. Kane, S. R., Ceja, A. Y., Way, M. J., & Quintana, E. V. 2018, ApJ, 869, 46 [NASA ADS] [CrossRef] [Google Scholar]
  16. Karman, T., Gordon, I. E., van der Avoird, A., et al. 2019, Icarus, 328, 160 [Google Scholar]
  17. Kasting, J. F. 1988, Icarus, 74, 472 [NASA ADS] [CrossRef] [Google Scholar]
  18. Kasting, J. F., Whitmire, D. P., & Reynolds, R. T. 1993, Icarus, 101, 108 [Google Scholar]
  19. Koll, D. D. B., & Cronin, T. W. 2019, ApJ, 881, 120 [Google Scholar]
  20. Komabayasi, M. 1967, J. Meteorol. Soc. Jpn. Ser. II, 45, 137 [NASA ADS] [CrossRef] [Google Scholar]
  21. Kopparapu, R. K., Ramirez, R., Kasting, J. F., et al. 2013, ApJ, 765, 131 [NASA ADS] [CrossRef] [Google Scholar]
  22. Kopparapu, R. K., Ramirez, R. M., SchottelKotte, J., et al. 2014, ApJ, 787, L29 [Google Scholar]
  23. Kopparapu, R. K., Wolf, E. T., Haqq-Misra, J., et al. 2016, ApJ, 819, 84 [NASA ADS] [CrossRef] [Google Scholar]
  24. Kopparapu, R. K., Wolf, E. T., Arney, G., et al. 2017, ApJ, 845, 5 [NASA ADS] [CrossRef] [Google Scholar]
  25. Leconte, J., Forget, F., Charnay, B., Wordsworth, R., & Pottier, A. 2013, Nature, 504, 268 [Google Scholar]
  26. Lefèvre, M., Turbet, M., & Pierrehumbert, R. 2021, ApJ, 913, 101 [CrossRef] [Google Scholar]
  27. Lovis, C., Blind, N., Chazelas, B., et al. 2022, Proc. SPIE, 12184, 121841Q [NASA ADS] [Google Scholar]
  28. Lustig-Yaeger, J., Meadows, V. S., & Lincowski, A. P. 2019, AJ, 158, 27 [NASA ADS] [CrossRef] [Google Scholar]
  29. Marconi, A., Abreu, M., Adibekyan, V., et al. 2022 Proc. SPIE, 12184, 1218424 [NASA ADS] [Google Scholar]
  30. Mlawer, E. J., Payne, V. H., Moncet, J.-L., et al. 2012, Philos. Trans. Roy. Soc. Lond. Ser. A, 370, 2520 [NASA ADS] [Google Scholar]
  31. Morley, C. V., Kreidberg, L., Rustamkulov, Z., Robinson, T., & Fortney, J. J. 2017, ApJ, 850, 121 [Google Scholar]
  32. Nakajima, S., Hayashi, Y.-Y., & Abe, Y. 1992, J. Atmos. Sci., 49, 2256 [NASA ADS] [CrossRef] [Google Scholar]
  33. Popp, M., Schmidt, H., & Marotzke, J. 2015, J. Atmos. Sci., 72, 452 [NASA ADS] [CrossRef] [Google Scholar]
  34. Popp, M., Schmidt, H., & Marotzke, J. 2016, Nat. Commun., 7, 10627 [NASA ADS] [CrossRef] [Google Scholar]
  35. Ramirez, R. M. 2014, PhD thesis, The Pennsylvania State University, USA [Google Scholar]
  36. Ramirez, R. M., Kopparapu, R. K., Lindner, V., & Kasting, J. F. 2014, Astrobiology, 14, 714 [NASA ADS] [CrossRef] [Google Scholar]
  37. Rothman, L. S., Gordon, I. E., Babikov, Y., et al. 2013, J. Quant. Spectrosc. Radiative Transf., 130, 4 [NASA ADS] [CrossRef] [Google Scholar]
  38. Selsis, F., Leconte, J., Turbet, M., Chaverot, G., & Bolmont, É. 2023, Nature, 620, 287 [NASA ADS] [CrossRef] [Google Scholar]
  39. Simpson, G. C. 1929, Q. J. Roy. Meteorol. Soc., 55, 73 [Google Scholar]
  40. Turbet, M., Leconte, J., Selsis, F., et al. 2016, A & A, 596, A112 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  41. Turbet, M., Bolmont, E., Leconte, J., et al. 2018, A & A, 612, A86 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  42. Turbet, M., Ehrenreich, D., Lovis, C., Bolmont, E., & Fauchez, T. 2019, A & A, 628, A12 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  43. Turbet, M., Bolmont, E., Ehrenreich, D., et al. 2020, A & A, 638, A41 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  44. Turbet, M., Bolmont, E., Chaverot, G., et al. 2021, Nature, 598, 276 [CrossRef] [Google Scholar]
  45. Way, M. J., Del Genio, A. D., Kiang, N. Y., et al. 2016, Geophys. Res. Lett., 43, 8376 [NASA ADS] [CrossRef] [Google Scholar]
  46. Wolf, E. T., & Toon, O. B. 2014, Geophys. Res. Lett., 41, 167 [Google Scholar]
  47. Wolf, E. T., & Toon, O. B. 2015, J. Geophys. Res. Atmos., 120, 5775 [NASA ADS] [CrossRef] [Google Scholar]
  48. Wordsworth, R. D., Forget, F., Selsis, F., et al. 2011, ApJ, 733, L48 [Google Scholar]
  49. Wunderlich, F., Godolt, M., Grenfell, J. L., et al. 2019, A & A, 624, A49 [CrossRef] [EDP Sciences] [Google Scholar]
  50. Yang, J., Cowan, N. B., & Abbot, D. S. 2013, ApJ, 771, L45 [Google Scholar]
  51. Yang, J., Leconte, J., Wolf, E. T., et al. 2019, ApJ, 875, 46 [NASA ADS] [CrossRef] [Google Scholar]
1

Named 1D reverse version of LMD-Generic in Chaverot et al. (2022).

2

Named LMDG in Yang et al. (2019)

3

The GCM outputs of the stable states (temperate and post-runaway) of each setup, as well as several snapshots along the runaway transition for setup W1 are available here: https://zenodo.org/record/8325631. Other data underlying this article will be shared on request to the corresponding author.

All Tables

Table 1

Parameters of the simulations.

In the text

All Figures

thumbnail Fig. 1

Evolution of the thermal emission during the onset of the runaway greenhouse. Bottom panel: values of the OLR function of the global surface temperature for each simulation setup described in Table 1. Colored dots are the stables states corresponding to various insolations, while the full curves are the instable state of the runaway greenhouse. Top panel: insolation values of each stable state. The colored arrows represent the insolation for which the runaway greenhouse arises for each setup. The OLR and temperature values are averaged over two years (for both the stable states and the runaway greenhouse) and the colored fills are the 1-sigma uncertainties due to variability on the OLR calculation.
In the text
thumbnail Fig. 2

Thermal emission and absorbed flux as a function of the mean surface temperature. Blue and purple solid lines are OLR and ASR, respectively, while the blue and purple dashed lines are their clear-sky equivalent (csOLR and csASR). Black dots are the stable states and the blue solid line is the runaway greenhouse for an insolation equal to 400 W m−2. The simulation setup is a waterworld with 1 bar of nitrogen without CO2 (W1). For comparison, the red line is the OLR computed with the 1D reverse version of the Generic-PCM which uses the same radiative transfer module and assume a fully saturated atmosphere (named the 1D reverse version of LMD-Generic in Chaverot et al. 2022). The green line is the OLR computed using a 1D reverse model assuming a subsaturated atmosphere (saturation mixing ratio less than 5%) from Goldblatt et al. (2013). The flux and temperature values are averaged over a two-year period and the colored areas are the 1 sigma uncertainties due to variability on the fluxes calculation.
In the text
thumbnail Fig. 3

Map of the evolution of the mean vertical profile of the total water cloud (liquid + ice) content as a function of the global surface temperature during the evaporation phase. The simulation setup is the waterworld with 1 bar of nitrogen without CO2 (W1). The profiles are averaged over one year.
In the text
thumbnail Fig. 4

Latitude-longitude map of the clouds (liquid + ice) distribution during the evaporation phase. The maps were calculated in the heliocentric frame, that is keeping the subsolar point at 0° longitude and 0° latitude, and using an annual average. The simulation setup is the waterworld with 1 bar of nitrogen without CO2 (W1).
In the text
thumbnail Fig. 5

Evolution of the wind circulation during the runaway greenhouse. The color map represent the zonal wind (positive values represent eastward winds) and the arrows the vertical and meridional winds averaged over 1 year. The first row is the evolution of the winds for the waterworld setup (W1) and the bottom row is considering the Earth’s continents (E1), both without CO2. The first panel of each row is the warmer stable state obtained for the two considered setups at I = 375 W m2. Others show the evolution during the runaway greenhouse for different temperatures. Wind speeds corresponding to the length of the arrows vary with temperature.
In the text
thumbnail Fig. 6

Evolution of the thermal emission when water vapor condense back into the surface ocean. The bottom panel is the OLR function of the mean surface temperature. The top panel is the insolation of each simulation. The blue curve is the warming simulation presented in Fig. 1 with an insolation equal to 400 W m−2. Black dots are stable states assuming a two-year average. Other colored curves (yellow, orange, brown, and red) are the evolution of the OLR for lower insolations and assuming a initial temperature equal to 369 K (the black star). Arrows represent the tendency of the evolution of the global surface temperature (warming or cooling) for different insolations. Colored crosses represent the stable stable found for cooling simulations (ISR=300 W m−2, 325 W m−2, and 350 W m−2, respectively). The top panel highlight the hysteresis loop induced by the clouds feedback during the runaway greenhouse. However, the bottom panel shows that there is no dependence of the OLR on the insolation for a given surface temperature. The simulation setup is a waterworld with 1 bar of nitrogen without CO2 (WCO2).
In the text
thumbnail Fig. 7

Temperature profile for different surface temperature during the dry transition with 1 bar of water vapor. Arrows indicate the transition between a dry adiabat in the lower part of the atmosphere (high pressures) and a wet one in the bottom part (low pressures) for the coldest (blue) and hottest (red) cases. Dry region is below the transition while moist one is above (Abe & Matsui 1988; Kasting 1988). The simulation setup is the waterworld with 1 bar of nitrogen without CO2 (W1).
In the text
thumbnail Fig. 8

Map of the evolution of the mean vertical profile of the total water cloud (liquid + ice) content as a function of the global surface temperature during the dry transition phase. The simulation setup is the waterworld with 1 bar of nitrogen without CO2 (W1) assuming 1 bar of water vapor. The profiles are averaged over one year.
In the text
thumbnail Fig. 9

Evolution of the OLR as a function of the mean surface temperature from temperate stable states to post-runaway stable states. For blue curves, the simulation setup is the waterworld, 1 bar of nitrogen without CO2 (W1). The red curve is taken from Turbet et al. (2021) assuming 10 bar of water vapor, 1 bar of nitrogen and no CO2. The insolation of the evaporation phase and the dry transition is equal to 400 W m−2. The vertical grey lines are the temperatures for which 1 bar and 1.5 bar of water vapor are evaporated. Stable post-runaway states from Turbet et al. (2021) assume different insolations. The OLR values are averaged over 2 years which explain the gap between the evaporation and the dry transition phases. Cyan, orange and green dotted lines are OLR curves from reference works using 1D models and assuming 273 bar of water (pure water from Goldblatt et al. 2013, Earth’s case from Kopparapu et al. 2013 and with 1 bar of N2 from Chaverot et al. 2022, respectively).
In the text
thumbnail Fig. 10

Latitude-longitude map of the cloud (liquid + ice) distribution during the dry transition phase with 1 bar of vapor. The maps were calculated in the heliocentric frame that is keeping the subsolar point at 0° longitude and 0° latitude, and using an annual average. The simulation setup is the waterworld assuming 1 bar of nitrogen without CO2 (W1).
In the text
thumbnail Fig. 11

Evolution of the cloud forcing during the runaway greenhouse. Solid lines are the evaporation and dry transition phases while black dots are stable states assuming a two-year average. Negatives values indicate a cooling effect of the clouds and positive ones a warming effect. In the longwave (blue curves), clouds have a cooling effect due to their high albedo reflecting the incoming stellar flux back to space. The red curve correspond to the warming contribution due to the greenhouse effect of the clouds (i.e., thermal mission absorbed by the clouds). The green curve is the net effect of the clouds, including shortwave and longwave contributions. The vertical grey lines are the temperatures for which 1 bar and 1.5 bar of water vapor are evaporated. These are used as initial conditions for the dry transition phase (colored dotted lines). Colored triangles are the values obtained in Turbet et al. (2021) for the same insolation (400 W m−2), but including 10 bar of water vapor. The fluxes values are averaged over two years, which explain the gap between the evaporation and the dry transition phases. The simulation setup is the waterworld with 1 bar of nitrogen without CO2 (W1).
In the text
thumbnail Fig. 12

Evolution of the bond albedo as a function of the mean surface temperature from temperate stable states to post-runaway stable states. The simulation setup is the waterworld with 1 bar of nitrogen without CO2 (W1). The insolation of the evaporation phase and the dry transition is equal to 400 W m−2. The vertical grey lines are the temperatures for which 1 bar and 1.5 bar of water vapor are evaporated. The albedo values are averaged over 2 years which explain the gap between the evaporation and the dry transition phases. Pink squares and green triangles are steady states of the Earth’s case from Leconte et al. (2013) and Wolf & Toon (2015) respectively. Dotted lines are albedo curves from reference works using 1D models (1 bar of N2 from Goldblatt et al. 2013 and Earth’s case from Kopparapu et al. 2013).
In the text
thumbnail Fig. 13

Literature comparison of the OLR and the insolation as a function of the temperature. The black dots are the stables states corresponding to various insolations, while the full curves are the unstable state of the runaway greenhouse. The top panel indicates the insolation values of each stable state. The colored arrows represent the insolation for which the runaway greenhouse arises for each setup. The OLR and temperature values are averaged over two years and the colored fills are the 1 sigma uncertainties due to variability on the OLR calculation. Pink squares and green triangles are steady states of the Earth’s case from Leconte et al. (2013) and Wolf & Toon (2015), respectively. The red dashed curve is a reference result from Goldblatt et al. (2013) for a pre-industrial atmosphere and using a 1D model, assuming saturation mixing ratio below 5%.
In the text
thumbnail Fig. A.1

Evolution of the water vapor mass mixing ratio and of the mean vertical profile of the water cloud (liquid + ice) content during the evaporation phase. The simulation setup is the waterworld with 1 bar of nitrogen without CO2 (W1). The profiles are averaged over one year. The dashed black line corresponds to the temperate stable state with the highest insolation (ISR=375 W m−2).
In the text
thumbnail Fig. A.2

Evolution of the water vapor mass mixing ratio and of the mean vertical profile of the water cloud (liquid + ice) content during the dry transition phase. The simulation setup is the waterworld with 1 bar of nitrogen without CO2 (W1) assuming 1 bar of water vapor. The dashed black line corresponds to the post-runaway stable state obtained after the runway greenhouse transition.
In the text
thumbnail Fig. A.3

Evolution of the latitude-altitude map of the water cloud (liquid + ice) distribution during the evaporation phase. The simulation setup is the Earth with 1 bar of nitrogen without CO2 (E1). The maps are averaged over two years. The A, B and C panels correspond to global surface temperatures of 320 K, 340 K, and 370 K, respectively.
In the text
thumbnail Fig. B.1

Cloud (liquid + ice) profiles and temperature profiles during the runaway greenhouse when the global surface temperature is equal to 370 K, from simulations assuming different insolations and a common initial condition. The simulation setup is the waterworld with 1 bar of nitrogen without CO2 (W1). The profiles are averaged over one year.
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.