Terminator Habitability: The Case for Limited Water Availability on M-dwarf Planets

We begin with a simple comparison of the water-abundant cases (Aq34, Aq30, and Aq25), examining the maximum and minimum surface temperatures. By definition, in order for terminator habitability to occur, a planet must sustain large day–night temperature gradients. However, the simulations show that there is a significant reduction in the surface temperature range as the planet moves closer to the star (Figure 2(a)), in agreement with the findings of Yang et al. (2019b), Haqq-Misra et al. (2018), and Noda et al. (2017). Comparing Aq34 and Aq30, we note that Aq30 is on average 25 K warmer, but the day–night temperature contrast (ΔT) is only 34 K (Table 3). This is unfavorable for terminator habitability, because as the planet's mean temperature increases, the nightside warms more than the dayside, and the day–night contrast becomes small. Therefore, at the inner edge of the habitable zone, we would expect planets to either be mostly habitable (e.g., Figure 3(b)) or, if their temperatures become sufficiently high, entirely uninhabitable, due to a global runaway greenhouse effect. In this section, we explore the mechanisms that lead to small day–night temperature differences in the water-abundant planets.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Increasing the mean planetary temperature can result in many changes to the radiative budget, including competing effects. For example, increasing the water vapor contributes to cloud formation and an increase in planetary albedo, which has a strong effect on the climate of M-dwarf planets (Yang et al. 2013; Yang & Abbot 2014). As can be noted in Figure 4(c), regions with abundant clouds reflect a larger fraction of incoming SW, which tends to reduce warming, particularly in the substellar region. Meanwhile, the increased water vapor also leads to stronger LW absorption and a stronger greenhouse effect. But, individually, neither of these effects supports a reduction in the day–night temperature contrast. While the presence of clouds would reduce the dayside warming, the cloud fraction is already high in Aq34, such that there is not a significant difference between the TOA albedo in Aq34 and Aq30. Meanwhile, the increased dayside greenhouse effect on its own, in the absence of any energy and moisture transport, would actually promote increased day–night temperature contrasts. Clearly, that is not what occurs, as can be noted in the surface temperatures (Figure 2(a)) and precipitable water (Figure 2(b)). To fully understand the relationship between the dayside and nightside climates, we have to explore the relationship between the radiative budget and the atmospheric energy transport.

Zoom In Zoom Out Reset image size

As can be observed for Aq34, in Figure 5(e), (j), and (o), the water vapor (Q) maximizes near the substellar region, and has large values even in the mid-troposphere. The water vapor leads to peaks in SW absorption in the mid- and upper atmospheric layers (Figures 5(c) and (h)), which, when combined with the reflection from clouds, prevents most of the SW from reaching the surface (Figures 5(b), (g), and (i)). The atmospheric substellar absorption of M-dwarf radiation on an Earth-like planet is stronger than for F- or G-dwarf instellation (Shields et al. 2019), such that in this simulation less than 50 W m⁻² of SW reaches the surface. Even with 400 W m⁻² of LW, the substellar surface is a radiative local minimum (Figure 5(d)). Near-surface downward radiative flux maxima occur in the "mid-latitudes" (roughly 50° from the substellar point), creating regions of strong evaporation. These are also regions of substantial sensible energy and moisture flux divergence, due to the large-scale overturning circulation's near-surface winds patterns, such that the "mid-latitude" temperatures are more moderate. Meanwhile, the resulting transport of sensible energy and moisture helps sustain the substellar region's temperature maxima, despite the weak radiative values.

Zoom In Zoom Out Reset image size

With little SW reaching the surface, and most stellar radiation being absorbed in the mid- and upper atmosphere, we would expect weak atmospheric lapse rates in the aquaplanet simulations. Figure 5(a) shows the atmospheric lapse rates for Aq34, averaged within a 30° radius of the substellar point. The near-surface lapse is roughly adiabatic (8.8 K km⁻¹), and it decreases above the boundary layer in the region of strong convection, as would be expected for a region with moist convection. In the region near σ = 0.5, where we note strong SW absorption (Figure 5(c)), the lapse rate decreases further, dropping below 4 K km⁻¹ at σ = 0.4. This is significantly lower than the local saturated adiabatic lapse rate of ∼7 K km⁻¹, creating a region of convective stability that limits the vertical flow (Figure 6(b)).

Zoom In Zoom Out Reset image size

Overall, the atmospheric circulation follows a structure typical for a planet in a slowly rotating regime (e.g., Merlis & Schneider 2010; Showman et al. 2013; Haqq-Misra et al. 2018). There is strong near-surface convergence at the substellar region (Figure 4(a)), which drives an upward flow. The flow then diverges in the mid-/upper troposphere (Figure 4(b)), and sinks in the nightside of the planet. But, due to the region of convective stability, the overturning circulation structure is shallow (6(c)), and a second smaller cell forms above the region of convective stability. The divisions between the cell layers are indicated with the black markers in Figure 6(c). If we considered only the cell's appearance, and the relatively small mass transport, we might erroneously assume that atmospheric transport plays a reduced role in water-abundant simulations.

While the atmospheric circulation intensity reflects the wind speeds and mass transport, it does not necessarily provide a good indication of the net energy transport. For a planet at steady state, in the absence of ocean energy transport, any TOA imbalance must be equivalent to the vertically integrated atmospheric energy transport, such that the energy budget (e.g., Neelin 2007) for a synchronously rotating planet without a seasonal cycle can be simplified to

$$\begin{eqnarray}&&{R}_{\mathrm{toa}}-{F}_{\mathrm{sfc}}={\rm{\nabla }}\cdot \langle \overline{{\boldsymbol{v}}h}\rangle ,\end{eqnarray} \tag{ 2 }$$

where R_toa is the TOA radiative fluxes and F_sfc is the surface fluxes, which include the radiative, sensible, and latent heat flux at the surface. The moist static energy term is defined such that h = c_p T + L_v Q + gZ, comprised of dry enthalpy c_p T, latent energy L_v Q, and potential energy gZ. 〈·〉 denotes a vertical integral, and $\overline{(\,\cdot \,)}$ denotes a temporal long-term average.

As is the case with Earth's mean meridional circulation, the near-surface convergence of warm air (in this case, at the substellar point) results in the net convergence of dry enthalpy (or sensible energy) and latent energy. The higher-altitude divergent flow (Figure 4(b)) tends to transport colder and dryer air, but with larger amounts of potential energy. On Aq34, both branches transport a significant amount of sensible energy (Figure 7), such that the sensible energy transport partially cancels out when vertically integrated. The latent energy transport is comparatively small, such that the potential energy transport of the upper branch ultimately determines the sign of the vertically integrated energy transport (Figure 8(a)). The small value of the latent energy flux divergence indicates that while water clearly impacts the overall climate outcome, the latent energy transport has only a small effect on the temperature gradients. The energy flux divergence (the positive values in Figures 7 and 8) indicates net energy transport away from the substellar point.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

As we increase the stellar flux (Aq30), the increased dayside SW (ΔSW_down) is slightly reduced, due to the reflection from the TOA albedo (ΔSW_up), and balanced, partially by the increased LW (ΔLW), but mostly by the increased energy transport divergence (ΔE). This can be noted in Figure 8(b), which compares Aq30 to Aq34. The net energy transport remains similar in shape (Figures 8(a) and (c)), but the net sensible energy transport is slightly reduced, in favor of latent and potential energy transport.

Many of the trends noted for Aq30 persist for Aq25. The simulations that do not achieve steady state, as is the case for Aq25, must be interpreted with caution. Prior to reaching numerical instability, the day-to-night temperature differences become nearly negligible (Figure 2(a)). The energy budget is not properly closed, as we might expect for a simulation that is not at steady state. But we notice that the sensible energy transport becomes small, likely due to the reduced temperature gradients, and the majority of the transport is done by the potential and latent energy transport, which increase to similar magnitudes.

In summary, we can attribute the reduction of the day–night temperature differences in our water-abundant planets to an increase in the net energy transport, which leads to additional energy divergence from the dayside, and additional energy convergence to the nightside. In other words, for increased stellar radiation, the transport increases, reducing the intensity of the dayside warming and enhancing the nightside warming. With higher surface and atmospheric temperatures, as well as increased moisture flux convergence, the nightside atmospheric water vapor increases (Figure 2(b)), allowing for additional nightside warming, due to an increase in the local greenhouse effect.

As previously mentioned, the atmospheric energy transport is responsible for resolving radiative imbalances (Equation (2)), such that these changes in the atmospheric transport can be more simply interpreted in the context of the TOA radiative budget. As described in Koll & Abbot (2016), for planets in a weak–temperature gradient regime, the circulation can be thought of as a heat engine driven by dayside heating, with cooling due to global thermal emissions to space. The aquaplanet simulations have high atmospheric water vapor content, resulting in optically thick dayside atmospheres, such that their dayside thermal emissions are determined by the upper-tropospheric temperature. In such cases, increasing the stellar flux has a small impact on the outgoing LW radiation (OLR; Yang & Abbot 2014), as can be noted in Figures 2(c) and 8(b). Meanwhile, the colder nightside has low humidity and a relatively transparent atmosphere (except for Aq25), such that warming is closely tied to increased OLR. Therefore, in a moist and opaque atmosphere, increasing the stellar radiation as we go from Aq34 to Aq30 results in a small dayside OLR increase and a larger nightside OLR increase, requiring a corresponding increase in the day-to-night energy transport. The extra SW absorbed in Aq30 is primarily balanced by a larger nightside OLR, which is only achievable through a strengthening of the atmospheric energy transport and an overall reduction in the day-to-night temperature gradients.

Given that atmospheric energy transport is essential for reducing the day-to-night temperature contrast, we might expect to see larger gradients on planets with thinner atmospheres. However, this is not observed in our aquaplanet simulations where we halve the background N₂ values. As can be noted in Figures 2(a) and 3, the temperature range remains almost unchanged when the surface atmospheric pressure is halved. Instead, we note overall increased wind speeds, such that the energy transport terms remain roughly similar. These results are in general agreement with the findings from Zhang & Yang (2020), which show that the background gas pressure has a relatively small impact on the inner edge of the habitable zone for synchronously rotating Earth-like planets. They show that the effects are nonmonotonic, and significantly less important than changes in the cloud scheme (Bin et al. 2018) and rotation rate (Kopparapu et al. 2017). Therefore, we would not expect the pressure alone to determine the viability of terminator habitability for the Earth-like planets that we are considering.

Here, we switch our focus to water-limited land planets, using ExoCAM with a land-planet setup, initialized with varying amounts of atmospheric water vapor (L34Qe3 and L34Qe4). Once the temperatures equilibrate, these simulations have 20% and 1% of Earth's atmospheric precipitable water, respectively, also with small amounts of water in the form of surface ice and soil water, concentrated on the nightside. We will begin by considering the differences between Aq34 and L34Qe4.

With reduced atmospheric water vapor, we expect a reduction in cloud coverage, especially for low-level clouds. This expectation is confirmed, resulting in a reduction of the dayside albedo, which becomes largely homogeneous with values between 0.2 and 0.26 in the water-limited simulations. There is still some condensation, precipitation, and high-altitude cloud formation, occurring near the substellar point. But with re-evaporation in the atmosphere, the surface precipitation becomes negligible. These precipitation patterns, noted in all our land-planet simulations, appear to follow the "transition regime" described in Ding & Wordsworth (2021). Despite having the same stellar flux as Aq34, L34Qe4 has a larger net SW, due to the reduced cloud coverage and lower albedo (negative ΔSW_up; Figure 8(d)). The resulting increase in the substellar net SW must result in one of two outcomes: either increased TOA LW emissions or increased day-to-night energy transport.

Near the substellar point, a much larger fraction of SW reaches the surface (Figure 9), with the dayside surface temperatures reaching maxima of 355K (L34Qe4). Though there is still water vapor SW absorption, the atmosphere's vertical temperature structure more closely resembles a dry region on Earth, with the dayside lapse rates remaining steeper than 6K km⁻¹ until σ = 0.25 (Figure 6). Without the low region of convective stability noted in Aq34, the water-limited simulations have deeper overturning cells, ascending at the substellar point (Figures 6(f) and (i)).

Zoom In Zoom Out Reset image size

Examining the energy flux divergence (Figure 8(e)), we note a strengthening of the individual energy transport terms, but a weakening of the net transport. The deepening of the atmospheric circulation, combined with strong high-altitude winds, leads to significantly larger values of potential energy transport away from the "eye," as can be noted in Figure 7, which shows the vertical profile of the energy flux divergence averaged across an area within 30° of the substellar point. Sensible energy fluxes are intensified in both the upper and lower branches of the circulation, but the lower branch increases significantly, in both intensity and depth. The boundary layer in the water-limited simulation is significantly deeper (Figure 7), and the lower branch of the circulation occupies its full depth, reaching up to σ = 0.4. This effect can also be noted in the upward shift of the cell's core (Figure 6). This "bottom-heavy" cell structure achieves higher values of sensible heat conversion at the substellar point, which largely cancels out the increased potential energy divergence, such that the net day-to-night energy transport is reduced relative to Aq34, despite the increased net SW. Comparing ΔSW_up to the change in net energy flux divergence (ΔE), we note that the effects of the reduced atmospheric energy transport are of comparable intensity to the effects of the reduced albedo.

With reduced albedo and reduced energy transport, the large TOA imbalance is resolved by an increase in the dayside TOA LW emissions (positive ΔLW; Figure 8(d)). It is interesting to note that in the water-limited cases, the substellar region becomes a region of LW maxima, rather than minima, as was the case for the aquaplanet simulations (Aq34 and Aq30). Meanwhile, the reduced nightside energy convergence results in a reduction of nightside temperatures. Together, these effects allow the water-limited case to achieve equilibrium with larger day–night temperature gradients, spanning 139 K (L34Qe4) and 122 K (L34Qe3). If we assume that surface temperatures surpassing 50°C are hostile to life, and temperatures nearing 100°C would be uninhabitable, both the L34Qe3 and L34Qe4 water-limited scenarios can be considered as cases of terminator habitability.

As with the aquaplanets, we can reframe these results in terms of the atmospheric radiative imbalances, to obtain a simpler intuition. On these land planets, the atmosphere is dryer, and therefore less cloudy and opaque. This allows the dayside to be closer to radiative equilibrium (Figure 8(d)), which therefore implies reduced energy transport to the nightside. The water-limited L34Qe4 absorbs a greater fraction of the incoming stellar radiation at the surface than Aq34, achieving higher surface temperatures, and higher OLR, which thus requires weaker energy transport (Figure 8(d)). In turn, the nightside receives less energy, and is colder than the aquaplanet counterpart. This results in significantly larger day–night temperature contrasts, such that a temperate terminator becomes not only easily achievable, but also a likely scenario for land planets in the habitable zone.

The simulations discussed so far illustrate the possibility of terminator habitability as an alternative climate configuration within the habitable zone (at an orbital period of 34 days). However, it is also worth emphasizing that water-limited planets can endure higher stellar fluxes, without entering a runaway greenhouse state. For example, while Aq25 and Aq25h become unstable due to runaway warming, an equivalent water-limited simulation (L25Qe4) achieves steady state and sustains a temperate terminator climate (Figure 3(f)). The increase in the substellar SW relative to L34Qe4 is balanced in roughly equal parts by an increase in the LW and an increase in the local energy flux divergence. The resulting climate is on average warmer than L34Qe4, and the day-to-night temperature range increases to 154 K (Figure 2). The L25Qe4 nightside temperatures remain just below freezing, the dayside temperatures reach scorching highs over 400 K, and the terminator remains within the 270–320 K range. While the fractional habitability (based on surface temperature) of L25Qe4 may be smaller than those of Aq34 and Aq30 (Table 3), the contrasting results of L25Qe4 and Aq25 illustrate how we might expect to observe planets with diverse surface climates and regions of temperate surface temperatures, even at the inner edge of the habitable zone.

The water-limited simulations show that terminator habitability is possible. However, based on the results discussed so far, we have not yet established whether they are ideal observational targets in the search for life beyond Earth. After all, these simulations depict planets with harsh climates, limited regions of habitability, and limited water availability, which may seem unappealing when compared to the climates of ocean worlds obtained with aquaplanet simulations. However, arid exoplanets offer observational advantages, because their reduced cloud coverage could facilitate the detection of water (Ding & Wordsworth 2022) by JWST, potentially making them more practical near-future targets. Also, a closer look at the aquaplanet moisture budget shows that we must be cautious of assuming long-term climate stability, which is crucial for increasing the odds of developing life.

Aquaplanet simulations provide the atmosphere with an unlimited water source. On the planet's dayside, not only is there water available everywhere, but there is no depletion in regions of negative net precipitation, nor resistance to evaporation, as would occur from drying soil (e.g., van de Griend & Owe 1994). For Aq34, which has dayside temperatures comparable to Earth's, this results in an Earth-like amount of atmospheric precipitable water (∼70% of Earth's atmospheric water). The "eye" is a region of negative net precipitation, and there is continuous transport of water vapor to the nightside, resulting in a total of 1.5 ×10¹⁶ kg s⁻¹ of snowfall. For scale, this rate of snowfall would cold trap a volume of water equivalent to Earth's oceans in 90,000 years. The presence of a deep global ocean circulation could, in some cases, help deglaciate part or all of the nightside ocean (Hu & Yang 2014). But in the absence of a global ocean with favorable properties, such as large depth (Hu & Yang 2014) and high salinity (Olson et al. 2020), we can assume a limited return flow for water deposited on the nightside. Even accounting for ice sheet dynamics, which would imply a down-gradient flow of ice toward the dayside, it is possible that over time a majority of the water from an ocean-covered Earth-like planet could become cold trapped (Leconte et al. 2013; Menou 2013).

While Aq34 could be at risk for cold trapping, the risk decreases significantly for planets closer to the star. The Aq30 nightside temperatures are higher, such that the equatorial region of the nightside remains above freezing. The snowfall is confined to the higher latitudes, and the global snowfall rate decreases by a factor of 100 relative to Aq34. However, the atmospheric precipitable water increases to approximately 3.5 times the Earth's atmosphere. In particular, we note an increase in water vapor content above the atmospheric cold trap, reaching values of 0.003 kg kg⁻¹ (∼0.005 mol mol⁻¹), which is 2 orders of magnitude larger than the Aq34 values, reaching the limit proposed by Kasting et al. (1984) and confirmed in a series of planetary studies (e.g., Kasting et al. 2015; Wolf & Toon 2015). This raises the possibility that such a planet would be vulnerable to continuous water vapor loss, especially during periods of strong flare activity. These losses are in addition to the water losses resulting from enhanced heating during the M dwarf's protracted pre–main sequence phase, which could impact all of these scenarios and cause the loss of large water reservoirs (e.g., Luger & Barnes 2015; Bolmont et al. 2017). Therefore, we have abundant reason to question water-rich M-dwarf scenarios.

Meanwhile, for the land planets L34Qe3, L34Qe4, and even L25Qe4, while there is still risk of water loss in their early history, their high-altitude moisture values are comparable to those of Aq34 and well below the Kasting et al. (1984) limit. Similarly, despite their extremely low nightside temperatures, at equilibrium there is no snow being deposited on the nightside. There is some snow and ice accumulation in L34Qe3 while the model spins up, which remains trapped. But the remaining moisture is subsequently retained in the atmosphere above the planetary boundary layer, primarily on the dayside. We have confirmed that simulations initialized with more water (e.g., Q = 0.01) have similar nightside water trapping, and slowly tend toward the climate in L34Qe3, but they would require hundreds more simulated years to fully reach equilibrium.

Based on our land-planet simulations, water-limited atmospheres appear to reach a more stable configuration than their aquaplanet counterparts, being less vulnerable to additional water vapor loss or nightside cold trapping. This also suggests that ocean-world configurations that are subject to significant water loss, without entering a runaway greenhouse state, could eventually reach a stable configuration with terminator habitability. These considerations, combined with the observational advantages of water-limited planets (Ding & Wordsworth 2022), suggest that despite their potentially limited or localized fractional habitability, land planets will be important observational targets in the coming years, and will play a prominent role in the early stages of exoplanet climate characterization.

Thus far, we have defined terminator habitability in terms of surface temperature. However, water is also a crucial ingredient for life as we know it, and could pose a problem for water-limited scenarios. In our simulations, rain is strongly concentrated in the substellar region (for both aquaplanet and land-planet simulations), and snow occurs across the nightside in some aquaplanet simulations. For the rotational configurations considered here, the terminator never receives significant amounts of precipitation. Therefore, even though terminator evaporation rates tend to be low, it is a region of weak negative net precipitation in our simulations. In other words, based on the atmospheric moisture budget, the terminator is a region that would tend to have an arid climate. However, water availability on land planets is not exclusively determined by the atmospheric moisture budget, but also depends on factors such as groundwater transport and glacier behavior. These effects are in turn dependent on many poorly constrained planetary properties, such as orography, soil structure, and geothermal heat flux. While we cannot quantify these effects with our climate model, we can discuss them qualitatively, to explore how they would impact the terminator water availability.

The role of glacier flow has been widely discussed, both in the context of a snowball Earth (e.g., Goodman & Pierrehumbert 2003; Pollard & Kasting 2005) and as a potential mechanism for releasing cold-trapped water on synchronously rotating planets (e.g., Leconte et al. 2013). Ice sheets can deform under their own weight, resulting in an ice flow down the ice thickness gradient. On a synchronously rotating planet, this flow could push ice toward the terminator, where the temperate climate would result in melting. The flow velocity depends on, among other things, the ice geometry, planetary gravity, and geothermal heat flux. For super-Earths, Yang et al. (2014b) argue that it could largely prevent nightside cold trapping on water-abundant planets.

Generally, the ice flow would be slower on Earth-sized planets, especially those with more moderate geothermal heat fluxes, but it could still provide a large enough source of meltwater to create a moist climate in the terminator region. For example, on Greenland and Antarctica, flow speeds and discharge into the ocean can reach rates over 1 km yr⁻¹ (Rignot et al. 2011; Rignot & Mouginot 2012). Currently, model representations of glacier dynamics are limited, and not fully suitable for planetary purposes. For CESM, even a configuration with an evolving ice sheet, where the glacier area and elevation are adjusted so as to conserve mass and energy, would not capture the dynamics relevant for the ice flow that we consider here. But, if we make a conservative estimate, considering only the slower flow rates observed at the Antarctic glacier edges (∼100 m yr⁻¹), and assuming that only a shallow ice layer reaches the terminator (≤10 m), this would imply water transport rates on the order of 10⁻² kg s⁻¹. This exceeds the terminator evaporation rates in the Aq30 and Aq34 simulations (see Figure 4(d)) by 2 orders of magnitude. Given that aquaplanet simulations have infinite water availability, we can cautiously treat their evaporation rates as an upper limit, such that the comparison of glacier flow to aquaplanet terminator evaporation rates indicates that glacier flow could sustain surface water near the terminator. Depending on the exact rates, this could result in swamp-like regions, oceans, or smaller lakes and rivers in regions of topographic lows. Even if the flow rates were significantly lower than expected, and the surface runoff resulted in a large evaporating area, soil evaporative resistance would tend to retain some local moisture, which could be beneficial for the development of life in the terminator region.

Of course, not all planets would have nightside glaciers. The presence of large ice deposits presumes a planet with abundant initial water, which was either retained throughout the early period of intense stellar emissions or delivered later on. It also presumes that ice would be stable on the nightside. For planets receiving sufficiently high instellation, the nightside ice can sublimate (Ding & Wordsworth 2020), such that water would likely only be present in vapor form. Our arid land-planet simulations are not ideal for exploring this limit, but based on the values from Ding & Wordsworth (2020), we would expect nightside sublimation to be important for planets in L25Qe4's orbit. Therefore, L25 planets would be more likely to have a temperate but dry terminator than their L30 or L34 counterparts.

A less recognized—but also potentially important—mechanism to consider for planets with stable surface water is groundwater transport. Faulk et al. (2020) have shown that surface and subsurface methane flows on Titan could help to shape the global hydrological cycle, transporting liquid from the low-latitude highlands to the lower polar regions, resulting in a moist polar climate. For synchronously rotating planets, Turbet et al. (2016) have argued that the alignment of large-scale gravitational anomalies and the star–planet axis is favored (Wieczorek 2007), such that the largest topographic basins would tend to be near the substellar point or antistellar point, which has served as motivation for studies of substellar continents on ocean worlds (Lewis et al. 2018). For a land planet, where precipitation is strongly concentrated at the substellar region, a substellar topographic high could result in a hydrological cycle that continuously transports water across a larger portion of the dayside, whereas a topographic low could further concentrate water near the substellar point. Therefore, additional explorations of topography, runoff, and subsurface transport are necessary to constrain the surface water availability for a land planet, especially those with significant large-scale topography.

Based on the above discussion of water availability, we might expect that some, but not all, planets with temperate terminators would also have moist terminator climates. We expect that future work, exploring a wider range of surface configurations, will help to better constrain the water availability for land planets. Model advancements, including glacier dynamics within the water budget, would further improve our understanding of these water worlds. Finally, an important caveat to this work is the fact that we have only considered planets with an Earth-like atmospheric composition. Varying atmospheric mass and composition would certainly impact these results. For example, Ding & Wordsworth (2020) have shown that nightside cold trapping might be avoided on planets with high CO₂ values, due to increased nightside temperatures. Constraining the exoplanet CO₂ levels remains a challenge, due to their dependence on outgassing and weathering rates (Walker et al. 1981). Therefore, future work may wish to consider both terminator habitability and high-CO₂ "eye" habitability scenarios, while exploring additional factors, such as the impact of groundwater flow and long-term climate sensitivity in the presence of M-dwarf flare activity, to quantify prospects for sustained habitability.