FAQ: Possible climate transitions from breakup of stratocumulus decks under greenhouse warming

There are several questions that have been asked frequently about our recent paper in Nature Geoscience. Here are answers to some of them.

  1. What are stratocumulus clouds?
    are low-lying clouds with tops typically below 2 km (7000 ft) altitude. They are the most frequent cloud type on Earth and, over oceans, often form nearly unbroken decks covering thousands of square kilometers. In the subtropics, they are especially prevalent over eastern ocean basins, for example, off the coasts of California, Peru/Chile, and Namibia/Angola.
  2. Why are they important for Earth’s climate?
    Stratocumulus clouds cover about 20% of the tropical ocean area (between 30°S and 30°N). Because they reflect much of the sunlight incident on them back to space, they shade and cool Earth’s surface. This makes stratocumulus clouds important for regulating Earth’s surface temperature globally. It has long been recognized that increasing the fraction of the globe they cover by a few percent can lead to substantial global cooling; conversely, decreasing the area they cover can lead to substantial global warming.
  3. How do stratocumulus clouds arise?
    While many other clouds, such as tropical cumulus clouds, are produced by sunlight absorbed at the surface driving air upwards, the turbulent overturning circulations that sustain stratocumulus clouds are driven by radiative cooling at the cloud tops. The condensed water in the clouds is an excellent absorber and emitter of thermal (infrared) radiation. The cloud tops emit thermal radiation upward, and they absorb thermal radiation downwelling from above. Because the downwelling radiation originates in higher layers of the atmosphere that are cooler than the cloud tops, the cloud tops cool radiatively. The cooled air masses sink, pick up moisture from the ocean surface, and then bring that moisture back up on their return trip to nourish and sustain the clouds. Stratocumulus decks break up when the radiative cooling of the cloud tops becomes too weak to propel air parcels to the surface, or when turbulence that entrains dry and warm air from above the clouds into the cloud layer becomes too strong.
  4. Why did you want to study stratocumulus clouds?
    Although stratocumulus clouds are known to be important for Earth’s climate, how they respond to climate change has remained unclear. Because they are sustained by a complicated interaction between radiation and turbulent air motions, there are good reasons to expect that they are impacted by increasing greenhouse gas concentrations. For example, if the greenhouse gas concentration in the atmosphere above the clouds increases, the cloud tops cannot cool as efficiently. We wanted to know more precisely what happens when greenhouse gas concentrations increase.
  5. What was your hypothesis about how stratocumulus clouds would respond to global warming?
    The physical mechanisms controlling stratocumulus clouds suggest that at high enough greenhouse gas concentrations, the cloud-top cooling may become so weak that the clouds cannot sustain themselves. Counterintuitively, the increase in evaporation from the surface in a warmer climate can also work against the existence of stratocumulus clouds by increasing the rate of turbulent mixing of dry air from above the clouds into the cloud layer, causing the clouds to dissipate (e.g., Bretherton and Wyant 1997). A previous study led by one of our graduate students, Zhihong Tan, had shown that stratocumulus cloud dissipation can occur at high greenhouse gas concentrations. As that happened, the surface warmed strongly and, crucially, this warming further amplified the mechanisms leading to breakup (reduced radiative cooling and enhanced evaporation). So our expectation was that stratocumulus decks would break up at high enough greenhouse gas concentrations. What we did not know was at which greenhouse gas concentrations this breakup occurs when the subtropical stratocumulus patches, which we can explicitly model, are coupled to the larger climate system. We also did not know what would happen when greenhouse gas concentrations are lowered again after stratocumulus breakup.
  6. What did you find?
    We found that stratocumulus decks over subtropical oceans become unstable and break up into scattered cumulus clouds at CO2 levels at least three times higher than those today (about 1,200 ppm or more of CO2 equivalent). When that happens, the ocean surface below the clouds warms because the cloud shading is diminished. The extra solar energy absorbed as stratocumulus decks break up (in an area estimated to cover about 6.5% of the globe) is assumed to spread uniformly across the globe, leading to around 8℃ global warming. After the stratocumulus decks have broken up, they only re-form once CO2 concentrations drop substantially below the level at which the instability first occurred. It is difficult to pinpoint precisely at which CO2 levels the clouds break up or re-form, because this depends on details of how we represent the global atmospheric circulation in the model. We found instability thresholds between 1,200 and 2,200 ppm in different simulations. But the mechanisms underlying the instability are well established and robust. Our study clearly points to, as the title suggests, the possibility of a previously undiscovered and potentially strong nonlinear feedback in the climate system.
  7. What are the implications?
    • Climate transitions that arise from the stratocumulus instability may have contributed importantly to abrupt climate changes and very warm climates in Earth’s geological past. For example, geological records indicate that around 50 million years ago (during the Eocene) Earth was very warm. It was unclear, however, how such a warm climate could arise given estimates that CO2 was less than 2,000 ppm. Current climate models do not reproduce warm enough climates with CO2 levels below 2,000 ppm. Our study suggests a possible mechanism of how such warm climates, and transitions between different climate states, may have arisen.
    • The study points to a blind spot in climate modeling: climate models do not simulate stratocumulus clouds and their climate change response well, and we urgently need to improve this.
    • It cannot be excluded that stratocumulus decks break up if CO2 levels continue to rise rapidly for over a century. However, it is unlikely that we will reach such high CO2 levels, and climate change would have severe and dangerous global impacts well before the loss of stratocumulus decks.
  8. What is new about your approach?
    Climate models simulate the large-scale dynamics of the atmosphere explicitly and represent the important smaller-scale dynamics of clouds semi-empirically. We inverted this approach: We simulated the dynamics of clouds explicitly and represented large-scale dynamics semi-empirically. (Unfortunately, representing all relevant scales simultaneously on the globe requires computers orders of magnitude faster than the fastest supercomputers we have. This is not an option for decades to come.)  As several other studies before ours did, we simulated the dynamics of the atmosphere over a small patch of a subtropical ocean, representative of an area off the coast of California, Peru/Chile, or Namibia in summer. A crucial new ingredient in our study was that we coupled the subtropical atmospheric domain (a) to the underlying ocean surface temperature through radiative energy fluxes (as in the previous study led by Zhihong Tan), and (b) to simple but physically plausible representations of large-scale atmosphere motions. This form of coupling  has a long history in the atmospheric sciences, for example, in models of tropical climate (e.g., Pierrehumbert 1995). But it is new in the context of explicit simulation studies of subtropical clouds. Because previous studies of subtropical clouds with high-resolution models, including “ultraparameterized” global models with embedded higher-resolution models, had prescribed surface temperatures and thus artificially suppressed the feedbacks between clouds and the surface, the stratocumulus instability had remained undiscovered, although the governing cloud-layer mechanisms were known.
  9. Why didn’t you just use a standard climate model?
    Climate models are notoriously poor at simulating stratocumulus clouds, because it is difficult to adequately capture the coupling of radiation with the meter-scale turbulent dynamics of clouds in models whose computational mesh is tens to hundreds of kilometers wide. For example, the upper panel of the figure below shows the observed cloud cover (black line) in a subtropical stratocumulus region and the cloud cover simulated with current climate models (colored lines).

    Annual cycle of (a) cloud cover and (b) sea surface temperature off the coast of subtropical South America from observations (black) and in climate models (colors). Data from Lin et al. (2014).

    Clearly, all models severely underestimate the prevalence of stratocumulus clouds, leading to large warm biases in their surface temperature (lower figure panel). This makes it challenging to use current climate models to study the stratocumulus response to warming: if you start with far too few clouds, it implies (a) that stratocumulus dynamics are not adequately captured in the models and (b) any amplifying feedback when their cover thins under warming will be muted. The same problems arise with recent “ultraparameterized” models that have two-dimensional, higher-resolution models embedded in a global model: the higher-resolution models are still too coarse and simulate far too few stratocumulus clouds. So studying the climate response of stratocumulus clouds with climate models is fraught with uncertainty. This is why the IPCC assesses the confidence in simulations of their climate change response to be low.

  10. What are the limitations of your study?
    Climate models do a good job of simulating large-scale dynamics but are inaccurate and uncertain in simulations of the smaller-scale cloud dynamics that, importantly, feed back onto the large scales. Conversely, our modeling setup does a good job of simulating small-scale cloud dynamics in a subtropical patch but is inaccurate and uncertain in its representation of larger scales. We have no explicit representation of spatial heterogeneity, seasonal variations, or large-scale weather. This makes it difficult to make precise quantitative statements about the CO2 level at which the instability occurs, or when stratocumulus clouds re-form after CO2 is lowered, or how abrupt the instability is (as a function of CO2 or radiative forcing). As we stated in the paper:

    The instability will probably occur first in regions and seasons in which the stratocumulus decks are close to the stability threshold, for example, at the margins of current stratocumulus regions. One may expect large transient fluctuations in cloud cover between the states with and without stratocumulus decks near the stability threshold—the flickering phenomenon common near critical transitions in complex dynamical systems. If the stratocumulus decks in different subtropical regions differ in their proximity to the stability threshold, the global effects of the instability as a function of CO2 levels may also be smoothed out.

    How much smoothing one gets from spatial heterogeneity is unclear, as there is also lateral coupling: stratocumulus loss in one part of the subtropics will affect the temperature and energy balance in other nearby parts, thereby affecting stratocumulus cover there. We also stated that “we expect the width of the hysteresis loop in nature to be reduced … by sources of noise neglected in our simulations, such as seasonal or synoptic [weather] variations.”

    Additionally, there is some uncertainty about the global warming triggered by the stratocumulus breakup, which depends on the global area fraction over which stratocumulus break up (we estimate this to be 6.5%); this uncertainty likely is small relative to the uncertainties about the CO2 thresholds because the global-mean warming is relatively insensitive to how the extra solar energy absorbed in the stratocumulus region is re-distributed across the globe.

    All of these aspects will have to await further study. However, although quantitative aspects of the stratocumulus instability remain to be investigated, the instability itself appears to be robust for the physical reasons we described and as seen in the simulations we presented.

  11. Given the limitations, how should I interpret your results?
    An analogy may help here. In the late 1960s, seminal studies pointed to the possibility of a runaway greenhouse effect, which is now widely thought to have occurred on Venus. The calculations then, and many since then, were based on relatively simple climate models, often one-dimensional models of an atmospheric column (e.g., Kasting et al. 1993). The simple models exposed a key nonlinearity that had not previously been understood: Greenhouse warming generally leads to increased water vapor concentrations in the atmosphere, which is itself a greenhouse gas and amplifies the warming. At a high enough total solar irradiance, this feedback can run away, leading to dramatic warming and the boiling away of all surface water. Because these models do not capture heterogeneity in the climate system and the runaway greenhouse warming is so extreme, it is difficult to be quantitatively precise about the conditions for a runaway greenhouse and the sharpness of the transition. Still, the physics of the nonlinear feedback processes is clear and widely accepted. It is no question that a runaway greenhouse can occur, and it very likely did occur on Venus. It would be mistaken to dismiss the runaway greenhouse models as “too simple”, for example, because they capture only one atmospheric column—such off-the-cuff reasoning does not give Venus its water back.

    Similarly, our work points to a nonlinear feedback mechanism that had not been discussed previously: Stratocumulus clouds can thin under greenhouse warming, leading to surface warming, which in turn fosters more stratocumulus thinning. At high enough greenhouse gas concentrations, this feedback can run away, leading to loss of the stratocumulus clouds. We cannot quantitatively pinpoint the critical CO2 threshold or how quickly this transition occurs. But the physical mechanisms underlying the essential nonlinearity are well established and appear robust. Our study shows how they can lead to a nonlinear response of the climate system to increasing CO2 concentrations—that is, a high climate sensitivity in hot climates, as Pierrehumbert suggested in 2013 to account for past hothouse climates. Our model exposes the underlying mechanisms and demonstrates that the instability is possible, although the model is not quantitatively precise.

  12. If your model is simple, why did it take 2 million core hours of computation to get the results?
    The model is conceptually relatively simple as far as the large-scale effects are concerned. But it is complex and state-of-the-art as far as the cloud dynamics are concerned. In fact, the model has over 10 million variables (e.g., temperature, humidity, and velocities on a computational grid with 2 million points). So if one takes the number of variables as the complexity metric, the complexity of our model is comparable with that of the most advanced of today’s atmospheric general circulation models. We chose to invest our computational budget into detailed cloud simulations, at the expense of a detailed representation of larger scales. Climate models do the opposite. Both have value, with complementary uncertainties.
  13. How can you be quantitatively more precise?
    One option is to have thousands of high-resolution simulations of stratocumulus clouds embedded in a global model, to capture interactions with large-scale dynamics and temporal and spatial heterogeneity. At the Climate Modeling Alliance, we are working on it. Stay tuned.