Improving climate projections by understanding how cloud phase affects radiation
Abstract
Whether a cloud is predominantly water or ice strongly influences interactions between clouds and radiation coming down from the Sun or up from the Earth. Being able to simulate cloud phase transitions accurately in climate models based on observational data sets is critical in order to improve confidence in climate projections, because this uncertainty contributes greatly to the overall uncertainty associated with cloud-climate feedbacks. Ultimately, it translates into uncertainties in Earth's sensitivity to higher CO2 levels. While a lot of effort has recently been made toward constraining cloud phase in climate models, more remains to be done to document the radiative properties of clouds according to their phase. Here we discuss the added value of a new satellite data set that advances the field by providing estimates of the cloud radiative effect as a function of cloud phase and the implications for climate projections.
Key Points
- The phase of a cloud (how much water and ice it contains) contributes to the cloud-climate feedbacks by modulating its radiation
- A recent observational data set better constrains the cloud phase-radiation relation
- This improved relation can help improve climate projections in models
1 Introduction
In Earth's atmosphere, water exists in three different thermodynamic states: gas, liquid, and solid. Most of the water is located in the lowest layer—the tropopause—in the form of water vapor. Under some circumstances, the air cools down to the point that water vapor condenses/deposits usually onto an aerosol particle, although sometimes onto a small nucleus of ice or water. Depending on the thermodynamic conditions (temperature, pressure) and the type of aerosols present in the atmosphere, the cloud particles forming may be liquid droplets, ice crystals, or a mixture of both. Whenever ice crystals and liquid droplets coexist, the ice crystals will grow at the expense of the surrounding droplets because of the higher saturation vapor pressures of water than ice (referred to as the Wegener-Bergeron-Findeisen (WBF) process [Bergeron, 1935; Findeisen, 1938]). Droplets may also collide with ice crystals and produce rimed particles like graupel or hail. All of the above processes together determine the thermodynamic phase of clouds or more simply cloud phase. While the cloud phase is easy to determine at temperatures above 0°C, because it will all be liquid water, and below −40°C, the approximate temperature at which all liquid droplets freeze spontaneously [Pruppacher and Klett, 1997], supercooled liquid droplets and ice crystals may coexist between these bounds (−40°C–0°C).
2 Why Should We Care About Cloud Phase?
Knowing precisely the phase of clouds is of particular interest as it gives information about how clouds interact with radiation flowing to and from Earth. For a given water content, a cloud made of liquid water contains more particles that are smaller in size than its frozen counterpart. As a consequence, the optical depth of the liquid cloud is larger than that of the ice cloud [Rogers and Yau, 1989], which, in turn, makes it reflect more solar radiation back to space (yellow arrows in Figure 1a). In addition, liquid clouds may also be more opaque to infrared radiation than ice clouds at the same temperature (red arrows in Figure 1a). Finally, the thermodynamic phase also influences the lifetime of clouds. Liquid clouds generally remain suspended longer than ice clouds, which precipitate more efficiently because of their larger size [e.g., Tao et al., 2014].
3 How Does Cloud Phase Relate to Climate Change?
For all of the above reasons, it is likely that cloud phase strongly influences the climate. In addition, because the overall phase and distribution of clouds may be different in a warmer climate, these processes need to be accurately incorporated in global scale climate models, also called general circulation models (GCMs). For example, Lohmann [2002] showed that replacing all mixed-phase clouds by supercooled liquid clouds in a GCM would lead to a difference as large as −13 W m−2 in the net cloud radiative effect at the top of the atmosphere (TOA). Similarly, a warmer atmosphere produced by increased greenhouse gas concentrations would contain more liquid and less ice, and therefore induce changes in radiation fluxes that would provide a negative feedback to warming [Li and Le Treut, 1992, Senior and Mitchell, 1993]. The strength of this feedback, however, depends on the distribution of cloud phase in the present atmosphere, which was, until recently, poorly constrained. This made quantification of climate feedbacks associated with phase changes in clouds challenging, which in turn contributed to the uncertainty in equilibrium climate sensitivity (ECS) predicted by GCMs [e.g., Knutti and Hegerl, 2008; Andrews et al., 2012; Flato et al., 2013]. The ECS is defined as the change in the near-surface global temperature per doubling of atmospheric CO2 concentration relative to pre-industrial levels, once a new equilibrium climate state is reached [Webb et al., 2006]. In other terms, the ECS characterizes the Earth's sensitivity to higher CO2 levels. It represents a useful metric for comparing the sensitivity to changes in CO2 concentrations across models and is currently estimated to lie in the range 2.1–4.7°C [Flato et al., 2013].
4 How Well Do Climate Models Simulate Cloud Phase?
Climate models constitute useful tools to study climate change and have also contributed to the understanding that cloud phase significantly influences radiation and therefore the climate system. However, the models do not perfectly replicate the reality, and conclusions made based on GCM simulations are therefore not automatically applicable to the real climate system. The models cannot resolve all the relevant cloud processes, especially those at small scale (i.e., <1 km). To overcome this issue, models use parameterizations that express averages of these subgrid processes at the typical GCM gridbox scale (~100 km). The microphysical mechanisms governing cloud phase transitions (i.e., the Wegener-Bergeron-Findeisen process, nucleation, precipitation formation) are particularly complex and difficult to implement in GCMs. As a result, climate models too often use oversimplified parameterizations relying mainly on temperature [e.g., Cesana et al., 2015; McCoy et al., 2014] and often underestimate the supercooled liquid clouds [e.g., Cesana et al., 2015; Komurcu et al., 2014; Tsushima et al., 2006]. However, better cloud phase schemes have been developed recently that account for more complex microphysics processes in GCMs [e.g., Forbes and Ahlgrimm, 2014]. Comparisons of simulations and observations, as well as the evaluation of these parameterizations, is essential for our confidence in GCM-simulated cloud feedbacks related to cloud phase. For that purpose, satellite observations are valuable [e.g., Cesana et al., 2015; Klein et al., 2013], given their broad geographic coverage and increased ability to probe clouds and sample radiation at the same time. Different types of cloud phase observations are available to the scientific community based on both passive and active sensor satellites (a passive sensor detects radiation; an active one sends out a signal and assesses its reflection [e.g., Rossow and Schiffer, 1999; Stephens and Kummerow, 2007; Cesana et al., 2016]). For GCM evaluation of cloud phase, active sensor satellites, such as Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) [Winker et al., 2010], appear to provide the best data. Compared to passive satellites, CALIPSO provides vertical measurements of the atmosphere and its lidar instrument is sensitive to supercooled liquid (thus mixed-phase clouds) as well as thin cirrus over both land and oceans.
5 Can We Improve Our General Understanding of Cloud Phase Processes?
In a recent review paper, Gettelman and Sherwood [2016] concluded that significant uncertainties remained in microphysical mechanisms (i.e., cloud phase changes, precipitation, and aerosols) and these contributed significantly to the spread in cloud feedbacks across models. They advocated that future studies should focus on better constraining these processes using observational data sets. A lot of effort has recently been made to improve cloud phase observations and modeling, some of which has been mentioned above. However, little has been done to document the radiative properties of clouds according to their phase. While previous studies have focused on specific regions such as the Arctic [i.e., Cesana et al., 2012; Morrison et al., 2011] or the Southern Ocean [Bodas-Salcedo et al., 2014; Forbes and Ahlgrimm, 2014; Kay et al., 2016; McCoy et al., 2016], only one study was for the globe as a whole [Matus and L'Ecuyer, 2017].
6 Observing the Link Between Cloud Phase and Radiation
In their paper, Matus and L'Ecuyer [2017], hereinafter ML17, discriminate clouds as a function of their phase and quantify their radiative impact using global scale observations of the A-train satellite constellation combined with reanalysis data. They provide a complete and global view of the cloud radiative effect at the surface, the top of the atmosphere, and within the atmosphere, which was not possible before the era of active sensor satellites. In the past, although we were able to precisely observe the fluxes at the TOA, surface fluxes were limited to sparse ground-based measurements and relied on many assumptions, adding large uncertainties to the estimates. ML17 obtained their pioneering results by taking advantage of the active sensing capability of the CALIPSO and CloudSat satellites, which enable global scale measurements at the surface. In addition, ML17 quantify cloud radiative effects as a function of cloud phase, which (i) gives new perspective on how cloud phase affects the climate and (ii) provides new valuable information that can be used to assess the relationship between cloud phase and radiation in GCMs. This represents an important step toward constraining cloud phase simulations in climate models and may ultimately reduce the uncertainties in cloud feedback processes and climate sensitivity estimates produced by GCMs.
ML17 show that clouds in general contribute to a negative cloud radiative effect at the TOA and the surface by reflecting more shortwave (SW) radiation than they trap longwave (LW) radiation, while they warm the atmosphere by absorbing SW radiation and by absorbing more LW radiation than they emit. ML17 show that mixed-phase clouds (8% occurrence) are less frequent than ice (20%) or liquid clouds (21%) and account for 10% of the total cloud frequency. However, their net radiative impact relative to all clouds is around 20%. Clouds categorized as “multilayered” (i.e., profiles that contain cloud layers of more than one phase) occurred 25% of the time. It is likely that some of the “multilayered” clouds were in fact mixed phased, but they were not detected as such by the instruments, and thus the cloud occurrences and radiative effects reported above could be viewed as lower bounds. In the tropics, the precipitating mixed-phase clouds dominate the overall mixed-phase cloud radiative effect, whereas nonprecipitating mixed-phase clouds are more important at middle and high latitudes. Overall, the atmosphere is cooled by the mixed-phase clouds in polar regions but warmed in the tropics (Figure 2a, blue text). On one hand, this modifies the general atmospheric circulation by enhancing the equator-to-pole temperature gradient, which generates a strengthening of the poleward atmospheric heat transport (Figure 2a, blue text). On the other hand, at the surface, mixed-phase clouds cool the oceans in the tropics while they warm it in the poles (Figure 2a, green text). This may contribute to a reduction of the oceanic poleward heat transport, although local heat exchange between the surface and the atmosphere will complicate this picture. Nevertheless, the above results confirm the influence of mixed-phase clouds on both the atmospheric and the oceanic circulation, rendering them particularly interesting for the climate system as a whole.
7 Implications for Climate Models
Supercooled liquid clouds, and by extension mixed-phase clouds, are known to be underestimated in the last generation of climate models. Given the findings of ML17, an underestimation of mixed-phase clouds in the tropics could cause a cooling bias within the atmosphere along with a warming bias at the surface, which would respectively limit the equator-to-pole heat transport and require compensating errors in GCMs in order to get the correct TOA fluxes. In polar regions, models are known to have problems reproducing observations when it comes to supercooled and mixed-phase clouds [e.g., Cesana et al., 2012; Klein et al., 2009]. Adding more of these clouds in the models would thus increase the cooling of the atmosphere in polar regions, strengthening the equator-to-pole temperature gradient. In addition, it would enhance the warming of the surface, which currently warms faster than the rest of the globe in response to a doubling of CO2, according to the last IPCC report.
Changing the cloud phase to allow more liquid clouds over ice clouds would increase the optical depth of clouds (how much clouds attenuate the solar radiation) and thus their albedo (how much solar radiation is reflected back to space) [e.g., Klein et al., 2009; Tsushima et al., 2006]. This change in cloud phase is expected to result from a warmer climate. More liquid droplets as opposed to ice crystals in clouds would lead to more SW reflection and cloud-top LW emission to space (cooling effect), along with slightly more cloud LW radiation emitted to the surface (warming effect) and less precipitation (Figure 2b, red). As a direct consequence of this phase change, the surface temperature increase would be weakened at high latitudes (Figure 2b), while it would be amplified over Greenland and Antarctica—the SW effect being less effective over high reflective surfaces. In the global mean, this cloud phase feedback is negative (cooling of the surface). However, an atmosphere with more supercooled liquid would also leave less possibility for this response of ice-to-liquid cloud phase change to warming, which would have weaker negative cloud phase feedback [Tsushima et al., 2006; Kay et al., 2014; McCoy et al., 2014; Zelinka et al., 2013]. Recent work by Tan et al. [2016] showed that allowing more supercooled liquid clouds to form at lower temperatures in a GCM, in better agreement with observations, could lead to a substantial increase of the model's climate sensitivity via this weakened cloud phase feedback. If that behavior was found to be robust in other GCMs, it would suggest that models might underestimate the increase of surface temperature in future climate, as most of them underestimate the amount of supercooled liquid clouds, particularly at low temperature (T < −15°C).
8 Open Questions and Future Work
8.1 Comparison With Climate Models
An obvious next step would be to analyze how cloud phase and radiation simulated by climate models compare with these new observations. This would help identify flaws in GCM simulations of the relation between cloud phase and radiation, and provide guidance for further model development toward a reduction of uncertainties in climate projections. Although a like-for-like comparison is challenging because the models do not simulate the same quantities that are measured by the A-train data sets, a qualitative comparison would still give new insights into GCMs ability to simulate mixed-phase clouds and their impact on radiation.
8.2 Supercooled Liquid- and Ice-Containing Clouds
While ML17 emphasized the importance of mixed-phase clouds in the Earth's radiation budget (particularly at middle and high latitudes), they did not explore completely the effect of all clouds containing supercooled liquid (i.e., clouds containing only supercooled liquid clouds in addition to mixed-phase clouds). The models fail to reproduce mixed-phase clouds mostly because they are not able to correctly simulate supercooled liquid clouds in the first place, in addition to struggling with the cloud phase transitioning processes (e.g., change from supercooled liquid to ice). Quantifying the importance of all clouds containing supercooled water would yield a more complete global picture. Similarly, a further division of ice clouds into two categories, one for the mixed-phase cloud temperature range and another for pure ice cloud regimes (i.e., colder than −40°C) would be tremendously helpful for the modeling community. This would make it easier to identify the cause of, and correct, biases identified in GCMs relative to the new data.
8.3 Effect of Cloud Phase on Precipitation
In addition to its value for the detection of radiation biases associated with cloud phase flaws in climate simulations, this data set could also be used to better understand precipitation biases in the tropics and subtropics, which is another long-standing problem in climate models. ML17 demonstrated the prevalence of mixed-phase clouds in convective tropical clouds. GCMs often use a different microphysics parameterization for convective and stratiform clouds and generally simplify the representation of phase transitions in convective clouds even more than for stratiform clouds. For example, one could certainly expect to affect the simulated precipitation and in turn the radiation by adding mixed-phase clouds in the tropics to better match cloud phase observations; but would that improve the simulations with respect to both precipitation and radiation observations?
8.4 Vertical Structure
The novelty of two out of the three satellites used by M17, i.e., CloudSat and CALIPSO, is their ability to probe the entire low atmosphere (0–20 km)—most of the time—and provide detailed vertical structure of the clouds. Coupled to other satellite information, reanalysis data, and forward models, it is therefore possible to compute radiative exchanges not only at the TOA, at the surface, or within the atmosphere but also for multiple levels from the ground to the tropopause. The analysis of vertical profiles of radiation using the A-train data set has already been accomplished in both observational and modeling studies [Oreopoulos et al., 2016; Li et al., 2016; Cesana et al., 2017], but none looked at the impact of the cloud phase on vertical profiles of radiation.
8.5 On the Importance of Synergetic Data Sets
Finally, the study of M17, among others [Stephens et al., 2012; Okamoto et al., 2010; Mace et al., 2009; Delanoë and Hogan, 2010, Zhang et al., 2010], highlights the need for synergistic methods to observe Earth's climate through satellite constellations. Combined measurements made it possible to observe the detailed vertical structure of the cloud phase. The strengths of some instruments tend to compensate for the weaknesses of the others and give us access to new information that could not be found using separate instruments alone. Satellite constellation missions have proven to significantly increase our knowledge regarding the climate, and similar multiinstrument space missions should be encouraged in the future. Upcoming cloud-oriented satellites (e.g., EarthCARE, launch scheduled in 2018 [Illingworth et al., 2015]) are expected to carry similar instruments and will thereby allow for long-term measurements and monitoring of cloud phase changes over time, ultimately bringing still new perspectives and discoveries to the field of climate science.
Acknowledgments
G.C. was supported by a CloudSat-CALIPSO RTOP at the Goddard Institute for Space Studies. T.S. was supported by the National Science Foundation under grant AGS-1352417. Both authors state that there is no conflict of interest. The observational data set mentioned here can be downloaded on the CloudSat website [http://www.cloudsat.cira.colostate.edu/data-products/level-2b/2b-flxhr-lidar].
