Report

From aerosol-limited to invigoration of warm convective clouds

See allHide authors and affiliations

Science  06 Jun 2014:
Vol. 344, Issue 6188, pp. 1143-1146
DOI: 10.1126/science.1252595

Invigorating convection in warm clouds

Atmospheric aerosols—tiny airborne particles—affect the way clouds form and how they affect climate. Koren et al. investigated how the formation of warm clouds, such as those that form over the oceans, depends on pollution levels (see the Perspective by Remer). Aerosols affect cloud formation in cleaner air disproportionately more than in more polluted air. Before the widespread air pollution of the industrial era, it seems, warm convective clouds may have covered much less of the oceans than they do today.

Science, this issue p. 1143; see also p. 1089

Abstract

Among all cloud-aerosol interactions, the invigoration effect is the most elusive. Most of the studies that do suggest this effect link it to deep convective clouds with a warm base and cold top. Here, we provide evidence from observations and numerical modeling of a dramatic aerosol effect on warm clouds. We propose that convective-cloud invigoration by aerosols can be viewed as an extension of the concept of aerosol-limited clouds, where cloud development is limited by the availability of cloud-condensation nuclei. A transition from pristine to slightly polluted atmosphere yields estimated negative forcing of ~15 watts per square meter (cooling), suggesting that a substantial part of this anthropogenic forcing over the oceans occurred at the beginning of the industrial era, when the marine atmosphere experienced such transformation.

How changes in aerosol concentrations (and properties) affect clouds and what are the derived cloud feedbacks are challenging questions that carry with them a substantial part of the uncertainty in our understanding of anthropogenic effects on climate (1, 2). The hypothesis of cloud invigoration by aerosols links the cloud’s vertical development to aerosol loading. A chain of processes and feedbacks ties the aerosol effect on the droplet-size distribution to dynamic effects, resulting in deeper and bigger clouds (36). Because the natural variation in convective systems is large and sensitive to environmental conditions, compelling evidence for the invigoration effect is difficult to obtain. Most studies that do suggest invigoration link the effect to deep convective clouds with a warm base and mixed-phase or cold top (4, 7). A few observational studies have suggested warm clouds’ invigoration by aerosols (8, 9), but most of the modeling studies suggest smaller warm clouds in a higher aerosol-loading environment resulting from enhanced evaporation (10, 11).

Cloud-drop formation requires aerosols that serve as cloud-condensation nuclei (CCN) to lower the energy barrier for activation. A hypothetical aerosol-free atmosphere would probably be mostly cloud-free. Therefore, theoretically, in such a clean environment, a small increase in aerosol loading could produce a very dramatic change from a cloud-free to partly cloudy atmosphere. When do clouds stop being aerosol-limited? Here, we argue that the aerosol-invigoration effect can be viewed as an extension of the concept of aerosol-limited clouds, where cloud development (by measures of liquid water mass, for example) is limited by the availability of CCN (1214). We propose that the two regimes can be combined and that the cloud-invigoration concept simply suggests that, in many cases, clouds are aerosol-limited up to much higher aerosol concentrations, well within the anthropogenic aerosol levels.

To reduce the inherent complexity and to focus on the transition from aerosol-limited to cloud invigoration, we looked for pristine areas with low variability of meteorological conditions that support the formation of warm-but-developed convective clouds. Our working hypothesis was that, despite the fact that we are focusing on areas with environmental conditions that favor deeper clouds, the forming clouds would often be less developed because of the clean conditions. Figure 1 shows that clean areas are more prevalent over the Southern Hemisphere. In particular, there is an atmospheric clean belt that marks the subtropical high in the vicinity of the so-called Horse Latitudes. The black contour marks clean areas that are characterized mainly by warm convective clouds. During June to August 2007, the area that showed the optimal conditions was over the southern Pacific (13°S to 22°S, 121°W to 130°W; Fig. 1, green box). The analysis was also performed over the southern Atlantic and Indian oceans (Fig. 1, yellow boxes, and fig. S4) (15).

Fig. 1 Averaged AOD over the oceans for June, July, and August 2007.

Areas marked by a black contour represent pristine oceanic regions with AOD < 0.1 and warm convective clouds. The green box marks the main study area over the Pacific, and the yellow boxes mark the study areas over the Atlantic and Indian oceans (all boxes are 9° by 9°). EQ, Equator.

Understanding the aerosol effect on developed warm convective clouds forming in a pristine region with relatively low cloud coverage and relatively steady meteorological conditions has several critically important qualities: (i) it provides a hint as to the relevant transition from preindustrial to industrial atmosphere; (ii) some of the key microphysical processes are nonlinear with respect to aerosol concentration, showing the highest sensitivity in the transition from clean to slightly polluted conditions; (iii) focusing on regions with relatively steady meteorology (fig. S1) can reduce the likelihood of changes in meteorological conditions explaining changes in both cloud and aerosol properties; (iv) measuring aerosol near clouds poses many challenges (16), such as cloud contamination (17), aerosol humidification (18), and cloud secondary illumination [three-dimensional (3D)] effects (19). Therefore, studying the aerosol effect on clouds in regimes that exhibit a small cloud fraction dramatically reduces the likelihood of encountering such challenges; (v) moreover, the dynamic interactions between clouds within a given cloud field are smaller in regimes with a small cloud fraction, further reducing complexity; (vi) last, one cannot understand aerosol effects on deep convective clouds with mixed and cold parts without fully understanding how aerosols change the warm processes that serve as the initial and boundary conditions for the upper parts.

Three types of global daily databases for 92 days between June and August 2007 were used. Moderate Resoluation Imaging Spectroradiometer (MODIS) Aqua data were used for cloud properties and aerosol optical depth (AOD) (20, 21), serving as a proxy for CCN concentration (22). Rain rates were obtained from the Tropical Rainfall Measuring Mission (TRMM) satellite measurements (23) and meteorological data from the Global Data Assimilation System (GDAS) (24). All data sets were projected to the Aqua passage time [1:30 pm local time (25)]. Cloud fraction, cloud top pressure, cloud top temperature (both are measures of vertical cloud development), and rain rate were sorted as a function of AOD and averaged, creating 100 scatter points (Fig. 2, top row).

Fig. 2 Associations between cloud properties and aerosol loading (estimated by AOD).

(Top) All data. (Middle) Data filtering by the 400-hPa geopotential height using only the lower one-third subset. (Bottom) Data filtering by using the upper one-third of the 400-hPa geopotential height subset. First (left) column shows rain rates versus AOD. Second column, cloud top pressure (P) versus AOD. Lower P values (or colder cloud top temperature, third column) indicate deeper clouds. Fourth (right) column, cloud fraction versus AOD.

Positive trends between aerosol loading and cloud properties are not always an indication of causality. Before declaring an aerosol effect on clouds, one should check that changes in the environmental conditions (meteorology) cannot explain the observed correlations. Which atmospheric variables can approximate well the dependency of clouds on the environmental conditions over the study areas? To answer this question, we analyzed 286 GDAS meteorological variables against the measured aerosol, cloud, and rain properties (fig. S2) (16). We found a range of geopotential heights (Zg, between 300 and 700 hPa) that correlated relatively well with cloud properties but not with aerosols (15). Restricting Zg to a narrow range can limit the variation in the meteorological conditions that control warm convective cloud formation in our study areas while applying no (or very small) restriction on AOD levels. The data set was divided into three equal-sized subsets of high, medium, and low Zg ranges at the 400 hPa level (Z400) (fig. S4). The analysis of AOD versus cloud and rain properties was performed for each subset separately. The middle and bottom rows in Fig. 2 show the trends for the subsets with mean Z400 of ~7530 m and ~7590 m, respectively. The former represents conditions that promote deeper clouds and consequently higher rain rates. An orthogonal microphysical effect is observed in all subsets.

By focusing on the transition from clean background aerosol conditions of AOD = 0.06 (26, 27) [equivalent to ~100 CCN/cm3 (22)] to slightly polluted conditions of AOD = 0.1 (~300 CCN/cm3; in many urban areas AOD = 0.1 is as clean as it gets), Fig. 2 reveals important changes in all cloud properties. For the convection-promoting subset (middle row), a significant decrease in the cloud top pressure is shown, suggesting an over-1-km increase in the cloud’s vertical extent with the increase in AOD. In addition, we see an increase in the cloud fraction from 0.3 to 0.6 and a significant increase in rain rate from almost no rain to ~0.2 mm/hour.* For the subset characterized by a larger geopotential height that dictates shallower clouds (bottom row in Fig. 2), we see an increase of a few hundreds of meters in the cloud top, an increase in the cloud fraction from ~0.2 to ~0.4, and no significant change in the low rain rates. Identical trends were observed over the Atlantic and Indian oceans (fig. S4). Cloud fraction, cloud top height, and rain rates all showed an increase with the increase in AOD.

To further explore the transition from aerosol-limited to cloud invigoration, we ran a detailed numerical experiment using a bin-microphysical model. The Tel Aviv University bin-microphysics axisymmetric model (TAU-CM) was used with detailed treatment of warm microphysical processes (15, 28, 29). We define here the effective terminal velocity (Ve) as the average, mass-weighted, terminal velocity of the water droplets within a given volume element. Similar measure was used before in the context of rain (30). An analysis of Ve for cloud droplets revealed how aerosol effects on the redistribution of droplet sizes on a microscale propagate to affect the vertical water distribution on the macroscale of the entire cloud. It can be shown that Ve measures the vertical displacement in time of the center of gravity (31) of the liquid water within the volume element for a zero updraft reference. Therefore, the superposition of Ve and the air vertical velocity estimates the center of gravity displacement taking into account all processes (15).

Conceptually, the aerosol’s microphysical effects can be divided into a bulk effect on vapor-consumption efficiency and a series of delays in the onset of collection-related processes. Clouds that form in a very clean environment exhibit limitations on the droplet surface area available for condensational growth, and therefore the ambient supersaturation is weakly (slowly) consumed, allowing the sparse droplets to grow relatively fast. The condensed mass is small, and there is early initiation of the collection process. The collecting drops quickly fall as a weak rain. An increase in the CCN concentration leads to a larger condensed mass (Fig. 3, A and B). The effect of aerosols on the droplet-size distribution drives the above-mentioned delays. The initial one is the delay in the onset of the collision-coalescence process that transforms the mass from smaller to larger droplet bins (Fig. 3C). This delay initiates an additional delay in the onset of significant Ve (Fig. 3, E and F), and both imply a delay in the onset of precipitation (Fig. 3D).

Fig. 3 Total cloud view of key processes as a function of time (t) for five aerosol levels.

(A) Total mass of the cloud, (B) total condensed mass per unit time (mass transferred from water vapor to liquid), (C) total collected mass per unit time (mass transferred from the smaller bins to the bigger ones), (D) rain rate, (E) effective terminal velocity, and (F) integrated view of the velocities as the sum of the air vertical velocity and the effective terminal velocity. Blue curves show results for 5 aerosols per cm3 (#/cc); red, 25; black, 125; magenta, 250; green, 500.

Driven by these delays, the invigoration of warm clouds occurs in the first stage of the cloud’s evolution, when polluted clouds condense more water mass, release more latent heat, and push this mass higher in the atmosphere before the onset of significant Ve and the development of significant (mass-driven) drag forces. Such delays provide the time for the extra cloud development. Later, once the collection process starts in the polluted cloud, more mass is transferred to the larger size bins, and Ve increases to larger falling velocities, leading to a stronger depletion of the cloud’s water as reflected by stronger rain rates (Fig. 3D). In addition, an increase in the entrainment process is expected in polluted clouds, driven by more efficient evaporation of the smaller droplets and stronger velocities [for further discussion on the modeling results, see (15)].

Starting our analysis from the extreme clean cases [concentrations below ~25 CCN/cm3 (3234)] makes it easier to argue that such clouds cannot condense or hold significant amounts of water. Indeed, the sharper changes in cloud properties, shown from both observational and numerical results (Figs. 2 and 3 and fig. S4) (6), occur when the aerosol loading shifts from extremely pristine to slightly polluted. Nevertheless, clouds do not stop being aerosol-limited in this range. The invigoration effect tends to saturate at much higher aerosol loading. Both the observational and modeling analyses suggest that the cloud’s liquid water mass, horizontal and vertical extent, and rain rates will continue to increase up to polluted conditions of a few hundred CCN/cm3 (AOD ~0.3). The model results demonstrate that polluted clouds exhibit a series of delays in the onset of collection processes, significant Ve and rain. All act to enhance the vertical and horizontal dimensions of the polluted cloud.

Clouds can be viewed as a complex system with many coupled processes acting together and orchestrated along a delicate timeline. The exact timing of each process’s initiation and its magnitude control the overall properties of the cloud. The use of a single-cloud model allowed us to separate key processes and to follow their magnitude and timing. However, such a model has some limitations, because it does not account for interactions between clouds or for the impact of clouds on their environment. Our satellite data analysis is also not completely free of residual cloud contamination on the aerosol retrieval and of some environmental effects that are not captured by the meteorological slicing. Nevertheless, we show a trend that survives restrictions on the meteorological variance in three different places over the pristine oceans. The effect is shown in the transition from extremely clean to slightly polluted conditions for low cloud fractions where cloud contamination and interactions between clouds are minor.

To understand the anthropogenic aerosol effect on the climate system, we have to know what the aerosol conditions were in the preindustrial era (3538). It has been recently argued that this is one of the toughest challenges we face (39). The average AOD of the pristine oceanic atmosphere today is estimated to be ~0.06 (26, 27). This by itself suggests that, over large marine areas, clouds are aerosol-limited. In fact, our results suggest that clouds forming in a pristine atmosphere may be the most sensitive to changes in aerosol loading (40). We hypothesize that, over the pristine oceans, higher pollution levels could push more clouds to cross the freezing level and to undergo ice processes.

Cloud invigoration produces two opposing radiative effects. On one hand, larger and thicker clouds increase the reflected shortwave (SW) radiation back to space and cool the atmosphere. On the other hand, deeper clouds have colder tops and therefore emit less longwave (LW) radiation to space, hence warming the atmosphere (41). By using the Clouds and the Earth’s Radiant Energy System (CERES) data (42), we analyzed trends in top-of-atmosphere SW and LW fluxes as a function of AOD. Indeed they show (fig. S7) a strong increase in the SW flux (cooling), partly compensated by a decrease in the LW flux. Specifically, in the all-data case the AOD transition from 0.05 to 0.15 produces a cooling effect of 27 W/m2 in the SW, whereas half of the forcing is counteracted by warming in the LW, yielding a net cooling effect of ~14 W/m2. These results obtained over pristine aerosol regimes suggest that a great portion of the anthropogenic forcing over the oceans occurred in the early stages of the industrial era, when the average marine atmosphere changed from pristine to slightly polluted (35). If this is true, it means that the preindustrial globe should be considered differently from today’s globe. At least over the oceans, the coverage of warm clouds should be regarded as having been much smaller than it is today.

*Correction (25 June 2014): The sentence was corrected to "~0.2 mm/hour" from "~2 mm/hour."

Supplementary Materials

www.sciencemag.org/content/344/6188/1143/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S7

References (4347)

References and Notes

  1. See supporting data on Science Online.
  2. Acknowledgments: The research leading to these results received funding from the European Research Council (ERC) under the European Union’s Seventh Framework Programme (FP7/2007-2013)/ERC Grant agreement no. 306965 (CAPRI).
View Abstract

Stay Connected to Science

Navigate This Article