Iodo

CCN number concentration JVCCH (m

Figure 3. Variations of the rate change of Nccn as a function of Nccn for different production strengths. The zero intercept represents an equilibrium condition, but only those marked by black dots are stable modes where any deviation from that Nccn value will eventually return to the stable conditions. (From Fu, 2002.)

CCN number concentration JVCCH (m

Figure 3. Variations of the rate change of Nccn as a function of Nccn for different production strengths. The zero intercept represents an equilibrium condition, but only those marked by black dots are stable modes where any deviation from that Nccn value will eventually return to the stable conditions. (From Fu, 2002.)

Baker (1993) proposed an interesting testability hypothesis to describe the possible relationship between CCNs and cloud drops. With a mixed-layer approach, she analyzed theoretically the variation of CCN and cloud drop number concentrations in marine stratiform clouds. Figure 3 is a reproduction of her analysis, but using a newer microphysical parameterization (Fu, 2002). Each curve represents the overall rate change of Nccn as a function of Nccn itself under a particular strength of CCN production, S. The microphysical processes included are Brownian coagulation of CCNs, self-collection of cloud drops, conversion of cloud drop to raindrop (autoconversion), precipitation fallout, and scavenging of CCNs by rain. On each curve one can see 1-3 intercepts at zero which represent the equilibrium (steady) states. However, only the intercepts at the right or left (marked by the red dots) are stable modes where any perturbation of Nccn from that value will cause a restoring force to bring it back to equilibrium. This means that the atmosphere has the capability of cleansing itself, but only when it is not heavily polluted (to the left of the central intercept). If all of a sudden too much CCN is put into the air (the value of Nccn greater than that at the central intercept), then Nccn will shift toward the other stable mode such that the polluted conditions will persist, and does not return to the clean mode unless the production strength S is reduced.

Furthermore, S must be reduced drastically to revert to the clean mode due to the hysteresis effect shown in Fig. 4(a). Because the two stable modes do not coexist all the time, the condition of jumping from clean to dirty modes occurs at a different production strength than that of jumping from dirty to clean modes. For example, the clean conditions may be sustained until S reaches about 700 m-3 s-1; but once it jumps to a dirty condition, aerosol production must be reduced to about 200 m-3 s-1 in order to return to the clean mode. Cloud albedo also exhibits such a hysteresis cycle, jumping between the clean value of about 0.2 and the dirty value of about 0.7 [Fig. 4(b)]. However, the simulation of Ackerman et al. (1994) did not show such a hysteresis effect. Fu (2002) pointed out that the discrepancy could be due to the transient effect as the time to reach equilibrium often takes several days, which might be much longer than the time of simulation spent by Ackerman et al. (1994). In addition, during the long time required to approach equilibrium, the properties of the cloud, such as the liquid water content and cloud depth, might have changed due to precipitation or dynamic adjustment, and yet Baker (1993) held-these parameters as constant. Although the bistability hypothesis needs further verification, one may nevertheless think of the two stable modes as attractors that the numbers of cloud drops and CCNs tend to approach. As the stratocumulus deck in the marine boundary layer (MBL) has a major impact upon the Earth's radiation budget by reflecting solar radiation, Baker's bistability hypothesis has important implications for the

Figure 4. (a) Equilibrium Nccn as a function of the production strength S. (b) Cloud albedo that corresponds to the equilibrium Nccn . The lower curves represent the scenario of shifting from clean to dirty conditions by sweeping S from low to high values for an initially low Nccn, whereas the upper curve is just the opposite. (From Fu, 2002.)

Figure 4. (a) Equilibrium Nccn as a function of the production strength S. (b) Cloud albedo that corresponds to the equilibrium Nccn . The lower curves represent the scenario of shifting from clean to dirty conditions by sweeping S from low to high values for an initially low Nccn, whereas the upper curve is just the opposite. (From Fu, 2002.)

limitation of atmospheric self-cleansing capability pertinent to the injection of anthropogenic pollutants.

The effect of aerosols on cloud drop concentration may occur due to not only anthropogenic pollution but also natural biogenic production. One of the main sources of natural condensation nuclei (CNs) over the ocean is believed to be dimethylsulfide (DMS) excreted by phy-toplankton and liberated into the atmosphere, where it is photochemically oxidized to form sulfate particles and become CCNs. By providing CCNs, DMS may determine the properties of the cloud. Thus, the production of DMS in oceans and its subsequent transformation into CCNs act as a negative feedback mechanism to counteract global warming (Shaw, 1983; Charlson et al., 1987; Easter and Peters, 1994), which is called the DMS-cloud-climate hypothesis. It is speculated that the productivity of DMS is temperature-dependent, such that any increase in the ocean surface temperature (expected outcome of greenhouse warming) would enhance CCN production, which, in turn, would increase cloud reflectance (albedo) and cause a cooling effect similar to the Twomey effect. Thus DMS (or anthropogenic SO2, for that matter) is sometimes referred to as the "anti greenhouse gas." The cooling of the sea surface or the blocking of sunlight by denser clouds could reduce the productivity of phytoplankton, thus forming a negative feedback loop. Recently, Meskhidze and Nenes (2006) analyzed satellite data and showed that cloud drop number concentration over the regions of the phytoplankton bloom was twice as high as that away from the bloom. The resulting change of short-wave radiative flux at the top of the atmosphere may be comparable to the aerosol indirect effect over highly polluted regions. However, the authors proposed that the changes were caused by secondary organic aerosols (also hygroscopic) formed from isoprene that was released by the phytoplankton. Yet, the DMS-cloud-climate hypothesis still cannot be discarded.

CCNs not only affect cloud-physical properties and albedo but may also influence precipitation formation. As cloud drops get smaller due to there being more CCNs, they collide less efficiently, such that the warm rain formation mechanism may be retarded. The sensitivity test of Teller and Levin (2006) nicely shows the decrease of total precipitation with increasing CCN concentration. Cheng et al. (2007) also showed that in more polluted conditions (i.e. higher CCN concentration) less rain is formed, whereas more cloud water is retained in the cloud (Fig. 5). Their results exemplify both the first and the second indirect effect of Twomey. Note that the influence of CCNs on precipitation might deviate from those discussed above when ice-phase processes are involved, because there could be stronger growth by riming and latent heat release in the cloud when more liquid water is retained. Giant CCNs discussed in the next section are an even clearer example of aerosols enhancing precipitation.

Quantitative estimates of indirect aerosol effects were provided in numerous articles (e.g. Lohmann and Feichter, 2005; IPCC, 2001, 2007). However, care must be taken that there exist large uncertainties in the estimation or verification of aerosol indirect effects. Besides our limited knowledge about the processes involved, a few other factors may contribute to the uncertainties. For example, for those models that are able to diagnose the mean cloud drop radius (usually volume-weighted), they also use it to represent the effective radius in determining the cloud radiative effects. Yet, these two radii may differ significantly due to the dispersive nature of the drop size distribution (Martin et al., 1994; Liu and Duam, 2002). Liu and Duam (2002) suggested that Twomey's first indirect effect of aerosols may be partially offset due to such dispersive effect, whereas Chen and Liu (2004) further suggested that the fraction of the offset should be about one-ninth, which is supported by the calculations of Cheng et al. (2007).

Twomey's second indirect effect seems to be a sound hypothesis. Yet, there are numerous examples where increases in CCNs do not lead to higher liquid water content clouds and longer lifetimes via the second indirect effect. For instance, Ackerman et al. (2004) demonstrated that the entrainment of overlying dry air above boundary layer stratocumulus may mitigate the second indirect effect and sometimes even results in cloud water decreases when the humidity of the overlying air is very low. In fact, more and smaller cloud drops may enhance evaporation at the downdraft region of cloud edges, which leads to a lower cloud fraction, cloud size, and depth (Teller and Levin, 2006; Xue and Feingold, 2006). Guo et al. (2007) also showed such a positive indirect effect due to stronger entrainment when large-scale subsidence is weak. Tao et al. (2007) pointed out another effect of enhanced evaporation. From simulations using a two-dimensional cloud resolving model, they found that in a tropical cloud more aerosols caused more but smaller raindrops, which evaporate faster and cause a stronger downdraft, in the stratiform region of the clouds. The stronger cold pool associated with the enhanced downdraft induced stronger low-level convergence and thus stronger convection, and this resulted in greater precipitation. But, for a continental convective cloud, increasing CCNs does result in less precipitation. More and smaller cloud drops may also suppress low-level precipitation, elevating the release of latent heat at the upper levels and the onset of precipitation, which result in more intense convection (Rosenfeld and Woodley, 2000). These examples caution us that the effects of aerosols on clouds and precipitation are not so straightforward. Whether the response follows Twomey's indirect effects depends on many intricate microphysical factors as well as dynamic feedbacks.

2.1.2. Giant CCNs

Some of the hygroscopic aerosols are inherently large in size, such as sea salt generated

Figure 5. Time series of the simulated vertical profiles (in a coordinates) of cloud drop fields averaged over the model domain. The rows from the top down are the number concentration Nc and the mixing ratio Qc of cloud drops, and the number concentration Nr and the mixing ratio Qr of raindrops. The three columns, from left to right, are for initial aerosol types — clean continental, averaged continental, and urban. (From Cheng et al., 2007.)

Figure 5. Time series of the simulated vertical profiles (in a coordinates) of cloud drop fields averaged over the model domain. The rows from the top down are the number concentration Nc and the mixing ratio Qc of cloud drops, and the number concentration Nr and the mixing ratio Qr of raindrops. The three columns, from left to right, are for initial aerosol types — clean continental, averaged continental, and urban. (From Cheng et al., 2007.)

by the breakup of air bubbles in the ocean or by wave tearing. These giant CCNs, with radii of a few microns or more, play very different roles than common CCNs in either the hydrological cycle or Twomey's indirect effects, mentioned in the previous subsection. They not only activate more readily into cloud drops but may also act as rain embryos to initiate the warm rain process because they grow, by condensation, rather quickly to exceed the so-called "Hocking limit" for collision-coalescence. Large, insoluble particles such as mineral dust may exhibit similar properties when coated with hygroscopic materials.

The role of rain embryo that giant CCNs play in cloud processes has been demonstrated to be important for warm rain initiation (Johnson, 1982; Feingold et al., 1999; Lasher-Trapp et al., 2001). This is particularly true for the statocumulus clouds in the marine boundary layer (MBL), where drizzle is the main form of precipitation. Studies have shown that sea salt particles are often present in concentrations similar to those of drizzle drops, implying that these giant CCNs may be the main source of drizzle production (see O'Dowd et al., 1997). The stratocumulus clouds in the MBL are particularly worthy of understanding, because they play an important role in the radiative balance of the Earth (Albrecht, 1989). It has been suggested that drizzle can significantly affect the thermodynamics and energetics of the MBL, leading to profound changes in the cloud amount and cloud structure (Paluch and Lenschow, 1991; Stevens et al., 1998). Strong drizzle may even lead to the collapse of the marine stratocumulus, as suggested by the observational evidence shown by Stevens et al. (2004).

Drizzle may be produced either via the so-called "autoconversion" process or from rain embryos. The former produces rain-sized droplets by a gradual collision-coalescence among cloud drops, whereas the latter immediately introduce droplets with sizes larger than the Hocking limit to initiate fast collision. Traditional bulkwater (also known as "Kessler type") microphysical schemes do not consider the effect of giant CCNs. The only way for them to produce rain in warm clouds is through the autoconversion process. Note that without considering aerosols, the traditional (one-moment) bulkwater schemes cannot simulate Twomey's indirect effects, not to mention the giant CCN effect.

However, whether giant CCNs are crucial to the initiation of drizzle or not depends on how high the cloud drop number concentration (CDNC) is, which in turn is driven largely by variability in the concentration of the smaller (e.g. accumulation mode) aerosol particles, as discussed previously. Under low CDNC, cloud drops are able to grow large enough by condensation to enable efficient autoconversion. Note that when the CDNC is too low, clouds could become optically thin, so that cloud-top radiative cooling could no longer drive vertical mixing, which is another possible cause of the collapse of the stratiform clouds in the MBL (Ackerman et al., 1993). Also note that low CCN and thus low CDNC may cause a stronger autoconversion process that may possibly contribute to the collapse of the marine stratocumulus deck (Khairoutdinov and Kogen, 2000). Under high CDNC (thus small droplet sizes), the autoconversion process tends to shut down due to the Hocking limit restriction, and so the formation of drizzle depends on the amount of giant CCNs present in the air. Feingold et al. (1999) found that giant CCNs are effective in enhancing precipitation in clouds with a CDNC of above 50cm~3 but are ineffective at a lower CDNC. Of course, there must be a sufficient amount of giant CCNs to start with. Figure 6 shows a similar effect of giant CCNs from simulations using the Mesoscale Meteorological Model, version 5 (MM5), with an improved bulkwater scheme that takes aerosols into account (Cheng et al., 2007). The setup of the case presented is the same as that for Fig. 6. One can see that with

Figure 6. Distribution of cloud water (top) and rainwater (bottom) under various amounts of ultragiant CCNs, (radius larger than 10¡1m). The amount of giant CCNs, from left to right is 0, 0.05 and 0.5cm~3, respectively. A greater amount of giant CCNs, causes more rains and retains less water in the cloud.

Figure 6. Distribution of cloud water (top) and rainwater (bottom) under various amounts of ultragiant CCNs, (radius larger than 10¡1m). The amount of giant CCNs, from left to right is 0, 0.05 and 0.5cm~3, respectively. A greater amount of giant CCNs, causes more rains and retains less water in the cloud.

the addition of giant CCNs, progressively more rain is produced while less water is retained in the cloud. Such an effect is particularly strong for polluted conditions where autoconversion is largely inhibited.

However, at least for not-very-polluted situations, the actual cause of such rain enhancement is not just for having more rain embryos. In fact, the introduction of giant CCNs may also enhance the autoconversion process. Because of their strong water-absorbing capabilities, giant CCNs consume water vapor during the activation stage, thus decreasing the maximum supersaturation and reducing the chance for smaller aerosols to be activated into cloud drops (Ghan et al., 1998; Cheng et al., 2007). As less cloud drops are activated, they can grow larger and have a higher efficiency of collision, thus enhancing autoconversion. Furthermore, the rain embryos may reduce the CDNC by accretion, as shown by Cheng et al., (2007). This has the long-term effect of reducing aerosols that can be recycled during cloud evaporation, and is particularly important for marine stra-tocumuli that normally go through many cloud cycles in a relatively short time. These concepts have actually been adopted for warm cloud seeding, where giant nuclei (typically composed of CaCl2) are artificially introduced into the updraft region below the cloud base to not only promote coalescence but also reduce cloud drops and cause those activated to grow bigger (Cooper et al., 1997; Mather et al., 1997). It is also worthwile to note that giant CCNs cause an anti-Twomey's first indirect-effect by reducing CDNC, and an anti-Twomey's-second-indirect-effect by enhancing precipitation. So, giant CCNs play an exactly opposite role to that of typical CCNs in the context of aerosol-cloud-climate interactions.

2.1.3. Effects on cloud, ice

Precipitation formation through ice-phase mechanisms is much more efficient than the warm rain process, provided that a sufficient amount of ice crystals can be formed in the cloud. Ice crystals can be formed most effectively by heterogeneous nucleation with the help of ice nuclei (INs). When there is a lack of INs, ice crystals can also be formed through homogeneous freezing of liquid droplets, but only at low temperatures (typically < — 35°C). It is also theoretically possible for water vapor to nucleate directly into ice crystals, but this would require extremely high supersaturations that are rarely found in the atmosphere. Due to the Raoult effect of freezing point depression, soluble compounds tend to inhibit the freezing of liquid aerosols and cloud droplets, and thus the formation of ice crystals. Yet, liquid droplets normally cannot form without these soluble chemicals. So, when the atmosphere is short of INs, the only viable way to form ice is to freeze the water in haze or cloud drops. In fact, ice clouds formed from homogeneous nucleation are very common, such as the cirrus formed by slow lifting of air or in orographic wave flow at high altitudes (Heymsfield and Sabbin, 1989; Heymsfield and Miloshevich, 1993), as well as the cirrus anvil generated in convective storms (Knollenberg et al., 1993). These ice clouds cover a large portion of the Earth's surface and have a major influence on the global energy budget (Liou, 1986). Simulations of cirrus anvils from convective outflow performed by Chen et al. (1997) using a column model showed that aerosols have strong influences on not only the microphysical structure but also the extent and lifetime of cirrus. As shown in Fig. 7, with five times the aerosol

10 2 10 1 10( 101 103 10' 10'1 10 3 10 1 10 a 10 1 10 1 10° 10 3 10 - 10 1 10" 101

Number Concentration, L1 Ice Water Content, gm3 Effective Diameter, mm

Figure 7. Simulation of ice microphysics in a tropical convective outflow anvil. The top panels show results of applying typical maritime aerosol size distribution, and the bottom panels show results with five times the aerosol concentration. The left, center and right panels display the number concentration, mass concentration and effective radius of ice particles. Different curves represent the vertical profiles at different times of evolution. (From Chen et al., 1997.)

Number Concentration, L1 Ice Water Content, gm3 Effective Diameter, mm

Figure 7. Simulation of ice microphysics in a tropical convective outflow anvil. The top panels show results of applying typical maritime aerosol size distribution, and the bottom panels show results with five times the aerosol concentration. The left, center and right panels display the number concentration, mass concentration and effective radius of ice particles. Different curves represent the vertical profiles at different times of evolution. (From Chen et al., 1997.)

concentration, the number concentration of ice particles may increase by more than one order of magnitude. With more aerosols, there will be more activation to form more and smaller cloud drops; smaller cloud drops also cause a reduction of droplet coalescence or accretion by ice, and thus more droplets will remain for eventual freezing into ice particles. As a consequence, more ice water but smaller effective radii are formed in the anvil, and the anvil becomes thicker and lasts longer. At 12 hours of simulation time, the vertical extent (as defined by the ice water content) increases from 3 km to 5.5 km as a result of the increase of aerosols.

The evolution from an aerosol particle to an ice crystal is illustrated in Fig. 8. Dry hygroscopic aerosols first obtain water when the humidity reaches a particular value called the deliquescence point, which is actually a discontinuity of the Köhler curve caused by the solubility limitation, as shown in Fig. 9 (Chen, 1994). Once deliquesced, the aerosol reaches an equilibrium size determined by the traditional Koöhler curve and stays in a haze state. Further lifting and cooling of the air causes the humidity to rise and the aerosol droplet to swell. If the air is not cold enough (but still colder than — 35°C), it needs to reach over

Was this article helpful?

0 0

Post a comment