1. Introduction
Atmospheric aerosols act as cloud condensation nuclei (CCN) and are therefore important in cloud formation processes. Anthropogenic CCN generally give rise to an increase of the cloud droplet number concentration (CDNC), which for a constant amount of liquid water reduces the average cloud droplet size and increases cloud albedo (first indirect effect; Twomey 1977). As a consequence, precipitation processes and cloud lifetime may be altered (second indirect effect; Albrecht 1989). For mixed-phase clouds, such as deep convective clouds, the effect of an increased aerosol concentration is more elaborate. For these clouds, aerosols may affect not only the droplet formation but also the ice formation processes within the cloud, directly through heterogeneous freezing and indirectly through changing the droplet size and thereby the freezing temperature. Some cloud-resolving model (CRM) studies of single (isolated) deep convective clouds show a decrease of precipitation with increasing aerosol concentration, whereas others show a precipitation increase with increasing aerosol concentration [for examples of both types of response, see Ekman et al. (2007) and Rosenfeld et al. (2008), and references therein]. Several of the model studies referenced above utilize a simplified description of the aerosol population chemistry and dynamics in which an initial aerosol size distribution/concentration is assumed and the initial aerosol composition is prescribed. In some of the models, the aerosol population is advected and scavenged throughout the simulation (e.g., Wang 2005a), whereas other models utilize a time-invariant aerosol concentration (e.g., Seifert and Beheng 2006). Different methods of activating aerosols into cloud droplets are also used, such as empirical formulations (e.g., Wang 2005a; van den Heever et al. 2006) or Koehler theory (e.g., Ekman et al. 2007; Teller and Levin 2006).
Rosenfeld et al. (2008) suggested that if the warm phase precipitation of a cloud is reduced, the amount of cloud water reaching the freezing level increases. This may in turn enhance the latent heat release and give higher updraft velocities, leading to more vigorous deep convection. On the other hand, smaller cloud droplets may also imply a shift of the heterogeneous freezing level to colder temperatures, which should decrease buoyancy and updraft velocity as more water mass has to be transported to higher levels. Based on this line of reasoning, an increase of the aerosol concentration should invigorate convection until an optimal aerosol loading is reached. Thereafter convection should be suppressed.
Fan et al. (2009) also noted the contradictory results between different studies of aerosol effects on deep convection and suggested that vertical wind shear and relative humidity play an important role in moderating the aerosol impact on deep convection. They argued that increased aerosol concentrations always suppress convection under conditions of strong vertical wind shear as the suppression resulting from the increased evaporative cooling dominates over the invigoration because of increased latent heat release. On the other hand, for weak vertical wind shear, the increase in latent heat release is at first larger than the increased evaporative cooling, so that convection is enhanced until an optimal aerosol loading is reached. In absolute terms, the effect of increasing aerosol concentration on convective strength was found in the study by Fan et al. (2009) to be the largest in environments with strong vertical wind shear and high humidity.
In the present study, we examine the sensitivity of a single deep convective cloud to varying aerosol concentrations and how this aerosol-induced sensitivity may depend on the complexity of the aerosol model as well as the aerosol activation parameterization. We also investigate whether the size of the aerosols may affect the aerosol-induced sensitivity. To simplify the analysis, only the aerosol effect on liquid droplet formation is considered [i.e., the number of aerosols available as ice nuclei (IN) is kept constant]. No direct radiative effects of aerosols are included. Furthermore, we constrain our study and simulate a range of aerosol concentrations where a monotonic increase of convective strength with increasing aerosol concentration should be seen according to the theoretical model proposed by Rosenfeld et al. (2008). Aerosol-induced changes on isolated deep convective clouds may result in dynamical feedbacks promoting or inhibiting further cloud formation (van den Heever and Cotton 2007). These types of effects have not been considered in the present study.
The paper is organized as follows: in section 2 we will present the model, the simulated case, and the different sensitivity simulation series. Thereafter follows a presentation of the results in section 3. In section 4 the results are summarized and discussed.
2. Model and simulated case
The dynamics–physics module of the CRM consists of the nonhydrostatic momentum equations, the continuity equations for water vapor and air mass density, the thermodynamic equation, and the equation of state (Wang and Chang 1993a). Also included are prognostic equations for the mixing ratios Q as well as number concentrations N of cloud droplets, raindrops, ice crystals, and graupel particles. All unrimed ice crystals and snowflakes are considered as one group (ice crystals), whereas the graupel category considers rimed ice particles as well as frozen raindrops. Every hydrometeor category has a unique spectrum described by the prognostic values of Q and N at each time step. The microphysical transformations are formulated based on a two-moment scheme incorporating the size spectra of particles (Wang and Chang 1993a; Wang et al. 1995). We note that a two-moment microphysics scheme provides certain limitations compared to a size-resolved bin microphysical model; for instance, the transfer of hydrometeors within the cloud droplet category to the rain drop category has to be parameterized as well as the evaporation of all hydrometeors. However, Seifert et al. (2006) examined the characteristics of a maritime deep convective cloud using both a bin microphysics model and a two-moment bulk scheme. They found that the sensitivity of the deep convection to varying CCN concentrations was similar when comparing the bulk and the bin microphysical method.
Nucleation of ice crystals occurs both through homogeneous freezing of liquid particles and by heterogeneous nucleation caused by aerosol particles. All cloud droplets present in a model grid with temperature below −40°C are allowed to freeze immediately through homogeneous freezing. The shape of these ice crystals is given as a solid column (cf. Pruppacher and Klett 1997) and the minimum stable size of them is given as 20 μm on the basis of observations (e.g., McFarquhar and Heymsfield 1996). Thus, the number concentration of ice crystals formed through the condensation/freezing process can be derived using the condensed water content of frozen cloud droplets. Heterogeneous nucleation of ice crystals is in the model described using the parameterization by Cotton et al. (1986). The impact of droplet size on homogeneous and heterogeneous ice crystal formation is not included in the model. Aggregated ice crystals are assumed to be converted to graupel if the ice crystals collect liquid cloud droplets or raindrops and if the dimension Di ≥ D* = 300 μm. The efficiency of the graupel formation process is also dependent on the liquid hydrometeor size. Ice crystals and graupel particles start melting at 0°C and the melting process is a function of particle size according to Wang and Chang (1993a).
A δ-four-stream radiation module based on Fu and Liou (1993) is incorporated in the CRM using predicted concentrations of gases (including H2O and O3) and hydrometeors to calculate radiative fluxes and heating rates. Six bands for the solar and 12 bands for the thermal part of the radiation spectrum are utilized.
The meteorological–chemical part of the CRM has been applied in several studies of the dynamics, microphysics, and chemistry in continental deep convection (e.g., Wang and Chang 1993a,b; Wang and Crutzen 1995) and oceanic deep convection over the Pacific (Wang et al. 1995; Wang and Prinn 1998, 2000; Wang 2005a,b). The chemistry submodule predicts atmospheric concentrations of 25 gaseous and 16 aqueous (in both cloud droplets and raindrops) chemical compounds including important aerosol precursors, such as sulfate and nitrate, undergoing more than 100 reactions as well as transport and microphysical conversions. Results of the chemical and dynamical parts of the model have been compared with available observations including aircraft, radar, and satellite data (Barth et al. 2007). The 3D CRM coupled with the explicit aerosol module has been described and evaluated against observations in Ekman et al. (2004, 2006, 2008) and Engström et al. (2008). In the present version of the model, the horizontal resolution of the model is set to 2 km and the vertical grid interval to 400 m. The model domain covers 400 × 400 × 50 km3.
a. Aerosol module
The evolution in time and space of the aerosol population is described using a multimodal aerosol model originally developed by Wilson et al. (2001) and modified as described in Engström et al. (2008) and references therein. In the present study, four different modes are used to represent the aerosol population. These four modes are nucleation mode aerosols (here defined by d ≤ 23 nm), Aitken mode aerosols (here defined by 23 ≤ d ≤ 100.0 nm), accumulation mode aerosols (here defined by 100 ≤ d ≤ 900.0 nm), and coarse mode aerosols (here defined by d ≥ 900.0 nm). The size distribution within each aerosol mode is assumed to be lognormal and is described by three parameters: number, mass, and standard deviation. To reduce the computational burden, the standard deviation is prescribed. Both number concentrations and mass mixing ratios of the four aerosol modes (i.e., all together eight variables) are incorporated in the cloud-resolving model as prognostic variables undergoing transport, mixing, dry deposition, and nucleation as well as impaction scavenging, in addition to aerosol microphysical processes. The advection scheme used to calculate the transport of these aerosol variables is a revised Bott scheme as described in Wang and Chang (1993a). The nucleation aerosol mode has a continuous source through the formation of new aerosols from H2SO4 (supplied by SO2 oxidation calculated in the chemistry module of the model) and H2O (Vehkamäki et al. 2002). The condensation coefficient as well as the intra- and intermodal coagulation coefficients for each aerosol mode is determined from the theory of Fuchs (1964), using the geometric mean radius of each mode.
In addition to the nucleation scavenging of aerosols through the formation of cloud droplets (cf. section 2), aerosols can also be scavenged through collision with falling raindrops, graupel, or ice crystals [i.e., precipitation (impaction) scavenging]. In the present version of the model, graupel and ice scavenging are only considered for nucleation and accumulation mode aerosols whereas impaction scavenging by rain is considered for all aerosols. In the model, the collision efficiency of aerosols with raindrops E varies with size and is prescribed for the different aerosol bins (cf. Ekman et al. 2004). The removal by raindrops is efficient for either small or large particles whereas the collision efficiency for particles in the 0.1- to 1.0-μm size range is relatively low. Recycling of aerosols through evaporation of cloud particles is not treated by default in the CRM; that is, the aerosols are assumed to be scavenged when they are in droplets or ice crystals. However, in Engström et al. (2008), recycling of aerosols was found to be an important process for representing the aerosol size distribution in the air surrounding a deep convective cloud. The free tropospheric mean and median accumulation mode aerosol concentration increased by 90% and 140%, respectively, when recycling of aerosols was considered. Thus, we also examine the aerosol-induced sensitivity of deep convective strength to this process in a separate sensitivity series.
b. Simulated case and sensitivity simulation series
The simulated case is a cumulonimbus cloud with extended anvil over the Indian Ocean observed during the Indian Ocean Experiment (INDOEX) campaign (Ramanathan et al. 2001) and described further in Engström et al. (2008). Following the definition by Fan et al. (2009), this is a case with relatively weak vertical wind shear and high relative humidity. Initial aerosol concentrations are based on observations and shown in Fig. 1. Sensitivity simulations are conducted for high (as in Fig. 1), medium (50% of the aerosol concentration shown in Fig. 1), and low (25% of the aerosol concentration shown in Fig. 1) pollution levels. The aerosol composition is assumed to be 100% (NH4)2SO4. All gas concentrations are initialized as in Engström et al. (2008). The total integration time is 5 h.
Vertical profiles of initial aerosol number concentrations (high aerosol concentration case).
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Vertical profiles of initial aerosol number concentrations (high aerosol concentration case).
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Vertical profiles of initial aerosol number concentrations (high aerosol concentration case).
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
To examine how the aerosol-induced sensitivity of deep convection may depend on the complexity of the aerosol model as well as the aerosol activation parameterization, four versions of the CRM are used:
(i) Emp-Const: aerosol concentration is given as number of CCN. All aerosols in the accumulation and Aitken mode are assumed available as CCN. The aerosols are not advected or scavenged (i.e., aerosol concentrations are constant during the simulation). All newly formed cloud droplets are assumed to have the same size (radius = 1 μm). The number of activated CCN NCCN is calculated using the empirical formula 
(ii) Emp-Adv: as in simulation Emp-Const, but the aerosols are advected and scavenged by precipitation as in Wang (2005a).
(iii) Koehler-Aero: fully interactive aerosol–cloud model as described in section 2a and in Ekman et al. (2007).
(iv) Koehler-Aero-Eva: as in Koehler-Aero, but the recycling of aerosols through evaporation/sublimation of hydrometeors is considered as in Engström et al. (2008).
We also conduct two additional sets of sensitivity simulations to investigate if the size of the aerosols may affect the aerosol-induced sensitivity:
(v) Koehler-Aero-Acc: same setup as in Koehler-Aero, but all changes in aerosol concentration are assumed to take place within the accumulation mode.
(vi) Koehler-Aero-Ait: same setup as in Koehler-Aero, but all changes in aerosol concentration are assumed to take place within the Aitken mode.
Note that the total change in aerosol number concentration is the same in Koehler-Aero-Ait and Koehler-Aero-Ait; it is only the size that is different.
3. Results
All variables are evaluated after 3 h of simulation and the average over the whole time period and whole model domain is used. During its maximum size of extent, the liquid part of the cloud covers less than 100 × 100 km2 (i.e., about 25% of the model domain). All changes discussed are significant at a 95% confidence level (using a Student’s t test) if nothing else is stated. If any significant discrepancy/change can be seen for other evaluation time periods than 3 h, this will be discussed. Figures 2a and 2b illustrate that the cloud development is slightly different in the different sensitivity series, but that the main part of the convective event is encompassed when 3 h is used as evaluation time.
Horizontally averaged vertical wind speed (≥1 m s−1) during the 5-h simulation using high aerosol concentrations.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Horizontally averaged vertical wind speed (≥1 m s−1) during the 5-h simulation using high aerosol concentrations.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Horizontally averaged vertical wind speed (≥1 m s−1) during the 5-h simulation using high aerosol concentrations.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
a. Cloud droplet number concentrations
The activation of aerosols to cloud droplets begins at approximately 800-m altitude in the model (cf., e.g., Figs. 4a and 4b), which is also an altitude where there is a large change in aerosol concentration for the different sensitivity simulations (cf. Fig. 1). Comparing the different sets of sensitivity simulations, the one with constant aerosol concentration (Emp-Const) clearly activates the largest number of aerosols (Figs. 3 and 4a,b). This result is not surprising because no aerosols are removed by precipitation in this simulation series. The activation of more aerosols in Emp-Const leads to higher liquid water content compared to the other sensitivity series (cf. section 3c), in particular at levels above 4–5 km, which in turn is reflected in a higher average updraft velocity at higher altitudes as displayed in Fig. 2a. The set of simulations including an explicit aerosol cycle (Koehler-Aero) activates a larger number of aerosols than the Emp-Adv simulations. The empirical formulation requires 1% supersaturation to activate all available CCN (cf. section 2a). Using the Koehler equation for calculating the effective radius at which aerosols are activated results in all accumulation mode aerosols and a large part of the Aitken mode aerosols (due to their (NH4)2SO4 soluble composition) being activated at a lower supersaturation (∼0.7%) in Koehler-Aero.
Domain- and time-averaged cloud droplet number concentration after 3 h of simulation.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Domain- and time-averaged cloud droplet number concentration after 3 h of simulation.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Domain- and time-averaged cloud droplet number concentration after 3 h of simulation.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Horizontally averaged cloud droplet number concentration (cm−3) during 3 h of simulation using high aerosol concentrations.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Horizontally averaged cloud droplet number concentration (cm−3) during 3 h of simulation using high aerosol concentrations.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Horizontally averaged cloud droplet number concentration (cm−3) during 3 h of simulation using high aerosol concentrations.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Increasing the number of aerosols from low to medium and high concentrations results in an average increase of cloud droplets in all simulations (Table 1). Interestingly, when adding aerosols into the Aitken mode (Koehler-Aero-Ait), a higher number of activated aerosols is generated after 3 h of simulation compared with when the additional aerosols are assumed to be of accumulation mode size (Koehler-Aero-Acc; cf. Fig. 4b). This is also true when comparing Koehler-Aero-Ait with the series where recycling of aerosols from hydrometeor evaporation/sublimation is considered (Koehler-Aero-Eva). For a given updraft velocity, a larger aerosol will consume more water vapor as it grows, both before and after activation. Hence, the supersaturation right below and above cloud base is lower in Koehler-Aero-Acc than in Koehler-Aero-Ait when using medium and high aerosol concentrations. In addition, the aerosol module does not include scavenging of Aitken mode aerosols by graupel whereas accumulation mode aerosols are efficiently scavenged (cf. section 2a). As a result, the CDNC concentration increases at all levels where liquid water is present in the Koehler-Aero-Ait simulations whereas the increase of CDNC in Koehler-Aero-Acc (as well as the other sensitivity simulations) is mainly restricted to the lowest 2 km of the model.
Percent change in cloud droplet number concentration (compared with simulation with low aerosol concentration) for the different sensitivity simulations.
The larger increase of aerosols activated in Koehler-Aero-Ait in combination with the higher supersaturation generates more latent heat release which results in higher updrafts and even more activated aerosols. Thereby more cloud water can reach the freezing level in the Koehler-Aero-Ait simulations using medium and high aerosol concentrations, which in turn generates an additional increase of latent heat release and higher updrafts (cf. section 3b).
b. Updraft
The average updraft velocity (≥1 m s−1) generally increases with increasing aerosol concentration in all sensitivity simulations (Fig. 5; Table 2). This is in agreement with the theory presented by both Rosenfeld et al. (2008) and Fan et al. (2009) (for a case with relatively low vertical wind shear and high humidity). However, there are two exceptions that do not follow the conceptual model presented by Rosenfeld et al. (2008).
Domain- and time-averaged vertical wind speed (≥1 m s−1) after 3 h of simulation.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Domain- and time-averaged vertical wind speed (≥1 m s−1) after 3 h of simulation.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Domain- and time-averaged vertical wind speed (≥1 m s−1) after 3 h of simulation.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Percent change in updraft (compared with simulation with low aerosol concentration) for the different sensitivity simulations.
First, for the Koehler-Aero-Acc series, the increase in vertical wind speed between low and medium aerosol concentration is not significant. In addition, the number of grid points with updraft velocities (≥1 m s−1) is lower when using medium compared to low aerosol concentrations. After 2 h of integration there is a small but significant increase in updraft velocity for the Koelher-Aero-Acc simulations using medium and high aerosol concentrations, but after this time period the efficient depletion of accumulation mode aerosols, mainly due to impaction scavenging by graupel, inhibits the formation of new cloud droplets and thereby the release of latent heat. The results are similar restricting the analysis to updraft velocities larger than 5 m s−1. Examining updraft velocities larger than 0.1 m s−1, there is a significant decrease in updraft velocity for Koehler-Aero-Acc when using medium compared to low concentrations of aerosols.
Second, for the simulation Koehler-Aero-Eva using high aerosol concentrations, the average updraft is actually slightly lower than for Koehler-Aero-Eva using medium aerosol concentrations (the difference is still significant at a 95% confidence level). The number of grid points with updraft velocities (≥1 m s−1) is also lower. The difference in updraft velocity for the Koehler-Aero-Eva simulations using medium and high aerosol concentrations is most pronounced at the beginning of the simulation. Both simulations display higher updraft velocities than the simulation with low aerosol concentration, as more latent heat is released because of more condensation of cloud water. However, the simulation with high aerosol concentrations has more cloud water present than the simulation with medium aerosol concentration (due to the higher CDNC), and this reduces the buoyancy.
Figure 6 shows that for medium aerosol concentrations, Koehler-Aero, Koehler-Aero-Acc, and Koehler-Aero-Ait all display an increase in latent heat release between 1 and 6 km altitude after 20 min of simulation. However, after 40 min of simulation, the evaporative cooling is larger than the increase in latent heat release in the Koehler-Aero-Acc simulation, which suppresses the convection.
Domain-averaged difference in temperature increase/decrease after 20-, 40-, and 60-min integration times for the simulations using medium (dashed lines) and high (solid lines) aerosol concentration (compared with the case with low aerosol concentrations). Koehler-Aero (light gray line), Koehler-Aero-Acc (dark gray line), and Koehler-Aero-Ait (black line). Note the different scales on the x axis.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Domain-averaged difference in temperature increase/decrease after 20-, 40-, and 60-min integration times for the simulations using medium (dashed lines) and high (solid lines) aerosol concentration (compared with the case with low aerosol concentrations). Koehler-Aero (light gray line), Koehler-Aero-Acc (dark gray line), and Koehler-Aero-Ait (black line). Note the different scales on the x axis.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Domain-averaged difference in temperature increase/decrease after 20-, 40-, and 60-min integration times for the simulations using medium (dashed lines) and high (solid lines) aerosol concentration (compared with the case with low aerosol concentrations). Koehler-Aero (light gray line), Koehler-Aero-Acc (dark gray line), and Koehler-Aero-Ait (black line). Note the different scales on the x axis.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Interestingly, the increase in average updraft between the three types of activation formulation varies between 5% and 7% for medium aerosol concentrations and between 11% and 14% for high aerosol concentrations. These differences are all significant, except for the difference between the Emp-Adv (medium aerosol concentration) and Koehler-Aero (medium aerosol concentration) simulations. The simulation where recycling of aerosols is considered (Koehler-Aero-Eva) generates a significantly higher (∼5%) average updraft velocity for the case with low aerosol concentrations compared to the other Koehler-Aero simulations. On the other hand, the percentage increase in updraft velocity with increasing aerosol concentration is lower in Koehler-Aero-Eva than in Koehler-Aero. After approximately 1 h of simulation, evaporation is larger in the Koehler-Aero-Eva series than in, for example, Koehler-Aero, which stabilizes the stratification below 5 km and limits the increase in updraft velocity (Fig. 7). In addition, the increased number of cloud droplets in the beginning of the Koehler-Aero-Eva simulation (cf. Fig. 4b) may also reduce the buoyancy in the air parcel, which further limits the increase in updraft velocity.
As in Fig. 6, but for Koehler-Aero-Eva (black line) and Koehler-Aero-Acc-Noimpact (gray line). Note the different scales on the x axis.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
As in Fig. 6, but for Koehler-Aero-Eva (black line) and Koehler-Aero-Acc-Noimpact (gray line). Note the different scales on the x axis.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
As in Fig. 6, but for Koehler-Aero-Eva (black line) and Koehler-Aero-Acc-Noimpact (gray line). Note the different scales on the x axis.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
To more carefully examine the impact of graupel scavenging on the accumulation mode aerosols, we conducted an additional set of sensitivity simulations omitting all impaction scavenging of accumulation mode aerosols (simulation series Koehler-Aero-Acc-Noimpact. For this set of simulations, the updraft does increase monotonically with increasing aerosol concentration (4.9% and 6.3% for medium and high aerosol load, respectively; cf. Table 2). It is worthwhile noticing that the increase is still significantly different (and lower) compared to the Koehler-Aero-Ait series. Figure 7 shows that as graupel scavenging of accumulation mode aerosols is removed, enough aerosols are present for new cloud droplets to form (and freeze), and thereby the latent heat release is larger than the evaporative cooling.
c. Cloud water
Cloud water content increases with increasing aerosol concentration for all sets of simulations, except in Koehler-Aero-Eva where a slight decrease in cloud water content is present after 3 h of simulation when comparing high and medium aerosol concentrations (Fig. 8; Table 3). However, after only 1 h of integration, there is a monotonic increase of cloud water content with increasing aerosol concentration for this simulation series as well. For Koehler-Aero-Acc, the increase in cloud water when comparing medium and low aerosol concentrations is not significant. If impaction scavenging of accumulation mode aerosols is removed (series Koehler-Aero-Acc-Noimpact), the formation of graupel increases significantly with increasing aerosol concentrations.
Domain- and time-integrated sum of cloud water after 3 h of simulation.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Domain- and time-integrated sum of cloud water after 3 h of simulation.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Domain- and time-integrated sum of cloud water after 3 h of simulation.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Percent change in cloud water (compared with simulation with low aerosol concentration) for the different sensitivity simulations.
In general, the increase is substantially larger (43%–76%, comparing high and low aerosol concentrations) for the simulations including a full aerosol cycle (Koehler-Aero, Koehler-Aero-Eva, Koehler-Aero-Acc, and Koehler-Aero-Ait) than for the simulations using a simplified description of the aerosol population (13%–23% increase). The simulations Emp-Const and Emp-Adv contain more cloud water in the simulation with low aerosol concentration than Koehler-Aero. This can be explained by the fact that for the empirical simulations, all cloud droplets have the same equilibrium size (cf. section 2a). Koehler-Aero-Ait displays more cloud water in the reference (low aerosol concentration) simulation than Koehler-Aero and Koehler-Aero-Acc, which is reasonable considering the larger number of aerosols activated (cf. Fig. 3).
An increase in the number of cloud droplets and in the overall cloud water content should imply that more supercooled water is formed in the cloud and that the supercooled droplets are smaller. Figures 9b and 9d show that indeed the droplets are smaller with increasing aerosol concentration until approximately 40–80 min of simulation. The smallest supercooled droplet radii are found in the Emp-Const simulation because of the constant aerosol concentrations (and thus high CDNC). Emp-Const is also the only sensitivity series that shows a continuous decrease in droplet size with increasing aerosol concentrations throughout the whole simulation period. Consistent with the smaller supercooled droplet radii, Figs. 9a and 9c show that in almost all simulations the supercooled water content increases with increasing aerosol concentration until approximately 60–80 min of simulation. However, for the Koehler-Aero-Eva series, the simulation with high aerosol concentrations actually displays somewhat lower supercooled water content than the simulation with medium aerosol concentration, which is consistent with the results for the updraft. Similarly, the supercooled liquid water content for the Koehler-Aero-Acc simulations with low and medium aerosol concentrations are almost indistinguishable.
Time evolution of (a),(c) domain-integrated sum of supercooled water and (b),(d) domain-averaged supercooled droplet radius: low (dashed–dotted lines), medium (dashed lines), and high (solid lines) aerosol concentrations. (a),(b) Emp-const (black lines), Emp-Adv (dark gray lines), and Koehler-Aero (light gray lines). (c),(d) Koehler-Aero-Eva (black lines), Koehler-Aero-Acc (dark gray lines), and Koehler-Aero-Ait (light gray lines). Note that the Koehler-Aero-Acc and Koehler-Aero-Ait simulations are the same for low aerosol concentrations.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Time evolution of (a),(c) domain-integrated sum of supercooled water and (b),(d) domain-averaged supercooled droplet radius: low (dashed–dotted lines), medium (dashed lines), and high (solid lines) aerosol concentrations. (a),(b) Emp-const (black lines), Emp-Adv (dark gray lines), and Koehler-Aero (light gray lines). (c),(d) Koehler-Aero-Eva (black lines), Koehler-Aero-Acc (dark gray lines), and Koehler-Aero-Ait (light gray lines). Note that the Koehler-Aero-Acc and Koehler-Aero-Ait simulations are the same for low aerosol concentrations.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Time evolution of (a),(c) domain-integrated sum of supercooled water and (b),(d) domain-averaged supercooled droplet radius: low (dashed–dotted lines), medium (dashed lines), and high (solid lines) aerosol concentrations. (a),(b) Emp-const (black lines), Emp-Adv (dark gray lines), and Koehler-Aero (light gray lines). (c),(d) Koehler-Aero-Eva (black lines), Koehler-Aero-Acc (dark gray lines), and Koehler-Aero-Ait (light gray lines). Note that the Koehler-Aero-Acc and Koehler-Aero-Ait simulations are the same for low aerosol concentrations.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
d. Graupel
All simulation series, except for Koehler-Aero-Acc and Koehler-Aero-Eva, display an increase of graupel with increasing aerosol concentrations (Figs. 10a,b and 11; Table 4). The increase is small but significant when using high compared to medium aerosol concentrations for Emp-Const and Koehler-Aero. For the Koehler-Aero-Eva series, a slight decrease can be noted when comparing high versus medium aerosol concentrations, which is consistent with the slight decrease in updraft velocity and cloud water formation.
Domain-averaged difference in graupel after 1-, 2-, and 3-h integration times for the simulations using medium (dashed lines) and high (solid lines) aerosol concentration (compared with the case with low aerosol concentrations). (a) Emp-Const (light gray line), Emp-Adv (black line), and Koehler-Aero (dark gray line). (b) Koehler-Aero-Eva (light gray line), Koehler-Aero-Acc (black line), and Koehler-Aero-Ait (dark gray line).
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Domain-averaged difference in graupel after 1-, 2-, and 3-h integration times for the simulations using medium (dashed lines) and high (solid lines) aerosol concentration (compared with the case with low aerosol concentrations). (a) Emp-Const (light gray line), Emp-Adv (black line), and Koehler-Aero (dark gray line). (b) Koehler-Aero-Eva (light gray line), Koehler-Aero-Acc (black line), and Koehler-Aero-Ait (dark gray line).
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Domain-averaged difference in graupel after 1-, 2-, and 3-h integration times for the simulations using medium (dashed lines) and high (solid lines) aerosol concentration (compared with the case with low aerosol concentrations). (a) Emp-Const (light gray line), Emp-Adv (black line), and Koehler-Aero (dark gray line). (b) Koehler-Aero-Eva (light gray line), Koehler-Aero-Acc (black line), and Koehler-Aero-Ait (dark gray line).
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Domain- and time-integrated sum of graupel after 3 h of simulation.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Domain- and time-integrated sum of graupel after 3 h of simulation.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Domain- and time-integrated sum of graupel after 3 h of simulation.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Percent change in graupel (compared with simulation with low aerosol concentration) for the different sensitivity simulations.
For Koehler-Aero-Acc there is a significant decrease in graupel amount when using medium aerosol concentrations. Comparing Koehler-Aero-Acc and Koehler-Aero, the development of graupel is relatively similar up to approximately 1–1.5 h of integration. After this time period, the convection in Koehler-Aero-Acc is suppressed (cf. section 3b) and thereby also the graupel formation. If impaction scavenging of accumulation mode aerosols is removed, then the graupel content increases with increasing aerosol concentration (Table 4).
e. Rain
For Emp-Const and Emp-Adv, the amount of rain at the lowest model level decreases with increasing aerosol concentration, despite the increase in vertical wind speed and graupel formation (Fig. 12; Table 5). The decrease in rain rate occurs because of more efficient evaporation of raindrops below the freezing level and, to some extent, the less efficient warm rain formation. Above 5 km, the rain drop mass actually increases with increasing aerosol concentration (not shown).
Domain- and time-averaged sum of rainwater at model level 1 after 3 h of simulation.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Domain- and time-averaged sum of rainwater at model level 1 after 3 h of simulation.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Domain- and time-averaged sum of rainwater at model level 1 after 3 h of simulation.
Citation: Journal of the Atmospheric Sciences 68, 4; 10.1175/2010JAS3651.1
Percent change in rainwater (compared with simulation with low aerosol concentration) at model level 1 for the different sensitivity simulations.
The rain formation at the lowest model level increases with increasing aerosol concentration in Koehler-Aero and Koehler-Aero-Ait, but the increase peaks at medium aerosol concentrations. The difference in rain formation between medium and high aerosol concentration is not significant for Koehler-Aero-Ait and only significant at a 90% confidence level for Koehler-Aero. For high aerosol concentrations, the rain formation above 5 km still increases in both these simulations, but the evaporation of raindrops below the freezing level increases. A higher aerosol load is required for a significant increase in rain formation to occur in Koehler-Aero-Acc compared to Koehler-Aero and Koehler-Aero-Ait. For medium aerosol concentrations, the amount of rain at the lowest model level actually decreases compared to the simulation with low aerosol concentrations in Koehler-Aero-Acc, which is consistent with the lower rate of graupel formation. If impaction scavenging of accumulation mode aerosols is removed, then the rain rate will increase with increasing aerosol concentrations (by 9% and 25% for medium and high aerosol concentrations, respectively).
4. Summary and conclusions
A cloud-resolving model including explicit aerosol physics and chemistry has been utilized to examine the influence of the complexity of the aerosol model, the size of the aerosols, and the aerosol activation parameterization on simulated aerosol-induced deep convective cloud sensitivity. The study considers an isolated cloud case with relatively weak vertical wind shear and high relative humidity. To simplify the analysis, only the aerosol effect on liquid droplet formation has been considered (i.e., the number of aerosols available as ice nuclei has been kept constant). In total, six sets of sensitivity simulations using low, medium, and high aerosol concentrations have been conducted. The applied range of aerosol concentrations corresponds to a region where a monotonic increase of convective strength with increasing aerosol concentration should be seen (Rosenfeld et al. 2008).
The sensitivity simulations show that the complexity of the aerosol model and the choice of aerosol activation parameterization significantly affect the aerosol-induced convective cloud sensitivity. When using different aerosol activation parameterizations, the increase in vertical wind speed (comparing high and low aerosol concentrations) varies between 11% and 14% if vertical wind velocities of at least 1 m s−1 are considered and between 10% and 19% if vertical wind velocities of at least 5 m s−1 are considered. Regarding the complexity of the aerosol model, graupel scavenging of aerosols appears to be a crucial process since it efficiently removes aerosols at a critical time point of the cloud evolution. However, this process is currently poorly understood and often described in a simplified manner in models. Recycling of aerosols through cloud droplet and ice crystal evaporation/sublimation is also found to be an important process for the aerosol-induced convective cloud sensitivity. For example, when aerosol recycling is considered, the convective strength reaches a maximum at medium instead of high aerosol concentrations and the sensitivity of the deep convective strength is significantly smaller.
An increased number of aerosols in the Aitken mode (here defined by 23 ≤ d ≤ 100.0 nm) may result in a larger impact on the convective strength compared to an increased number of aerosols in the accumulation mode (here defined by 100 ≤ d ≤ 900.0 nm). As accumulation mode aerosols are activated and grow in the beginning of the cloud cycle, the supersaturation near the cloud base is lowered. Thereby, more aerosols in the Aitken mode are activated compared to in the accumulation mode during the active phase of the convective cloud development, which in turn results in a larger increase of the convective updraft.
For the analysis of the model simulations, the average updraft is taken as a measure of the deep convective strength. We find that, in general, the model study supports the conceptual model that increased aerosol concentrations result in stronger convection as outlined by Rosenfeld et al. (2008). However, there are two main exceptions where the updraft does not increase monotonically with increasing aerosol concentrations. Both these exceptions are in line with the study of Fan et al. (2009) and illustrate the need to more carefully examine the effect of changes in evaporation processes when considering aerosol effects on convective strength and precipitation. In addition, one sensitivity series shows that graupel scavenging may efficiently deplete the number of aerosols available as CCN, which for the present case limits the formation of cloud droplets and thereby the release of latent heat. It is also worthwhile noting that a decrease in buoyancy, resulting from an increase in aerosol concentration and thereby cloud water, may take place already during the initial stages of cloud development before any freezing is initiated.
The sensitivity simulations show that the change in graupel and rain formation may not always be directly proportional to the change in updraft velocity. Several of the sensitivity simulations display a decrease of the rain amount at the lowest model level with increasing updraft velocity. This decrease was generally caused by an increased evaporation of raindrops. These results are consistent with previous studies by, for example, Carrió et al. (2010). As the response of the convection (both in terms of strength and in terms of precipitation) was found to be significantly different when using different complexity of the aerosol model and aerosol activation parameterization, it underlines the importance of better understanding the two-way interaction between aerosols and clouds. The results may also to some extent explain why different cloud-resolving model studies have shown different results regarding aerosol-induced deep convective cloud sensitivity. It should be noted that the present study has been performed using a two-moment bulk microphysics scheme, which inherently includes certain assumptions regarding the hydrometeor size distribution. We therefore recommend complementary future studies using bin microphysical models.
The relatively small differences in convective strength obtained for all sensitivity simulations comparing medium and low aerosol concentrations suggest that single deep convective clouds developing in an environment with weak vertical wind shear and high relative humidity are relatively insensitive to changes in aerosol concentration. This finding corroborates the results by Fan et al. (2009) and suggests that aerosol effects on isolated deep convective clouds are more likely to be distinguished in humid environments with large vertical wind shear.
Acknowledgments
The work of A. Engström was funded by the Swedish National Space Board.
REFERENCES
Albrecht, B. A., 1989: Aerosols, cloud microphysics, and fractional cloudiness. Science, 245, 1227–1230.
Barth, M. C., and Coauthors, 2007: Cloud-scale model intercomparison of chemical constituent transport in deep convection. Atmos. Chem. Phys., 7, 4709–4731.
Berry, E. X., 1967: Cloud droplet growth by collection. J. Atmos. Sci., 24, 688–701.
Carrió, G. G., W. R. Cotton, and W. Y. Y. Cheng, 2010: Urban growth and aerosol effects on convection over Houston. Part I: The August 2000 case. Atmos. Res., 96, 560–574, doi:10.1016/j.atmosres.2010.01.005.
Cotton, W. R., G. J. Tripoli, R. M. Rauber, and E. A. Mulvihill, 1986: Numerical simulation of the effects of varying ice crystal nucleation rates and aggregation processes on orographic snowfall. J. Climate Appl. Meteor., 25, 1658–1680.
Ekman, A. M. L., C. Wang, J. Wilson, and J. Ström, 2004: Explicit simulations of aerosol physics in a cloud-resolving model: A sensitivity study based on an observed convective cloud. Atmos. Chem. Phys., 4, 773–791.
Ekman, A. M. L., C. Wang, J. Ström, and R. Krejci, 2006: Explicit simulation of aerosol physics in a cloud-resolving model: Aerosol transport and processing in the free troposphere. J. Atmos. Sci., 63, 682–696.
Ekman, A. M. L., A. Engström, and C. Wang, 2007: The effect of aerosol composition and concentration on the development and anvil properties of a continental deep convective cloud. Quart. J. Roy. Meteor. Soc., 133, 1439–1452, doi:10.1002/qj.108.
Ekman, A. M. L., R. Krejci, A. Engström, J. Ström, M. D. Reus, J. Williams, and M. O. Andreae, 2008: Do organics contribute to small particle formation in the Amazonian upper troposphere? Geophys. Res. Lett., 35, L17810, doi:10.1029/2008GL034970.
Engström, A., A. M. L. Ekman, R. Krejci, J. Ström, M. de Reus, and C. Wang, 2008: Observational and modelling evidence of ;tropical deep convective clouds as a source of mid-tropospheric accumulation mode aerosols. Geophys. Res. Lett., 35, L23813, doi:10.1029/2008GL035817.
Fan, J., and Coauthors, 2009: Dominant role by vertical wind shear in regulating aerosol effects on deep convective clouds. J. Geophys. Res., 114, D22206, doi:10.1029/2009JD012352.
Fu, Q., and K. N. Liou, 1993: Parameterization of the radiative properties of cirrus clouds. J. Atmos. Sci., 50, 2008–2025.
Fuchs, N. A., 1964: The Mechanics of Aerosols. Pergamon Press, 408 pp.
Kessler, E., 1969: On the Distribution and Continuity of Water Substance in Atmospheric Circulation. Meteor. Monogr., No. 32, 84 pp.
Liu, Y., P. H. Daum, R. L. McGraw, M. A. Miller, and S. Niu, 2007: Theoretical expression for the autoconversion rate of the cloud droplet number concentration. Geophys. Res. Lett., 34, L16821, doi:10.1029/2007GL030389.
McFarquhar, G. M., and A. J. Heymsfield, 1996: Microphysical characteristics of three anvils sampled during the Central Equatorial Pacific Experiment. J. Atmos. Sci., 53, 2401–2423.
Pruppacher, H. R., and J. D. Klett, 1997: Microphysics of Clouds and Precipitation. 2nd ed. Kluwer Academic, 954 pp.
Ramanathan, V., and Coauthors, 2001: Indian Ocean Experiment: An integrated analysis of the climate forcing and effects of the great Indo-Asian haze. J. Geophys. Res., 106 (D22), 28 371–28 398
Rosenfeld, D., U. Lohmann, G. B. Raga, C. D. O’Dowd, M. Kulmala, S. Fuzzi, A. Reissell, and M. O. Andreae, 2008: Flood or drought: How do aerosols affect precipitation? Science, 321, 1309–1313, doi:10.1126/science.1160606.
Seifert, A., and K. D. Beheng, 2006: A two-moment cloud microphysics parameterization for mixed-phase clouds. Part 2: Maritime vs. continental deep convective storms. Meteor. Atmos. Phys., 92, 67–82, doi:10.1007/s00703-005-0113-3.
Seifert, A., A. Khain, A. Pokrovsky, and K. Beheng, 2006: A comparison of spectral bin and two-moment bulk mixed-phase cloud microphysics. Atmos. Res., 80, 46–66, doi:10.1016/j.atmosres.2005.06.009.
Simpson, J., and V. Wiggert, 1969: Models of precipitating cumulus towers. Mon. Wea. Rev., 97, 471–489.
Teller, A., and Z. Levin, 2006: The effects of aerosols on precipitation and dimensions of subtropical clouds: A sensitivity study using a numerical cloud model. Atmos. Chem. Phys., 6, 67–80.
Twomey, S., 1977: The influence of pollution on the shortwave albedo of clouds. J. Atmos. Sci., 34, 1149–1152.
van den Heever, S. C., and W. R. Cotton, 2007: Urban aerosol impacts on downwind convective storms. J. Appl. Meteor. Climatol., 46, 828–850.
van den Heever, S. C., G. G. Carrió, W. R. Cotton, P. J. DeMott, and A. J. Prenni, 2006: Impacts of nucleating aerosol on Florida storms. Part I: Mesoscale simulations. J. Atmos. Sci., 63, 1752–1775.
Vehkamäki, H., M. Kulmala, I. Napari, K. E. J. Lehtinen, C. Timmreck, M. Noppel, and A. Laaksonen, 2002: An improved parameterization for sulfuric acid-water nucleation rates for tropospheric and stratospheric conditions. J. Geophys. Res., 107, 4622, doi:10.1029/2002JD002184.
Wang, C., 2005a: A modeling study of the response of tropical deep convection to the increase of cloud condensation nuclei concentration: 1. Dynamics and microphysics. J. Geophys. Res., 110, D21211, doi:10.1029/2004JD005720.
Wang, C., 2005b: A modeling study of the response of tropical deep convection to the increase of cloud condensation nuclei concentration: 2. Radiation and tropospheric chemistry. J. Geophys. Res., 110, D22204, doi:10.1029/2005JD005829.
Wang, C., and J. S. Chang, 1993a: A three-dimensional numerical model of cloud dynamics, microphysics, and chemistry. 1. Concepts and formulation. J. Geophys. Res., 98 (D8), 14 827–14 844.
Wang, C., and J. S. Chang, 1993b: A three-dimensional numerical model of cloud dynamics, microphysics, and chemistry. 2. A case study of the dynamics and microphysics of a severe local storm. J. Geophys. Res., 98 (D8), 14 845–14 862.
Wang, C., and P. J. Crutzen, 1995: Impact of a simulated severe local storm on the redistribution of sulfur dioxide. J. Geophys. Res., 100 (D6), 11 357–11 367.
Wang, C., and R. G. Prinn, 1998: Impact of the horizontal wind profile on the convective transport of chemical species. J. Geophys. Res., 103 (D17), 22 063–22 071.
Wang, C., and R. G. Prinn, 2000: On the roles of deep convective clouds in tropospheric chemistry. J. Geophys. Res., 105 (D17), 22 269–22 297.
Wang, C., P. J. Crutzen, V. Ramanathan, and S. F. Williams, 1995: The role of a deep convective storm over the tropical Pacific Ocean in the redistribution of atmospheric chemical species. J. Geophys. Res., 100 (D6), 11 509–11 516.
Wilson, J., C. Cuvelier, and F. Raes, 2001: A modeling study of global mixed aerosol fields. J. Geophys. Res., 106 (D24), 34 081–34 108.
