modeling the impact of aerosols on tropical overshooting...

Modeling the impact of aerosols on tropical overshootingthunderstorms and stratospheric water vapor

Baojun Chen1 and Yan Yin2

Received 4 January 2011; revised 8 July 2011; accepted 15 July 2011; published 8 October 2011.

[1] Overshooting deep convection plays an important role in regulating the water vaporcontent of the tropical tropopause layer and is also an important mechanism for transportingwater vapor into the lower stratosphere (LS). The aim of this study is to examine the effect ofaerosols as cloud condensation nuclei (CCN) on the water vapor content of the LS via singleisolated overshooting thunderstorms. The development of a severe Hector thunderstormin northern Australia observed during the Stratospheric‐Climate Links with Emphasis on theUpper Troposphere and Lower Stratosphere/Aerosol and Chemical Transport in TropicalConvection (SCOUT‐O3/ACTIVE) campaign is simulated using a three‐dimensionalnonhydrostatic convective cloud model with a double‐moment bulk microphysics scheme.The results show that the ice hydrometeors account for over 50% of the total condensatemass, indicating that ice processes play an important role in regulating the structure ofthunderstorms in the tropics. A large number of ice particles occurring in the LS are notformed in situ but are transported upward by convective overshooting with subsequentmixing. Sensitivity tests show that the increase in cloud droplet numbers induced byincreasing CCN concentrations would increase the number concentrations of the ice crystalstransported to the LS, which had the effect of reducing the sizes and fall speeds of theice crystal, thereby causing a moistening of the LS by sublimation of the injected iceparticles. This result suggests that aerosols in the boundary layer can affect stratosphericwater vapor via overshooting deep convection.

Citation: Chen, B., and Y. Yin (2011), Modeling the impact of aerosols on tropical overshooting thunderstormsand stratospheric water vapor, J. Geophys. Res., 116, D19203, doi:10.1029/2011JD015591.

1. Introduction

[2] Changes in water vapor in the upper troposphere andlower stratosphere (UT/LS) play an important role in modu-lating the Earth’s climate since it is the most powerfulgreenhouse gas in the atmosphere. An increase of about 1% inwater vapor per year in the stratosphere was observed over thepast 50 years [Oltmans et al., 2000; Rosenlof et al., 2001].Determining what affects the stratospheric water vapor con-tent is important for better understanding the greenhouseeffect and global climate change.[3] Water vapor in the stratosphere has two main sources.

One is transport of water vapor from the troposphere via deepconvection to the tropical tropopause layer (TTL), from thereto the stratosphere by the Brewer‐Dobson circulation. Theother is the oxidation of methane, which occurs mostly in theupper stratosphere. The tropics are the main source regions ofwater vapor in the Earth’s atmosphere. Deep convection isoccurring very frequently in these regions, and some of the

deepest convection can penetrate up to the cold point tropo-pause (CPT) and have significant overshoots. In recent years,the contribution of overshooting deep convection to the watervapor content of the TTL and lower stratosphere (LS) hasattracted considerable attentions [Romps and Kuang, 2009;Toon et al., 2010]. Previous studies suggest that overshootingdehydrates the stratosphere caused by the irreversible mixingof colder, drier air into the ambient environment [Danielsen,1982, 1993; Sherwood and Dessler, 2000, 2001]. Morerecently, however, measurements [Nielsen et al., 2007; Cortiet al., 2008; de Reus et al., 2009; Khaykin et al., 2009;Iwasaki et al., 2010] and model simulations [Chaboureauet al., 2007; Chemel et al., 2009; Grosvenor et al., 2007;Liu et al., 2010; Wang, 2003] suggest that overshootingconvection has rather a hydrating than dehydrating effect bysublimation of the injected ice particles above the tropopause.Furthermore, Liu et al. [2010] and Grosvenor [2010] havemade quantitative estimates of the impact of overshootingconvection on the stratospheric water vapor content. Theirresults showed that the mass of water vapor transported intothe stratosphere was associated with the severity of theovershoot. Hassim and Lane [2010] reported the importanceof TTL relative humidity in governing the stratosphericmoisture, and showed that overshooting tended to hydrate(dehydrate) the stratosphere when the TTL was alreadysubsaturated (supersaturated).

1Key Laboratory of Mesoscale Severe Weather, MOE, and School ofAtmospheric Sciences, Nanjing University, Nanjing, China.

2CMA Key Laboratory for Atmospheric Physics and Environment,Nanjing University of Information Science and Technology, Nanjing,China.

Copyright 2011 by the American Geophysical Union.0148‐0227/11/2011JD015591

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 116, D19203, doi:10.1029/2011JD015591, 2011

D19203 1 of 15

http://dx.doi.org/10.1029/2011JD015591

[4] Tropical deep convection mainly occurs over threeregions, namely South America, Central Africa, and theMaritime Continent, which includes the Indonesian archi-pelago and northern Australia [Chemel et al., 2009]. Duringthe Australian premonsoon and monsoon‐break periodsbetween November and February, severe thunderstorms,known colloquially as “Hector,” occur almost daily overnorthern Australia, particularly in the area of Darwin, as aresult of the strong convergence of warm, moist sea and landbreezes [Crook, 2001]. Owing to its unique location, theDarwin area has hosted a number of field campaigns devotedto tropical deep convection, including the Island Thunder-storm Experiment [Keenan et al., 1989], the Maritime Con-tinent Thunderstorm Experiment [Keenan et al., 2000], theStratospheric‐Climate Links with Emphasis on the UpperTroposphere and Lower Stratosphere/Aerosol and ChemicalTransport in Tropical Convection (SCOUT‐O3/ACTIVE)[Vaughan et al., 2008], and many others. SCOUT‐O3/ACTIVE was held in November–December 2005 and wasaimed at determining the impact of deep convection on thecomposition of the UTLS region and its relative importancecompared with large‐scale ascent. The Hector thunderstormswere a particular focus of the SCOUT‐O3/ACTIVE. One ofthe objectives of the SCOUT‐O3/ACTIVE was to investigatethe aerosol effects on deep convection.[5] Atmospheric aerosols, formed from natural and

anthropogenic processes [Zhang, 2010], as cloud condensa-tion nuclei (CCN), can markedly affect the development ofclouds both microphysically and dynamically [Andreae et al.,2004; Carrió et al., 2007; Li et al., 2008, 2009; Khain et al.,2005, 2008; Khain and Lynn, 2009; May et al., 2009;Sherwood, 2002; van den Heever et al., 2006; Zhang et al.,2007]. These will potentially alter the ability of convectiveclouds to transport of water and vapor into the TTL and LS.However, there have been relatively few studies of the impactof aerosols on stratospheric water vapor through its effect ontropical deep convection. Recently, Grosvenor et al. [2007]employed the UK Met Office LEM (Large Eddy Model)with bulk microphysics scheme to simulate the developmentof a storm complex over the Bauru area in Brazil under dif-ferent aerosol conditions. Their results showed that anincrease in the CCN number increased the ice water contentof the TTL and the water vapor content of the LS. It is notedthat the CCN number in Grosvenor et al.’s work actuallyrefers to the activated droplets, and is assumed constantthroughout the cloud’s lifetime.[6] Darwin in northern Australia is a natural laboratory for

the study of deep tropical convection responding to the var-iation in aerosols. There is a significant seasonal variation inaerosol characteristics in the Darwin region [Allen et al.,2008]. In the premonsoon periods, the atmosphere is rela-tively smoky owing to widespread biomass burning innorthern Australia, resulting in a high concentration ofaerosols, while in the monsoon periods, the atmosphere isback to clean maritime air owing to the cessation of thebiomass burning source. The aim of this study is to determinethe possible impact of severe thunderstorms that developedover the Darwin region under different aerosol conditions onthe water vapor content of the LS. Two problems areaddressed. One is whether aerosols affect the development ofa single isolated Hector thunderstorm. The other is whetherthese changes in the clouds induced by the aerosols affect the

stratospheric water vapor content. To address these issues, weemploy a three‐dimensional nonhydrostatic convective cloudmodel with a double‐moment bulk microphysics scheme andexplicit droplet activation from CCN (detailed description insection 2), in contrast to the hybrid bulk microphysics schemeemployed byGrosvenor et al. [2007], whichwas onemomentfor liquid and rain and two moment for ice, snow, and grau-pel. Furthermore, the current work differs in that single iso-lated severe thunderstorms over the tropics are simulated.Although there have been several numerical studies focusingon the aerosol effects on cloud structures and precipitation ofisolated thunderstorms [e.g., Cui et al., 2006; Yin and Chen,2007; Lerach et al., 2008; Khain and Lynn, 2009], few werebased on the tropics and they did not focus on the impact ofaerosols on how severe tropical thunderstorms affect strato-spheric water vapor.[7] The rest of the paper is organized as follows: Themodel

description is provided in section 2, followed by the simu-lation initialization in section 3. The dynamics and micro-physics of the simulated thunderstorms developed undertwo CCN conditions are investigated in section 4, along witha change of water vapor and ice content in the LS inducedby CCN. Finally, section 5 presents a summary and theconclusions.

2. Model Description

[8] The cloud model used in this study is a three‐dimensional compressible nonhydrostatic cloud model with adouble‐moment bulk microphysics scheme [Kong et al.,1990; Hong and Fan, 1999; Chen and Xiao, 2010]. Themodel domain is on a standard spatially staggered meshsystem. A conventional time‐splitting integration technique,proposed by Klemp and Wilhelmson [1978], is used in themodel. The large and small time steps are 2 s and 0.25 s,respectively. The spatial differential terms are of second‐order accuracy except for the advection term that has fourth‐order accuracy. All other derivatives are evaluated withsecond‐order centered differences. The radiation boundaryconditions of Klemp and Wilhelmson [1978] are used for thelateral boundaries while the top and bottom boundaries areassumed as a rigid wall. A Rayleigh friction zone is also usedto absorb vertically propagating gravity waves near the top ofthe domain. The damping depth frommodel top is set to 6 km.The model includes a conventional first‐order closure forsubgrid turbulence and a diagnostic surface boundary layerbased on Monin‐Obukhov similarity theory.[9] Using bulk microphysics parameterization, the mixing

ratios and number concentrations of seven hydrometeor typesare predicted in the model: cloud water, cloud ice, rain, snow,graupel, frozen drops and hail. The cloud droplets, ice crystalsand raindrops are assumed to follow gamma size distribu-tions, while other hydrometeors are represented by inverseexponential size distributions [Hong and Fan, 1999]. Cor-responding microphysical processes include (1) nucleation ofcloud droplets and ice crystals; (2) autoconversion of cloudwater to rain, cloud ice to snow or graupel, snow to graupel,and graupel and frozen drops to hail; (3) homogeneousfreezing of cloud water to cloud ice, rain to graupel or frozendrops, probabilistic freezing of rain to graupel or frozendrops, collisional freezing of rain with cloud ice and snow tograupel or frozen drops; (4) accretion of cloud water by rain,

CHEN AND YIN: AEROSOL IMPACT ON LS WATER VAPOR D19203D19203

2 of 15

cloud ice, snow, graupel, frozen drops, and hail; (5) accretionof rain by graupel, frozen drops, and hail; (6) rain shed fromgraupel, frozen drops, and hail; (7) melting of cloud ice,snow, graupel, frozen drops, and hail; (8) accretion of cloudice by snow, graupel, frozen drops, and hail; (9) accretion ofsnow by graupel, frozen drops, and hail; (10) self‐accretionof cloud ice and snow; (11) condensation/evaporation ofcloud water and rain; (12) deposition/sublimation of cloudice, snow, graupel, frozen drops, and hail; and (13) rime‐splintering of snow, graupel, and frozen drops. A completedescription of the microphysical parameterization is given byHong and Fan [1999] and Chen and Xiao [2010]. Only thekey processes pertinent to the present work are describedhere.[10] A power law activity spectrum of CCN is applied for

droplet nucleation. The number of activated CCN, NCCN, forsupersaturation with respect to water Sw (in percent) can bewritten as

NCCN ¼ CSkw; ð1Þ

where C and k are constant parameters. Owing to the lack ofdirect measurements of CCN concentration in the Darwinregion, the coefficients were set toC = 100 cm−3 and k = 0.462for maritime conditions, or C = 1260 cm−3 and k = 0.308 forcontinental conditions, on the basis of the work ofKhain et al.[2005]. The CCN concentrations are assumed to decreaseexponentially with altitude with a scale height of 2.5 km.Activation only occurs in regions of upwardmotion andwhenthe supersaturation resolved on the model grid is positive.Nucleation of droplets may occur at cloud base or within thecloud. The peak supersaturation at the cloud base is diag-nosed analytically [Twomey, 1959], and the potential numbersof droplets activated are given by Rogers and Yau [1989],

NCCN ¼ 0:88C2= kþ2ð Þ�0:07w3=2�k= kþ2ð Þ

; ð2Þ

where NCCN is in cm−3 and w is the vertical velocity in cms−1. Within the cloud, the supersaturation resolved on themodel grid is then applied for droplet nucleation. Calcula-tion of the supersaturation during the time step is based onthe approach of Morrison et al. [2005, equations (3)–(7)],but the present work neglects the impact of radiative transferand microphysical processes associated with hail on super-saturation. In this scheme, the evolution of the supersatu-ration due to condensation/deposition (evaporation/sublimation), the Bergeron‐Findeisen mechanism and adia-batic (vertical air motion) cooling is predicted over longertime steps (>1 s). The phase relaxation time scales associ-ated with each hydrometeor species (cloud water, rain, cloudice, snow, graupel and frozen drops) are given by Morrisonet al. [2005, equation (4)], but the spectral shape index pcused herein is specified a constant value, particularly, pc = 0for snow, graupel and frozen drops, 5 for cloud droplets, 2 forraindrops, and 1 for ice crystals [Hong and Fan, 1999]. Theaverage radius of newly nucleated droplets or initial dropletsize distribution is assumed to be a function of the supersat-uration—the higher the supersaturation, the smaller the aver-age radius, using the empirical formula and specificparameters given by Reisin et al. [1996]. This approach causes

a relatively wide spectrum of droplets for maritime cloudscompared to continental clouds, leading to drop collisionsearly in the maritime clouds.[11] Rainwater can be initiated by autoconversion of cloud

droplets to raindrops, melting of precipitating ice, or sheddingof excess water drops accreted by graupel, frozen drops andhail in the wet growth regime. The autoconversion representsa key process in cloud development, in which raindrops areinitiated by collision and coalescence of cloud droplets.Although various parameterizations of the autoconversionprocesses are available [see Li et al., 2008], it is beyond thescope of this study to examine in detail the impact of differentschemes. We also note that Li et al. [2008] found a similarvariation of cloud properties with CCN concentration pro-duced by different types of autoconversion parameteriza-tion, especially when the CCN concentration was less than5000 cm−3. A sensitivity test of different autoconversionparameterizations has been performed, by running thescheme proposed by Khairoutdinov and Kogan [2000], andthe results are very similar to the present results except forthe total mass of water and vapor in clouds. Autoconversionof cloud droplets to rain used herein is parameterized fol-lowing Lin et al. [1983] except that the constant relativedispersion has been replaced by a predicted value basedon the concentration of cloud droplets. The autoconversionterm PRAUT is expressed by

PRAUT ¼ �a lCW � lW0ð Þ2

� 1:2� 10�4 þ 1:569� 10�12 NC

D0 lCW � lW0ð Þ� ��1

; ð3Þ

where lCW is the mixing ratio of cloud water, NC (cm−3) is the

number concentration of cloud droplets, andD0 is the relativedispersion. The lW0 is the threshold for autoconversion with avalue of 2 × 10−3 g g−1. In this study, the relative dispersionD0 is calculated from the predicted NC using the relation[Grabowski, 1999]

D0 ¼ 0:146� 5:964� 10�2 ln NC=2000ð Þ: ð4Þ

Furthermore, we add the mass of a droplet of a critical size,Dw,auto, as an additional threshold for autoconversion fol-lowing Phillips et al. [2007], except that Dw,auto is set equalto 28 mm to prevent unrealistically fast conversion for mari-time clouds, in which the largest droplet is about 22 mm indiameter.[12] The natural cloud ice is produced through condensa-

tion freezing and deposition nucleation, and homogeneousfreezing of cloud droplets below −40°C. The number con-centration of active natural ice nuclei is parameterized fol-lowing Cooper [1986]. Once formed, cloud ice grows via thedepositional and riming processes. Ice multiplication, or thesecondary ice generation mechanism, for riming of snow andgraupel/frozen drops at temperatures between −3°C and −8°Cis based on the work of Hallett and Mossop [1974] and isparameterized following Hu and He [1988]. Snow can beformed by the Bergeron–Findeisen process and autoconver-sion of cloud ice. Frozen drops can be initiated by homoge-neous or probabilistic freezing of raindrops, or collisionsbetween rain and cloud ice or snow only when the raindrop


3 of 15

diameter is greater than 1 mm. If the raindrop diameter issmaller than 1 mm, a frozen raindrop is converted to graupel.Graupel may be also created via a parameterized form of theBergeron process, or by aggregation of ice crystals andsnowflakes. The autoconversion rate coefficient for cloud iceto form snow and for snow to form graupel is based on thework of Lin et al. [1983]. Graupel and frozen drops convert tohail when their diameters are greater than or equal to 5mmviaautoconversion [Hu and He, 1988].

[13] Additionally, it should also be noted that the termscontinental and maritime relate in this study to CCN distri-bution and concentration only.

3. Model Initialization

[14] The numerical simulations presented here were basedon a case study day of the 2005 SCOUT‐O3/ACTIVE fieldcampaign. The day chosen was 30 November. That day was

Figure 1. (a) Skew‐T log‐p diagram plotted from the sounding taken in Darwin at 1430 local time,30 November 2005. The dotted line represents undiluted parcel ascent from the surface to the liftingcondensation level (LCL) and then follows the moist adiabatic condition. The heavy solid and dashed linesrepresent temperature and dew point profile, respectively. (b) The corresponding u (westerly, solid) andv (southerly, dashed) components of the horizontal wind.


4 of 15

also a “golden Hector day” of the campaign [Brunner et al.,2009]. Hector that developed on that day was much tallerand stronger than in other days of the campaign. This Hectorproduced a significant overshoot of convective turretsreaching up to 18 km into the stratosphere. Very light upperlevel winds caused the anvil to flow radially out from thestorm with a slight northeastward drift. The CPT was situatedat 17.3 km altitude during that day [de Reus et al., 2009].According to de Reus et al. [2009], the ice crystals wereobserved at altitudes between 18 and 18.7 km, at potentialtemperatures level between 386 and 414 K, which corre-sponds to 0.7 to 1.4 km above the CPT. The presence of theseice particles would likely lead to a moistening of the LS. Theeffect of overshooting on hydrating the LS via sublimationof the injected ice particles from the troposphere has beenfurther confirmed through numerical simulations by Chemelet al. [2009].

[15] The model was initialized on the basis of the rawin-sonde sounding shown in Figure 1 taken from Darwin, on30 November 2005 at 1430 LT. The sounding shows largewater vapor availability with mixing ratios of 17 g kg−1 nearthe surface and a warm cloud base with a temperature of morethan 20°C at about 1.5 km, while the 0°C freezing tempera-ture is at 5.1 km. The 380 K potential temperature level issituated at ∼17.3 km, which corresponds to the 87 hPa pres-sure level (not shown in Figure 1). The atmospheric envi-ronment has a very large convective available potentialenergy (CAPE) of approximately 3500 J kg−1. The windhodograph shows a low‐level northwesterly flow with lowshear, the flow above being westerly up to about 12 km andeasterly aloft. This synoptic environment favors the trigger-ing of Hector thunderstorms over the islands, just as sug-gested, for instance, by Crook [2001].[16] All simulations were integrated to 4200 s using a grid

size of 0.5 km horizontally and 0.3 km vertically, with cor-

Figure 2. Modeled radar reflectivity from the maritime CCN simulation at (a) 32 and (b) 40 min.Observations made by Bureau of Meteorology Research Centre–National Center for Atmospheric ResearchC‐band polarimetric (BMRC/NCAR C‐POL) radar at (c) 0510 and (d) 0540 UTC.


5 of 15

responding dimensions of 41 and 24 km for the modeldomain. A domain moving method, where the grids movewith the center of the total hydrometeor mass, is used to keepthe simulated storm within the computational domain. Con-vectionwas initiated by awarm thermal bubble of 20 kmwideand 3 km deep, which was centered at 1.5 km above groundlevel in a horizontally homogeneous environment. Themaximum thermal perturbation was 2.0 K in the center of thebubble with the heating rate decreasing according to a cos2

relationship with radial distance from the center. Such heatingresulted in a relatively strong updraft and formation of theconvective clouds. However, we note that Grosvenor et al.[2007, hereinafter G07] and Grosvenor [2010, hereinafterG10] found that the moistening potential of the stratosphereexhibited a large variation with bubble temperature pertur-bations, showing an increasing stratospheric water vapor

content with increasing perturbations. In our simulations,with a smaller temperature perturbation and bubble width thesimulated convection was relatively weak and failed to pro-duce convective overshoots. On increasing the temperatureperturbation and width, the simulated convection was con-siderably stronger when compared with the radar observa-tions. In addition, similarly to G07 and G10, the stratosphericwater content was increased with increasing temperatureperturbation. On the basis of these sensitivity tests, we chooseamaximum temperature perturbation of 2K and bubble widthof 20 km to initialize the convection. The following analysisis based on these choices. Figure 2 shows a comparison of theobservational radar data with the simulated radar reflectivitywhich is calculated from the sum of the reflectivity for allhydrometeors except for cloud water [Hogan et al., 2006].The characteristic structure of observed thunderstorms hasbeen well reproduced by the simulations. The simulated radarreflectivity field generally agrees with the observations. Thehorizontal and vertical scales of convection in the simulationsare similar to the radar observations. However, the simulatedradar reflectivity values are somewhat higher than theobservations, probably owing to the over prediction ofgraupel, frozen drops and hail (Figure 3), which is likelycaused by the microphysical schemes and the initialization ofthe warm bubble. Overall, these choices of bubble width andtemperature perturbation lead to the best comparison with theradar observations.

4. Results

4.1. Simulation of Convective Clouds Under MaritimeAerosol Conditions

[17] The dynamical and microphysical characteristics ofsimulated convective clouds in the maritime aerosol condi-tions will be first examined. Figure 4 presents the time evo-lution of the maximum updraft velocities in the modeldomain. One can see that updrafts tend to increase with

Figure 3. The mass mixing ratios (g kg−1) of (a) 32 minand (b) 40 min, showing the presence of graupel/frozendrops and hail in high radar reflectivity corresponding toFigures 2a and 2b. Blue, cyan, yellow, red, and greenlines represent mixing ratio contours of cloud ice, snow,graupel/frozen drops, hail, and rain, respectively. The blackline with the value of 0.001 g kg−1 indicates the cloudboundary, which is calculated from the sum of the massdensity for all hydrometeors. Temperature contours aresuperimposed at 20°C intervals.

Figure 4. Time‐height plots of the maximum updraft undermaritime CCN conditions. Data are obtained from instanta-neous model output every 1 min. Contours are in intervalsof 4 m s−1 starting at 4 m s−1. The corresponding potentialtemperature (K) for significant altitude derived from thesounding data is plotted on the right axis.


6 of 15

height, with peak values of greater than 40 m s−1 appearing atapproximately 12–14 km altitude. This result agrees withobservation that peak updrafts in deep tropical convection arealmost always above the 10 km level [e.g., May andRajopadhyaya, 1999; Heymsfield et al., 2010]. The simu-lated maximum updraft velocity is 43.1 m s−1. This valueseems to be reasonable on the basis of the prestorm envi-ronment shown in Figure 1. Using the value of 3500 J kg−1 forCAPE and the relation Wmax =

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2CAPE

p, the theoretically

maximum updraft velocity turns out to be 84 m s−1. Whenconsidering precipitation loading, mixing effects of ambientair and perturbed vertical pressure gradients, that valueshould be reduced by roughly 50% [Bluestein, 1993], giving amaximum updraft velocity of 42 m s−1, which is fairly con-sistent with the value simulated by our model. After 32 minthe maximum vertical velocities decrease with time.[18] The first updraft penetrates the TTL (hereinafter

defined as the region between approximately 14 and 17.4 kmin altitude) after 27 min and about 4 min later the updraftshave overshot the CPT to reach into the LS. The updraftswithin the TTL and LS persist for over 20 min and ∼10 min,respectively, allowing time for ice and vapor to be transportedupward across the tropopause and into the stratosphere. Notethat a strong convection appears to detach from the top ofthe main convection during the time period 37–50 min. Thecenter of this convection is situated at the 18.3 km level, thepeak updraft velocity is 16.8 m s−1, and the top reaches over22 km. This strong updraft plays an important role in deter-mining the entry of the tropospheric ice particles into the LS.According to Lane et al. [2003] and Wang [2003], whensignificant wind shear exists near the tropopause, a conditionto favor gravity wave breaking, turbulence is generated atcloud top owing to gravity wave breaking. In this case, thevertical wind shear in the vicinity of the tropopause is rela-tively weak (Figure 1b), such that the new convection in theLS is not generated by the breaking of gravity waves.[19] Overshooting is defined as clouds with their tops

located higher than the CPT. The general structure and evo-lution of the simulated storm associated with the overshoot-ing plume are shown in Figure 5 in detail with a series ofvertical cross sections along the domain center. At 30 min, anupdraft from 6 to 18 km has fully developed, pushing theconvective dome (herein using a total water mixing ratiothreshold of 10−3 g kg−1) close to the CPT, while a low‐leveldowndraft has established owing to precipitation loading.Temperature of the middle‐low troposphere in cloud is higherthan the environment owing to the diabatic heating resultingfrom the latent heat released by freezing of liquid waterdroplets within the updraft. Note that air parcels in the uppertroposphere become progressively colder with height thanthe environment, and eventually form a cold trap region nearthe tropopause. This cold trap is likely due to cooling near thetropopause caused by a combination of adiabatic ascent andirreversible diabatic mixing of air overshooting its level ofneutral buoyancy [Sherwood, 2000;Robinson and Sherwood,2006]. Sherwood [2000] further argued that this cooling nearthe tropopause would drive descent above vigorous convec-tion and lead to compensating ascent through the tropopauseoccurring elsewhere. An unusual circulation thus occurs inthe troposphere and stratosphere. In our simulations, thisstratosphere‐troposphere circulation is established within

several minutes after the cold trap (Figure 5b). Subsequently,a new convection occurs 1–2 km above the tropopause anddevelops rapidly. At 40 min a strong convection has estab-lished well in the LS, just above the main storm (Figure 5c).The associated cloud cover extends upward into the LS andthe cloud top reaches 20 km altitude, whereas the mainupdraft and total water content below the CPT are decreasingrapidly. The vertical motion structure in the troposphere isalso changed, with one inner updraft region and one outerupdraft region between 10 and 16 km in altitude. The innerregion is around the center of the storm, and the outer region isadjacent to the edge of the storm. The outer updrafts extendthe cover of the cloud anvils, pushing more ice at a largerhorizontal extent upward into the TTL even the LS. The massand thermal exchanges between the storm and the environ-mental air are enhanced at the locations of the outer updrafts.At 50 min the updraft has decreased to 2 m s−1 leaving a weakperturbed circulation and the cloud dissipated by precipita-tion and mixing (Figure 5d). These results indicate that theovershooting could be an important mechanism for trans-porting water into the stratosphere. After 50 min rather icewater of 0.001–1 g kg−1 is still remained in the TTL(Figures 5e and 5f).[20] Lin et al. [2005] found that the ratio of ice mass to

liquidmass in themaritime subtropical thunderstorms is ∼1:1.This demonstrates that ice processes are highly important tocloud structure even in thunderstorms occurring in warmclimatic regimes. In our simulations, in a time‐averagedsense, the hydrometeor mass partitioning in the domain is8.2% for cloud water, 41.5% for rainwater, 18.7% for cloudice, 3.6% for snow, 8.6% for graupel, 12.9% for frozen dropsand 6.5% for hail, so that the ice phase accounts for about51% of the total hydrometeor mass. This percentage is verysimilar to that of the subtropical thunderstorm presented byLin et al. [2005]. Note also that cloud ice accounts for about19% of the total hydrometeor mass, contributing 40% of thetotal ice mass.[21] Corti et al. [2008] investigated this event in details and

found that the observed ice water content in the stratospherewas much too high to be formed in situ. He suggests that theice particles in the stratosphere likely result from overshoot-ing convection of the Hector. In our simulations, with themean over the whole 41 × 41 km domain and over the whole70 min duration, 5.6 ppmv and 3.1 ppbv of total ice in theform of cloud ice are found at 17.4 km and 20.1 km in altitude,respectively (Figure 6). The maximum cloud ice mixing ratioreaching these two levels is 2186.8 ppmv occurring at 34 minand 6.7 ppmv at 44 min, respectively. How are these highmixing ratios of stratospheric ice particles produced? Thereare two potential mechanisms for cloud ice formation in thesimulation: In situ nucleation and upward transportation fromthe cloud top. In the model, there are three generationmechanisms of ice crystals: primary nucleation, rime splin-tering and homogeneous freezing of cloud droplets. On thebasis of the simulation conducted here, we found that thesethree processes occurred primarily below 14 km owing to thelimitation of water vapor and temperature, as shown inFigure 7a, while depositional growth and sublimation pro-cesses of cloud ice crystals occurred in the stratosphereindeed (Figures 7b and 7c). After having excluded the in situformation, the simulated cloud ice particles in the stratosphere


7 of 15

Figure 5. Vertical cross sections of total water content larger than 10−3 g kg−1 (white), updraft greater than2 m s−1 (black), and the temperature field (shaded, °C) along x = 20 km under maritime CCN conditions forthe time after (a) 30min, (b) 35min, (c) 40min, (d) 50min, (e) 60min, and (f) 70min. The red line representsthe 380 K potential temperature isotherm.


8 of 15

are solely transported upward by convective overshootingwith subsequent mixing. This is similar to the conclusionmade by Corti et al. [2008] based on observations of thiscloud. The upward transporting ice crystal across the tropo-pause is found to occur during the time period 30–34 min and38–50min. Through calculating the vertical flux of ice crystalconcentrations, we found that the contribution induced by thesecond stage was over 75% of the total, so that the secondstage was very important for ice production in the LS. After48 min, evaporation of cloud ice is almost the only micro-physical process occurring in the LS (Figure 7c).[22] Figure 8 shows the time evolution of the mean water

vapor mixing ratios at each level above 14 km altitude. Onecan see that a significant dry layer is present between 16 and18 km in altitude at the early stage of the simulations. After27min, the water vapor mixing ratio within the TTL is rapidlyincreased, indicating a moistening of the TTL due to deepconvection. Meanwhile, the water vapor in the LS is alsoprogressively enhanced. At the end of the simulation, the totalmass of water vapor is ∼4140 tons (1000 kg) for the TTL and1366 tons for the LS, increased by about 128% and 5%,respectively, compared to the simulation start. By the endof the simulation, accumulated total mass of water vapor inthe LS is 92990 tons, during the 40 min integration periodcovering the overshoot, the total mass of water vapor is53618 tons, thereby resulting in an extra 14247 tons of watervapor entering the LS owing to the overshoot (Table 1). Theseresults clearly demonstrate that ice particles injected byovershooting have a moistening effect on the LS.[23] During the time period covering the overshoot, the

total mass of cloud ice entering the stratosphere is 56104 tonswith a corresponding increase rate of 23 tons s−1. Note thatrather total water is still left in the cloud at the end of thesimulation (Figure 5f). The amount of cloud ice that remainsin the stratosphere is about 2500 tons. This suggests that the

Figure 6. Vertical profiles of mean cloud ice mixing ratiounder maritime CCN conditions. Means are taken over theentire domain for the whole duration of the simulation.

Figure 7. Time‐height plots of the domain‐averaged cloudice produced by different microphysical processes (units in10−3 ppmv) under maritime CCN conditions: (a) nucleation,including primary nucleation, rime splintering, and homoge-neous freezing of cloud droplets; (b) depositional growth; and(c) evaporation.


9 of 15

moistening of the LS is potentially increased further givenmore time.

4.2. Simulation of Convective Clouds UnderContinental Conditions

[24] While clouds in both the maritime and the continentalcases began to form after 7 min of simulations, the clouds inthe latter case produced much more cloud water than in theformer. The maximum cloud droplet concentration in themaritime case is only 82 cm−3 while 814 cm−3 were producedin the continental case. Cloud water exists primarily belowthe melting level (5.1 km) in the maritime case (Figure 9a),whereas in the continental case, the cloud water exists as highas 10 km, at temperatures near −40°C. The core of the cloudwater is at 5–6 km or 0°C to −5°C (Figure 9b) with themaximum LWC of 7.5 g m−3. In comparison, only 4.7 g m−3

was obtained in the maritime case.[25] Higher droplet concentrations in the continental case

lead to stronger competition for available water vapor duringdroplet growth via diffusion and therefore restrict the size ofcloud droplets. As a consequence, droplet growth by thecollision‐coalescence process is less efficient in the conti-nental case than in the maritime case, resulting in a time delay(∼6 min) of raindrop formation in the continental case(Figures 9c and 9d). The initial raindrop formation is found inthe continental case at a much higher altitude than in themaritime case (5.1 versus 2.4 km). Although the accretion ofcloud water is the dominant growth mechanism for rainduring the first 30 min in both cases, accounting for about90% of the rain production, in the case of high CCN con-centrations, rain develops more slowly and less effectively ascompared to the case of low CCN concentrations, resulting inmuch lower rainwater content. After 30 min, nearly all of therainwater is below the 0°C level in both cases, which indi-cates that the melting of ice hydrometeors is the main sourceof the rain production. The largest source of the rainwater isthe melting of frozen drops in the maritime case, and meltingof graupel in the continental case. Melting of hail is the sec-ond largest source in both cases. The maximum rainwater

content is 14.5 gm−3 in themaritime case versus 12.8 gm−3 inthe continental case, indicating that the increase in aerosols inthe continental thunderstorm produces more cloud waterbut much less rain, as compared to the maritime thunder-storm. Correspondingly, the accumulated precipitation at theground is also reduced with increasing CCN concentrations(Table 1).[26] The continental clouds have more ice crystals and less

snow compared to the maritime clouds (Figures 9e–9h). Mostof cloud ice is located generally above 11 km, at temperaturesbelow −40°C, and are formed by homogeneous freezing ofcloud droplets. The maximum concentration reaches 60 cm−3

in the continental case and 30 cm−3 in the maritime case, bothoccurring at 33 min. The mass of ice crystals in the conti-nental case is also slightly higher than that in the maritimecase, with a maximum in the domain averaged values of0.984 g m−3 versus 0.978 g m−3. The higher ice crystal con-centration in the continental case leads to stronger competi-tion for available water vapor during depositional growthand therefore restricts the size of ice crystals. As a result, thesnow process is less efficient in the continental case.[27] In the model, freezing raindrops can convert into

graupel or frozen drops depending on the raindrop diameterDfr. The destination particle is then classified as graupel whenDfr < 1mm, and as frozen drop whenDfr > 1mm. The fractionof small raindrops with Dfr < 1 mm increases with increasingCCN concentrations. Subsequently, the concentration ofgraupel particles in the continental case (with a maximumof ∼60 L−1) is significantly higher than that in the maritimecase (6 L−1). The graupel mass content in the continental caseis also significantly higher than that in the maritime case, witha maximum of 12.0 g m−3 versus 1.0 g m−3. But the maritimeclouds have many more frozen drops (with the maximum of900 m−3 in the concentration and 6.6 g m–3 in mass content)compared to the continental clouds (760 m−3 and 4.4 g m−3).[28] The high CCN concentration leads to more hails in the

continental case than in the maritime case (Figures 9m and9n). The maximum values are 20 m−3 for number concen-tration and 8.8 g m−3 for mass content in the continental case,while only 10m−3 and 4.7 gm−3 appears in the maritime case.The increase in the hail is mainly due to the increase in

Figure 8. Time‐height plots of the domain‐averaged watervapor mixing ratio (ppmv) at each altitude under maritimeCCN conditions.

Table 1. Cloud Properties Under the Maritime (Case‐M) andContinental (Case‐C) CCN Conditionsa

Case‐M Case‐C

Maximum updraft in the cloud (m s−1) 43.1 37.2Maximum updraft in the LS (m s−1) 16.8 14.5Accumulated total mass of

condensate (KT)54515 56787

Ratio of ice mass to liquid mass ∼5:5 ∼6:4Total increase in stratospheric ice

caused by overshoot (tons)56104 64193

Accumulated total mass ofstratospheric vapor (tons)

92990 93086

Total increase in stratospheric vaporcaused by overshoot (tons)

14247 14343

Accumulated total mass of liquidprecipitation (KT)

1686 1550

Accumulated total mass of solidprecipitation (KT)

97 70

aMaximum updraft is given by an instantaneous value. The accumulatedtotal mass is the overall amount produced throughout the simulation.


10 of 15

Figure 9. Time‐height plots of domain‐averaged hydrometeormass density (gm−3) under (a, c, e, g, i, k, m)maritime and (b, d, f, h, j, l, n) continental CCN conditions.


11 of 15

autoconversion of graupel and accretion of cloud water byhail, as shown in Figures 9i and 9j.[29] From Figure 10 and Table 1, it can be found that in the

high CCN concentration case (the continental case), totalcondensate mass for the entire thunderstorm system isincreased by 4.2%, as compared to that in the low CCNconcentration case (the maritime case). This result indicatesthat the total condensate mass in tropical thunderstormsincreases with an increase in the CCN concentration. Fur-thermore, the partitioning of hydrometeors in thunderstormsis also related to the CCN concentration. As shown inFigure 10, the continental thunderstorms have more cloudwater, cloud ice, graupel and hail mass, but less rainwater andfrozen drop mass than in the maritime case. Overall, icehydrometeors account for about 57% of the total hydrometeormass, showing that the fraction of ice phase in the thunder-storms increases with increasing CCN concentrations.[30] Figure 11 shows the time evolution of the maximum

updraft velocity at each level during each 1 min interval.Compared to the maritime case, the clouds that developed inthe high CCN concentration case have larger updraft veloc-ities before 28 min, occurring at the middle levels. As shownin Figures 9a and 9b, most of the cloud water develops before30 min in the simulations. But the continental case producesmore latent heat during condensation and freezing because ofthe higher droplet concentration and the higher levels that thedroplets reach, leading to the increase in the updraft velocity.In the maritime case, low CCN concentrations lead to theformation of raindrops at lower levels. In the absence ofvertical wind shear (Figure 1b), the effects of the latent heat ofcondensation on the cloud dynamics are offset by the pre-cipitation loading mainly in the form of rain, resulting insmaller vertical velocities in the maritime case than in thecontinental case. After 28 min, however, the maximumupdraft velocities in the maritime case in the upper levels ofthe cloud are higher than in the continental case which pro-duced a peak updraft of 37.2 m s−1. On the basis of theaforementioned microphysical analysis, one can see that thecontinental clouds have more precipitation mass in the upper

levels during the mature stage of thunderstorm, especially thegraupel (Figures 9i and 9j) and hail (Figures 9m and 9n). Theloading effect of precipitation mass suppresses the upwardmotion and causes the storm to have a weaker updraft ascompared to the case of low maritime CCN concentrations.Correspondingly, a weaker cold trap region is formed in thecontinental case, thus resulting in a weaker descent above thecloud top and a weaker convection in the LS with the maxi-mum of 14.5 m s−1. This result indicates that increasing CCNconcentration weakens the intensity of overshooting tropicalthunderstorms.

4.3. Sensitivity of Ice and Water Vapor Contentin the LS to CCN Concentrations

[31] Considering that the cold point tropopause was situ-ated at 17.3 km altitude for this event and the vertical grid sizeused in the simulations, the LS boundary in our work isdefined at 17.4 km. Figure 12 shows the stratospheric meanchanges in total ice and water vapor content between the

Figure 10. The percentage contributions of each hydrometeor to the total condensate mass, obtained fromthe integration over the entire domain and duration time under (a) maritime CCN conditions and (b) conti-nental CCN conditions. The net total mass (microphysical source minus sink) is also shown.

Figure 11. Same as Figure 4 but for the continental CCNcase.


12 of 15

continental and maritime cases. One can see that the increasein CCN leads to increases in both the ice mass and numberconcentrations and vapor content in the LS. For water vapor,maximum increments in the domain and time averaged valuesof 4.4 ppbv were found in the LS occurring at 17.7 km alti-tude. This change is directly associated with homogeneousfreezing of cloud droplets and subsequent transportingupward by overshooting. The increase in CCN concentrationcause a large increase in the number of small ice particlesformed owing to droplet homogeneous freezing process,since the more numerous small droplets form more ice par-ticles when they reach the homogeneous freezing zone andthen freeze. This subsequently results in more ice particlesbeing advected up into the LS, so that the ice crystal numberconcentrations within the LS are eventually increased. Cor-responding to the significant increase in the number con-centration of ice crystals, which would have the effect ofreducing the sizes and fall speeds of the ice crystal, the vapormass is increased subsequently by sublimation of the injectedcloud ice particles.[32] To better understand the effect of the increasing CCN

concentrations on ice and water vapor in the LS, Figure 13presents the time evolution of the difference between thecontinental and maritime cases in accumulated total mass.The result shows that the rapid increase in ice water contentoccurs after convective overshooting especially for the timeperiod 30–34 min when the thunderstorm enters the maturestage. This increase is attributed to a large number of theinjected ice crystals by strong updrafts. Within the followingseveral minutes, the increasing effect is reduced. As men-tioned before, during 35–37min the vertical motion above thecloud top is dominated by descent, thus some ice mass in the

LS are transported downward across the tropopause and backinto the troposphere. After 38 min, the ice mass in the LSbegins increasing slowly when the stratosphere‐troposphereconvection develops above the cloud top. However, theincrease in ice water mass during the vigorous stratosphere‐troposphere convection is smaller, when compared to the timeperiod 30–34 min. After 45 min, the ice water mass in the LSis progressively increased although there is little verticaltransporting of ice crystals across the tropopause when thestratosphere‐troposphere convection decayed. This increasein the ice mass is wholly caused by less precipitation loss ofice by lower sedimentation. In the continental case, the sizeand fall speed of the ice crystals is smaller than that in themaritime case. Overall, the mean diameter and fall speed ofthe ice crystal in the LS for the continental clouds has beenreduced by 40% and 22%, respectively. Lower fall speedslead to less precipitation loss of ice by sedimentation andallow more ice to remain in the LS. Overall, after 50 minwhen there is no vertical motion near the tropopause, theincrease in ice mass is 5570 tons, accounting for 70% of thetotal increase, so that the sedimentation effect dominatesthe ice mass in the LS. Furthermore, slow sedimentation ofthe ice leads to increased evaporation upon mixing with thesurrounding air, thereby causing an increase in the strato-spheric water vapor content.[33] By the end of the simulation, the accumulated

increases caused by increasing CCN in the LS are 8089 tonsfor cloud ice and 96 tons for water vapor, with a corre-sponding increase of 14.4% and 0.1%, respectively. At theend of the simulation, the amount of cloud ice that remains inthe stratosphere is 2774 tons in the continental case. Thatvalue is higher than in the maritime case. This suggests thatthe stratospheric water vapor would be potentially increasedfurther given more time.

5. Summary and Conclusions

[34] In this paper we investigate the effects of aerosolsacting as CCN on tropical overshooting thunderstorms and

Figure 12. Differences between the continental and mari-time cases for mean total ice concentrations dNi (1000 L−1)and mixing ratios dQi (ppmv) and water vapor mixing ratiosdQv (ppbv). Means are taken over the domain and wholeduration period.

Figure 13. Time series of the differences between thecontinental and maritime cases in accumulated total icemass (blue) and water vapor mass (red) in the region above17.4 km.


13 of 15

stratospheric water vapor using a three‐dimensional non-hydrostatic convective cloud model with a double‐momentbulk microphysics scheme and explicit droplet activation. Asevere Hector thunderstorm that developed on 30 November2005 during the SCOUT‐O3/ACTIVE campaign was simu-lated under maritime and continental aerosol scenarios.[35] The results show that ice‐phase hydrometeors account

for over 50% of the total hydrometeor mass of the tropicalthunderstorms. This is consistent with the subtropicalthunderstorms simulated by Lin et al. [2005] and furtherdemonstrates that ice processes are highly important indetermining cloud structure in thunderstorms occurring inwarm climatic regimes. Ice crystals are highly important andcontribute about 19% of the total hydrometeor mass in thissevere tropical thunderstorm. It is also found that the cloud iceparticles in the LS are not formed in situ but are transportedupward by convective overshoots with subsequent mixingoccurring. The penetration of convective overshoots has amoistening effect on the LS by sublimation of the injected iceparticles.[36] It is shown that aerosols significantly influence the

cloud microphysics and dynamics of the simulated tropicalthunderstorms. An increase in the CCN concentration leadsto more cloud water, cloud ice, graupel and hail, but lessrainwater and frozen drops. Total cloud condensate mass isenhanced as well. Moreover, the partitioning to ice‐phasehydrometeors in the storms increases with an increase in theCCN concentration. With more ice crystals being transportedupward into the LS, which would have the effect of reducingthe sizes and fall speeds of the ice crystal, the stratosphericwater vapor content is increased by sublimation of theinjected ice particles. This result suggests that aerosols in theboundary layer do affect the entry of water vapor in the LS,in agreement with the work of Sherwood [2002].[37] Note that the simulated overshooting convection in the

high CCN concentration case was found to be weakenedsomewhat owing to the loading effect of increased preci-pitation mass at the middle and upper levels, when com-pared to the low CCN concentration case. In this work,however, microphysical effects induced by aerosol increasesdominate, resulting in a moistening effect on the strato-sphere. In contrast, a reduced moistening effect might beproduced if dynamical effects play a dominant role. If thiswas the case then this would suggest that the dependenceof the water vapor content of the LS on aerosol concentra-tions might be nonlinear. Connolly et al. [2006] and Li et al.[2008] found nonmonotonic change of the convectionintensity with aerosols, because of the complicated couplingbetween cloud microphysics and dynamics. The presentwork only investigated one case and two aerosol scenarios,more cases and sensitivity tests are needed to determinewhether aerosols increase or decrease the stratosphericwater vapor.

[38] Acknowledgments. We thank the ACTIVE team for providingus with the measurement data and the three anonymous reviewers for their con-structive comments and suggestions, which helped us to clarify some of theimportant issues and to improve the quality of the manuscript. This workwas jointly supported by the National Natural Science Foundation of China(grants 40775005, 41175118, 40675005, 41030962, and 41075029) and the973 Program of China through grants 2010CB428504, 2010CB428602, and2011CB403406.

ReferencesAllen, G., et al. (2008), Aerosol and trace‐gas measurements in the Darwinarea during the wet season, J. Geophys. Res., 113, D06306, doi:10.1029/2007JD008706.

Andreae, M. O., D. Rosenfeld, P. Artaxo, A. A. Costa, G. P. Frank, K. M.Longo, and M. F. Silvas‐Dias (2004), Smoking rain clouds over theAmazon, Science, 303, 1337–1342, doi:10.1126/science.1092779.

Bluestein, H. B. (1993), Observations and Theory of Weather Systems,vol. 2, 594 pp., Oxford Univ. Press, Oxford, U. K.

Brunner, D., et al. (2009), The SCOUT‐O3 Darwin Aircraft Campaign:Rationale and meteorology, Atmos. Chem. Phys. , 9 , 93–117,doi:10.5194/acp-9-93-2009.

Carrió, G. G., S. C. van den Heever, and W. R. Cotton (2007), Impacts ofnucleating aerosol on anvil‐cirrus clouds: A modeling study, Atmos. Res.,84, 111–131.

Chaboureau, J.‐P., J.‐P. Cammas, J. Duron, P. J. Mascart, N. M. Sitnikov,and H.‐J. Voessing (2007), A numerical study of tropical cross‐tropopause transport by convective overshoots, Atmos. Chem. Phys., 7,1731–1740, doi:10.5194/acp-7-1731-2007.

Chemel, C., M. R. Russo, J. A. Pyle, R. S. Sokhi, and C. Schiller (2009),Quantifying the imprint of a severe Hector thunderstorm duringACTIVE/SCOUT‐O3 onto the water content in the upper troposphere/lower stratosphere, Mon. Weather Rev., 137, 2493–2514, doi:10.1175/2008MWR2666.1.

Chen, B., and H. Xiao (2010), Silver iodide seeding impact on the micro-physics and dynamics of convective clouds in the High Plains, Atmos.Res., 96, 186–207, doi:10.1016/j.atmosres.2009.04.001.

Connolly, P. J., T. W. Choularton, M. W. Gallagher, K. N. Bower, M. J.Flynn, and J. A. Whiteway (2006), Cloud‐resolving simulations ofintense tropical Hector thunderstorms: Implications for aerosol–cloudinteractions, Q. J. R. Meteorol. Soc., 132, 3079–3106, doi:10.1256/qj.05.86.

Cooper, W. A. (1986), Ice initiation in natural clouds, in PrecipitationEnhancement: A Scientific Challenge, Meteorol. Monogr., vol. 43,pp. 29–32, Am. Meteorol. Soc., Boston, Mass.

Corti, T., et al. (2008), Unprecedented evidence for overshooting convec-tion hydrating the tropical stratosphere, Geophys. Res. Lett., 35,L10810, doi:10.1029/2008GL033641.

Crook, N. A. (2001), Understanding Hector: The dynamics of islandthunderstorms, Mon. Weather Rev., 129, 1550–1563, doi:10.1175/1520-0493(2001)129<1550:UHTDOI>2.0.CO;2.

Cui, Z., K. S. Carslaw, and Y. Yin (2006), A numerical study of aerosoleffects on the dynamics and microphysics of a deep convective cloudin a continental environment, J. Geophys. Res., 111, D05201,doi:10.1029/2005JD005981.

Danielsen, E. F. (1982), A dehydration mechanism for the stratosphere,Geophys. Res. Lett., 9, 605–608, doi:10.1029/GL009i006p00605.

Danielsen, E. F. (1993), In situ evidence of rapid, vertical, irreversibletransport of lower tropospheric air into the lower tropical stratosphereby convective cloud turrets and by large‐scale upwelling in tropicalcyclones, J. Geophys. Res., 98, 8665–8681, doi:10.1029/92JD02954.

de Reus, M., S. Borrmann, A. Bansemer, A. J. Heymsfield, R. Weigel,C. Schiller, V. Mitev, and W. Frey (2009), Evidence for ice particlesin the tropical stratosphere from in‐situ measurements, Atmos. Chem.Phys., 9, 6775–6792, doi:10.5194/acp-9-6775-2009.

Grabowski, W. W. (1999), A parameterization of cloud microphysics forlong‐term cloud‐resolving modeling of tropical convection, Atmos.Res., 52, 17–41, doi:10.1016/S0169-8095(99)00029-0.

Grosvenor, D. (2010), Interactive comment on “Water vapour budget asso-ciated to overshoots in the tropical stratosphere: Mesoscale modellingstudy of 4–5 August 2006 during SCOUT‐AMMA” by X. M. Liu et al.,Atmos. Chem. Phys. Discuss., 10, C1348–C1351.

Grosvenor, D. P., T. W. Choularton, H. Coe, and G. Held (2007), A studyof the effect of overshooting deep convection on the water content of theTTL and lower stratosphere from Cloud Resolving Model simulations,Atmos. Chem. Phys., 7, 4977–5002, doi:10.5194/acp-7-4977-2007.

Hallett, J., and S. C. Mossop (1974), Production of secondary ice particlesduring the riming process, Nature, 249, 26–28, doi:10.1038/249026a0.

Hassim, M. E. E., and T. P. Lane (2010), A model study on the influence ofovershooting convection on TTL water vapour, Atmos. Chem. Phys., 10,9833–9849, doi:10.5194/acp-10-9833-2010.

Heymsfield, G. M., L. Tian, A. J. Heymsfield, L. Li, and S. Guimond(2010), Characteristics of deep tropical and sub‐tropical convection fromnadir‐viewing high‐altitude airborne Doppler radar, J. Atmos. Sci., 67,285–308, doi:10.1175/2009JAS3132.1.

Hogan, R. J., M. P. Mittermaier, and A. J. Illingworth (2006), The retrievalof ice water content from radar reflectivity factor and temperature and itsuse in evaluating a mesoscale model, J. Appl. Meteorol. Climatol., 45,301–317, doi:10.1175/JAM2340.1.


14 of 15

Hong, Y., and P. Fan (1999), Numerical simulation study of hail cloud,Part I: The numerical model, Acta Meteorol. Sin., 13, 188–199.

Hu, Z., and G. He (1988), Numerical simulation of microphysical processesin cumulonimbus, Part I: Microphysical model, Acta Meteorol. Sin., 2,471–489.

Iwasaki, S., T. Shibata, J. Nakamoto, H. Okamoto, H. Ishimoto, andH. Kubota (2010), Characteristics of deep convection measured by usingthe A‐train constellation, J. Geophys. Res., 115, D06207, doi:10.1029/2009JD013000.

Keenan, T. D., B. R. Morton, M. J. Manton, and G. J. Holland (1989),The Island Thunderstorm Experiment (ITEX): A study of tropicalthunderstorms in the Maritime Continent, Bull. Am. Meteorol. Soc., 70,152–159, doi:10.1175/1520-0477(1989)070<0152:TITESO>2.0.CO;2.

Keenan, T. D., et al. (2000), The Maritime Continent Thunderstorm Exper-iment (MCTEX): Overview and some results, Bull. Am. Meteorol. Soc.,81 , 2433–2455 , do i :10 .1175 /1520-0477(2000)081<2433 :TMCTEM>2.3.CO;2.

Khain, A., and B. Lynn (2009), Simulation of a supercell storm in clean anddirty atmosphere using weather research and forecast model with spectralbin microphysics, J. Geophys. Res., 114, D19209, doi:10.1029/2009JD011827.

Khain, A., D. Rosenfeld, and A. Pokrovsky (2005), Aerosol impact on thedynamics and microphysics of deep convective clouds, Q. J. R. Meteorol.Soc., 131, 2639–2663, doi:10.1256/qj.04.62.

Khain, A., N. Cohen, B. Lynn, and A. Pokrovsky (2008), Possible aerosoleffects on lightning activity and structure of hurricanes, J. Atmos. Sci.,65, 3652–3677, doi:10.1175/2008JAS2678.1.

Khairoutdinov, M. F., and Y. L. Kogan (2000), A new cloud physicsparameterization in a large‐eddy simulation model of marine stratocumu-lus, Mon. Weather Rev., 128, 229–243, doi:10.1175/1520-0493(2000)128<0229:ANCPPI>2.0.CO;2.

Khaykin, S., J.‐P. Pommereau, L. Korshunov, V. Yushkov, J. Nielsen,N. Larsen, T. Christensen, A. Garnier, A. Lukyanov, and E. Williams(2009), Hydration of the lower stratosphere by ice crystal geysers overland convective systems, Atmos. Chem. Phys. , 9 , 2275–2287,doi:10.5194/acp-9-2275-2009.

Klemp, J. B., and R. B. Wilhelmson (1978), The simulation of three‐dimensional convective storm dynamics, J. Atmos. Sci., 35, 1070–1096, doi:10.1175/1520-0469(1978)035<1070:TSOTDC>2.0.CO;2.

Kong, F., M. Huang, and H. Xu (1990), Three‐dimensional numericalsimulation of ice phase microphysics in cumulus clouds, Part I: Modelestablishment and ice phase parameterization, Chin. J. Atmos. Sci., 14,441–453.

Lane, T. P., R. D. Sharman, T. L. Clark, and H.‐M. Hsu (2003), An inves-tigation of turbulence generation mechanisms above deep convection,J. Atmos. Sci., 60, 1297–1321, doi:10.1175/1520-0469(2003)60<1297:AIOTGM>2.0.CO;2.

Lerach, D. G., B. J. Gaudet, and W. R. Cotton (2008), Idealized simulationsof aerosol influences on tornadogenesis, Geophys. Res. Lett., 35, L23806,doi:10.1029/2008GL035617.

Li, G., Y. Wang, and R. Zhang (2008), Implementation of a two‐momentbulk microphysics scheme to the WRF model to investigate aerosol‐cloud interaction, J. Geophys. Res., 113, D15211, doi:10.1029/2007JD009361.

Li, G., Y. Wang, K.‐H. Lee, Y. Diao, and R. Zhang (2009), Impacts ofaerosols on the development and precipitation of a mesoscale squall line,J. Geophys. Res., 114, D17205, doi:10.1029/2008JD011581.

Lin, H., P. K. Wang, and R. E. Schlesinger (2005), Three‐dimensional non-hydrostatic simulations of summer thunderstorms in the humid subtropicsversus High Plains, Atmos. Res., 78, 103–145, doi:10.1016/j.atmosres.2005.03.005.

Lin, Y. L., R. D. Farley, and H. D. Orville (1983), Bulk parameterization ofthe snow field in a cloud model, J. Clim. Appl. Meteorol., 22, 1065–1092, doi:10.1175/1520-0450(1983)022<1065:BPOTSF>2.0.CO;2.

Liu, X. M., E. D. Rivière, V. Marécal, G. Durry, A. Hamdouni, J. Arteta,and S. Khaykin (2010), Stratospheric water vapour budget and con-vection overshooting the tropopause: Modell ing study fromSCOUT‐AMMA, Atmos. Chem. Phys., 10, 8267–8286, doi:10.5194/acp-10-8267-2010.

May, P. T., and D. K. Rajopadhyaya (1999), Vertical velocity characteris-tics of deep convection over Darwin, Australia, Mon. Weather Rev., 127,1056–1071, doi:10.1175/1520-0493(1999)127<1056:VVCODC>2.0.CO;2.

May, P. T., G. Allen, G. Vaughan, and P. Connolly (2009), Aerosol andthermodynamic effects on tropical cloud systems during TWPICE andACTIVE, Atmos. Chem. Phys., 9, 15–24, doi:10.5194/acp-9-15-2009.

Morrison, H., J. A. Curry, and V. I. Khvorostyanov (2005), A new double‐moment microphysics parameterization for application in cloud and cli-mate models, Part I: Description, J. Atmos. Sci., 62, 1665–1677,doi:10.1175/JAS3446.1.

Nielsen, J. K., N. Larsen, F. Cairo, G. Di Donfrancesco, J. M. Rosen,G. Durry, G. Held, and J. P. Pommereau (2007), Solid particles in thetropical lowest stratosphere, Atmos. Chem. Phys., 7, 685–695,doi:10.5194/acp-7-685-2007.

Oltmans, S. J., H. Vömel, D. J. Hofmann, K. H. Rosenlof, and D. Kley(2000), The increase in stratospheric water vapor from balloonborne,frostpoint hygrometer measurements at Washington, D.C., and Boulder,Geophys. Res. Lett., 27, 3453–3456, doi:10.1029/2000GL012133.

Phillips, V. T. J., L. J. Donner, and S. Garner (2007), Nucleation processesin deep convection simulated by a cloud‐system‐resolving model withdouble‐moment bulk microphysics, J. Atmos. Sci., 64, 738–761,doi:10.1175/JAS3869.1.

Reisin, T., Z. Levin, and S. Tzvion (1996), Rain production in convectiveclouds as simulated in an axisymmetric model with detailed micro-physics, Part 1: Description of the model, J. Atmos. Sci., 53, 497–519,doi:10.1175/1520-0469(1996)053<0497:RPICCA>2.0.CO;2.

Robinson, F. J., and S. C. Sherwood (2006), Modeling the impact ofconvective entrainment on the tropical tropopause, J. Atmos. Sci., 63,1013–1027, doi:10.1175/JAS3673.1.

Rogers, R. R., and M. K. Yau (1989), A Short Course in Cloud Physics,3rd ed., 293 pp., Pergamon, New York.

Romps, D. M., and Z. Kuang (2009), Overshooting convection in tropicalcyclones, Geophys. Res. Lett., 36, L09804, doi:10.1029/2009GL037396.

Rosenlof, K. H. I., E.‐W. Chiou, W. P. Chu, D. G. Johnson, K. K. Kelly,H. A. Michelsen, G. E. Nedoluha, E. E. Remsberg, G. C. Toon, and M. P.McCormick (2001), Stratospheric water vapor increases over the pasthalf‐century, Geophys. Res. Lett., 28, 1195–1198, doi:10.1029/2000GL012502.

Sherwood, S. C. (2000), A stratospheric “drain” over the maritime conti-nent, Geophys. Res. Lett., 27, 677–680, doi:10.1029/1999GL010868.

Sherwood, S. C. (2002), A microphysical connection among biomassburning, cumulus clouds, and stratospheric moisture, Science, 295,1272–1275, doi:10.1126/science.1065080.

Sherwood, S., and A. E. Dessler (2000), On the control of stratospherichumidity, Geophys. Res. Lett. , 27 , 2513–2516, doi:10.1029/2000GL011438.

Sherwood, S., and A. E. Dessler (2001), A model for transport across thetropical tropopause, J. Atmos. Sci., 58, 765–779, doi:10.1175/1520-0469(2001)058<0765:AMFTAT>2.0.CO;2.

Toon, O. B., et al. (2010), Planning, implementation, and first results of theTropical Composition, Cloud and Climate Coupling Experiment (TC4),J. Geophys. Res., 115, D00J04, doi:10.1029/2009JD013073.

Twomey, S. (1959), The nuclei of natural cloud formation: The supersatu-ration in natural clouds and the variation of cloud droplet concentration,Geofis. Pura Appl., 43, 243–249, doi:10.1007/BF01993560.

van den Heever, S. C., G. G. Carrió, W. R. Cotton, P. J. DeMott, and A. J.Prenni (2006), Impact of nucleating aerosol on Florida storms, Part 1:Mesoscale simulations, J. Atmos. Sci., 63, 1752–1775, doi:10.1175/JAS3713.1.

Vaughan, G., C. Schiller, A. R. MacKenzie, K. Bower, T. Peter, H. Schlager,N. R. P. Harris, and P. T. May (2008), SCOUTO3/ACTIVE: High‐altitudeaircraft measurements around deep tropical convection, Bull. Am.Meteorol. Soc., 89, 647–662, doi:10.1175/BAMS-89-5-647.

Wang, P. K. (2003), Moisture plumes above thunderstorm anvils and theircontributions to cross‐tropopause transport of water vapor in midlati-tudes, J. Geophys. Res., 108(D6), 4194, doi:10.1029/2002JD002581.

Yin, Y., and L. Chen (2007), The effects of heating by transported dustlayers on cloud and precipitation: A numerical study, Atmos. Chem.Phys., 7, 3497–3505, doi:10.5194/acp-7-3497-2007.

Zhang, H., G. M. McFarquhar, S. M. Saleeby, and W. R. Cotton (2007),Impacts of Saharan dust as CCN on the evolution of an idealized tropicalcyclone, Geophys. Res. Lett., 34, L14812, doi:10.1029/2007GL029876.

Zhang, R. (2010), Getting to the critical nucleus of aerosol formation,Science, 328, 1366–1367, doi:10.1126/science.1189732.

B. Chen, Key Laboratory of Mesoscale Severe Weather, MOE, NanjingUniversity, Nanjing 210093, China. ([email protected])Y. Yin, CMA Key Laboratory for Atmospheric Physics and Environment,

Nanjing University of Information Science and Technology, Nanjing210044, China.


15 of 15

modeling the impact of aerosols on tropical overshooting...

Documents