Books - Non-fiction

A Study of the Aerosol Effect on a Cloud Field with Simultaneous Use of GCM Modeling and Satellite Observation

The indirect effect of aerosols was simulated by a GCM for nonconvective water clouds and was compared with remote sensing results from the Advanced Very High Resolution Radiometer (AVHRR) satellite-borne sensor for January, April, July, and October
of 16
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  15 J ANUARY  2004  179 S U Z U K I E T A L .   2004 American Meteorological Society A Study of the Aerosol Effect on a Cloud Field with Simultaneous Use of GCMModeling and Satellite Observation K ENTAROH  S UZUKI AND  T ERUYUKI  N AKAJIMA Center for Climate System Research, University of Tokyo, Tokyo, Japan A TUSI  N UMAGUTI Graduate School of Environmental Earth Sciences, Hokkaido University, Hokkaido, Japan T OSHIHIKO  T AKEMURA  Research Institute for Applied Mechanics, Kyushu University, Fukuoka, Japan K AZUAKI  K AWAMOTO  Research Institute for Humanity and Nature, Kyoto, Japan A KIKO  H IGURASHI  National Institute for Environmental Studies, Tsukuba, Japan (Manuscript received 3 September 2002, in final form 1 July 2003)ABSTRACTThe indirect effect of aerosols was simulated by a GCM for nonconvective water clouds and was comparedwith remote sensing results from the Advanced Very High Resolution Radiometer (AVHRR) satellite-bornesensor for January, April, July, and October of 1990.The simulated global distribution of cloud droplet radius showed a land–sea contrast and a characteristicfeature along the coastal region similar to the AVHRR results, although cloud droplet radii from GCMcalculationsand AVHRR retrievals were different over tropical marine regions due to a lack of calculation of cloud–aerosolinteraction for convective clouds in the present model and also due to a possible error in the satellite retrievalcaused by cirrus and broken cloud contamination. The simulated dependence of the cloud properties on thecolumn aerosol particle number was also consistent with the statistics obtained by the AVHRR remote sensingwhen a parameterization with the aerosol lifetime effect was incorporated in the simulation. The global averageof the simulated liquid water path based on the parameterization with the aerosol lifetime effect showed aninsignificant dependence on the aerosol particle number as a result of a global balance of the lifetime effect andthe wash-out effect. This dependence was contrary to the results of simulations based on the Sundqvist’sparameterization without aerosol lifetime effect; that is, the simulated cloud liquid water path showed a decreasingtendency with the aerosol particle number reflecting only the wash-out effect. 1. Introduction Clouds play an important role in the earth’s climatethrough their effect on the atmospheric radiative pro-cesses and the global hydrologic cycle. One of the im-portant properties of clouds significant for the climaticeffect is their microphysical structure, because this de-termines the cloud optical properties and the precipi-tation production efficiency of the cloud system. It hasbeen known since the 1950s that clouds over land and Corresponding author address:  Teruyuki Nakajima, 4-6-1 Ko-maba, Meguro-ku, Tokyo 153, Japan.E-mail: ocean have systematically different microphysicalstruc-tures associated with different abundances of aerosolsacting as cloud condensation nuclei (CCN; Squires1958) as well as with the difference in the dynamicalcondition. Two kinds of indirect effects of atmosphericaerosols acting as CCN have been recognized to modifythe cloud microphysical and radiative properties. Oneis the so-called Twomey effect or the first kind of in-direct effect in which an increase in the aerosol particlenumber concentration causes an increase in the clouddroplet number concentration and a reduction in thedroplet radius when the cloud liquid water path is as-sumed constant. The increase in the cloud particle num-  180  V OLUME  61J O U R N A L O F T H E A T M O S P H E R I C S C I E N C E S ber further induces an increase in the cloudopticalthick-ness and hence the cloud reflectivity (Twomey 1974).The other effect is the so-called lifetime effect or thesecond kind of indirect effect in which smaller clouddroplets have a longer lifetime before growing into rain-drops, leading to an enhancement of the cloud liquidwater path. The aerosol-induced reduction in the clouddroplet size is followed by suppression of rainfall pro-duction from the cloud and reduced scavenging effi-ciency of the aerosols (Albrecht 1989).Large scale signatures of the aerosol effects on cloudoptical properties have been observed in the ship trailcloud phenomenon and in the cloud modification byoutflow of continental airmasses over coastal regions(Conover 1966; Coakley et al. 1987; Radke et al. 1989;Nakajima and Nakajima 1995). They found that marinecloud reflectivity was enhanced along the ship trajectoryand in the continental air flow due to the indirect effectof aerosols resulting from the input of large amounts of CCN into the clean maritime atmosphere. Smoke aero-sol particles emitted by biomass burning have also beendetected to produce an influence on the cloud opticalproperties over land as pointed out by Kaufman andNakajima (1993) for thick clouds and Kaufman andFraser (1997) for thin clouds. The global features of theland–ocean contrast of cloud optical properties havebeen revealed by Han et al. (1994) from Advanced VeryHigh Resolution Radiometers (AVHRR) global analy-sis. More recently, aerosol impacts on the precipitationefficiency of cloud system have also been observed withthe Tropical Rainfall Measuring Mission (TRMM). Ro-senfeld (1999) reported that the rainfall from tropicalconvective clouds was inhibited by forest fire smokeover Indonesia and provided the first direct evidence of the effect of aerosols on the precipitation from convec-tive clouds. Similarly, urban and industrial air pollutionwas also detected to suppress precipitation (Rosenfeld2000). Masunaga et al. (2002) have pointed out that thevertical stratification of the effective particle radius issignificantly modified by the CCN amount.The above discussion suggests that the human-in-duced increase in aerosols has produced a significantchange in the cloud microphysical structure. The radi-ative forcing of the anthropogenic indirect effect hasbeen considered to be comparable to but with the op-posite sign of the manmade greenhouse gas forcing(Twomey et al. 1984; Charlson et al. 1992). The indirectradiative forcing of anthropogenic aerosols, therefore,needs to be evaluated quantitatively in order to under-stand and predict global climate change. Some attemptsto estimate the aerosol indirect forcing with general cir-culation models (GCM) have been made in the last sev-eral years (Jones et al. 1994; Boucher and Lohmann1995; Jones and Slingo 1996; Chuang et al. 1997; Loh-mann and Feichter 1997; Rotstayn 1999; Lohmann etal. 2000). Satisfactory estimates, however, have not yetbeen achieved. The radiative forcing of the first kind of indirect effect ranges from 0 to   2 W m  2 (Houghtonet al. 1996, 2001), and the second kind of effect hasbeen hardly understood. Uncertainty in the estimate canbe attributed to the problems and ambiguities regardingthe treatment of aerosols and clouds in the models.Motivated by this difficulty in understanding theaero-sol indirect effect, several observational research proj-ects on this issue have been developed. Han et al. (2002)investigated the correlation between cloud liquid waterpath and cloud droplet number concentration based onglobal satellite-retrieved data to find three different be-haviors of positive, neutral, and negative correlationsbetween the two parameters. They suggested that thenegative correlation is associated with reduction of wa-ter vapor supply and cloud liquid water when cloudsare decoupled from the boundary layer. Wetzel andStowe (1999) showed that the cloud droplet size fromAVHRR observations is inversely correlated with theaerosol optical thickness over the global ocean, sug-gesting the critical value of aerosol optical thicknessabove which the cloud properties did not change. Morerecently, Nakajima et al. (2001) analyzed AVHRR-re-trieved cloud and aerosol microphysicalparametersoverthe global ocean. They found that the cloud particleeffective radius and the cloud optical thickness havenegative and positive correlations with the column aero-sol particle number, respectively, while the liquid watercontent is approximately constant and independent of the aerosol particle number on a global scale.There were studies relating the satellite-derivedcloudproperties with model-calculated aerosol amount. Cha-meides et al. (2002) examined the correlation betweencloud optical depths derived from the International Sat-ellite Cloud Climatology Project (ISCCP) and aerosolburden calculated by a regional climate/chemical trans-port model over East Asia and found a similarity in thedistributions of model-calculated anthropogenic aero-sols and ISCCP-derived cloud optical depths. Schwartzet al. (2002) made an effort to detect the enhancementof cloud optical properties caused by aerosols by in-vestigating how the cloud optical depth or albedo de-pends on liquid water path using model-calculated sul-fate loading in mid–North Atlantic region.These researches have thrown a light on the mech-anism of the aerosol and cloud interaction phenomena,but to our knowledge there has been no full comparisonbetween satellite-derived aerosol–cloud interactionglobal statistics with the corresponding statistics sim-ulated by a global model implemented with the aerosol–cloud interaction process. In the present study, we re-investigated the results by Nakajima et al. (2001) bycomparing those observed aerosol and cloud parameterswith simulated values from the Center for Climate Sys-tem Research (CCSR)/National Institute for Environ-mental Studies (NIES) atmospheric general circulationmodel (AGCM) that has been implemented with a newparameterization of cloud–aerosol interaction.Themod-el used in this study is described in section 2, wherethe conditions and design of the simulation are also  15 J ANUARY  2004  181 S U Z U K I E T A L .T ABLE  1. Particle density and radius of sulfate and carbonaceous aerosols assumed for calculation of aerosol particle numberconcentration.Species Sulfate CarbonaceousOrigin Forest fire(tropical)Forest fire (other) Fossil fuel Fuel wood Agriculture TerpeneDensity (g cm  3 ) 1.769 1.473 1.468 1.442 1.462 1.468 1.5  Radius (  m) 0.07 0.1 presented. Satellite-derived data are introduced in sec-tion 3. Section 4 presents the simulated results to becompared with the satellite observation. Discussion andconclusions are given in sections 5 and 6, respectively. 2. Model simulation The GCM used in this study is the CCSR/NIESAGCM (Numaguti 1993; Numaguti et al. 1995). A. Nu-maguti (1999, personal communication) has updated themodel to introduce the aerosol indirect effect into thecondensation process representing nonconvectiveclouds whose temperature is above 273 K, whereas con-vective clouds including those whose top is lower than273 K isotherm level or ice clouds are simulatedwithoutthe aerosol effect. Detailed description of how the aero-sol indirect effect was incorporated into the model ispresented below.In the present model, the number concentration of aerosol particles acting as CCN at each model level, n a (  z ), is used for calculating the cloud droplet numberconcentration  n c (  z ) at the corresponding level. Theglob-al distribution of   n a (  z ) used in this study was calculatedby the global three-dimensional aerosol transport modelSpectral Radiation-Transport Model for AerosolSpecies(SPRINTARS), developed by Takemura et al. (2000,2002), which simultaneously treats all the major aerosolspecies, that is, sulfate, carbonaceous, sea salt, and min-eral dust aerosols. This aerosol model has been imple-mented in the CCSR/NIES AGCM. The aerosol numberconcentration  n a (  z ) is given as the sum of contributionsfrom sulfate, carbonaceous and sea salt aerosols: n  (  z )    n  (  z )    n  (  z )    n  (  z ). a  sulfate carbon seasalt Mineral dust is excluded from the evaluation of   n a (  z )because dust itself is insoluble and regarded as ineffec-tive as CCN, although mixture with other aerosol spe-cies such as sulfate or carbonaceous may provide CCN.Simplified treatment of dust aerosol can be a source of error for inferring the cloud particle radius since dustaerosol is abundant in mass concentration in the at-mosphere. The number concentration of sulfate aerosoltype  n sulfate (  z ) is evaluated as   (  z ) q  (  z ) sf  n  (  z )    , sulfate3    4   r   /3 sf sf  where    (  z ) denotes the air density, and  q sf  (  z ),    sf  , and r  sf   the mass mixing ratio, particle density, and moderadius of sulfate aerosol, respectively.Here q sf  (  z )iseval-uated by the aerosol transport model with the values of    sf   and  r  sf   presented in Table 1. The value of mode radiusis that of dry particle, which is same value as those usedin the aerosol transport model (Takemura et al. 2002,Table 4). Carbonaceous aerosols are treated as an in-ternal mixture of black carbon (BC) and organic carbon(OC) with the ratio of OC to BC (OC/BC) dependingon the srcin of the carbonaceous aerosols (Takemuraet al. 2000, 2002), and hence the particle density is alsosomewhat different for each srcin as presented in Table1. Then  n carbon (  z ) is given by the sum of the contributionsfrom various srcins:   (  z ) q  (  z ) cb, i n  (  z )    ,  carbon3    4   r   /3 i  cb, i  cb, i where the suffix  i  identifies the srcin of carbonaceousaerosols. The mass mixing ratio  q cb, i (  z ) is evaluated bythe aerosol transport model. The values of the particledensity    cb, i  and the mode radius  r  cb, i  are also presentedin Table 1. Mode radius is a dry particle radius, whichis same value as used in srcinal aerosol transportmodel(Takemura et al. 2002, Table 4). The number concen-tration of sea salt aerosol  n seasalt (  z ) is calculatedsimilarlyto the above two species, but as a function of windvelocity over the ocean surface based on the empiricalformula of Erickson et al. (1986) with a mode radiusestimated by the relationship suggested by Erickson andDuce (1988) (Takemura et al. 2000). The global distri-bution of   n a (  z ) is calculated as above for each time step,assuming the background minimum value of 3    10 6 m  3 . In the present study, the aerosol distribution wassimulated by SPRINTARS, an aerosol transport modelimplemented to CCSR/NIES AGCM with the standardcloud process parameterization without aerosol indirecteffect. In this calculation with SPRINTARS, the wetdeposition was driven by cloud and precipitation basedon the parameterization of Sundqvist (1978) withoutaerosol indirect effect and thus not affected by aerosollifetime effect. This simplification will not produce asignificant error in our results, because the CCSR/NIESAGCM standard version has been validated to producerealistic cloud and precipitation fields that are furtherused in the simulation of aerosol distribution by aerosoltransport model (SPRINTARS). In this simulation of aerosol distribution, we further applied a nudging tech-nique with National Centers for Environmental Predic-tion–National Center for Atmospheric Research  182  V OLUME  61J O U R N A L O F T H E A T M O S P H E R I C S C I E N C E S (NCEP–NCAR) reanalysis data to secure a realisticsim-ulation of the meteorological fields such as wind, tem-perature, and humidity, which are used for simulatingthe global aerosol distribution properly for our targetyear of 1990.In order to calculate the cloud microphysicalstructurechange due to the aerosol indirect effect, the CCSR/ NIES AGCM is rerun with the aerosol indirect effectparameterization using the simulated value of   n a (  z ) asexplained above. This second run with indirect effectis performed without nudging technique to investigatehow aerosols affect the cloud field in GCM calculationwith various atmospheric processes. The cloud dropletnumber concentration  n c (  z ) is determined diagnosticallyby the parameterization with the empirical relationshipdeveloped by A. Numaguti (1999, personal communi-cation):  n  (  z ) n a m n  (  z )    , (1) c  n  (  z )    n a m where  and  n m  are prefixed constants. According to thisformula,  n c  is proportional to  n a  when  n a  is small com-pared with  n m , whereas  n c  approaches the saturationvalue  n m  with increasing  n a  when  n a  is comparable orsuperior to  n m . The range of small  n a  corresponds to aclean condition (e.g., over the ocean) and the range of large  n a  corresponds to a polluted condition (e.g., overthe continent). The values of    1 and  n m  4  10 8 m  3 are used to approximate the relationship observedby Martin et al. (1994).The CCSR/NIES AGCM classifies the total waterintoliquid water and water vapor assuming a subgrid scaledistribution of the total water following the scheme of Le Treut and Li (1991). Based on the liquid water con-tent  l  determined in this way, the rate of rain production P  is calculated as dl lP      , dt      p where     p  is the relaxation time constant, which repre-sents the time scale of the autoconversion rate  l  /     p . Thetime constant     p  has been classically expressed by thebulk parameterization of cloud physics in several ways.One is the classical formula given by Kessler (1969):   0         ( l )    ,  p p 1    l  /  l c or removing the threshold effect (Sundqvist 1978):   0         ( l )    . (2)  p p 2 1    exp[  ( l  /  l  ) ] c CCSR/NIES AGCM uses this formula (2) in its standardversion. According to Sundqvist’s formula in Eq. (2),the time constant     p  depends only on the cloud watercontent  l  in such a way that an increase in  l  shortensthe cloud lifetime through decreasing     p . Since     p  doesnot depend on the cloud particle size in Eq. (2), theaerosol lifetime effect is not included in this parame-terization. The present study performs a simulation withthis formula for comparison with other types of param-eterization presented below. Values of the constants ap-pearing in Eq. (2) are set as  l c    10  4 kg kg  1 and    0   10 4 s.There are other types of parameterization that incor-porate the effect of cloud droplet number concentration n c  as well as cloud liquid water content  l.  The classicalexample for such kind of parameterization is the formulagiven by Berry (1967), which includes a CCN effect, as n c         l         ( l, n  )    , (3)  p p c   l where     is the air density and    ,     ,  and     are constants.Berry’s formula, contrary to that of Kessler or Sund-qvist, takes into account the dependency of the con-version rate upon the cloud droplet size as well as theliquid water content and thus includes the lifetime effectof cloud particles due to increasing aerosols. Accordingto this formula, more abundant  n c  reduces the dropletmass    l  /  n c , leading to a longer     p . We use Eq. (3) with     0.35,       0.12, and       5.7    10  12 followingLohmann and Feichter (1997).Another example of parameterization including theeffect of cloud droplet number concentration  n c  is theformula suggested by Khairoutdinov and Kogan (2000).They assumed the form,  n c         ( l, n  )    , (4)  p p c  l with nondimensional parameters    1.79 and     1.47determined from many numerical experiments with a dropspectrum resolving microphysical model. The conversionrate of cloud water into rain water is evaluated with eitherparameterization, Eqs. (2), (3), or (4) in the present study,to derive the cloud water content  l (  z ) remaining after pre-cipitation. Based on the cloud droplet number concentra-tion  n c (  z ) and the cloud water content  l (  z ) thus calculated,the cloud droplet radius is calculated as 1/3 3    l (  z ) r   (  z )    , (5) c [ ] 4     n  (  z ) w c where    w  is the liquid water density.Cloud particle effective radius  r  e (  z ) can be empiri-cally related to  r  c (  z ) as  1/3 r   (  z )    k r   (  z ), e c where the parameter  k   depends on the cloud droplet sizedistribution function. According to Martin et al. (1994),the value of   k   is 0.67  0.07 and 0.80  0.07 over theland and ocean, respectively; thus  k   1/3 is equal to 1.14(land) and 1.07 (ocean). We choose 1.1 for the value of  k   1/3 commonly over land and ocean. Using the liquidwater content  l  and the effective radius  r  e , the liquid  15 J ANUARY  2004  183 S U Z U K I E T A L . water path (LWP) and the cloud optical thickness    c  arecalculated by their definitions: top LWP      (  z ) l (  z )  dz  and  bottomtop 3 1    (  z ) l (  z )      dz. c   2     r   (  z ) w e bottom Based on the model presented above, we performeda numerical simulation of the global cloud–aerosolchar-acteristics with horizontal resolution of about 5.6   5.6  (T21) and with 11 levels (    levels at 0.995, 0.980,0.950, 0.900, 0.815, 0.679, 0.513, 0.348, 0.203, 0.092,and 0.021). Time integration was performed for 13months from December 1989 to December 1990 with atime step of 30 min and with the initial condition of temperature and wind fields taken from NCEP–NCARreanalysis data of 1 December 1989. The simulated re-sults for one year, 1990, are adopted for comparisonwith satellite observation without those for first month(December 1989) considering the spinup of the model.The year 1990 was selected for the present study tocompare the simulation result with the satellite remotesensing results retrieved in the year of 1990 as describedin the next section. 3. Satellite data In the present study, the model-simulated results arecompared with the aerosol and cloud microphysical pa-rameters derived from satellite remote sensing. Thecloud optical thickness    c  and particle effective radius r  e  in 1990 were derived by the method of Kawamotoet al. (2001) for water clouds whose top temperature ishigher than 273 K. Kawamoto et al. (2001), which isan extension of Nakajima and Nakajima (1995) to aglobal algorithm, developed the algorithm for globaldetermination of the water cloud optical thickness andthe particle effective radius from reflected solar spectralradiances in channels 1 and 3 of AVHRR, with an al-gorithm similar to that of Nakajima and King (1990),after removing thermal radiation emission usingchannel3 and 4 radiances. Retrieved  r  e  and    c  by Kawamoto etal. (2001) are further used to calculate the cloud liquidwater path as2LWP        r   . (6) w c e 3Also used in this study are the aerosol optical thick-ness    a  and A˚ngstro¨m exponent    over the global oceanin 1990 also retrieved from AVHRR by the algorithmof Nakajima and Higurashi (1998) and Higurashi andNakajima (1999), which are the first global satelliteanalysis of the A˚ngstro¨m parameters utilizing reflectedsolar spectral radiances in the red and near-infraredchannels (channels 1 and 2) of AVHRR. This two-chan-nel algorithm retrieves the peak volume values of theassumed bimodal size distribution of aerosol particlesand then evaluates the aerosol optical properties    a  at areference wavelength of 500 nm and A˚ngstro¨m expo-nent    ,  the mean wavelength index of the optical thick-ness spectrum over 369 to 1050 nm. Since the aerosoloptical thickness represents the vertical total amount of the aerosol cross section and the A˚ngstro¨m exponentprovides information on the aerosol particle size, anestimate of the column particle number  N  a  is acquiredfrom the approximation proposed by Nakajima et al.(2001). 4. Comparison of GCM simulation with satelliteobservation a. Global distribution of cloud particle radius Figure 1 shows the global map of AVHRR-retrieved(top) and GCM-simulated cloud particle effective radiiwith Berry’s (middle) and Khairoutdinov’s (bottom) pa-rameterization for a 4-month (January, April, July, andOctober) mean condition in 1990. Since the particleradius was mainly determined from the near-infraredradiance of AVHRR reflected by the cloud particle nearthe cloud top, AVHRR-derived cloud particle effectiveradii represent those near the cloud top. To match theAVHRR retrieval, the cloud top height is also deter-mined in the GCM simulation by finding the highestlevel where the liquid water content is larger than acritical value and the temperature is higher than 273 Kto avoid ice particles for which the remote sensing al-gorithm cannot retrieve a precise value of effective par-ticle radius due to nonspherical scattering by irregularice particles. Shown in the middle and bottom of Fig.1 is the global distribution of simulated cloud particleeffective radii at the cloud top level determined in sucha way. The satellite and model values are found to agreewith each other regarding the land–ocean contrast. Par-ticle radii over the continent are systematically smallerthan those over the ocean as also revealed by previoussatellite retrievals such as those of Han et al. (1994) andKawamoto et al. (2001). The simulated particle radiusis significantly reduced even over some coastal oceanregions such as the east coasts of ChinaandNorthAmer-ica, the Persian Gulf, the west coast of Africa, and anarea off California, as also found by satellite remotesensing. On the other hand, a large effective radius issimulated by GCM over the Northern Hemispheric At-lantic Ocean of mid to high latitudes, where satellite-retrieved values are also large. The zonal belt of largedroplet radii over mid to high latitudes in the SouthernHemisphere detected by AVHRR is also found in theGCM calculation. GCM also produced large particleradii over the Amazon basin and equatorial Africa con-sistently with satellite observation. GCM-simulated re-sults with Berry’s and Khairoutdinov’s parameteriza-tions, shown in the middle and bottom of Fig. 1, re-spectively, depict the common characteristics such as
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks