JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 102, NO. D3, PAGES 3805-3818, FEBRUARY 20, 1997
Modeling sea-salt aerosols in the atmosphere 1. Model development S. L. Gong and L. A. Bartie AtmosphericEnvironment Service, Downsview, Ontario, Canada
J.-P. Blanchet Earth SciencesDepartment, University of Quebec at Montreal, Canada
Abstract. A simulationof the processesof sea-saltaerosol generation, diffusive transport, transformation,and removal as a function of particle size is incorporatedinto a one- dimensionalversion of the Canadiangeneral climate model (GCMII). This model was then run in the North Atlantic betweenIceland and Ireland during the period of January- March. Model predictionsare comparedto observationsof sea-saltaerosols selected from a review of availablestudies that were subjectedto strict screeningcriteria to ensure their representativeness.The number and masssize distributionand the wind dependencyof total sea-saltaerosol mass concentrations predicted by the model comparewell with observations.The modeled dependenceof sea-saltaerosol concentration in the surface layer(X,/xg m -3) on 10-mwind speed (U•0, m s-•) is givenby X = beaU•ø.Simulations showthat both a and b changewith location.The value a and b range from 0.20 and 3.1 for Mace Head, Ireland to 0.26, and 1.4 for Heimaey, Iceland. The dependenceof X on surfacewind speedis weaker for smallerparticles and for particlesat higher altitudes.The residencetime of sea-saltaerosols in the first atmosphericlayer (0-166 m) rangesfrom 30 min for large particles(r = 4-8/xm) to -60 hoursfor smallparticles (r = 0.13-0.25/xm). Although somerefinements are required for the model, it forms the basisfor comparing the simulationswith long-term atmosphericsea-salt measurements made at marine baselineobservatories around the world and for a more comprehensivethree-dimensional modelingof atmosphericsea-salt aerosols.
1. Introduction can only be performed at certain times and locations,applica- tion of the experimentalresults is limited. It is vital to develop Sea-salt aerosolsplay a very important role in a variety of an aerosol transport model to predict its concentration and processesin the atmosphere.They influenceradiative transfer size distribution on a regional scale as a function of atmo- directlyby scatteringsolar radiation and indirectlyby altering sphericprocesses. Various modelshave been proposedin the cloud droplet size distributionand concentration,thereby in- literature [Gathman, 1982; Fairall and Davidson, 1986; Erick- fluencingthe albedo of marine boundary layer clouds.In ad- son and Duce, 1988; Fitzgerald,1992; van Eijk et al., 1992]. A dition, sea-saltaerosol particles are chemicalcarriers of spe- summaryof major functionsand assumptionsin these models ciescontaining C1, Br, I, and S and therefore play a role in the is shown in Table 1. All of these models are limited to the atmosphericcycles of these important elements.The halogens marine boundary layer and do not addressthe long-range Br and C1, once mobilized by heterogeneousreactions from transport of sea salt. sea-saltinorganic forms to reactive gaseousforms (e.g., Br2, A comprehensivesea-salt aerosol model coupledwith a one- C12)[e.g., Mozurkiewcz, 1995], can play a role in atmospheric dimensionalclimate model (FIZ-C) [Therrien,1993] is pre- ozonedepletion and destructionof light hydrocarbons[Jobson sentedin this paper. Using the meteorologicalconditions gen- et al., 1994]. erated by the FIZ-C, the model includes the following Numerous experimental investigationsof sea-salt aerosols processes:(1) sea-saltgeneration due to surfacewind; (2) have been conductedaround the globe in order to determine vertical transportby turbulenceand convection;(3) dry depo- their abundance and physical/chemicalproperties [Jacobs, sitionand gravitationalsettling; and (4) wet removalprocesses 1937; Woodcock,1953; Toba, 1965a, b; Lovett, 1978; Prospero, which includeboth in-cloudand below-cloudscavenging. The 1979; Podzimek,1980; Parungoet al., 1986; Marks, 1990; Ike- model doesnot at this time includechemical or physicaltrans- gami et al., 1994;McGovern et al., 1994]. Sea-saltaerosol con- formationthrough interactions with other aerosoltypes such as centrationis a strongfunction of the state of the sea surface, acidic sulphur. The particle size distribution is modeled by which is in turn determinedby meteorologicalconditions, es- representingthe sizespectrum as a seriesof discretesize bins. pecially surfacewind speed.Sea-salt particles are distributed from 0.02 to 60 txm with a bimodal size distribution in the 2. Physical Model submicronportion [Fitzgerald,1991]. Since the observations The fate of sea-saltaerosols, once they are injected into the Copyright1997 by the American GeophysicalUnion. atmospherefrom the ocean source,is governedby a seriesof Paper number 96JD02953. physicalprocesses such as transport,coagulation, dry and wet 0148-0227/97/96JD-02953509.00 removal,and chemicaltransformation. Transport of aerosolsis
3805 3806 GONG ET AL.: ATMOSPHERIC SEA-SALT MODELING, 1
Table 1. Summaryof Sea-SaltAerosol Models
Removal Size Authors Dimension Domain Spectrum Transport Source Dry Wet
Gathman [1982] ' one MBL no eddy Blanchard[1963] D • BC* IC* Fairall and Davidson[1986] one MBL yes eddy Fairall and Davidson[1983] BC? G? IC? Burk [1986] one MBL no eddy Monahahet al. [1986] D? BC* G? IC* Stramska[1987] one MSL yes eddy Monahah et al. [1986] D* BC* G? IC* Ericksonand Duce [1988] two surface no eddy empirical D? BC? G? IC* FitzgeraM[1992] zero MBL yes eddy Monahah et al. [1986] D* BC? G* IC? van Eijk et al. [1992] zero MBL yes eddy empirical D* BC* G* IC*
MBL, marine boundarylayer; D, dry deposition;G, gravitationalfall; BC, below-cloudscavenging; IC, in-cloudscavenging. *Not included. ?Included.
realized by turbulent diffusion,advection, and vertical convec- In (2), the aerosolconcentration change has been dividedinto tion of the atmosphere,which is representedin this model by tendenciesfor dynamics,surface, clear air, dry deposition,in- the FIZ-C parameterizationswhich are analogousto thosefor cloud and below-cloudprocesses. The dynamicsincludes re- moisture transport in the FIZ-C. A generalizedprognostic solvedmotion as well as subgridturbulent diffusionand con- massbalance equation for size i of typej aerosolparticle can vection.The surfaceprocesses include surface emission rate of be written as both natural and anthropogenicaerosols and serveas bound- ary conditionsfor the model. Particle nucleation,coagulation, Opxu + divpxuU = Su + I u (1) and chemicaltransformation are includedin clear-airprocess. Ot In the current version,coagulation is not includedfor sea-salt whereX'q is the mass mixing ratio of thei th sizerange aerosols aerosols.However, for purposesof predictingtotal mass of (kg/kgair),U isthe horizontal wind velocity vector (m s-•), p is sea-saltaerosols for comparisonwith observations,this is not a theair density (kg m-3), and Sq is the source and sink terms major drawback, since Na and C1 which dominate sea-salt whichmay includefollowing contributions: (1) surfacesource aerosolmass are not affectedsubstantially by coagulation.The (natural and anthropogenic);(2) clear air processes(particle aerosolmodel is structuredin such a way that any process which affects the concentration and size distribution of sea-salt nucleation, particle coagulation, and chemical transforma- aerosols can be modified or added to the model as new or tion); (3) in-cloud processes(activation of aerosols,attach- ment to clouds, and removal of aerosol-attachedclouds by better parameterizationsfor suchprocesses become available. hydrometeors);(4) dry deposition;and (5) precipitationscav- Somedetailed physical parameterizations for sea-saltaerosols enging,where particle nucleation,and chemicaltransforma- will be presentedin the followingsections. tion are processesnot requiredfor sea-saltaerosols. Term I u representsthe rate of intersectionaltransfer processwhich 2.1. Aerosol Growth With Relative Humidity moves the aerosol mass from one size bin to another. The Experimental evidence [Fitzgerald,1991] has shown that formation of new aerosol particles through coagulationor there exists a size distribution of sea-salt aerosols of radii breakingof the existingparticles is one intersectionaltransfer rangingfrom 0.02 to 60 •m. A knowledgeof size distribution process.Since only dry massof aerosolsis entered into the is necessaryin order to calculatethe effect of the aerosol on massbalance equation (1), the condensation/evaporationof climate as well as on chemicalreactions in the atmosphere. water on sea-saltaerosol particles will not contribute to the In this model, the aerosolsize distributionis simulatedby a intersectionaltransfer rate. However, condensation/evapora- method similar to that developedby Gelbardand Seinfeld tion may contribute to intersectional transfer rate of other [1980].The whole sizerange of aerosolparticles is dividedinto typesof aerosols,for example,sulphate aerosols when sulphu- a number of sections(or size bins), and each size sectionis ric acid vapor condenseson the existingaerosols. representedby one prognosticequation (i) which is a mass In accordancewith the FIZ-C convention,the prognostic continuityequation averagedfor that size range. The calcula- equation is rewritten in the form of tion of sea-salt generation is carried out by integrating the production over each size bin while the physicalproperties 0-7-= at +-bT +7F suchas dry depositionand wet removalrates are calculatedby dynamics surface clear air dry usingthe averagesize in each size bin. It is noted that in the model, the prognosticequation is written for dry size only. OXu OXu Beingvery hygroscopic,sea-salt aerosols will changesize as the •n-cloud below-clouds ambientrelative humiditychanges. By assumingthat the par- GONG ET AL.' ATMOSPHERIC SEA-SALT MODELING, 1 3807 ticles are alwaysin stable equilibrium with the surrounding (b) relative humidity,the actual size of sea-saltaerosols is related to their dry sizeby the empiricalrelationship of Gerber[1985]'
r • C3rdC4Clr•__ log 2 S + r• ] 1/3 (3) wherer d is the dry particleradius (centimeters), S the satura- tion ratio (or relativehumidity), and C1, C2, C3, and C4 are parametersfor differenttypes of the aerosolparticles (Table 2). >4 ms" wind > 10 rn s" Equation (3) can be usedin a relative humidityranging from 0% to 100%. Physicaltendencies, for example,coagulation, dry deposi- tion, gravitationalsettling, or wet removal, are computedfor each dry size bin using the aerosol size in equilibrium with ambient water vapor. The relative humidity profile is calcu- lated from temperatureand moisturepredicted by the climate model. As the particlesgrow with relative humidity, the densityof the particlesis also altered by incorporatingmore water. The average densityof sea-saltaerosols in each size bin is calcu- Figure 1. Mechanisms for sea-salt aerosol generations lated based on the size differencebetween the dry and wet (adaptedfrom Monahah et al. [1986]). Two mechanismsare particles.The radius of the particlesin equilibriumwith am- presented(a) indirect productionby bubblesand (b) direct bient moistureis estimatedby (3). productionby spumes.
2.2. Generation of Sea-Salt Aerosols The generation of sea-saltaerosols has been attributed to variousphysical processes [Blanchard, 1983]. The most prom- inent mechanism is believed to be the entrained air bubbles Toba and Chean, 1973] and wave tank measurementsof the burstingduring whitecap formationsdue to the surfacewind sea-saltaerosol productionrate per unit area of whitecaps, [Monahah et al., 1986]. Precipitation(both rain and snow) Monahah et al. [1986] have substantiallyrefined the formula- contributesto the productionbut only on an intermittent and tion for the sea-saltaerosol generation and includedboth in- local scale.A summaryof various mechanismsare shown in direct (bubbles)and direct (spume)mechanisms in the model. Figure 1. The densityfunction dF/dr (particlesm -2 s--• /•m-•), which Physically,the production of sea-saltaerosols by wind is expressesthe rate of sea-saltdroplet generationper unit area proportionalto the whitecapcoverage. Continuous supply of of seasurface, per incrementof dropletradius, is given(1) for excessiveenergy by the wind to the sea surfaceresults in wave indirectmechanism (through bubbles): breaking. Wu [1986] developed a relationship between the dFo whitecap coverageand wind speed by consideringthat the --=dy 1.373U264•r-3(1+ 0.057rl'øs) X 10 (Sa) energylost by wave breakingis balancedby the energygained from the wind: where B = (0.380 - log r)/0.650 and (2) for direct mech- anism (throughspume)' E: ß = (CoVo)(C •0 ,•,0) - (4) 0 r < 10/zm where W is the fraction of the sea surfacewith whitecap cov- erage,E isthe energy supply rate by the wind to a unitarea of dF• 8.60x 10-6e2'øSUlør-2 10/zm_< r -< 75 (5b) the sea surface, r is the wind stress,V is the sea surface drift dr 4.83 x 10-2e2'øsU•"r-4 75/zm-
Aerosol A total flux of sea-saltaerosols for each sizebin is obtainedby Model C 1 C 2 C 3 C 4 integratingthis formula over the size range of this bin. This sea-saltgeneration model is of essentialimportance in setting Sea salt 0.7674 3.079 2.573 x 10-1• - 1.424 Urban 0.3926 3.101 4.190 x 10- • - 1.404 up the lower boundary conditionsfor our sea-salt aerosol Rural 0.2789 3.115 5.415 x 10 -• -1.399 model. It is noted that the applicablerange of radiusfor (5a) (NH4)2504 0.4809 3.082 3.110X 10-• -1.428 is 0.8-10/zm (at 80% relativehumidity (RH)) withinwhich the empiricalrelationships are based.It may introduce some un- Reproducedfrom Table 1 of Gerber [1985]. The temperature de- pendenceof (3) is found primarily in C 3 due to the sensitivityof the certaintiesif it is usedfor the generationof submicronsea-salt Kelvin effectto temperature;C3 can be temperaturecorrected with the aerosols.For particleslarger than 10/zm, (5b) shouldbe used expression:C•(T) : C311 + 0.004(298 - T)], T in kelvin. with (5a). However,it generatestoo muchbig sea-saltparticles 3808 GONG ET AL.: ATMOSPHERIC SEA-SALT MODELING, 1
at high wind speedscompared to observationsand is therefore 2.3. Wet Removal of the Marine Aerosols neglectedin the simulation. For modeling atmosphericaerosols, the inclusion of wet Another mechanismof marine aerosol generation, which removalprocess, that is, the scavengingof particlesin the air by has not receivedmuch attention during the last 3 decades,is precipitation and by clouds, is essential.This self-cleaning the productionof sea-saltaerosols by wet precipitation.This mechanism maintains the balance between the sources and mechanism is mentioned here for two reasons. First, since it is sinksof atmosphericaerosol particles. In the presentmodel, well recognizedthat a major self-cleaningprocess of the atmo- both below-cloudand in-cloudscavenging processes are con- sphereis the removalof particlesby cloudsand precipitation, sideredwith a simplifiedapproach suitable for use in a large- any attempt to model aerosolparticles in the marine atmo- scaletransport model. spherewithout incorporatingwet removal mechanismis not 2.3.1. Below-cloudscavenging. Below-cloud scavenging realistic.Second, in the case of wet precipitation,if only the as referred to here is the processof aerosolremoval from the removal processis considered,which is the case for all the atmospherebetween cloud-base and the groundby precipita- presentmarine aerosolmodels, aerosol concentrations may be tion (i.e., rain or snow).The captureof aerosolparticles by underestimated,especially near the sea surface. falling hydrometeorstakes place by BrownJanand turbulent Unfortunately,literature on this subjectis very limited. Only shear diffusion,inertial impaction,diffusiophoresis, thermo- two papershave been found that deal exclusivelywith it [Blan- phoresis,and electricaleffects. Detailed studiesof thesepro- chard and Woodcock,1957; Marks, 1990]. In the first paper, cesses[GreenfieM, 1957; Pilat, 1975; Wanget al., 1978; Slinn, Blanchardand Woodcock[1957] suggesttwo mechanismsof 1984;Herbert and Beheng,1986] have revealed that there exists marine aerosolgeneration by raindrops:(1) directproduction a minimum in the collectionefficiency of aerosolparticles through the impact of the rain drops and (2) from bubbles betweensizes ranging from 0.5 to 1 txmradius, which is some- producedin the surfacewater as a result of impact. This pro- times called the "Greenfieldgap." Accordingto Slinn [1984], cessoccurs when a rain drop hits the sea surface,it generates the removal rate of aerosolsper unit volume can be written as not only sea-saltaerosols and bubblesbut alsosecondary drops due to the splash.The secondarydrops fall back into the sea L (7) and generatemore bubbles.It is believedthat in both casesthe where r i is the averagedactual radius of ith sizebin and ½is the productionof bubblesis a functionof droplet size [Blanchard scavengingrate which is computedby integrationof the fol- and Woodcock,1957]. The generationof sea-saltaerosol par- lowing equationover all hydrometeorsizes: ticles follow the burstingof bubbles. Marks [1990] carried out an experimentalinvestigation of the marine aerosolsunder both dry and wet (rain) conditions ½(r•)= E(r•, a)Ax[Vt(a) - Vg(r•)]N(a)da from the Dutch researchplatform Noordwijk,9 km offshorein the North Sea. Total sea-salt concentrations were measured at three heightsof 4.5, 12, and 18.3 m and sea-saltsize distribu- wherea is the dropradius, Vt and Vg are the gravitational tions at 12-m elevation.Compared to dry conditions,a reduc- settlingvelocity of hydrometeorsand particles,E (ri, a) is the tion in the aerosolconcentration at 4.5 m by rain was found at collectionefficiency of aerosolsof sizer i by hydrometeorsof windspeeds of 8 and 10 m s-•. In otherwords, the net effect sizea, andN(a) andAx are the hydrometeornumber density of rain on the total aerosolconcentrations at thesewind speeds and effectivecross-sectional area. However, the requirementof is negative,an indication that the wet removal processsur- N(a) and large computertime for the integrationmakes it passesthe generationof sea-saltaerosols by rain. When the unrealisticto use in a large-scaleaerosol model. An approxi- wind speedat 4.5 m reached15 m s-•, the sea-saltaerosol mate expression[Slinn, 1984] for the rain scavengingrate is concentrationduring rainy weather becamegreater than under used: dry conditions.This meansrain generatesmore sea-saltaero- cpE(r•, gin) sols than it scavengesprovided that wind-mediatedsea-salt production remains unchangedduring dry or wet conditions. ½(lrt)= grn (8) One interestingpoint from the experimentwas that for 12- and wherec is a numericalfactor of order unity (0.5) and the mean 18.3-melevations at a windspeed of approximately15 m s-•, collisionefficiency E(ri, Rm) is evaluatedby using a mean the sea-saltaerosol concentrations were not influencedby rain, drop size. The precipitationrate p is also used to obtain the meaningthat a balancebetween production and washoutof mean drop radius' sea-saltaerosols by rain occurredunder this condition. If this is a true characteristicof the marine boundary layer sea-salt aerosols,it might be used to parameterizeprocesses of rain- Rm--' 0.3 5 mm 1 mm/h (9) mediated generationof sea-saltaerosol. In this respect,wet scavengingmechanism which have been studied extensively For snowscavenging, [Wanget al., 1978;Herbert and Beheng,1986] would play an Tp• (r,, A) important role. qt(rt)= Dm (10) Since the rain intensitywas not reported in the paper, a quantitativerelationship between the rain-mediatedproduc- where X is the characteristiccapture length, D m is the charac- tion and scavengingand wind speedcan not be inferred from teristiclength, and T is a constantof order unity (0.6). De- the paper. Therefore this mechanismof sea-saltgeneration is pendingon temperature,X and D m may have differentvalues neglectedin our model. One justificationfor thiswould be that [Slinn,1984] because of differenttype of snows(Table 3). in comparisonto wind-generatedsea-salt aerosols, this source 2.3.2. In-cloud processes. In-cloud scavengingprocesses is likely a minor one, especiallyconsidering the frequencyof are reviewedby Bartie [1991]. Scavengingof chemicalpollut- the precipitation. ants in cloudsby three phasesof water was discussedin that GONG lET AL.' ATMOSPHERIC SEA-SALT MODELING, 1 $809 paper. A three-processmodel was developedto describethe Table 3. Parametersfor SnowScavenging in-cloudprocesses of the aerosolparticles. The first processis the activation of sea-salt aerosols, that is, served as cloud CrystalType Din, cm condensationnuclei (CCN), dependingon the critical super- Graupelparticles 1.4 X 10-2 saturationand particledry size.The mechanismof thisprocess Rimedplates and stellardendrites 2.7 x 10-3 has been studied in detail by cloud physicists[e.g., Twomey, Powdersnow and spatialdendrites 1.0 x 10-3 Plane dendrites 3.8 X 10 -4 1977] and is representedby the so-calledK6hler curves.For Needles 3.8 x 10 -3 sea salt (NaC1) and sulphate((NH4)2504) , the relationship between the critical radius (rc) of dry soluble particle and CrystalType X,/am a supersaturation(So in percent) is given approximatelyby sleet, graupel 1000 2/3 [Twomey,1977] rimed crystals 100 2/3 powder snow 50 2/3 Sea salt dendrites 10 1 tissuepaper 50 rc = 1.16 x 10-6 S•-2/3 camera film 1000
Sulphate
r•.= 1.40 x 10-6 S•-2/3 d Xia dt :--(OttT q- Ai)Xta -- •/iaXia which simplydescribes the minimum dry radiuswhich would (•3) enablenucleation to occurupon it at a supersaturationSo. The d Xic activationprocess has two major impacts:(1) becauseof the --=dt (o•,r+ A,)X,.- 13,x,• larger size of CCN they will be subjectto a different removal schemefor particleswith sizegreater than the critical radiusr c where miais the interstitialconcentration of the aerosolparti- and hence the size distributionspectrum of sea-saltaerosols cles of i th size bin, Xic is the concentrationof the aerosolsin will be modifiedand (2) a link is establishedbetween aerosols cloud water, and •/ia is the rate of direct aerosolremoval by and their indirect effect on climate for the activation will alter large hydrometeorswhich can usuallybe ignoredwhen com- pared to OliTand [3i. The activationrate A i is determinedby the cloud number density,liquid water content, and possibly cloud albedo. the critical radius of aerosolparticles. The total aerosolcon- centration,Xit = Xia q- Xic, is obtainedby solving(13) for The secondin-cloud process is the attachmentof the aerosol subjectto xitl•=o = ?rio. particlesin size range i to the existingcloud dropletsat a rate a,r. The attachment of aerosolsto the cloud droplets is as- sumedto be by two processes:Brownian and turbulent diffu- X,,=/3,- (c•,r + A,) [/3'e-("•+^')' - (0litq- A')e-•"] (14a) sion;that is, a,r = a,• + a•r•. The contributionby Brownian diffusionto a•r is estimatedby [Dingleand Lee, 1973] In the case of nonactivation,that is, r i < r c, A i = O,
mto --• [ [3ie - •"rt -- ottre - t3't] (14b ) OltB= kT• (--+l + Fclal/r, l + Ftal/rc,) (r, + r•)M• (11) Xit--[3 i __Otir and for the activationcase, that is, ri -> rc, if we assumeA i -- 0% where r i is the averagedaerosol particle radiusat sizebin i and r• is the cloudequivalent droplet radius; a is the Cunningham ,)('it-- Xioe-t3,t (14C) correctionfactor; I is the mean free path of air molecules;M c In deriving(14c), an implicit assumptionis made that oncean is the number density of cloud droplets; k is Boltzmann's aerosolparticle is activated,all the particleswith the same dry constant;T is the absolutetemperature; and • is the air vis- radius will form cloud droplets (Ai = o•), and they will be cosity.The turbulent contribution [Levich, 1962] is approxi- removed from the atmosphereif there is precipitation.The mated as removalrate by large hydrometeors,/37,is calculatedby setting /3i = ½(rc/) in (8). If no precipitationis experiencedduring the a,rB= 14.1(r, + r•) 3 M• (12) time step,the aerosolparticle remainsin the atmosphereand total concentrationdoes not change. 2.3.3. Collection etficiency(E). In order to calculate the where • is the rate of energydissipation and v is the kinematic scavengingrate of aerosolsor cloudsby both rain and snow viscosity.One approximationin formulating(11) and (12) is (equations(8) and (10)), the collectionefficiency of aerosol that the cloudsize spectra is replacedby an equivalentdroplet particlesby hydrometeors(E) is required.The collectionef- radius (rcl), which is estimatedby the cloud schemeof the ficiencyis definedas the ratio of the total number of collisions climate model. occurringbetween dropletsand particlesto the total number The final process involves the removal of the aerosol- of particlesin an area equal to the droplet's effectivecross- containingcloud dropletsproduced by the first two processes sectionalarea. Accordingto previousinvestigations [Slinn and by large falling hydrometeorsat a rate /3i [Dingle and Lee, Hales, 1971;Pilat, 1975; Wanget al., 1978;Herbert and Beheng, 1973;Slinn, 1984]. Assumingthat an activationrate is Ai for 1986],E is a result of the combinedaction of Brownian diffu- size range i and that a,r and /3i are constantwithin the inte- sion,inertial impaction,interception, thermophoresis and dif- gration interval, the in-cloud processescan be expressedby fusiophoresis,and electric forces. Detailed formulation of followingordinary differentialequations: theseprocesses in an atmosphericaerosol model is a prohibi- 3810 GONG ET AL.: ATMOSPHERIC SEA-SALT MODELING, 1
•h ture of the surfaceroughness are consideredonly in the inter- faciallayer which is from surfaceto height 8. Assumingsteady state,the dry depositionflux of aerosolparticles through the MeanWindSpeed,u Earth-atmosphere is expressedby Giorgi [1986]: SURFACELAYER Sea-salt concentration, Zi/ 4i)h --- l•'depFl h (17) KsL•turbulent transfervelocity ß ß ß Ka,drift velocity Frictionvelocity, u, h wherenh is the particle concentration at height h andl•'dep, the particledry depositionvelocity, is givenby Giorgi [1986] as' •'dep= Kdep+ gdh Bounce-offfactor, f•o I
where the definitionof Kdh is given in Figure 2 with Ksœ(KizfBo+ Kda- gdh) / INTERFACIAL LAYER Zo-surface roughness Kdep= KsL+ KizfBo + Kd• (18)
K,L- interfaciallayer transfervelocity K• - drift velocity where KsL and KiL are the transfer velocity in surfaceand interfacial layer, respectively,and are assumedto be of the Figure 2. Schematicillustration of two-layermodel used for followingforms [Giorgi,1986]' dry depositionof aerosolparticles on water surface.It is as- sumed a zero surface concentration and a bounce-off factor fBO. Ksz•= CoUhAp •.•DOr.•/2_ C•2 (19) Ki/_,= C•02u,G tively difficult undertaking.Neglecting phoresis and electric effects,Slinn [1984] derived the following semiempiricalex- where Co and Coo are the drag coefficientsrelative to h and pressionfor the particle/dropcollection efficiency: 8, Uh is the wind speedat h. Cr• is predictedin the climate model as the drag coefficientfor the first layer above the 4 surfaceand Coo is calculatedin the dry depositionmodule E = ReSc[1 + 0.4Re•/2Scm+ 0.16Re•/2Scm] accordingto the surfacetype, that is, Zo. The model distin- guishesbetween the followingsurfaces: water (oceanor lake), sea-ice, snow and vegetative canopies, and computes the + 4•f[to-• + (1+ 2Re•/2)•f]+ [St- St-S* S,+ 2/3]3/2 (15) roughnesslength Zo for them respectively.The frictionvelocity where Re = R mVt/p is the Reynoldsnumber of hydrometeors, u, is calculated as Sc = •,/D andSt = [r(Vt - Vg)]/Rm are the Schmidtand lg, '-- C•2/Uh Stokesnumbers of particles.The coefficientr is the character- istic relaxationtime of the particles,D is the diffusioncoeffi- G is a function accountingfor Brownian diffusion,intercep- cientof particles,Vt and Va are the terminalvelocity of hy- tion, and impactionof particleson the surfaceelements. Given drometeorsand particles, and R m and r• are the radius of the surfacetype, wind speed, air density,temperature at the hydrometeorsand particles,respectively. •15 = ri/Rm, to = ground level, and the surfacemomentum drag coefficient,the IXw/•',•,and/Xw and/x, are the viscosityof water and hydrom- dry depositionvelocity of aerosolparticles can be estimated. eteors. Slinn and Slinn [1981] deriveda similarexpression for dry 12/10 + (1/12)Ln(1 + Re) depositionof particleson a naturalwater surface.Hygroscopic S, = 1 +Ln(1 +Re) growth of particleswas taken into accountin the saturated viscoussublayer near the surface.As a first approximation,this For snowscavenging, the efficiencyis estimatedas effectwas neglected because of highrelative humidity (>80%) in open oceanmarine boundarylayer. E= •-• + 1-exp[-(1+Re}/2)]•j 2.5. Vertical Turbulent Transport St- S, Vertical transport of aerosol particles in the free tropo- + St- S, + 2/3 (16) spheredue to turbulenceis realized usingeddy diffusivityfor- mulationsin the free atmosphereand drag coefficientformu- Depending on the type of snowcrystals, a and X valuesfrom lations at the surface. This method is used in the Canadian Table 3 are usedin (16). climatemodel for the transportof momentum,heat, and mois- 2.4. Dry Deposition ture [McFarlaneet al., 1992].The eddydiffusivities for momen- tum and heat are of the form In modelingparticle dry deposition,a two-layerhypothesis is usuallyassumed [Slinn and Slinn, 1981;Giorgi, 1986]. Figure 2 (gm,gh) = 1210V/Ozl(fm,fh) (20) showsa typical two-layer model for the water surface.The transportof particlesin the top layer, that is, the surfacelayer where V is the horizontalvelocity vector and z the distance extendingbetween h and 8, is dominatedby turbulenceand abovethe local terrain. The Richardsonnumber (Ri) depen- gravitationalsettling, respectively. Viscous effects and the na- dent functions(fm, fh) have the form GONG ET AL.: ATMOSPHERIC SEA-SALT MODELING, 1 3811
(fm, • /10øW • - •0 Ril/(• + •0 Ri/87 1/2) Ri < 0 = (1-5eRi)2/(1 + 10(1- e)Ri) O<-Ri<- 1/5• (21) Greenla•/•/ ocati 0 Ri > 1/5• where I = kz/(1 + kz/X) wherek is the von K/trm/tnconstant and the asymptoticmixing / lo øNr length(X) is specifiedas 100m. Like moisturein the climate model,the aerosolparticles have a diffusivityequal to that for heat. The flux in the free atmosphereA p-r is written as Avr- - pKaOz (22) / whereK a -- K e = Kh; Xi and9 aredefined as in (1). The flux at and near the surface(A •r)x is expressedas Figure3. The geographiclocation for the test run in the North Atlantic. (Avr)s: -9• VL (XL+i/2- xg)Ca (23) whereCa = Ch;,•z• + •/: is thelowest prognostic level aerosol massmixing ratio. The dragcoefficient (C•) is relatedto the advectionof aerosolsis neglected. In otherwords, it is assumed surface-layerbulk Richardsonnumber (RIB) by following that thereis an infinitehomogeneous domain in the horizontal equations: and the solutionfor aerosolconcentration represents a quasi- Ct• = CDHFt• steadystate balance between local production and localre- moval. where A gridpoint from the GCMII in theNorth Atlantic (latitude 61.2ø north, longitude352.5 ø) was selectedfor the test run F•(RiB) (Figure3). Thesimulation was run for 90 daysstarting January 1 - 10 RiB/(1 + 10 RiB/(87A•)1/:), RIB<0 1. The aim was to evaluatethe performanceof the modelby comparingmodel predictions with availableexperimental ob- 0 --
0-166m ..... 166 - 722 rn measuredby Woodcock[1953], Extortet al. [1986], Gras and a) FourLayer Sea-salt --- 722- 1767rn Ayers[1983], and Marks [1990]by differenttechniques. Sea-salt --- 1767 - 3594 rn 103 ...... content was obtained by analyzing the sodium [Woodcock, 1953] and by using the sea-saltdensity and particle volume [Extortet al., 1986;Gras and Ayers, 1983]. The latter technique ,•'10' .',' I for obtainingsea-salt contents may be subjectto interference ••' 10ø' :': ",.':' ", \ll • "",' "';':: ø'-:':" "::':" :" ': • '•-:: ' '" I , by non sea salt in the particles.An important assumptionin this particle size to sea salt masscalculation is that the col- lected aerosolsample is closeto 100% marine origin. 10"j / '"' The locationof experimentcan stronglyaffect the represen- tativenessof observationsof open marine aerosolsmodeled ß .' here. Measurementsconducted on a shipover the ocean[Tsu- lO-4 nogaiet al., 1972;Lovett, 1978;Marks, 1990] are closerto ideal 10 -5 .... 10 20 30 40 50 60 70 80 90 marine aerosolsthan thoseperformed inland [Kulkamiet al., JulianDay 1982]or on a coast[Gras and Ayers, 1983; Extort et al., 1986]. b) WindSpeed & Precipitation The inland (1.8 km) observationby Kulkarni et al. was cer- 25 tainly affectedby coastalsurf as well as by dry depositional • 20 lossesand hence is not appropriatefor comparisonwith our '• 15 model. Even if the measurementwas made over open ocean, the particle size is not completelycharacteristic of sea-salt
•._ 10 aerosols.Anthropogenic sulphates and formation of DMS ox- E o 5 idation productscan contribute substantialinto non-sea-salt
0.0006 0 fine particles,especially during high biogenicactivity periods.
• 0.0005 Therefore the total sea-saltmass from particle size measure-
•E 0.0004 ments [Extortet al., 1986; Gras and Ayers,1983] may be sub- jected to some uncertainties.It also seemsthat the measure- .• 0.0003 ments by Gras and Ayers and Exton et al. may suffer from •. 0.0002 beach surf effects, sedimentation,and turbulent deposition o. 0.0001 lossesto the beach as their probes were several hundred 0.0000 10 20 30 40 50 60 70 80 90 meters away from the tide line. Julian Day Most of the measurements in Table 4 were made below Figure 4. Simulationresults for the test location.(a) Varia- 100m, exceptWoodcock [1953] whose data were taken at cloud tion of total sea-salt aerosol concentrations in the first four levels(-600 m). Accordingly,the observationsby Tsunogaiet model layersextending from the surfaceto 3594 m. (b) The al. [1972],Lovett [1978], and Marks [1990]are chosenfor com- wind speedand precipitationfor the sameperiod. parisonwith our modelpredictions for the surfacelayer (0-166 m), while that by Woodcock[1953] is usedfor comparisonto simulationin layer2 (167-722 m). The total sea-saltby sodium and Marks [1990] measuredthe total sea-saltmass as a func- measurementsand marine test locationsby theseinvestigators tion of wind speed.Their methodsinvolve the collectionof match closelyto the conditionsused in the simulation. marine particleson a filter and then analyzethe sodiumcon- tent on it. While Marks obtained the total sodiumby neutron 3.2. Size Distribution of Sea-Salt Aerosols activation method, the others measured only the soluble so- The particle radius rangingbetween 0.03 /am and 8/am is dium. The total Na measurement eliminates the artifacts dividedinto 8 sizebins accordingto Table 5. In this sizerange, causedby non-sea-saltparticles. Particle size distributionwas the main mechanismof sea-saltgeneration is attributed to the
Table 4. PublishedConstants for the RelationshipsBetween Airborne Sea-SaltConcentration X and Wind Speed U•o by Using Equation (25)
Maximum Method of Locationand Sampling Wind, a, b, Authors Latitude Measurements Height m s-• s m-• /•g m-3
Woodcock[1953] 20øN glassslide (Na and size) cloudbase over PacificOcean 35 0.16 2.57 (600 m) Tsunogaiet al. [1972] 40øN-50øN membranefilter (na) shipon PacificOcean 18 0.62 0.33 (12-14 m) Lovett [1978] 50øN-60øN membranefilter (na) weatherships in Atlantic Ocean 20 0.16 4.26 (5-15 m) Kulkarniet al. [1982] 19øN membranefilter (na) W. Indian coast1.8 km inland 8 0.27 5.35 Extonet al. [1986] 57øN ASAS and CSAS (size) Hebrideanbeach facing Atlantic 20 0.17 14.30 (15 m) Grasand Ayers [1983] 41øS roundjet collector(size) Cape Grim (94 m) 20 0.124 2.52 Marks [1990] N/A neutronactivation (impactor Na) shipon North Sea (12 m) 24 0.23 1.10 Marks [1990] N/A neutronactivation (filter Na) shipon North Sea (12 m) 24 0.23 1.13 Concentrationis in/•g m-3; windspeed is in metersper second. GONG ET AL.: ATMOSPHERIC SEA-SALT MODELING, 1 3813
Table 5. Aerosol Size Discretization Scheme indirect scheme,that is, via bubbles.The predictedmass size distributionof sea-saltaerosol (Figure 5a) is comparedwith Bin Size Range experimentalresults of Marks [1990] in the North Sea and 1 0.03-0.06 McGovern et al. [1994] at Mace Head on the west coast of 2 0.06-0.13 Ireland. Severalfeatures are revealedby the comparison.The 3 0.13-0.25 4 0.25-0.50 size distribution (1) of the sea-saltmass concentration com- 5 0.50-1 pares well within the observationaldata (2) in terms of the 6 1-2 shapesof the distributioncurves and the magnitudesof the 7 2-4 8 4-8 massconcentrations (Figure 5c). This agreementindicates that the parameterizationschemes adapted in the model are per- The radius is in micrometers. formed reasonablywell. It is also seen from Figure 5a that
a) PREDICTIONS b) MARKS OBSERVATIONS
10 2 -i .... i .... i .... t .... i .... i-
•... .. /r 22rn s-1 /
• 101 E 11ms
8ms '1 • 100 10 ø
10-1 10-1
i I .... I .... I .... I .... I .... I
0 1 2 3 4 5 0 1 2 3 4 5
Median Radius [Izm] Dry Radius [gm]
c) PREDICTION vs McGOVERN et al
10 2
101
• 100 •.._..______-••• WindSpeed I Prediction 10-11ms' // ---.--7ms '1s' 10-1 Prediction[--•--170r•1:1• • Obse•ation{•' •lm•-• i I = 14ms'1 10-2 , , , , , ,•,l , •[ , • • ,,[t • • , , • • 0.01 0.1 1 10
Radius [l•m]
Figure 5. The sea-saltaerosol size distribution of massconcentrations. (a) Model predictionsat three wind speeds;(b) experimentalobservations [Marks, 1990] for both dry and wet conditions;(c) comparisonwith McGovernet al. [1994]. The vertical dashedline in Figure 5c indicatesthe smallestexperimental radius at which (5) was based. 3814 GONG ET AL.' ATMOSPHERIC SEA-SALT MODELING, 1
a) PREDICTION b) SURFACEFLUX
109 _---109 108 • 108
• 107 • 07 • 106 • 106
• 105 19.ms'•q '•.•- 10 5 Q 19 ms-• • • 103 • 10• 8.9 m s'•
I• 4.710.8 mms '•s '• z 10• I 10• 10• I 10 o ...... i , ,I...... I ...... I 100 0.01 0.1 1 10 0.01 0.1 1 10 O0 DryRadius [lu.m], rd Dry Radius[!•n], rd c) Predictionsvs Observations
101o
10 9 •-'•_ .-•. • I Prediction E .--. 10 8 '-•" "',"•/ *• I at 19 ms"
o :• 107 ' • '._.•-•ß
d]ogr,-• 106 .m • 10 5 Prediction E lO4 atlOms"• I z I 10 3 ! '. I 102 I Observations ! 101 !
10 o ! 0.01 0.1 1 10 100
Radius [[tm] rd Figure 6. (a) As a functionof wind speed,the numbersize distribution of sea-saltaerosol particles predicted by the model. (b) The predictedsea-salt surface flux distribution(equation (5)) correspondingto resultsin Figure 6a. (c) Comparisonof observations[from Fitzgerald,1991, Figure 1] with predictionsat two wind speeds.The verticaldashed line in Figures6b and 6c indicatesthe lowestexperimental radius at which(5) was based. some detailed structures of the distribution at submicron tion. Coagulation would also strengthenthe concentration rangesare resolvedby the simulation.There is a hint of bi- near 0.1 /•m and thereby enhancethe bimodal structure.An- modal mass size distribution in the model results. This feature other effect is that other non-sea-salt aerosols also contribute is even more pronouncedwhen the predictionsare plotted as to the distribution measured in the remote marine environ- number size distribution(Figure 6). The bimodal distribution ment. However, in the radius range where sea-saltdominates from the model is not as clear as for the observations. This is (0.5 to 10/•m) the model and experimentalresults agree rea- likely due to lower size resolutionin the model than observa- sonablywell. A comparisonof Figures6a and 6b illustratethat tions as well as the fact that processesof cyclingthe aerosol the shapeof the number size distributioncurves is strongly through cloud formation and evaporationare not considered influencedby the sea surfaceflux size distributionof sea-salt in the presentone-dimensional model. They will, however,be dropletswhich is the only sourceof sea-saltaerosols. As seen taken into account in the three-dimensional model to be run from Figure 6b, the flux curve also does not show a distinct later. bimodal distribution. Furthermore,coagulation of aerosolsis ignoredin thisver- It shouldbe pointedout that the empiricalsea-salt produc- sion of the model. As a result, at fine particle size ranging tion functionby Monahahet al. [1986] (equation(Sa)) is based below 0.1 /•m, the predictionyielded a higher concentration on observationsof sea-saltgeneration for particleslarger than than the observation for both mass and number size distribu- about0.3/•m in radius(dry). This is the lowerlimit of particles GONG ET AL.: ATMOSPHERIC SEA-SALT MODELING, 1 3815 detected in the experimentswith which the production func- paper that the samplingtime was from 20 to 50 hours. During tion was comparedwell. Extrapolationbeyond the range re- the long experimentalperiod, the wind speedmight undergo sults in uncertaintiessince the empirical function begins to substantialchange. Because of the exponentialrelationship deviatefrom the experimentalobservations below 0.8 tzm ra- between sea salt and wind speed, the averagedsea-salt con- dius(80% RH). From this model simulation,use of Monahah centrationat the averagedwind speedcould be higher than an et al.'s [1986]production function seems to reasonablysimulate averaged sea-saltat the same averagedwind for a short time productiondown to a radiusof about 0.1 tzm (dry). However, whenthe wind speed would not change substantially. This is below 0.1 tzm, a large overestimateof both massand number clearly shownin Figure 8 where the wind speedis assumedto concentrationsoccurs (e.g., the predictioncurves on the left of undergo a sinusoidalchange from 0ø to 180ø with a speed of a dashedline in Figures5c and 6c). 0-15-0 m s-1 cyclefor a designatedtime andan exponential relationshipis assumedbetween sea salt and wind speed.For 3.3. Wind Speed Dependence of Sea-Salt Aerosols samplingthrough 0ø-180ø with every degree a sample (dura- tion 180), the root mean square (rms) wind speed for the The effect of wind speed on the sea-salt aerosol ambient durationis 10.6 m s-•, and the rms sea-saltconcentration is concentration(X) has been investigatedexperimentally by 29.6/•g m-3. For a shortersampling duration (51) whichre- many researchers[Woodcock, 1953; Lepple et al., 1983;Lovett, sultsin thesame rms wind speed (10.6 m s-•), therms sea-salt 1978; Gras and Ayers, 1983; Extort et al., 1986; Marks, 1990]. concentrationis reducedto 22.71 tzg m-3. When the wind Generally,the dependenceis expressedas a log-linearfit to the speedis constant at 10.6m s- • for theduration (180 or 51),the experimentaldata betweenconcentration and wind speed: sea-saltconcentration is 16.37m s-• whichis givenby the In X' = In (b) + auto (25) assumedexponential relationship. Furthermore, even if the windspeed only changes slightly in a 15-16-15m s-• cycle,the In [•'icr•rp '7h rntaclpl nrpctir, t•r•nc c•f textual rn•acc •arp r,c•rnn•arpct long .... •i•,,, ,4.... ,i,•, (!8{)) still ..... •t• in ...... tlrn•toct experimental observationsin the surface layer. A general rms sea-salt concentrationwhen compared with a sampling agreementfor the wind effectis obtainedbetween the predic- duration of 51 for the samerms wind speed.Therefore, if wind tionsand experiments(Figure 7b). Bestfitting curvesfrom the speed changes,the long-time average of sea-salt and wind model resultsare alsoshown in Figure 7a. A slope(a) 05 0.26 speedby Tsurlogaiet al. [1972] actuallyoverestimates the real and 0.13 and a correlation coe•cient of 0.82 and 0.23 are relationship between the sea salt and wind speed, that is, predictedfor total and fine particle (r, - 0.13 to 0.25 /•m) higher slope,compared to the samplingtime of 0.5 to 3 hours mass,respectively. Using a five-stageimpactor and filter sam- of Marks [1990] and 1-2 to severalhours of Lovett [1978]. pler to measurethe sea-saltaerosols at variouswind speeds, The sea-saltand wind speedrelationship given by Woodcock Marlcs[1990] obtainedthe dependence05 sea-saltaerosols on [1953]was obtained at cloudlevels (about 600 m) in the Pacific wind speedfor five •ize rangesand œortotal sea salt, respec- with a slope(a) of 0.16. AlthoughLovett [1978] obtainedthe tively. Under dry nonprecipitatingconditions, except for the same value for an annual averagebetween August 1974 and dinmeter of 0.•9-0.95 •m where the correlation coe•cient is July 1975 in the Atlantic, the measurementswere performed at 0.31, the slopesfor all other four size stagesranging from 0.95 5-15 m. It is expectedthat as altitude increasesthe dependency to 30 /•m (diameter) was about 0.23 [Marlcs,1990, Table 2], of sea salt on wind becomes weaker. This feature is reflected in which is very close to the model simulatedresults of 0.26 for the model which gives a smaller slope of 0.15 in the layer of total mass.The slopeis 0.12 for the observationof particlesin 166-722 m than that of 0.26 in the surfacelayer. It is closerto diameter of 0.49-0.95/•m and 0.13 for the simulationof par- 0.16 of Woodcock[1953]. ticles in the radii ranging from 0.13 to 0.25/•m. The constantb in (25) is referred asthe backgroundsea-salt Exton et al. [1985, Table 3] found the sametrends for three loading. It is the sea-salt concentration when wind speed size categoriesof sea-saltaerosols based on the field measure- reacheszero. From Figure 7a, our predictionyields a value of ments off the northwestcoast of Scotland.For the ultra large 1.4/•gm -3 for Heimaey.Compared to theobservations (Table sea-salt(r = 8-16/•m) aerosols,the slopeand the correlation 4), b is 1.13/•gm --• forMarks [1990], 2.57/•g m --• for Wood- coefficientwere 0.235 (average) and 0.59 (average) respec- cock[1953], 0.33/•g m -'• for Tsunogaiet al. [1972],and 4.26/•g tively.For the large (r = 0.3-8/•m) aerosols,they changedto m-3 forLovett [1978]. There are two factors which may explain 0.13 and 0.56 respectively.For the small (r = 0.1-0.3 /•m) the observationaldifference. (1) The three experimentswere aerosols,the slopewas reduced even more to 0.08 while the made at the North Atlantic [Lovett, 1978], the North Sea correlation coefficient was 0.4. [Marks, 1990] and the Pacific [Tsunogaiet al., 1972] respec- Tsunogaiet al. [1972] measuredsea-salt concentration from tively. Different meteorologysuch as wind speedand precipi- a cruise ship on the Pacific Ocean and a slope of 0.62 was tation at these siteswill result in different patternsfor sea-salt obtainedfor the wind dependence.It is muchlarger than other generationand removalwhich influence the constanta and b observation's.They attributed it to two reasons:(1) the pro- in (25). Our simulationalso shows the variabilityof constanta ductionof sea-saltparticles > 10/•min radiusor 10-s g in mass and b at varioussites (Table 6). The salinityof the oceanmay is more dependenton the wind speedthan the productionof play a role in the differencebut may not accountfor the big smallerones; (2) the residencetime of the larger particlesis so discrepancybetween 4.26 and 0.218 since the differencebe- short that they contribute only to the weight concentrationof tween the highestsalinity of 35.5%e in the North Atlantic and sea-saltparticles near the surfaceof the ocean.Lovett [1978] the lowest of 34.2%e in the North Pacific cannot contribute attributed the large value to the movingship which may gen- such a big differencein sea-saltconcentration; (2) the mea- erate sea-saltparticles. This effect is unlikely true sincesam- surement techniquesused, for example, the sampling time, ples were discardedwhen a tail wind was recorded. In our averagemethods and equipment.The samplingduration prob- opinion, it may be due to the experimentalmethod used to lem has been discussedabove with Tsunogaiet al.'s [1972] constructthe sea-saltwind dependence.It was reported in the measurements.When a size distribution is measured by an 3816 GONG ET AL.' ATMOSPHERIC SEA-SALT MODELING, 1
a) PREDICTIONS generatedby wind will depositback to the surfacevery quickly. At any time and location, the sea-saltconcentration is main-
10 3 taincdmainly by the localwind speed.Hence a strongerwind speeddependency is observed.On the other hand, the smaller Total Sea-salt, a=0.262 particleshave a residencetime of severaldays. This implies • 102 that at any time the aerosolconcentration of smallerparticles is controllednot only by the local wind speed but also by • 10' ß e ß ß long-rangetransport of fine sea-saltparticles. The longerres- ß ß ß v • v ß .9 ß ß ß vv vl ßß ß •,,! ß ß vv ß ß idencetime yieldsa smallerslope and correlationco½!ficicnt c 100 between sea salt and wind speed. In addition, non Sea-salt
o biogcnicsulphur aerosolcan contributesubstantially to fine o *-' particles.Since their concentrationis lessdependent on wind speedthan that of seasalt, the correlation with wind speed is less. 10 -2 ß Accordingto our simulations,the dependenceof total sea- saltaerosols also varies with geographiclocations. For four 10 -3 ßi , 'i ß , 'i i ' ßi ';i i i Sea-salti i i [r0=0.13-0.25i i i i •m],i a=0.129i i i 5 10 15 20 marine locationssimulated in this study,the slope a ranges 10 rnWind Speed [m S '1] from 0.20 to 0.26 (Table 6). It seemsthat the climatepattern in a specificsite may regulatethe removalprocesses and hence b) PREDICTIONS vs OBSERVATIONS the dependence.Even though different measuringmethods mayyield a differencein the absolutevalue of sea-saltconcen- 10 5 trations,the dependenceon wind speedshould be relatively less affected.The variant slopesin Table 4 are indicativeof ,+ßWoo,dcock 1953 such differences.A global relationshipused for the sea-salt 10 4 --a-- Tsunogaieta11972 --,•--ß Lovett 1978 dependenceon wind speed[Erickson and Duce, 1988] may be ,•--G•--Kulkarni eta11982 subjectto large uncertainties. • Extoneta11986 ,-•)--Gras & Ayers1983 3.4. Residence Time of Sea-Salt Aerosols I::: 10 3 --a-- Marks Residencetime of theparticles is an importantindicator of
._ particlecycles in the atmosphere.Factors regulating residence (• 102 time include the production, accumulationrate, removal and growth of the aerosols[Junge, 1974; Bolin and Rodhe, 1973; o Slinn,1984]. Except for horizontaladvection, all majorphysical processesare includedin this aerosolmodel. Therefore it is an 10' ideal tool to investigatethe effect of local processeson the residencetime of the atmosphericaerosols. The residencetime of particlesin each size bin is dynami- 10 0 cally calculatedaccording to the followingformula:
X 10-1 , • , , I , , • I , • • , I , , ,,1 0 5 10 15 WindSpeed [m Figure 7. Comparison of sea-salt mass concentration as a functionof wind speed;(a)•mod½l prediction for total massand 35 radii of 0.13-0.25 •m in the first modellayer at H½imacyof Icelandand (b) observationswith the modelpredictions at two sites. + SamplingDuration 180 • -' SamplingDuration 51 • E 25
• 20 impactor, isokincticconditions at the inlet have to be main- taincd to reduce the inlet lossesof large particles[Marple and Willeke,1976]. • 15 it is interestingto note that for both observationsand pre- dictions,as the particle size decreases,the slopeand correla- tion of the sea-saltvs. wind speedcurve decreases.This indi- 5 catesthat for largesea-salt particles (e.g., r d = 4-8 /•m) the local surfacewind is the dominantdriving force in determining 0 .... i i i .... i .... i , , • • the concentration.For smallparticles (e.g., rd = 0.1-0.3/•m), 0 2 4 6 8 10 however,other factorsbegin to affect the concentration.One WindSpeed [rn s-'] important parameter to characterizethis phenomenonis the Figure 8. The effectof samplingduration on the rms sea-salt residencetime. For large particles,the residencetime is on the concentrationassuming a sinecycle change of wind speedand order of hours(see next sectionfor details).Any suchparticles an exponentialrelationship between sea salt and wind speed. GONG ET AL.: ATMOSPHERIC SEA-SALT MODELING, 1 3817
Table 6. Total Sea-SaltMass ConcentrationX Dependence (a) ResidenceTime of Sea-saltAerosols [ro=4-8 tzm] on 10-mWind SpeedU•o' X = beaU•ø
10 2 ...... i ...... i ...... i ...... i ...... i ...... i ...... a, b, Location s m- • gg m-3 r 2
Mace Head 0.20 3.1 0.76 Hawaii 0.21 3.4 0.58 Bermuda 0.22 2.5 0.68 o 10 • Heimaey 0.26 1.4 0.82 E ,_ 1:•~ 30 Minutes
'• 100 where X'is the concentrationand [Ox/Ot] the tendencydue to either addition or removal processescalculated at each time step. Under steadystate conditions,this tendencyis identical for both of them. In a dynamic model where steady state is hardly reached,a choicehas to be made as to which tendency 10-,...... i ...... i ...... i ...... i ...... i ...... i ..... i i i 0 10 20 30 40 50 60 is used.Fortunately, in terms of the averagedtendency over a long period, a quasi-steadystate is maintained, that is, an Julian Day averagedtendency is equal for both addition and removal. Consequently,an averagedresidence time is calculatedby us- (b) ResidenceTime of Sea-saltAerosols [rd=0.13-0.25 gm] ing the removaltendency. Figure 9 showsthe residencetimes for particlesin two size bins for the atmosphericsurface layer (0-166m). For particles in bin8 (ra = 4-8 •m), theaveraged 103 residencetime is about 30 min (Figure 9a). Becauseof a large dry depositionand settlingvelocity, this categoryof aerosols will not reachthe upperatmosphere and engagein longrange 1:1~ 60 hours transports.Smaller particles (r a = 0.13-0.25 •m) have a residencetime of about 60 hours (Figures 9b) in the first atmosphericlayer (0=166 m) ß The longerresidence time of m 10 • smallparticles allows them to reach high altitudesand partic- ipate in long-rangetransport. It shouldbe pointed out that the residencetime in Figure 9 is for the first atmosphericlayer only (0-166 m). This is the reservoirwhere the removal tendencyis calculated.It is antic- ...... i ...... i ...... i ...... t ...... i ...... •111111111 ipatedthat the residencetime will vary at differentlayers as the 0 10 20 30 40 50 60 removal tendency and concentrationof aerosolswill be con- Julian Day trolled by different conditions. Figure 9. The residencetime of sea-saltaerosols in the first 4. Conclusions atmosphericlayer (0-166 m) for (a) large and (b) small par- ticles.The radii of sea-saltaerosols are shownfor dry particles. The applicationof this aerosolmodel to sea-saltaerosols revealedvery good agreementbetween predictions and obser- vations in terms of particle number density, size distribution and wind effectson sea-saltaerosols. The sourcefunction by explainsthe less dependencyof sea-saltaerosols on the wind Monahah et al. [1986] producesreasonably sea-salt emission speed. rate for particlesdown to 0.1 •m. However, below 0.1 •m, a large overestimateof both mass and number concentrations occursprobably due to the extrapolationof the Monahan et Acknowledgments. The authors would like to express sincere thanksto Filippo Giorgi for discussionon dry depositioncalculation, al.'s formula beyond the particle size range within which the P. K. Wang and Beheng on the scavengingof aerosolparticles. formula is defined. It is found that the total mass of atmo- sphericsea-salt aerosols is mainly governedby the local wind speed.A regressionof the simulatedresults as a function of References --1 wind speedgives a slope (a) rangingfrom 0.20 to 0.26 s m Barrie, L. A., Snow formation and processesin the atmospherethat and an interceptcorresponding to backgroundsalt loading(b) influenceits chemicalcomposition, Seasonal Snowpacks, edited by from 1.4 to 3.4 •g m-3 for the four locationsmodeled, indi- T. Davies et al., NATO ASI Set., Set. G28, 1-20, 1991. catingthat the sea-saltwind correlationdepends on geographic Blanchard,D.C., Electrificationof the atmosphereby particlesfrom bubbles in the sea, in Progressin Oceanography,vol. 1, edited by locations.Small particle concentrationsare lessdependent on M. Sears,pp. 73-197, Pergamon,New York, 1963. wind speedthan large particles.This was true in both experi- Blanchard,D.C., The production,distribution, and bacterial enrich- ments and this simulation. The residence time of sea-salt aero- ment of the sea-salt aerosol, in Air-Sea Exchangeof Gases and solsis found rangingfrom 30 min (r a = 4-8 •m) to 60 hours Particles,edited by P.S. Liss and W. G. N. Slinn, pp. 407-454, D. Reidel, Norwell, Mass., 1983. (r a = 0.13-0.25 •m) in the first atmosphericlayer (0-166 m) Blanchard, D.C., and A. H. Woodcock, Bubble formation and mod- dependingon the size of the particlesand local meteorological ification in the sea and its meteorologicalsignificance, Tellus, 9, conditions.The longer residence time for smaller particles 145-158, 1957. 3818 GONG ET AL.: ATMOSPHERIC SEA-SALT MODELING, 1
Bolin, B., and H. Rodhe, A note on the conceptsof age distribution Canadian climate centre second-generationgeneral circulation and transit time in natural reservoirs, Tellus, 25, 305-308, 1973. model and its equilibrium climate,J. Climate,5, 1013-1044, 1992. Burk, S. D., The generation,transport, and depositionof marine aero- McGovern, F. M., A. Krasenbrink,S. G. Jennings,B. Georgi, T. G. sols, a turbulent modeling study, in Oceanic Whitecaps,edited by Spain,M. Below, and T. C. O'Connor, Massmeasurement of aerosol E. C. Monahan and G. Mac Niocaill, pp. 267-268, D. Reidel, Nor- at Mace Head, on the westcoast of Ireland,Atmos. Environ., 28(7), well, Mass., 1986. 1311-1318, 1994. Dingle, A. N., and Y. Lee, An analysison in-cloudscavenging, J. Appl. Monahan, E. C., Oceanic whitecaps,J. Phys. Oceanogr.,1, 139-144, Meteorol., 12, 1295-1302, 1973. 1971. Erickson,D. J., and R. A. Duce, On globalflux of atmosphericsea salt, Monahan, E. C., D. E. Spiel, and K. L. Davidson,A model of marine J. Geophys.Res., 93(Cll), 14,079-14,088,1988. aerosol generationvia whitecapsand wave disruption, in Oceanic Exton, H. J., J. Latham, P.M. Park, M. H. Smith, and R. R. Allan, The Whitecaps,edited by E. C. Monahan and G. Mac Niocaill, pp. 167- productionand dispersalof maritime aerosol,in OceanicWhitecaps, 174, D. Reidel, Norwell, Mass., 1986. edited by E. C. Monahan and G. Mac Niocaill, pp. 175-193, Mozurkiewcz,M., Mechanismsfor the releaseof halogensfrom sea D. Reidel, Norwell, Mass., 1986. salt particlesby free radical reactions,J. Geophys.Res., 100(D7), Fairall, C. W., and K. L. Davidson,Dynamics and modelingof aerosols 14,199-14,207, 1995. in the marine atmosphericboundary layer, in Oceanic Whitecaps, Parungo,F. P., C. T. Nagamoto,J. Rosinski,and P. L. Haagenson,A edited by E. C. Monahan and G. Mac Niocaill, pp. 195-208, studyof marine aerosolsover the PacificOcean, J. Atmos. Chem.,4, D. Reidel, Norwell, Mass., 1986. 199-226, 1987. Fairall, C. W., and S. E. Larsen, Dry deposition, surface production Pilat, M. J., Collection of aerosolparticles by electrostaticdrop spray and dynamicsof aerosolsin the marine boundary layer, Atmos. scrubbers,J. Air Pollut. Control Assoc., 25, 176-178, 1975. Environ., 18, 69-77, 1983. Podzimek, J., Advances in marine aerosol research,J. Rech.Atmos., 14, Fitzgerald,J. M., Marine aerosols:A review,Atmos. Environ., Part A, 35-61, 1980. 25, 533-545, 1991. Prospero,J. M., Mineral and seasalt aerosolconcentrations in various Fitzgerald,J. M., Numerical simulationof the evolutionof the aerosol ocean regions,J. Geophys.Res., 84, 725-730, 1979. size distributionin a remote marine boundarylayer, in Nucleation Slinn, W. G. N., Precipitationscavenging, in AtmosphericScience and and AtmosphericAerosols, edited by N. Fukuta and P. E. Wagner, PowerProduction, edited by D. Randerson,pp. 466-532, Tech. Inf. pp. 157-160, A. Deepak, Hampton, Va., 1992. Cent., Off. of Sci. and Techn. Inf., Dep. of Energy, Washington, Gathman, S. G., A time-dependent oceanic aerosol profile model, D.C., 1984. NRL Rep. 8536, Natl. Res. Lab., Washington,D.C., 1982. Slinn, W. G. N., and J. M. Hales, A re-evaluation of the role of Gelbard, F., and J. H. Seinfeld,Simulation of multicomponentaerosol thermophoresisas a mechanismof in- and below-cloudscavenging, dynamics,J. ColloidInterface Sci., 78, 485-501, 1980. J. Atmos. Sci., 28, 1465-1471, 1971. Gerber, H. E., Relative-humidityparameterization of the Navy aerosol Slinn, S. A., and W. G. N. Slinn, Modeling of atmosphericparticulate model (NAM), NRL Rep.8956, Natl. Res. Lab., Washington,D.C., depositionto natural waters, in AtmosphericPollutants in Natural 1985. Waters,edited by S. J. Eisenreich,pp. 23-53, Ann Arbor Sci., Ann Giorgi, F., A particle dry-depositionparameterization scheme for use Arbor, Mich., 1981. in tracer transport models,J. Geophys.Res., 91(D9), 9794-9806, Stramska,M., Vertical profiles of sea-saltaerosol in the atmospheric 1986. surfacelayer: A numericalmodel, Acta Geophys.Pol., x:rrv(1),87- Gras, J. L., and G. P. Ayers, Marine aerosolat southernmid-latitudes, 100, 1987. J. Geophys.Res., 88(C15), 10,661-10,666,1983. Therrien, D., Le modale de circulation g6n6rale atmosph6riqueCa- Greenfield,S., Rain scavengingof radioactiveparticulate matter from nadien en versioncolomne: FIZ-C, MSc. thesis,123 pp., Univ. of the atmosphere,J. Meteorol.,14, 115-125, 1957. Quebec, Montreal, 1993. Herbert, F., and K. D. Beheng, Scavengingof airborne particlesby Toba, Y., On the giant sea-saltparticles in the atmosphere,I, General collisionwith water drops--Model studieson the combinedeffect of features of the distribution, Tellus, 17, 131-145, 1965a. essential microdynamic mechanisms,Meteorol. Atmos. Phys., 35, Toba, Y., On the giant sea-saltparticles in the atmosphere,II, Theory 201-211, 1986. of the vertical distributionin the 10-m layer over the ocean, Tellus, Ikegami, M., K. Okada, Y. Zaizen, and Y. Makino, Sea-saltparticles in 17, 365-382, 1965b. the upper tropical troposphere,Tellus, Set. B, 46, 142-151, 1994. Toba, Y., and M. Chaen, Quantitative expressionof the breakingof Jacobs,W. C., Preliminaryreport of a studyof atmosphericchlorides, wind waveson the sea surface,Rec. Oceanogr.Works Jpn., 12, 1-11, Mort. Weather Rev., 65, 147-151, 1937. 1973. Jobson,B. T., H. Niki, Y. Yokouchi, J. Bottenheim, F. Hopper, and Tsunogai,S., O. Saito, R. Yamada, and S. Nakaya, Chemicalcompo- R. Leaitch, Measurementsof C2-C6 hydrocarbonsduring the Polar sition of oceanicaerosols, J. Geophys.Res., 77, 5283-5291, 1972. Sunrise1992 Experiment:Evidence for C1atom and Br atom chem- Twomey,S., AtmosphericAerosols, Elsevier Sci., New York, 1977. istry,J. Geophys.Res., 99(D12), 25,355-25,368,1994. Wang, P. K., S. N. Grover, and H. R. Pruppacher,On the effect of Junge, C. E., Residence time and variability of tropospherictrace electriccharge on the scavengingof aerosolparticles by cloudsand gases,Tellus, 26, 477-488, 1974. small raindrops,J. Atmos. Sci., 35, 1735-1743, 1978. Kulkarni, M. R., B. B. Adiga, R. K. Kapoor, and V. V. Shirvaikar,Sea Woodcock, A. H., Salt nuclei in marine air as a function of altitude and salt in coastalair and its depositionon porcelaininsulators, J. Appl. wind force, J. Meteorol., 10, 362-371, 1953. Meteorol., 21, 350-355, 1982. Wu, J., Wind stressand surface roughnessat air-sea interface, J. Lepple, F. K., D. S. Bressan,J. B. Hoover, and R. E. Larson,Sea salt Geophys.Res., 74, 444-555, 1969. aerosol,atmospheric radon and meteorologicalobservations in the Wu, J., Whitecaps,bubbles, and spray,in OceanicWhitecaps, edited by westernSouth Atlantic Ocean (February1981), NRL Memo. Rep., E. C. Monahan and G. Mac Niocaill, pp. 113-124, D. Reidel, Nor- 5153, 1983. well, Mass., 1986. Levich, V. G., Physico-chemicalHydrodynamics, Prentice-Hall, Engle- van Eijk, M. J. Anderson,and G. de Leeuw, Modeling aerosolparticle wood Cliffs, N.J., 1962. size distribution over the North Sea, J. Geophys.Res., 97(C9), Lovett, R. F., Quantitative measurement of airborne sea-salt in the 14,417-14,429, 1992. North Atlantic, Tellus, 30, 358-363, 1978. Marks, R., Preliminaryinvestigation on the influenceof rain on the L. A. Barrie and S. L. Gong, Atmospheric Environment Service, production,concentration, and vertical distributionof seasalt aero- 4905 Dufferin St., Downsview, Ontario, M3H 5T4 Canada. sol,J. Geophys.Res., 95(C12), 22,299-22,304, 1990. J.-P. Blanchet,Earth SciencesDepartment, Universityof Quebec at Marple, V. A., and K. Willeke, Inertial impactor:theory, designand Montreal, Station "Centre Ville," 515 Ste.-Catherine St., Montreal, use, in Fine Particles,Aerosol Generation,Measurement, Sampling, Quebec, H3C 3P8 Canada. and Analysis,edited by B. Y. H. Liu, pp. 411-446, Academic, San Diego, Calif., 1976. (ReceivedOctober 12, 1995;revised July 12, 1996; McFarlane, N. A., G. J. Boer, J.-P. Blanchet, and M. Lazare, The acceptedAugust 26, 1996.)