Modeling of aerosol properties related to direct ... - Wiley Online Library

JOURNAL OF GEOPHYSICAL

RESEARCH, VOL. 103, NO. D14, PAGES 17,009-17,032, JULY 27, 1998

Modeling of aerosolproperties related to direct climate forcing Sotiria Koloutsou-Vakakis

and Mark

J. Rood

EnvironmentalEngineeringand ScienceProgram,Departmentof Civil and EnvironmentalEngineering, Universityof Illinois, Urbana-Champaign

Athanasios

Nenes

EnvironmentalEngineeringDepartment,California Instituteof Technology,Pasadena

Christodoulos

Pilinis

EnvironmentalScienceDepartment,University of the Aegean, Mytilene, Greece

Abstract. A long-termlocalexperimentwasdesignedwith thepurposeto accuratelyquantify aerosolparameters neededin orderto estimateaerosolclimateforcingat an anthropogenically

perturbed continental site.Totallight-scattering O'•,,s pandbackscattering O'•,,bsp coefficients at wavelength •, thehygroscopic growthfactorswith respectto scattering, f(RH)x,s,andthe backscatter ratiobxaretheparameters considered in thepaper.Referenceandcontrolledrelative humiditynephelometry measurements weretakenat a groundlevelfieldsamplingstation,

located nearBondville Illinois (40ø03'12" N,W 88ø22'19" W).Aerosol particle chemical

composition andmassparticlesizedistributions werealsomeasured. The targetparameters were alsoestimatedfrommodels.The modelingapproachinvolveda two-stepprocess.In thefirst step,aerosolproperties wereparameterized with anapproach thatmadeuseof a modified thermodynamic equilibriummodel,publishedlaboratory measurements of singlehygroscopic particleproperties, andempirical mixingrules.In thesecond step,theparameterized aerosol

properties wereusedasinputsintoa codethatcalculate O'•,,s pandO'•,,bsp asfunctions of )•,RH, particlesize,andcomposition. Comparison between themeasured andthemodeled results showed thatdepending ontheassumptions, thedifferences betweenthemodeledandobserved results werewithin5 to 28%forf(RH)•,s andwithin22-35%forbxatlowRH and0-20%forbx athighRH. Thetemporal variation of theparticlesizedistribution, theequilibrium stateof the particles, andthehygroscopicity of thematerialcharacterized asresidual werethemajorfactors limitingthepredictiveabilityof themodels. 1.

Introduction

(blue: 450 nm), g (green= 550 nm) or r (red = 700 nm)). Pilinis et al. [1995] and Boucher and Anderson [1995] examinedthe Estimatesof the direct aerosol forcing due to tropospheric sensitivityof climateforcingto differentfactors,and they found sulfateaerosolparticlesrange between-0.2 and -0.8 W/m2, that relative humidity (RH) is the most important factor in determiningaerosolradiativeforcing. InternationalPanel on Climate Change[(IPCC), 1996, chap.2]. Aerosol parametersneeded to validate these conclusionsas To reduce the uncertainties of these estimates, ambient aerosol propertiesandrelatedphysicalandchemicalprocesses needto be well as the predictions of the magnitude of aerosol climate accuratelyquantified[Lacisand Mishchenko,1995; Schwartzet forcingby globalradiativetransfermodels[e.g., Charlsonet al., al., 1995]. In assessingthe direct radiative forcing of aerosol 1991;Kiehl and Briegleb, 1993] have not yet beenevaluatedin a particles, interaction of aerosol particles with water vapor systematic or extensive enough manner to allow accurate (H20[g]) is veryimportant. Theeffectsof hygroscopicity onthe quantificationof the magnitudeof the aerosolradiative forcing optical propertiesof the aerosolhas been studied by many [Penneret al., 1994; Ogren, 1995; National ResearchCouncil, investigators[Pilat and Charlson, 1966; Covert et al, 1972; (NRC), 1996]. Here we focus on the following parameters:(1) Charlsonet al., 1974; Weisset al., 1977; Wolff et al., 1981; the light-scattering hygroscopicgrowth factorf(RH)x,.,,which Groblickiet al., 1981; Tang et al., 1981; Waggoneret al., 1983; represents the ratio of crx,,or crx, b, at RH > RHre f..... to O',•,s p Sloaneand Wolff, 1985; Rood et al., 1987; 1989; McMurry and or crx,•,, respectively, at RH = RHref• .... ; (2) the hemispheric Stolzenburg,1989; Pilinis, 1989; Koloutsou-Vakakis and Rood,

backscatter ratiobx,whichis theratioof crx,•,overcrx,.•p for the 1994;Zhanget al., 1994].H20[g ] uptakeby aerosolparticles same • and RH, bx is used as a proxy for the asynm•etry results in changes of the particle size distribution and compositionof the particlesand subsequent changesin total

light-scattering (crx, sp)andbackscattering coefficients (crx, b,) at differentwavelengthsof light • (• in this papercorresponds to b

Copyright1998 by the AmericanGeophysicalUnion. Papernumber 98JD000680 0148-0227/98/98JD-00068509.00

parameter gxthatcanbe usedto estimatethe upscatter fractionfix usedin radiativetransfermodels [Wiscombeand Grams, 1976; Marshall et al., 1995]. Measurementsalso need to be coupled with models. The combinedmeasurement- modeling approachis needed (1) to examine how much of the measured parameters can be adequately predictedby the modelsgivenpracticalconstraints in data acquisitionof ambientaerosolpropertiesand (2) to specify the limitationsof either the measurements or the models,as they are identified by the explicit comparisonapproachwe follow

17,009

17,010

KOLOUTSOU-VAKAKIS ET AL.: MODELING OF AEROSOLPROPERTIES

Subsequentto the one-stageimpactors,the aerosol flowed througha referenceintegratingnephelometer (with similaroptics to MRI Inc., model 1560) [Rood et al., 1987] which measured

here. Comparisonsbetween results from measurementsand results from models are necessarilypart of a feedback loop, where each iteration is a step into understanding the observed processes and obtainingmore accuratepredictions.Sincethere are limitations in the spatial and temporal resolutionof the measurements,models are needed to interpolate aerosol properties andto estimateaerosolparameters thatarenotdirectly measurablesuch as fix or mass scatteringefficiencies, aX,s,i (definedas the changeof crxfor a unit changein the mass concentrationof chemical compound i). On the other hand, improvements in the measurement methodsandprocedures used in the field and the laboratory can lead to more accurate parameterizationof aerosol propertiesused as inputs to the

crg,s pof the ambientaerosolat RH < 40%. The one-stage impactorswere usedto investigatedifferencesbetweenfine and fine pluscoarsemodeparticlescattering. In the secondpath, aerosol flow alternatedevery 7.5 min between a 10 or a 1 gm cutpoint impactor upstreamof a humidifier. The humidifier was designedto allow continuous

automated measurements of thedependence of crx,,andcrx, bsp on RH [Koloutsou-Vakakis, 1996].The humidifierconsisted of two concentric tubes. An outer 2.54 cm O.D. stainless steel tube and an inner 1.90 cm O.D. flexible Teflon membrane tube which was

models.

supported by a stainless steelwiremeshtube.Themembrane was

In this paper, we presentsuch a measurement- modeling approach.The purposeof the measurements was quantification of the dependenceof light-scatteringon 2, RH, chemical composition,and direction of scatteringfor anthropogenic aerosolat a continentalsite in the northernhemisphere.Existing and appropriatelymodified modelswere then used to obtain predictions of suchdependence. The experimental andmodeling proceduresare described,and the resultsare comparedand

permeable by H20[g ] but not H20[l ] at normaloperating conditions. Transport of H20[g ] into the aerosolstreamwas controlled by controlling thetemperature of H20[l]. The system

evaluated.

throughout the samplingsystem,andheatingof theaerosolwas

2. Experimental Procedure

limitedto 5øC,asit passedthroughthehumidifier. A three-wavelengthnephelometerwith backscattershutter (TSI, model 3563) locateddownstreamof the humidifierwas

allowed accurate RH control of the aerosol sample. The

advantages of thishumidifierwerethatit wasrelativelysimpleto

operate, its designeliminated the needfor dilutingthe aerosol streamwith dry air to adjustthe RH to the requiredvalue,the 1.59 cm internal diameter of the sample line was constant

usedto measure crx, sp andcrx, b, of theaerosol. Plotsof thecrx, sp andcrx,•sp dependence on RH, referred to as "humidograms,"

An aerosolfield samplingsite was developedfor this study,

southof Bondville, in eastcentral Illinois,at 40ø03'12" N, wereobtainedevery30 min, alternatingbetweenthe fine andthe 88ø22'19" W,and229mabove sealevel[Semonin etal.,1987]. fine plus coarsemodeparticles.ContinuousRH scansoccurred between40% g RH g 90%, over a period of 7.5 min. From

The site is located within the region with the predictedhighest

of rrx, spandcrx,•sp, bxcanbeestimated asfunction SO42burdens andradiative forcingdueto anthropogenic SO42- measurements of 2 and RH. Measurementsobtained from this path were

[Chadsonet al., 1991]. Ambient aerosolwas sampledfrom 10 m abovegroundlevel. The aerosol flowed through an inlet stack and split into four paths(Figure 1). A flow of • 1 lpm of the aerosolpassedthrough a condensation nuclei (CN) counter(CNC) (TSI model 3010) to

combined with the low RH measurements obtained from the

reference nephelometer to determine f(RH)x,,. In the third path, the aerosolstreamflowed througha lowpressureimpactorwith seven stages.The 50% aerodynamic particlediametercutpointswere at 0.125, 0.25, 0.5, 1.0, 2.0, 4.0 and 8.0 pm, at an aerosolflow rate of 30 lpm [Berneret al., 1979; Wangand John, 1988; Quinnet al., 1995]. The particles

monitor CN in the inlet aerosol stream.

In the first path, the aerosolstreamflowed througheither of two one-stageimpactorswith upper cutpointsat 10 and l gm aerodynamicdiameter, respectively. An automatic ball valve alternatedevery 7.5 min to allow samplingby each impactor.

inlet stack

were collected on Tedlar films and a Teflon-coated quartz

backupfilter. Thesefilms and filter were subsequently analyzed

reference nephelometer []

MFC

I [path 1]=:• relativehumidity MFC controll•!d nephelometer

I

/,,__ [..,,•

[path 21

MFC

,'",";• [ impactor with cutpoints

,('/

I

MFC

intheO.1to4 pmrange

impactors with

1and10pm [

[path3]

cutpoints I

impactor with

/

pump

1 pmcutpoi?if ....'l'--'•

filters

........................

MFC

i:• filter

[path 4]=i•'•"

MFC: mass flow controller CNC: condensation nuclei counter

Figure 1. Schematicof the instrumentation.

atmosphere

KOLOUTSOU-VAKAKIS

ET AL.' MODELING OF AEROSOL PROPERTIES

17,011

with ion chromatography (IC), (Dionex model DX-300) to assess 3.1.Estimateof crx. spand O'A, bs p the H20 solubleinorganicion compositionof the aerosolwith

respect toNa+ K + Mg 2+ Ca2+ NHI SO42-C1- Br- and NO• [Hahn,1995]. In the fourthpath, the aerosolstreamflowed througha l gm impactor,before it was filtered with Teflon-coatedquartz fiber filters. The 24-hr filter sampleswere measuredgravimetrically (Cahn C-31 microbalance)before and after sampling.Then the filter samples were analyzed with IC to assessthe ionic compositionof the aerosolwith respectto the sameions listed for the impactorfilms.

ValuesforC%pand O'A, bs p due to a population of aerosol particles with number concentrationN and size distribution

dN/dlog dp,aregivenby

dmdlogalp cr=I C(dp,n,A) dlogdp 0

(1)

[Hinds,1982],wherearepresents eithercrx, sp or O')[,bsp andC(dp, n, g) is the scatteringcrosssectionof a singleparticleof diameter

dpandrefractive indexn atwavelength g. C isdetermined by

3. Models

øø2

(2) C=2•rI I (dp,n,A,O)-•-sin(O)dO

Thepurpose of modeling in thisstudywasto estimate crX. sp andCrX. b. asfunctions of composition, size,RH, and/LModeling consistedof two main steps'(1) parameterization of the aerosol

properties neededas inputsfor estimating C%pand crx. b. as

d2

whereI(dp,n, A, O) isthenormalized intensity of thescattered or backscattered lightat distance d/2 fromthecenterof theparticle

and •9• = 0 for total scattering,and functionsof/• and RH and (2) calculationof light-scattering at scatteringangle 0, t92 = 71; coefficients at the appropriateangularranges(Figure2). •9• = x/2 for backscattering[Kerker, 1969]. C is equal to the

24hrambient filter samples ) I

IC analysis

,n9

i•ambicnt measured

,

themodynamie

•paflicle size distributions)

equilibrium m•eling •• Wi, ni, Pt, Adp i• 240%) (SEQ•Lm+)

•

Inmi•,Pmi•,ad,mi (• •

i

, 'pl m

[molecular composition i

measured

modeled

propetites of dual mate•al includo•ff [ paniclos ofknown [ mo•stablo

predictionof o.•,spandøk,bsp

using •o thoo• and

fiR.H) l,,bsp

ß•

ncphclomoter simulation

AS•E

E•IE

f(RH) •,sP ]"---i•,sp-( measure •'• •P..,•)I fOt3'I) •"bsp

fOti-I) •,,sp

com•sition

rr

f(RH •.)

b•.

( 24hrambient filter samples ) I

• and) Oreblent mcasur• •articlc size distdbutions•

•ravimetric

•C analysis

[•] I••

I

each •

I

value I

v

•.•

measured modeled comparison

f(RH) implicitlycontains thedependence on particleecruposition andsize

Iicsidual material included? I yes

no

I

prediction of•,sp •e theo• •LSIE)

Figure 2. Flowchartof modelingstepsandcomparisons betweenmeasured andmodeledresults.

17,012

KOLOUTSOU-VAKAKIS

ET AL.: MODELING OF AEROSOL PROPERTIES

optical scatteringefficiency Q multiplied by the particle's NO j existin the ambientaerosolare not explicitpartsof the input. As a result, the aerosolparametersneededfor lightNephelometers do not typicallydetectlight scattered overthe scatteringcalculationsare estimatedby consideringa set of entire [0ø, 180ø] angular range. The finite length of the indicator ions, and then assumptionsare made about the nephelometer restrictsthe angularrangeto detecttotalscattering molecularforms of these ions for ambientaerosolparticles in the 7 ø < 8 < 170ø range and backscattering in the 90ø < 8 < [Sloane,1984, 1986;Sloaneand Wolff,1985;Quinnet al., 1995] geometriccross-sectionalarea.

170ørange. Also,duringcrx, bsp measurements thenephelometer's depending onindicators suchastheNH4+/SO 42-moleratio.For

shutterdoesnot perfectlytruncatethe light at 8=90 ø. In this if NH •/SO 42= 2, thendensity andrefractive indexfor paper,we usethe approachof Andersonet al. [ 1996]to account example, 4areusedtocalculate O'2, spandO'2,bs p. for the "truncation"and "shutter" errorsof the nephelometer. (NH4)2SO

Two angularsensitivity functions f•p(8)andf•,(8) areusedto estimate the scattering efficiencies QspandQb,, respectively,3.3. Effect of Compositionon Light Scattering, This Work whichcorrespond to thenephelometer measurements; f•p(8)and f•,(8) thatreplace sin(8)in equation (2) asdefined below The approachfollowedin this paperis an attemptto make parameterizationof aerosolpropertiesmore systematic.We use a 0

lsn00=2

0

lfbsp (6})d/9= lsinOdO=l

0

zt/2

(3a)

(3b)

3.2. Effect of Composition on Light Scattering, Previous Work

Equation(1) canbe rewrittenin the form

recentlydevelopedthermodynamic equilibriummodelto predict the molecular compositionof the ambient aerosol. The main motivationbehindthe approachproposedhereis to minimizethe useof "typical" aerosolpropertiesby quantifyingthe physical, chemical,and opticalpropertiesof the specificaerosolbasedon the field measurements and appropriatemodels.This approach couldalsobe usefulin assessing how adequate are singleparticle and/or singlesolutepropertiesin representingambientaerosol properties. Initially, anionandcationconcentrations measuredin the field are used to predict dry aerosolcompositionas well as aerosol

H20[1 ] contentas a functionof RH underthe assumption of thermodynamic equilibrium. The predicted dry aerosol compositionis thenusedto provideparameters(particlesizeand refractiveindex)neededfor the light-scattering calculations. Apportionmentof light-scatteringto different components (4) requiresestimatesof scatteringefficiencieswith and without a wherecrrepresents eithercrx,,or O')[,bsp , depending ontherange certain species by appropriately shifting the particle size of 8, and i denotesparticlesof compositioni [Sloane,1986]. The distribution. In thiswork,the component of interest is H200]. integralin equation(4) is equivalentto the integral in equation Therefore, first, H200]content,and,second,the shiftof the (1), when consideringsphericalparticles,with a specifiedmass particle sizedistribution dueto thepresence of H20[• ] needto be size distribution. Contributions to scattering by different estimated. AerosolH20[• ] content is a measurable quantity, but chemicalspeciesare assumedlinearly additive. due to the complexity/of such measurement,it is not usually Equation(4) explicitly includesthe effect of compositionon known.Twoapproaches for estimating H200]content havebeen crX, sp ando'x,b•p. Severalapproaches havebeenpresented in the usedin this study.One is basedon the assumptionthat aerosol literature for separating the light-scatteringcontributionsof exists at thermodynamicequilibrium;the secondis basedon aerosolof different compositions.Sloane's [1983, 1984, 1986] experimental datadescribingmetastableequilibrium. approach, incorporated into the Elastic Light-scattering In the first approach,both the aerosolchemicalcomposition InteractiveEfficiency(ELSIE) computerprogram,considersfive andthe H20[• ] contentare estimated with the assumption that aerosolcomponents: SO42-, NO j, organiccarbon,elemental thermodynamicequilibriumgovernsthe distributionof volatile carbon, and residual material. Aerosol particle compositionis inorganicspeciesandH20, betweenthe gaseousand solidand/or treatedas homogeneous or distributedbetweenan insolublecore liquid phasesof the aerosol.The assumptionof thermodynamic andan aqueouscell whenH20 insolublecomponents arepresent. equilibriumin a closedsystemunder constanttotal pressureP

Er=y. mi[ 7 3Q(n'O'g'dp) 1 dM•i ,dlogdp ], i•:H20 i 0 2dp,oi Mi dlogdp

In the latter case,Mie light- scatteringtheoryis appliedto and temperatureT takesthe form of a constrainedminimization estimateQ for the case of concentricspheres.The angular problemof the form

integrationis between0ø_2), thenone "section"

is assumed to exist.

In this work, SEQUILIB+ is usedto answerthe following gaseousphase

NH3, HC1,HNO3, andH20 liquid phase

H20, NH •, SO42-, NO3', H+,Na+,CI', HSO 4, andH2SO4 _

solidphase

question:given the ambient RH and T as well as the ambient + , Na+, and CI-, concentrations of particulate SOl-, NO j, NH 4 what is the physical state and molecularform of the ambient aerosol at thermodynamic equilibrium? The model simultaneously estimatesthe gas phaseconcentrations of NH 3,

HNO3,andHC1aswell asthe H2001whicharein equilibrium with the measuredambientaerosolparticulatematerialand the ambient RH.

AlthoughSEQUILIB + providesa tool for obtainingmolecular composition from the measured ionic composition for the ambient aerosol particles in a systematicway, there is no certaintythat the ambient aerosolcompositionis the one that The solid phaseconstituents are presentas long as RH is correspondsto thermodynamicequilibrium. The existenceof belowthedeliquescence relativehumidity(DRH) for the specific metastableparticleshas been verified in the laboratory[Tang mixtureof components present.The aerosolis assumed to be dry 1980; Spann and Richardson, 1985] and in the ambient for RH values less than the smallest DRH value of the environment[Rood et al., 1989]. For this reason,in this paper, thermodynamically possiblesalts.Then, dependingon the molal SEQUILIB+ is usedin combinationwith experimentaldata from ratio RM = ( Mo,NH • + Mo,Na.)!Mo,SO 4 •' the aerosol is Tang and Munkelwitz [1991, 1994]. Becausethe experimental characterized as sulfaterich (RM< 2 ) or sulfatedeficient(R• dataof TangandMunkelwitz [1991,1994]consider H2001 in the _>2). For the former,when 1 < R• < 2, H2SO 4 is partially aerosolat RH < DRH for the examinedchemicalcompounds,it neutralized,while the remainingH2SO4 formsHSO •. The ratio is implicit that thesedata correspondto metastableequilibrium conditions.The underlyingassumptionin the approachfollowed

Na2SO4,NaHSO4,NaC1,NaNO3,NH4C1, NH4NO3,(NH4)2SO4, NH4HSO4,and (NH4)3H(SO4)2

of HSO• to SO42-is controlled by thermodynamic equilibrium.

WhenR• < 1, the solutionis very acidic,so H2SO4 is not even partially neutralized;the remainingfree H2SO4 dissociatesto HSO X and H +. SEQUILIB predictsthe concentrations of all the thermodynamicallypossible speciesin the system, as listed

above,as well as the total aerosoland "unbound"H20[11 concentrations, when the total compositionin both gaseousand solid/liquidphasesis known.The total compositionis then used as input, and SEQUILIB partitionsthe volatile speciesbetween the gaseousandthe solid/liquidphases,dependingon T andRH.

This is not the casein the currentstudy,where only the solubleinorganicparticulatecompositionis known from IC analysisof the filter/impactorsamples.In this case,the measured concentrations of NH 4, + NO j , and C1-in the aerosolparticles constrain the initial problem solved by SEQUILIB. The constrained problemis solvedby a new algorithm,SEQUILIB+, whichtakesasinputonlythemeasured particlecomposition with respect to NH 4, + NO j , CI-, as well as ions corresponding to nonvolatilespecies,i.e., Na+ and SO]-. SEQUILIB+ uses multiplebisectiontechniquesto solve the systemof algebraic equationsthat correspondto the thermodynamicequilibrium assumption. In SEQUILIB+ the aerosolis assumedto consistof different

in this paperis that the dry aerosolcomposition corresponds to thermodynamic equilibriumaspredictedby SEQUILIB+ at RH < 40%, but the particlescould be either at thermodynamic or at

metastable equilibrium with respect to H2001content at higher RH values.

For the caseof metastableparticles,Tang and Munkelwitz [ 1991,1994]havepresented polynomialfits of theform N

W=•"fi',,a• n=0

(7)

N

p=

n=0

(8)

for the water activity an and densityp of concentrated solutions for a numberof single-saltaerosolparticles(Table 1). B'n andAn are the coefficientsof the fitted polynomial,N is the orderof the polynomial, and W is the percent by weight of solute in the droplet. These polynomial equations have been based on measurementscharacterizingmass changewith changesin RH for single particles of known compositionsuspendedin an electrodynamicparticle balance. When W as a function of RH

"sections,"where a "section" denotesa certainaerosoltype. Specifically,if the aerosolis sulfaterich (R• < 2), then in the particleareknown,the H2001 case that NO j is also presentin the measuredaerosol,the andthe dry massof the aerosol content of the droplet can be determined at a certainRH. aerosol is likely externally mixed. This has been concluded For specieswhere polynomialexpressions (equations(7) or empirically, after repeated runs of SEQUILIB [Pilinis and (8)) do not exist, W at variousRH valuesis estimatedby using Seinfeld,1987], which have shownthat internallymixed aerosol solubility data of bulk solutionswhich give the corresponding is unlikelyto contain NOj in SO42richcases. ThuswhenRM< molalityM and an[Robinsonand Stokes,1959; Hamer and Wu, 2, the aerosolis composed of at leasttwo "sections."When there 1972; Goldberg,1981]. Cohenet al. [1987] havecombinedbulk is enoughSO24 - to yieldfreeH2SO 4 (RM < 1), thenthe first informationwith measurementson particles suspendedin an " section' ' is composedof H2SO4,NH 4, + Na+ and HSO X, while electrodynamicparticle balance, and they have presented the secondsectionconsistsof the remainingNH + 4, CI-, and equationsto describean as a functionof M. For a certainan (or NO j. If all H2SO 4 is partiallyneutralized (1 _< RM < 2), then equivalentlyRH), W is estimatedin the followingway: The ratio

17,014

KOLOUTSOU-VAKAKIS

ET AL.: MODELING OF AEROSOL PROPERTIES

Table 1. PolynomialCoefficientsfor Estimatingthe PercentWeight,Density,andReal RefractiveIndex of SelectedCompounds. H2SO4a

(NH4)2SO4

NH4HSO4

(NH4)3H(SO4) 2

Valid W range (%)

0-90

0-78

0-97

0-78

Valid RH range (%)

0-100

39-100

2-100

B'0

0.72

-156.58

B'•

-0.74

B'2

Na2SO4

Na2SO4

NaHSO4

0-40

40-67

0-95

39-100

82-100

58-82

2-100

97.80

-377.00

-321470.12

3915.76

96.03

2717.65

-33.77

4726.43

1240988.24

-20991.68

-54.94

-0.92

-12143.08

-154.94

-19557.34

-935551.03

43056.84

82.50

B'3

8.25

27748.02

617.13

41912.82

-2328552.43

-39402.00

-448.54

B'4

-19.56

-35117.49

-1339.22

-49785.38

5019071.64

13516.72

882.16

19.91

23359.52

1345.77

31070.56

-3586166.41

.....

840.12

B'6

-7.646

-6408.25

-532.66

-7989.94

911680.11

....

282.82

Ao

0.998

0.9971

0.9971

0.9971

8.871e-3

7.56e-3

A•

0.557

5.92e-3

5.87e-3

5.66e-3

3.195e-5

2.36e-5

A2

1.24

-5.036e-6

-1.89e-6

2.96e-6

2.28e-7

2.33e-7

A3

-3.79

1.024e-8

1.763e-7

6.68e-8

A4

6.07

As

-3.23

Source:Tang and Munkelwitz [ 1991, 1994] Coefficientsrefer to metastableparticles. aBray,[1970]

of the weight of salt over the weight of solvent at M corresponding to a givenRH is

1000

Rso • = R•+ (R2 - R•) y

(12)

andVso I is givenby

(9)

1

[MW•+(MW2- MW•)y]

(13)

Pso•

Thenthepercentweightof solutein solutionis givenby

wheretOsol is the densityof the solution[Tang and Munkelwitz,

W= 1'•--•100 3.4. Estimate

of the Refractive

(10) 1994].

Index

3.5. Estimate

of Particle

Size as a Function

of RH

The refractive index n changes when the chemical

In additionto changes in n, H20m uptakealsoresultsin composition changes, asisthecasewiththeaddition of H20[l I to changesin the size of the aerosolparticles.From a material

the aerosol. The refractive

index at different

RH

values is

balanceon the particle, the particle's growth under increasing estimatedby usingthe conceptof molal refractionR [Stelson, RH conditionsis describedby the following equation: 1990],whichis definedfor a condensed phaseas /,/2--1

R= Vn2+2

dp, wet /3 Po = 100

• dp, dry

(14)

WPsol

where v is the molal volume and n is the refractive index. For a

binary aqueoussolutionof solutemole fractiony, the molal refractionRso • may be expressedas the sumof the partialmolal refractions of the solventR• andthe soluteR2:

wheredp,we t isthediameter of thedroplet, dp,dry isthediameter of the dry particle,andPo is the densityof the dry salt [Tang and Munkelwitz,1977].

KOLOUTSOU-VAKAKIS

ET AL.: MODELING OF AEROSOL PROPERTIES

3.6. Predictionof Chemical,Physical,and Optical Properties of Ambient

17,015

The compositereal refractive index for the dry and wet aerosol particles, nrnixt, is estimated afterStelson[ 1990]fromthe

Aerosol

Ambientaerosolis a mixtureof many components. Because Rmixt wet , (15) of the lack of datathat describepropertiesof chemicalmixtures •'mixt, wet •'mixt, wet relevantto atmospheric aerosolparticles,certainassumptions and mixing rules are needed. These assumptionsfacilitate the where Rrnixt, wetand 1,'rnixt, wetare estimatedafter Tang and estimates of nmixt and changesin particlesizedistribution,for the Munkelwitz,[ 1994]' ambientaerosol,as functionsof Rid. The subscript"mixt" indicatesthe propertiesof a mixtureof chemicalcompounds. Rmixt, wet =RH20 +(Rmixt, dry--gH20)Ymixt, wet (16) Data describingsingle-saltaerosolare usedin combinationwith mixing rules to estimatethe necessaryparametersfor mixtures. 1 wet= •(MWH20 + (MWdry,mixt - MWH20)Ymixt, wet) AerosolSpecificThermodynamic and MetastableEquilibrium ¾mixt, JOmixt, wet (ASTME) code was developedto perform the necessary calculations.ASTME is used to parameterizethe aerosol

nmixt, wet(1+2

' )l/2(l_Rmixtwet)-l/2

properties to determine cr,•,s p andcr,•,bsp for theambient aerosol,

dryandMWmixt, dryareestimated, respectively, asthe after the dry compositionof the aerosolhas been determined whereRmixt, moleandmassweightedaveragesof Ri andMW i in themixture. from SEQUILIB+. Estimate of H20[• 1content at selectRH valuesis thefirststep Implicit in this approachis the assumptionof a homogeneous to determinenmixt and changesin the midpointdiameterin each internalmixture of all the aerosolparticlesolutes,where in the of H20[•],the mixtureis treatedas a binarysolution, sizebinof theparticle sizedistribution Adp.If thedryaerosol is presence anexternal mixture,ni andAd•,,i for eachchemical compound i with the solutehavingthe averagepropertiesof the dry mixture areestimated fromequations (11) and(14),respectively; crx,•p,iof chemicalcompounds. andcr,•,b,i arethenestimated foreachsinglechemical compound As previouslymentioned,the particlesize distributionfor the i; cr,•,•p and cr,•,•, for the entireaerosolare determined by ambientaerosolwasmeasuredduringintensivesamplingperiods low-pressure cascadeimpactor.Aerodynamic summingcr,•,sp, i and cr•,•p,ivaluesat the respective As, as with a seven-stage diameters dp, awereconverted to Stokes diameter dpassuming a suggested by White[ 1986]. However, if the dry aerosol is an internal mixture, then

dryparticle density pp= 1.7g/cm 3,whichis representative of

inorganic compounds. additional assumptions areneededsothat Wm•xt, ,Omixt, nmixt, and NH • andSO,•- containing

Adp,mixt canbeestimated. To estimate Wm•xt, thedrymassof the mixture is assumed to be the sum of the masses of the individual

Subsequently,the observedsize distributionswere fitted to a

lognormaldistributionusing an 2'2 minimizationtechnique

describedby Whitby [1978]. One or more modes are fitted dependingon the shapeof the discreteimpactordistribution.If the aerosolwas assumedto be externallymixed, then the size distributionthat corresponded to eachcomponentwas considered separately. If an internal mixture was considered, then all differentRH values.Suchan approachhas limitationsbecause the differentcompounds comprising the internalmixturemay componentswere assumedto follow the same particle size distribution.Each mode of the distributionis describedby the dry compoundsand the wet massof the mixture is the sumof the individualwet massesof each of the compoundsat each RH. Similarassumptions are madeto estimate,Omixt as the ratio of the totalmassoverthe totalvolumeof the mixture[Stelson,1990] at

reactwith eachotheras well as with gaseous compounds. The

presence of additional H20[1 ] [Winkler, 1973,1988]athigherRH values

further

facilitates

such

reactions.

Therefore

the

modalvolume/mass ornumberconcentration, Vmoa½ (gm3/cm3 air)

orNmoa½ (particles/cm 3 air),thegeometric meandiameter dgvor

standard deviation of themodearound the compositionof the mixture may change to include new dg,,andthegeometric mean diameter, s•. Conversion between total volume/mass and components that may have differenthygroscopicproperties.

However,lackof availabledatanecessitates suchan assumption. totalnumberparametersis doneby Parametersfor single salts that are used in this study are

Indgv= Indgn+ 3 In2Sg

presentedin Table 2.

Table2.Parameters UsedforthePrediction of EY2. spandEY)[,bsp inthisStudy Species

Molecular

RealRefractive

Dry Density

Molal

Weight (MW)

Index(n)a

(,/o9) g/cm 3

Refraction (R)

132.14 115.11 247.14 142.04 120.06 84.99 80.04 18.01

1.53 !.47 1.52 1.48 1.49b 1.46 • 1.55 1.33

1.77 1.78 1.79 2.68 2.44 2.26 1.73 !.00

23.50 18.38 20.95 15.13 14.38 11.22 15.13 3.717

....

1.50a

1.00a

(NH4)2SO4 NH4HSO 4

(NH4)3H(SO4)2 Na2SO4 NaHSO4 NaNO3 NH4NO 3 H20

Nonhygroscopic

___

component

Sources:Weast[1988], $telson[1990], and TangandMunkelwitz,[1991, 1994]. aRefractiveindicesarereportedat 2 ofNa+ (589.3nm). bEstimated frommolalrefraction, • Mean of reportedvalues

aConsistent withcommonly observed nonhygroscopic organic compounds in theatmosphere

(18)

17,016

KOLOUTSOU-VAKAKIS ET AL.: MODELINGOFAEROSOLPROPERTIES Table 3. MeasuredArithmeticMean Values s and o'xb., ß . ofo o'x ,P , P. for AmbientAerosol

Particles (d•,_