size-dependent ccn closure during icartt 2004 1 cloud...

Size-dependent CCN Closure during ICARTT 2004 1

Cloud Condensation Nuclei (CCN) closure during theICARTT 2004 campaign: a) effects of size-resolvedcomposition.

Jeessy Medina

School of Chemical & Biomolecular Engineering, Georgia Institute of Technology

Athanasios NenesSchools of Earth & Atmospheric Sciences and Chemical & Biomolecular Engineering, GeorgiaInstitute of Technology

Rafaella-Eleni P. Sotiropoulou

School of Earth & Atmospheric Sciences, Georgia Institute of Technology

Laura D. Cottrell

Institute for the Study of Earth, Oceans, and Space, University of New Hampshire

Luke D. Ziemba


Pieter J. Beckman


Robert J. GriffinInstitute for the Study of Earth, Oceans, and Space, and Department of Earth Sciences,University of New Hampshire

Abstract.Measurements of Cloud Condensation Nuclei (CCN), aerosol size distributionand chemical composition were obtained at the UNH-AIRMAP ThompsonFarms site, during the ICARTT 2004 campaign. This work focuses on theanalysis of a week of measurements, during which semi-urban and continentalair were sampled. Predictions of CCN concentrations were carried out using“simple” Kohler theory; the predictions are subsequently compared with CCNmeasurements at 0.2%, 0.3%, 0.37%, 0.5% and 0.6% supersaturation. Usingsize-averaged chemical composition, CCN are substantially overpredicted (by35.8±28.5%). Introducing size-dependent chemical composition substantiallyimproved closure (average error 17.4±27.0%). CCN closure is worse duringperiods of changing wind direction, suggesting that the introduction ofaerosol mixing state into CCN predictions may sometimes be required.Finally, knowledge of the soluble salt fraction is sufficient for description ofCCN activity.

1. Introduction

It is well accepted that aerosol-cloud-climate inter-actions constitute a major source of uncertainty for as-sessments of anthropogenic climate change [Intergov-ernmental Panel on Climate Change (IPCC), 2001;Charlson et al., 1992; Haywood and Boucher , 2000;Ramanathan et al., 2001]. Understanding what con-trols the ability of ambient atmospheric aerosol to actas cloud condensation nuclei (CCN) is at the heartof this problem, as all physically-based treatments ofcloud droplet formation in global climate models [e.g.,Abdul-Razzak and Ghan, 2000, 2002; Nenes and Sein-feld , 2003; Fountoukis and Nenes, 2005] rely on predic-tions of CCN concentrations.

It is necessary to expose the particle to a supersat-uration long enough for it to reach its critical radius.Each aerosol, when exposed to a water vapor super-saturation above a characteristic “critical” supersatu-ration, sc, experiences unconstrained growth or, acts asa CCN and forms a cloud droplet. sc depends on theaerosol dry size and chemical composition, and is com-puted from thermodynamic arguments first introducedby H. Kohler [Kohler , 1921, 1936]. This “Kohler the-ory”, originally developed to describe the CCN activityof salt particles, has since been refined to address thebehavior of multicomponent atmospheric aerosol. As aresult, theory can be quite complex and may require theknowledge of poorly constrained parameters. This is es-pecially true if the aerosol contains substantial amountsof water-soluble organic carbon (WSOC). WSOC canact as a surfactant [Decesari et al., 2003], potentiallydepressing droplet surface tension [Facchini et al., 1999]and the condensation rate of water [e.g., Nenes et al.,2002]. WSOC may also contribute solute [Shulmanet al., 1996], although low solubility [e.g., Nenes et al.,2002], dissolution kinetics [Asa-Awuku and Nenes, in re-view] and the presence of soluble salts can diminish itsimportance. Since each of these effects alone can eitherenhance or diminish the CCN activity of the aerosol,comprehensively treating the effect of organics on CCNactivation can be quite complex and subject to substan-tial uncertainty.

“CCN closure”, or comparison of predictions withobservations of CCN concentrations, is the ultimate testof Kohler theory and has been the focus of numerousstudies. Liu et al. [1996] made ground measurements ofCCN at Chebogue Point in 1993 as part of the NorthAtlantic Regional Experiment (NARE) intensive. CCNat 0.4% were measured using a static diffusion cloudchamber from DH Associates. CCN were predicted us-

ing size distributions from a PMS passive cavity aerosolspectrometer probe (PCASP-100X). Aerosol composi-tion was obtained from filter samples. Two distinctgroups of aerosols were identified, one likely from cloudprocessing; despite this heterogeneity, good closure (i.e.,within experimental uncertainty) was achieved for 75%of the time.

Covert et al. [1998] conducted ground-based CCNmeasurements at 0.5% supersaturation using a ther-mal gradient diffusion cloud chamber during the FirstAerosol Characterization Experiment(Cape Grim, Tas-mania, 1995). Differential mobility particle sizers wereused to measure the size distribution. On average,Covert et al. [1998] obtained good closure, with a slight(20%) tendency for overprediction.

During the second Aerosol Characterization Exper-iment (ACE-2) over the northeast Atlantic (Tener-ife, Spain), Chuang et al. [2000a] made airborne CCNmeasurements onboard the CIRPAS Pelican. Duringthis study, the Caltech continuous-flow CCN counter[Chuang et al., 2000b] measured CCN at 0.1% supersat-uration; size distributions were obtained with the Cal-tech Automated Classified Aerosol Detector (ACAD)[Collins et al., 2000] along side a Particle MeasuringSystems PCASP-100X. Chemical composition was notmeasured in-situ but constrained from ground-based fil-ter samples. Closure was not achieved, as the CCNconcentration was overpredicted by more than tenfold.

Ground-based CCN measurements were performedby Cantrell et al. [2001] at the Kaashidhoo ClimateObservatory during the Indian Ocean Experiment (IN-DOEX) intensive field phase (February - March, 1999).Aerosol chemical composition were obtained using anaerosol time-of-flight spectrometer (ATOFMS) and fourmicroorifice uniform deposit cascade impactors (MOUDIs);aerosol size distributions were measured using a TSISMPS. CCN spectra were obtained using a CCN Re-mover [Ji et al., 1998], operated in series with theSMPS. Closure was achieved in 8 out the 10 cases; dis-crepancies increased with organic mass fraction.

Roberts et al. [2003] performed measurements duringthe Cooperative LBA (Large scale Biosphere-AtmosphereExperiment in Amazonia) Airborne Regional Exper-iment 1998 (CLAIRE-98). Ground measurements ofCCN spectra were made using a thermal diffusion cloudchamber for supersaturations between 0.15 and 1.5%.Additional instrumentation included a multi-stage cas-cade impactor to determine chemical composition anda scanning mobility particle sizer to determine aerosolsize distributions. The CCN activity of the LBA aerosoldepended primarily on the aerosol soluble fraction; po-

2


tential impact of surface tension depression from thepresence of WSOC (assessed by applying the correla-tion of Facchini et al. [1999]), was found to be impor-tant only at low supersaturations.

Snider et al. [2003] and Dusek et al. [2003] analyzedaerosol samples for CCN closure during the ACE-2 cam-paign. Snider et al. [2003] compared observations for 5study days collected at a coastal site and on an aircraftover the eastern Atlantic Ocean. After accounting fordifferences between the mobility and sphere-equivalentdiameters, CCN agreed with predictions for two sam-pling days unaffected by continental pollution. Duseket al. [2003] analyzed data collected in southern Portu-gal and showed that WSOC constituents at the locationconstituted less than 10% of the total aerosol mass andneglected their effect on CCN activity. Despite this,calculated CCN were overestimated on average by 30%,comparable to the measurement and prediction uncer-tainty.

VanReken et al. [2003] performed airborne CCNmeasurements during the Cirrus Regional Study ofTropical Anvils and Cirrus Layers-Florida Area CirrusExperiment (CRYSTAL-FACE) field campaign (July,2002). Using two continuous-flow streamwise tempera-ture gradient chambers [Roberts and Nenes, 2005], CCNwere measured at 0.2 and 0.85% supersaturation. Theaerosol was most frequently marine; assuming a com-position of ammonium sulfate, closure was achieved towithin 5% at 0.2% supersaturation, and 9% at 0.85%supersaturation.

Rissman et al. [2006] performed an inverse aerosol–CCN closure and explored its sensitivity to chemicalcomposition and mixing state. The observations wereobtained during the 2003 atmospheric radiation mea-surement (ARM) aerosol intensive observational period(IOP) at the southern great plains (SGP) site in Ok-lahoma. Optimum closure was achieved when the pop-ulation of aerosol is treated as an external mixture ofparticles, with the insoluble material preferentially dis-tributed in particles less than 50 nm diameter.

Broekhuizen et al. [2006] measured CCN concen-trations using a continuous thermal-gradient diffusionchamber in Toronto, Canada. The CCN chamber mea-sured CCN concentrations at 0.58% for 4 days. Aerosolsize distributions were measured using a TSI SMPS andan APS system; chemical composition was simultane-ously measured with an Aerodyne aerosol mass spec-trometer (AMS). Considering the size-dependant com-position, and assuming internally-mixed aerosol, closure(on average) was achieved to within 4%.

Stroud et al. [in press] measured aerosol size distri-

bution with a TSI SMPS, chemical composition withan AMS and CCN concentration with a University ofWyoming thermal-gradient static-diffusion cloud cham-ber (CCNC-100A) [Delene and Deshler , 2000] duringthe Chemical Emission, Loss, Transformation and In-teractions with Canopies (CELTIC) field program atDuke Forest in North Carolina. A numerical modelof the CCN instrument [Nenes et al., 2001] was usedto perform an aerosol-CCN closure study, and to con-strain the value of the water-vapor mass uptake coeffi-cient. CCN predictions were within a factor of two ofthe observations. Sensitivity simulations suggest thatassuming either an insoluble organic fraction or exter-nal aerosol mixing were both sufficient assumptions toreconcile the model bias. The chamber model best re-produced the kinetics of droplet growth (i.e. timescalerequired for the scattering signal in the instrument viewvolume to peak) when the water vapor mass uptake co-efficient was set equal to 0.07; this conclusion was in-sensitive to the assumption of chemical composition (orthe degree of closure).

Despite the numerous CCN studies, a comprehen-sive CCN climatology is still lacking. There is a strongneed for CCN closure studies that cover a wide rangeof seasons and aerosol types; this precludes a robustquantification of CCN prediction uncertainty, especiallyfor simplifying assumptions (e.g., aerosol mixing state,variation of composition with size and affinity of car-bonaceous material with water) commonly taken inglobal models. This study, together with the work ofBroekhuizen et al. [2006] and Stroud et al. [in press] isa step towards addressing this problem. We performa detailed CCN closure study using concurrent mea-surements of aerosol size distribution, chemical compo-sition and CCN concentrations during the New Eng-land Air Quality Study – Intercontinental Transportand Chemical Transformation 2004 (NEAQS – ITCT2004) as part of the International Consortium for At-mospheric Research on Transport and Transformation(ICARTT). ICARTT was a series of coordinated exper-iments, focused on understanding the properties of re-gional aerosol and their effects regional air quality andclimate. NEAQS – ITCT 2004 took place during themonths of July and August and focused on air qual-ity along the Eastern Seaboard and the aging of NorthAmerican emissions as they are transported out intothe North Atlantic. This paper focuses on the impor-tance of having size-resolved chemical composition onCCN closure; a companion paper [Nenes and Medina,in preparation] presents measurements of size-resolvedCCN activity which are then used to gain insight on the

4 Medina et.al

characteristics of the aerosol organic species, infer theCCN mixing state and explore compositional effects ondroplet growth kinetics. The main difference betweenthis study and previous work [e.g., Broekhuizen et al.,2006; Stroud et al., in press], is that it refers to a differ-ent location (in space and time) with aerosol exposedto a different mix of influences.

2. Observational Dataset

2.1. Study location

The measurements were obtained at the Univer-sity of New Hampshire (UNH) Thompson Farm (TF)AIRMAP Observing Station (http://airmap.unh.edu/).The site is located in Durham, NH, approximatelytwo miles south of the University of New Hampshire(43.11N, 70.95W, elevation 75ft). TF is characterizedby fields and surrounding forest with some local an-thropogenic influence. The aerosol at this location is acomplex mixture of organic and inorganic material andis ideal for a CCN closure study. Measurements for thiscampaign were done from late July to mid August. Thisstudy focuses on a week of measurements, from August05 to August 11, 2004.

2.2. Instrumentation Setup

Figure 1 is a schematic of the field instrumentationsetup. A Droplet Measurement Technologies (DMT)streamwise thermal–gradient cloud condensation nucleicounter [Roberts and Nenes, 2005; Lance et al., 2006]was used to measure CCN concentrations. A scanningmobility particle sizer (SMPS, TSI 3080) composed ofa condensation particle counter (CPC, TSI 3010) and along differential mobility analyzer (DMA, TSI 3081L)measured the dry aerosol size distribution. An Aero-dyne Aerosol Mass Spectrometer (AMS) concurrentlymeasured aerosol chemical composition. Air was drawnfrom the top of a 60 ft tower with a 2.5 µm URG cycloneat 10 L min−1 through a 1/2” OD copper tube. Inlet airsampled from atop a 60 ft tower was split into two flows,one for measurements of size-resolved chemical compo-sition (AMS) and the other for measurements of sizedistribution (SMPS) and CCN activity (CFSTGC).

2.3. AMS chemical composition measurements

The AMS [Jayne et al., 2000] can measure the aerosolsize distribution (aerodynamic diameter) and chemicalcomposition. Aerosol entering the instrument passesthrough a critical orifice and are aerodynamically fo-cused into a narrow particle beam. The particles subse-

Figure 1. Schematic of the setup used for measuringthe aerosol size distribution (SMPS), chemical compo-sition (AMS) and CCN concentration (CFSTGC).

quently pass through a rotating chopper and are sizedaccording to their time-of-flight across a vacuum cham-ber. After sizing, the particle beam is directed on resis-tively heated surface; particles are then vaporized andionized with a 70 eV electron impact ionization source.The ions are filtered and detected with a quadrupolemass spectrometer and secondary electron multiplier.During ICARTT, the AMS was operated with a 130 µmdiameter critical orifice and a 1% chopper rotating atapproximately 100 Hz. The vaporizer was maintainedat 550˚C. The chemical composition and size distrib-ution of the aerosol was alternately measured every 30seconds and averaged data was saved every 10 minutes.

Mass loadings for sulfate, nitrate, ammonium and or-ganics were calculated according to Allan et al. [2003].Based on findings from other aerosol instruments withsimilar size cuts [e.g., Alfarra et al., 2004; Zhang et al.,2005], a 50% collection efficiency is assumed (the rest ofthe particles being lost to particle bouncing). The col-lection efficiency used is consistent with concurrent 24hour bulk filter measurements [Cottrell et al., in prepa-ration]. The collection efficiency is likely the largestsource of uncertainty for AMS measurements [Alfarraet al., 2004; Zhang et al., 2005], as it has been foundto vary with factors such as composition, relative hu-midity, and aerosol acidity. The relative ionization ef-ficiency of different compounds has been characterized


in other studies [Jimenez et al., 2003; Alfarra et al.,2004] and is thought to be related to the electron im-pact cross section of each compound. In this study,relative ionization efficiencies of 1.4, 1.2, and 4.0 wereadapted for organics, sulfate, and ammonium, respec-tively. Size calibrations were performed using NISTtraceable polystyrene latex spheres at the start of thecampaign. Mass calibrations were performed usingclassified ammonium nitrate aerosol every 3 to 5 daysthroughout the campaign. AMS configuration, opera-tion, and data are detailed by Cottrell et al. [in prepa-ration].

2.4. Dry Size Distribution Measurements

The TSI 3080 SMPS is an electrical-mobility basedsystem capable of high-resolution size distribution mea-surements. Figure 1 illustrates its major components.Dried ambient aerosol is charged using a Kr-85 neutral-izer (TSI 3077A). The charged aerosol stream entersa DMA where aerosols are classified based upon theirelectrical mobility and counted with a TSI 3010 con-densation particle counter (CPC). The DMA voltagecan be scanned and a particle size distribution is ob-tained for mobility diameters between 10 and 300 nm.Throughout this study, size distribution scans were ob-tained every 135 seconds (120 seconds for the voltage“upscan” and 15 seconds for the voltage “downscan”).The sample flow rate was set to 1 L min−1 and thesheath-to-aerosol flow was maintained at 10:1.

2.5. CCN measurements

The DMT CCN counter [Roberts and Nenes, 2005;Lance et al., 2006] is a cylindrical continuous-flow stream-wise thermal-gradient diffusion chamber. A constantstreamwise temperature gradient is applied at the in-strument walls; a constant centerline supersaturationdevelops from the difference between thermal and watervapor diffusivity. The instrument is able to generate su-persaturations between 0.07% and 3% through precisecontrol of the air flow and the streamwise temperaturegradient. Figure 1 illustrates the components and flowdiagram of the CCN instrument. The inlet flow is firstsplit into a “sheath” and a “sample” flow. The “sample”is directed the center of the flow column, whereas the“sheath” flow is filtered and humidified prior to its entryin the chamber, resulting in a “blanket” of humidifiedclean air around the “sample” flow. Both flows travelthrough the column exposing aerosol to a constant su-persaturation. Aerosol whose sc is below the instrumentsupersaturation will activate into droplets. An OpticalParticle Counter (OPC) counts and sizes the droplets at

Figure 2. HYSPLIT backtrajectories for characteristictypes of airmasses sampled during the period of interest.

the outlet of the column, using a 50 mW, 658 nm laserdiode. All scattered light between 35˚ to 145˚ is col-lected and detected with a photodiode. A multichannelanalyzer bins the detected droplets into 20 size classes(0.75 µm to 10 µm). The particle counts are averagedover 1 second. Throughout the measurement period,the CCN counter operated at a flow rate of 0.5 L min−1

with a sheath-to-aerosol flow ratio of 10:1. CCN con-centrations were measured at 0.2%, 0.3%, 0.37%, 0.5%and 0.6% supersaturation. Concentrations were mea-sured at each supersaturation for 6 minutes, yielding aCCN spectrum every 30 minutes. We used calibrationswith (NH4)2SO4 aerosol to characterize the instrumentsupersaturation; we also operated under conditions forwhich the droplets had enough time to grow and becomedistinguishable from the interstitial aerosol.

3. Data Overview

3.1. Airmasses sampled

Analysis of backtrajectories from the HYSPLIT(http://www.arl.noaa.gov/ready/hysplit4.html) modelsuggests that the airmasses sampled at TF from Au-gust 5 thru 8 resided mostly within the boundary layerthroughout its transit from the north (Figure 2a). FromAugust 9 to 10, boundary layer air from the west wassampled (Figure 2b, c), while in the evening of August10, the wind direction shifted, transporting southerlyboundary layer air (Figure 2d). The daily wind direc-

6 Medina et.al

Table 1. Origin of airmasses sampled during theperiod of interest.

Date Wind Airmassdirection origin

08/05 N Boundary Layer08/06 N Boundary Layer08/07 N Boundary Layer08/08 N Boundary Layer08/09 W-WNW Boundary Layer08/10 W-WSW Boundary Layer08/11 SW-SSW Boundary Layer

tions are given in Table 1.

3.2. Aerosol Measurements

Figure 3 shows a timeseries of the total mass loadings(µg m−3) of sulfate and organic measured by the AMS.The inorganic fraction is always dominated by sulfateand ammonium, their relative amounts suggesting thatthe aerosol is primarily acidic. As a result, nitrate tendsto partition in the gas phase, hence the low levels ofaerosol nitrate. On August 8 (around 12:00 am), thesulfate mass fraction decreased, leading to partial neu-tralization of the aerosol and small increase in aerosolnitrate (not shown).

Size-resolved chemical composition was also obtained,for aerosol larger than 80nm (Figure 4); composition atsmaller aerosol sizes was too uncertain, as there was notenough mass to obtain a strong enough signal in thespectrometer. The size-resolved measurements showthat the sulfate ranges between 20-30% for most of theperiod (Figure 4). The sulfate mass fraction substan-tially increased between August 10 and 11 (Figure 4),accompanied by a shift of the aerosol size distributiontowards larger sizes (Figure 5). The observed shift inchemical composition and size distribution is consistentwith the HYSPLIT backtrajectories; for most of the pe-riod, polluted continental air was sampled, with signif-icant amounts of freshly condensed secondary organicaerosols (SOA) mixing with roughly equal amounts ofprimary organic carbon (POA) [Cottrell et al., in prepa-ration]. From August 10 to 11, the airmass was primar-ily semi-urban (i.e., rural aerosol influenced by urbansources), hence the significant increase in sulfate with asize distribution consistent with aged aerosol.

Figure 6 presents the timeseries of total aerosol (CN)

Figure 3. Time series of the total mass loadings fromthe AMS.

Figure 4. Average sulfate mass fraction size distribu-tion. The shaded region indicates where measurementuncertainty is large.


Figure 5. Daily average aerosol size distributions,characteristic of the airmasses sampled.

as measured with the SMPS, between August 5 and11. For comparison, we present CN measurements fromthe AIRMAP site CPC (TSI 3010). Overall both in-struments follow each other very closely (on average towithin 10%).

The total condensation nuclei measurements duringthe entire campaign averaged about 5000 cm−3. Theaerosol concentrations tend to be higher at night andearly morning, when the boundary is shallow. Occa-sionally there are large “spikes” in aerosol concentra-tion, going to as high as 12000 cm−3, likely from the in-fluence of fresh, local combustion emissions (since theseevents often coincided with large increases in sulfatemass), perhaps with a combination of new particle for-mation. Throughout most of the measurement period(periods “A” and “B” in Figure 6), the number distri-bution peaks somewhere between 70 and 80 nm (Fig-ure 5); this changed in the latter part of the campaign(periods “C” in Figure 6), where the aerosol exhibitsthe characteristic bimodal character of aged aerosol.

3.3. CCN measurements

The timeseries of the CCN measurements are super-imposed on Figure 6. As expected, the CCN and CNqualitatively track each other, with concentrations be-ing lower as supersaturation decreases. Some of theCN spikes (e.g., around noon on August 9) are not ac-companied by a corresponding peak in the CCN obser-vations, suggesting that a fresh combustion source or

Figure 6. Time series of the CN and CCN measure-ments throughout the period of interest.

nucleation event may be responsible for the sudden in-crease in CN. The subsequent aging of the aerosol andincrease in CCN also support this hypothesis.

The CCN concentration timeseries can be dividedinto three periods (Figure 6), which coincide with thechange in airmasses sampled at TF (Figure 2). CCNconcentrations increase substantially during period “C”,while total CN do not; this is consistent with the changein the airmasses sampled (Figure 4c,d), as semi-urbanair tends to have more soluble material (i.e., sulfatemass fraction) than continental air, hence are more ef-ficient CCN.

Following the procedure outlined in Sotiropoulouet al. [2006], the CCN observations are fit to the“modified power law” CCN spectrum of Cohard et al.[1998, 2000],

F (s) =κCsκ−1

(1 + ηs2)λ

(1)

where F (s) is the CCN concentration at supersatura-tion s (i.e., the CCN “spectrum”), C is the total aerosolconcentration (cm−3), and κ, η and λ are unitless co-efficients determined from the fitting. The numeratorin equation 1 is the “power law” expression of Twomey[1959]; Cohard et al. [1998, 2000] introduced the denom-inator so that F (s) asymptotes to C at high s (≈10%for ambient aerosol). Table 2 displays the daily andperiod averages of the CCN spectra. Of all the pa-rameters, C (which scales with total CN) exhibits thelargest variability. The other parameters do not varymuch at all, suggesting that the normalized CCN spec-trum, F (s) = F (s)/C, is fairly constant throughout thesampling period. Although quite remarkable (given thesignificant shifts in aerosol size distribution and chem-ical composition observed between August 10 and 11),

8 Medina et.al

Table 2. Daily averages for CCN spectra. Parameters in table correspond to Equation 1.

Date C(×1000 cm−3) κ η λ

8/05/2004 4.21±1.01 3.18±0.22 3.13±0.29 1.07±0.088/06/2004 3.59±0.59 3.34±0.34 3.14±0.14 1.12±0.128/07/2004 4.14±1.49 3.04±0.09 3.01±0.03 1.02±0.038/08/2004 2.74±1.00 3.00±0.00 3.00±0.00 1.00±0.008/09/2004 2.77±0.69 3.09±0.18 3.04±0.07 1.03±0.068/10/2004 4.75±0.96 3.03±0.07 3.01±0.02 1.01±0.038/11/2004 3.20±0.32 3.00±0.00 3.00±0.00 1.00±0.00

Period average 3.63±1.17 3.10±0.21 3.05±0.14 1.03±0.07

this is largely due to F (s) scaling linearly to C and sub-linearly to all the other constants in Equation 1. Fur-thermore, even if the daily averages did not vary much,their uncertainty range became quite large during 8/5,8/6, 8/9 and 8/10.

4. CCN Closure

Before each closure calculation, the CCN measure-ments were screened to eliminate biased observations.Each 6-minute supersaturation segment is examined fori) minimal fluctuations in the flow chamber tempera-ture gradient, and, ii) stability of the flows. Fluctua-tions in temperature, if small enough, have a minimalimpact on instrument supersaturation and closure [Er-vens et al., in review]. If both criteria were satisfied,then the CCN concentrations are averaged over the last2-3 minutes of the supersaturation segment (to elimi-nate the effects of transients when switching supersat-uration) and a closure calculation is done. The SMPSsize distributions were averaged over each supersatura-tion segment (we have included only those scans thatencompass the times for which CCN measurements wereused); instead of fitting the measurements to lognor-mal distributions (as is often done) we directly use thebinned distribution obtained from the SMPS inversionroutine. Half-hour averaged aerosol chemical compo-sition is obtained from the AMS data, so there is onecomposition measurement for each supersaturation cy-cle.

When predicting CCN concentrations, we assumethat organics do not contribute solute (i.e., are insol-uble) nor act as a surfactant. The inorganic fraction ofthe aerosol is assumed to be ammonium sulfate with anaverage effective van’t Hoff factor, νs (which includesan estimate of the osmotic coefficient), of 2.5, obtained

from Pitzer activity coefficients [Brechtel and Kreiden-weis, 2000] for ammonium sulfate CCN with criticalsupersaturation between 0.2 - 0.6%. The influence ofother inorganic species (e.g., nitrate) is negligible. Thecritical supersaturation, sc for each particle with drysize d, is calculated from Kohler theory [Seinfeld andPandis, 1998],

sc =

[

256

27

(

Mwσ

RTρw

)3 (

Ms

ρs

) (

ρw

Mw

)

d3

εsνs

]1/2

(2)

where R is the universal gas constant, T is the ambienttemperature, σ is the surface tension of the CCN at thepoint of activation (assumed here to be equal to that ofwater), Ms is the molar mass of the solute and Mw, ρw

are the molar mass and density of water, respectively.The volume fraction of solute, εs, can be calculated asa function of the mass fraction of solute, ms and itsdensity, ρs,

εs =

ms

ρs

ms

ρs

+ 1−ms

ρi

(3)

where ρi is the density of the insoluble organic (hereassumed 1500 kg m−3). Particles are counted as CCNwhen their sc is less or equal than the supersaturationof the CCN counter. The predicted and measured CCNare then compared to assess closure.

4.1. Size-averaged chemical composition

The CCN closure using size-averaged chemical com-position (which is computed by weighing the compo-sition of each CCN by their concentration) is shownin Figure 7. The individual closure calculations are


Figure 7. Calculated vs. measured CCN concentra-tions. Size-averaged εs is used for the calculations.

color-coded by supersaturation. Figure 7 clearly showsthat there is a tendency for overprediction; for 0.2%,0.3%, 0.37%, 0.5% and 0.6% supersaturation, the pre-diction error is 37.2±28.2%, 22.0±21.4%, 16.9±23.1%,37.7±20.8% and 64.2±22.6%, respectively (average er-ror 35.8±28.5%). The CCN closure worsens (average er-ror > 40%) during August 5,9 and 10 (Table 3), whichcorrespond to days with shifting wind direction (Ta-ble 1). If the organic fraction is assumed to be soluble,the CCN prediction error would be even larger. Not ac-counting for size-resolved composition and aerosol mix-ing state are prime sources of CCN prediction error.The former is addressed below.

4.2. Size-Dependant chemical composition

The size-dependent CCN closure is shown in Fig-ure 8. The individual closure calculations are color-coded by supersaturation. Compared to the size-averagedclosure (Figure 7), specifying size-dependent chemi-cal composition substantially improves predictions; for0.2%, 0.4%, 0.47%, 0.5% and 0.6% supersaturation, theprediction error is 21.3±30.0%, 5.2±17.0%, 1.5±18.4%,15.6±19.8% and 43.3±26.0%, respectively (average er-ror 17.4±27.0%). The closure at 0.6% is considerablyworse than for the other supersaturations; this is likelyfrom lack of chemical composition measurements below80nm size (many of the CCN at 0.6% are from smallersizes).

Compared to the size-averaged closure, specifyingsize-dependent chemical composition substantially im-

Figure 8. Calculated vs. measured CCN concentra-tions. Size-dependent εs is used for the calculations.

proves the closure for all days except August 11 (Ta-ble 3). CCN closure is worst (average error > 25%)during August 9, 10 and 11 (Table 1). The average er-ror on August 11 is somewhat high, largely from thediscrepancy in CCN predictions at 0.6%. Like in thesize-averaged closure, August 9, 10 correspond to dayswith shifting wind direction (Table 1). If the organicfraction is assumed to be soluble, the CCN predictionerror would increase. Since size-dependent chemicalcomposition is considered in the closure, the assump-tion of internally-mixed aerosol is likely a key source ofthe residual error. This is addressed in a companionmanuscript [Nenes and Medina, in preparation].

5. Summary and Conclusions

Concurrent measurements of CCN (at 0.2%, 0.3%,0.37%, 0.5% and 0.6% supersaturation), aerosol size dis-tribution and chemical composition were carried out atthe UNH-AIRMAP Thompson Farms site, during theICARTT 2004 campaign. This work focuses on a weekof measurements (August 5 - August 11, 2004) for as-sessment of CCN closure. The aerosol sampled in thisstudy was either polluted continental from the GreatLakes area, or semi-urban from south of the samplingsite. The aerosol measured exhibits the characteristicsexpected for each type of aerosol; the polluted continen-tal air tends to have high concentrations of small parti-cles, and contain large amounts of organics. Semi-urbantends to have larger CCN, and higher sulfate mass frac-

10 Medina et.al

Table 3. Daily CCN prediction error (%) assuming uniform and size-dependant chemical composition.

Date % Error using uniform εs % Error using size-dependant εs

8/05/2004 41.37±21.32 16.66±19.188/06/2004 36.92±27.03 3.07±23.428/07/2004 26.95±26.81 -0.11±21.288/08/2004 32.75±25.19 21.23±23.298/09/2004 43.71±33.68 28.06±30.208/10/2004 46.76±34.21 30.32±28.138/11/2004 21.92±28.86 37.93±25.88

Period average 35.71±28.54 17.41±27.05

tion. The aerosol was assumed to be composed of aninsoluble and soluble fraction.

CCN concentrations are calculated using the mea-sured aerosol size distribution and chemical compositioncoupled with “simple” Kohler theory. Predictions arethen compared with measurements for supersaturationsranging between 0.2% and 0.6% and a variety of air-masses sampled. Using size-averaged chemical composi-tion, CCN are overpredicted on average by 35.8±28.5%.Introducing size-dependent chemical composition sub-stantially improved closure (average error 17.4±27.0%),although CCN concentrations still tend to be overpre-dicted. CCN closure is worse during periods of chang-ing wind direction; this suggests that aerosol mixingstate may need to be considered to improve CCN predic-tions. Introducing organic solubility and surface tensiondepression would worsen CCN closure. To the extentwhere our analysis applies, this suggests that organicsmay not contribute substantial amounts of solute.

Overall our study suggests that “simple” Kohler the-ory can be used to predict CCN concentrations; mostof the discrepancies are addressed if size-dependent sol-uble salt fraction is known. Although organics can po-tentially have a strong influence on cloud droplet for-mation (e.g., in biomass burning aerosol), it seems thatknowledge of the soluble salt fraction is sufficient fordescription of CCN activity at Thompson Farms. Thelatter finding is consistent with similar closure studiesconducted in the past [e.g., Liu et al., 1996; Broekhuizenet al., 2006] and those released during the ICARTT [Er-vens et al., in review; Fountoukis et al., in review].

Our measurements can be used to constrain the un-certainty associated with common assumptions in GCMmodeling studies of the aerosol indirect effect. For ex-ample, Sotiropoulou et al. [2006], using the observations

presented herein, estimate a 25% error in global clouddroplet number arising from the application of “sim-ple” Kohler theory. This suggests that the estimateduncertainty in indirect forcing is -0.5 W m−2, or abouta third of the total indirect forcing [IntergovernmentalPanel on Climate Change (IPCC), 2001]. A full assess-ment requires the application of a global model and willbe the focus of a future study.

Acknowledgments. This work was supported by NOAAunder grant NA04OAR4310088, an NSF CAREER awardand a Blanched-Million Young Faculty Fellowship. JIM ac-knowledges support from a NASA Earth System Science Fel-lowship. UNH acknowledges funding of NOAA AIRMAPgrants NA03OAR4600122 and NA04OAR4600154. We alsoacknowledge the support from Dr.G.Kok (DMT), and Dr.G.Hueyand his research group (Georgia Tech).

References

Abdul-Razzak, H., and S. Ghan, A parameterization ofaerosol activation 2. Multiple aerosol types, J. Geophys.Res., 105 , 6837–6844, 2000.

Abdul-Razzak, H., and S. Ghan, A parameterization ofaerosol activation 3. Sectional representation, J. Geophys.Res., 107 , 2002.

Alfarra, R., H. Coe, J. Allan, K. Bower, H. Boudries,M. Canagaratna, J. Jimenez, J. Jayne, A. Garforth, S. Li,and D. Worsnop, Characterization of urban and regionalorganic aerosols in the Lower Fraser Valley using twoAerodyne aerosol mass spectrometers, Atmos. Environ.,38 , 5745–5758, 2004.

Allan, J., J. Jimenez, H. Coe, K. Bower, P. Williams,and D. Worsnop, Quantitative sampling using an aero-dyne aerosol mass spectrometer. part 1: Techniques ofdata interpretation and error analysis, J. Geophys. Res.,108(D3), 4doi:10.1029/2002JD002,358, 2003.

Asa-Awuku, A., and A. Nenes, Ccn predictions: is theory


sufficient for assessments of the indirect effect?, J. Geo-phys. Res., in review.

Brechtel, F. J., and S. M. Kreidenweis, Predicting parti-cle critical supersaturation from hygroscopic growth mea-surements in the humidified TDMA. Part I: Theory andsensitivity studies, J. Aerosol Sci., 57 , 1854–1871, 2000.

Broekhuizen, K., R. Chang, W. Leaitch, S. Li, and J. P. D.Abbatt, Closure between measured and modeled cloudcondensation nuclei (CCN) using size-resolved aerosolcompositions in downtown Toronto, Atmos. Chem. Phys.,6 , 2513–2524, 2006.

Cantrell, W., G. Shaw, G. Cass, Z. Chowdhury, L. Hughes,K. Prather, S. Guazzotti, and K. Coffee, Closure be-tween aerosol particles and cloud condensation nucleiat Kaashidhoo climate observatory, J. Geophys. Res.,106(D22), 28,711–28,718, 2001.

Charlson, R., S. Schwartz, J. Hales, R. Cess, J. Coakley,J. Hansen, and D. Hofmann, Climate forcing by anthro-pogenic aerosols, Science, 255 , 6423–430, 1992.

Chuang, P., D. Collins, H. Pawlowska, J. Snider, H. Jons-son, J.L.Brenguier, R. Flagan, and J. Seinfeld, CCN mea-surements during ACE-2 and their relationship to cloudmicrophysical properties, Tellus B , 52 , 843–867, 2000a.

Chuang, P., A. Nenes, J. Smith, R. Flagan, and J. Seinfeld,Design of a CCN instrument for airborne measurement,J. Atmos. Ocean. Tech., 17 , 1005–1019, 2000b.

Cohard, J., J. Pinty, and C. Bedos, Extending Twomey’sanalytical estimate of nucleated cloud droplet concentra-tions from CCN spectra, J. Atmos. Sci., 55 , 3348–3357,1998.

Cohard, J., J. Pinty, and K. Suhre, On the parameteriza-tion of activation spectra from cloud condensation nucleimicrophysical properties, J. Geophys. Res., 105 , 11,753–11,766, 2000.

Collins, D., A. Nenes, R. Flagan, and J. Seinfeld, The scan-ning flow DMA, J. Aerosol Sci., 31 , 1129–1144, 2000.

Cottrell, L., R. Griffin, L. Ziemba, P. Beckman, L. Nielsen,B. Sive, R. Talbot, J. Jimenez, and D. Worsnop, Chemicalcharacterization of submicron aerosol at thompson farmduring ICARTT 2004, J. Geophys. Res., in preparation.

Covert, D., J. Gras, A. Wiedensohler, and F. Stratmann,Comparison of directly measured CCN with CCN mod-eled from the number-size distribution in the marineboundary layer during ACE 1 at cape grim, tasmania,J. Geophys. Res., 103(D13), 16,597–16,608, 1998.

Decesari, S., M. Facchini, M. Mircea, F. Cavalli, andS. Fuzzi, Solubility properties of surfactants in at-mospheric aerosol and cloud/fog water samples, J. Geo-phys. Res., 108(21), doi:10.1029/2003JD003,566, 2003.

Delene, D., and T. Deshler, Calibration of a photometriccloud condensation nucleus counter designed for deploy-ment on a balloon package., J. Atmos. Ocean. Tech., 17 ,459–467, 2000.

Dusek, U., D. Covert, A. Wiedensohler, C. Neususs,D. Weise, and W. Cantrell, Cloud condensation nucleispectra derived from size distributions and hygroscopicproperties of the aerosol in coastal south-west portugalduring ace-2, Tellus B , 55 , 35–53, 2003.

Ervens, B., M. Cubison, E. Andrews, G. Feingold, J. Ogren,J. Jimenez, P. DeCarlo, and A. Nenes, Prediction of CCNnumber concentration using measurements of aerosol sizedistributions and composition and light scattering en-hancement due to humidity, J. Geophys. Res., in review.

Facchini, M., M. Mircea, S. Fuzzi, and R. Charlson, Cloudalbedo enhancement by surface-active organic solutes ingrowing droplets, Nature, 401 , 257–259, 1999.

Fountoukis, C., and A. Nenes, Continued developmentof a cloud droplet formation parameterization forglobal climate models, J. Geophys. Res., 110(D11212),doi:10.1029/2004JD005,591, 2005.

Fountoukis, C., A. Nenes, N. Meskhidze, R. Bahreini,F. Brechtel, W. Conant, H. Jonsson, S. Murphy,A. Sorooshian, V. Varutbangkul, R. Flagan, and J. Sein-feld, Aerosol - cloud drop concentration closure for cloudssampled during ICARTT, J. Geophys. Res., in review.

Haywood, J., and O. Boucher, Estimates of the direct andindirect radiative forcing due to tropospheric aerosols: Areview., Rev.Geoph., 38(4), 513–543, 2000.

Intergovernmental Panel on Climate Change (IPCC), Cli-mate Change (2001): The Scientific Basis, CambridgeUniversity Press, UK, 2001.

Jayne, J., D. Leard, X. Zhang, P. Davidovits, K. Smith,C. Kolb, and D. Worsnop, Development of an aerosolmass spectrometer for sizevand composition analysis ofsubmicron particles, Aerosol Sci. Technol., 33 , 49–70,2000.

Ji, Q., G. Shaw, and W. Cantrell, A new instrument formeasuring cloud condensation nuclei: Cloud condensa-tion nucleus “remover”, J. Geophys. Res., 103 , 28,013–28,019, 1998.

Jimenez, J., J. Jayne, Q. Shi, C. Kolb, D. Worsnop,I. Yourshaw, J. Seinfeld, R. Flagan, X. Zhang, K. Smith,J. Morris, and P. Davidovits, Development of an aerosolmass spectrometer for size and composition analysisof submicron particles, J. Geophys. Res., 108(D7),doi:10.1029/2001JD001,213, 2003.

Kohler, H., Zur kondensation des wasserdampfe in der at-mosphare, Geofys. Publ., 2 , 3–15, 1921.

Kohler, H., Trans. Faraday Soc., 32 , 1152, 1936.Lance, S., J. Medina, J. Smith, and A. Nenes, Mapping

the operation of the DMT continuous flow CCN counter,Aerosol Sci. Technol., 40 , 242–254, 2006.

Liu, P., W. Leaitch, C. Banic, S. Li, D. Ngo, and W. Megaw,Aerosol observations at chebogue point during the 1993north atlantic regional experiment: Relationships amongcloud condensation nuclei, size distribution, and chem-istry, J. Geophys. Res., 101 , 28,971–28,990, 1996.

Nenes, A., and J. Medina, Cloud condensation nuclei (CCN)closure during the ICARTT 2004 campaign: b) mixingstate, chemical composition and droplet growth kineticsobtained from scanning mobility CCN analysis, J. Geo-phys. Res., in preparation.

Nenes, A., and J. Seinfeld, Parameterization of cloud dropletformation in global climate models, J. Geophys. Res.,108(D7), doi: 10.1029/2002JD002,911, 2003.

12 Medina et.al

Nenes, A., P. Chuang, R. Flagan, and J. Seinfeld, A theo-retical analysis of cloud condensation nucleus (CCN) in-struments, J. Geophys. Res., 106 , 3449–3474, 2001.

Nenes, A., R. Charlson, M. Facchini, M. Kulmala, A. Laak-sonen, and J. Seinfeld, Can chemical effects on clouddroplet number rival the first indirect effect?, Geophys.Res. Lett., 24 , doi: 10.1029/2002GL015,295, 2002.

Ramanathan, V., P. Crutzen, J. Kiehl, and D. Rosenfeld,Aerosols, climate, and the hydrological cycle, Science,294 , 2119–2124, 2001.

Rissman, T., T. VanReken, J. Wang, R. Gasparini,D. Collins, H. Jonsson, F. Brechtel, R. Flagan, and J. Se-infeld, Characterization of ambient aerosol from measure-ments of cloud condensation nuclei during the 2003 at-mospheric radiation measurement aerosol intensive obser-vational period at the southern great plains site in okla-homa, J. Geophys. Res., 111(D5), ISI:000235027200,002,2006.

Roberts, G., and A. Nenes, A continuous-flow streamwisethermal-gradient CCN chamber for atmospheric measure-ments, Aerosol Sci. Technol., 39 , 206–221, 2005.

Roberts, G., A. Nenes, M. Andreae, and J. H. Sein-feld, Impact of biomass burning on cloud propertiesin the Amazon Basin., J. Geophys. Res., 108 , doi:10.1029/2001JD000,985, 2003.

Seinfeld, J., and S. Pandis, Atmospheric chemistry andphysics: from air pollution to climate change, John Wi-ley, New York, 1998.

Shulman, M., M. Jacobson, R. Charlson, R. Synovec, andT. Young, Dissolution behaviour and surface tension ef-fects of organic compounds in nucleating cloud droplets,Geophys. Res. Lett., 23 , 277–280, 1996.

Snider, J., S. Guibert, J. Brenguier, and J. Putaud, Aerosolactivation in marine stratocumulus clouds: 2. kohler andparcel theory closures studies, J. Geophys. Res., 108 ,doi:10.1029/2002JD002,692, 2003.

Sotiropoulou, R., J. Medina, and A. Nenes, Ccn predictions:is theory sufficient for assessments of the indirect effect?,Geophys. Res. Lett., 33 , doi:10.1029/2005GL025,148,2006.

Stroud, C., A. Nenes, J. Jimenez, P. DeCarlo, J. Huffman,R. Bruintjes, E. Nemitz, A. Delia, D. Toohey, A. Guen-ther, and S. Nandi, Cloud activating properties of aerosolobserved during CELTIC, J. Atmos. Sci., in press.

Twomey, S., The nuclei of natural cloud formation. II. Thesupersaturation in natural clouds and the variation ofcloud droplet concentration., Geofisica pura e applicata,43 , 243–249, 1959.

VanReken, T., T. Rissman, G. Roberts, V. Varut-bangkul, H. Jonsson, R. Flagan, and J. Seinfeld, To-ward aerosol/cloud condensation nuclei (CCN) closureduring CRYSTAL-FACE, J. Geophys. Res., 108(D20),ISI:000186088500,001, 2003.

Zhang, Q., M. Canagaratna, J. Jayne, D. Worsnop, andJ. Jimenez, Time and size resolved chemical composi-tion of submicron particles in Pittsburgh – implicationsfor aerosol sources and processes., J. Geophys. Res., 110 ,doi:10.1029/2004JD004,649, 2005.

J. Medina, School of Chemical & Biomolecular Engi-neering, Georgia Institute of Technology, Atlanta, GA30332 (e-mail: [email protected])

A. Nenes, Schools of Earth & Atmopsheric Sciencesand Chemical & Biomolecular Engineering, GeorgiaInstitute of Technology, Atlanta, GA 30332 (e-mail:[email protected])

R.E.P. Sotiropoulou, School of Earth & AtmopshericSciences, Georgia Institute of Technology, Atlanta, GA30332 (e-mail: [email protected])

L.D. Cottrell, L.D. Ziemba and P.J. Beckman,Institute for the Study of Earth, Oceans, andSpace, University of New Hampshire (e-mail: [email protected], [email protected], [email protected])

R.J. Griffin, Institute for the Study of Earth, Oceans,and Space, and Department of Earth Sciences, Univer-sity of New Hampshire (e-mail: [email protected])Received N/A; revised N/A; accepted N/A.

size-dependent ccn closure during icartt 2004 1 cloud...

Documents