Characterization of ambient aerosol from measurements of cloud condensation nuclei during the 2003 Atmospheric Radiation Measurement Aerosol Intensive Observational Period at the Southern Great Plains site in Oklahoma

Characterization of ambient aerosol from measurements of cloud

condensation nuclei during the 2003 Atmospheric Radiation

Measurement Aerosol Intensive Observational Period at the Southern

Great Plains site in Oklahoma

T. A. Rissman,

T. M. VanReken,

J. Wang,

R. Gasparini,

D. R. Collins,

H. H. Jonsson,

F. J. Brechtel,

1,5

R. C. Flagan,

and J. H. Seinfeld

Received 13 December 2004; revised 15 March 2005; accepted 19 May 2005; published 28 January 2006.

[

]

Measurements were made by a new cloud condensation nuclei (CCN) instrument

(CCNC3) during the Atmospheric Radiation Measurement (ARM) Program’s Aerosol

Intensive Observational Period (IOP) in May 2003 in Lamont, Oklahoma. An inverse

aerosol/CCN closure study is undertaken, in which the predicted number concentration of

particles available for activation (

) at the CCNC3 operating supersaturations is

compared to that observed (

is based on Ko

̈hler Theory, with assumed and inferred

aerosol composition and mixing state, and the airborne aerosol size distribution measured

by the Caltech Dual Automatic Classified Aerosol Detector (DACAD). An initial

comparison of

and

, assuming the ambient aerosol is pure ammonium sulfate

((NH

)

), results in closure ratios (

) ranging from 1.18 to 3.68 over the duration

of the IOP, indicating that the aerosol is less hygroscopic than (NH

)

and

are

found to agree when the modeled aerosol population has characteristics of an external

mixture of particles, in which insoluble material is preferentially distributed among

particles with small diameters (<50 nm) and purely insoluble particles are present over a

range of diameters. The classification of sampled air masses by closure ratio and aerosol

size distribution is discussed in depth. Inverse aerosol/CCN closure analysis can be a

valuable means of inferring aerosol composition and mixing state when direct

measurements are not available, especially when surface measurements of aerosol

composition and mixing state are not sufficient to predict CCN concentrations at altitude,

as was the case under the stratified aerosol layer conditions encountered during the

IOP.

Citation:

Rissman, T. A., T. M. VanReken, J. Wang, R. Gasparini, D. R. Collins, H. H. Jonsson, F. J. Brechtel, R. C. Flagan, and J. H.

Seinfeld (2006), Characterization of ambient aerosol from measurements of cloud condensation nuclei during the 2003 Atmospheric

Radiation Measurement Aerosol Intensive Observational Period at the Southern Great Plains site in Oklahoma,

J. Geophys. Res.

111

D05S11, doi:10.1029/2004JD005695.

1. Introduction

[

] One of the largest uncertainties in aerosol radiative

forcing is associated with the indirect effect, which results

from the relationship between atmospheric aerosols and

cloud formation, properties, and lifetime [

Intergovernmental

Panel on Climate Change

, 2001].

Twomey

[1977] postulated

that an increase in the number concentration of atmospheric

aerosol particles would increase the number of cloud

droplets formed for a given air mass. For fixed liquid water

content, the cloud droplets would also be smaller than those

formed under conditions with lower particle concentrations.

This increase in number and decrease in mean diameter of

cloud droplets would have two indirect effects on climate.

Cloud albedo is greater for clouds with more numerous,

smaller droplets; this has been termed the first indirect

climatic effect of aerosols. Also, the lifetime of a cloud with

smaller cloud droplets is greater than that of a cloud with

larger droplets because the rain forming mechanisms are

less efficient [

Albrecht

, 1989]. This is referred to as the

second indirect climatic effect of aerosols. Both effects

create clouds that are more reflective and more persistent,

leading to the cooling of the Earth’s surface [

Twomey

1977].

[

] The relationship between atmospheric particles that

are capable of activating into cloud droplets, known as

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 111, D05S11, doi:10.1029/2004JD005695, 2006

Department of Chemical Engineering, California Institute of Technol-

ogy, Pasadena, California, USA.

Brookhaven National Laboratory, Upton, New York, USA.

Department of Atmospheric Sciences, Texas A&M University, College

Station, Texas, USA.

Center for Interdisciplinary Remotely Piloted Aircraft Studies, United

States Naval Postgraduate School, Marina, California, USA.

Brechtel Manufacturing, Inc., Hayward, California, USA.

D05S11

1of20

cloud condensation nuclei (CCN), and aerosol size distri-

bution and composition, in addition to meteorological

conditions, are central to the indirect climatic effect of

aerosols. For a given particle composition and size, the

supersaturation above which the particle undergoes sponta-

neous condensational growth (activation) into a cloud

droplet, the so-called critical supersaturation, is described

by Ko

̈hler Theory. The activation diameter, the dry diameter

at which a particle of known or assumed composition will

activate, can also be calculated for a given supersaturation.

Prediction of aerosol activation from Ko

̈hler Theory is very

successful for aerosols composed of soluble, inorganic salts,

such as ammonium sulfate ((NH

)

), sodium chloride

(NaCl), and ammonium bisulfate (NH

HSO

). However,

̈hler Theory needs to be augmented when considering

chemical components, such as organic compounds, that are

partially soluble, insoluble, or affect the surface tension of

the aqueous solution. The chemical composition of atmo-

spheric aerosol can be complex and include many different

chemical species, which may affect aerosol activation in

competing ways.

[

] The ability of Ko

̈hler Theory to predict ambient

CCN concentrations can be studied by comparing atmo-

spheric CCN measurements at a given supersaturation

with CCN concentrations calculated using aerosol size

distribution and composition measurements. This type

of study, called an aerosol/CCN closure, compares the

observed CCN concentration (

) at the operating super-

saturation of the CCN instrument to that predicted from

the aerosol size distribution and composition (

)ina

closure ratio, defined here as

is determined

from the aerosol size distribution by summing the

concentration of particles with diameters greater than

the activation diameter calculated from Ko

̈hler Theory

[

VanReken et al.

, 2003]. When direct measurements of

aerosol composition are unavailable,

can be calculated

using an assumed aerosol composition or one that is

inferred from other available measurements. An ‘‘inverse’’

aerosol/CCN closure study (explained further in section

5.2) can be undertaken to determine aerosol composition

and mixing states that are most consistent with CCN

observations when direct measurements of these aerosol

characteristics are unavailable.

[

] In this paper, data measured by a new CCN

instrument (CCNC3) during the Atmospheric Radiation

Measurement (ARM) Aerosol Intensive Observational

Period(IOP)inMay2003areusedinaninverse

aerosol/CCN closure study of the midcontinental aerosol

sampled near the Southern Great Plains (SGP) Central

Facility (CF) to determine aerosol composition and mix-

ing states consistent with the CCN measurements at the

operating supersaturations of the instrument. The Texas

A&M Differential Mobility Analyzer/Tandem Differential

Mobility Analyzer (DMA/TDMA) data from the surface

are used to infer the mixing state and insoluble fraction

of the aerosol population as a function of dry diameter.

This information is used with the airborne CCN and

aerosol size distribution measurements to determine con-

ditions under which aerosol composition and mixing

states inferred from surface measurements are able to

reproduce CCNC3 measurements at altitude. Aerosol

properties, categorized by closure ratio and aerosol size

distribution shape, during pollution and smoke events are

also discussed.

2. ARM Aerosol IOP

[

] The ARM Aerosol IOP occurred from 5 to 31 May

2003 at the Department of Energy’s (DOE’s) ARM SGP CF

in Lamont, Oklahoma. There were a total of 16 science

flights, with a total of 60.6 flight hours, conducted by the

Center for Interdisciplinary Remotely-Piloted Aircraft Stud-

ies’ (CIRPAS) Twin Otter aircraft on 15 days during this

period. The ARM Aerosol IOP flight tracks for flights 6–10

and 12–17 (the flights for which there are CCN data) are

shown in Figure 1, which illustrates that most flights took

place over or near the SGP CF. The last flight, flight 17, was

coordinated with the Moderate Resolution Imaging Spec-

troradiometer (MODIS) overpasses of four ARM sites

(SGP CF, Extended Facility (EF)-12, EF-20, and EF-19 in

Figure 1).

[

] The ARM SGP site is located in a midlatitude,

continental area, surrounded by agricultural land and dirt

roads. The site is influenced by local emissions from nearby

industrial and power plants and local aerosol sources, such

as vehicle and agricultural aerosols. Sulfur dioxide (SO

)

emissions from nearby oil refineries and power plants, such

as the Conoco and Ponca City Power Plants, are major local

sources of sulfate aerosols over the SGP site. Anthropogenic,

agricultural related aerosol sources include local fertilizer

application and production, field burning, and animal

byproducts. Local biomass burning is greatest from May

through July [

Iziomon and Lohmann

, 2003]. The particles

that are commonly found at the ARM SGP site are a

complex mix of these aerosol types, with smoke- and

dust-dominated events, which are commonly characterized

by decreased aerosol hygroscopicity, occurring occasionally

[

Sheridan et al.

, 2001]. Routine condensation particle

counter (CPC) measurements from the SGP site generally

show a strong, diurnal cycle of aerosol number concentra-

tion, with peak concentrations in the afternoon and

early evening. Over the 4-year period from July 1996 to

June 2000, the daily average condensation nuclei (CN)

concentration ranged from less than 1000 cm

to about

20,000 cm

, with a mean around 5000 cm

. The hourly

average CN concentration ranged from about 4000 cm

from 1100 to 1500 UTC (Universal Time Coordinated) to

about 18,000 cm

from 1800 to 2000 UTC [

Sheridan et

al.

, 2001]. (The difference between local time (LT) at the

ARM site and UTC is 5–6 hours, depending on daylight

savings. During the ARM Aerosol IOP, LT was 5 hours

behind UTC, so that 1200 LT corresponds to 1700 UTC in

this paper.) These high CN concentrations could result from

buildup and advection of pollutant aerosols from local

sources, photochemical particle production [

Sheridan et

al.

, 2001], coagulation, or the evolution of the boundary

layer [

Iziomon and Lohmann

, 2003].

[

] Most of the Twin Otter flights during the ARM

Aerosol IOP were conducted under clear or partly cloudy

skies to assess aerosol impacts on solar radiation. Additional

flights targeted mostly cloudy conditions to assess aerosol/

cloud interactions, test theoretical understanding of aerosol

activation, and to test surface remote sensing of the indirect

effect. Ground and airborne measurements, which included

D05S11

RISSMAN ET AL.: CCN DURING THE 2003 ARM AEROSOL IOP

2of20

D05S11

aerosol absorption, scattering, extinction, size distribution,

and CCN concentration, are compared in a variety of

closure studies to help resolve differences in measurements

and models. Routine ARM SGP aerosol measurements

(absorption, total scattering and hemispheric backscattering,

light scattering as a function of relative humidity, total

condensation particle concentration, number concentration

of particles with diameters from 0.l to 10

m, vertical

aerosol optical thickness, etc.) continued throughout the

IOP.

3. Instrument Descriptions

3.1. Twin Otter Inlet System

[

] In order to minimize sampling losses, the aerosol inlet

on the CIRPAS Twin Otter is designed to admit the air

sample prior to any bending of flow lines and slow the

sample down before transport to the instruments. The intake

extends forward from the roof of the cabin to a position

1.2 m directly above the aircraft’s nose. The sampled air is

initially slowed down by a factor of 5 by means of an

aerodynamically lipped diffuser. A second diffuser, posi-

tioned at the centerline of the first diffuser, reduces the flow

speed by another factor of two, while excess flow from the

first diffuser exits along the sides of the second diffuser. The

sample then flows down a 7.62 cm diameter duct and enters

the cabin after a 45

bend. Inside the cabin the duct is

straightened out again with another 45

bend, and samples

are drawn to the various instruments from ports mounted on

the side of the duct. The ports are flush inside the duct, but

extend outward at 45

angle to the flow. At an aircraft speed

of 50 m/s, approximately 1000 liters per minute (lpm) flow

down the duct. Air not ciphered off to the instruments is

dumped out of the cabin.

3.2. CCN Instrument (CCNC3)

[

] The CCN instrument (CCNC3) deployed on the

Twin Otter consists of three columns, each of which is

physically modeled after a previous CCN instrument col-

umn design [

Chuang et al.

, 2000] with some changes to the

physical design. Three temperature-controlled sections are

used to create the desired temperature gradient, instead of

fourteen sections, as in the work by

Chuang et al.

[2000].

The instrument was designed to be fully automated and

software controllable in normal operation and to operate

with a different supersaturation in each column. CCNC3

operating conditions, some aspects of which differ from

those given by

Chuang et al.

[2000], are discussed in the

following paragraphs. Only column 1 operated properly

during the ARM Aerosol IOP, the first field mission in

which the CCNC3 was deployed; thus CCN data were

obtained at one supersaturation per flight.

[

] Each CCNC3 column (Figure 2) consists of a stain-

less steel growth tube 0.4 m in length with a 1.9 cm outer

diameter and a 1.6 mm wall thickness. The inner wall of the

growth tube is lined with filter paper, which is rewetted by a

small peristaltic pump every 90 min. Three temperature

controlled, movable sections are in contact with the outer

wall of the growth tube to create the desired temperature

gradient, and, thus, the desired operating supersaturation,

inside the growth tube. For the ARM Aerosol IOP, a linear

temperature profile was used to develop a constant super-

Figure 1.

ARM Aerosol IOP flight paths for flights with CCN data. PNC is the Ponca City, Oklahoma,

airport, where the Twin Otter was based. The other sites are ARM ground measurement sites. The insert

shows the position of the counties (in pink) within the continental United States. The axes of the insert are

in the same units as those in the main plot.

D05S11

RISSMAN ET AL.: CCN DURING THE 2003 ARM AEROSOL IOP

3of20

D05S11

saturation at the centerline of the growth tube [

Rogers and

Squires

, 1977;

Roberts and Nenes

, 2005]. (A linear tem-

perature gradient was also used in the airborne CCN instru-

ments during the Cirrus Regional Study of Tropical Anvils

and Cirrus Layers – Florida Area Cirrus Experiment

(CRYSTAL-FACE) [

VanReken et al.

, 2003;

Roberts and

Nenes

, 2005].) The sampled aerosol is focused onto the

centerline of the growth tube and introduced to the column

with filtered sheath air; the droplets that form within the

growth tube are then counted by a detector.

[

] The CCNC3 detector consists of a laser, an optics

tube, and an avalanche photo-diode (APD) module, and was

designed on the basis of that of Laser Particle Counter

(LPC) Model 3755 (TSI, Incorporated), scaled down for

CCN application. A 670 nm, 10 mW Lasiris MFL Micro-

Focus Laser is positioned at one end of an optics tube, so

that the droplet inlet to the optics tube is at the 30 mm

working distance of the laser. When a droplet falls through

the laser beam, the laser light is scattered in the forward

direction, and a pair of aspherical condenser lenses collects

the scattered light and focuses it into a fiber optic at the

other end of the optics tube. The signal from the scattered

laser light is sent to a Hamamatsu Photonics C5460-01 APD

module, which sends the resulting digital pulse to the data

acquisition system.

[

] During the IOP, the column operated at three differ-

ent linear temperature gradients, with one temperature

gradient per flight. The total flow rate of the column was

about 0.56 lpm with a sheath to sample ratio around 10.

(NH

)

calibrations for column 1 at its different linear

temperature gradients are shown in Figure 3. The activation

diameters and supersaturations associated with each linear

temperature gradient are given in Figure 3. For each linear

temperature gradient calibration, a solution of (NH

)

was atomized to create droplets that were then dried and

introduced into a differential mobility analyzer (DMA).

Certain dry diameters were selected using the DMA and

then split to the CCNC3 inlet and the inlet to a CPC (TSI,

Inc., Model 3010). The activated ratio (

) was calculated

as the ratio of the number concentration of CCN measured

by the CCNC3 (

CCN

) to the number concentration of total

particles measured by the CPC (

CPC

). The activation

diameter (

act

) is the dry diameter at which 50% of the

particles are activated (

= 0.5). The uncertainty limits

given on the calibration curves in Figure 3 result from the

uncertainty in the diameter produced by the DMA (gener-

ally taken to be ±5%) and the combined uncertainties

associated with the concentrations measured by the CPC

and the CCNC3. These uncertainties associated with the

column calibrations are folded into the overall measurement

uncertainty, which is estimated for each flight on the basis

of criteria explained in section 4.3.

[

] The supersaturation corresponding to dry (NH

)

particles with diameter

act

, and thus the operating super-

saturation of the column, was calculated theoretically by

̈hler Theory. Droplet density is calculated from data of

Tang and Munkelwitz

[1994]; the full Pitzer model [

Pitzer

1973;

Pitzer and Mayorga

, 1973] is used to calculate the

osmotic coefficient; and values from

Pruppacher and Klett

[1997] are used for surface tension. The model calculates

the critical supersaturation for particles that contain certain

soluble salts, certain organics, and generalized insoluble

material [

Brechtel and Kreidenweis

, 2000a, 2000b].

Figure 2.

A schematic of a CCNC3 column. A second

layer of insulation covers the growth tube and heating/

cooling sections (to prevent temperature transfer to the

outside air) and the detector (to prevent condensation within

the optics tube).

Figure 3.

(NH

)

calibration curves for column 1 of

the CCNC3 at the temperature gradients used during the

ARM Aerosol IOP. The column was not functioning

properly for flights 1–5 and 11. The activation ratio is

defined as the ratio of the number concentration of CCN

measured by the CCNC3 to the number concentration of

particles measured by the CPC.

D05S11

RISSMAN ET AL.: CCN DURING THE 2003 ARM AEROSOL IOP

4of20

D05S11

[

] The calibrated activation diameters and the operating

supersaturations for column 1 during the ARM Aerosol IOP

are given in the legend in Figure 3 and in Table 1. The

instrument did not operate as expected from available

instrument models and previous CCN instrument experi-

ence. Thermal contact between the temperature control

sections and the growth column was improved in the design

of the CCNC3. Therefore the CCNC3 requires a smaller

temperature gradient than that required in previous CCN

instruments to develop a similar supersaturation within the

growth column. In-field calibrations were not available

because the calibration CPC had been flooded with water.

Therefore the operating supersaturations of the CCNC3

during the ARM Aerosol IOP were higher than would

usually be desired for ambient aerosol studies. The con-

sequences of these high supersaturations are further dis-

cussed in the Conclusion (section 7).

3.3. Dual Automatic Classified Aerosol Detectors

(DACAD)

[

] The Caltech Dual Automatic Classified Aerosol

Detectors (DACAD) have been deployed in several previ-

ous airborne experiments, and their characteristics are well

documented [

Wang et al.

, 2002, 2003;

VanReken et al.

2003]. The DACAD consists of two DMA systems operated

in parallel. One of the DMA systems measures the dry

aerosol size distribution, while the other measures the

aerosol size distribution at ambient relative humidity (RH)

by using an active RH controller [

Wang et al.

, 2003]. The

aerosol wet and dry size distributions are obtained separately

and independent of each other; therefore no size-resolved

information is obtained. The main components of the mea-

surement system are a cylindrical DMA (TSI Inc., Model

3081) and a CPC (TSI Inc., Model 3010), which has a 50%

counting efficiency at 10 nm. Using the scanning mobility

technique, each DMA system generates a size distribution

for particle diameters from

17 to

720 nm every 72.5 s

[

Wang et al.

, 2002]. (This scanning time was reduced from

103 s [

VanReken et al.

, 2003] prior to the ARM Aerosol

IOP.)

3.4. Tandem Differential Mobility Analyzer

(DMA/TDMA)

[

] The Texas A&M DMA/TDMA measured aerosol

size distributions and size-resolved hygroscopic growth at

the SGP Cloud and Radiation Testbed (CART) trailer at the

CF during the IOP [

Gasparini et al.

, 2006]. The main

measurement section of the DMA/TDMA consists of two

Aerosol Dynamics, Inc., high-flow DMAs (HF-DMAs

[

Stolzenburg et al.

, 1998]), a charger, two Nafion tubes,

and a CPC. One DMA/TDMA measurement sequence

(

1 hour) consists of two different operational modes to

obtain both the aerosol size distribution and the size-resolved

hygroscopic growth. For both modes, the sample air is first

dried in a Nafion tube and then introduced to a charger

before entering the first DMA. During the single DMA mode

(

5 min), the aerosol size distribution is measured by

scanning the voltage applied to the first DMA [

Wang and

Flagan

, 1990] and counting the size-selected aerosol particle

concentration with the CPC. In TDMA mode, the voltage

supplied to the first DMA is fixed to produce a monodisperse

aerosol of known particle size. The monodisperse aerosol

is exposed to a controlled, elevated relative humidity

(RH; 85% in this IOP) before entering the second DMA.

The humidified aerosol is classified by scanning the

voltage applied to the second DMA and the size- and

hygroscopicity-resolved aerosol particle concentration is

observed with the CPC. The second mode sequence is

repeated for other particle dry diameters by changing the

voltage applied to the first DMA [

Gasparini et al.

, 2004]. A

third mode was implemented during the IOP, in which the

dry monodisperse aerosol was exposed to a wide range of

RH to determine the deliquescence properties of the aerosol

[

Gasparini et al.

, 2006].

[

] The aerosol particle soluble fraction and mixing state

is inferred by comparing the dry diameter separated by the

first DMA with the hydrated size distribution measured with

the second DMA. The comparison results in a normalized

distribution expressed in terms of the hygroscopic growth

factor (

(

)), which is defined as the ratio of the hydrated

particle diameter to that of the dry particle. The aerosol is

then divided into four hygroscopicity-based categories: pure

insoluble, pure soluble, mixed insoluble, and mixed soluble.

The full technique used to determine the relative contribu-

tion of soluble and insoluble components to the dry particle

composition is described by

Gasparini et al.

[2004].

3.5. PILS-IC, TEOM, and Integrating Nephelometer

[

] At the SGP site, the aerosol ionic composition

(species: NH

,SO

,NO

,Ca

,Mg

,Na

, and Cl

)

Table 1.

Flight Summary With Operating Conditions and Uncertainty Limits for CCNC3 Column 1

Flight

Day in

May

Flight Begin

Time,

UTC

Flight End

Time,

UTC

Flight

Length,

Hours

Activation

Diameter,

Operating

Supersaturation,

Upper

Limit,

Lower

Limit,

1553

2019

4.4

15 ± 0.8

2.8 ± 0.2

2124

2248

1.4

15 ± 0.8

2.8 ± 0.2

1634

1909

2.6

15 ± 0.8

2.8 ± 0.2

1402

1805

4.0

13 ± 0.6

3.6 ± 0.4

N/A

1543

1745

2.0

15 ± 0.8

2.8 ± 0.2

N/A

1551

1847

2.9

18 ± 0.9

2.1 ± 0.2

1325

1813

4.8

18 ± 0.9

2.1 ± 0.2

1852

2212

3.3

18 ± 0.9

2.1 ± 0.2

1420

1929

5.2

18 ± 0.9

2.1 ± 0.2

1824

2205

3.7

18 ± 0.9

2.1 ± 0.2

1411

1751

3.7

18 ± 0.9

2.1 ± 0.2

The difference between local time (LT) and UTC was 5 hours during the ARM Aerosol IOP, so that 1200 LT corresponds to 1700 UTC.

Uncertainty limits are reported as a percent of the measured concentration.

D05S11

RISSMAN ET AL.: CCN DURING THE 2003 ARM AEROSOL IOP

5of20

D05S11

was measured using the Particle-Into-Liquid Sampler–Ion

Chromatography (PILS-IC) technique [

Weber et al.

, 2001],

and the aerosol total mass concentration was measured

using the Tapered Element Oscillating Microbalance

(TEOM) technique [

Patashnick and Rupprecht

, 1986].

The PILS-IC technique has a

0.05

limit of

detection for all of the ions. The TEOM exhibited a

temperature-related oscillation behavior during the IOP that

resulted in an uncertainty of ±40%. The time resolution of

the PILS-IC and TEOM were 8 min and 30 min, respec-

tively, during the IOP. A sharp cut cyclone and two glass

honeycomb denuders in series remove particles greater than

m diameter and water-soluble vapor phase species,

respectively, from the sample flow to both the PILS-IC

and the TEOM. To prevent condensation of water within the

tapered element, the TEOM inlet was maintained at 35

[

Pahlow et al.

, 2006].

[

] Routine aerosol light scattering coefficient (

)

and light scattering hygroscopic growth factor (

(

))

measurements for total and fine (submicron) mode aerosol

continued at the ground site during the IOP. Measure-

ments of

and

(

) complement those from the PILS-

IC and TEOM. Total scattering and backscattering were

measured with a three-wavelength integrating nephelom-

eter (TSI Inc., Model 3563) [

Sheridan et al.

, 2001]. The

light scattering hygroscopic growth factor,

(

), is

different from the hygroscopic growth factor (

(

))

determined from DMA/TDMA data.

(

) is based on

the increase in particle diameter with increasing RH (see

section 3.4), whereas

(

) is defined as the change in

aerosol light scattering with changing RH [

Covert et al.

1972;

Rood et al.

, 1987] and is determined as the ratio of

at 80% to that at 40% RH.

(

) was measured using

a humidograph [

Carrico et al.

, 1998], which consists of

two nephelometers separated by a humidity control sys-

tem that ramps up the RH in the second nephelometer.

Data from the humidograph are used to relate

at any

RH to that at 40% RH with a three-parameter curve-

fitting model [

Sheridan et al.

, 2001].

4. Data Trends During the ARM Aerosol IOP

4.1. Particle Composition (PILS-IC, TEOM, and

DMA/TDMA) and Aerosol Mixing State (DMA/TDMA)

[

] Of the major inorganic ionic species (NH

,SO

,Ca

,Mg

,Na

,Cl

) measured at the SGP site

during the IOP, NH

and SO

were dominant. NO

was

observed in

10% of the samples and only when SO

was

completely neutralized by NH

, a tracer for biomass

burning, and Ca

, a tracer for dust, were occasionally

observed in appreciable levels. On average, the ratio of

to SO

was 2, with ratios significantly greater than 2

observed on 20 May (no flight) and 22 May (flight 13) and

less than 2 observed on 21 May (flight 12) [

Pahlow et al.

2006].

[

] Aerosol organic fraction increased continuously

during the daytime hours, while the inorganic concentra-

tion remained fairly constant. Overall, the aerosol organic

content, which is estimated as the difference between

total mass (TEOM) and total inorganic mass (PILS-IC),

accounted for >40% of the aerosol mass. Increases in

organic fraction, values of which were as high as 80%,

were accompanied by a lowering of

(

), suggesting

that the particulate organic fraction had a lower hygro-

scopicity than the inorganic fraction. Also, the amount of

aerosol organic mass correlated strongly with the amount

of black carbon inferred from the aerosol absorption

coefficient [

Pahlow et al.

, 2006], which was measured

using a filter-based light absorption photometer [

Sheridan

et al.

, 2001].

[

]

Gasparini et al.

[2006] inferred size-resolved aerosol

particle composition and hygroscopic growth properties

from data collected by the Texas A&M DMA/TDMA at

the SGP site using the technique from

Gasparini et al.

[2004]. The

(

) of the observed aerosol during the IOP

was found to increase with increasing dry diameter, and the

average median

(

) at 85% RH was 1.20 at 12 nm and

1.37 at 300 nm, which indicates that the smallest analyzed

particles were largely composed of carbonaceous com-

pounds. The largest particles were either very hygroscopic

or slightly hygroscopic, but rarely exhibited hygroscopicity

in between these extremes. At times during the IOP, a

nonhygroscopic mode with median

(

) less than 1.10

was observed at 450 and 600 nm and, more rarely, at 200

and 300 nm [

Gasparini et al.

, 2006].

4.2. Aerosol Size Distribution (DACAD-Airborne)

[

] The aerosol size distributions measured by the

DACAD were analyzed for flights 6–10 and 12–17,

for which CCN data were collected. Aerosol size distri-

butions differed greatly between flights, as well as at

different altitudes within a single flight. For example,

Figure 4 shows DACAD aerosol size distributions from

six level legs and five different altitudes from flight 16.

At the lower altitudes, 351 m and 720 m AGL (Above

Ground Level; all reported altitudes are AGL), the aerosol

size distributions are multimodal with median diameters

less than the cutoff diameter of the DACAD (

17 nm),

which may indicate that the sampled aerosols are freshly

emitted or formed locally in the atmosphere. The aerosol

populations at 1028 m and 1337 m have similar size

distributions that are multimodal with median diameters

that may be smaller than the cutoff diameter of the

DACAD, which may indicate mixed aerosol layers that

are a combination between fresh and aged aerosol pop-

ulations. The aerosol population at 2281 m during flight

16 is multimodal but with a less pronounced second

mode at larger diameters, which is indicative of back-

ground aerosol. Also note that the

dlogD

values are

greater for lower-altitude level legs than higher legs,

which is a typical feature of ambient aerosol. Aerosol

size distributions such as those shown in Figure 4 were

seen throughout the IOP flights, along with some simple

unimodal distributions. The differences in aerosol size

distributions are consistent with sampled aerosol layers

that had different sources and/or were of different ages.

[

] Rapidly changing aerosol size distributions were

observed frequently during the IOP. Since the DACAD size

distribution scan time is 72.5 s, some characteristics of a

rapidly changing aerosol size distribution may not have

been resolved. The CCNC3 has a sampling rate of 1 s and,

therefore, is capable of observing rapid changes in the

aerosol size distribution. For this reason, predicted and

observed CCN concentrations may not agree well during

D05S11

RISSMAN ET AL.: CCN DURING THE 2003 ARM AEROSOL IOP

6of20

D05S11

times of rapidly changing aerosol characteristics. This will

be discussed further in section 6.

4.3. CCN Concentration (CCNC3-Airborne)

[

] The trends in the CCN concentration data follow

those seen in both the airborne CPC and airborne DACAD

data. When present, homogeneous boundary layers are

easily discerned from plots of CCNC3, CPC, and DACAD

data, e.g., Figure 5 for flight 14. A more specific CCN trend

analysis is discussed in section 6.

[

] The airborne CPC had a cutoff diameter of 13 nm,

and the DACAD scanned particle sizes down to

17 nm.

Figure 4.

Aerosol size distributions from different level legs during flight 16. The legend gives the

average altitudes (AGL) and times analyzed for each level leg. The times given are UTC; 1700 UTC

corresponds to 1200 LT.

Figure 5.

Flight traces of CPC, CCNC3, and DACAD data for flight 14. The DACAD traces give the

total concentrations of particles with diameters greater than those noted in the legend. A well-mixed

boundary layer is clearly shown from about 2010 to 2045. The times given are UTC; 1700 UTC

corresponds to 1200 LT.

D05S11

RISSMAN ET AL.: CCN DURING THE 2003 ARM AEROSOL IOP

7of20

D05S11

The CCN number concentrations are always less than the

total particle number concentrations measured by the CPC

or the DACAD in any of the eleven flights, even though

the activation diameters for pure (NH

)

particles at the

operating supersaturations of the CCNC3 are similar to the

cutoff diameters for the CPC and DACAD. To determine

the uncertainty in the CCN measurements, the CPC con-

centration was used as a high-end limit for CCN concen-

tration on all flights. The concentration measured by the

CCNC3 column 2, when operating, was used as the lower

limit for the column 1 CCN concentration. The operating

supersaturation of column 2 was unable to be determined

with any certainty, but it was always operating at some

supersaturation lower than that of column 1. Therefore

measurements by column 2 are used as estimates of the

lower uncertainty limit for the measurements of column 1,

when both columns were operating, but are not used in any

other part of the analysis. Table 1 gives the uncertainty

limits for column 1.

5. Inverse Aerosol/CCN Closure Study: Results

and Discussion

[

] Comparison of the CCN data to the aerosol size

distribution data allows certain conclusions to be drawn

about the characteristics of the measured CCN and the

aerosol population. Certain aerosol population properties,

such as particle insoluble fraction or mixing state, can be

estimated as those that lead to closer agreement between

predicted and observed CCN concentrations at the operating

supersaturations of the CCN instrument. These estimated

properties can be compared with those inferred from or

measured by other instruments to determine whether the

estimates are reasonable.

[

] For this study, the CCN data sets were averaged over

the scan time of the DACAD (72.5 s) to correspond with a

single aerosol size distribution. Scans during which the

standard deviation of the CCN data exceeded 15% of the

mean of the CCN data were removed from the inverse

closure analysis. This helps to remove intervals during

which the aerosol size distribution may be changing suffi-

ciently rapidly during the 72.5 s scan time of the DACAD

that the changes are not resolved in aerosol size distribution

measurements. One further constraint: only data measured

during level legs of flight patterns were included to ensure

that the CCNC3 inlet pressure was steady and did not cause

any rapid changes or instability in the CCN sample flow

rate, although this artifact was not observed even during

rapid ascents and descents. Data measured within clouds

were also removed from the inverse aerosol/CCN closure

analysis to exclude the potential influence of artifact par-

ticles resulting from cloud droplet shattering in the Twin

Otter inlet. The DACAD did not operate at altitudes above

about

3000 m (except during flight 14), so the inverse

Figure 6.

Predicted CCN concentration (

; assuming pure (NH

)

) versus observed CCN

concentration (

) for ARM Aerosol IOP. Only flights for which CCN data are available are shown.

is determined from the DACAD size distribution for all flights except flight 9, for which

is determined

from the CPC total particle (>13 nm) concentration. The linear fit in log-log space for all flight closure

data is shown, and corresponding fit parameters are given. The mean and standard deviation of the ratio

averaged over all flights are also given. The fit parameters and mean and standard deviation

of the ratio of

are also shown for the ‘‘Insoluble Limit’’ case, for which

is calculated using

the insoluble fractions calculated in section 5.2.1.

D05S11

RISSMAN ET AL.: CCN DURING THE 2003 ARM AEROSOL IOP

8of20

D05S11

aerosol/CCN closure analysis generally does not include

data above this altitude.

5.1. Inverse Aerosol/CCN Closure: Initial Comparison,

Assuming 100% (NH

)

[

] Figure 6 shows an initial comparison of the predicted

CCN number concentrations (

) to those observed by the

CCNC3 instrument (

) for all eleven flights and includes

all the data points from each flight that fit the criteria

outlined in section 5. For this initial comparison,

determined as the sum of the number concentration of

particles in the DACAD size distributions with diameters

greater than the activation diameter of an aerosol consisting

of pure (NH

)

for the operating supersaturation of the

CCN instrument. Pure (NH

)

is assumed as a starting

point because PILS-IC data from the ground indicated that

and SO

were dominant species and that the average

ratio of NH

to SO

was near 2 during the ARM Aerosol

IOP. For flight 9,

is determined from the CPC total

particle concentration, since the activation diameter of the

CCNC3 column is about the same as the cutoff diameter for

the CPC. The average value of

(

)) for all

flights is 1.92, with a standard deviation (

(

)) of 1.29,

and Table 2 shows this ratio for each flight. (

(

)is

used when values of the ratio,

, have been averaged

over more than one DACAD scan.) The average value of

for each flight, and thus for all flights averaged

together, is greater than unity. This indicates that fewer

particles activated than would be expected if the particles

were composed of pure (NH

)

. The average

also differs from flight to flight, which indicates that the

aerosol CCN activity properties changed from flight to

flight.

[

] The

versus

data for each flight were fit to a

straight line in log-log space, and the results for the slope

and intercept of those fits are given in Table 2. The fit for all

data is shown in Figure 6, and Figure 7 shows the fit for

flights 8, 12, 14, and 17. Nonunity slopes and nonzero

intercepts of these linear fits in log-log space offer some

insight as to the characteristics of the ambient aerosol

population through its deviation from the activation prop-

erties of pure (NH

)

particles.

[

] The slopes of the log-log fit of

versus

for

the IOP flights are generally not unity, which indicates

that the chemistry and mixing state of the aerosol

population does not affect the CCN ability at all concen-

trations equally. This imbalance results from sampling

different air masses with different aerosol properties,

which lead to different

ratios, during the same

flight. Aerosol populations with low CCN concentration,

which often result from those with low total particle con-

centration (

Total

), tend to have ‘‘clean’’ aerosol sources

or be well-aged aerosols, in which particle coagulation has

decreased

Total

and cloud processing has likely increased

soluble fractions. These types of air masses with low

CCN concentrations would have

ratios nearer to

unity. High

Total

, on the other hand, is often indicative of

fresh, polluted aerosol sources, possibly with high levels

of insoluble material and a greater degree of external

mixing, which could lead to higher values (greater than

unity) of

. In Figure 7, the clustering of flight

12’s closure data points at low, medium, and high CCN

concentrations could indicate that three different air

masses were sampled. The air mass with midlevel CCN

concentrations could be a mixture of the air masses with

high- and low-level CCN concentrations, instead of a

distinct air mass. Figure 8 illustrates trends of

with altitude as a function of longitude for flights 12

(Figure 8a) and 14 (Figure 8b). The high CCN concen-

tration data points (Figure 7) for flight 12 correspond to

the

points (Figure 8) at altitudes of about 600 m

(

= 2.77), 850 m (

= 1.99–2.42), and

1380 m (

= 2.00); the mid CCN concentration data

points correspond to the

ratios at about 550 m

(

= 1.51–2.02), 600 m (

= 1.77, 1.80), and

1380 m (

= 1.67); and the low CCN concentration

points correspond to the

ratios at about 2100 m

(

= 1.23–1.53).

[

] The situation is quite different for flight 14 (see

Figures 7 and 8), for which the generalized high and low

Total

classifications do not hold. All of the closure data

points from flight 14 are clustered at CCN concentrations

less than 2000 cm

, and the slope is less than unity.

Contrary to what was observed in flight 12, the sampled

aerosol with lower

Total

has a greater deviation from

idealized, pure (NH

)

particles than those with higher

Total

. As will be discussed later in section 6, flight 14 was

apparently influenced by a Siberian smoke event [

Schmid et

al.

, 2004]. Smoke aerosols that have been transported a long

distance are likely to have low particle concentrations

through particle coagulation and cloud processing. However,

the aged smoke particles at low

Total

in flight 14 did not

Table 2.

Average (

(

)) and Standard Deviation (

(

)) of the Ratios of Predicted CCN Concentration to Observed CCN

Concentration for All Flights for Which There Were CCN Data

Flight

Day in May

(

)

(

)

Number of Data Points

1.30

0.18

0.63

1.32

0.63

1.29

0.13

0.81

0.77

0.96

1.18

0.07

1.05

0.10

0.96

1.29

0.35

0.85

0.54

0.96

1.34

0.08

0.97

0.21

0.98

1.79

0.36

1.14

0.22

0.99

1.78

0.51

0.91

0.51

0.90

1.84

0.48

0.77

0.88

0.96

1.87

0.45

1.14

0.18

0.97

3.68

2.96

1.40

0.68

0.65

3.45

0.86

0.93

0.74

0.77

All

1.92

1.29

0.90

0.54

0.74

517

Slope (

) and intercept (

) values for linear fits of

in log-log space are also given. Assumes pure (NH

)

in the calculation of

D05S11

RISSMAN ET AL.: CCN DURING THE 2003 ARM AEROSOL IOP

9of20

D05S11

become more CCN active than the background aerosol,

which is probably the type of aerosol sampled under about

1400 m (see Figure 8).

[

] An intercept that deviates from zero indicates that

some fraction of the aerosol population is unavailable for

activation at the measurement supersaturation, although this

fraction would activate under the assumption of pure

(NH

)

particles. This nonactive fraction could consist

of particles from an internal or external mixture that have

some fraction of insoluble, hydrophobic material that pre-

vents activation. Even if the slope were near unity, the

intercept can, as explained above, deviate from zero if

particles exist that do not exhibit predicted CCN activity

characteristics. Flights 8 and 17 (Figure 7) have slopes near

unity but have very different intercepts, even though their

observed CCN concentrations are within a similar range.

HYSPLIT [

Draxler and Rolph

, 2003;

Rolph

, 2003] 3-day

back trajectories for flights 8 and 17 (Figures 9 and 10,

respectively) show that the sampled air masses from flights

8 and 17 originated in different areas, with those from flight

8 being more influenced by marine conditions in the Gulf of

Mexico and those from flight 17 being more influenced by

inland continental conditions. DACAD size distributions

from the analyzed times during flight 8 are bimodal, with no

indication of sampled pollution events, and

dlogD

values were less than 5000 cm

. Flight 17, however, was

dominated by pollution events, as is indicated by its

primarily multimodal size distributions that are dominated

by particles with diameters smaller than 50 nm. Measured

dlogD

values for flight 17 were as high as 60,000 cm

but most were under 10,000 cm

. It makes sense, then, that

the deviation of flight 17’s intercept from zero is greater

than that of flight 8 because fewer particles are available for

activation under polluted than clean conditions. When the

slope is not near unity, the intercept can be affected by

differences in CCN activities among the sampled air masses

and may not have a simple interpretation.

[

] On the basis of this initial comparison of

and

determined from DACAD size distributions with the

assumption of 100% (NH

)

, it is apparent that more

detailed aerosol properties, such as chemical composition

and the mixing state, are needed to predict the CCN

measurements at the operating supersaturations of the

CCNC3. Because such composition and mixing data were,

however, unavailable on the Twin Otter during the IOP, an

inverse closure study is performed, in which the comparison

and

based on the assumption of pure (NH

)

used to infer the aerosol composition and mixing state that

are consistent with CCN observations.

5.2. Inverse Aerosol/CCN Closure: Inferring

Deviations From 100% (NH

)

[

] An inverse aerosol/CCN closure evaluates the extent

to which particle composition alone can explain the dis-

crepancy between

and

. Particle composition can be

modeled as an internal or an external mixture and can

consist of completely soluble and insoluble material or

include partially soluble material, as well. For the purposes

of the study presented here, particle components are con-

sidered to be either completely soluble or completely

Figure 7.

(assuming pure (NH

)

) versus

and linear fits in log-log space for flights 8, 12, 14,

and 17. High aerosol concentrations (from the CPC) were 10,400 cm

, 37,000 cm

, 3550 cm

, and

2890 cm

for flights 8, 12, 14, and 17, respectively. High CCN concentrations were 2882 cm

, 16,941

, 2143 cm

, and 7454 cm

for flights 8, 12, 14, and 17, respectively. High concentration values

were determined as the maximum concentrations measured after the Twin Otter had achieved sampling

altitude after takeoff, regardless of whether the data were obtained during constant altitude legs.

D05S11

RISSMAN ET AL.: CCN DURING THE 2003 ARM AEROSOL IOP

10 of 20

D05S11

insoluble, and both internally and externally mixed aerosol

populations are considered. Although the customary defini-

tion of an internal mixture is that all particles of the same

size have the same chemical composition, a more simpli-

fied, size-independent internal mixture is used here in the

absence of airborne measurements of size resolved chemis-

try. Thus, under the assumption of a size-independent

internally mixed aerosol population, all particles of all

diameters have the same fractional insoluble/soluble com-

position in this analysis. An external mixture can include

purely soluble particles, purely insoluble particles, and

particles composed of both insoluble and soluble material.

5.2.1. Estimating Aerosol Particle Composition From

CCNC3 and DACAD Measurements

[

] The initial comparison of

and

(from

section 5.1) can be used to determine the particle volu-

metric fraction of insoluble material (

ins

) that would be

necessary for

(

) to approach unity, assuming that

the ambient aerosol is a size-independent internal mixture,

with particles of all sizes having insoluble fraction

ins

.In

order to calculate

ins

for each flight, the cumulative

aerosol size distribution is determined, with each size

bin containing the concentration of all particles with

diameters greater than the bin diameter. For each analysis

period, the DACAD midbin diameter for which the

cumulative aerosol concentration approaches the measured

CCN concentration is chosen as the cutoff diameter,

defined as the dry diameter above which sampled ambient

particles would be able to activate under the assumption

of size-independent internal mixing. A value for

ins

determined by inputting the cutoff diameter and different

values of

ins

(with the balance (NH

)

) into the

̈hler Theory model, until the particle’s calculated

critical supersaturation approaches the operating supersat-

uration of the CCNC3.

[

] Since the sampled air mass properties typically

varied widely during a flight, this analysis is carried out

for each level leg during each flight. A level leg is defined

as one for which consecutive static pressure measurements

are constant within ±3 mbar. Some level legs are further

divided if it appears that drastically different air masses

were sampled during the same level leg. This usually occurs

for level legs that cover long distances but was observed

even during some shorter-distance legs. Table 3 gives the

range and mean of

ins

values for each flight. Most of the

values of

ins

that are required for

(

) to approach

unity exceed 0.90, with only 14 out of the 113 analyzed

level legs having

ins

values less than 0.90. While insoluble

mass fractions in aerosol particles in continental areas have

been found as high as 0.98, values less than 0.6 are more

Figure 8.

Altitude versus longitude trace for flights (a) 12

and (b) 14. Closure data points are indicated by the solid

circles. The color of the closure data points indicates the

magnitude of

, assuming pure (NH

)

, for the

corresponding DACAD scan.

Figure 9.

Three-day HYSPLIT back trajectories for

15 May 2003 (flight 8) at the SGP site at altitudes of 450,

650, and 1000 m. The times given are UTC; 1700 UTC

corresponds to 1200 LT.

D05S11

RISSMAN ET AL.: CCN DURING THE 2003 ARM AEROSOL IOP

11 of 20

D05S11

common [

Pruppacher and Klett

, 1997]. Lower insoluble

fractions would be expected in the ambient aerosol,

especially if the aerosol were truly internally mixed

independently of particle size. Under the assumption of a

size-independent internal mixture of particles, the values of

ins

that would be required to achieve agreement between

and

exceed the average fraction of insoluble material

commonly found in continental aerosols. Other aerosol

properties likely contributed to the discrepancy between

and

5.2.2. CCN and DACAD Derived Insoluble Fraction

(Flight 10)

[

] During the IOP, there was one flight (flight 10)

during which CCN concentrations and DACAD wet and

dry aerosol size distributions were all measured. From the

wet and dry aerosol size distributions and assuming size-

independent internally mixed aerosol population, an esti-

mate of

ins

can be obtained by comparing the cumulative

particle concentrations of the wet and dry size distributions

at each diameter. The average difference in wet and dry

Figure 10.

Three-day HYSPLIT back trajectories for 29 May 2003 (flight 17) (a) at the SGP site at

altitudes of 400, 800, and 3000 m and (b) at site EF-19 at altitudes of 650, 800, and 1000 m. The times

given are UTC; 1700 UTC corresponds to 1200 LT.

Table 3.

Insoluble Volumetric Fractions (

ins

) and Fraction of Particles Unavailable for Activation (

) for Which

(

)

act

, nm

Flight

ins

(Range

)

ins

(Mean

)

(Range

)

(Mean

)

2.8

0.66–0.96

0.90

0–0.39

0.22

2.8

0.93–0.96

0.94

0–0.27

0.21

2.8

0.97–>0.99

0.98

0.13–0.37

0.15

2.8

>0.99 (all)

>0.99

0.24–0.26

0.25

3.6

0.91–>0.99

0.98

0.04–0.61

0.22

2.1

0.78–0.96

0.92

0.26–0.64

0.42

2.1

0.95–0.98

0.96

0.17–0.61

0.40

2.1

0.98–>0.99

>0.99

0.24–0.63

0.42

2.1

0.92–>0.99

0.98

0.36–0.70

0.44

2.1

>0.99 (all)

>0.99

0.40–0.91

0.63

2.1

0.87–>0.99

0.98

0.50–0.76

0.69

Size-independent internal mixing is assumed in the determination of

ins

The ranges of values for

ins

and

are determined from those averaged over each of the level legs during a flight.

The mean values of

ins

and

are averaged over all analysis points.

is determined under the assumption of pure (NH

)

in the determination of

D05S11

RISSMAN ET AL.: CCN DURING THE 2003 ARM AEROSOL IOP

12 of 20

D05S11