Full text of DOI:10.3402/tellusb.v52i2.16102

Tellus B: Chemical and Physical Meteorology
ISSN: (Print) 1600-0889 (Online) Journal homepage: https://www.tandfonline.com/loi/zelb20
Shipboard measurements of concentrations and
properties of carbonaceous aerosols during ACE-2
T. Novakov, Timothy S. Bates & Patricia K. Quinn
To cite this article: T. Novakov, Timothy S. Bates & Patricia K. Quinn (2000) Shipboard
measurements of concentrations and properties of carbonaceous aerosols during ACE-2, Tellus B:
Chemical and Physical Meteorology, 52:2, 228-238, DOI: 10.3402/tellusb.v52i2.16102
To link to this article: https://doi.org/10.3402/tellusb.v52i2.16102
©
2000 The Author(s). Published by Taylor &
Francis.
Published online: 15 Dec 2016.
Submit your article to this journal
Article views: 15
Citing articles: 21 View citing articles
Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=zelb20

T ellus (2000), 52B, 228–238
Printed in UK. All rights reserved
Copyright © Munksgaard, 2000
TELLUS
ISSN 0280–6509
Shipboard measurements of concentrations and properties
of carbonaceous aerosols during ACE-2
By T. NOVAKOV*, TIMOTHY S. BATES and PATRICIA K. QUINN, Environmental Energy
T echnologies Division, L awrence Berkeley National L aboratory, 1 Cyclotron Rd., MS-73, Berkeley,
CA 94720, USA; NOAA, Pacific Marine Environmental L aboratory, 7600 Sand Point Way NE, Seattle,
WA 98115, USA
(
Manuscript received 25 January 1999; in final form 20 September 1999)
ABSTRACT
Mass concentrations of total, organic and black carbon were derived by analyzing the super-
micron and submicron aerosol fractions of shipboard collected samples in the eastern Atlantic
Ocean as part of the second Aerosol Characterization Experiment (ACE-2). These analyses
were complemented by experiments intended to estimate the water-soluble fraction of the sub-
micron carbonaceous material. Our results can be summarized as follows. Depending on the
sample, between 35% and 80% of total aerosol carbon is associated with the submicron fraction.
Total submicron carbon was well correlated with black carbon, a unique tracer for incomplete
combustion. These correlations and the approximately constant total to black carbon ratios,
suggest that the majority of submicron total carbon is of primary combustion derived origin.
No systematic relationship between total submicron aerosol carbon and sulfate concentrations
was found. Sulfate concentrations were, with a few exceptions, significantly higher than total
carbon. Our experiments have demonstrated that water exposure removed between 36% and
7
2% of total carbon from the front filter, suggesting that a substantial fraction of the total
submicron aerosol organic carbon is water-soluble. An unexpected result of this study is that
water exposure of filter samples caused substantial removal of, nominally insoluble, submicron
black carbon. Possible reasons for this observation are discussed.
1
. Introduction
is that characterizing carbonaceous material is
more difficult because of the chemical and physical
Atmospheric aerosol mass is composed of com- complexity of organic aerosol material and the
plex mixtures of chemical species, including sul- large uncertainties caused by sampling artifacts.
fates, other water-soluble inorganic compounds, Analyses of aerosol samples collected during
carbonaceous material, seasalt, and mineral par- the second International Global Atmospheric
ticles. The optical and nucleation properties, Chemistry
(IGAC)
Program’s
Aerosol
sources, and spatial distributions of sulfate aero- Characterization Experiment (ACE-2) provide an
sols are relatively well-known. In contrast, data opportunity to broaden our knowledge of the
on the concentrations, sources, formation mechan- concentrations and properties of carbonaceous
isms, and radiative and nucleation properties of aerosols in the ACE-2 region.
the carbonaceous component of anthropogenic
In this paper, we report results on mass concen-
and natural aerosols are much scarcer. One reason trations of supermicron and submicron aerosol
carbon material obtained by analyzing shipboard
* Corresponding author.
e-mail: TNovakov@lbl.gov
collected ACE-2 samples. For the supermicron
aerosol fraction we present data on total carbon
Tellus 52B (2000), 2


 
229
TC (>1 mm), defined as the sum of organic and impactor has 50% aerodynamic cutoff diameters
black carbon. For the submicron fraction, concen- of 1.0 and 10 mm. Supermicron particles were
tration data on total TC (<1 mm), organic OC collected on an Al foil and submicron particles on
(<1 mm), and black carbon BC (<1 mm) are given. a quartz filter. The impactor also included a
Mass concentration measurements were comple- second, backup, quartz filter the purpose of which
mented by experiments intended to estimate the is to assess the magnitude of sampling artifacts
water-soluble fraction of the submicron carbona- commonly encountered with filter sampling of
ceous material. This paper is structured as follows: organic aerosols. The Al foils and quartz filters
we first describe the sampling and analytical were combusted immediately before use at 600°C
methods, this is followed by measurement results, for 4 h to remove residual organic contaminants.
and by conclusions and discussion.
Foils and filters were loaded and unloaded into a
second impactor without air flow to serve as a
sampling blank. Foils and filters were stored frozen
in tightly capped glass vials until analysis to inhibit
possible microbial activity.
2. Sampling and analytical methods
2
.1. Sampling
Aerosol samples for chemical analysis were col-
lected aboard the RV Vodyanitskiy during the
ACE-2 project. The experiment took place off the
2
.2. Analytical methods
Carbonaceous submicron aerosol material is
coasts of Portugal and North Africa from 20 June principally composed of an organic component,
997 to 25 July 1997. During this time, the ship commonly referred to as ‘‘organic carbon’’ (OC),
1
transited back and forth through the region and a light absorbing component referred to as
extending from 29° to 41° north and 7° to 15° ‘‘black carbon’’ (BC). ‘‘Total carbon’’ (TC) desig-
west. All references to sampling time are reported nates the sum of OC and BC. Carbonate carbon
here in UTC. Dates are given as Day of Year from mineral dust is usually confined to larger
(
3
DOY) where noon on 1 February equals DOY particles and is not considered here. Total carbon
2.5. content in a sample and be accurately determined
Aerosol particles were sampled at 10 m above by any of the combustion methods. Accurate
sea level through a heated mast. The mast determination of BC by thermal and optical
extended 6 m above the aerosol measurement con- methods is a nontrivial problem. Uncertainties in
tainer and was capped with a rotating cone-shaped BC and OC determinations by the commonly
inlet nozzle that was positioned into the relative used methods have been demonstrated by an
wind. Air was pulled through this 5 cm diameter intercomparison study (Shah and Rau, 1990)
inlet nozzle at 1 m3 min−1 and down the 20 cm which showed that most methods agree for TC
diameter mast. The lower 1.5 m of the mast were but significantly differ for OC and BC determina-
heated to dry the aerosol to a relative humidity tions. While the measurement of total carbon on
(
RH) of 55%. 15, 1.9 cm diameter conductive the filter is analytically straight forward, the un-
tubes extending into this heated zone were used certainties associated with positive and negative
to subsample the air stream for the various aerosol sampling artifacts are substantial and cannot be
instruments at flows of 30 l min−1. rigorously calculated from the measurements con-
Samples were collected for chemical analysis ducted during ACE-2. Our attempt to correct for
only when the wind speed was greater than the positive sampling artifact is discussed below.
3
m s−1, the wind direction was forward of the
The definitions of OC and BC used in conjunc-
beam, and the concentration of particles greater tion with thermal methods are based on presumed
than 15 nm in diameter indicated the sample air volatilization and combustion properties of these
was free of contamination from the Vodyanitskiy classes of carbonaceous materials and are opera-
or passing ships. One of the fifteen 1.9 cm diameter tional and method dependent. Uncertainties in
tubes, constructed of stainless steel, was used to black carbon determinations by thermal methods
supply ambient air to a specially designed 2-stage are caused, in large part, by some of the organic
multijet cascade impactor (Berner et al., 1979) for materials present in the sample. These may either
collecting samples for organic analysis. The burn at temperatures close to that of BC, or may
Tellus 52B (2000), 2

2
30
.   .
pyrolize during heating producing a charred mat- ents collected on impactor foils and on quartz
erial having a combustion temperature similar to filters were determined by a thermal — evolved
that of actual BC. Therefore, we refer to the BC gas analysis — (EGA) method (Novakov, 1981)
determined solely by combustion methods as the and, following the rationale outlined above, sup-
apparent black carbon BC(app). Because the plemented by analyses of solvent extracted
organic component is the main interfering mat- samples. In EGA an aliquot of the filter or
erial, its removal prior to BC analyses is an impactor foil sample is progressively heated at a
obvious option for a more accurate BC determina- rate of 12.5°C min−1 in an oxygen atmosphere
tion (Novakov and Corrigan, 1995). The details from 50°C to 600°C. The carbon-containing gases
of these procedures adopted in this study are evolving from the sample as a result of volatiliza-
described below.
tion, decomposition and combustion of the car-
Application of optical transmission measure- bonaceous material are converted to CO over a
2
ments to BC estimation is based on empirical catalyst. The resulting CO concentrations are
2
results that assume the attenuation of visible light measured by a non-dispersive infrared analyzer. A
by the filter deposit is proportional to the surface plot of the rate of carbon evolution versus temper-
concentration of black carbon. The accuracy of ature constitutes the ‘‘thermogram’’ of the sample.
optically determined black carbon concentrations The area under the thermogram is equal to the
critically depends on the choice of the proportion- total carbon content of the sample. This method
ality constant s. Results from a number of field is quantitative for TC within about 10% with a
experiments show that s values may differ by a reproducibility of 3–5% (Dod et al., 1979; Gundel
factor of 5 (Liousse et al., 1993; Niessner and et al., 1984).
Petzold, 1995). Because these values are derived
Both the original and the extracted quartz filters
from calibrations against independently deter- were also characterized by an optical transmission
mined BC concentrations, uncertainties in BC method similar to that described by Rosen et al.
determination could be a major cause of the (1980). This method compares the transmission of
observed variability in the s values. Furthermore, white light through a loaded filter relative to that
theoretical considerations suggest that s may of the blank filter. The relationship between the
depend on whether the BC is internally or extern- optical attenuation, ATN, and the BC concentra-
ally mixed with non-absorbing species such as tion (per unit filter area) is given by ATN=sBC,
sulfates.
where ATN=−100 ln(I/I ), I and I are the light
The disposition of filter samples for EGA ana- intensities transmitted through the loaded and
0
0
lyses was as follows. One half of each filter sample blank filters, and s is a proportionality constant.
(and the corresponding backup filter) was analyzed Because of the uncertainties in the s values (as
to characterize the carbonaceous aerosol in its discussed above) we used ATN measurements
original state. The other halves of filter samples solely to monitor the possible loss of BC from the
were divided into 2 parts. One of these aliquots filter samples during exposure to solvents and not
was first treated with acetone prior to analysis, to quantify the BC concentrations.
and the other aliquot was used for measurements
Analyses of filter and impactor samples give the
involving water extraction as described below. The mass loading of TC, OC and BC present in the
extraction procedures consisted of immersing sep- samples. Volumetric mass concentrations, particu-
arate filter aliquots in 30 ml acetone for about larly of TC and OC, may not be always derived
3
0 min and in 80 ml of de-ionized water for 30 min, from their mass loading unless corrections for
both without agitation. This procedure efficiently sampling artifacts are made. The reason being that
removes most of the organic material but leaves the sampled air contains organic material in both
BC intact (Novakov and Corrigan, 1995). This the particle and gas phase. During filter sample
allows for an interference-free BC determination collection, gaseous organic species may adsorb
and eliminates the need for thermogram peak onto filter material and thus increase the organic
deconvolution. Experiments were also conducted content of the filter deposit (‘‘positive’’ artifact). A
to investigate the effect of water exposure on the positive artifact is most pronounced with filter
carbonaceous aerosol material. sampling because of high surface area of the
Mass concentrations of these carbon compon- filter matrix. Positive artifacts should be less
Tellus 52B (2000), 2


 
231
pronounced with impactor samples because of the
low surface area of the sampling substrates (such
as aluminum foils used in this study) and the fact
that with impactor stages the air is not drawn
through the sampling substrate.
To approximately correct for the positive
artifact we used 2 quartz filters in series as sug-
gested by Fitz (1990). This approach assumes that
the front filter efficiently retains all particles and
that organic gases are adsorbed roughly equally
on both front and back filters. Consequently, if
positive artifacts dominate then the organic aero-
sol concentration is determined from the difference
between the OC loadings on the front and back
filters and the sampled air volume. This technique
has been intercompared in a polluted urban loca-
tion with a number of other techniques and found
to be largely consistent with them (Hering et al.,
1
990). It should be noted, however, that the same
intercomparison study has shown that no tech-
nique for aerosol carbon mass can yet be consid-
ered ideal.
Positive artifacts, however, may not be the only
source of uncertainty in deriving the aerosol
carbon. Some of the semivolatile organic aerosol
species may desorb from the particulate phase
during sampling. This ‘‘negative’’ artifact is especi-
ally pronounced when relatively unpolluted air is
sampled, when it may result in significant underes-
timation of the particulate OC concentration
(Eatough et al., 1993, 1996). Negative artifact has
been demonstrated with filter sampling but it may,
possibly, also affect impactor foil samples.
Fig. 1. Examples of EGA thermograms of (a) super-
micron particles collected on impactor foils, (b) sub-
micron particles collected on front filter and (c) adsorbed
gas-phase species collected on backup filter (sample
set 12).
3
. Results
3
.1. T otal aerosol carbon
The primary analytical data used to derive total
carbon associated with supermicron TC (>1 mm)
and submicron TC (<1 mm) concentrations were region. These loadings are considerably lower than
obtained by EGA analyses of impactor foils and those of exposed foils and filters (Table 1).
front and backup filter samples (examples of ther-
Volumetric mass concentrations of aerosol
mograms are shown in Fig.1). Total carbon load- carbon (in mg m−3) can be directly calculated from
ings determined on impactor foils TC(I), front the air volumes and the carbon loadings (Table 1)
filters TC(F), and back filters TC(B) for all only if the latter are unaffected by sampling
samples (expressed in mg C per filter or foil) are artifacts. As mentioned above samples collected
shown in Table 1. Total carbon loadings on blank on impactor foils are less likely to be affected, at
impactor foils were less than 2 mg and about 5 mg least by positive artifacts, than filter samples.
on blank filters with 40% in the most volatile Consequently, the mass concentrations of total
Tellus 52B (2000), 2

2
32
.   .
Table 1. Sample numbers, sampling start and end,
Total submicron aerosol carbon concentrations
total carbon mass loadings determined on impactor TC (<1 mm) can be derived from TC(F) loading
foils T C(I), front T C(F) and backup filters T C(B), only if a correction for sampling artifacts is made.
and air volumes used for sample collections
To accurately account for sampling artifacts
refined denuder-based methods, such as those
devised by Eatough et al. (1993, 1996), should be
used. These, however, were not available in the
present study. Instead we had to use an approxi-
mate approach based on TC(F) and TC(B) con-
centration differences, as mentioned above. That
this approach is applicable to the samples analyzed
in this study can be argued by the data in Table 1
and by the examples of thermograms shown in
Fig. 1 as discussed below.
As it is seen from Fig. 1 the shapes and struc-
tures of thermograms of the impactor foil and the
front and backup filters (corresponding to sample
set 12) are different and reflect the differences in
the composition of carbonaceous material col-
lected on the 3 sampling substrates. The thermo-
gram of the front quartz (submicron) filter shows
2 low temperature peaks at ~180°C and ~230°C
and a larger and wider structure extending to
mg C per sample
Sampling
start–enda) TC(I) TC(F) TC(B)
Air
volume
(m3)
Sample
no.
1
3
4
5
6
7
8
9
0
1
2
3
174.5–176.7
181.4–184.2
184.3–187.2
187.3–188.2
188.3–189.2
190.6–193.2
193.2–194.2
194.3–195.5
158.8–198.0
199.2–199.7
200.1–203.8
201.3–203.8
32
NA
47
35
29
64
20
41
90
20
44
54
104
54
149
168
58
98
45
75
73
33
31
31
26
20
23
14
18
25
9
48
105
109
38
38
102
38
49
85
17
47
1
1
1
1
45
75
103
24
28
88
a) Times are in UTC. Dates are given as Day of Year
DOY) where noon on 1 February equals DOY 32.
NA: not analyzed.
(
supermicron aerosol carbon TC (>1 mm) could be 600°C. (The latter contains BC as discussed
calculated from TC(I) and the sampled air volume below.) Most of the carbon on the backup quartz
assuming that negative artifacts can be neglected. filter is confined to a single peak at ~180°C
TC (>1 mm) concentration derived in this manner coinciding with the most volatile peak seen with
are listed in Table 2.
the front filter. In contrast, the impactor foil
Table 2. Mass concentrations of supermicron total carbon T C (>1 m) particles, and submicron total
T C (<1 m), black BC (<1 m), organic carbon OC (<1 m), and sulfate; also shown are %s of total, and
black carbon mass removed by water
Water removed fraction (%)
mg C m−3
BC (<1 m)
Sample
no.
BCa)
(EGA)
BCb)
(ATN)
TC (>1 m)
TC (<1 m)
OC (<1 m)
SO 42 −
TC
1
3
4
5
6
7
8
9
0
1
2
3
0.7
NA
0.4
1.0
0.8
0.6
0.5
0.8
1.1
1.2
0.9
0.6
1.5
0.2
1.1
3.9
1.0
0.7
0.8
1.2
0.6
2.2
1.1
0.9
0.6
0.1
0.3
1.1
0.4
0.2
0.3
0.4
0.3
0.8
0.3
0.3
0.9
0.1
0.8
2.8
0.6
0.5
0.5
0.8
0.3
1.4
0.8
0.6
4.7
1.6
2.5
3.9
4.2
4.8
6.1
9.6
6.3
6.2
6.1
5.6
58
36
53
70
64
72
50
59
53
58
49
72
60
<4.0
5
64
70
52
52
50
45
12
83
#80
67
54
83
1
1
1
1
53
23
57
72
a) From EGA analysis.
b) From ATN measurements.
NA: not analyzed.
Tellus 52B (2000), 2


 
233
thermogram shows a broad structure extending
from ~200°C to above 500°C, with a peak at
~
480°C most likely due to BC. Noticeably absent,
however, are any low temperature peaks such as
those seen with filter samples. This comparison
suggests that the species giving rise to the 180°C
peak, seen with both front and back filters, are
gas phase organic species adsorbed on the quartz
filter matrix. (Particles are not expected to penet-
rate the front filter.) Absence of volatile species in
the foil thermogram is consistent with this sugges-
tion because adsorption of gaseous organics is not
favored on the low surface area impactor foil.
Assuming that positive artifacts dominate and
all the carbon on the backup filter is caused by
such artifacts, then the difference in the carbon
loading on the front and back filters can be used
to derive an approximate measure of total submic-
ron aerosol carbon. A condition for the validity
of this assumption is that both front and back
filters are similarly saturated with adsorbed
species. That this condition may be satisfied is
suggested by comparing TC(F) and TC(B) load-
ings shown in Table 1. The spread in TC(B) values
is much smaller than the spread in TC(F) values.
With the exception of sample 11, collected with
the lowest air volume, TC(B) values range from
about 14 to 33 mg and TC(F) loadings range
between about 45 to 168 mg. TC (<1 mm) concen-
trations derived in the manner described are listed
in Table 2. We note, however, that if negative
artifacts are significant then TC (<1 mm) and
possibly T C(>1 mm) would be underestimated.
The present data, however, do not allow for the
quantification of this underestimation.
Fig. 2. Thermograms of (a) untreated, (b) acetone and
(c) water exposed filter aliquots (sample set 5) taken from
the same front quartz filter.
3
.2. Black carbon
The difficulty of BC determination by thermal
properties alone is illustrated by thermograms no OC, which is a very unlikely possibility.
sample set 5) shown in Fig. 2. The high temper- Alternatively, the complexity of the high temper-
(
ature thermogram region of the front filter is ature region is caused by a mixture of BC and
broad and apparently composed of several indi- interfering organic materials such as the pyrolysis
vidual peaks some of which undoubtedly are due products (i.e. charred material) and high molecular
to black carbon. It is clear that taking the entire weight organics (i.e. tarry materials) both having
high temperature feature as the representation of combustion temperatures close to those of true
BC will seriously overestimate its concentration. black carbon. Consequently, removal of most of
As can be seen from Fig. 2a, the BC, or more the organic material from the sample prior to
appropriately the BC(app) if defined in this analysis, for example by exposure to acetone,
manner, will account for most of the carbonaceous should give a more accurate measure of BC
material and the sample would contain essentially (Novakov and Corrigan, 1995). The effect of
Tellus 52B (2000), 2

2
34
.   .
acetone exposure on thermograms is illustrated by (Fig. 2a) and water exposed samples show that
Figs. 2a,b, where thermograms of the untreated water exposure resulted in the loss of about 70%
and the acetone exposed aliquot taken from the of the original total carbon. The results of such
same front quartz filter (sample 5) are shown. As measurements performed on all samples and sum-
expected acetone removed most of the organics marized in Table 2 show that depending on the
and revealed a well defined Gaussian BC peak at sample, between 36% and 72% of TC loading on
~480°C. The small peak at the lowest temperature the front filters is apparently water soluble. These
is due to residual solvent. results are not unexpected in view of the fact that
That acetone exposure does not physically some of the organic aerosol species are known to
remove BC from the sample was shown by optical be water soluble (Saxena and Hildemann, 1996),
attenuation measurements performed on the and contribute to cloud condensation nuclei
samples before and after acetone exposure. Ratios (Novakov and Penner, 1993; Rivera-Carpio
of optical attenuation (ATN) on all samples before et al., 1996).
and after acetone exposure ranged from 0.94 to
1
An unexpected result of these analyses is that
.16 with an average of 1.04± 0.08. These results water exposure appears to significantly affect the
show that no appreciable loss of BC occurs upon black carbon loading. As seen in Fig. 2 water
acetone exposure. The slight decrease in optical exposure affected the entire thermogram temper-
transmission of 4% could be due, in part, to the ature region, including the highest temperature
removal of high molecular weight slightly absorb- peak which, as shown above, contains the BC.
ing organic material. Furthermore, the filter back- The peak at about 550°C seen with the water
ing used for sample collection consisted of a exposed sample represents the apparent black
perforated disk. As a consequence the deposit on carbon, BC(app), due both to BC and, possibly,
the filters appeared in a similar pattern of dots, some thermally stable organic matter not removed
making the precise alignment in the ATN appar- by water exposure. (The shift of the apparent
atus difficult, and thus the ATN results less accur- black carbon peak relative to the BC peak in the
ate. We note, however, that all EGA measurements acetone extracted sample is caused by the removal
were normalized to the number of dots on the of water-soluble catalysts, such as Na and K
sample analyzed.
containing compounds, that promote combustion
Submicron BC (<1 mm) concentrations deter- (Lin and Friedlander, 1988; Novakov and
mined in this manner are given in Table 2, where Corrigan, 1995). A comparison of BC(app) area
OC (<1 mm) concentrations derived from TC and the BC peak observed in the thermogram of
(<1 mm)–BC (<1 mm) differences are also shown. the acetone extracted sample (Fig. 2) shows that
water exposure resulted in about 64% loss of
apparent black carbon. (Both BC areas were
determined by a Gaussian peak fit assuming a
3
.3. EVect of water exposure on OC and BC
The original purpose of EGA measurements of symmetrical peak shape.) The apparent BC loss
water exposed filter samples was to estimate the for all samples determined in this manner by EGA
water soluble fraction of organic aerosol compon- is shown in Table 2.
ent. We have arbitrarily chosen a water volume of
8
During the analyses of the first batch of 5
0 ml to simplify the analytical procedure. We samples, it became obvious that most water
realize water solubility is a relative term and do exposed samples showed distinctly less coloration
not mean to imply that the water soluble fraction than the corresponding untreated and acetone
measured here would be equal to the water soluble exposed samples. Because BC is the principal light
fraction at the liquid water content of humid air absorbing aerosol species and mainly responsible
or a cloud droplet.
for the deposit coloration, the observed decrease
The approach was to compare the thermograms in sample coloration indicated BC removal.
and total carbon content of identical aliquots of Consequently, the remaining original and water
untreated and water exposed front filter samples. exposed samples were also analyzed by optical
A thermogram obtained with a water exposed attenuation (transmission) measurements. The
filter (sample 5) is shown in Fig. 2c. Comparison reduction in ATN values upon water exposure is
of the areas under the thermograms of untreated an alternative qualitative measure of BC loss. BC
Tellus 52B (2000), 2


 
235
losses estimated from ATN measurements are
shown in Table 2. Black carbon losses estimated
from ATN changes are generally larger than those
estimated from EGA analyses. This discrepancy is
not unexpected because BC(app) is not necessarily
a true measure of BC, and ATN derived BC may
be subject to unknown effect of sample composi-
tion. A more accurate analytical procedure would
require application of acetone extraction to water
exposed samples. Limited sample material, how-
ever, prevented us from doing this and other
possible experiments to quantify the BC loss.
Therefore, data on BC losses derived both by
EGA and ATN should be viewed as semi-quanti-
tative estimates.
Fig. 3. Plot of total submicron aerosol carbon, corrected
for positive artifacts, TC (<1 mm) versus black carbon
BC (<1 mm) concentrations.
4
. Discussion and conclusions
are distinctly different reflecting the compositional
differences of the 2 fractions (Fig. 1).
Black carbon, a unique indicator of incomplete
In the following we discuss our results on:
(
1) the partitioning of total carbon between combustion, was detected in all samples collected
supermicron and submicron particles; (2) relation- during the cruise. In Fig. 3 total submicron carbon
ships between submicron total and black carbon concentrations TC (<1 mm) are plotted versus the
and sulfate concentrations obtained in this study, BC concentrations (Table 2). Linear regression
and compare these with previously obtained data of these data resulted in relation TC (<1 mm)=
over the anthropogenically influenced western 3.37 (BC)−0.14, with a correlation coefficient
Atlantic Ocean during the Tropospheric Aerosol r2=0.94. High correlation coefficients and the
Radiative Forcing Experiment (TARFOX) (Hobbs low intercept suggest that most of the aerosol
et al. 1996); and (3) removal of total and black carbon is combustion derived.
carbon from filters upon water exposure. Table 2
summarizes the pertinent data obtained in this from our data is 0.35± 0.07. For comparison, a
study. mean submicron BC/TC ratio of 0.43 was meas-
The average submicron BC/TC ratio derived
Before proceeding with the data discussion we ured at Aveiro on the coast of Portugal (Nunes
emphasize that the ACE-2 Vodyanitskiy carbon and Pio, 1993). Measurements of Kuhlbush et al.
aerosol data set must be considered as a regional (1998) BC/TC ratios for central Europe of
representation of concentrations during this time 0.36–0.38 agree with our value. The fact that these
period as sampling intervals did not correspond ratios are comparable (considering that different
to a single air mass back trajectory. Hence, mul- sampling and analytical procedures were used) is
tiple air mass sources could have contributed to consistent with the notion that the carbonaceous
one carbon aerosol sample.
aerosol collected during the Vodyanitskiy cruise is
The data show that total carbon in supermicron dominated by continental combustion sources.
particles, TC (>1 mm), is at times comparable to Substantial secondary organic and non-combus-
submicron total carbon, TC (<1 mm). Depending tion derived organic aerosols material result in
on the sample, between 35% and 80% of total lower BC/TC ratios than expected for primary
(sub- and supermicron) carbon is associated with combustion particles (Turpin and Huntzicker,
the submicron fraction. Here we report only the 1995). The lack of substantial deviations of these
TC (>1 mm) concentrations because supermicron ratios measured from the average suggests that
BC was not determined. It should be noted, how- the majority of submicron total carbon in the
ever, that the shapes and structure of the thermo- cruise region was of primary origin. Our data
grams of the supermicron and submicron samples alone, however, are insufficient to quantify the
Tellus 52B (2000), 2

2
36
.   .
During TARFOX the SO2− concentration aver-
4
aged over 30 flights was 5.7± 5.9 mg m−3, covering
the range from 0.1 to 19.5 mg m−3. Although the
ments were comparable the maximum values
average SO2− concentrations from the 2 experi-
4
in TARFOX were about 2× higher than found
in this study. The average submicron TC con-
centration measured during the cruise was
1.3± 0.9 mg m−3, ranging from 0.2 to 3.9 mg m−3.
In contrast, the average TARFOX total carbon
was 4.9± 2.8 mg m−3 (range 0.6 to 9.9 mg m−3) or
3
.8 times higher during TARFOX than found in
the present study. The average SO2−/TC ratio of
4
5
.3± 2.9 for ACE-2 (using TC<1 mm) is about
Fig. 4. Time series of submicron nss sulfate and TC
5× higher than the ratio of 1.0± 0.6 averaged for
(<1 mm) concentrations. Dates are given as Day of Year
(
DOY) where noon on 1 February equals DOY 32.5. all TARFOX flights.
Sampling times are in UTC. Numbers correspond to
carbon sample number.
This comparison, however, may not be justified
for 2 reasons. First, The ACE-2 concentrations
were derived from shipboard (surface) collected
contribution of secondary organics to total sub- samples, while the TARFOX data were obtained
micron aerosol carbon. with samples collected at different altitudes.
An accurate comparison of aerosol sulfate and Second, in contrast to ACE-2 experiment when
carbon concentrations must take into account the both submicron and supermicron aerosol samples
different sampling times involved. The samples for were obtained, the TARFOX aerosol intake
carbon analyses were collected over much longer system collected particles with diameters smaller
time periods than the sulfate samples so that the than ~5 mm. Therefore,
a more meaningful
carbon samples average multiple air masses. approach is to compare the SO2−/TC ratios by
4
Variability in the measured TC and SO2− concen- using the sum of submicron and supermicron TC
4
trations for the <1 mm aerosol fraction is shown for ACE-2 and the TARFOX ratios corresponding
in Fig. 4.
to the lowest sampling altitudes (<0.3 km). We
Submicron aerosol carbon concentrations were note that Quinn et al. (this issue) have shown that
measured at 2 land-based sites during ACE-2, the nss SO2− contribution to the supermicron size
4
Sagres located on the southwest tip of Portugal range is small. The SO2−/TC ratios defined in this
4
and Punto del Hidalgo located on Tenerife. The manner are 2.9± 1.3 (range 0.8–4.8) for ACE-2
average submicron OC concentration measured and 1.6± 0.7 (range 0.5–2.3) for TARFOX. With
on the ship during polluted conditions these assumptions the ACE-2 ratio is 1.8 times
(
0.89 mg m−3) was very similar to the average higher than the TARFOX ratio. For comparison
polluted value measured at Sagres (0.75± 0.32 the SO2−/TC ratio for TARFOX samples taken
4
mg m−3) and Punto del Hidalgo (0.52± 0.27 at altitudes above 2.5 km is reduced to 0.6± 0.3,
possible as a purely marine sample was not col- and carbon mass fraction observed in that
mg m−3). A comparison of marine values is not consistent with the altitude dependence of sulfate
lected onboard the ship.
It is instructive to compare the aerosol carbon
experiment.
The lower values of SO2−/TC ratio for
4
and sulfate concentrations from the eastern TARFOX suggest that the aerosol over the west-
Atlantic obtained in this study with the aircraft ern Atlantic is enriched in TC relatively to that
measurements over the western Atlantic during found during ACE-2. A possible explanation is
TARFOX, conducted off the Virginia coast of the suggested by the corresponding TC/BC ratios.
United States in July 1996 (Novakov et al., 1997; The average BC/TC ratio determined in this study
Hegg et al., 1997). The SO2− concentration meas- was 0.35 which is considerably higher than the
4
ured during the cruise (Table 2) ranged from 1.6 BC/TC ratio of 0.1 estimated from the TARFOX
to 9.6 mg m−3, with an average of 5.1± 2.0 mg m−3. data. (We note, however, that the TARFOX black
Tellus 52B (2000), 2


 
237
carbon fraction was inferred from TC concentra- carbon–sulfate mixtures, especially when the sulf-
tions and concurrently measured aerosol absorp- ate is the dominant species. Transmission electron
tion coefficients.) This TARFOX ratio is roughly microscopy measurements during ASTEX-MAGE
consistent with ground based measurements in and ACE-2 confirm that BC particles over the
rural areas of the eastern US which gave the Atlantic Ocean generally occur within sulfates
BC/TC ratio for the PM aerosol of about 0.15 (Posfai et al., 1999).
2.5
(
Malm et al., 1994). These low ratios suggest that
Rapid dissolution of sulfate may dislodge the
a significant contribution of organic material over small BC particles occluded within the larger
the eastern Atlantic may, at least in part, be of sulfate–carbon agglomerate. Consequently, a frac-
biogenic or secondary anthropogenic origin.
tion of the BC will be introduced into the water
Finally, we turn to a discussion of our results in the form of a suspension. The removed fraction
on the removal of TC and BC from filter samples should vary depending on the relative amounts of
upon contact with water. Our experiments demon- BC and sulfate and their state of mixing. Indeed,
strate that between 36% and 72% of total carbon the smallest BC loss occurred in samples having
is removed from the front filter as the result of the lowest SO2− concentrations (samples 3 and
4
water exposure (Table 2). This result is not unex- 4). Finally, the possibility that nominally hydro-
pected because the complex aerosol organic mat- phobic BC may be rendered hydrophilic by chem-
erial does contain a water-soluble fraction. An ical reactions occurring during atmospheric
unexpected result obtained in this study pertains transport cannot be excluded. A better under-
to the effect of water exposure on black carbon. standing of the BC loss indicated by our experi-
Water exposure of filter samples caused substantial ments warrants more study. It is clear that
removal of submicron BC from front filters. The a mechanism similar to that affecting BC could
fraction of BC removed ranged from about 5 to also influence the organic aerosol component.
7
0% and 12 to 80%, estimated from EGA analyses However, as the data show 36 to 53% of TC was
and optical measurements, respectively (Table 2). removed from the 2 samples (3 and 4) which
These finding are unexpected because BC is not showed the smallest BC loss. Therefore, a substan-
water soluble. In contrast to ACE-2 samples, water tial part of the TC or, more accurately, its OC
exposure of source dominated aerosol samples component, appears to be intrinsically water sol-
(collected in a highway tunnel and at a suburban uble. The solubility of carbonaceous aerosol par-
traffic influenced site), affected the OC fraction ticles may influence their transport and removal.
but not the BC. Because of the inherent BC
insolubility its removal must involve an indirect
process.
At this time we cannot offer an exact explana-
tion of this phenomenon. It may be that BC
5. Acknowledgements
This work was supported by the Director, Office
removal depends on whether fresh or aged aerosol of Science, Environmental Sciences Division,
is sampled. In addition, the physico-chemical US Department of Energy, under contract
properties of both the aerosol particles and the DE-AC03-76SF00098, and by NOAA under an
filter deposit will be different depending on interagency agreement NA97AANAGO187. Work
whether the BC is externally or internally mixed at PMEL was funded by the Aerosol Component
with water-soluble species such as sulfate. In an of the NOAA Climate and Global Change
external mixture, carbonaceous particles will exist Program. This research is a contribution to the
separate from water-soluble species. If BC particles International Global Atmospheric Chemistry
in such samples strongly adhere to each other and (IGAC) Core Project of the International
to the filter matrix, then water exposure will not Geosphere-Biosphere Programme (IGBP) and is
dislodge the BC from the filter. However, aged part of the IGAC Aerosol Characterization
aerosol may be largely composed of internal Experiments (ACE).
Tellus 52B (2000), 2

2
38
.   .
REFERENCES
Berner, A., Lurzer, C., Pohl, F., Preining, O. and
Wagner, P. 1979. The size distribution of the urban
aerosol in Vienna. Sci. T ot. Environ. 13, 245–261.
Dod, R. L., Rosen, H. and Novakov, T. 1979. Lawrence
Berkeley Laboratory Report LBL-8696.
particle concentration and optical extinction in the
United States. J. Geophys. Res. 99, 1347–1370.
Niessner, R. and Petzold, A. 1995. Intercomparison study
on soot-selective methods — field study results from
several polluted areas in Germany. J. Aerosol Sci. 26
(suppl. 1), 393–394.
Eatough, D. J., Wadsworth, A., Eatough, D. A., Craw-
ford, J. W., Hansen L. D. and Lewis, E. L. 1993. A Novakov, T. 1981. Microchemical characterization of
multiple-system, multi-channel diffusion denuder sam-
pler for the determination of fine-particulate organic
material in the atmosphere. Atmos. Environ. 27A,
aerosols. In: Nature, aim and methods of microchemistry
(eds. Grasserbauer, M. M. and Belcher, R.). Springer,
New York, pp. 141–165.
1
213–1219.
Novakov, T. and Penner, J. E. 1993. Large contribution
of organic aerosols to cloud–condensation–nuclei con-
centrations. Nature 365, 823–826.
Eatough, D. J., Eatough, D. A., Lewis, L. and Lewis,
E. L. 1996. Fine particulate chemical composition and
light extinction at Canyonlands National Park using Novakov, T. and Corrigan, C. E. 1995. Thermal charac-
organic particulate material concentrations obtained
with a multisysytem, multichannel diffusion denuder
sampler. J. Geophys. Res. 101, 19,515–19,531.
Fitz, D. R. 1990. Reduction of the positive organic
artifact on quartz filters. Aerosol Sci. T echnol. 12,
terization of biomass smoke particles. Mikrochimica
Acta 119, 157–166.
Novakov, T. and Corrigan, C. E. 1996. Cloud condensa-
tion nucleus activity of the organic component of bio-
mass smoke. Geophys. Res. L ett. 23, 2141–2144.
Novakov, T, Hegg, D. A. and Hobbs, P. V. 1997. Air-
borne measurements of carbonaceous aerosols on the
East Coast of the United States. J. Geophys. Res. 102,
30,023–30,030.
Nunes, T. V. and Pio, C. A. 1993. Carbonaceous aerosols
in industrial and coastal atmospheres. Atmos. Environ.
27A, 1139–1346.
1
42–148.
Gundel, L. A., Dod, R. L., Rosen, H. and Novakov, T.
984. The relationship between optical attenuation
1
and black carbon concentration for ambient and
source particles. Sci. T otal Environ. 36, 197–202.
Hegg, D. A., Livingston, J., Hobbs, P. V., Novakov, T.
and Russell, P. 1997. Chemical apportionment of
aerosol column optical depth off the Mid-Atlantic Posfai, M., Anderson, J. R., Buseck, P. R. and Siever-
Coast of the United States. J. Geophys. Res. 102,
5,293–25,303.
Hering, S. V. et al. 1990. Comparison of sampling
methods for carbonaceous aerosols in ambient air.
Aerosol Sci. T echnol. 12, 200–213.
Hobbs, P. V., Russell, P. B. and Stowe, L. L. 1996.
Tropospheric Aerosol Radiative Forcing Observa-
tional Experiment (TARFOX). In: Atmospheric aero-
sols: a new focus of the international global atmospheric
chemistry project. Int. Global Atmos. Chem. Project
Office, MIT Press, Cambridge, Mass.
ing, H. 1999. Soot and sulfate aerosol particles in the
remote marine troposphere. J. Geophys. Res. 104,
21,685–21,693.
2
Rivera-Carpio, C. A., Corrigan, C. E., Novakov, T.,
Penner, J. E., Rogers, C. F. and Chow, J. C. 1996.
Derivation of contributions of sulfate and carbona-
ceous aerosols to cloud condensation nuclei from mass
size distributions. J. Geophys. Res. 101, 19,483–19,493.
Rosen, H., Hansen, A. D. A., Dod, R. L. and Novakov, T.
1980. Soot in urban atmospheres: determination by
an optical absorption technique. Science 208, 741–744.
Kuhlbusch, T. A. J., Hertlein, A.-M. and Sch u¨ tz, L. W. Saxena, P. and Hildemann, L. M. 1996. Water-soluble
1
998. Sources, determination, monitoring, and trans-
organics in atmospheric particles. A critical review of
the literature and application of thermodynamics to
identify candidate compounds. J. Atmos. Chem. 24,
57–109.
port of carbonaceous aerosols in Mainz, Germany.
Atmos. Environ. 32, 1097–1110.
Lin, C. and Friedlander, S. K. 1988. A note on the use
of glass fiber filters in the thermal analysis of carbon Shah, J. J. and Rau, J. A. 1990. Carbonaceous species
containing aerosols. Atmos. Environ. 22, 605–607.
Liousse, C., Cachier, H. and Jennings, S. G. 1993. Optical
and thermal measurements of black carbon aerosol
content in different environments: variation of the
specific attenuation cross-section. Atmos. Environ. 27A,
1203–1211.
Malm, W. C., Sisler, J. F., Huffman, D., Eldred, R. A.
and Cahill, T. A. 1994. Spatial and seasonal trends in
methods comparison study. Inter laboratory round robin
interpretation of results. Final report to Research Divi-
sion, California Air Resources Board, Contract no.
A832–154.
Turpin, B. J. and Huntzicker, J. J. 1995. Identification of
secondary organic aerosol episodes and quantification
of primary and secondary organic concentrations
during SCAQS. Atmos. Environ. 29, 3527–3544.
Tellus 52B (2000), 2