Filetype PDF | Posted on 02 Oct 2022 | 3 years ago
Authentication
digital resources
163x Filetype PDF File size 0.53 MB Source: acp.copernicus.org
Atmos. Chem. Phys., 10, 4625–4641, 2010 Atmospheric
www.atmos-chem-phys.net/10/4625/2010/ Chemistry
doi:10.5194/acp-10-4625-2010
©Author(s) 2010. CC Attribution 3.0 License. andPhysics
Organic aerosol components observed in Northern Hemispheric
datasets from Aerosol Mass Spectrometry
1 1 2,* 3,4 2 3,4 1,5
N. L. Ng , M. R. Canagaratna , Q. Zhang , J. L. Jimenez , J. Tian , I. M. Ulbrich , J. H. Kroll ,
3,4 6 3,7 7 6 8 8
K.S.Docherty ,P.S.Chhabra ,R.Bahreini ,S.M.Murphy ,J.H.Seinfeld ,L.Hildebrandt ,N.M.Donahue ,
3,9,10 10 ´ ˆ 10 11 11 1
P. F. DeCarlo , V. A. Lanz , A. S. H. Prevot , E. Dinar , Y. Rudich , and D. R. Worsnop
1Aerodyne Research, Inc. Billerica, MA, USA
2Atmospheric Sciences Research Center, State University of New York, Albany, NY, USA
3CIRES,University of Colorado, Boulder, CO, USA
4Department of Chemistry and Biochemistry, University of Colorado, Boulder, CO, USA
5Department of Civil and Environmental Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA
6Department of Chemical Engineering, California Institute of Technology, Pasadena, CA, USA
7NOAA,EarthSystemResearchLaboratory, Boulder, CO, USA
8Center for Atmospheric Particle Studies, Carnegie Mellon University, Pittsburgh, PA, USA
9Department of Atmospheric and Oceanic Science, University of Colorado, Boulder, CO, USA
10Laboratory of Atmospheric Chemistry, Paul Scherrer Institut, Villigen, Switzerland
11Department of Environmental Sciences, Weizmann Institute of Science, Rehovot 76100, Israel
*nowat: Department of Environmental Toxicology, University of California, Davis, CA, USA
Received: 27 November 2009 – Published in Atmos. Chem. Phys. Discuss.: 23 December 2009
Revised: 14 April 2010 – Accepted: 29 April 2010 – Published: 20 May 2010
Abstract. In this study we compile and present results from ponent mass spectrum) and lower f (ratio of m/z 43 to
43
the factor analysis of 43 Aerosol Mass Spectrometer (AMS) total signal in the component mass spectrum) than SV-OOA.
datasets (27 of the datasets are reanalyzed in this work). A wide range of f and O:C ratios are observed for both
44
The components from all sites, when taken together, pro- LV-OOA(0.17±0.04,0.73±0.14)andSV-OOA(0.07±0.04,
vide a holistic overview of Northern Hemisphere organic 0.35±0.14) components, reflecting the fact that there is a
aerosol (OA) and its evolution in the atmosphere. At most continuum of OOA properties in ambient aerosol. The OOA
sites, the OA can be separated into oxygenated OA (OOA), components (OOA, LV-OOA, and SV-OOA) from all sites
hydrocarbon-like OA (HOA), and sometimes other compo- cluster within a well-defined triangular region in the f vs.
44
nents such as biomass burning OA (BBOA). We focus on f space, which can be used as a standardized means for
43
the OOA components in this work. In many analyses, the comparing and characterizing any OOA components (labo-
OOA can be further deconvolved into low-volatility OOA ratory or ambient) observed with the AMS. Examination of
(LV-OOA) and semi-volatile OOA (SV-OOA). Differences the OOA components in this triangular space indicates that
in the mass spectra of these components are characterized OOAcomponentspectrabecomeincreasinglysimilartoeach
+
in terms of the two main ions m/z 44 (CO ) and m/z 43 other and to fulvic acid and HULIS sample spectra as f (a
2 44
(mostly C H O+), which are used to develop a new mass surrogate for O:C and an indicator of photochemical aging)
2 3
spectral diagnostic for following the aging of OA compo- increases. This indicates that ambient OA converges towards
nents in the atmosphere. The LV-OOA component spectra highly aged LV-OOA with atmospheric oxidation. The com-
have higher f (ratio of m/z 44 to total signal in the com- mon features of the transformation between SV-OOA and
44
LV-OOAatmultiple sites potentially enable a simplified de-
Correspondence to: M. R. Canagaratna scription of the oxidation of OA in the atmosphere. Com-
(mrcana@aerodyne.com) parison of laboratory SOA data with ambient OOA indicates
Published by Copernicus Publications on behalf of the European Geosciences Union.
4626 N. L. Ng et al.: Organic aerosol components in Northern Hemispheric datasets
that laboratory SOA are more similar to SV-OOA and rarely mitz et al., 2008; Ulbrich et al., 2009). In those studies the
become as oxidized as ambient LV-OOA, likely due to the OOA-1representedthemoreoxidized,agedaerosolsandthe
higher loadings employed in the experiments and/or limited OOA-2 represented the less oxidized, fresher secondary or-
oxidant exposure in most chamber experiments. ganic species. Temporal correlations with sulfate and nitrate
(Lanz et al., 2007; Ulbrich et al., 2009) as well as direct
volatility measurements (Huffman et al., 2009) in those stud-
1 Introduction ies further showed that OOA-1 is less volatile than OOA-2.
Since the OOA-1 and OOA-2 terminology does not convey
Organic aerosols (OA) constitute a substantial fraction (20- the knownphysiochemicalpropertiesoftheseOOAsubcom-
90%) of submicron aerosols worldwide and a full under- ponents, in the following discussion we will refer to these
standing of their sources, atmospheric processing, and prop- subcomponents as low-volatility OOA (LV-OOA) and semi-
erties is important to assess their impacts on climate, human volatile OOA (SV-OOA),respectively. It is important to note
health, and visibility (Kanakidou et al., 2005; Zhang et al., that the assignment of LV-OOA and SV-OOA to the compo-
2007; Kroll and Seinfeld, 2008; Hallquist et al., 2009). The nents identified at each site is not absolute, meaning that the
Aerodyne AMS provides quantitative data on inorganic and LV-OOAatonesitedoesnothavethesamecompositionasin
organic aerosol species in submicron non-refractory aerosol another site. This is an expected result since factor analysis
particles with high-time resolution. In recent years, several has been applied separately to each site. Thus here the ter-
techniqueshavebeenemployedtodeconvolvethemassspec- minology for the OOA subtypes is relative for each site (i.e.,
at each site the component with a higher f is referred to
tra of the organic aerosols acquired with the AMS includ- 44
as LV-OOA and the component with a lower f is referred
ing custom principal component analysis (CPCA) (Zhang 44
to as SV-OOA regardless of the absolute values of f ). An
et al., 2005a), multiple component analysis (MCA) (Zhang 44
et al., 2007), hierarchical cluster analysis (Marcolli et al., absolute volatility scale for LV-OOA and SV-OOA is being
2006), positive matrix factorization (PMF) (Paatero and Tap- investigated (e.g. Faulhaber et al., 2009; Cappa et al., 2009)
per, 1994; Paatero 1997; Lanz et al., 2007; Nemitz et al., but requires a better understanding of the volatility measure-
2008; Aiken et al., 2008; Aiken et al., 2009b; Ulbrich et al., ments of ambient aerosols (e.g., with thermodenuders).
2009),andtheMultilinearEngine(ME-2)(Lanzetal.,2008). ThegoalofthisstudyistocompareandcontrasttheOOA
components identified in multiple field studies in order to
Multivariate analysis by Zhang et al. (2007) showed better characterize the sources and evolution of OA in the at-
that OA at multiple sites can be described by two main mosphere. We present results from the factor analysis of 43
components: hydrocarbon-like organic aerosol (HOA) and AMSdatasets. 27 of the datasets, which encompass a major-
oxygenated organic aerosol (OOA). Biomass burning OA ity of the sites in Zhang et al. (2007), are reanalyzed as part of
(BBOA)andother local primary sources have also been ob- this work. We focus mainly on the OOA component and the
served (Jimenez et al., 2009). OOA accounts for a large frac- reanalysis allows for further deconvolution of the total OOA
tion (72±21%) of the total organic mass at many locations componentreportedbyZhangetal.(2007)intoLV-OOAand
(Zhang et al., 2007; Jimenez et al., 2009). Studies from mul- SV-OOA components. The OA components resulting from
tiple locations show that the HOA component correlates well this work were used by Jimenez et al. (2009) to form the ba-
with primary tracers such as CO and NOx (e.g. Zhang et al., sis of a modeling framework that links oxidation and volatil-
2005b; Lanz et al., 2007; Aiken et al., 2009b; Ulbrich et ity to capture the evolution of OA in the atmosphere. In this
al., 2009) and can be considered as a surrogate of combus- manuscript we combine the factor analysis results from the
tion primary OA (POA). BBOA correlates with acetonitrile, ambientdatasetstogethertoobtainaholisticviewofhowthe
levoglucosan, and potassium, and can be considered a surro- AMSambient component mass spectra change across envi-
gate of BB POA (Aiken et al., 2009a, b). The OOA compo- ronments with different sources and aerosol processes. The
nent has been shown to be a good surrogate of secondary OA common features of the component spectra are used to de-
(SOA) in multiple studies, correlating well with secondary velop a new mass spectral diagnostic for following the atmo-
species such as O and Ox (de Gouw et al., 2005; Zhang et spheric aging of OA components in the atmosphere. Finally,
3
al., 2005a, b, 2007; Volkamer et al., 2006; Lanz et al., 2007; since the AMS has been employed in many laboratory ex-
Herndon et al., 2008). periments over the years, a series of chamber data (both pub-
In many analyses, two types of OOA have been identi- lished and unpublished) are also integrated and compared to
fied. The two broad subtypes differ in volatility and degree the ambient data. Chamber data provide the basis for sim-
of oxidation (Jimenez et al., 2009), as indicated by the ratio ulating SOA formation in the atmosphere. Hence, it is im-
of m/z 44 to total signal in the component mass spectrum portant to evaluate whether the results from chamber exper-
(f ). The more oxidized component (higher f ) has pre- iments are representative of the atmosphere; similarities and
44 44
viously been referred to as OOA-1 while the less oxidized differences between ambient OOA and laboratory SOA are
component (lower f ) has previously been referred to as examined and discussed.
44
OOA-2 (Lanz et al., 2007; Aiken et al., 2008, 2009b; Ne-
Atmos. Chem. Phys., 10, 4625–4641, 2010 www.atmos-chem-phys.net/10/4625/2010/
N. L. Ng et al.: Organic aerosol components in Northern Hemispheric datasets 4627
2 Method analysis. The analysis and input error calculations are per-
formed following the procedures described by Ulbrich et
The organic aerosol data have been obtained with the Aero- al. (2009). The optimal number of PMF components is de-
dyne quadrupole mass aerosol spectrometer (Q-AMS), the termined by carefully examining the scaled residuals, evalu-
compact time-of-flight mass spectrometer (C-ToF-AMS), or ating the component’s diurnal cycles and factor correlations
the high-resolution ToF-AMS (HR-ToF-AMS). The instru- with external tracers (including CO, O3, NOx, NOy, SO2,
ment design and operation of each version of the AMS has − 2−
NO , and SO , when available), and comparing the com-
been described in detail by Jayne et al. (2000), Drewnick 3 4
et al. (2005), and DeCarlo et al. (2006), respectively, and ponent spectra with source mass spectra from the AMS mass
reviewed by Canagaratna et al. (2007). Factor analyses of spectra database (Ulbrich et al., 2009). The PMF2 optimiza-
the data from Riverside, CA, Whistler (Canada), and Mex- tion algorithm starts from random initial conditions, which
ico City (both ground and aircraft data) are based on high can be changed by varying the value of the SEED input pa-
resolution (HR) data from HR-ToF-AMS, while analyses for rameter. Multiple solutions generated with different SEED
all other locations are based on unit mass resolution (UMR) values are carefully examined to explore the possibility of
data. The details (locations, times, previous publications, multiple local minima in the solutions. The uncertainty in
etc.) of all datasets included in this study are given in the component mass spectra and time series due to rotation am-
supplementary material. biguity is also examined by performing PMF analysis over
In this work we present factor analysis results from 43 a range of FPEAK values. Overall, the effect of positive
AMS datasets. As part of this work, we performed PMF FPEAKistocreate more near-zero values in the mass spec-
(Paatero and Tapper, 1994; Paatero, 1997) based factor anal- tra and decrease the number of near-zero values in the time
ysis of the organic aerosol mass spectra observed at 27 of the series; negative FPEAK values have the opposite effect (Ul-
sites. For some of the urban downwind/rural/remote sites, a brich et al., 2009). For all sites, improved correlations with
hybrid of PMF/MCA approach is employed (Cottrell et al., external tracers or mass spectra are not found for FPEAK
2008; Nemitz et al., 2008). The factor analyses of the re- valuesdifferent from 0. Thus the FPEAK=0solutions, which
maining 16 sites were performed previously and the results bound the observed data most closely, are chosen for all the
are discussed in more detail in the corresponding publica- sites analyzed in this study.
tions: Beijing (Sun et al., 2009), Riverside (Docherty et al., Rotational ambiguity can be explored by examining the
2008;Huffmanetal.,2009),MexicoCity(Aikenetal.,2008, appearance and disappearance of zero values in the mass
2009a, b; DeCarlo et al., 2009), Pittsburgh (Zhang et al., spectra and time series of the factors (Paatero, 2008). In
2005a, b; Ulbrich et al., 2009), Thompson Farm, NH (Cot- typical ambient datasets, a priori information about compo-
trell et al., 2008), Zurich (Lanz et al., 2007), Egbert (Slowik nent time points or fragment ions with true zero values is
et al., 2010), Crete (Hildebrandt et al., 2010), and the great not known; thus, the appearance of unrealistic zero values
Alpine region (Lanz et al., 2009). For the sites in Lanz et in the mass spectra and time series of the solutions can be
al. (2009), both PMF and ME-2 are used. In contrast to PMF, used to evaluate the most reasonable limits of the FPEAK
ME-2 allows for a priori constraints (partial or total) on the parameter (Ulbrich et al., 2009). For a few sites where com-
massspectra and/or time series of the factors (Paatero, 1999; ponent mass spectra or time series are highly correlated, ro-
Lanz et al., 2008). tational ambiguity is more significant. The change in mass
PMFisamultivariate factor analysis technique developed spectra with FPEAK is more dramatic in components with
by Paatero and Tapper (1994) and Paatero (1997) to solve a smaller mass fraction, as they can change more without
the bilinear factor model x =6 g f +e wherex are causing large changes in the residuals. For instance, in the
ij p ip pj ij ij Pittsburgh dataset (acquired in September 2002) studied by
the measured values of j species in i samples, P are factors Ulbrich et al. (2009), the variation in f and f in the
comprised of constant source profiles (f , mass spectra for 44 43
j SV-OOA component across the range of retained solutions
AMSdata) with varying contributions over the time period (FPEAK −1.6 to 1.0) was ∼30%, for the LV-OOA compo-
of the dataset (g , time series), without any a priori assump-
i nent ∼2%,andfortheHOAcomponent∼5%(relativetothe
tions of either mass spectral or time profile (Lanz et al., 2007; solution with FPEAK=0). In general, it is found that the
Ulbrich et al., 2009). PMF computes the solution by min- component mass spectra and time series for most sites ana-
imizing the summed least squares errors of the fit weighted lyzed in this work do not vary drastically over the reasonable
withtheerrorestimatesofeachdatapoint. Solutionsarealso rangeofFPEAKchosen. Forexample,therelativeuncertain-
constrained to have non-negative values. The error weight- ties in OOA component mass spectra for f and f (these
ing and non-negativity constraint result in more physically 44 43
meaningful solutions that are easier to interpret compared to fragments will be discussed in detail in the following sec-
other receptor models. tions) are typically <5%. For the Lanz et al. (2009) sites,
similar rotational uncertainties are observed except for sites
The PMF2 executable version 4.2 in the robust mode with low f in SV-OOA, where an absolute uncertainty of
(Paatero, 1997) is used together with a custom software tool 44
(PMF Evaluation Tool (PET), Ulbrich et al., 2009) in this ±5%isestimated.
www.atmos-chem-phys.net/10/4625/2010/ Atmos. Chem. Phys., 10, 4625–4641, 2010
4628 N. L. Ng et al.: Organic aerosol components in Northern Hemispheric datasets
1 Fig. 1. 28 Fig. 2.
44 O:C atomic ratio
l0.12 43 HOA l0.12 OOA
a a 14
n 55 n
0.10 g0.10
g i 43 0.2 0.4 0.6 0.8 1.0 1.2
i s
s 57
41 l
l a0.08 12
a0.08 t
t o s
o t
t
f0.06 e
f0.06 o t
o i 10
n s
n o
o0.04 i0.04 HOA
i t f
t c o
c a 8 OOA
a r
r0.02 f0.02 r
f e
0.00 0.00 b 6
0 20 40m/z60 80 100 0 20 40 m/z 60 80 100 m
u 4
N
44 80x10-3 2
l0.12 LV-OOA l 43 44 SV-OOA
a a 0
n0.10 n
g g 60
i 43 i
s s
l l 0.00 0.05 0.10 0.15 0.20 0.25 0.30
a0.08 a
t t
o o
t t
40
f0.06 f f
o o 44
n n 20
o0.04 o
i i
t t 20
c c LV-OOA
a a
r0.02 r
f f s SV-OOA
0.00 0 e
t
i 15
0 20 40 60 80 100 0 20 40m/z 60 80 100 s
m/z
2 f
o
3
r
4 e 10
5 Fig. 1. Example mass spectra of the HOA, total OOA, LV-OOAand b
m
6 SV-OOAcomponentsidentifiedfromthePittsburghdataset (Zhang u
7 et al., 2007; Ulbrich et al., 2009). Note that the total OOA spectrum N 5
8 is not the average of the LV-OOA and SV-OOA because LV-OOA
9 accounts for a much larger fraction (59%) of OA than SV-OOA 0
10 (10%) in Pittsburgh. 0.00 0.05 0.10 0.15 0.20 0.25 0.30
11
12 29 f44
13 30
14 3 Results and discussion 31
15 Fig. 2. Histograms showing the distribution of f and estimated
16 32 44
33
17 O:Catomicratios observed across the multiple sites. The top panel
3.1 Overviewoforganicaerosol components in the 34
18 corresponds to sites where only one OOA component is obtained.
Northern Hemisphere 35
19 ThebottompanelcorrespondstositeswherebothLV-OOAandSV-
36
20 37
OOAareresolved.
21 Figure 1 shows example mass spectra of the HOA, total38
22 39
23 OOA, LV-OOA, and SV-OOA components identified from40
24 41
25 the Pittsburgh dataset (Zhang et al., 2005b; Ulbrich et al., end of the SV-OOA range and the low end of the LV-OOA
26 2009). The HOA component is distinguished by the clear range occurs because the names SV-OOA and LV-OOA are
2
27 hydrocarbonsignaturesinits spectrum, which are dominated relative for each site. Since an absolute scale to define volatil-
+ + ity is not available, it is likely that the volatilities of the SV-
by the ion series C H and C H (m/z 27, 29, 41,
n 2n+1 n 2n−1
43, 55, 57, 69, 71, 83, 85, 97, 99...) that are typical of hy-1 OOAandLV-OOAcomponentsintheoverlappingregionare
drocarbons. The (total) OOA component is distinguished by not very different from each other.
+
the prominent m/z 44 (CO ) in its spectrum and the lower The different f and O:C values observed for the OOA
2 44
relative intensity of higher mass fragments. Figure 2 shows components in Fig. 2 reflect the fact that there is a contin-
the distribution of f for the HOA, total OOA, LV-OOA, uumofOOApropertiesinambientaerosol. At each site this
44
andSV-OOAcomponentsobservedacrossthemultiplesites. continuum is discretized into SV-OOA and LV-OOA com-
The top axis shows O:C ratios estimated using the f of ponents according to the details of the ambient OOA that is
44
the PMF-resolved factor mass spectra and the correlation de- observed at that particular site. Figure 3 explicitly shows the
rived by Aiken et al. (2008). As seen in Fig. 2, both average O:C atomic ratios and f of the OOA, LV-OOA, and SV-
44
f and O:C for HOA components are generally very low. OOAcomponentsateachsite. SimilartoZhangetal.(2007),
44
The OOA components, on the other hand, have higher f the sites have been grouped according to their location as ei-
44
and O:C values of 0.14±0.04 and 0.62±0.15. The f and ther being primarily urban or urban downwind/rural/remote.
44
O:CratiosforalltheLV-OOAandSV-OOAfallintotwodis- For some sites only one OOA factor is obtained while for
tinctive groups. The average f and O:C ratio for SV-OOA others the range of oxidation is represented by both LV-
44
components are 0.07±0.04 and 0.35±0.14, while those for OOAand SV-OOA components. For sites where both LV-
the LV-OOA components are 0.17±0.04 and 0.73±0.14. It OOAandSV-OOAareresolved, the average OOA O:C ob-
is important to note that a wide range of f and O:C is ob- served at any given time point can be reconstructed as a
44
served around the average values for both the LV-OOA and mass-weightedaverageoftheLV-OOAandSV-OOAO:C.In
SV-OOA components across all sites. This underscores the Fig. 3, the mass-weighted average OOA component over the
fact that neither the total OOA nor OOA subtypes are identi- entire campaignisalsoshownforsitesinwhichLV-OOAand
cal across the different sites. Some overlap between the high SV-OOA are both resolved. The sites within each location
Atmos. Chem. Phys., 10, 4625–4641, 2010 www.atmos-chem-phys.net/10/4625/2010/
no reviews yet
Please Login to review.
