Discussion paper

46 Pages·2016·0.83 MB·English

Atmos.Meas.Tech.Discuss.,doi:10.5194/amt-2015-332,2016 D
ManuscriptunderreviewforjournalAtmos.Meas.Tech. isc AMTD
u
Published:14January2016 s
s
©Author(s)2016.CC-BY3.0License. io doi:10.5194/amt-2015-332
n
P
a
p
Thisdiscussionpaperis/hasbeenunderreviewforthejournalAtmosphericMeasurement e Sampling strategies
r
Techniques(AMT).PleaserefertothecorrespondingfinalpaperinAMTifavailable. and post-processing
|
methods for OA
Sampling strategies and post-processing D measurement
is
c
u
methods for increasing the time ss R.L.Modiniand
io
n S.Takahama
resolution of organic aerosol P
a
p
e
measurements requiring long sample r
TitlePage
|
collection times Abstract Introduction
D
is
c Conclusions References
u
s
R. L. Modini and S. Takahama s
io Tables Figures
n
ENAC/IIESwissFederalInstituteofTechnologyLausanne(EPFL),Lausanne,Switzerland P
a
p J I
e
Received:30October2015–Accepted:13December2015–Published:14January2016 r
J I
Correspondenceto:S.Takahama(satoshi.takahama@epfl.ch) |
D Back Close
PublishedbyCopernicusPublicationsonbehalfoftheEuropeanGeosciencesUnion.
is
c
u FullScreen/Esc
s
s
io
n
Printer-friendlyVersion
P
a
p
e InteractiveDiscussion
r
1 |
Abstract D
is
c AMTD
u
The composition and properties of atmospheric Organic Aerosols (OAs) change on s
s
timescales of minutes to hours. However, some important OA characterization tech- io doi:10.5194/amt-2015-332
n
niquestypicallyrequiregreaterthanafewhoursofsamplecollectiontime(e.g.Fourier P
a
5 Transform Infrared (FTIR) spectroscopy). In this study we have performed numerical pe Sampling strategies
r
modelingtoinvestigateandcomparesamplecollectionstrategiesandpost-processing and post-processing
methods for increasing the time resolution of OA measurements requiring long sam- | methods for OA
ple collection times. Specifically, we modeled the measurement of Hydrocarbon-like D measurement
is
OA (HOA) and Oxygenated OA (OOA) concentrations at a polluted urban site in Mex- c
u
10 ico City, and investigated how to construct hourly-resolved time series from samples ssio R.L.Modiniand
collectedfor4,6,and8h.Wemodeledtwosamplingstrategies–sequentialandstag- n S.Takahama
P
gered sampling – and a range of post-processing methods including interpolation and a
p
deconvolution. The results indicated that relative to the more sophisticated and costly e
r
staggered sampling methods, linear interpolation between sequential measurements TitlePage
|
is a surprisingly effective method for increasing time resolution. Additional error can
15 Abstract Introduction
D
be added to a time series constructed in this manner if a suboptimal sequential sam- is
pling schedule is chosen. Staggering measurements is one way to avoid this effect. cu Conclusions References
s
Thereislittletobegainedfromdeconvolvingstaggeredmeasurements,exceptatvery s
io Tables Figures
low values of random measurement error (<5%). Assuming 20% random measure- n
P
ment error, one can expect average recovery errors of 1.33–2.81µgm−3 when using a
20 p J I
e
4–8h long sequential and staggered samples to measure time series of concentration r
−3 J I
values ranging from 0.13–29.16µgm . For 4h samples, 19–47% of this total error
|
can be attributed to the process of increasing time resolution alone, depending on the
D Back Close
method used, meaning that measurement precision would only be improved by 0.30– is
c
0.75µgm−3 if samples could be collected over 1h instead of 4h. Devising a suitable u FullScreen/Esc
25 s
s
sampling strategy and post-processing method is a good approach for increasing the io
n
time resolution of measurements requiring long sample collection times. P Printer-friendlyVersion
a
p
e InteractiveDiscussion
r
2 |
1 Introduction D
is
c AMTD
u
Organic Aerosols (OAs) comprise 20–90% of total, dry, sub-micrometer atmospheric s
s
aerosol mass, and therefore have important influences on air quality and aerosol- io doi:10.5194/amt-2015-332
n
climate effects (Jimenez et al., 2009; Fuzzi et al., 2015). OAs can be emitted directly P
a
5 into the atmosphere (Primary Organic Aerosol, POA), or formed in the atmosphere pe Sampling strategies
r
from the oxidation products of precursor gases (Secondary Organic Aerosol, SOA). It and post-processing
is critical to distinguish between POA and SOA since they result from different (natu- | methods for OA
ral and anthropogenic) emission and transformation processes, and therefore require D measurement
is
separate control and regulation strategies. This separation is complicated by the fact c
u
10 thatOAsarecomplexmixturesofthousandsofdifferentindividualorganiccompounds. ssio R.L.Modiniand
AkeyfeatureofOAisthatitscompositionandpropertieschangeandevolvecontin- n S.Takahama
P
ually in time (Jimenez et al., 2009). These changes happen on timescales of minutes a
p
to hours. OA evolution occurs because organic compounds are subject to continual e
r
oxidation throughout their lifetime in the atmosphere, while also mixing with freshly TitlePage
|
emitted OA. Oxidation changes basic OA molecular properties such as size and de-
15 Abstract Introduction
D
gree and type of functionalization. These basic molecular properties determine OA is
volatility, solubility and hygroscopicity, which in turn determine OA concentrations and cu Conclusions References
s
the ability of OA to take up water. These effects combined are relevant for assessing s
io Tables Figures
aerosol impacts on health and climate. Observation of OA composition over time also n
P
permitssourceresolutionimportantforidentifyingmajorcontributorstotheOAburden a
20 p J I
e
in the atmosphere (Corrigan et al., 2013). To capture the evolution of OA composition r
J I
and properties in the atmosphere it is necessary to measure OA at high time resolu-
|
tion(Jimenezetal.,2009).Wedefinetimeresolutionhereasthenumberofmeasured
D Back Close
values per unit time. is
c
DuetotheircomplexityOAscannotbecompletelycharacterizedbyanysinglemea- u FullScreen/Esc
25 s
s
surement technique. The full OA picture can only be captured by combining a range io
of different measurement techniques. Depending on analytical detection limits, some nP Printer-friendlyVersion
a
techniques require long sample collection times (typically greater than a few hours) to p
e InteractiveDiscussion
r
3 |
collect enough aerosol mass for analysis; these samples are often analyzed off-line in D
is
a laboratory facility rather than in the field. Examples of analytical techniques requir- c AMTD
u
s
ing longer sample collection times at atmospherically-relevant aerosol concentrations s
io doi:10.5194/amt-2015-332
include: Fourier Transform Infrared (FTIR) spectroscopy (4–24h; Russell et al., 2011; n
P
5 Frossard et al., 2014; Corrigan et al., 2013); and Nuclear Magnetic Resonance (NMR) a
p
spectroscopy (8–48h; Finessi et al., 2012; Matta et al., 2003; Decesari et al., 2006). e Sampling strategies
r
In contrast, measurement integration times can be as short as minutes (aerosol mass and post-processing
|
spectrometry) to 1h (online GC-MS), and these are often associated with on-line (or methods for OA
D
in-situ) instruments. measurement
is
c
10 Measurements with longer collection times still provide molecular- and functional- u
s R.L.Modiniand
group-levelinformationthatarevaluableforOAcharacterization(Corriganetal.,2013). s
io
Therefore, to obtain diverse and detailed chemical information at high time resolution, n S.Takahama
P
new approaches are desired. One approach is to develop new instrumentation and a
p
e
hardware for rapid sample collection and analysis. For example, an online GC-MS in- r
TitlePage
strument has been developed (Williams et al., 2006). Additionally, aerosol can be con-
15 |
centratedinaparticleconcentratorpriortosampling,whichcandecreaseFTIRsample Abstract Introduction
D
collectiontimesfromafewhoursto1h(Mariaetal.,2002).However,duetothecosts, is
c Conclusions References
complexities,andpracticallimitationsinvolved(e.g.aerosolconcentratorsrequirevery u
s
s
largeflowratesandvirtualimpactorsaresensitivetooperatingconditions),instrument io Tables Figures
n
development is not always a viable approach to improving time resolution. As an alter-
20 P
a
native or complement to hardware design, it is possible to devise sampling strategies p J I
e
and post-processing methods for constructing higher time resolution measurements r
J I
from a set of low resolution samples. This is the approach that we investigate in this |
work. D Back Close
We performed numerical modeling to compare the effectiveness of sampling strate- is
25 c
u FullScreen/Esc
giesandpost-processingmethodsforachieving1htimeresolutionwithmeasurements s
s
requiring 4, 6 and 8h of sample collection time. We modeled two sampling strategies: io
n
Printer-friendlyVersion
sequentialsampling,wheresuccessivemeasurementsarecollectedoneafteranother, P
a
and staggered sampling, where each new measurement is regularly initiated before pe InteractiveDiscussion
r
4 |
termination of the previous measurement. The time resolution of a sequentially mea- D
is
sured time series can be controlled (and increased) by interpolating between mea- c AMTD
u
s
surements. The resolution of a time series obtained by staggered sampling can be s
io doi:10.5194/amt-2015-332
controlled through the choice of the staggering interval between samples. A time se- n
P
5 riesresultingfromstaggeredsamplingisarunningaverageofthetruetimeseriesone a
p
seeks to measure. In the ideal case, mathematical deconvolution can be used to re- e Sampling strategies
r
trievetheoriginaltimeseriesattheresolutionofthestaggeringratherthansamplecol- and post-processing
|
lectioninterval.Foractualmeasurements,theprocessofdeconvolutioniscomplicated methods for OA
D
by unavoidable perturbations to measurement signals due to random measurement measurement
is
c
10 errors. Regularization techniques are required. u
s R.L.Modiniand
We examined two concentration time series with contrasting diurnal patterns. s
io
Hydrocarbon-like organic aerosol (HOA) and oxygenated organic aerosol (OOA) are n S.Takahama
P
majorcontributorstoOAasidentifiedbyAMSandfactoranalyticdecomposition(Zhang a
p
e
et al., 2011). HOA is generally associated with primary organic aerosol (POA) emis- r
sions and follows diurnal trends of traffic patterns in urban areas (i.e., early morning TitlePage
15 |
and late afternoons during weekdays). OOA is associated with SOA formed from pho- Abstract Introduction
D
tochemical oxidation in combination with aged background aerosol (de Gouw et al., is
c Conclusions References
2009), and exhibits a peak close to solar noon. The data set we used are AMS mea- u
s
s
surements of HOA and OOA reported by Aiken et al. (2009) at a polluted urban site in io Tables Figures
n
Mexico City, Mexico (T0 site MILAGRO field campaign; Molina et al., 2010). The data
20 P
a
set is described fully in Sect. 2. p J I
e
Section 3 formerly introduces and describes the different sampling strategies and r
J I
post-processing methods we investigated. Section 4 describes the numerical model- |
ing used to apply these sampling strategies and post-processing methods to the test D Back Close
data. The modeled conditions were designed primarily to represent the measurement is
25 c
u FullScreen/Esc
of functional groups representing HOA and OOA by aerosol FTIR spectroscopy, since s
s
thisistheprimarymeasurementtechniqueofourresearchgroup.However,theresults io
n
Printer-friendlyVersion
should be applicable to any type of environmental sampling that can be characterized P
a
with parameters falling within the ranges that we modeled. pe InteractiveDiscussion
r
5 |
The numerical modeling results were grouped into three major categories: sequen- D
tial (sequential sampling + interpolation), smeared (staggered sampling with no data isc AMTD
u
processing), and recovered (staggered sampling + deconvolution). In Sects. 5 and 6 ss
io doi:10.5194/amt-2015-332
thebestpost-processingmethodsareidentifiedforthesequentialandrecoveredcate- n
P
5 gories, respectively. An overall comparison of the best-case sequential and recovered a
p
solutions with the smeared solution is made in Sect. 7. The advantages and disad- e Sampling strategies
r
vantages of each method are discussed, taking into account the attainability of the and post-processing
|
modeled best-case scenarios and the practical costs involved. Section 8 discusses methods for OA
the differences between the HOA and OOA results. Finally in Sect. 9 we discuss the D measurement
is
10 interpretation of the error results. cu
s R.L.Modiniand
s
io
n S.Takahama
2 Test case: HOA and OOA concentration time series P
a
p
e
To test different methods of increasing time resolution we used time series of HOA r
TitlePage
and OOA concentrations originally measured at high time resolution by aerosol mass |
Abstract Introduction
spectrometry at the T0 site in central Mexico City in 2006 during the MILAGRO field D
15 campaign.TheMILAGROcampaignandT0sitearedescribedbyMolinaetal.(2010). iscu Conclusions References
The aerosol mass spectrometer measurements and the Positive Matrix Factorization s
s
(PMF) analysis used to derive the HOA and OOA profiles and concentrations are de- io Tables Figures
n
scribed by Aiken et al. (2009). P
a
p J I
The HOA and OOA concentration time series are displayed in Fig. 1a. The origi- e
r
20 nal measurements were collected over the period from 10 March to 31 March 2006. J I
|
To avoid gaps in the time series greater than 1h we only used the measurements
from 23:00 19 March 2006 to 10:00LT 29 March 2006, which amounts to a total pe- D Back Close
is
riod of 228h. This period was chosen because 228 has many factors (7 greater than cu FullScreen/Esc
12), which was desirable for numerically modeling the effect of the time series period ss
io
measured (see Sect. 4). The original measurements were averaged over 1h intervals n
25 Printer-friendlyVersion
P
to generate hourly-resolution data for the inverse modeling and to smooth out some a
p
of the high-frequency perturbations due to random measurement uncertainties. The e InteractiveDiscussion
r
6 |
hourly-resolution data certainly still contain measurement noise, but for the purposes D
is
of our modeling we assume that these signals represent the true changes in HOA and c AMTD
u
s
OOA concentrations at the T0 site over this time period. s
io doi:10.5194/amt-2015-332
BoththeHOAandOOAconcentrationtimeseriesdisplayedstrongandregulardaily n
P
5 peaks. The diurnally averaged profiles shown in Fig. 1b indicate that HOA concentra- a
p
tionspeakedinthemorningsaround07:00LT.TheseHOApeakswerecoincidentwith e Sampling strategies
r
the occurrence of a morning vehicle rush hour period and low atmospheric boundary and post-processing
|
layerheights.ThispeaktimingsuggeststheHOAwaspredominantlyprimaryOAemit- methods for OA
D
ted from combustion sources that was able to build up to high concentrations in the measurement
is
c
10 shallow morning boundary layers (Aiken et al., 2009). The daily OOA concentration u
s R.L.Modiniand
peakswerebroader,beginningaround08:00andextendingto15:00LT.Thispeaktim- s
io
ing suggests that the OOA concentration peaks were the result of photochemistry and n S.Takahama
P
SOA formation (Aiken et al., 2009). a
p
e
ThetwotimeseriesinFig.1werechosenforthisanalysisbecausetheirdailypeaks r
TitlePage
wereseparatedbyonlyafewhours.IftheseHOAandOOAconcentrations(orthecon-
15 |
centrations of functional groups or specific molecules representing these OA classes) Abstract Introduction
D
weremeasuredatpoortimeresolution(>4h),thedifferencesbetweenthedailypeaks is
c Conclusions References
would not be clearly resolved. In that case it would not be possible to easily recog- u
s
s
nize that the concentration peaks resulted from two distinct processes: primary parti- io Tables Figures
n
cle emission and secondary aerosol formation. Therefore, the ability to clearly resolve
20 P
the daily HOA and OOA concentration peaks provided an ideal test case for different ap J I
e
methods of obtaining hourly time resolution data from measurements requiring longer r
J I
sample collection times. |
We note that it is not possible to measure HOA or OOA concentrations directly with D Back Close
FTIR spectroscopy. FTIR spectroscopy is used to measure the absorption spectra of is
25 c
u FullScreen/Esc
aerosolsamples.OrganicfunctionalgroupandtotalOAconcentrationscanbederived s
s
from these measured spectra (Russell et al., 2009; Takahama et al., 2013). The ideal io
n
Printer-friendlyVersion
conditions we have modeled in this study could represent, for example, the measure- P
a
ment of organic functional groups that represent HOA and OOA. Factor analysis can pe InteractiveDiscussion
r
7 |
alsobeusedtocalculatetheFTIR-equivalentofHOAandOOAspecies(Corriganetal., D
is
2013). In this case the relevant time series would be multivariate (many wavelengths c AMTD
u
s
or functional group abundances considered together) rather than univariate (concen- s
io doi:10.5194/amt-2015-332
trations of individual species). The theory developed in Sect. 3 can be extended to n
P
5 the multivariate case. The multivariate extension is the topic of future work and is not a
p
covered in the present study. For the current, univariate case we chose to model the e Sampling strategies
r
measurement of HOA and OOA concentrations because these species display con- and post-processing
|
trasting diurnal profiles and because they illustrate the variations in OA that can be methods for OA
D
captured at high time resolution. measurement
is
c
u
s R.L.Modiniand
s
10 3 Sampling strategies and post-processing methods for increasing ion S.Takahama
measurement time resolution P
a
p
e
r
Two simulated sampling strategies were applied to the HOA and OOA test data: se-
TitlePage
quential and staggered sampling. A variety of different post-processing methods for |
Abstract Introduction
increasing measurement time resolution were investigated with the two sets of simu- D
is
15 lated measurements. Figure 2 lists each of the methods applied and each method is cu Conclusions References
explainedinfurtherdetailbelow.Foreachmethod,thebest-casescenariowasconsid- s
s
ered in order to determine the theoretically optimal combination of sampling strategy ion Tables Figures
and data processing method for increasing measurement time resolution. P
a
p J I
e
r
3.1 Sequential sampling
J I
|
20 Aerosol samples (and most other environmental samples) are typically collected se- D Back Close
quentially, one after another. We refer to this as sequential sampling. Sequential mea- is
c
u FullScreen/Esc
surements are separated by an interval of time (δτ) equal to the individual sample s
s
collection or measurement integration time (∆τ). Post-measurement, the resolution of io
n
Printer-friendlyVersion
sequentially collected measurements can be increased by interpolating between suc- P
a
25 cessive points with some chosen function. Here we consider two interpolation meth- pe InteractiveDiscussion
r
8 |
ods: step function interpolation (Fig. 3a), and linear interpolation (Fig. 3b). Although D
is
it seems likely that linear interpolation will better represent the original time series c AMTD
u
s
we have tested step interpolation as this case is often assumed (at least implicitly). s
io doi:10.5194/amt-2015-332
For both interpolation cases we represented a single measurement by the midpoint of n
5 agivensample:eachmeasurementoccursattimetmid=tstart+∆τ/2=tend−∆τ/2).It Pap
is also possible to represent individual measurements by the start (t ) or endpoints e Sampling strategies
start r
(t ) of each sample. We do not consider those options here because the modeled and post-processing
end |
results do not represent the original time series as well as the simulations with t . methods for OA
mid
D
measurement
is
3.2 Staggered sampling c
u
s R.L.Modiniand
s
10 Aerosol sample collection can also be staggered, such that each new sample is regu- ion S.Takahama
larlyinitiatedbeforeterminationoftheprevioussample.Byseparatingsuccessivemea- P
a
surements by a staggering interval δτ less than the individual sample collection time pe
r
∆τ, it is possible to increase measurement time resolution. The principle of combin- TitlePage
|
ingmultiple,overlapping,lower-resolutionsamplesinordertoconstructhigherspatial-
Abstract Introduction
and temporal-resolution information has been used extensively for image processing D
15 is
(Borman and Stevenson, 1998; Shechtman et al., 2005). c Conclusions References
u
Staggeredsamplingeffectivelyappliesarunningaveragetoatimeseriesofaerosol ss
io Tables Figures
concentrations,whichproducesasmearedversionoftheoriginalsignal,denotedhere n
P
asg(t).Iff(t)representsthetruechangeinaerosolconcentrationsatsomepointinthe a
p J I
atmospherefromtimet=0toT,g(t)istheproductoftheconvolutionofaboxcarkernel e
20 r
function h(∆τ) and f(t). This is a specific example of a Fredholm integral equation of J I
|
the first kind:
D Back Close
T is
Z c
g(t)= h(∆τ)f(t)dt. (1) us FullScreen/Esc
s
io
0 n Printer-friendlyVersion
P
In the case of measured data a smeared signal is more appropriately represented ap
e InteractiveDiscussion
by a finite series of n measurement points g separated by δτ than by the continuous r
25
9 |
function g(t). In addition, all measurements are subject to some amount of measure- D
is
ment uncertainty (cid:15). A discrete formulation of Eq. (1) that more accurately reflects the c AMTD
u
s
actual measurement process is the matrix equation: s
io doi:10.5194/amt-2015-332
n
g=Hf+(cid:15), (2) P
a
p
e Sampling strategies
where H is a convolution matrix and f is a finite series of m data points representing r
5 and post-processing
f(t). The temporal resolution of f is δτ, the temporal resolution of g. For staggered |
methods for OA
samples, the convolution matrix H is an n-by-m toeplitz matrix. Each of the n rows of D
measurement
H contains a shifted copy of a boxcar function with k =∆τ/δτ non-zero values equal isc
u
to 1/k. In general, n=m+k−1. Figure 5 displays examples of a true time series f of s R.L.Modiniand
s
10 HOAconcentrationsandcorrespondingsmearedtimeserieswithout(Fig.5a)andwith ion S.Takahama
(Fig. 5c) measurement error. P
a
Equation (2) suggests two post-processing methods for recovering a higher time pe
resolution estimate fˆof the true time series f from staggered measurements: r TitlePage
|
1. The measured time series is taken as an approximation of the true time series. D Abstract Introduction
No further data processing is applied. is
15 c Conclusions References
u
2. One attempts to recover fˆthrough a deconvolution operation. For example, if H+ ssio Tables Figures
is the pseudo-inverse matrix of H one can solve the following inverse problem n
P
a
fˆ=H+g. (3) pe J I
r
J I
|
In principle, the true aerosol concentrations f can be recovered precisely from a set
D Back Close
20 of staggered measurements g and solution of Eq. (3) (Fig. 5b). However, in practice is
c
the problem is ill-posed. The small perturbations (cid:15) to g due to random measurement u FullScreen/Esc
s
uncertainty are strongly amplified in fˆ. One can only ever hope to find a solution fˆthat sio
n
is a good approximation of f (Fig. 5d and e). Printer-friendlyVersion
P
A variety of different deconvolution methods exist for finding the inverse solution of ap
e InteractiveDiscussion
Eq. (2). For example, the convolution theorem (Arfken and Weber, 2005) states that r
25
10 |