Elsevier Editorial System(tm) for Atmospheric Environment Manuscript Draft Manuscript Number: Title: VOLATILE ORGANIC COMPOUNDS SOURCES IN PARIS IN SPRING 2007. Part I: QUALITATIVE ANALYSIS Article Type: Research Paper Keywords: VOC; Ile de France; emission; diurnal variation; oxygenated compounds Corresponding Author: Dr. Valerie Gros, Corresponding Author's Institution: CNRS First Author: Valerie Gros Order of Authors: Valerie Gros; Cecile Gaimoz; Frank Herrmann; Tom Custer; Jonathan Williams; Bernard Bonsang; Stephane Sauvage; Nadine Locoge; Odile d'Argouges; Roland Sarda-Esteve; Jean Sciare Abstract: Fast measurements of volatile organic compounds (VOC) have been performed in Paris city centre in spring 2007. The measurements were influenced by two main air mass origins, 1) from the Atlantic Ocean and 2) from north-eastern Europe. While non-methane hydrocarbons (NMHC) and CO baseline levels were only slightly impacted by changes in the air mass origin, oxygenated compounds such as acetone and methanol showed much higher baseline levels during continentally influenced air masses. This suggests that NMHC and CO mixing ratios were mainly influenced by local-to-regional scale sources whereas oxygenated compounds had a more significant continental scale contribution. Similar behaviour was observed in the chemical composition of the aerosols (aerodynamic diameter < 2.5 µm), with carbonaceous aerosol variations indicating predominantly local sources and inorganic aerosols (sulphate and nitrate aerosols) broader continental sources. This highlights the importance of measuring VOCs other than NMHC in source classification studies. The period of Atlantic air influence was used to characterize local pollution which was dominated by traffic related emissions, although traffic represents only one third of total VOCs emissions in the local inventory. In addition to traffic related sources (from exhaust and evaporation), additional sources were identified, in particular emissions from dry cleaning activities were identified by the use of a specific tracer (i.e. tetrachloroethylene).
Cover Letter
Dr. Valérie GROS Tel : + 33 1 69 08 79 67 Fax : + 33 69 08 77 16 E-mail :
[email protected]
Paris, 04/12/09
Dear Editor
Plese find attached the manuscript “Volatile organic compounds sources in Paris in spring 2007, Part I: qualitative analysis” by Gros et al. for submission in Atmospheric Environment. Please note that a companion paper “Volatile organic compounds sources in Paris in spring 2007, Part II: source apportionment using positive matrix factorization” by Gaimoz et al. is submitted at the same time. As these papers report original VOCs data for the city in Paris along with a carfeul examination of VOCs sources, we think they may desserve publication in your journal.
Best regards
Dr. Valérie Gros,
Unité Mixte de Recherche CEA-CNRS LSCE-Orme - Bât. 709 - Orme des Merisiers - 91191 Gif-sur-Yvette Cedex LSCE-Vallée - Bât. 12 - avenue de la Terrasse - 91198 Gif-sur-Yvette Cedex
Tél. : 01 69 08 77 11 - Fax : 01 69 08 77 16 Tél. : 01 69 82 35 23 - Fax : 01 69 82 35 68
*Manuscript Click here to download Manuscript: Gros_manuscript.doc
1
VOLATILE ORGANIC COMPOUNDS SOURCES IN PARIS IN SPRING 2007.
2
Part I: QUALITATIVE ANALYSIS
3 4
Valérie Gros (1)*, Cécile Gaimoz (1), Frank Herrmann (2), Tom Custer (2) Jonathan
5
Williams (2), Bernard Bonsang (1), Stéphane Sauvage (3,4), Nadine Locoge (3,4),
6
Odile d’Argouges (1), Roland Sarda-Estève (1), and Jean Sciare (1).
7 8
(1) LSCE, Laboratoire des Sciences du Climat et de l’Environnement, unité mixte
9
CEA-CNRS-UVSQ, Gif-sur-Yvette, France
10
(2) Max Planck Institute for Chemistry, Mainz, Germany
11
(3) Université de Lille Nord de France, F-59000 Lille, France
12
(4) Ecole des Mines Douai, Département Chimie environnement, F-59508 Douai,
13
France
14 15
*corresponding author
16
[email protected],
17
LSCE, Orme des Merisiers, Bat 701
18
91191 Gif sur Yvette cedex, France
19
Tel + 33 1 69 08 79 67
20
Fax + 33 1 69 08 77 16
21 22 23 24 25 26 1
1 2
Abstract
3
Fast measurements of volatile organic compounds (VOC) have been performed in
4
Paris city centre in spring 2007. The measurements were influenced by two main air
5
mass origins, 1) from the Atlantic Ocean and 2) from north-eastern Europe. While
6
non-methane hydrocarbons (NMHC) and CO baseline levels were only slightly
7
impacted by changes in the air mass origin, oxygenated compounds such as acetone
8
and methanol showed much higher baseline levels during continentally influenced air
9
masses. This suggests that NMHC and CO mixing ratios were mainly influenced by
10
local-to-regional scale sources whereas oxygenated compounds had a more
11
significant continental scale contribution. Similar behaviour was observed in the
12
chemical composition of the aerosols (aerodynamic diameter < 2.5 µm), with
13
carbonaceous aerosol variations indicating predominantly local sources and
14
inorganic aerosols (sulphate and nitrate aerosols) broader continental sources. This
15
highlights the importance of measuring VOCs other than NMHC in source
16
classification studies. The period of Atlantic air influence was used to characterize
17
local pollution which was dominated by traffic related emissions, although traffic
18
represents only one third of total VOCs emissions in the local inventory. In addition to
19
traffic related sources (from exhaust and evaporation), additional sources were
20
identified, in particular emissions from dry cleaning activities were identified by the
21
use of a specific tracer (i.e. tetrachloroethylene).
22 23
Keywords : VOC, Ile de France, emission, diurnal variation, oxygenated compounds
24 25 26
I Introduction 2
1 2
Measurements of gaseous and particulate pollutants in cities have been
3
historically associated with (local) air quality issues. However, more recently it has
4
been recognised that the large amounts of pollutants emitted in urban areas also
5
have a chemical and radiative impact on a larger scale, through the transport and
6
chemical processing of the pollution plume. This is especially relevant for megacities
7
which emit huge quantities of gaseous and particulate pollutants from a rather small
8
area (Gurjar and Lelieveld, 2005). The city of Paris and its surrounding region
9
(named “Ile de France”) with about 12 Millions inhabitants constitutes one of the few
10
megacities in Europe and concentrates 20% of the whole French population. Its
11
geographical location -a relatively small area basin surrounded by rural areas- makes
12
it a favourable place to study local pollution when air masses come from the clean
13
marine western sector, and to characterize the impact of the European contribution
14
when air masses come from the eastern sector. In order to determine the chemical
15
impact of a megacity, a first but important step is the characterization of pollutant
16
emissions and variability at the source point (the city itself). Among the pollutants
17
playing an important role in urban areas, volatile organic compounds (VOCs) are key
18
constituents, because their oxidation leads to formation of ozone and, for some
19
VOCs, of secondary organic aerosols (SOA). Although it was found that the chemical
20
regime over the urban area of Paris and within plumes is clearly VOC sensitive on
21
the average over two summers (Deguillaume et al., 2008) there are surprisingly very
22
few VOCs measurements reported for Paris and its region. Most of the atmospheric
23
VOCs measurements in Paris and its region have so far been dedicated to exposure
24
studies (Pernelet-Joly et al., 2009) and (Vardoulakis et al., 2002). Traffic is known to
25
be an important source of VOC in Paris (Gros et al., 2007; Vardoulakis et al., 2005)
26
but the local emission inventory from AIRPARIF, the local air quality network, 3
1
suggests that solvent (from industries and from residential sectors) is the dominant
2
source of VOC (AIRPARIF, 2009). Uncertainty in the spatial and temporal source
3
types suggests the need for pioneer groundtruthing experiments in order to better
4
constrain the sources of VOC in Paris and its surroundings. We note that in this
5
paper the term “local” will be used for Paris and Ile de France region (about 100 km
6
wide). Results from a campaign organized in spring 2007 (May 25-June 13) which
7
included time resolved measurements of VOCs, CO, O3, NOx and chemical
8
composition of the aerosols are presented here. The first part of the paper is
9
dedicated to the qualitative analysis of VOC sources and variability including a study
10
of the dependence on air mass origin, a first identification of VOCs sources and an
11
examination of diel variation of compounds. A second accompanying paper (Gaimoz
12
et al., 2009) proposes a source apportionment study based on results from Positive
13
Matrix Factorization (PMF) simulations. For consistency of the two papers, we use as
14
main unit μg m-3 as it is the unit requested by PMF simulations. However, for
15
comparison purpose with literature data, the unit ppb (nmol mol-1) is used as well in
16
this paper.
17 18
II Experimental
19 20 21
II.1 Measurement site The measurement campaign took place in spring 2007 (May 25- June 13) in
22
Paris, on the terraced roof (~15 m a.g.l.) of the « Laboratoire d’Hygiène de la Ville de
23
Paris » (LHVP), located in the southern part of Paris inner city. This station, which
24
belongs to the air quality network AIRPARIF, has been classified as representative
25
for “urban background” and has hosted in the past several experiments (see (Favez
26
et al., 2009) and references therein). According to the AIRPARIF criteria, “urban” 4
1
means that the population density is at least 4000 inhabitants per km2 within a 1 km
2
radius of the station, and “background” implies that no major traffic road is located
3
close by (within 300m). This station is a permanent monitoring station of AIRPARIF
4
for the measurement of ozone and nitrogen oxides (NOx). During the intensive
5
campaign, additional measurements were installed at the station and these are
6
described below.
7 8 9
II.2 Non-methane hydrocarbons on-line measurements by gas chromatographs Non-methane hydrocarbons (NMHC) in ambient air were measured with two
10
portable gas chromatographs equipped with Flame Ionization Detector (GC-FID,
11
Chromatotec, France). The instruments were installed in a small room located
12
directly on the roof terrace (about 14m high). The first analyser, ChromaTrap,
13
allowed the measurement of compounds constituted of two to six carbon atoms (C2-
14
C6 analyser) and the second, AirmoBTX, the measurement of compounds constituted
15
of six to ten carbon atoms (C6-C10 analyser). For the ChromaTrap instrument, for
16
each sample, 180 mL of air was drawn through a 1/8” diameter 6m long stainless-
17
steel line with a flow rate of 18 mL/min. First, ambient air was passed through a
18
Nafion Dryer to reduce the water content and then hydrocarbons were
19
preconcentrated at -8°C, on a 2.25 mm internal diameter (i.d.), 8 cm length glass trap
20
containing the following adsorbents : Carboxen 1000 (50 mg), Carbopack B (10 mg)
21
and Carbotrap C (10 mg). Then the trap was heated rapidly to 220°C for three
22
minutes, so that the preconcentrated VOCs were thermally desorbed onto a Plot
23
Column (Al2O3 / Na2SO4, 25 m * 0.53 mm). During the two first minutes of the
24
analytical procedure, the oven temperature was raised from 38 to 40°C, and
25
thereafter with a constant heating rate of 20°C/min, so that the temperature reached
5
1
199°C by the end of the analysis. We refer to (Bonsang et al., 2008 and references
2
therein) for more details about the performance of this instrument.
3
For each sample taken by the AirmoBTX instrument, 660 mL of air were drawn
4
through through a 1/8” diameter 6m long stainless-steel line with a flow rate of 60
5
mL/min. The hydrocarbons were preconcentrated at ambient temperature on a glass
6
trap containing the adsorbent Carbotrap C. Then the trap was heated to 380°C over
7
two minutes, to desorb the preconcentrated VOCs into a separating column
8
(MXT30CE, 30 m * 0.28 mm). For the first two minutes of the analysis, the oven
9
temperature was raised from 38 to 40°C, then with a constant heating rate of 2°C/min
10
to 50°C, then with a constant heating rate of 10°C/min to 80°C, and finally with a
11
constant heating rate of 15°C/min until the temperature reached 199°C by the end of
12
the analysis. Due to technical problems, the instrument was working only during the
13
second part of the campaign (June, 2-14) and will be used in this study only to verify
14
the quality of the aromatic compound measurements made by Proton Transfer Mass
15
Spectrometer (see next section).
16
For both GC instruments, the sampling time was ten minutes, analysis time was
17
20 min and therefore measurements were performed with a time of resolution of 30
18
min. Data were then aggregated into an hourly mean (average of two samples).
19
During the campaign, a small (2 ml) of a calibrated gas mixture was injected once
20
every day. This standard gas contained 1 ppmv levels of 56 C2-C12 VOCs (Restek,
21
Spectra Gases) including the analyzed hydrocarbons. This analysis allowed the
22
confirmation of compound retention times and the calculation of a daily response
23
factor (one average response factor per instrument). For the calibration of the
24
campaign measurements, an average response factor was calculated as the mean of
25
all daily response factors.
6
1
Tests performed in the laboratory have shown a repeatability of the measurement
2
better than 5% for all compounds (except for ethane and ethene, for which a
3
repeatability of 6.4% and 5. 1% was found respectively). Taking into account
4
additional uncertainties associated with the standard precision (+/- 5%) and with a
5
small memory effect due to the trapping phase (affecting mostly ethane and tri-
6
methyl benzene) the overall uncertainty is estimated as better than 15%.
7 8 9
II.3 VOC measurements by PTR-MS
10
The air inlet for the Proton Transfer Mass Spectrometer (PTR-MS) was located on
11
the roof, close to the inlet used by the GC-FID. Air was sampled through a 50 m long,
12
3/8” o.d. (1/4” i.d.) sheathed Teflon line that ran from the roof of the building to the
13
instrument. A Teflon filter (Pore size diameter: 5 m) was installed at the head of the
14
inlet to prevent from large aerosol particles, insects and other miscellaneous debris
15
entering into the sample lines. Air flow through the main sampling line was 16 L/min
16
and was maintained by a membrane pump. The resulting residence time of air in the
17
line was 0.5 min. The time resolution for measurements was approximately two
18
minutes (the time required for measurement of 55 different ions at 2 seconds per
19
ion).
20
The PTR-MS instrument, (Ionicon Analytik, Austria), has been described in detail
21
elsewhere (Lindinger et al., 1998) Briefly, a stable flow of air and high concentrations
22
of H3O+ ions are continuously sampled into a drift tube held at 2.2 millibar pressure.
23
Here, compounds with a proton affinity greater than that of water, including a large
24
selection of OVOCs, undergo efficient proton-transfer reactions with the H3O+ ions to
25
produce protonated organic product ions which can be detected by a mass
26
spectrometer. Here the PTR-MS drift tube was operated at 2.2 mbar and 50°C with a 7
1
drift field of 600 V cm-1. Sample air flow into the drift tube itself was constant
2
(approximately 15 mL min-1) with flow through a short length of tubing between the
3
main 50 m sampling line and the PTR-MS maintained at a flow of 300 mL min-1.
4
Instrument response was determined through measurement of a standard gas
5
diluted to mixing ratios in the range~0.1 ppbv and 8 ppbv and containing chemicals
6
including methanol, acetaldehyde, acetone, acetonitrile, methyl vinyl ketone,
7
benzene, toluene, o-xylene and 1,3,5 trimethyl benzene. The calibration factors (the
8
slope of the mixing ratio with respect to product ion signal normalized to m/z 21*500,
9
pressure, and temperature) were applied to the measurements (see Figure 1). For
10
those chemicals not found in the calibration gas, mixing ratios were calculated based
11
on ion/molecule reaction kinetics using the measured reaction conditions of the flow-
12
drift tube of the PTRMS (pressure, temperature, drift field, drift tube length) and
13
reasonable values for reaction rate coefficients and ion transmission as a function of
14
m/z as generally described in (Lindinger et al., 1998).
15
Instrumental background signal was determined through periodic sampling of air
16
scrubbed through a catalytic converter (platinum coated wool heated to 350° C) to
17
remove residual VOCs. Residual signals observed at all m/z during sampling of
18
scrubbed air are assumed to be purely instrumental in origin (e.g. we are detecting
19
chemicals outgassing from tubing/surfaces within the PTRMS and electronic noise).
20
Any data below one standard deviation of the background signal have been
21
removed. Background values were averaged, linearly interpolated onto all other
22
measurements, and subtracted from them prior to final conversion to mixing ratios.
23
Reproducibility of these measurements was assessed using a collation of
24
calibration data taken over a period of 2 years. The relative standard deviation of
25
these measurements was then plotted versus the mixing ratio itself (Figure 2). This
26
data was then fitted to an exponential function of the form y=A+Be^(x-x0)/C where A, 8
1
B, C, and x0 are fit parameters, x is the mixing ratio in ppbv, and y is a measure of
2
the “reproducibility” expressed as a percent of the signal. Combining this value with
3
the stated accuracy of the chemical mixing ratios in the standard gas (generally 5%),
4
the root sum of squares of the reproducibility and this accuracy was then calculated
5
and applied to all data to produce a combined uncertainty for each observed mixing
6
ratio data point.
7
With PTRMS measurements, unambiguous identification of chemical species is
8
not possible. For instance, isomers such as acetone and propanal or acetic acid and
9
isopropanol have the same nominal m/z using the quadrupole mass spectrometer
10
employed here. Other forms of chemical cross talk (characterized by situations where
11
a single ion may have multiple precursors with different time evolutions) might arise
12
from production of ions that fragment or cluster with water molecules (in high
13
humidity situations). Keeping in mind this limitation, it is also important to realize that
14
many chemical assignments have been confirmed over years of use in the field,
15
especially as it concerns ambient atmospheric sampling, where secondary
16
measurement techniques have been on hand for confirmation and can be considered
17
reliable (see(Blake et al., 2009) and references therein).
18 19 20
II.4 Off-line VOC measurements by cartridge sampling and GC-MS analysis During the campaign, cartridges filled with 250 mg of Tenax TA and previously
21
conditioned for 8 hours at 300°C under nitrogen (quality 6.0) flow were sampled
22
automatically with a Smart Automatic Sampling System (Tera-Environnement).
23
Different sampling times were selected (depending on the period and on the time of
24
the day) and ranging from 2h to 4h. Cartridges were then stored in the fridge (+4°C)
25
and analyzed at the laboratory within two months following sampling. Cartridges were
26
analyzed by using an Automated Thermo-Desorber (Perkin-Elmer) coupled with a 9
1
gas chromatograph mass spectrometry system (GC-MS CP3800-Saturn 200, Varian)
2
equipped with a capillary VF-5ms column (30m, 0,25 mm i.d.) and an ion trap.
3
Helium was used as carrier gas at a flowrate of 1 mL/min and the GC oven program
4
was set as follows: 50°C for 5 minutes, then 5°C/min until 110°C, and finally
5
15°C/min until the final temperature of 250°C. The parameters of the analytical
6
system have been optimized in order to measure cresols and therefore analyses
7
have been performed in the Single Ion Storage mode. Detection limits lower than 1
8
ppt were reached for the determination of cresols. In addition, toluene (a cresol
9
precursor) and selected chlorinated compounds (tetrachloroethene) were measured.
10
Calibrations were performed once or twice a day by analyzing cartridges filled with a
11
known volume of a calibrated mixture.
12 13 14
II.5 VOC intercomparison Benzene and toluene were two compounds measured by both in-situ instruments
15
(GC-FID and PTR-MS) and were therefore chosen to cross-check the quality of the
16
results. Figures 3a and 3b show the correlation of benzene and toluene hourly mean
17
measured by PTR-MS and GC-FID during the period June 2-14 (the only period of
18
the campaign when the GC-FID was running). High correlation coefficients (R2 of
19
0.84 for benzene and of 0.95 for toluene, N= number of points=185) with slopes
20
close to 1 are observed for both compounds. Almost no offset is observed for
21
toluene, whereas a small offset of 0.25 μg m-3 (~ 77 ppt) is observed for benzene.
22
This offset of 77 ppt (for an average of 350 ppt during the campaign) has to be kept
23
in mind when comparing absolute values of benzene with other field campaigns
24
results. However, this will not play a role on the analysis of variability which is made
25
later in this paper.
26 10
1
II.6 Additional compounds measurements
2
Carbon monoxide measurements were performed every minute with a 48i-TL
3
instrument (Thermo-Environment, USA) and used the same main inlet line as the
4
PTR-MS, with a short length (
