Introduction
The contribution of anthropogenic aerosol particles is thought to be of the
order of 10 Tg C yr-1; however, that of natural biogenic aerosols has
been estimated to be as much as 90 Tg C yr-1, having an important
effect on climate in both populated and remote areas of the world (IPCC,
2007; Hallquist et al., 2009). Our knowledge of how primary emissions from
anthropogenic and natural sources contribute to the formation of secondary
aerosols and their evolution in the atmosphere continues to improve with
considerable advances in numerical simulations. However, discrepancies
between simulations and measurements still exist and are more apparent over
remote and forested environments than over anthropogenic environments
(Ganzeveld et al., 2008; Lelieveld et al., 2008). The most commonly emitted
biogenic volatile organic compounds (BVOCs) include isoprene and
monoterpenes, with isoprene emissions accounting for approximately 44 %
(Kesselmeier and Staudt, 1999; Arneth et al., 2008). These species can be
difficult to characterize because of their high temporal and spatial
variability. Studies have shown that the formation yields of secondary organic aerosol (SOA) from
biogenic emissions alone are relatively low compared to those from
anthropogenic sources, but when emissions from both biogenic and
anthropogenic sources are combined, the resulting yield for SOA formation is
much higher than either anthropogenic or biogenic emissions alone (Day et
al., 2009; Bryan et al., 2012; Shilling et al., 2013; Hu et al., 2015).
The increasing improvement of instrumentation (namely aerosol mass
spectrometry) available for the detection of different biogenic species has
led to an increase in the characterization of biogenic SOA in rural
(Schwartz et al., 2010; Slowik et al., 2010; Setyan et al., 2014),
boreal
(Kulmala et al., 2000; Allan et al., 2006), Amazonian (Martin et al., 2010),
and some other tropical and subtropical forests (Capes et al., 2009; Robinson
et al., 2011). Using aerosol mass spectrometry, a number of studies have
identified specific signatures for isoprene-derived SOA (Allan et al., 2014;
Budisulistiorini et al., 2015). Hu et al. (2015) showed, through comparison
with model simulations, that the global distribution of a particular SOA
formation route from isoprene to epoxydiols is largely focused in the
Southern Hemisphere and over Siberian forests far from anthropogenic
emissions. The occurrence of these species in the Northern Hemisphere has
been documented in several studies (Budisulistiorini et al., 2015), but in
general the contribution is less than that reported in the South Hemisphere.
The Mediterranean region is thought to be extremely sensitive to climate
change and is influenced by air masses from the Atlantic, continental Europe,
and northern Africa, as well as increasing emissions from biomass burning,
intense shipping, and from the increasing population density in Mediterranean
coastal cities (e.g. Sciare et al., 2003; Barnaba and Gobbi, 2004; Lyamani et
al., 2006; Alados-Arboledas et al., 2011; Mallet et al., 2013). Several
studies have shown that during the summer months the aerosol radiative effect
within the Mediterranean is one of the most significant in the world
(Markowitz et al., 2002; Papadimas et al., 2012).
Sartelet et al. (2012) modelled aerosol loading in Europe and North America
and retrieved high concentrations of ozone and SOA over the Mediterranean
Sea. It was estimated that biogenic emissions contributed to the formation of
up to 72–88 % of the SOA over Europe. In order to better characterize
the sources of SOA and its precursors over the Mediterranean, the SAFMED
(Secondary Aerosol Formation in the Mediterranean) experiment took place in
the Mediterranean as part of the Chemistry-Aerosol Mediterranean Experiment
(ChArMEx; http://charmex.lsce.ipsl.fr/, last access: 22 January 2017) during Summer 2014. In this work, we present
observations from four research flights over the forested Mediterranean
region. The objectives of these flights were to combine both aerosol and
gas-phase measurements to investigate the origin of SOA over these forested
areas.
Methodology
ATR42
All airborne measurements were performed aboard the ATR42 research aircraft,
run by SAFIRE (French aircraft service for environmental research;
http://www.safire.fr, last access: 12 July 2016). The ATR (Avion de Transport Régional) is a turbo propeller
aircraft approximately 23 m long and 25 m wide, with a payload of
about 4.6 t (www.atraircraft.com, last access: 22 October 2017). The aircraft was based in Avignon, France.
Aircraft flight plans were decided depending on forecasts from meteorological
and air quality models made available on a dedicated operational web server
called the ChArMEx Operation Centre (ChOC; http://choc.sedoo.fr, last
access: 22 April 2018). A series of different
standard meteorological parameters were measured aboard the ATR42 including
temperature, pressure, relative humidity, turbulence, wind speed,
direction, and downward and upward radiances.
Online aerosol chemical and physical properties
In order to sample aerosol particle species, a forward facing aerosol inlet
was fitted in place of a side window. This inlet is designed with an outer
sleeve for channelling air and an inner tube with a large diameter and low
curvature to limit particle losses due to deposition. This inlet is both
isokinetic and isoaxial and has a 50 % sampling efficiency for aerosol
particles with diameters of 4.5 µm (Crumeyrolle et al., 2013). From
the aerosol inlet, the sampled aerosols are directed through a manifold to a
number of different instruments. Aerosol particle number concentrations were
measured using a condensation particle counter (CPC, cut-off diameter 5 nm)
and scanning mobility particle sizer (SMPS) with 162 size channels for
particle diameters ranging from 17 nm up to 400 nm, with a time resolution
of 84 s. Measurements of aerosol chemical properties were performed using a
compact aerosol time-of-flight mass spectrometer (cToF-AMS) (Drewnick et al.,
2005). This instrument was operating with a time resolution of 40 s in order
to ensure that the maximum amount of spatial information (aircraft covers
approximately 5 km in 40 s) could be obtained while maintaining a high
enough signal-to-noise ratio. Prior to being sampled into the cToF-AMS,
aerosol particles passed through a pressure-controlled inlet. This
inlet maintained a constant pressure of about 400 hPa throughout the
duration of the flight (Bahreini et al., 2008). However, all
reported concentrations are in standard temperature and pressure (used here
22 ∘C, 950 hPa). In order to provide quantitative information on
aerosol mass concentrations, a collection efficiency (CE) must be applied to
the aerosol mass concentrations. This is based on the principle that the
cToF-AMS aerodynamic inlet is designed to sample dry spherical particles and
that particles with non-spherical shapes will not be as efficiently sampled.
In addition to this, sampled aerosol particles can sometimes be lost in the
instrument as a result of particle bounce on the heating filament. This CE
correction is chemical dependant (Middlebrook et al., 2012); however since
the contribution of nitrate and sulfate remained lower than 25 % at all
times, the CE remained at 50 % throughout the sampling period. The total
mass measured by the cToF-AMS (added to that from the black carbon (BC) measurements) was
compared to the total aerosol concentration measured by the SMPS set-up
(Fig. S1 in the Supplement). BC concentrations were obtained using
a single-particle soot photometer (SP2, Droplet Measurement Technologies). Full details of this instrument are available in Baumgardner et
al. (2007).
Typical flight track travelling (a) west (RF15 and RF21)
and (b) east (RF20 and RF23) of Avignon (black circle). Points of
the flight track are coloured by organic aerosol concentrations.
Gas-phase measurements
Gas-phase species were sampled on-board through a rear-facing 1/4 inch Teflon
tube. Ozone and CO were measured using ultraviolet and infrared analysers
(Thermo Fisher environmental instruments) (Nedelec et al., 2003). The NO and NOx
measurements were performed using an ozone chemiluminescence instrument
(Environment SA AC42S instrument). The quantification of NO2 is obtained
by converting NO2 to NO using a molybdenum converter at 320 ∘C.
As some NOy is also converted into NO in the molybdenum oven, the
NO2 and NOx concentrations can be overestimated. In this work, these
measurements will be referred to as NOw, and represent
NO + NO2 + an unquantified NOy. For measurements of volatile
organic compounds (VOCs), a unit mass resolution proton-transfer-reaction mass
spectrometer (PTR-MS) from Ionicon Analytik (Innsbruck, Austria) was used,
with a time resolution of 19 s. Full details of the PTR-MS configuration
on-board and operating conditions are provided in Borbon et al. (2013).
During the so-called biogenic flights, 16 protonated masses were monitored.
Compounds of interest are
VOCs of biogenic origin (BVOCs) and their first generation oxidation
products including m/z 69 (isoprene), m/z 71 (sum of methyl vinyl
ketone,
MVK; methacroleine, MACR; and isoprene hydroxyhydroperoxides, ISOPOOH),
m/z 137 (sum of monoterpenes);
anthropogenic volatile organic compounds including
m/z 79 (benzene), m/z 93 (toluene), m/z 107 (C8 aromatics), and
m/z 121 (C9 aromatics);
oxygenated VOCs including m/z 33 (methanol), m/z 45
(acetaldehyde), and m/z 59 (acetone).
Detection limits, defined as 3σ of background mixing ratios ranged
from 0.05 to 2.70 ppbV over a 1 s dwell time. Instrumental background
signal was determined through periodic air sampling (triplicates) of ambient
air scrubbed through a custom-built catalyst converter (platinum-coated steel
wool) heated to 250 ∘C. Three complete calibrations over a
0.1–20 ppb range were performed before, during, and after the campaign. The
standard gas used was provided by Ionimed (Innsbruck, Austria) and contained
several VOCs including isoprene, α-pinene, benzene, toluene, and
o-xylene at 1 ppmV certified at ±5 %. A second parts-per-billion-level gaseous
standard from NPL (UK) was used to cross-check the quality of the calibration
and to perform regular one-point calibration control for isoprene and C6–C9
aromatics (4 ± 0.8 ppbV). A relative difference of less than 10 %
was measured. The calibration factor for all major VOCs (the slope of the
mixing ratio with respect to product ion signal normalized to H3O+)
ranged from 2.35 (m/z 137) to 18.9
(m/z 59) counts s-1.
Statistical analysis
Detailed analysis of the organic aerosol mass spectra measured using the
cToF-AMS was performed using positive matrix factorization (PMF) (Paatero and
Tapper, 1994). The PMF2 software package was
used in conjunction with the PMF evaluation tool (version 2.04; Ulbrich
et al., 2009). Recommended procedures of down-weighting for certain m/z
values were followed (Ulbrich et al., 2009) as well as removal of several
m/z values due to low (m/z 19 and 20) or high signal (m/z 29). In
this particular case, m/z values from inorganic ions (SO4, NO3)
were equally combined with the organic matrices to better separate different
factors. Error values for all m/z values were calculated in the same way
using the
SQUIRREL software (version 1.53). The number of factors was determined using
correlations with external factors (temporal series of VOC measurements).
The reported correlations used later on in the discussion were calculated
using simple linear regression.
Contribution of the different measured gas-phase species aboard the
ATR42: (a) RF15, (b) RF20, (c) RF21,
(d) RF23. The pie charts illustrated in each figure represent the
contribution of all VOC species except those of methanol and
acetone.
Electron microscopy analysis
Aerosol particles were collected on transmission electron microscope (TEM)
grids using a sampler consisting of two impactor stages. The 50 % cut-off
of each of these stages was calculated to be 1.6 and 0.2 µm, with a flow rate of approximately 1.0 L min-1. The
samples were collected only when the aircraft was travelling at a constant
altitude, usually lasting between 15 and 20 min. The TEM grids on the
submicron stage of the impactor were then analysed using a 120 kV TEM
(JEM-1400, JEOL) to provide detailed information on individual aerosol
compositions and shapes. The advantage of having TEM analysis is the ability
to detect both refractory and non-refractory aerosol particles. Composition
of each sample collected was analysed using energy-dispersive X-ray
spectrometer (EDS) with scanning-TEM mode (Adachi et al., 2014). There were
at least 230 particles analysed from each grid. Particles were classified
into five aerosol categories based on their compositions: organic aerosol (C
dominant), sulfate (S dominant), sulfate + organic (C and S dominant),
sea salt (Na dominant), and other (e.g. mineral dust (Si is dominant, but also
included traces of Ca and K)). Morphological differences between aerosol
particles also allowed us to determine the extent of internal and external
mixing. Organic aerosol generally had an amorphous morphology with no
evidence of a crystal structure. EDS analysis of the homogeneously mixed
amorphous particles showed that these particles contained C, O, and S without
any evidence of crystalline structure. Sulfate particles (likely
(NH4)2SO4) had a crystalline structure and were sensitive to the
electron beam (evaporation). Internal mixtures of organic and sulfate were
mostly crystalline (NH4)2SO4 surrounded by an amorphous carbon
material.
Back trajectory analysis
In order to determine the history of air masses prior to being sampled by the
aircraft, air mass trajectories were calculated for a 24 h period using the
Lagrangian model HYSPLIT (http://ready.arl.noaa.gov/HYSPLIT.php,
last access: 1 February 2018). These
air mass trajectories were calculated for intervals of 5 min along the
flight track and provide information on the dominant air mass sources during
the flight. Back trajectories of 24 h provided enough information to
determine whether air masses were slow moving and local or fast moving and
arriving over larger distances (Fig. S2). For all flights, the air mass
trajectory path was constant along the flight track at low altitudes, showing
that air masses of the same origin were measured during each flight. 72 h
back trajectories were also computed (although not shown) in order to
determine possible aerosol sources over longer timescales.
Overview of flights
Four research flights (RF) dedicated to biogenic emissions were carried out:
30 June (RF15), 3 July (RF20), 5 July (RF21), and 7 July 2014
(RF23). Each flight was approximately 3.5 h in duration, and the aircraft
flew over forested areas with elevations varying from 250 to 500 m above
ground level (a.g.l.) during straight and level runs. The flight plan
consisted of the aircraft leaving the city of Avignon (southern France) and
travelling east or west for about 50 km before starting a vertical sounding.
Vertical soundings were performed from around 100 up to 3300 m above sea
level (a.s.l.). Using the vertical profiles of VOC concentrations and
relative humidity, the atmospheric boundary layer height was determined for
each flight, and varied from 1300 m a.s.l. (RF20) up to 1900 m a.s.l.
(RF15) (Fig. S3). Two of the flights (RF15, RF21) flew west over the
Puéchabon Mediterranean national forest region (northwest of
Montpellier, Fig. 1a), where the principle type of vegetation is evergreen
oaks (Quercus ilex) and Alpine pines (halepensis). The evergreen oak is known
to emit several different types of monoterpene species but mainly α-pinene (Loreto et al., 1996). The other two flights (RF20, RF23) flew east
over the Oak Observatory field site at Observatoire de Haute-Provence (O3HP,
https://o3hp.obs-hp.fr, last access: 14 July 2016, Fig. 1b). This area is dominated by downy oak trees (Quercus pubescens) but also contains Montpellier maple (Acer monspessulanum) and
smokey bushes (Cotinus coggygria) in a lower canopy stage. Since Quercus pubescens is the dominant type of vegetation, it makes this region a strong
isoprene-emitting region and an ideal area to study isoprene chemistry and
its relation to aerosol particles (Zannoni et al., 2015).
Mean concentrations of the different gas-phase species measured
during low and constant altitude of each flight. The error represents
±1σ on all the measurements.
Flight
Date
Isoprene
MVK+ MACR
Monoterpenes
Toluene
Benzene
C8 + C9
O3
CO
NOw
NO
(2014)
(pptV)
+ ISOPOOH
(pptV)
(pptV)
(pptV)
aromatics
(ppbV)
(ppbV)
(ppbV)
(ppbV)
(pptV)
(pptV)
RF15
30 Jun
583 ± 290
214 ± 91
117 ± 82
146 ± 59
93 ± 61
200 ± 85
40 ± 8.8
118 ± 27
4.2 ± 0.8
0.17 ± 0.3
RF20
3 Jul
1240 ± 527
756 ± 287
205 ± 107
149 ± 82
102 ± 42
196 ± 79
53 ± 4.0
136 ± 46
7.9 ± 2.3
0.31 ± 0.2
RF21
5 Jul
600 ± 262
365 ± 182
179 ± 128
147 ± 176
108 ± 53
135 ± 39
31 ± 8.0
79 ± 11
5.86 ± 0.7
0.29 ± 0.3
RF23
7 Jul
392 ± 197
230 ± 159
119 ± 87
74 ± 34
88 ± 18
125 ± 36
52 ± 3.0
88 ± 9
5.6 ± 2.1
0.29 ± 0.9
For the flight RF20, temperatures were stable, varying from 18 to
19 ∘C in the boundary layer, and wind speeds were always less than
5 ± 1 m s-1, originating from a southeasterly direction.
HYSPLIT air mass back trajectories show that for a 24 h period prior to the
measurements, air masses were slow moving and remained within a 200 km
radius of the sampling site (Fig. S2b). This, together with the clear skies
and relatively high temperatures, made ideal conditions to study local
biogenic emissions. RF23 had similar temperatures to those recorded on RF20
(17 to 20 ∘C), but with some cloud cover. Wind speeds ranged between
2 and 4 m s-1; air masses arrived from a southerly direction passing
over the coast line prior to being sampled along the flight track (Fig. S2d).
For the two westerly flights, average temperatures were slightly higher, with average values of 23 ± 1 ∘C. Wind speeds were low (3 ± 1 m s-1). Air
masses travelled much greater distances over the western (RF15) and northwestern
(RF21) parts of France prior to being sampled (Fig. S2a, c).
Results
Gas-phase properties
Contribution of the non-refractory aerosol chemical species aboard
the ATR42: (a) RF15 3006 (b) RF20 0307, (c) RF21
0507, (d) RF23 0707.
The principal VOC species measured with the PTR-MS during all flights were
acetone (m/z 59) and methanol (m/z 33), followed by isoprene
(m/z 69) and its oxidation products (MVK + MACR + ISOPOOH)
(m/z 71) and then VOC species representative of monoterpene emissions
(m/z 137) (Fig. 2). Isoprene and its oxidation products showed a high
temporal variation during flights, suggesting a more local influence of these
VOC species. Monoterpene VOC, with a short atmospheric lifetime were
measured in low concentration with little temporal evolution. Anthropogenic
VOC species (m/z 93 (toluene), m/z 79 (benzene), and C8 and C9
aromatics) never contributed more than 5 % to the total VOCs measured
(Table 1, Fig. 2). Despite this, we cannot ignore the presence of the
anthropogenic VOC species measured during all flights. During westerly
flights (RF15 and RF21), air masses arrived from the north (Fig. S2),
possibly transporting accumulated anthropogenic emissions from over mainland
France. Easterly flights (RF20 and RF23), being principally influenced by
local or southerly air masses, are likely impacted by anthropogenic
activities over the Marseille and Fos sur Mer (Fos-Berre) industrial area.
Concentrations (µg m-3) of the different chemical
species measured aboard each flight during low-altitude legs. Error
values are standard deviations calculated on the mean values of the
measurements.
Flight
Date
NR-PM1
Org
SO4
NO3
BC
(2014)
µg m-3
µg m-3
µg m-3
µg m-3
µg m-3
RF15
30 Jun
0.70 ± 0.08
0.48 ± 0.23
0.08 ± 0.07
0.013 ± 0.01
0.03 ± 0.02
RF20
3 Jul
2.70 ± 1.10
1.79 ± 0.70
0.48 ± 0.2
0.10 ± 0.06
0.11 ± 0.03
RF21
5 Jul
0.99 ± 0.50
0.72 ± 0.40
0.20 ± 0.09
0.03 ± 0.02
0.11 ± 0.04
RF23
7 Jul
1.64 ± 0.70
0.96 ± 0.60
0.30 ± 0.18
0.06 ± 0.07
0.08 ± 0.06
Aerosol chemical properties
In the following section we will report average values for different chemical
species measured during low and constant altitude parts of the flights (below
the boundary layer height determined from the vertical profiles shown in
Fig. S3). For all flights, the aerosol composition measured by the cToF-AMS
instrument shows that the organic compounds contributed a significant
fraction to the total aerosol concentration (with average values of 72
(±36) % for RF20 and 71 (±30) % for RF23) (Table 2, Fig. 3).
The second most dominant species measured were sulfate and ammonium aerosol
particles, with a combined contribution of up to 25 ± 10 %. The
contribution from nitrate species was on average 3 (±1.5) %. BC measured using the SP2 never exceeded 5 % of the total
PM1 mass (Fig. 3). O : C values were 1.05 (±0.05) for the total
organic aerosol, with high f44 > 0.2 and corresponding low
f43 < 0.6. These mass spectral
signatures suggested that the majority of the organic aerosol was secondary,
with little influence from fresh primary organic aerosol.
Panel (a) shows (i) an example of an amorphous particle
deposited on a carbon substrate. EDS mapping analysis showing signals for
(ii) carbon, (iii) oxygen, and (iv) sulfur.
(b) Internally mixed amorphous particles with signals
(i) before and (ii) after electron beam damage. EDS
analysis showing signals for (iii) carbon, (iv) oxygen,
and (v) sulfur.
As described in Sect. 2.5, the chemical composition of aerosol particles
collected on TEM grids was determined using EDS. At least 230 particles were
analysed during each flight providing information of particle size and
composition. The absolute number of particles analysed using offline electron
microscopy is small in comparison to what is measured by online particle
counters; however this technique provides us with a qualitative snapshot
into particle mixing state, morphology, and composition. Only filters from the
submicron stages are discussed here and showed that at least 35
(±5) % of all aerosol particles measured were made up of
homogeneously mixed amorphous (no evidence of a crystal structure) particles.
EDS analysis showed that these amorphous particles were composed of
homogeneously distributed C, O, and S (Fig. 4a.i, ii, iii). The molecular
structure of these compounds is unknown. Externally mixed crystalline
sulfate particles contributed 15 % (±5 %), and 10 % were internally mixed
amorphous C and crystalline sulfate (likely ammonium sulfate) species
(Figs. 4b, S4). The remaining fractions contained signals for sea salt (Na
Cl) and dust (Si, Ca) particles.
Aerosol size distribution measured with the SMPS for (a) RF15
(b) RF20, (c) RF21, and (d) RF23 from 17 up to
400 nm. The colour scale indicates aerosol concentration
dN/dlogDp. Altitude is illustrated as the black line and
is represented on the right-hand axis.
Aerosol physical properties
Aerosol number concentrations and size distributions were measured using a
CPC and SMPS (respectively) during the four biogenic flights (Fig. 5). During
the westerly flight RF15, when air masses were travelling from the northwest
of France, particle concentrations were on average
1500 ± 300 cm-3 and the principle size mode was less than 90 nm
at altitudes of around 500 m. At higher altitudes aerosol number
concentration decreased to 600 ± 200 cm-3 with modal diameters of
around 30 ± 20 nm (Fig. 5a). The measured particle concentrations
during the other westerly flight, RF21, were on average
1895 ± 1707 cm-3. The fraction of fine particles, < 40 nm in
diameter (F40), measured during these flights was high, explaining the lower
aerosol mass measured using the cToF-AMS instrument (Table 2). During the two
easterly flights, average aerosol number concentrations were considerably
higher at 3332 ± 1920 cm-3 than the westerly flights (Fig. 5b and
d). The size distribution of the aerosol had a single mode at around 100 nm.
However, during periods with increased aerosol concentrations, the size
distribution spectra showed an additional mode between 20 and 40 nm
(nucleation-mode particles). Calculating the difference in aerosol particle
concentrations measured by the CPC (cut-off 5 nm) from that measured by the
SMPS, we were able to determine the contribution of nucleation-mode
particles.
The increases in fine-mode particles measured at lower altitudes
(∼ 500 m a.g.l.) during all flights are likely linked to new particle
formation. Observations of new particle formation from biogenic emissions
have been reported over Boreal forests (Sihto et al., 2006), European
coniferous forests (Held et al., 2004), and African savannah forests (Laakso et
al., 2013), as well as during laboratory studies (Kiendler-Scharr et al.,
2009). Monoterpene oxidation products were shown to produce new particles more efficiently by
nucleation than the isoprene oxidation products (Spracklen
et al., 2008; Bonn et al., 2014). Some of these studies have also shown that
high concentrations of isoprene relative to monoterpenes can inhibit new
particle formation (Kiendler-Scharr et al., 2009; Kanawade et al., 2011),
although the underlying processes are not yet clear. Calculating the ratio of
isopreneC / monoterpeneC (carbon associated with
isoprene / monoterpene) and comparing it to the number concentration of
nucleation-mode particles (Fig. 6), this relationship between biogenic VOC
species and nucleation-mode particles was investigated. As a result of the
low time resolution of the SMPS, we were limited to a small number of points
per flight. Data were combined for all flights, giving average ratios of
isopreneC / monoterpeneC varying between 0.05 and 8 (average
3 ± 1), with the lowest values corresponding to the highest fractions of fine-particle concentrations. Although the variation among points is high, the
general trend of these observations is in agreement with previous field
studies over mixed deciduous forests (Kanawade et al., 2011) and with
laboratory studies in controlled environments showing that high
concentrations of monoterpenes relative to isoprene can favour new particle
formation (Kiendler-Scharr et al., 2009). The average ratios of
isopreneC / monoterpeneC measured during these Mediterranean flights were
higher than those reported in Finnish forests (ratios of 0.18) and lower than
the ones measured in Michigan (ratios of 26.4) and Amazonian forests (ratios
of 15.2) (Kanawade et al., 2011). In general, high ratios are associated with
a very low or a suppressed number of new particle formation events.
Secondary organic aerosol
From cToF-AMS measurements, average mass concentrations measured during the
two easterly flights were approximately
2.0 ± 0.5 µg m-3, whereas those measured by the
westerly flights were considerably less at approximately
1 µg m-3, making more detailed analysis of aerosol chemical
properties difficult. For this reason the remaining analysis is focused on
the two easterly flights.
For both easterly flights, increases in organic aerosol concentrations were
observed in the valley area between the two high-elevation zones (between
43.6 and 43.8∘ N) during horizontal transects of the flight. For
RF20, these increases were accompanied by significant increases in the fine
particulate matter between 20 and 40 nm, and also those at 90 nm. In
addition, increases in the ratio of isoprene oxidation products to isoprene
were observed in the same region, implying a more oxidized air mass. A time
series plot of total organic aerosol (OA) with MACR + MVK + ISOPOOH
shows a good relationship (Fig. S5a), and plotting the OA concentration
against the ratio of MACR + MVK + ISOPOOH / isoprene provides us with
a means to observe the evolution of the organic aerosol with the relative age
of the air mass with respect to biogenic emissions (Fig. S5b). The ratios of
MACR + MVK + ISOPOOH / isoprene measured during this flight are
comparable to those measured over this forested area (0.4 to 0.8) during
ground-based measurements (Zannoni et al., 2016). We observe a reasonable
correlation (r=0.46) and positive slope (b=1.1) with increasing OA as
the relative air mass age increases, suggesting that SOA formation is likely
to have originated from biogenic precursors. Similar plots were prepared
using anthropogenic precursor gases toluene and benzene (Fig. S6), showing a
negative correlation with increasing organic mass concentration of r=0.35
and a slope of -0.56. However, as the toluene and benzene concentrations
are both close to the detection limit, care needs to be taken when
interpreting these ratios. Generally, although anthropogenic precursor
species are present, the VOC concentrations and trends measured suggest that
the increases in OA concentrations are primarily related to biogenic
emissions.
Ratios of isoprenec / monoterpeneC plotted as a function of the
nucleation-mode particles (difference between the CPC (cut-off 5 nm) and the
SMPS (cut-off 17 nm)). Values for the four biogenic flights, as
well as average values calculated over a number of
isoprenec / monoterpeneC ratios are included (size bins of 0.5). Error bars represent
±1σ of the average CPC5nm – SMPS values. The black line
represents the linear correlation fit.
In order to extract additional information on the OA measured during the
flights, we performed PMF analysis. Since the temporal evolution of both the
organic and inorganic concentrations was similar, we chose to combine the
mass spectral signatures for SO4 and NO3 into the PMF matrix
alongside those of the organic species. Mass spectral signatures of NH4
were not included since higher “noise” was associated with these m/z
values. Adding the inorganic signals into the PMF matrix allows us to
separate different aerosol sources and not only those related to organic
compounds. For both flights a two-factor solution (f-peak 0) was chosen to
best describe the sources of the aerosol particles and those two factors have
maximum correlations with external species (Table 3). Additional details of
the PMF analysis are included in Figs. S7 to S9 as well as Table S1 in the
Supplement. The two resolved factors include (i) a more-oxidized organic
aerosol (MOOA, contributing 55 % to the resolved factors), containing
high contributions from m/z 44 and associated with inorganic peaks
(m/z 30, 46 (NO3), and 48, 64, and 80 (SO4)), and (ii) a less-oxidized organic aerosol species (LOOA, contributing 45 %) with little
contribution from inorganic m/z (Fig. 7a). These two factors MOOA and LOOA
are very similar to the OOA-1 and OOA-2 species identified from ground-based
measurements during a biogenic event over a forested area in Canada (Slowik
et al., 2010). Slowik et al. (2010) showed similar trends with the two
identified oxidized organic aerosols, where one was associated with inorganic
aerosols and the other was not correlated with inorganic aerosols, while the
other was well correlated with biogenic VOC species.
(a) A two-factor solution determined from PMF analysis of
the biogenic research flights. (i) The more-oxidized organic aerosol
(MOOA) associated with inorganic peaks for sulfate (red) and nitrate (blue);
(ii) the less-oxidized organic aerosol (LOOA) with a lower
contribution of inorganic peaks. (b) Variations of these two species
with aging air mass (using MACR + MVK + ISOPOOH as a proxy for
photochemical age of air mass). The delta values (Δ) are calculated
based on background concentrations measured outside of the study region.
During the flights, as the valley area is approached, we observe the sampled
air masses becoming gradually more oxidized with respect to biogenic emissions,
providing us with a well-defined sample area to evaluate the contribution of
biogenic SOA on background and/or regional air masses. In order to isolate the
formation of OA resulting from the oxidation of VOC species, the change in
the OA concentrations above the background was calculated (ΔOrg). The
background values were determined based on measurements during transects of
the flight between the valley area and the airport. During this time, aerosol
concentrations were low with little temporal variation. Particle size
measurements display a single mode at 100 nm with average particle
concentrations of 3000 cm-3. Measurements of VOC species during this
background period result in average concentrations of isoprene of
1544 ± 696 pptV and lower concentrations of longer-lived
species MACR + MVK + ISOPOOH of 661 ± 239 pptV
(Table S2).
Pearson r (Pr) correlations for different time series during RF20 and RF23.
MOOA
LOOA
RF20
RF23
RF20
RF23
Pr (n=91)
Pr (n=144)
Pr (n=91)
Pr (n=169)
PTR-MS m/z 69 (isoprene)
0.11
0.41
0.51
0.67
PTR-MS m/z 71 (MVK + MACR + ISOPOOH)
0
0.28
0.64
0.71
PTR-MS m/z 137
0.59
0.58
0.29
0.49
PTR-MS m/z 93
0.15
0.25
0.27
0.38
PTR-MS m/z 79
0.04
0.21
0
0.15
BC
0.52
0.48
0.61
0.48
CO
0.37
0.48
0.34
0.60
NOw
0.56
0.44
0.51
0.60
For the resolved PMF factors, LOOA and MOOA, background values were
determined to be 0.27 and 0.41 µg m-3, respectively,
(Table S2). Organic factors corrected for background concentrations are
referred to as Δ-LOOA and Δ-MOOA. Plotting these two factors
against the ratio of MACR + MVK + ISOPOOH / isoprene (relative air
mass age) (Figs. 7 and S10), we observe a significant increase in the Δ-LOOA species with air mass age until a maximum is reached at ratios of
0.65. Given that MOOA does not change with the relative air mass age in the
measured area, and that it is associated with SO4 and NO3 species,
it is reasonable to suggest that the MOOA is associated with long-range-transported aerosol. A slower increase in concentrations of LOOA at higher
ratios suggests that as the relative photochemical age of the air mass
increases, LOOA becomes more oxidized and is converted to MOOA, as has
recently been illustrated in chamber experiments by Palm et al. (2018).
Plotting these two factors as a function of air mass age using anthropogenic
VOC species (ratio of toluene / benzene), we observe a relatively flat
and decreasing trend (Fig. S10a). These observations would suggest the
contribution of toluene and benzene, although not insignificant, plays a lesser
role in the formation of the SOA measured during these flights.
Given the good correlation between the LOOA and isoprene and its oxidation
products, we investigated the possibility of identifying isoprene-derived SOA.
Several recent publications have identified signature peaks in aerosol mass
spectrometry for isoprene-derived SOA using the m/z 53 and m/z 82
(ionization products of IEPOX) (Allan et al., 2014; Budisulistiorini et al.,
2015; Zhang et al., 2017). In this study, the contribution of these peaks in
both types of spectra was very low (fraction of signal < 0.004), although
somewhat more pronounced for LOOA. These contributions are similar to
background contributions of f82 (fraction of m/z 82 to the total
organic signal) observed globally by Hu et al. (2015) and Lee et al. (2016),
ranging from 0.0002 to 0.0035, and would lead us to believe that we have no
significant contributions of f82 in our aerosol mass spectra. Factors
influencing the formation of isoprene SOA include aerosol acidity and the
presence of NOx and sulfate (Nguyen et al., 2014), with the highest yields
of isoprene SOA being measured under low-NOx conditions (< 30 pptV)
and in the presence of acidic aerosols (Gaston et al., 2014). Aerosol
concentrations measured with the cToF-AMS appear to be fully neutralized with
little evidence of acidity (Fig. S11), and the NOw concentrations
measured during these flights varied from 6 up to 10 ppbV; however the
average concentrations of NO were 0.30 ± 0.2 ppbV, suggesting that the
real contribution of NOx (see Sect. 2.3) is also likely to be low, but
still higher than parts-per-trillion-level concentrations measured in truly remote forested
areas. There have been some reports of isoprene-derived SOA formation
(hereafter isoprene SOA) in high-NO regions but the contribution of this
pathway is considered to be much smaller (Jacobs et al., 2014).
Other sources of biogenic SOA can originate from the oxidation of monoterpene
and sequesterpene VOC species, or additionally from isoprene SOA, that do not
follow the IEPOX route. In both cases, the contribution of m/z 91 in the
cToF-AMS mass spectra, often identified as being the C7H7+
fragment (Lee et al., 2016; Riva et al., 2016), would be enhanced. This
m/z 91 was present in all OA mass spectra and was higher for the LOOA
(f91 = 0.007). However, in previous studies these f91 values
are considered background (Hu et al., 2015; Lee et al., 2016), hence
making it difficult to associate the measured SOA with these formation
routes. It should be noted that m/z 91 can also be associated with
fragments of primary anthropogenic OA, and the contribution of anthropogenic
aerosols from the industrial zone (Fos-sur-Mer) south of the flight area
cannot be ruled out.
In general, the yield of formation of SOA from the isoprene VOC precursor is
relatively low compared to other biogenic species such as monoterpenes, and
also compared with aromatic precursors (Ait-Helal et al., 2014). Since the
measured aerosol particles are neutralized (Fig. S11) and the measured NO
concentrations are still reasonably high (0.30 ppbV), we assume that
isoprene-derived SOAs, following the IEPOX formation route, do not contribute
significant amounts to the OA measured during these flights. Given the
increase in OA with the relative biogenic air mass age, we could suspect
that additional sources of SOA could originate from other isoprene SOA
formation routes and/or terpene precursors. This is also coherent with the
increase in the number concentrations of fine particles at the lower
isopreneC / monoterpeneC ratios discussed in Sect. 3.3.
Measured (green) and modelled (red) organic concentration during the
(a) RF20 and (b) RF23 flights. The concentrations are
averaged on the vertical layers of the model and variations around the
average are indicated by the horizontal error bars.
Model evaluation of secondary organic aerosol formation
In order to evaluate the relative contribution of the different gaseous
precursors to SOA formation over these forested regions, two simulations were
performed using the Polyphemus model. Full details of the model set-up are
available in Chrit et al. (2017). The domain of the air quality simulation
has a horizontal resolution of 0.125∘ × 0.125∘,
while the vertical is modelled with 14 layers with interface heights at 0,
30, 60, 100, 150, 200, 300, 500, 750, 1000, 1500, 2400, 3500, 6000,
and 12 000 m a.s.l. (Fig. S12). Biogenic emissions are
computed using MEGAN (Guenther et al., 2006), and anthropogenic emissions are
computed
using HTAP-v2 (http://edgar.jrc.ec.europa.eu/htap_v2/, last access:
11 December 2017). Initial and boundary conditions
are obtained from a larger-scale simulation (over Europe), as detailed in
Chrit et al. (2018). For gaseous chemistry, a carbon-bound approach model is
used (CB05; Yarwood et al., 2005). Aerosol dynamics is modelled with a
sectional approach (SIREAM; Debry et al., 2007), and for SOA modelling, a
surrogate approach is used (Couvidat et al., 2012). The modelling of SOA
formation is based on smog chamber experiments, which provide information on
SOA yield as a function of organic mass concentration for each precursor
using an Odum approach (Odum et al., 1996). Stoichiometric coefficients of
SOA surrogates and their saturation vapour pressures are selected to fit data
from smog chambers. Candidates for SOA surrogates are estimated from the
literature (Couvidat et al., 2012). Biogenic precursors are isoprene,
monoterpenes (with α-pinene and limonene as surrogates), and
sesquiterpenes, with low-NOx and high-NOx oxidation regimes. Isoprene
may form two surrogates, amongst which are methyl methyl dihydroxy
dihydroperoxide under low NOx, and methyl glyceric acids under high
NOx. Monoterpenes may form pinonaldehyde, norpinic acid, pinic acid,
3-methyl-1, 2, 3-butanetricarboxylic acid under low-NOx conditions, and
organic nitrate, as well as extremely low-volatility organic carbons (ELVOCs)
or highly oxidized multifunctional organic compounds (HOMs) by ozonolysis.
Anthropogenic precursors are toluene, xylene, and intermediate or semi-volatile
organic compounds (I/SVOCs). Gas-phase emissions of I/SVOCs are estimated by
multiplying emissions of primary organic aerosols by a factor 1.5 (Zhu et
al., 2016). Partitioning between the gas and aerosol phases is carried out
with a
secondary organic aerosol processor model (Couvidat and Sartelet,
2015) for organics and inorganic aerosol model ISORROPIA for inorganics
(Nenes et al., 1998). Maps of the simulated submicron organic matter
(OA1) are shown in Fig. S12a and b for the two easterly flights RF20 and
RF23, respectively.
Figure 8 compares the vertical profiles of measured and modelled OA1
during the RF20 and the RF23 flights. The concentrations averaged over the
vertical layers of the model and the standard deviations around the mean
concentrations are shown. The measured concentrations have higher standard
deviations than the modelled concentrations due to the coarse horizontal
model resolution (0.125∘ × 0.125∘). For both flights
there are some differences between the model and the measurements. However,
this discrepancy may be due to difficulties in representing the vertical
distribution of pollutants above the canopy. Although the mean vertical
concentrations of OA1 tend to be underestimated over 1000 m, they are
on average well modelled under 1000 m within the boundary layer for both
flights.
Modelled averaged composition of OA1 along the flight path
during the (a) RF20 and (b) RF23 flights. This averaged
composition is obtained by averaging concentrations along the flight path at
altitudes below 1000 m.
Although isoprene emissions are 2.5 times higher than those of monoterpenes
and 11.6 times higher than those of sesquiterpenes over that region during
the period of simulation, isoprene-derived SOA represents about 15 to
35 % of the simulated OA, which is lower than the monoterpene-derived SOA
that represents 35 to 40 %. Sesquiterpene-derived SOA represents about
10 %. Amongst those monoterpene-derived SOAs, 4 to 7 % are monoterpene
products (first-generation semi-volatile organic compounds: pinic acid,
norpinic acid, and pinonaldehyde), 9 to 14 % are ELVOCs or HOMs, and 17 to
23 % are organic nitrate. In total, biogenic-derived OA represents about
66 to 80 % of OA. The rest is made up by aromatic-derived OA (2 to
3 %) and anthropogenic intermediate and semi-volatile organic compounds
(17 to 31 %) (Fig. 9). The contribution of organic nitrate modelled is not
reflected in the measurements where less than 5 % of the total measured
mass was nitrate aerosol. This difference may be due to hydrolysis not being
accounted for in the model. Under ambient conditions hydrolysis could
eliminate the organic nitrate functionality, allowing nitric acid to
evaporate from the particles (Rindelaub et al., 2016).
The measured organic matter is highly oxidized during both flights with an
average measured O : C ratio of 1.05 below 1000 m during the RF20 flight
and 1.1 during the RF23 flight. This ratio is very well represented by the
model with average values of 1.07 (RF20) and 1.17 (RF23). In the model,
these high O : C ratios arise because of organic compounds from isoprene
oxidation, which all have an O : C ratio greater than 0.8, as well as some
ELVOC compounds (monomers) from monoterpene oxidation. We can conclude from
these observations that the low-volatility products (ELVOCs) from
monoterpene oxidation as well as isoprene oxidation products may therefore
correspond to the measured LOOA concentrations. Although the SOA contribution
of anthropogenic VOC precursors is low (Couvidat et al., 2013; Sartelet et
al., 2018), the results of the model show a high contribution of
anthropogenic compounds (up to 30 %). These anthropogenic compounds could
correspond to the regionally transported SOA, potentially identified as MOOA.
Conclusion
This paper characterizes aerosol and gas-phase physical and chemical
properties over two different forested areas in southern France. During four
dedicated flights, aerosol particles and gas-phase composition were measured
using a cToF-AMS and a PTR-MS, respectively, with the principle objective of
characterizing biogenic emissions. Aerosol particle physical properties were
measured using a number of different techniques characterizing particle size
and number concentrations. Using a combination of aerosol size distributions
coupled with VOC concentrations we observe that although new particle
formation seems to occur over all types of vegetation (mainly
isoprene-emitting species or mainly monoterpene-emitting species), that low
isopreneC / monoterpeneC ratio can favour the formation of fine aerosol
particles. These VOC species likely condense on pre-existing particles
that can then be chemically analysed.
During eastern flights, in valley areas, high concentrations of organic
aerosol and biogenic VOC species were measured (isoprene and its oxidation
products MACR + MVK + ISOPOOH). PMF analysis of the organic mass
spectra separated two organic factors, namely a more-oxidized organic aerosol
(MOOA) and a less-oxidized organic aerosol (LOOA). The MOOA species were
strongly associated with SO4 species whereas the LOOA species were not
related to inorganic species but correlated with temporal evolutions of
biogenic oxidation products (MACR + MVK + ISOPOOH). Correlation with
other precursor biogenic or aromatic VOC species was very weak.
A lack of direct evidence of IEPOX SOA (m/z 82 C5H6O+) in
the cToF-AMS measurements leads us to conclude that the formation of SOA,
following an IEPOX formation route from isoprene precursor species was not
dominant during this measurement period. The Polyphemus model determines a
contribution of isoprene SOA, formed through alternative pathways, of the
order of 15 to 35 %. Therefore, although not possible to accurately
identify the formation pathway of the measured SOA, we can, based on its
correlation with the oxidation products of isoprene, propose that it is at
least partly associated with biogenic isoprene VOC species. The model also
illustrates that although the emission of monoterpene and sesquiterpene
species is low compared to that of isoprene, the yield of SOA formation from
these precursor species is important. This is in agreement with recent
observations by Zhang et al. (2018), who showed that SOA is principally
formed from monoterpene emissions in the southern USA.
The model results estimate an overall contribution of 66 % biogenic
species and approximately 30 % anthropogenic influence to the formation
of SOA. The model can successfully replicate the measured OA during the
flights, as well as the OA oxidation properties. However, the detailed
molecular information obtained in the model (isoprene SOA, monoterpene SOA, organic nitrate) was not easily comparable to measurements.
The model resolved organic nitrate contributions up to 17 to 23 %. This
high contribution of organic nitrate is not reflected in the cToF-AMS
measurements where nitrate contributed less than 5 % to the measured
PM1 mass. This difference is possibly due to nitrate hydrolysis that is not
considered in the mode.
This study takes advantage of measurements sampling regional air masses that
were gradually enriched with biogenic compounds, allowing us to evaluate the
contribution of biogenic SOA in ambient environments. These measurements are
compared directly with model simulations, highlighting that there are
several atmospheric processes that cannot be neglected by atmospheric models
(e.g contribution of ELVOC), as well as emphasizing important processes that
need to be implemented into future model simulations (e.g. hydrolysis of
organic nitrates).