Volatile organic compounds (VOCs) contribute to air
pollution through the formation of secondary aerosols and ozone and extend
the lifetime of methane in the atmosphere. Tropospheric VOCs originate to
90 % from biogenic sources on a global scale, mainly from forests. Crops
are also a potentially large yet poorly characterized source of VOCs (30 %
of the VOC emissions in Europe, mostly oxygenated). In this study, we
investigated VOC fluxes over a winter wheat field by eddy covariance using a
PTR-Qi-TOF-MS with high sensitivity and mass resolution. The study took
place near Paris over a 5-week period and included flowering, crop maturity
and senescence. We found a total of 123 VOCs with fluxes 3 times above
the detection limit. Methanol was the most emitted compound with an average
flux of 63
Volatile organic compounds (VOCs) are key compounds for atmospheric
chemistry that contribute to the production of harmful pollutants to human
health, among which are ozone (O
Of the 760 Tg (C) yr
Reported VOC emission rates from crops are variable over a wide range of values as shown by Gonzaga Gomez et al. (2019) and Bachy et al. (2016). Thus, more data are needed to obtain reliable regional emission estimates. Studies measuring ecosystem-scale fluxes are particularly useful, since they integrate VOC sources and sinks throughout the canopy. For example, Bachy et al. (2018) showed that methanol can be emitted from bare agricultural soils in comparable quantities to plants. The emergence of highly sensitive time of flight PTR-MS (PTR-TOF-MS) (Sulzer et al., 2014) enables the detection of a lot more VOCs than with previous quadrupole PTR-MS and allows ecosystem-scale measurement of their fluxes by eddy covariance. This opens the possibility of obtaining a much more complete spectrum of VOC fluxes from crops than in previous studies. The objective of this study was therefore to quantify the fluxes of VOC exchanged between a winter wheat field and the atmosphere at the ecosystem scale using the eddy covariance method over periods of flowering, grain filling and senescence, using a highly sensitive PTR-Qi-TOF-MS. The methodology to compute the fluxes and their uncertainties are presented in detail, and the emission and deposition fluxes are discussed in terms of their magnitude and timing.
Flux measurements took place at the Integrated Carbon Observation System
(ICOS) FR-Gri site (Grignon, 48
Winter wheat was sown on 20 October 2015 at a density of 2 500 000 plant ha
Satellite image of the site showing the ICOS FR-GRI field (yellow line), the ICOS flux station (red dot) the VOC eddy covariance sampling site (blue dot) and the VOC profile sampling site (green dot). The dotted red line shows the farm buildings where animals are mainly stored, on the southeast of the field. © Google Earth.
VOCs were measured in 30 min cycles. The eddy covariance flux was
recorded over 20 min, while the last 10 min were devoted to vertical
profiles (5 min) and chamber (3 min) mixing ratio measurements and zero
checks (2 min) by sampling through a hydrocarbon filter. In this study, the
profiles and chamber measurements are not presented. A 16-way Sulfinert
coated valve, located 5 cm from the drift tube of the PTR-Qi-TOF-MS and
heated to 80
The VOC eddy covariance mast comprised a sonic anemometer (R3-50, Gill, UK)
placed at the reference height of 2.7 m aboveground and an open-path IRGA (infrared gas analyser) for
CO
The eddy covariance VOC measurement setup, showing
A PTR-Qi-TOF-MS (Ionicon Analytik GmbH, Innsbruck, Austria) was used for
continuous online measurements of VOC mixing ratios at 10 Hz. The analyser
has been described in detail before (Abis et al., 2018; Sulzer et al., 2014).
The drift tube was maintained at 3.5
A data acquisition software was developed on Labview to synchronize the
measurements of the PTR-Qi-TOF-MS with the sonic anemometer and the other
fast-response instruments and to store the data. This software was based on
the acquisition of a list of ion peaks integrated online by the TOFDaq
software (TOFWERK, SW) using shared variables that are exchanged between the
acquisition computer and the PTR-Qi-TOF-MS computer via a local Ethernet
connexion. The desired ion peak list was set up at the start of the
experiment based on a list of known compounds tuned to the site. The
acquired ion peak integral counts per seconds (cps) were synchronized at 20
Hz with the ultrasonic anemometer. Every 5 min, a mass calibration was
performed by the TOFDaq software using masses
The pre-processing steps provided averages and standard deviation counts
per seconds (cps) for the ion peaks selected
The PTR-Qi-TOF-MS calibration factor was measured five times during the
experiment. To that purpose, we used a standard calibration mixture cylinder
containing 102 ppb of benzene, 104 ppb of toluene, 130 ppb of
ethylbenzene and 336 ppb of xylene (122 ppb ortho, 121 ppb meta, 123 ppb
para; Messer). The gas mixture from this cylinder was diluted with
synthetic air (Alphagaz 1, Air Liquide, France), filtered with a hydrocarbon and
humidity filter (Filter Super Clean, final purity
The background mixing ratio was determined during each calibration using
synthetic air passed through a hydrocarbon filter (Supelco ref 22445-12) for
2 min and keeping the last 30 s of the record. The background was
also determined every 30 min by passing ambient air through the same
filter to account for the ambient humidity in the zero calibration. The
background mixing ratio was determined as the minimum between a 10 d
moving minimum of the zero measured every 30 min and the zero measured
during the calibrations. This procedure was used since the zero air
concentration was sometimes much larger than the one measured with filtered
ambient air. The zero was then withdrawn from the uncalibrated mixing ratio,
providing the zero corrected mixing ratio
The difference in
The relative uncertainty of the calibration factors
The fluxes were computed as the covariance between the vertical component of
the wind velocity and the mixing ratio in dry air
Since the PTR measures a mixing ratio
The magnitude of high-frequency losses was evaluated as the difference between the cross-spectrum of the first water cluster and air temperature in the high-frequency domain, based on the methodology of Ammann et al. (2006). High-frequency losses could not be computed for VOCs due to too high a noise-to-signal ratio, which made the high-frequency part of the spectrum non-exploitable for computing the high-frequency losses. The loss of signal was starting at a frequency around 0.2 Hz, and the signal was halved at around 2 Hz (Fig. S5). Since the power-spectral frequency was at around 0.2 Hz at around 14:00 h UTC, the high-frequency loss appears just after the spectral peak during the day, but the decrease was gentle until 2 Hz and most of the signal energy was contained below 2 Hz. The high-frequency losses were evaluated by comparing the integrated co-spectra (co-ogives) for water vapour cluster and air temperatures. High-frequency losses were evaluated as being less than 5 %. See Ammann et al. (2006) for details. They were therefore not corrected in the following.
The limit of detection for fluxes (LOD
Since the PTR-MS only measured a mass-to-charge ratio
Meteorological measurements included wind speed, air and soil temperatures,
and humidity as well as rainfall and global, net and photosynthetic active
radiation. The vertical profile of air temperature and wind speed was
measured with five two-dimensional ultrasonic anemometers (Wind Sonic, GILL,
UK) and shielded thermocouples (HMP155, Vaisala, Finland) placed at 0.5,
1.0, 2.0, 3.0 and 5.0 m above the ground. The CO
CH
The FIDES concentration and flux footprint model (Flux Interpretation by
Dispersion and ExchangeS) was used to evaluate the flux footprint and to
infer the contribution of the VOC fluxes to the VOC concentrations measured
at the site (Loubet et al., 2010; Carozzi et al., 2013; Loubet et al.,
2018). The flux footprint
All data were merged into a common dataset according to their date and time and averaged at hourly time steps. The dataset is available as an R format file in the Supplement. All statistics were performed under R (R version 4.0.1, 6 June 2020).
The experiment started after wheat anthesis, during the grain-filling period
(Ripening stage, Fig. 3). During that period, the grain was
progressively filled up to reach a biomass of around 500 g DW m
The campaign started a week after a major flooding event in the Parisian
area characterized by a 3-week rain period that might have
affected the crop functioning. However, the CO
Field flux footprint, friction velocity (
The typical friction velocity of the site was around 0.3 m s
The major VOCs at the site were methanol, formaldehyde, ethanol, furan,
acetic acid, acetone, hydroxyacetone, acetaldehyde, isoprene and
monoterpenes (Table S3). The mixing ratios of the most emitted and
deposited VOCs showed no marked daily patterns but some similarities in their
weekly patterns (Fig. S7). During the first week, mixing ratios
were lower and diurnal variations smaller than during the rest of the
experimental period. This weak corresponded to a rainy period with westerly
winds typical of oceanic influence at the site (Fig. S6). The
second week showed an increase in the mixing ratios of all compounds that
mostly lasted for 3 weeks. This period corresponded to the end of the rainy
period, a sudden change in the wind direction to the east and an increase in
temperature and water vapour content of the atmosphere (Fig. S6).
This event also brought polluted air masses from the Parisian area as shown
by the increase in NO
This period also showed a concentration increase in several oxygenated,
mainly deposited compounds like hydroxyacetone, 4-oxopentanal, methylfuran
or propyne (Fig. S7). Week 3 corresponded to the windy period
(Fig. 4) with a well-mixed boundary layer both during day and
night and to wind blowing from the farm. Week 4 is marked by maximum
concentrations of methanol and acetaldehyde but also of most of the
oxygenated compounds, the daily pattern of these being more marked during
this week. This period, which corresponded to the end of the senescence
period, was also characterized by high air temperatures (up to 30
In total, 123 VOCs had fluxes greater than 3 times the flux detection
limit (LOD
The most emitted compound was methanol (
Whole ecosystem net fluxes of the six most emitted VOCs. Each week
shows the diel cycle with its mean (line) and standard deviation (ribbons).
The
The most net-deposited VOCs were formaldehyde (
Whole ecosystem fluxes of the six most deposited VOCs. Each week
shows the diel cycle with its mean (line) and standard deviation (ribbons).
The
The sum of emissions of all VOCs was 1.02
Stacked daily averages of the six most emitted
Because of the air density fluctuations, the Webb–Pearman–Leuning (WPL) correction needs to be applied to the eddy covariance flux (Webb et al., 1980). This correction accounts for dilution by water vapour and temperature-induced density variations in the analysed air. Because of the long and heated sampling lines and large thermal mass of the drift tube, temperature variations can be neglected. Water vapour dilution however needs to be accounted for, since the water vapour fluctuations are not directly measured in the PTR-MS. Eq. (S9) shows the WPL correction specific to the PTR-Qi-TOF-MS. We found that most of the time this correction was smaller than a few percent (Fig. S1) because of the relatively small mixing ratios of the measured VOCs in ambient air.
Another issue that, to our knowledge, was not discussed before concerns the
effect of normalizing the cps by the cps of H
High-frequency losses were evaluated to be small and in line with the
expected values for the lag time of our sampling system (lower than 5 %)
(Ammann et al., 2006). These high-frequency losses were
estimated from the first water cluster (
Overall, we were able to identify more than 123 VOCs that had fluxes larger
than 3 times the flux limit of detection (LOD
In our study methanol (
More specifically, our study confirms the increase in methanol emissions during senescence (Fig. 5) as already observed by Bachy et al. (2020) and Gonzaga Gomez et al. (2021) for wheat and Mozaffar et al. (2018) for maize. These observations collectively suggest that cereal crops become a major source of methanol at the end of the cultural cycle with net emission rates possibly exceeding those during the vegetative growth phase. This seasonal peak can be explained by the demethylation of pectin in senescent plant tissues (Fall and Benson, 1996) along with the pronounced degradation of cellular components at the end of chlorosis (Keskitalo et al., 2005; Woo et al., 2019), both of which promote methanol production and its release through higher leaf porosity. In addition, methanol emissions were enhanced by high tissue temperatures during the senescence period due to warm weather conditions and reduced transpiration (Figs. S6 and 6).
We recorded rather small deposition rates at the end of the night, when the canopy was the wettest (Fig. 5). As a very soluble compound, methanol can be easily adsorbed in water layers and desorbed when the canopy dries out (Laffineur et al., 2012; Bachy et al., 2018). However, we did not find any clear desorption pattern in the morning, suggesting instead moderate net surface deposition at night in our study.
Ethanol (C
Furan emissions from crops (tentative identification of
Acetaldehyde (
Globally, In our study, acetone (
Fluxes and mixing ratios of the 9 most emitted VOCs found in this
study, together with isoprene and monoterpenes, compared to literature
values using different methods of measurement. VOC fluxes measured by eddy
covariance refer to the whole ecosystem including soil and are expressed per square metre of ground surface. Fluxes from chamber measurements refer to
projected surface and dry weight (DW) of the enclosed aboveground organ of wheat.
Means
F1988: Fall et al. (1988). K1995: Kanda et al. (1995). K2009: M. Karl et al. (2009). B2020: Bachy et al. (2020). G2019: Gonzaga Gomez et al. (2019). M2016: Morrison et al. (2016). In this
study, 18 T ha
To date dimethyl sulfide emissions (DMS,
Crops are generally regarded as low emitters of isoprene (
Monoterpenes (MT,
Phenol (
Formic acid (CH
The third most deposited ion was C
Hydroxyacetone was the fourth most deposited compound (
Fluxes and mixing ratios of the six most deposited VOCs
The fifth most deposited compound was acetic acid (
Methylfuran (
Diurnal cycle of deposition velocity of the six most deposited VOCs. Each week shows the diel cycle with its mean (line) and standard
deviation (ribbons). The
4-Oxopentanal (4-OPA,
C
Since many volatile organic compounds were bidirectionally exchanged, the
question arises whether the observed mixing ratios correspond to those of
the suburban region or were largely influenced by the local environment,
especially the nearby farm with livestock and the surrounding fields. By
analysing the mixing ratio frequency as a function of the wind direction and
wind speed, several compounds were clearly identified as coming from the
nearby animal farm (Figs. 9, S12–S14 and Table S3). Some of these VOCs are directly emitted by the farm (Kammer et al., 2020), namely acetic acid (
Three typical wind roses of compounds identified as coming from the
farm. Colours show the mixing ratios, and plots are binned by averaged wind
direction and wind speeds.
The VOC
By contrast, the VOC
Finally, VOCs
We found that the PTR-Qi-TOF-MS was sensitive enough to measure the fluxes of
264 VOCs by eddy covariance (77 emitting and 187 deposited, fluxes 3 times the LOD
A detailed analysis of the eddy covariance flux computation revealed that the
Webb–Pearman–Leuning (WPL) correction was negligible (most of the time lower
than 2 %). Similarly, normalizing by the H
Our measurements confirm previous findings showing that methanol is the most
emitted compound from wheat fields, representing 52 % of the total VOC
moles emitted by the crop. Acetone and acetaldehyde were also found to
contribute significantly to the summed VOC emissions. However, we detected
several other compounds not previously reported as emitted by wheat, in
particular, furan (
It is the first time 123 VOCs fluxes (move than 3 times LOD
Using a PTR-Qi-TOF-MS to measure VOC fluxes over an ecosystem for a long period by eddy covariance allows estimating the fluxes of a large number of compounds and establishing a net flux. The main difficulty continues to be the clear identification of the compounds corresponding to each ion as well as the calibration of a large number of gases.
The complete hourly dataset is available as an Rdata file
(COV3ER_2016_dataset.Rdata) containing the whole dataset, together with a data description file
(COV3ER_2016_dataset-description.xlsx) and a script for reading and plotting the data (COV3ER_2016_dataset_make_graphs.R). The data and code for visualization are available at
The Supplement includes a PDF document with supplementary figures
and tables and an Excel file containing the averaged fluxes, mixing ratios and LOD
This study was conceptualized by BL, who also developed the software for data acquisition and analysis with OZ. The experiment was supervised by all authors except LA, JK, SB and MS. The data were curated by FL, RC, PB, LGG, BL, FT, DB and VG. The formal analysis was performed by BL, LGG, LA, FL, PB and RC. All authors except LA, JK, SB and MS provided resources. BL wrote the draft of the manuscript, and all authors contributed to the writing and reviewing. BL and VG contributed to finding the funding resources.
The contact author has declared that neither they nor their co-authors have any competing interests.
Publisher’s note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We acknowledge funding by ADEME (COV3ER, no. 1562C0032) and ANAEE-FR services (ANR project no. 11-INBS-0001). The data analysis was supported by the regional funding R2DS and the ADEME projects DICOV (grant no. 1662c0020) and AGRIMULTUPOL (grant number 1703C0012). Sandy Bsaibes acknowledges the European Union's Horizon 2020 research and innovation programme under the Marie Sklodowska Curie grant agreement no. 674911-IMPACT. The measurements were performed on the ICOS FR-GRI site. We acknowledge Dominique Tristan field site. The field site FR-GRI is part of the ICOS Europe and ICOS France infrastructures.
This research has been supported by the Agence de l'Environnement et de la Maîtrise de l'Energie (grant no. COV3ER, 1562C0032) and the Agence Nationale de la Recherche (grant no. ANAEE-FR, 11-INBS-0001).
This paper was edited by Kyung-Eun Min and reviewed by two anonymous referees.