Introduction
During the last decades, many field measurement campaigns have reported
unusually high nitrous acid (HONO) concentrations during daytime for remote
(Zhou et al., 2002; Acker et al., 2006a; Sörgel et al., 2011a; Villena
et al., 2011; Oswald et al., 2015; Meusel et al., 2016), semi-urban (Neftel
et al., 1996; Staffelbach et al., 1997; Kleffmann et al., 2005; Su et al.,
2008; Li et al., 2012; Yang et al., 2014) and urban regions (Kleffmann et
al., 2002, 2003, Ren et al., 2003, 2006; Acker et al., 2006b;
Elshorbany et al., 2009, 2010; Hou et al., 2016; Lee et al., 2016). These
results stimulated laboratory investigations on potential HONO precursors
from which the most frequently discussed mechanisms are (i) the
photosensitized reduction of nitrogen dioxide (NO2) by organic
material, e.g. humic acids (George et al., 2005; Stemmler et al., 2006;
2007, Sosedova et al., 2011; Han et al., 2016), (ii) the photolysis of
adsorbed nitric acid (Zhou et al., 2003, 2011; Laufs and Kleffmann, 2016),
(iii) biogenic production of nitrite in soil (Su et al., 2011; Ostwald et
al., 2013; Maljanen et al., 2013; Oswald et al., 2015; Scharko et al., 2015;
Weber, 2015) and (iv) release of adsorbed HONO from soil surfaces after
deposition of strong acids (VandenBoer et al., 2013, 2014, 2015; Donaldson
et al., 2014). Another discussed source, the reaction of excited gaseous
NO2 with water (Li et al., 2008), is of minor importance as
demonstrated by laboratory (Crowley and Carl, 1997; Carr et al., 2009;
Amedro et al., 2011) and modelling studies (Sörgel et al., 2011b; Czader
et al., 2012). Also the photolysis of nitrophenols or similar compounds
(Bejan et al., 2006) is meaningful only in polluted areas, where
concentrations of these precursors are high. Finally, the gas-phase HONO
source by the reaction of HO2⋅ H2O complexes with
NO2, recently postulated by Li et al. (2014), could not be confirmed by
the same group in simulation chamber experiments (Li et al., 2015) and is
also in conflict with recent aircraft measurements (Ye et al., 2015).
Several field studies point to an atmospheric daytime HONO source by
heterogeneous photosensitized reduction of NO2 on organic substrates
(Kleffmann, 2007). In these studies, calculated daytime HONO sources,
determined from HONO levels exceeding theoretical photostationary state
(PSS) values, showed high correlations with the photolysis rate coefficient
J(NO2) or the irradiance and NO2 concentration (Vogel et al., 2003; Su
et al., 2008; Elshorbany et al., 2009; Sörgel et al., 2011b; Villena et
al., 2011; Wong et al., 2012; Lee et al., 2016; Crilley et al., 2016).
However, concentrations are not only controlled by the local ground surfaces
source processes but also depend on the convective mixing in the atmosphere,
leading to a potential misinterpretation of the correlation results. In
addition, the assumed PSS conditions may also not be fulfilled when HONO and
its precursors were measured close to their sources (Lee et al., 2013;
Crilley et al., 2016).
In contrast, flux measurements are able to give direct information about
ground surface production and loss processes and are potentially a better
tool to investigate HONO sources in the lower atmosphere. Nowadays, eddy
covariance (EC) is the most commonly applied method to measure fluxes
between the surface and the atmosphere. The lack of fast and sensitive
EC-HONO measurement systems, however, requires the use of indirect methods
like the aerodynamic gradient (AG) method that has been described by a
number of authors (Thom et al., 1975; Sutton et al., 1993) or the relaxed
eddy accumulation method (REA) that was also recently used for HONO (Ren et
al., 2011; Zhou et al., 2011; Zhang et al., 2012). Unfortunately, the
available flux observations indicate different HONO precursors. Harrison and
Kitto (1994) and Ren et al. (2011), for example, found a relationship of the
HONO flux with the NO2 concentration and also its product with light
intensity, whereas Zhou et al. (2011) observed a correlation of the HONO
flux with adsorbed nitric acid and short wavelength radiation. A campaign
above a grassland spread with manure (Twigg et al., 2011) found no evidence
for an NO2-driven mechanism producing upward HONO fluxes at a local
field site, although HONO and NO2 concentrations were coupled with one
another, which indicated a regional connection. Hence, the origin of the
ground surface daytime HONO source is still a topic of controversial
discussion.
The present study was part of the German–French (DFG/INSU-CNRS) PHOTONA
project (PHOTOlytic sources of Nitrous Acid in the atmosphere) with
laboratory and field investigations concerning HONO in the troposphere. In
the present work only the field campaigns aimed at elucidating unknown
sources of HONO by flux measurements above an agricultural field site are
described. The measurements were performed during different seasons of the
year and above different types of canopies using the AG method and the LOPAP (LOng Path Absorption Photometer) technique.
Methods
Field site
The measurement site is an agricultural field located in Grignon, around 40 km west from Paris, France (48.9∘ N, 1.95∘ E). This site, operated by INRA
(Institut National de la Recherche Agronomique) and hosted by AgroParisTech,
was part of the EC CarboEurope-IP, NitroEurope-IP, Eclaire-IP and INGOS-IP
European projects and is part of the ICOS and Fluxnet measurement networks.
The site is well documented (Laville et al., 2009; Loubet et al., 2011) as
are several experiments on reactive trace gases performed at the site (Bedos
et al., 2010; Loubet et al., 2012, 2013; Potier et al., 2015; Personne et al., 2015; Wu et al.,
2016; Vuolo et al., 2017). Briefly, measurements were
carried out on a 19 ha field with a fetch of 100–400 m depending on wind
direction. Roads with substantial traffic surround the site to the south
(700 m), east (900 m), north-west (200 m) and south-west (500 m). Other
agricultural fields surround the site to the north, south and east. The
small village Grignon is located to the west around 700 m away from the
measurement site. An animal farm with an average annual production of 210
cattle, 510 sheep and 900 lambs is situated 400 m to the south-south-west.
The soil on the field is a silt loam with 31 % clay, 62.5 % silt and
6.5 % sand and was managed with a maize, winter wheat, winter barley and
mustard rotation. The field is annually fertilized with nitrogen solution
and cattle manure at a rate varying between 100 and
300 kg N ha-1 yr-1, with manure usually applied every 2 to 3 years.
Experimental design
Three field campaigns were performed during the PHOTONA project over a range
of crop developments and types: PHOTONA 1 was carried out from 20 to
30 August 2009 over bare soil (for more details see Stella et al.,
2012). PHOTONA 2 was a spring campaign from 7 to 19 April 2010
following fertilization with nitrogen solution (composed of 50 % of
ammonium nitrate and 50 % of urea, with a nitrogen content of 39 % in
dry mass) on 17 March (60 kg N ha-1) and 6 April
(40 kg N ha-1) over a growing (from 0.2 to 0.3 ± 0.05 m height)
triticale (hybrid of wheat – Triticum – and rye – Secale) canopy seeded on 14 October
2009. PHOTONA 3 was a second summer campaign from 16 to 30
August 2011 over a well-developed maize canopy of about 2 m height, seeded
on 21 April 2011. The last fertilization of the site before the
campaign was on 17 March 2011 with cattle slurry application of
25 m3 ha-1. The slurry composition was 10.5 % dry matter,
28.9 g N kg-1 of dry matter (of which 12.4 g N kg-1 was
ammonium), corresponding to an amount of nitrogen of 33.4 kg N ha-1.
HONO aerodynamic gradient set-up during the PHOTONA
campaigns. Left: PHOTONA 1 (bare soil) and 2 (triticale canopy) with the two
external sample units fixed on a mast. Right: scaffold tower during PHOTONA
3 (maize canopy) with the LOPAPs placed at different levels on the tower.
During all campaigns HONO mixing ratios were measured at two heights above
the canopy using the LOPAP technique (QUMA Elektronik & Analytik GmbH,
Germany), which is explained in detail elsewhere (Heland et al., 2001;
Kleffmann et al., 2002). The LOPAP instrument allows detection of HONO down
to mixing ratios of 1 pptV and the instrument showed excellent agreement
with the DOAS (differential optical absorption spectroscopy) technique during
intercomparison studies (Kleffmann et al., 2006). Recently, 15 %
interference against HNO4 was inferred from laboratory experiments for
the LOPAP instrument (Legrand et al., 2014). However, because of the typical
high temperatures and low NO2 levels during daytime, low HNO4
levels (< 50 ppt) are expected for the present study, leading to no
significant overestimation of the HONO data. In addition, since HONO fluxes
are calculated only from the difference of both instruments, potential
HNO4 interferences are not considered important in the present study.
Two LOPAP instruments were placed in thermostated field racks, with the
external sampling units fixed at two heights on a mast in the open
atmosphere (see Fig. 1). Other trace gases measured during the campaigns
were NO2 that was detected with the sensitive Luminol technique (LMA-3D
or LMA-4, Unisearch Associates Inc., Ontario, Canada), NO (CLD780TR,
Ecophysics, Switzerland) and O3 (FOS, Sextant Technology Ltd, New
Zealand), which are based on chemiluminescence techniques. Since the
Ecophysics NOx instrument is relatively slow and measures NO and
NOx consecutively, only the NO channel was used while NO2 was
detected by the Luminol instrument. Here potential overestimation of
NO2 by interferences against peroxyacyl nitrates was ignored since
their concentrations were unknown. NO2, NO and O3 were sampled at
a flow rate exceeding 40 L min-1 with 7 m (9.24 mm internal
diameter; i.d.) plus 7 m (3.96 mm i.d.) long PFA (perfluoroalkoxy polymer) sampling
lines during PHOTONA 1, 7 m (9.24 mm i.d.) plus 3 m (3.96 mm i.d.) long PFA
sampling lines during PHOTONA 2 and 10 m long (9.24 mm i.d.) plus 2 m
(3.96 mm i.d.) sampling lines during PHOTONA 3. The residence times in the
sampling inlets were estimated to be between 1.6 and 3 s, which were short
enough to avoid significant chemical conversions, e.g. by the reaction of
O3 with NO.
The different canopy heights used during the three campaigns required slight
changes in the AG set-up. During PHOTONA 1 and 2, the external sampling units
of the LOPAPs were fixed on a small mast at heights of 0.15 and 1.5 m and
0.3 and 1.5 m, respectively (see Fig. 1). Note that the lowest height in
PHOTONA 2 was just at the top of the canopy towards the end of the campaign
but was always above the displacement height (0.24 m; for definition see
Sect. 2.3). The canopy was also quite heterogeneous at that time as shown
by the 15 % coefficient of variation of the canopy height and the area
around the LOPAP had, on average, a lower canopy height. For PHOTONA 3, a
scaffold tower of around 5.5 m height with two levels was installed on
the field; on each a LOPAP was mounted (inlet sampling heights
3.0 and 5.2 m). All other trace gases were measured at three heights
during PHOTONA 1 (0.2, 0.7, 1.6 m) and 2 (0.4, 0.6, 1.5 m) and one height
during PHOTONA 3 (5.0 m), using one instrument for each trace gas connected
to Teflon solenoid valves (NResearch, USA). Measurements were made at 30 s
intervals at all three heights (for details see Stella et al., 2012). During
all campaigns the sampling inlets were positioned facing away from the field
racks towards the prevailing wind direction in order to minimize turbulence
disruptions by the racks themselves. For EC measurements a sonic anemometer
(R3, Gill Inc., UK) was mounted on a nearby mast at a height of 3.17 m
during PHOTONA 1 and 2 and 5.0 m during PHOTONA 3.
Furthermore, meteorological parameters such as wind speed at different
heights (cup anemometer, Cimel, FR), wind direction (W200P, Campbell Sci.
Inc., USA), relative humidity (RH) and air temperature (Tair) (HMP-45,
Vaisala, FI) as well as soil parameters like the soil temperature
(Tsoil) (copper–constantan thermocouples) and soil water content at
different depths (TDR CS 616, Campbell Sci. Inc., USA) were measured
continuously. The photolysis frequency J(NO2) was measured using a filter
radiometer (Meteorologie Consult GmbH, Germany) during PHOTONA 1, 2 and 3
and a spectral radiometer (Meteorologie Consult GmbH, Germany) during
PHOTONA 3, by which also J(O1D) and J(HONO) were determined. During PHOTONA 1 and 2
J(HONO) was calculated from measured J(NO2) using the method described by Kraus and
Hofzumahaus (1998).
Aerodynamic gradient method
The HONO flux was calculated from the AG method by using a flux–profile
relationship based on the Monin–Obukhov (MO) similarity theory that
describes the non-dimensional gradient of a scalar χ (i.e. the
concentration of HONO, c(HONO)) as a universal function of the atmospheric
stability parameter (z-d)/L (e.g. Kaimal and Finnigan, 1994).
κ⋅(z-d)χ∗⋅∂χ∂z=ϕ(z-d)/L
Here, κ is the von Karman constant (0.41), χ the measured
scalar, χ∗ the scaling parameter of χ, z the measurement
height, d the displacement height and L the Obukhov length. The
displacement height accounts for the disturbance of the canopy on the flow
and was taken as 0.7 ⋅ hc (hc is the height
of canopy), a common parameterization in micrometeorology which was validated
for this field site by Loubet et al. (2013). During the 1960s and 1970s a lot
of effort was spent in the determination of the universal function
φ(z-d)/L and its primitive Ψ(z-d)/L (Swinbank, 1964, 1968;
Businger et al., 1971). In the actual work, the universal function for heat
ΨH,(z-d)/L as published by Businger (1966) was integrated with the
method of Dyer and Hicks (1970) for the unstable case. For the stable case
the universal function for heat as published by Webb (1970) was used (see
Supplement).
The flux of a scalar, which is equal to u∗⋅χ∗, can be deduced from Eq. (1), which leads for HONO to (see
Supplement for the development of this equation)
Fzref=-κ⋅u∗⋅∂c(HONO)∂[ln(z-d)-Ψ(z-d)/L].
Here, c(HONO) is the concentration of HONO. Fzref is representative of the flux
at the geometric mean height of the concentration measurements, zref,
which we hence define as zref=(z1-d)⋅(z2-d), where z1 and z2 are the measurement
heights above the ground. The friction velocity u∗ and the Obukhov
length L were calculated from EC measurements as explained in
detail in Loubet et al. (2011).
Data treatment
To interpret the flux data for each measurement campaign, first 30 min
(PHOTONA 1 and 2) or 5 min (PHOTONA 3) averages were formed from the
measurement data including the HONO fluxes. Only 52, 77 and
78 % of the campaign data could be used to determine HONO fluxes during
PHOTONA 1, 2 and 3, respectively. For the other periods, data from all
instruments were simultaneously not available (calibrations, zeros,
intercalibrations, malfunctions, etc.). Secondly, for each campaign a diurnal
average day using all this averaged data was calculated by the formation of
1 h means from the whole measurement period. Using this procedure the
errors of the individual measurements were reduced by averaging over a large
number of values. However, some filtering steps were also applied which
removed rain (PHOTONA 1: 24 August 2009; PHOTONA 3: 26 August 2011) and high-emission
(PHOTONA 2: 7 April 2010) events from the data. These events led to higher noise
in the daily patterns of the HONO flux and were therefore classified as
unusual conditions or artefacts that did not represent a common flux profile
of the studied agricultural field site. In addition, for PHOTONA 3 the flux
data for the first measurement period (16–21 August 2011) were also discarded,
caused by low quality of the first intercalibration. From all available flux
data, 97, 99 and 57 % were finally used to calculate the
average days for PHOTONA 1, 2 and 3, respectively.
Finally, for the correlation analysis of the diurnal average (see Sect. 3.4), weighted orthogonal regression fits (Brauers and Finlayson-Pitts,
1997) of the HONO flux against different variables were applied using the
standard error (SE) of the 1 h average for weighting (SE being the standard
deviation divided by the square root of n, the number of data). For the
correlations, the lower level data were used for PHOTONA 1 and 2 since they
better describe the proposed ground surface source processes. For PHOTONA 3
NOx data were available only for one level (5 m), which was used here.
To assess the goodness of these fits the merit function χ2 and
the goodness-of-fit parameter Q were determined (Brauers and Finlayson-Pitts,
1997). A small χ2 and a large Q indicate a strong linear
correlation of the analysed parameters.
HONO mixing ratios measured at the same height during
PHOTONA 1–3 (from left to right). The solid lines show linear weighted
orthogonal regressions (Brauers and Finlayson-Pitts, 1997) between the two
instruments. The slope (b) and intercept (a) are given with their (2σ) standard deviations.
Quality of the HONO flux
Estimation of the aerodynamic gradient uncertainty
The following main factors may influence the error of a flux calculation
using the AG method. First of all, flux gradient relationships have been
studied for quite some time and show good similarities for trace gases such
as CO2 or sensible and latent heat using the above-described universal
functions, but there is always some uncertainty when using an indirect method.
Moreover, for HONO the flux similarity has never been compared to other
techniques (e.g. the EC method). However, during PHOTONA 1 fluxes of nitric
oxide (NO) and ozone (O3) were measured additionally by EC and were in good agreement (O3), or at least comparable (NO) with
fluxes calculated by the AG method (Stella et al., 2012). This demonstrated
the applicability of the gradient method at the local homogeneous field
site.
For the calculation of the uncertainties of the HONO flux by Eq. (2), errors
of the gradient (σgradient) and of u∗ (σu∗) are of direct importance. During all campaigns, HONO was
measured at two heights using two LOPAPs. Hence, the quantification of the
gradient strongly depended on the accuracy of these two instruments. The
LOPAPs were therefore intercompared several times in the field by placing
the external sampling units beside each other and also by using a common PFA
inlet line and a T piece between the sampling inlets. In order to
estimate the error of both instruments, again weighted orthogonal
regressions (Brauers and Finlayson-Pitts, 1997) were applied, using the
precision errors of both LOPAPs for weighting (see Fig. 2). The
inter-comparisons showed excellent agreement during PHOTONA 1 and 2, with a
small intercept and a slope close to 1, demonstrating the capability of the
used method to calculate gradients. Not quite so good agreement was obtained
for PHOTONA 3, which may partly be explained by the lower HONO levels. To
reduce systematic errors in the flux calculation, the lower LOPAP was
harmonized using the linear regression fits shown in Fig. 2.
The error of the gradient was then calculated from the precision of the
instruments (σLOPAP) and the errors of the slope (Δb)
and the intercept (Δa) of the regression fit (see Fig. 2), using
95 % confidence intervals (2σ). The HONO concentration,
c(HONO), always refers to the higher value of both instruments in order to obtain
the maximum deviation.
σgradient=σLOPAP2+Δb2⋅c(HONO)2+Δa2
The uncertainty of the flux during PHOTONA 1 was finally calculated by error
propagation using σgradient and σu∗ (for
further details of the calculation of σu∗, see Stella
et al., 2012). For PHOTONA 2 σu∗ was calculated from 5 min data of u∗ (n= 6). For PHOTONA 3 the uncertainty of the flux
was not calculated, as only σgradient was available.
Influence of the roughness sublayer
The flux gradient similarity is not valid inside the roughness sublayer
(RSL), which ranges from the canopy top to around 2 times the canopy
height (Cellier and Brunet, 1992). In the present study, the flux divergence
caused by the RSL was analysed using the methods of Cellier and Brunet (1992) and Graefe (2004). However, the influence of the RSL during both
canopy campaigns was only of minor importance and therefore neglected for
further interpretation of the flux data.
Dealing with chemical reactions in the gas phase
The aerodynamic gradient method is strictly valid only for non-reactive
trace gases. However, in the present study, the photolysis and the production
of HONO (e.g. by NO + OH) in the gas phase below the measurement heights may
create artificial fluxes that need to be corrected for.
To check for chemical reactions during turbulent transport the so-called
Damköhler number (Da) has been used: Da =τtrans/τchem.
It compares the chemical reaction timescale (τchem,
see Eq. 4) with the transport timescale (τtransp, see Eq. 5)
to identify periods when chemical reaction may generate flux divergence. To
calculate the chemical timescale of HONO only its photolysis was taken into
account, which is the dominant destruction path of HONO during daytime:
τchem=1J(HONO).
In contrast, loss or production rates by the reactions HONO + OH and
NO + OH are typically more than an order of magnitude lower than the HONO
photolysis even when considering a typical maximum OH concentration of
5 × 106 cm-3 during daytime. For the correction of
chemistry, the transport timescale depends on (a) the location of the HONO
source and (b) the region where HONO chemistry starts to become important.
Since photolysis is diminished in the canopy due to shadowing by the leaves,
we only consider the transfer time between the canopy exchange height
(defined as d+z0′) and the reference height (zref). This leads to
the following definition of τtrans:
τtrans=1Ra⋅(zref-d-z0)+Rb⋅(z0-z0′),
where Ra is the aerodynamic resistance for transfer between
d+z0 (z0 is the roughness height for momentum) and zref.
Rb is the canopy boundary layer resistance for HONO accounting for the
transfer between the roughness height d+z0 and the canopy exchange
height located at d+z0′ (z0′ is the roughness height for the
scalar). Ra and Rb were estimated by
Ra=uzref(u∗)2,Rb=1.45⋅Re0.24⋅Sc0.8u∗.
Here uzref is the wind speed at zref, Re is the canopy Reynolds number,
Re = u∗⋅z0/νa, and Sc the Schmidt number,
Sc = νa/D(HONO), where νa is the kinematic viscosity of air
and D(HONO) the diffusivity of HONO in air (Garland et al., 1977). During PHOTONA 3,
for which the direct measured photolysis rate J(HONO) was available, the transport
time during daytime was typically of the order of a minute and much smaller
than the chemical lifetime of HONO of τchem≥ 10 min. Thus,
the influence of the photolytic loss to the overall HONO flux was always
below 10 % (Da < 0.1) and we considered a refinement of the
analysis by the stability corrections on Ra (see Stella et al., 2012) of
less importance. As no production terms for HONO were considered for the
calculation of the flux divergence, the influence of photolysis gives even
only an upper limit for the flux divergence. Similar results were obtained
for PHOTONA 1 and 2 using calculated J(HONO) data. For further analysis, errors by
chemical reactions were neglected, which will, however, not significantly
influence the interpretation of potential precursors and driving forces of
the HONO flux.
Time series of mixing ratios of the main species HONO,
NO2, NO and O3 (upper level
data), ΔHONO (including 2σ precision
errors), meteorological parameters and
J(NO2) during the
PHOTONA project (left to right: PHOTONA 1, 2 and 3). WD: wind direction.
Footprint area
The field site in Grignon is quite homogenous although with a slight slope
and some building and trees around 600 m to the west. To decide if the flux
is influenced by surfaces outside this area that may disturb homogeneity, a
footprint analysis, as described by Neftel et al. (2008), has been performed
using the model ART Footprint Tool version 1.0, which is available from
http://www.agroscope.admin.ch/art-footprint-tool. The influence
of the field site was > 92 % (median of all campaigns) and
very comparable to other flux measurements at this location. For example,
Loubet et al. (2011) stated that up to 93 % of the field was inside the
mast footprint (3.17 m height) during summer/spring campaigns.
Results and discussion
General observations
Main standard meteorological measurements and mixing ratios from all
campaigns are presented in Fig. 3. During PHOTONA 1, maximum daytime
temperatures ranged from 21 to 28 ∘C and daytime relative
humidities were around 30 to 40 %. Minimum night-time temperatures ranged
from 10 to 17 ∘C with relative humidities between 60 and 95 %.
Dry conditions generally prevailed with only one moderate rain event on the
24 August (1.2 mm h-1). During most of the campaign the wind
came from the south-west with only a short period of 2 days dominated by
north-easterly winds starting on 22 August. Maximum wind speeds
during the day varied from 2 to 8 m s-1, with friction velocities up to
0.46 m s-1. Minimum HONO mixing ratios during the day varied from 5 to
120 pptV with maximum morning peaks of up to 700 pptV. Minimum daytime
mixing ratios of NO2 were around 1 ppbV with maximum mixing ratios
during the morning hours of up to 25 ppbV. Nocturnal mixing ratios of
NO2 varied typically from < 1 ppbV in the middle of the night
and up to 25 ppbV in the late evening.
Maximum daytime temperatures during PHOTONA 2 varied from 10 to
19 ∘C with relative humidities in the range 31–65 %. Minimum
nocturnal temperatures, reached before early morning, ranged from 2.4 to
7.7 ∘C with relative humidities between 76 and 93 %. Moderate
rain events (up to 1.2 mm h-1) occurred during the beginning (8
April) and in the middle (13 April) of the campaign. Maximum wind
speeds during the day varied from 3 to 9 m s-1, i.e. comparable to
PHOTONA 1. However, the canopy generated turbulence, which is expressed in
higher friction velocities with maximum values up to 0.6 m s-1. HONO
mixing ratios during early afternoon varied from 60 to 450 pptV and reached
up to 900 pptV during the night. For NO2 one short emission event with
air mass coming from the Paris region occurred during the first day with
mixing ratios of up to 60 ppbV. Thereafter, the mixing ratios varied from
< 20 ppbV during morning hours to around 1 ppbv in the afternoon.
Diurnal average data of the HONO flux,
F(HONO), and its potential precursors and driving
factors HONO, u∗, Tsoil,
c(NO2) and
J(NO2) during the
three PHOTONA campaigns.
For the first half of the PHOTONA 3 campaign, maximum daytime temperatures
reached values up to 31 ∘C but decreased to around 16 ∘C during the second half of the campaign. A similar trend was observed for
the nocturnal temperatures with minimum values in the range 16 to
22 ∘C in the beginning and 6 to 12 ∘C at the end of
the campaign. Minimum relative humidities in the afternoon ranged from 40 to
80 % and maximum humidities during early morning were in the range of 80 to
97 %. Light to strong rain events (0.1–8 mm h-1) occurred on
21, 22, 26 and 27 August and led to decreases
in the HONO mixing ratios, possibly caused by the high effective solubility
of HONO in water. Friction velocities reached values up to 0.6 m s-1
and were comparable to those of PHOTONA 2. HONO mixing ratios reached values
up to 600 pptV during the early morning and decreased to around 20 pptV in
the afternoon. Maximum NO2 mixing ratios with values up to 30 ppbV were
reached in the night or during early morning hours and minimum mixing ratios
down to < 1 ppbV were observed in the afternoon.
Diurnal average HONO flux
The HONO flux showed similar profiles in the summer campaigns PHOTONA 1 and
3 but was different in the spring campaign PHOTONA 2 (see Fig. 4). Minimum
emissions, or even depositions (PHOTONA 2), occurred at night and emissions
were observed during daytime with a morning peak at around 08:00 UTC
(coordinated universal time). During daytime of the summer campaigns
(PHOTONA 1 and 3) continuously decreasing HONO fluxes were observed after
the morning peak, whereas during the spring campaign the flux rapidly
decreased after a strong morning peak and stayed more or less constant
throughout the rest of the day. For PHOTONA 1 and 3 the HONO flux then
decreased again to a minimum around midnight or slightly earlier. The
magnitudes of the observed daytime fluxes in the range 0.1 to
2.3 ng N m-2 s-1 (0.05–1 × 1014 molec m-2 s-1, see Fig. 4) are comparable to
measurements of other studies in suburban/rural regions. Ren et al. (2011),
for example, found fluxes during daytime with a maximum around
1.4 ng N m-2 s-1 on average during CalNex 2010 in California and
Zhou et al. (2011) obtained maximum daytime HONO fluxes during noon and
early afternoon of around 2.7 ng N m-2 s-1 on average on the
PROPHET tower in Michigan. The observed morning peak is also in agreement
with another study, where the measurements were performed in and above a
forest canopy (He et al., 2006) and were explained by dew evaporation.
The range of daytime HONO fluxes measured in this study is also of the order
of magnitude of the laboratory derived “optimum HONO emission flux” by
biological processes for the soil collected from the Grignon field site,
which was 6.9 ng N m-2 s-1 (Oswald et al., 2013). In these
experiments optimum HONO emissions were derived during drying of the soil
surface in a dark chamber by flushing with completely dry air. The cited
maximum emission for the Grignon soil was obtained at a soil temperature of
25 ∘C and at a low soil humidity of around 10 % of the water
holding capacity (whc), which corresponds to a gravimetric soil water content
of 5.5 % (whc = 54.9 % in gravimetric humidity). Multiplying by the soil
density at the surface (1.3 ± 0.5 kg L-1) gives the corresponding
soil water volume content of 7.1 % much lower than those at the present
field site, where soil water contents and soil temperatures at 5 cm depth of
13.2 ± 0.4 % and 22.6 ± 9.7 ∘C in PHOTONA 1,
27.1 ± 2.0 % and 10.1 ± 4.2 ∘C in PHOTONA 2 and
27.7 ± 1 % and 18.4 ± 4.1 ∘C in PHOTONA 3 were
observed. According to the soil humidity and temperature response curves
reported by Oswald et al. (2013), biological emissions of HONO are expected
to be lower than 5 ng N m-2 s-1 in PHOTONA 1 and lower than
0.001 ng N m-2 s-1 in PHOTONA 2 and PHOTONA 3. Hence we expect
the biological source as evaluated by Oswald et al. (2013) to be negligible
in PHOTONA 2 and 3, while it could be comparable to the measured HONO flux
in PHOTONA 1.
Goodness of the weighted orthogonal regressions of hourly
average daytime data (06:00 to 20:00 UTC) of
F(HONO) against different variables for the three
PHOTONA campaigns. The numbers represent χ2/Q(R2) values for which lower
χ2 and higher
Q and R2
values indicate better correlations (for definition see Brauers and
Finlayson-Pitts, 1997). Bold numbers represent the strongest correlations
observed for each campaign.
J(NO2)
J(O1D)
Tsoil
u∗
J(NO2) ⋅ c(NO2)
PHOTONA 1
27.5/0.004
not measured
50.2/6 × 10-7
23.4/0.016
7.27/0.78
(0.47)
(0.22)
(0.41)
(0.79)
PHOTONA 2
12.1/0.28
not measured
9.33/0.50
5.66/0.84
12.4/0.26
(0.27)
(0.019)
(0.37)
(0.37)
PHOTONA 3
53.7/3 × 10-7
79.8/5 × 10-12
121/5 × 10-20
62.7/7 × 10-9
3.26/0.994
(0.38)
(0.17)
(0.03)
(0.20)
(0.85)
Correlation between fluxes and concentrations of HONO
When plotting the night-time data of the HONO flux against the HONO
concentration for PHOTONA 2 (and to lesser extent for PHOTONA 1) a
significant positive correlation is observed (PHOTONA 2: R2= 0.92;
PHOTONA 1: R2= 0.43) with negative HONO fluxes at HONO mixing ratios
< 0.43 and < 0.13 ppb for PHOTONA 2 and 1, respectively.
This observation indicates a significant impact of HONO deposition on the
net HONO flux and is in agreement with the observed negative net HONO fluxes
observed in the early morning of PHOTONA 2 (see Fig. 4). In contrast, for
the night-time data of PHOTONA 3 and the daytime data of all three campaigns
there was no significant correlation between the HONO flux and its
concentration. The missing daytime correlation offers support that deposition is
of minor importance compared to the more important HONO source terms during
daytime.
Correlation between fluxes of HONO and potential precursors
As the major aim of the present study was to explain the origin of HONO
sources during daytime, the following section concentrates on the flux and
its correlations with potential precursors using only data from 06:00 to
20:00 UTC. Campaign averaged HONO fluxes (see Fig. 4) were plotted against
different potential precursors and controlling parameters. Correlations of
the HONO flux with the product of the photolysis frequency and concentration
of NO2, J(NO2) ⋅ c(NO2), were observed for all campaigns (see
Table 1). Especially the HONO fluxes during PHOTONA 1 and 3 were well
explained by NO2 and UV-A light intensity expressed by J(NO2), which is
presented exemplarily for PHOTONA 1 in Fig. 5.
While a correlation between the daytime HONO flux and the product of
J(NO2) ⋅ c(NO2) was observed for all three campaigns, especially
during the two summer campaigns PHOTONA 1 and 3, an additional correlation
with the friction velocity was observed during the spring campaign PHOTONA
2 (see Table 1). Reasons for the different correlation results and the
different diurnal shapes of the HONO fluxes between the two summer and the
spring campaigns (see Fig. 4) are still not fully clear. A potential
explanation could be the higher influence of HONO deposition during the
colder spring campaign (see below) masking the correlation with the main
proposed source precursors NO2 and radiation. Since deposition fluxes
will depend on the turbulent vertical mixing this could explain the higher
correlation with the friction velocity. Alternatively, decoupling between
the regimes above and below a dense canopy will also depend on the vertical
turbulent mixing (Sörgel et al., 2011a) and may have influenced the HONO
flux from the soil source region to the measurements heights above the
canopy. Finally, stomatal uptake of HONO by the leaves of the triticale
canopy, especially during daytime (Schimang et al., 2006), may have caused
the lower daytime fluxes during PHOTONA 2 (see Fig. 4) compared to the
other campaigns.
The finding of a light- and NO2-dependent HONO flux is in good agreement
with the study of Ren et al. (2011), where daytime HONO fluxes above an
agricultural field also correlated well with the product of NO2
concentrations and incident solar radiation during the CalNex 2010 campaign.
Also the very weak correlation of the HONO flux with the NO2
concentration above a forest canopy at the PROPHET site (Zhou et al., 2011;
Zhang et al., 2012) can be attributed to an influence of the canopy.
Correlations of HONO with its precursors are expected to become worse when
measurements are carried out above high trees as at the PROPHET site, which
are able to fully decouple the ground surface from the air above the canopy
(Sörgel et al., 2011a; Foken et al., 2012). The results from the present
study are qualitatively also in good agreement with former studies in which
the daytime source of HONO was quantified using the PSS approach and in
which also a strong correlation of the daytime source with radiation and/or
NO2 was observed (Elshorbany et al., 2009; Sörgel et al., 2011b;
Villena et al., 2011; Wong et al., 2012; Lee et al., 2016; Meusel et al.,
2016). This observation may imply a mechanism of HONO formation by the
reduction of NO2 with organic photosensitizer materials like humic
acids as proposed in laboratory studies (George et al., 2005; Stemmler et
al., 2006, 2007; Han et al., 2016).
Another HONO source, microbiological formation of nitrite in the soil, as
proposed by Su et al. (2011) and Oswald et al. (2013), should strongly
depend on the soil temperature and the soil surface water content due to
the temperature dependence of the solubility of HONO in soil water and/or
the adsorption of HONO on the soil surface and the biological activity of
the soil. Here, the HONO fluxes are expected to increase with increasing
temperature and decreasing humidity. However, except for PHOTONA 2 the
correlations of the HONO flux were much weaker with the soil temperature
compared to those with J(NO2) and with the product
J(NO2) ⋅ c(NO2) (see Table 1). In addition, the HONO fluxes showed
no significant correlation with the soil water content, the relative
humidity of the air or its inverse. Also based on the observed diurnal shape
of the HONO flux, microbiological formation of nitrite/HONO on the soil
surfaces seems to be unlikely, since the highest fluxes would be expected at
low soil water content and high temperature, leading to a maximum of the
HONO flux in the early afternoon, when the soil surface is at its driest and
warmest due to irradiation from the sun. In contrast, the highest fluxes
were observed during the morning in the present study (see Fig. 4). And
finally, the expected optimum HONO fluxes were much lower in PHOTONA 3
compared to PHOTONA 1 due mainly to the different soil water contents (see
Sect. 3.2), while the measured fluxes were very comparable (see Fig. 4).
Thus, although the laboratory derived optimum HONO fluxes were in the same
range as those observed in the present field study during PHOTONA 1 (see
Sect. 3.2), the different diurnal shapes and seasonal variability of the
expected and measured HONO fluxes do not support the microbiological soil
mechanism proposed by Su et al. (2011) and Oswald et al. (2013) as a major
HONO source at the present field site. This result is in good agreement with
another recent field study in which the daytime HONO source could also not
be explained by a biological soil source but showed a strong correlation
with the radiation (Oswald et al., 2015), similar to that observed in the
present study. It should be stressed that in the Oswald et al. (2013) study
the experimental conditions were not representative for the present field
site. While in these laboratory studies the upper soil surface was flushed
by completely dry air, leading to optimum HONO emissions only at very dry
conditions, the relative humidity never decreased below 26, 31 and
43% in PHOTONA 1, 2 and 3, respectively. More work is desirable to
reconcile HONO field data with incubation experiments as performed by Oswald
et al. (2013). Finally, we cannot completely exclude this source here, as
we observed a small positive intercept in the correlation plots of the HONO
flux against J(NO2) ⋅ c(NO2) in all campaigns (see Fig. 5 for
PHOTONA 1). Since the biological soil source is expected to be light and
NO2 independent (Su et al., 2011; Oswald et al., 2013) this intercept
may reflect the magnitude of this source and/or other light-independent
sources. However, the small magnitude of the intercept compared to the
observed HONO fluxes, especially for PHOTONA 1 and 3, suggests that
light-independent sources are of minor importance during daytime.
Correlation of the diurnal HONO flux (06:00 to 20:00 UTC)
with the product
J(NO2) ⋅ c(NO2) during
PHOTONA 1 with a weighted orthogonal regression fit (Brauers and
Finlayson-Pitts, 1997).
The lack of information about nitrate surface concentrations during the
present study does not allow us to directly exclude a HNO3 photolysis
mechanism as proposed by Zhou et al. (2011), who observed a HONO flux that
is positively correlated with leaf surface nitrate loading and light
intensity. However, in the present study a better correlation of the HONO
flux with J(NO2) (near UV-A) of R2= 0.38 than with J(O1D) (UV-B) of
R2= 0.17 was observed for PHOTONA 3 for which a spectroradiometer
was used to measure both photolysis frequencies (see Table 1). Since
HNO3 photolysis is expected to be active mainly under short wavelength
UV radiation, while the photosensitized conversion of NO2 on humic acid
surfaces works well already in the visible and near UV-A (Stemmler et al.,
2006; Han et al., 2016), the latter mechanism seems to be a more likely HONO
source at the present field site compared to photolysis of adsorbed
HNO3. This is confirmed by the high correlation of F(HONO) with the product
J(NO2) ⋅ c(NO2) of R2= 0.85 (see Table 1). In addition, for a potential
nitrate photolysis source a maximum of the HONO flux would be expected in
the afternoon due to a number of contributing factors: (i) the main
HNO3 source during daytime is the reaction of NO2 with OH, (ii) the typical diurnal profiles of the OH concentration and (iii) the
subsequent deposition of gas-phase HNO3 onto ground surfaces. In
contrast, the campaign averaged HONO fluxes showed asymmetric diurnal
profiles with a maximum in the morning, which can be explained by the higher
NO2 morning levels compared to the afternoon (see Fig. 4). Finally,
in a recent laboratory study on the photolysis of adsorbed HNO3 only a
very low upper-limit photolysis frequency of J(HNO3→ HONO) = 2.4 × 10-7 s-1 (0∘ SZA, 50 % RH) was
determined (Laufs and Kleffmann, 2016), which is too low to explain any
significant HONO formation in the atmosphere.
Another recently discussed mechanism, the acid displacement of HONO by
deposition of strong acids (e.g. VandenBoer et al., 2015), also seems to be
unlikely for the present field site. This proposed source should maximize in
the afternoon because of the daytime formation of the main strong acid
HNO3 and its subsequent deposition on ground surfaces (see discussion
above and see Fig. 4c in VandenBoer et al., 2015). In contrast, for any
NO2-dependent photochemical source a maximum HONO flux during morning
hours is expected (see Ren et al., 2011, and Fig. 4c in VandenBoer et al.,
2015) since the highest NO2 concentrations occur in the morning and not
in the afternoon (see Fig. 4). Only if the majority
of the soil acidity results from night-time dry deposition of
N2O5, the higher morning fluxes of HONO might be explained by the
acid displacement mechanism. Here flux measurements of HNO3 and
N2O5 are necessary in the future. However, since we expect a
higher contribution of HNO3 uptake to the soil acidity, the asymmetric
HONO flux profile with higher values in the morning indicates that the acid
displacement is of less significance for the present field site (and also
for the data shown in Fig. 4c in VandenBoer et al., 2015).
Average diurnal HONO fluxes of all individual campaigns
as a function of the soil temperature. The black line presents the
regression fit using the exponential function
F(HONO) = exp(ΔsolH/RTsoil + C),
where R is the universal gas constant
(8.314 J mol1 K-1) and
Δsol H is the experimental enthalpy of solvation
(-41.2 kJ mol-1).
Comparison of all campaigns
In order to find parameters that control the HONO flux in a kind of manner
that is not visible using the individual campaign data, we tried to find
parameters that affect the HONO flux using the data from all three
campaigns. Figure 6 shows HONO fluxes during PHOTONA 1, 2 and 3 as a
function of the soil temperature. Although HONO fluxes of the individual
campaigns correlated better with J(NO2) ⋅ c(NO2) (see Table 1), an additional
positive correlation of all the data with the soil temperature is obvious.
With increasing soil temperature the net HONO flux increases, which may be
explained by a temperature-dependent adsorption/solubility process (Su et
al., 2011), which becomes more important at lower temperatures compared to
the HONO source reactions. In the present study, only net HONO fluxes were
quantified, which are controlled by typically smaller negative deposition
fluxes and stronger positive formation by heterogeneous processes on the
soil surface. When plotting the logarithm of the positive HONO fluxes
against the inverse temperature a formal activation enthalpy for HONO
formation of 41.2 kJ mol-1 can be derived (see Fig. 6). Assuming that
HONO formation by NO2 conversion on the soil surface is controlled by
the temperature-dependent HONO solubility in the soil water, this activation
enthalpy is in good agreement with the value of the enthalpy of solvation of
HONO in water of ΔsolH=-40.5 kJ mol-1 (Park and Lee,
1988). The different signs of the two enthalpies are explained by the
different reference points describing the same process, for which increasing
solubility leads to a decrease in the HONO flux.
In conclusion, positive daytime HONO fluxes are explained in the present
study by a NO2- and light-dependent source, i.e. by the photosensitized
conversion of NO2 on soil surfaces (Stemmler et al., 2006) additionally
controlled by the temperature-dependent HONO adsorption on the soil or its
solubility in soil water.
Parameterization of the HONO flux
The above results of the above correlation study were used to set up a
simple parameterization that describes the HONO flux for all campaigns. As
the strongest correlation was observed for the HONO flux with the product of
NO2 concentration with light intensity, a proposed photosensitized
HONO source (Stemmler et al., 2006) was parameterized by the term
A ⋅ J(NO2) ⋅ c(NO2) (see Eq. 8). To also describe the
night-time HONO flux, which would have been zero when considering only this
light-dependent source, an additional slower dark formation of HONO by
heterogeneous NO2 conversion on soil surfaces (e.g. Finlayson-Pitts et
al., 2003, or Arens et al., 2002) was introduced by using a second source
term B ⋅ c(NO2). Because of the observed temperature dependence of
the HONO fluxes, both sources were multiplied by a Boltzmann term, for which
the negative value of the experimental solvation/adsorption enthalpy of HONO
of -41.2 kJ mol-1 (see Fig. 6) was used.
The two proposed sources are in agreement with results from several field
and laboratory studies (Kleffmann, 2007) but would result in only positive
modelled HONO fluxes. However, during PHOTONA 2 also net HONO deposition was
observed in the early morning at the low soil temperatures of the spring
campaign (see Fig. 4). To describe this net HONO uptake on ground surfaces
an additional temperature-dependent HONO deposition term was included, i.e.
the product of the HONO concentration measured at the lower sampling height
with a temperature-dependent deposition velocity, ν(HONO)T. Finally,
since the magnitude of HONO sources and sinks is expected to positively
correlate with humidity (Finlayson-Pitts et al., 2003; Stemmler et al.,
2006; Han et al., 2016; Su et al., 2011), all variables were optimized for a
reference RH of 50 % and were scaled linearly with
humidity (RH / 50 %), leading to the final Eq. (8):
F(HONO)mod=A⋅J(NO2)⋅c(NO2)+B⋅c(NO2)⋅expΔsolHR⋅Tsoil-c(HONO)⋅ν(HONO)T⋅RH50%.
The constants A and B were adjusted to obtain (i) a slope of one, (ii) an
intercept of zero and (iii) a high correlation between modelled and measured
HONO fluxes (R2= 0.68), resulting in final values for A and B of
2.9 × 106 m and 2.0 × 10-3 m s-1,
respectively. When considering also for the Boltzmann and humidity terms,
the final value for A is in good agreement with the average experimental
value of 2.5 × 106 m determined from the correlation plots of
the three campaigns (see Sect. 3.3.2). This indicates that the photolytic
HONO source is mainly controlling the net HONO fluxes during daytime. The
second term B multiplied by the Boltzmann and humidity terms can be
described as the effective deposition velocity of NO2 to form HONO in
the dark on ground surfaces. The measured overall deposition velocity of
NO2 during PHOTONA 1 varied from 0.002 m s-1 during night time to
0.0055 m s-1 before noon (calculated from the diurnal average data of
the whole campaign; see Stella et al., 2011). Dividing the effective
deposition velocity for HONO formation in the dark (B multiplied by the
Boltzmann and humidity terms) by the overall measured deposition velocity of
NO2, resulted in campaign averaged ratios in the range 2.0 % (day)
to 4.4 % (night), i.e. only 2–4.4 % of the NO2 uptake on ground
surfaces leads to HONO production by the heterogeneous dark conversion of
NO2. This range of values is comparable with night-time observations of
Stutz et al. (2002), who calculated a conversion efficiency to form HONO
from NO2 deposition of 3 ± 1 %.
Diurnally averaged measured HONO fluxes in comparison
with modelled values (Eq. 8) during the three PHOTONA campaigns.
When comparing the two proposed sources, the dark conversion of NO2
contributed only ∼ 10 % to the HONO fluxes around noon,
while it was the only source during night time by definition. When
integrating over the whole day (24 h), the dark conversion contributed 23, 28 and 30 % to the total heterogeneous HONO production, while
the photochemical source was 3.3, 2.6 and 2.3 times larger during PHOTONA 1,
2 and 3, respectively. These results are in general agreement with former
field studies using the more simple PSS approach in which the photochemical
HONO source also dominates daytime production (Kleffmann, 2007, and
references therein), while the dark conversion of NO2 controls
the night-time build-up of HONO and the OH radical production in the early
morning after sunrise (e.g. Alicke et al., 2002).
To describe also the negative HONO fluxes during the PHOTONA 2 spring
campaign (see Fig. 4), the temperature-dependent effective HONO deposition
velocity (ν(HONO)T, see Eq. 8) was adjusted to values of 0.02 m s-1 at 0 ∘C decreasing exponentially to non-significant
values at 40 ∘C (ν(HONO)T= exp(23920/T-91.5)). The higher
end deposition velocity used here is in agreement with published upper-limit
values in the range 0.005 m s-1 (Stutz et al., 2002), 0.017 m s-1
(Harrison and Kitto, 1994; Trebs et al., 2006) and 0.06 m s-1 (Harrison
et al., 1996). Based on this model adjustment, HONO deposition became more
significant towards the end of the night, especially during PHOTONA 2, when
modelled deposition fluxes were up to 4 times larger compared to the
sources. In contrast, during daytime, deposition fluxes were less
significant and made up only a few percent at maximum compared to the source
reactions, in agreement with the missing correlation of the HONO flux with
its concentration during daytime (see Sect. 3.3).
The measured HONO fluxes were well described by Eq. (8) especially
during PHOTONA 1 and 3 (see Fig. 7). However, during PHOTONA 2, the
campaign with the triticale canopy, the daytime HONO fluxes were
overestimated by the model, which may be explained by additional stomatal
uptake of HONO by the leaves (Schimang et al., 2006) during transport of
HONO from the proposed soil surface source region to the sampling positions
above the dense canopy. In addition, the sharp measured morning peak of the
HONO flux during PHOTONA 2 is also not well represented by the model. This
morning peak may be explained by dew evaporation of accumulated nitrite
(formed e.g. by dark reactions of NO2 or uptake of HONO) from
vegetation surfaces when the temperature increased in the morning, which is
in agreement with results from other field studies (Rubio et al., 2002; He
et al., 2006).