Measurements of nitrogen oxides and ozone fluxes by eddy covariance at a meadow : evide ce for an internal leaf resistance to NO 2

Nitrogen dioxide (NO 2 ) plays an important role in atmospheric pollution, in particular for tropospheric ozone production. However, the removal processes involved in NO 2 deposition to terrestrial ecosystems are still the subject of ongoing discussion. This study reports NO 2 flux measurements made over a meadow using the eddy covariance method. The measured NO 2 deposition fluxes during daytime were about a factor of two lower than a priori calculated fluxes using the Surfatm model without taking into account an internal (also called mesophyllic or sub-stomatal) resistance. Neither an underestimation of the measured NO 2 deposition flux due to chemical divergence or an in-canopy NO 2 source nor an underestimation of the resistances used to model the NO 2 deposition explained the large difference between measured and modelled NO 2 fluxes. Thus, only the existence of the internal resistance could account for this large discrepancy between model and measurements. The median internal resistance was estimated to be 300 s m −1 during daytime, but exhibited a large variability (100–800 s m −1 ). In comparison, the stomatal resistance was only around 100 s m −1 during daytime. Hence, the internal resistance accounted for 50–90% of the total leaf resistance to NO 2 . This study presents the first clear evidence and quantification of the internal resistance using the eddy covariance method; i.e. plant functioning was not affected by changes of microclimatological (turbulent) conditions that typically occur when using enclosure methods.


Introduction
Nitrogen oxides (NO x , the sum of nitric oxide, NO, and nitrogen dioxide, NO 2 ) play an important role in the photochemistry of the atmosphere.By controlling the levels of key radical species such as the hydroxyl radical (OH), NO x are key compounds that influence the oxidative capacity of the atmosphere.In addition, NO x are closely linked with tropospheric ozone (O 3 ) production.NO is rapidly oxidized to NO 2 , which is photo-dissociated to NO and groundstate atomic oxygen (O( 3 P)) that reacts with O 2 to form O 3 (Crutzen, 1970(Crutzen, , 1979)).O 3 is a well-known greenhouse gas responsible for positive radiative forcing, i.e. contributing to global warming, representing 25 % of the net radiative forcing attributed to human activities since the beginning of the industrial era (Forster et al., 2007).Moreover, due to its oxidative capacities, O 3 is also a harmful pollutant responsible for damages to materials (Almeida et al., 2000;Boyce et al., 2001), human health (Levy et al., 2005;Hazucha and Lefohn, 2007) and plants (Paoletti, 2005;Ainsworth, 2008).In natural environments, O 3 may lead to biodiversity losses, while in agro-ecosystems, it induces crop yield losses (Hillstrom and Lindroth, 2008;Avnery et al., 2011a, b;Payne et al., 2011).

P. Stella et al.: Measurements of nitrogen oxides and ozone fluxes by eddy covariance at a meadow
NO x is also responsible for the production of nitric acid and organic nitrates, both acid rain and aerosol precursors (Crutzen, 1983).In addition, it influences the formation of nitrous acid (HONO), which is an important precursor for OH radicals in the atmosphere.
The important impacts of NO, NO 2 and O 3 on both atmospheric chemistry and environmental pollution require establishing the atmospheric budgets of these gases.Therefore, it is necessary (i) to identify the different sources and sinks of NO, NO 2 and O 3 , and (ii) to understand the processes governing the exchange of these compounds between the atmosphere and the biosphere.To achieve this goal, several studies were carried out in the last decades over various ecosystems to identify the underlying processes controlling the biosphere-atmosphere exchanges of NO (e.g.Meixner, 1994;Meixner et al., 1997;Ludwig et al., 2001;Laville et al., 2009;Bargsten et al., 2010), NO 2 (e.g.Meixner, 1994;Eugster and Hesterberg, 1996;Hereid and Monson, 2001;Chaparro-Suarez et al., 2011;Breuninger et al., 2012), and O 3 (e.g.Zhang et al., 2002;Rummel et al., 2007;Stella et al., 2011a).
It is now well established that soil biogenic NO emission depends on several factors, such as the amount of soil moisture, soil temperature, and soil nitrogen (Remde et al., 1989;Remde and Conrad, 1991;Ludwig et al., 2001;Laville et al., 2009).Ozone is deposited to terrestrial ecosystems through dry deposition (Fowler et al., 2009).The different O 3 deposition pathways are well identified and the variables controlling each pathway are well understood: the cuticular and soil ozone deposition pathways are governed by canopy structure (canopy height, leaf area index) and relative humidity at the leaf and soil surface (Zhang et al., 2002;Altimir et al., 2006;Lamaud et al., 2009;Stella et al., 2011a), while stomatal ozone flux is controlled by climatic variables responsible for stomata opening such as radiation, temperature and vapour pressure deficit (Emberson et al., 2000;Gerosa et al., 2004).
However, the processes governing the NO 2 exchange between the atmosphere and the biosphere still remain unclear.While it is well recognized that NO 2 is mainly deposited through stomata, with the cuticular and soil fluxes being insignificant deposition pathways for NO 2 (Rondón et al., 1993;Segschneider et al., 1995;Pilegaard et al., 1998;Geßler et al., 2000;Ludwig et al., 2001), the existence of an internal resistance (also called mesophyllic or sub-stomatal resistance in previous studies) limiting NO 2 stomatal uptake is still under discussion.Previous studies reported contrasting results: Segschneider et al. (1995) and Geßler et al. (2000Geßler et al. ( , 2002) ) did not find an internal resistance for sunflower, beech and spruce, whereas the results obtained by Sparks et al. (2001) and Teklemariam and Sparks (2006) for herbaceous plant species and tropical wet forest suggested its existence.In addition, the importance of this internal resistance for the overall NO 2 sink is not well established.Current estimates range from 3 to 60 % of the total resistance to NO 2 uptake (Johansson, 1987;Gut et al., 2002;Chaparro-Suarez et al., 2011).Nevertheless, all the previous studies explored the processes of NO 2 exchange using enclosure (chamber) methods under field or controlled conditions, which may affect the microclimatological conditions around the plant leaves.This issue is of particular concern since the biochemical processes probably responsible for the internal resistance are linked with leaf functioning (Eller and Sparks, 2006;Hu and Sun, 2010).In addition, the aerodynamic resistance and the quasi-laminar boundary layer resistance above the plant leaves may be modified when applying enclosure methods.
In this study we present results of the SALSA campaign (SALSA: German acronym for "contribution of nitrous acid (HONO) to the atmospheric OH budget"; for details see Mayer et al., 2008).Turbulent fluxes of NO, NO 2 and O 3 were measured at a meadow below the Meteorological Observatory Hohenpeissenberg (MOHp) using the eddy covariance method.These measurements were accompanied by a comprehensive micrometeorological setup involving vertical profiles of trace gases and temperature as well as by eddy covariance measurements of carbon dioxide (CO 2 ) and water vapour fluxes.In the present work, (i) the influence of chemical divergence was estimated above and within the canopy, (ii) the existence of an NO 2 compensation point mixing ratio was explored, (iii) the impact of the soil resistance to modelled NO 2 deposition was discussed and (iv) the internal resistance for NO 2 was quantified in order to understand the processes governing the NO 2 exchange.

Site description
The field study was made at a meadow in the complex landscape around Hohenpeißenberg (southern Germany) within the framework of the SALSA campaign (see Mayer et al., 2008;Trebs et al., 2009).The site consists in a managed and fertilized meadow located at the gentle lower (743 m a.s.l.) WSW slope (3-4 • ) of the mountain Hoher Peißenberg (summit 988 m a.s.l.), directly west of the village Hohenpeißenberg in Bavaria, southern Germany (coordinates: 47 • 48 N, 11 • 02 E).The surrounding pre-alpine landscape is characterized by its glacially shaped, hilly relief and a patchy land use dominated by the alternation of cattle pastures, meadows, mainly coniferous forests and rural settlements.The meadow is growing on clay-rich soil that can be classified as gley-colluvium with very small patches of marsh soil.Furthermore, it was characterized by its relatively low plant biodiversity and consisted mainly of perennial ryegrass (Lolium perenne L.), ribwort (Plantago lanceolata L.), dandelion (Taraxacum officinale), red clover (Trifolium pratense L.), white clover (Trifolium repens L.), common cow parsnip (Heracleum sphondylium L.), sour dock (Rumex acetosa L.), daisy (Bellis perennis L.), and cow parsley (Anthriscus sylvestris (L.) Hoffm.).
The experiment was carried out from 29 August to 20 September 2005.The meadow was mown just before the instrument setup.The canopy height (h c ) and leaf area index (LAI) increased from 15 cm and 2.9 m 2 m −2 (at the beginning of the campaign) to 25 cm and 4.9 m 2 m −2 at the end of the experiment, respectively.The roughness length (z 0 = 0.1 h c ) ranged from 1.5 to 2.5 cm and the displacement height (d = 0.7 h c ) varied between 10.5 and 17.5 cm.These values were confirmed by estimates of z 0 and d from flux and profile records for 10-15 September.
The setup consisted of five measurement stations, (all located in an area of 400 m 2 , with a distance of 20-30 m to each other).The stations recorded meteorological conditions ("MET 1" and "MET 2" from the Bayreuth University (UBT) and the Max Planck Institute for Chemistry (MPIC), respectively), mixing ratio profiles ("PROFILE" from the MPIC) and turbulent fluxes ("EC 1" and "EC 2" from the UBT and the MPIC, respectively) (see Table 1).The detailed measurement setups are described in Table 1 and the following sections.

Meteorological measurements
The following standard meteorological variables (and vertical profiles) were recorded: global radiation (G r ) and net radiation, relative humidity (RH), air temperature (T a ), wind speed (u) and direction, and rainfall.The photolysis rate of NO 2 (j NO 2 ), soil temperature (T soil ) and soil water content (SWC) were also measured (for details see Table 1).

Trace gas profile measurements
Profile measurements of NO, O 3 , and NO 2 mixing ratios were made in order to investigate the chemistry of the NO-O 3 -NO 2 triad above and within the canopy.The profile system consisted of six measurement levels: one inside the canopy (0.05 m above ground level), one at the canopy top (first in 0.20 m, later moved to 0.28 m), and four above the canopy (0.50, 1.00, 1.65 and 3.00 m).The NO, O 3 , and NO 2 analysers were located in an air-conditioned container about 60 m northeast from the air inlets.The profile system was described previously by Mayer et al. (2011).Briefly, air samples from all heights were analysed by the same analyser consecutively and the levels were switched automatically by a valve system directly in front of a Teflon ® diaphragm pump.The length of the opaque inlet lines made of PFA (perfluoroalkoxy copolymer) ranged from 62 to 65 m (depending on the sampling height).All non-active tubes were continuously flushed by a bypass pump.To avoid condensation of water vapour inside the tubes, they were insulated and heated to a few degrees above ambient temperature.Pressure and temperature in the tubes were monitored continuously.The individual heights were sampled with different frequencies: ambient air from the inlet levels at 0.50 and 1.65 m were sampled ten times, other levels five times per 60 min (with each interval consisting of three individually recorded 30 s subintervals).Data from the first 30 s interval at each level were discarded to take into account the equilibration time of tubing and analysers.Measured mixing ratios were corrected for the gas-phase chemistry during the residence time of the air inside the sampling system according to Beier and Schneewind (1991).
NO was measured by red-filtered detection of chemiluminescence -generated by the NO + O 3 reaction -with a CLD 780TR (EcoPhysics, Switzerland).Excess O 3 was frequently added in the pre-reaction chamber to account for interference of other trace gases.For the conversion of NO 2 to detectable NO, photolysis is the most specific technique (Kley and Mc-Farland, 1980;Ridley et al., 1988).Thus, NO 2 in ambient air was photolytically converted to NO by directing every air sample air through a blue light converter (BLC, Droplet Measurement Technologies Inc.).Here, the light source was an UV diode array, which emits radiation within a very narrow spectral band (385-405 nm), making the NO-to-NO 2 conversion more specific and the conversion efficiency more stable in time than conventional converters based on photolysis of a broad spectral continuum (Pollack et al., 2011).The NO 2 mixing ratio can be determined from the difference between the NO mixing ratios measured with BLC and bypassing the BLC, respectively.The NO analyser was calibrated by diluting a certified NO standard gas (5.0 ppm, Air Liquide).The detection limit of the CLD 780TR was 90 ppt (3σ -definition).The efficiency of the photolytic conversion of NO 2 to NO was determined by a back titration procedure involving the reaction of O 3 with NO using a gas-phase titration system (Dynamic Gas Calibrator 146 C, Thermo Environmental Instruments Inc., USA).Conversion efficiencies were about 33 %.Ozone mixing ratios of the ambient air samples were measured by an UV absorption instrument (49 C, Thermo Environment, USA).

Eddy covariance measurements
Eddy covariance (EC) has been extensively used during the last decades to estimate turbulent fluxes of momentum, heat and (non-reactive) trace gases (Running et al., 1999;Aubinet et al., 2000;Baldocchi et al., 2001;Dolman et al., 2006;Skiba et al., 2009).It is a direct measurement method to determine the exchange of mass and energy between the atmosphere and terrestrial surfaces without application of any empirical constant.The theoretical background for the eddy covariance can be found in the existing literature (e.g.Foken, 2008;Foken et al., 2012a;Aubinet et al., 2012) and will not be detailed here.
The turbulent fluxes of momentum (τ ), sensible (H ) and latent (LE) heat, CO 2 , NO, NO 2 and O 3 were measured by two EC stations (Table 1 ).The NO detection principle of the CLD 790SR-2 is identical to that of the CLD 780TR described above.However, the sensitivity is a factor of 10 higher than that of the CLD 780TR, and due to the presence of two channels the concentrations of NO and NO 2 can be measured simultaneously with high time resolution (see Hosaynali Beygi et al., 2011).The accuracy of the CLD790SR-2 is about 5 % and the NO detection limit for a one-second integration time is 10 ppt (3σ -definition).The instrument was also located in the air-conditioned container, about 60 m NE from the sonic anemometer.The trace gas inlets were fixed 33 cm below the sound path of the anemometer without horizontal separation at a three-pod mast.Air was sampled through two heated and opaque PFA tubes with a length of 63 m and an inner diameter of 4.4 mm.While the first sample line and CLD channel was used for measuring NO, a BLC was placed just behind the sample inlet of the second channel in a ventilated housing mounted at a boom of the measurement mast.Despite the low volume of the BLC (17 mL), conversion efficiencies γ for NO 2 to NO of around 41 % were achieved.Consequently this channel detected a partial NO x signal (denoted here as NO * x ) corresponding to Flow restrictors for both channels of the CLD790SR-2 were mounted into the tubing closely after the corresponding inlets (after the BLC in the second channel) in order to achieve short residence times of the air samples inside the tubing (9 ± 0.4 s and 13 ± 0.4 s for NO and NO 2 , respectively) and fully turbulent conditions.The EC fluxes for the two analyser channels were first calculated independently and the NO 2 flux was then determined as Simultaneously, eddy covariance fluxes of O 3 were measured with a surface chemiluminescence instrument (Table 1) (Güsten et al., 1992;Güsten and Heinrich, 1996), which was mounted on the three-pod mast with its inlet mounted directly next to that of NO and NO 2 .
The 5 Hz signals of both CLD790SR-2 channels, referenced to the frequently calibrated NO and NO 2 measurements at 1.65 m from the trace gas profile system, were used for the final calculation of NO and NO 2 fluxes for 30 min time intervals.The O 3 flux calibration was done according to Müller et al. (2010).The quality of the derived fluxes was evaluated with the quality assessment schemes of Foken and Wichura (1996) (see also Foken et al., 2004), which validate the development state of turbulence by comparing the measured integral turbulence characteristics.Flux calculations included despiking of scalar time series (Vickers and Mahrt, 1997), planar fit coordinate rotation (Wilczak et al., 2001), linear detrending, correction of the time lag induced by the 63 m inlet tube, and correction for flux losses due to the attenuation of high-frequency contributions according to Spirig et al. (2005) based on ogive analysis (Oncley, 1989;Desjardins et al., 1989).The high-frequency losses were typically 12-20 % for NO, 16-25 % for NO 2 and 6-8 % for O 3 .Since pressure and temperature were held constant by the instruments and the effect of water vapour fluctuations was negligible, corrections for density fluctuations (WPL corrections, Webb et al., 1980) were not necessary for NO, NO 2 and O 3 .

Soil NO emission from laboratory
A composite soil sample (0-5 cm depth) was taken from the Hohenpeißenberg meadow site at the end of September 2005, and biogenic NO emission of the meadow soil was subsequently quantified in the soil laboratory of MPIC.Applying a method which is described in full detail by Feig et al. (2008) and Bargsten et al. (2010), sub-samples (80 g) of the composite soil sample were sieved through a 2 mm mesh and were incubated (at soil temperatures of 15 and 25 • C) and fumigated (with zero and 58 ppb NO) over the full range of 0.05 to 0.6 gravimetric soil moisture (in steps of 0.002).These laboratory studies resulted in the determination of the so-called net potential soil NO flux as function of soil temperature and moisture.From that, the actual surface net NO flux of the meadow soil is calculated using soil temperature (2 cm depth) and soil moisture (5 cm depth) data obtained by continuous measurements at the meadow site during the field experiment.

Resistance model parameterizations
The transfer of heat and trace gases can be assimilated into a resistance network with analogy to Ohm's law (Wesely, 1989;Wesely and Hicks, 2000).It includes the turbulent resistance above (R a ) and within (R ac ) the canopy, the quasilaminar boundary layer (R b ), the stomatal (R s ) and internal (R int ) resistances, the cuticular resistance (R cut ) and the soil resistance (R soil ).
In order to investigate the processes governing the exchanges of NO 2 and O 3 , we used the Surfatm model developed to simulate exchanges of heat and pollutant between the atmosphere and the vegetation (Personne et al., 2009;Stella et al., 2011b).It is a multi-resistance soil-vegetationatmosphere transfer (SVAT) model which couples (i) a trace gas exchange model and (ii) an energy budget model allowing to estimate the temperature and humidity of the leaves and of the soil to calculate the resistances to trace gas exchange.It comprises one vegetation layer and one soil layer.This model was initially developed to simulate the ammonia exchange, it was validated over grasslands by Personne et al. (2009), and it was recently adapted to estimate O 3 deposition to several maize crops by Stella et al. (2011b).In the following, we will only focus on the specific resistances to NO 2 and O 3 deposition.However, more details and explanations concerning the resistive scheme and the resistance parameterizations can be found in Personne et al. (2009) and Stella et al. (2011b).
The resistive scheme for NO 2 and O 3 deposition is shown in Fig. 1.Turbulent resistances above and within the canopy are identical for both NO 2 and O 3 , and were expressed as where k (= 0.4) is the von Kármán constant; z ref is the reference height; d is the displacement height; z 0T and z 0M are the canopy roughness length for temperature and momentum, respectively; z 0s (= 0.02 m; Personne et al., 2009) igure 1: Resistive scheme used in the Surfatm model for pollutant exchange.χ is the gas R cut and R soil are aerodynamic resistance above the canopy, aerodynamic resistance within the canopy, leaf quasi-laminar boundary layer resistance, soil quasi-laminar boundary layer resistance, stomatal resistance, internal resistance, cuticular resistance and soil resistance, respectively.Indexes i, z ref , z 0 , z 0 , z 0s and "surf" indicate the gas considered, the reference height, the canopy roughness height for momentum, the canopy roughness height for scalar, the soil roughness height for momentum, and the soil surface, respectively.
inside the canopy (Raupach et al., 1996); K M (h c ) is the eddy diffusivity at the canopy height; and M (( and H ((z ref − d)/L) are dimensionless stability correction functions for momentum and heat, respectively (Dyer and Hicks, 1970).
The canopy (R bl ) and soil (R bs ) quasi-laminar boundary layer resistances depend on the trace gas i considered and were expressed following Shuttelworth and Wallace (1985) and Choudhury andMonteith (1988), andHicks et al. (1987), respectively, as where a is a coefficient equal to 0.01 s m −1/2 (Choudhury and Monteith, 1988) = 1.19 for NO 2 ; Erisman et al., 1994); and u * ground is the friction velocity near the soil surface calculated following Loubet et al. (2006) as where u * is the friction velocity above the canopy.The stomatal resistance was not modelled but used as input.It was inferred from water vapour flux measurements by inverting the Penman-Monteith equation (Monteith, 1981): where E is the water vapour flux (kg m −2 s −1 ), δ w the water vapour density saturation deficit (kg m −3 ), β the Bowen ratio, s the slope of the saturation curve (K −1 ) and γ the psychrometric constant (K −1 ).However, R s PM can be defined as the stomatal resistance if E represents plant transpiration only; i.e. the influence of soil evaporation and evaporation of liquid water (rain, dew) that may be present at the canopy surface has to be excluded.Thus, our estimation of stomatal resistance was corrected for water evaporation as proposed by Lamaud et al. (2009): for dry conditions (RH < 60 %, for which liquid water at the leaf surface is considered to be completely evaporated) R s PM was plotted against gross primary production (GPP, estimated on a daily basis following Kowalski et al., 2003Kowalski et al., , 2004)).The corrected stomatal resistance (R s ) for all humidity conditions is then given by where α (= 7465) and λ (= −1.6) are coefficients given by the regression between R s PM and GPP under dry conditions.The soil and cuticular resistances to O 3 deposition were expressed following Stella et al. (2011a, b) as where R soil min (= 21.15 s m −1 ) is the soil resistance without water adsorbed at the soil surface (i.e. at RH surf = 0 %), k soil (= 0.024) is an empirical coefficient of the exponential function, R cut max (= 5000/LAI) is the maximal cuticular resistance calculated according to Massman (2004), RH 0 (= 60 %) is a threshold value of the relative humidity, k cut (= 0.045) is an empirical coefficient of the exponential function taken from Lamaud et al. (2009), and RH surf and RH z 0 are the relative humidity at the soil and leaf surface, respectively, calculated by the energy balance model.Concerning the NO 2 cuticular resistance, several studies have shown that this deposition pathway did not contribute significantly to NO 2 deposition and could be neglected (Rondón et al., 1993;Segschneider et al., 1995;Gut et al., 2002).Consequently, R NO 2 cut was set to 9999 s m −1 .Since an empirical parameterization for the soil resistance to NO 2 deposition is currently not available, a constant value (R NO 2 soil = 340s m −1 ) reported by Gut et al. (2002) for a soil in the Amazonian rain forest was used.
Finally, many trace gases entering into plants through stomata can react with compounds in the sub-stomatal cavity and the mesophyll.For O 3 , there is evidence that R int is usually zero (Erisman et al., 1994).However, for NO 2 there is currently no consensus concerning the existence of an internal resistance, and the uncertainty of the magnitude of its contribution to the overall surface resistance is large.Due to this insufficient knowledge, R int was also set to zero for NO 2 in the "a priori" model parameterization.
The total deposition flux of the scalar i (F i ) is the sum of deposition flux to the soil (F i soil ) and the deposition flux to the vegetation (F i veg ): In analogy to Ohm's law and following the resistive scheme of the Surfatm model (Fig. 1), total, vegetation and soil fluxes can be expressed as The deposition flux to soil can also be expressed as

Chemical reactions and transport times
In contrast to inert gases such as CO 2 and H 2 O, the fluxes of NO, NO 2 and O 3 could be subject to chemical reactions leading to non-constant fluxes with height (vertical flux divergence).According to Remde et al. (1993) and Warneck (2000), the main gas-phase reactions for the NO-O 3 -NO 2 triad are where k r is the rate constant of R1 (Atkinson et al., 2004) and j NO2 is the photolysis frequency for R2.
The chemical reaction time for the NO-O 3 -NO 2 triad (τ chem in s) gives the characteristic timescale of the set of R1 and R2.It was estimated following the approach of Lenschow (1982): In addition, the characteristic chemical depletion times for NO, O 3 and NO 2 were calculated according to De Arellano and Duynkerke (1992): The comparison of characteristic chemical reaction times with characteristic turbulent transport times indicates whether or not there is a significant vertical divergence of the turbulent flux of reactive trace gases.The transport time (τ trans in s) in one layer (i.e.above the canopy, between the measurement height and the canopy top, or within the canopy) can be expressed as the aerodynamic resistance through each layer multiplied by the layer thickness (Garland, 1977): The ratio between τ trans and τ chem is defined as the Damköhler number (DA) (Damköhler, 1940): According to Damköhler (1940), the divergence of a reactive trace gas flux is negligible if DA 1 (conventionally DA ≤ 0.1), i.e. the turbulent transport is much faster than chemical reactions, and consequently the reactive trace gas can be considered as a (quasi-)passive tracer.For DA > 0.1 measured reactive trace gas fluxes have to be corrected for the influence of (fast) chemical reactions to obtain correct turbulent fluxes of the reactive trace gas.The measured NO 2 -O 3 -NO fluxes were corrected for chemical reactions occurring between the canopy top and the measurement height using the method proposed by Duyzer et al. (1995) Duyzer et al. (1995) demonstrated that the general form of the flux divergence is The factor a i is calculated for NO 2 , NO and O 3 as where ϕ X = ϕ NO = ϕ O 3 = ϕ NO 2 = ϕ H is the stability correction function for heat (Dyer and Hicks, 1970).As shown by Lenschow and Delany (1987), the flux divergence at higher levels approaches zero.The factor b i was calculated for NO 2 , NO and O 3 as b i = −a i ln(z ref ), assuming that at z ref = 2 m the flux divergence was zero.For each compound, the corrected flux (F i, corr ) is then approximated as 3 Results and discussion

Meteorological conditions and mixing ratios
During the experimental period, the median value of the mean diel course of global radiation, G r , reached its maximum of ∼ 700 Wm −2 at noon (Fig. 2a).The air temperature followed the same diel cycle (Fig. 2b) with median daytime maxima of 21 • C. Relative humidity, RH, decreased during the morning to reach its minimum of 65 % after noon (Fig. 2b).The meteorological conditions were different during the first half of the experiment (29 August to 9 September 2005) and the second half (10-20 September 2005).While the former period was sunny and warm and characterized by easterly flows, the latter was dominated by rainy, cold, and overcast conditions governed by westerly winds.This resulted in considerable variability of the meteorological conditions during the experiment: maximal G r and T a ranged between 200 and 800 W m −2 , and 15 and 25 • C, respectively, and minimal RH varied between 80 and 50 % (Fig. 2a and b).
Mean diel courses of NO 2 , NO and O 3 mixing ratios measured at 1.65 m above ground level (profile system) are shown in Fig. 2c.Median NO mixing ratios were close to zero during the major part of the experiment and slightly increased during the morning to about 1 ppb.These elevated NO values occurred when the NO 2 mixing ratio began to decrease due to photolysis.In addition, some NO was most likely advected from roads passing the site at a distance of 2 km NE from the experimental site.Highest mixing ratios of NO 2 were on average about 6 ppb during the early morning and 4 ppb during the late afternoon, but increased occasionally up to 8 ppb.During the rest of the day, NO 2 mixing ratios were around 2-3 ppb.The diel trend of NO 2 was linked with photochemistry: during sunrise, NO 2 photolysis led to the decrease in NO 2 mixing ratios, while during nighttime the absence of photolysis and the stable stratification induced an accumulation of NO 2 in the lower troposphere.O 3 mixing ratios exceeded NO and NO 2 mixing ratios and varied from 10 to 20 ppb during nighttime and from 40 to 60 ppb during daytime.During the morning, turbulent mixing in the planetary boundary layer led to entrainment of O 3 from the free troposphere (Stull, 1989).In addition, photochemical O 3 production (in the presence of NO x and volatile organic compounds) caused the increase of O 3 mixing ratios during the morning, reaching its maximum in the early afternoon.The O 3 removal by dry deposition processes and the reduced entrainment of O 3 from the free troposphere as a result of thermally stable stratification and low-wind conditions induced a decrease in O 3 mixing ratio during late afternoon and particularly during the night (cf.Coyle et al., 2002).Overall, NO 2 and O 3 mixing ratios were higher from 29 August 2005 to 9 September 2005 than from 10 to 20 September 2005.

Footprint analysis and measured fluxes
Since the three-pod mast, the laboratory container and some rural settlements were potentially distorting the flow in the north and in the eastern sector of our site, we performed a footprint analysis according to Göckede et al. (2004Göckede et al. ( , 2006)).Owing to the extended fetch in western, southern and southeastern directions, the major part of the fluxes measured by the EC systems originated from the experimental field, independently of the stability conditions (Fig. 3).However, the surrounding areas contributed to the total fluxes mainly in the NNW/NE sectors, due to (i) the limited fetch and (ii) the rural settlements disturbing the flow in these directions.In addition, the footprint area increased with atmospheric stability.In order to ensure that only those measured fluxes were used for subsequent analyses which orig-inated from the experimental field (and not from the surrounding areas), we considered only those 30 min flux data for which at least 95 % of the total footprint area could be attributed to the experimental field.
NO 2 and O 3 fluxes were directed downward; i.e.net deposition fluxes were observed (Fig. 2d).Both NO 2 and O 3 deposition fluxes were close to zero during nighttime and typically increased during the morning to their maximum.Maximum deposition fluxes of NO 2 occurred in the early morning and ranged on average from about −0.3 nmol m −2 s −1 to −0.6 nmol m −2 s −1 .The deposition fluxes of O 3 were about 10 to 20 times higher than NO 2 fluxes, ranging on average from −7 nmol m −2 s −1 to −12 nmol m −2 s −1 at noon.The calculated deposition velocities for NO 2 and O 3 exhibited a similar diel course and increased during the morning, reaching their maximum and decreasing during the afternoon.Despite similar deposition velocities during nighttime (∼ 0.1 cm s −1 ), the maximal median deposition velocity for NO 2 was two times lower than for O 3 during daytime (around 0.3 cm s −1 for NO 2 and 0.6 cm s −1 for O 3 ) (Fig. 2e).NO fluxes measured by EC during the field experiment were close to zero during nighttime and were directed upward during daytime, i.e. indicating net emission, with maxima of 0.05-0.1 nmol m −2 s −1 during daytime (see Fig. 2d).

Model vs. measurements: fluxes and mixing ratios
The O 3 fluxes estimated using the Surfatm model agreed well with those measured during the whole experimental period.The linear regression showed that the model underestimated the measured fluxes by only 2 % on average (Fig. 4a).We attempted another step of validation of the Surfatm model by comparing measured and model-derived O 3 mixing ratios at two crucial levels, namely at z 0 and z 0s .For that O 3 mixing ratios were estimated (a) at z 0 from Eq. ( 13) using the measured O 3 flux, the measured O 3 mixing ratio at z ref and modelled R a , and (b) at z 0s from Eq. ( 16) using the modelled O 3 soil flux, the measured O 3 mixing ratio at 20 cm (later moved at 28 cm) and modelled R ac values.In Fig. 4b and c these O 3 mixing ratios are shown in comparison (a) to the O 3 mixing ratio measured at 20-28 cm assuming that 20-28 cm was representative of z 0 , and (b) to the measured O 3 mixing ratio at 5 cm assuming that this level was representative of z 0s .At least during daytime, the modelled O 3 mixing ratios just above the canopy and the soil agree very well with the measurements, which validates the applied values of R a and R ac (necessary to estimate transport times above and within the canopy; see Sect.2.7).This result is indeed justified also by the fact that O 3 mixing ratios modelled with ±50 % of R a and R ac (red dashed lines in Fig. 4b,c) largely deviate from measured mixing ratios.The good agreement for O 3 indicates that the resistances used to model O 3 fluxes were valid and consequently represent the O 3 exchange processes quite well.The turbulent resistances (i.e.R a and R ac ) used to model NO 2 deposition fluxes are identical to those used for modelling the O 3 fluxes (only modulated by different molecular diffusivities; see Sect.2.6).Thus, the good agreement between measured and modelled O 3 fluxes and mixing ratios would suggest applying resistances R a , R ac , R bl , R bs , and R s also for the simulation of NO 2 deposition fluxes.
However, a priori modelled NO 2 deposition fluxes (with R NO 2 int = 0) do not agree well with the measured NO 2 fluxes during the SALSA campaign (Fig. 5).The relationship between measured and modelled NO 2 fluxes showed a signifi- cant scatter (R 2 = 0.45) and a large deviation (slope = 1.22) from the 1:1 line (Fig. 5a).The NO 2 fluxes during nighttime were quite well reproduced by the model with an absolute difference varying around zero (Fig. 5b).However, this small absolute difference caused a large relative difference between measured and modelled fluxes, indicating an underestimation by the model of around 50 %, which was due to the small NO 2 fluxes during nighttime (Fig. 2d).Nevertheless, during daytime the NO 2 deposition was significantly overestimated.The difference between measured and modelled NO 2 fluxes increased during the morning, reached its maximum at noon and decreased during the afternoon (Fig. 5b).At noon, the modelled NO 2 fluxes were typically two times larger than the measured NO 2 fluxes, and this overestimation could occasionally reach a factor of three (Fig. 5b).
It is now required to understand the reasons responsible for this substantial overestimation of the a priori modelled NO 2 deposition.These reasons could be separated into two categories: (i) the measured NO 2 fluxes were not only caused by turbulent transport of NO 2 towards the surface and/or (ii) the resistances to NO 2 deposition used in the model were underestimated.On one hand, the EC method measures the flux at a specific height (z ref = 2 m).For reactive species such as NO 2 , chemical reactions in the air column within or above the canopy could induce a flux divergence with height, meaning that the flux at the measurement height is different than the flux close to the surface, which is in contrast to inert species such as water vapour or CO 2 (e.g.Kramm et al., 1991Kramm et al., , 1996;;Galmarini et al., 1997;Walton et al., 1997).If the characteristic turbulent transport times (see Eq. 20) are not significantly shorter than characteristic chemical reaction times (see Eq. 18), these processes could also induce lower deposition fluxes measured at a height of 2 m.In addition, the EC flux measurements represent the net exchange resulting from the balance between emission and deposition processes.In case the NO 2 fluxes were bi-directional, which would imply that a surface source for NO 2 exists, then the deposition flux estimated by the model would be larger than the measured net flux.On the other hand, the model could also overestimate NO 2 deposition, which implies that the applied resistance parameterizations in the model might be not complete.However, as explained previously, this was not the case for R a , R ac , R bl , R bs , and R s since they were validated owing to the good agreement between measured and modelled O 3 fluxes.Thus, if we presume that the cuticular deposition is negligible (i.e.R NO 2 cut = 9999 s m −1 ) as shown previously (see above), only the remaining resistances R soil and R int for NO 2 could be underestimated.In the following, each reason that may explain the overestimation of NO 2 deposition by the model is explored and discussed.

Impact of chemical reactions on NO 2 fluxes
Transport and chemical reaction times for the NO-O 3 -NO 2 triad were estimated above and within the canopy in order to determine to what extent chemical depletion or production in the air column could affect the measured NO 2 fluxes.
Characteristic transport times (τ trans ) for both above and within the canopy followed a diurnal cycle (Fig. 6a).It was larger during nighttime and decreased during the morning to reach its minimum in the early afternoon.It then increased during the afternoon until sunset.Despite the difference of the layer height (above the canopy: z ref − d = 1.60 m and 1.50 m at the beginning and the end of the experiment, respectively; within the canopy: d − z 0s = 0.10 and 0.19 m at the beginning and the end of the experiment, respectively), τ trans was comparable above the canopy and within the canopy.It was about 200 s during nighttime and decreased to about 55 s above the canopy and to 80 s within the canopy at noon.The lower turbulence and stable atmospheric conditions during nighttime induced a slower turbulent transport, while the unstable atmospheric conditions and turbulent mixing enhancement reduced τ trans .Although τ trans was comparable above and within the canopy, it must be kept in mind that the layer height was different, being 1.50 m above and only 0.20 m within the canopy.This implies that the "transfer velocity" was significantly lower within the canopy than above.
Characteristic chemical reaction times were calculated above and within the canopy.Above the canopy, τ chem was calculated using Eq. ( 18), i.e. taking into account both NO 2 photolysis and NO 2 production by the reaction between O 3 and NO.However, j NO2 was not measured inside the canopy; hence, τ chem could not be calculated using Eq. ( 18).Since j NO 2 is closely related to G r (see Trebs et al., 2009), which typically sharply decreases in a dense canopy, NO 2 photolysis was assumed to be negligible.In addition, the measured O 3 mixing ratio at 0.05 m above ground level was about 10 times larger than the measured NO mixing ratio in the early morning and up to 30 times larger during the afternoon and nighttime (data not shown).The reaction between NO and O 3 is a second-order reaction, but can be approximated by a pseudo-first-order reaction because O 3 was in excess compared to NO.The pseudo-first-order reaction rate constant is defined as k r = k r × O 3 (in s −1 ), and τ chem inside the canopy can be approximated as the chemical depletion time for NO (Eq. 19a).The chemical reaction time followed the same diurnal cycle above and within the canopy: it reached its maximum in the early morning, progressively decreased to reach a minimum in early afternoon, and increased from the early afternoon to the early morning (Fig. 6b).In spite of the comparable diurnal cycle above and within the canopy, τ chem above the canopy was usually faster than inside the canopy.The chemical reaction time above the canopy peaked at 300 s and decreased to 80 s, whereas inside the canopy it reached 600 s and decreased to only 150 s (Fig. 6b).
The DA values calculated from Eq. ( 21) were usually lower than unity, implying that in general turbulent transport was faster than chemical reactions, although DA was occasionally close to unity (Fig. 6c).In addition, DA was larger above the canopy than within the canopy due to the faster chemical reaction time above the canopy.DA values varied between 0.3 and 0.7 within the canopy and ranged from 0.5 to unity above the canopy.Damköhler (1940) stated that a trace gas can be treated as a non-reactive tracer for DA 1.However, it is now generally accepted by the scientific community that a gas can be treated as non-reactive only for DA < 0.1, and that chemical divergence could be of minor importance for 0.1 < DA < 1.For example, Stella et al. (2012) demon-strated that chemical reactions induced a flux divergence for O 3 and NO accounting for 0-25 % of the measured fluxes for 0.1 < DA < 1.
Consequently, the impact of chemical reactions for the NO-O 3 -NO 2 triad above the canopy on measured NO 2 fluxes was evaluated using the method proposed by Duyzer et al. (1995).According to this method, chemistry between NO, NO 2 and O 3 above the canopy could induce only a small divergence.The median difference between the measured and the corrected NO 2 fluxes varied between ±0.025 nmol m −2 s −1 , which corresponded to a relative difference of ±10 % (Fig. 7a), whereas the difference between measured and modelled NO 2 fluxes was about 20 times larger (absolute difference ≈ 0.40 nmol m −2 s −1 , ratio ≈ 2 during daytime; see Fig. 5b and Sect.3.3).Hence, chemistry above the canopy did not explain the large overestimation of NO 2 deposition fluxes by the model.In addition, similarly to O 3 , the NO 2 mixing ratio was estimated at z 0 from Eq. ( 13) using the measured NO 2 flux, the measured NO 2 mixing ratio at z ref and modelled R a , and compared with the NO 2 mixing ratio estimated at 20-28 cm (Fig. 7b).Since the resistance analogy implies the absence of chemical reactions, the good agreement between measured and modelled NO 2 mixing ratio above the canopy also confirmed the non-significance of chemistry above the canopy, at least during daytime.Nevertheless, during nighttime, discrepancies occurred between measured and modelled NO 2 mixing ratios, meaning that fast chemistry cannot be discarded These methods could not be used to estimate the influence of chemical reactions inside the canopy since (i) the method proposed by Duyzer et al. (1995) is based on mass conservation of the NO-O 3 -NO 2 triad and it does not integrate the different emission or deposition processes that could occur inside the canopy, and (ii) the comparison of measured and modelled NO 2 mixing ratios inside the canopy (i.e. at 5 cm) requires knowledge of the modelled soil NO 2 flux, or at least the vegetation flux (to deduce the soil flux from the difference between total and vegetation NO 2 flux), which cannot be estimated without knowledge of the NO 2 internal resistance.However, our results suggest that the impact of NO-O 3 -NO 2 chemistry inside the canopy could be negligible.The calculated DA numbers did not indicate that chemistry was dominating the exchange inside the canopy.In addition, the DA number inside the canopy was lower than above the canopy (Fig. 6c), which implies that chemistry inside the canopy was probably even less important than above the canopy.
It also has to be mentioned that besides NO-O 3 -NO 2 chemistry, other reactions could induce chemical divergence, especially those involving biogenic volatile organic compounds (BVOCs).BVOCs are emitted from vegetation (Guenther et al., 2000;Karl et al., 2001;Beauchamp et al., 2005;Goldstein and Galbally, 2007), including a large variety of compounds (e.g.isoprene, monoterpenes, sesquiterpenes, acetone, methanol, ethanol) with highly variable reactivity (Atkinson and Arey, 2003;Bamberger et al.,  Fig. 8. Mean diurnal cycle of in-canopy mixing ratio differences for NO, NO 2 , and O 3 between 5 cm and 20-28 cm (increasing canopy top) above ground.Solid lines show median values and dashed lines the interquartile range, respectively, for the entire measurement period.Ruuskanen et al., 2011).As indicated in Atkinson and Arey (2003), the lifetime of BVOCs for the reaction with O 3 ranges from few minutes (e.g.α-Terpinene, α-Humulene, β-Caryophyllene) to several hours/months (e.g.isoprene, acetone, methanol).Bamberger et al. (2010) reported that only methanol exhibited consistent fluxes above a grassland.Since the lifetime of methanol for reaction with O 3 is very long (> 4.5 yr; Atkinson and Arey, 2003), we expect a negligible impact of BVOC chemistry on NO, O 3 and NO 2 .This hypothesis is also supported by the good agreement between measured and modelled NO 2 mixing ratio above the canopy (Fig. 7b).

Near-soil NO 2 source and compensation point for NO 2
In the following we discuss the possibility of the existence of a significant NO 2 source near the soil surface that would cause a difference between the observed above-canopy NO 2 flux and the total NO 2 deposition.It would imply the existence of a non-zero canopy or soil compensation point in the resistance model.The potential reason for an NO 2 source is a soil NO emission that is higher than the NO eddy covariance flux observed above the canopy (Fig. 2).There are no direct in situ measurements of soil NO emissions available in the present study, but we estimated the soil emission potential by laboratory incubation measurements (Sect.2.5).For the period of the field experiment, the laboratory-derived soil NO flux ranged from 0.08 to 0.35 nmol m −2 s −1 (median: 0.2 nmol m −2 s −1 ).The values are on average higher than the corresponding above-canopy flux, and a large part of it may have been converted to NO 2 already in the lower part of the canopy (see Mayer et al., 2011;Foken et al., 2012b).However, it has to be considered that the laboratory measurements have been performed with sieved soil.The absence of the usually dense active grass roots (as a competitive sink for mineral nitrogen) may have enhanced the soil microbial processes and led to an overestimation of NO emission compared to an intact plant-soil system, similarly to the effect of grassland tillage (see e.g.Pinto et al., 2004).Another argument against a significant NO 2 source in the lower canopy is the observed in-canopy gradients between 5 cm and 20-28 cm.As shown in Fig. 8, the NO 2 concentration always increased with height, indicating a general downward flux inside the canopy.This is even true for the chemically conserved NO x concentration, indicating that the soil and the air layer above (0-5 cm) were generally a net sink for NO x .It cannot be discarded that chemical conversion occurs just above or in contact to the soil surface, but it obviously does not significantly affect the present analysis.
In addition to these findings, the existence of a canopy compensation point (the NO 2 mixing ratio just above the vegetation elements at which consumption and production processes balance each other) was empirically explored.Figure 9 shows the measured NO 2 fluxes corrected for chemical reactions above the canopy versus the measured NO 2 mixing ratios.Only data for G r > 400 W m −2 were considered, a threshold above which stomatal conductance is supposed to be constant.The linear regression between the NO 2 flux and the NO 2 mixing ratio did not show an intersection of the regression line with the x axis (NO 2 mixing ratio) within the error of the regression at the 95 % confidence interval.Hence, these results do not suggest the existence of a canopy compensation point, and thus indicate the non-existence of an NO 2 emission flux at the meadow.In addition, this result also supports the small influence of chemical NO 2 production inside the canopy, as stated previously.The existence of the NO 2 compensation point, as well as its magnitude, is currently subject to debate (Lerdau et al., 2000).Numerous studies carried out over several ecosystems such as forests, croplands and grasslands reported NO 2 compensation points on the leaf or branch level ranging from less than 0.1 to 1.5 ppb (Johansson, 1987;Weber and Rennenberg, 1996;Gebler et al., 2000Gebler et al., , 2002;;Hereid and Monson, 2001;Teklemariam and Sparks, 2006).However, these studies used (i) non-specific NO 2 detection techniques using molybdenum or iron sulphate converters and (ii) chamber methods to measure the exchange of NO 2 at the leaf level.These methods could lead to an overestimation of the NO 2 compensation point estimation due to (i) overestimation of the NO 2 mixing ratio (Parrish and Fehsenfeld, 2000;Dunlea et al., 2007;Dari-Salisburgo et al., 2009) and (ii) underestimation of the NO 2 deposition flux due to chemistry inside the chambers as discussed by Meixner et al. (1997), Pape et al. (2009), Chaparro-Suarez (2011) and Breuninger et al. (2012).Our results underline the findings of Gut et al. (2002) on Amazonian forest trees and by Segschneider et al. (1995) on sunflower.In addition, Chaparro-Suarez et al. ( 2011) and Breuninger et al. (2012), who made measurements on pine, birch, beech and oak using a specific NO 2 converter (see Sect. 2.3) and performed corrections for chemical reactions inside the chamber, did not find a compensation point for NO 2 .

Model sensitivity to soil resistance for NO 2
A sensitivity analysis of the Surfatm model to R NO 2 soil was made in order to evaluate to what extent a potential underestimation of the NO 2 soil resistance could explain the overestimation of the a priori modelled NO 2 deposition fluxes.The NO 2 deposition flux was modelled using four differ-ent soil resistances (R NO 2 soil = 500 s m −1 , R NO 2 soil = 1000 s m −1 , R NO 2 soil = 2000 s m −1 , and R NO 2 soil = 9999 s m −1 ) and compared to the reference case (i.e.R NO 2 soil = 340 s m −1 ).The modelled NO 2 deposition decreased when R NO 2 soil increased (Fig. 10).However, the sensitivity of the model result to R NO 2 soil was dependent on the time of the day.The relative decrease of the modelled NO 2 deposition flux with increasing R NO 2 soil was less marked during daytime than during nighttime.It was around 1.5, 4, 8.5, and 16 % during daytime for R NO 2 soil equal to 500, 1000, 2000, and 9999 s m −1 , respectively, whereas during nighttime the increase of R NO 2 soil caused a decrease of the modelled NO 2 deposition flux of around 4, 13, 25, and 240 % for the four cases considered (Fig. 10).
This diurnal variation was due to the change of the NO 2 deposition pathways during the course of the day.During daytime, NO 2 is deposited through stomatal and soil pathways, the former representing the main NO 2 removal pathway (Rondón et al., 1993;Gut et al., 2002).Since NO 2 soil deposition represents only a small part of the total deposition, any increase of R NO 2 soil does not induce a large modification of the modelled NO 2 deposition flux.Conversely, the soil pathway represents the only sink for NO 2 during nighttime.Thus, the sensitivity of the modelled NO 2 flux to R NO 2 soil is larger.Obviously, a potential underestimation of R NO 2 soil did not explain the observed discrepancy between measured and modelled NO 2 fluxes.For realistic values of R NO 2 soil (500 s m −1 and 1000 s m −1 ) the modelled NO 2 fluxes were only less than 5 % lower during daytime than the fluxes modelled with R NO 2 soil = 340 s m −1 , whereas the model overestimated measurements by about a factor of two (Fig. 5b).Even if we assume that the soil deposition was zero (i.e.R NO 2 soil = 9999 s m −1 ), that would only explain a model overestimation of 13 %.
Consequently, neither an underestimation of R NO 2 soil nor chemical divergence within and above the canopy or NO 2 emission from vegetation explained the large overestimation of the NO 2 deposition fluxes by the model during daytime.In addition, R a , R ac , R bl , R bs , and R s were already validated owing to the good agreement between measured and modelled O 3 fluxes (see Sect. 3.3).These facts prove that the only process that could explain the overestimation of the modelled NO 2 deposition flux is the existence of an internal resistance for NO 2 , which was ignored in the modelling approach.

Internal resistance for NO 2
In the a priori model parameterization presented above the internal resistance for NO 2 was set to zero.According to the pervious results, only the existence of a significant internal resistance could explain the large discrepancy between measured and modelled NO 2 fluxes.In order to estimate the magnitude of R NO 2 int , NO 2 fluxes were modelled including several values of R NO 2 int (i.e.50-500 s m −1 , with  < 500 s m −1 ) were included.steps of 50 s m −1 ).The results are summarized in Table 2. Following this analysis, it is not clear what was the best value for R NO 2 int .The best slope of the regression (0.94) was found for R NO 2 int = 100 s m −1 , but the lowest RMSE (0.21 nmol m −2 s −1 ) was found for a value of R NO 2 int = 150 s m −1 .Hence, we also deduced R NO 2 int in an alternative empirical approach from the NO 2 flux measurements by invert-ing the resistive scheme (leaving all other resistances as described above for the a priori approach).For large R NO 2 s values that have a high relative uncertainty, this calculation procedure may lead to errors and sometimes even to negative values ofR NO 2  int .Hence, only data for 1/R NO 2 s > 0.2 cm s −1 (R NO 2 s < 500 s m −1 ) were considered.The magnitude of R NO 2 int was highly variable throughout the day (Fig. 11a).It was close to zero during the early morning and progressively increased to 200 s m −1 at noon.The maximal median of R NO 2 int was prevailing during the early afternoon and was about 300 s m −1 .The averaged R NO 2 int was 165 s m −1 , but the magnitude of the estimated R NO 2 int varied considerably and ranged from 100 to 800 s m −1 (interquartile range).In comparison, R NO 2 s was around 400 s m −1 during the early morning and progressively decreased to 100 s m −1 .It then increased again during the early afternoon (Fig. 11a).The contribution of R NO 2 int to the total leaf resistance varied during the day.The contribution was close to zero during the early morning but increased to represent between 50 and 90 % (interquartile range), with the median contribution of R NO 2 int to the total leaf resistance estimated to be 75 % during the early afternoon (Fig. 11b).
Contrary to the results obtained by Segschneider et al. (1995) for sunflower and Geßler et al. (2000and Geßler et al. ( , 2002) ) for beech and spruce, we found the existence of an internal leaf resistance for NO 2 .The results obtained during this study confirmed those obtained by Jonhansson (1987) and Gut et al. (2002), who reported significant values of R NO 2 int ranging from 10 to 2000 s m −1 .As reported in these previous studies R NO 2 int contributed significantly to the total leaf resistance.Nevertheless, its contribution was slightly larger than reported by Jonhansson (1987), who indicated that R NO 2 int represented between 3 and 60 % of the total leaf resistance, and by Gut et al. (2002) and Chaparro-Suarez (2011), who both estimated that R NO 2 int accounted for 40 % of the total leaf resistance.
Both R NO 2 int and its contribution to the total leaf resistance exhibited a diurnal cycle: they increased during the morning but did not decrease in the same proportion during the afternoon.The underlying processes responsible for R NO 2 are the reactions involving NO 2 with apoplastic ascorbate and nitrate reductase (Eller and Sparks, 2006;Teklemarian and Sparks, 2006;Hu and Sun, 2010).The higher the concentrations of ascorbate and nitrate reductase are, the higher is the depletion of NO 2 in the sub-stomatal cavity and the lower is R NO 2 int .However, these reactions are irreversible, and ascorbate and nitrate reductase are not immediately regenerated.Thus, the dynamics of R NO 2 int and its contribution to the total leaf resistance probably reflect these biological processes: the pool of apoplastic ascorbate and nitrate reductase progressively decreased during the morning due to the reactions with NO 2 , leading to the increase of R NO 2 int in the afternoon.Since these substances are not regenerated immediately, R NO 2 int remained at its maximum value during the afternoon.Finally, during nighttime when stomatal closure prevented NO 2 from entering into the sub-stomatal cavity (and thus did not react with apoplastic ascorbate and nitrate reductase), the pool of ascorbate and nitrate reductase was regenerated leading to minimum R NO 2 int values in the morning.

Conclusions
This study reports about measurements of NO, NO 2 and O 3 exchanges between a meadow and the atmosphere using eddy covariance, a method without disturbance of the micrometeorological conditions and without impacts on plant functioning.Initially, our a priori NO 2 deposition fluxes modelled with the Surfatm model did not consider any internal resistance.In this case, the modelled NO 2 deposition flux exceeded the measured NO 2 deposition flux by a factor of two.In order to identify the processes responsible for this overestimation, (i) the influence of a chemical divergence above the canopy, (ii) the existence of an NO 2 emission flux from vegetation, (iii) the potential underestimation of the resistances used in the model, and (iv) the existence of an internal resistance for NO 2 were explored.
The results did not suggest a considerable influence of chemical reactions above (and within) the canopy.In addition, the non-existence of a canopy compensation point for NO 2 excluded the presence of an NO 2 emission flux from vegetation.Moreover, the sensitivity of the model to the soil resistance to NO 2 only accounted for a small difference between measured and modelled flux, which was 13 % during daytime if the soil deposition was assumed to be zero.The other resistances were implicitly validated owing to the good agreement between measured and modelled O 3 fluxes.
Consequently, only the existence of an internal resistance limiting NO 2 stomatal uptake could explain the overestimation by the Surfatm model.The median internal resistance for NO 2 was estimated from the NO 2 flux measurements and from the modelled resistances to be about 300 s m −1 , while the median for the stomatal resistance was only around 100 s m −1 during daytime.Consequently, the internal resistance represented between 50 and 90 % of the total leaf resistance.
This study proved the existence of a large and significant internal resistance for NO 2 for the grass species present at the meadow.For the first time, this type of investigation was made without an alteration of the microclimatological conditions that may occur when using the chamber method.This topic is particularly relevant for estimating dry deposition of NO 2 over terrestrial ecosystems.An internal resistance is currently not taken into account in global models such as the EMEP model (Tsyro, 2001;Simpson et al., 2003) or the MOZART model (Horovitz et al., 2003), or strongly underestimated such as in the MATCH-MPIC model, in which the internal resistance is assumed to be half of the leaf stomatal resistance (Ganzeveld and Lelieveld, 1995;Shepon et al., 2007).These issues could lead to a large overestimation of the terrestrial NO 2 sink.Nevertheless, further studies at other ecosystems are required to establish a parameterization of the internal resistance as a function of vegetation type that can be implemented in global chemistry and transport models.
R soil are aerodynamic resistance above the anopy, aerodynamic resistance within the canopy, leaf quasi-laminar boundary layer esistance, soil quasi-laminar boundary layer resistance, stomatal resistance, internal esistance, cuticular resistance and soil resistance, respectively.Indexes i, z ref , z 0 , z 0' , z 0s and urf indicate the gas considered, the reference height, the canopy roughness height for omentum, the canopy roughness height for scalar, the soil roughness height for momentum, nd the soil surface, respectively.

Fig. 1 .
Fig.1.Resistive scheme used in the Surfatm model for pollutant exchange.χ is the gas concentration.R a , R ac , R bl , R bs , R s , R int , R cut and R soil are aerodynamic resistance above the canopy, aerodynamic resistance within the canopy, leaf quasi-laminar boundary layer resistance, soil quasi-laminar boundary layer resistance, stomatal resistance, internal resistance, cuticular resistance and soil resistance, respectively.Indexes i, z ref , z 0 , z 0 , z 0s and "surf" indicate the gas considered, the reference height, the canopy roughness height for momentum, the canopy roughness height for scalar, the soil roughness height for momentum, and the soil surface, respectively.

P.
Stella et al.: Measurements of nitrogen oxides and ozone fluxes by eddy covariance at a meadow 2.8 Estimation of NO-O 3 -NO 2 flux divergences above the canopy

Fig. 2 .
Fig. 2. Diel courses of (a) global radiation; (b) air relative humidity (blue line) and temperature (red line); (c) nitrogen dioxide (blue line), nitric oxide (green line) and ozone (red line) mixing ratios at 1.65 m above ground level; (d) nitrogen dioxide (blue line), nitric oxide (green line) and ozone (red line) fluxes; and (e) deposition velocities for nitrogen dioxide (blue line) and ozone (red line) determined by EC from 29 August to 20 September 2005.Solid lines represent half-hourly medians and dotted lines represent interquartile ranges.Fluxes were not corrected for chemical reactions.Only those data have been considered for which footprint analysis indicated that at least 95 % of the fluxes have originated from the experimental field (see Fig. 3).

45Figure 3 :Fig. 3 .
Figure 3: Averaged cumulative footprint contours showing the footprint areas for 80 % (solid line) and 95 % (dotted line) of the total flux measured by eddy covariance for (a) all, (b) unstable, (c) neutral, and (d) stable conditions.x-axis and y-axis are distances from the mast (in meter).The analysis was performed for all data from 29 August to 20 September 2005.

Figure 4 :Fig. 4 .
Figure 4: Comparison between measured and modelled (a) O 3 fluxes, and O 3 mixing ratios (b) above and (c) within the canopy.Shown are median values from 29 August to 20 September 2005.Blue lines are measured mixing ratios, solid red lines are modelled mixing ratios and dotted red lines are modelled mixing ratios with an uncertainty of ± 50 % for the aerodynamic resistances.For details see text.
Fig. 5. (a) Comparison between measured and modelled NO 2 fluxes.(b) Half-hourly median (solid lines) and interquartile range (dotted lines) of the difference (blue lines) and ratio (red lines) between measured and modelled NO 2 fluxes from 29 August to 20 September 2005.

Figure 6 :Fig. 6 .
Figure 6: Half hourly medians of (a) transport times, (b) chemical reaction times, and (c) Damköhler numbers above (red symbols) and within (blue symbols) the canopy from 29 August to 20 September 2005.

Figure 7 :Fig. 7 .
Figure 7: (a) Half hourly median (solid line) and interquartile range (dotted lines) of the difference (blue lines) and the ratio (red lines) between measured NO 2 fluxes at z = 2.0 m and NO 2 fluxes corrected for chemical reactions above the canopy from 29 August to 20 September 2005.(b) Comparison between measured (blue line) and modelled (red lines) NO 2 mixing ratio above the canopy Dotted lines are mixing ratios modelled with an uncertainty of ± 50 % for the aerodynamic resistance.For details see text.

Figure 8 :
Figure 8: Mean diurnal cycle of in-canopy mixing ratio differences for NO, NO 2 , and O 3 between 5 cm and 20-28 cm (increasing canopy top) above ground.Solid lines show median values and dashes lines the interquartile range, respectively, for the entire measurement period.

Fig. 9 .
Fig. 9. Measured NO 2 flux as a function of the NO 2 mixing ratio (z = 2.0 m) from 29 August to 20 September 2005.Solid and dotted lines are the regression line and its 95 % confidence interval, respectively.NO 2 fluxes were corrected for chemical reactions above the canopy and averaged for NO 2 mixing ratio bins of 0.1 ppb.Only data for G r > 400 W m −2 were included.

Fig. 10 .
Figure 10: Half hourly median of the response of the modelled NO2 deposition flux to the soil resistance for 2 NO soil R = 500 s m -1 (solid red line), 2 NO soil R = 1000 s m -1 (solid blue line), 2 NO soil R =

Fig. 11 .
Figure 11: Half hourly medians (solid lines) and interquartile range (dotted lines) of (a) NO 2 internal (blue lines) and stomatal resistances (red lines) and (b) the relative contribution of

Table 2 .
Comparison of measured and modelled NO 2 fluxes for different values of the internal resistance.Only data for 1/R