Eddy covariance fluxes and vertical concentration gradient measurements of NO and NO2 over a ponderosa pine ecosystem: observational evidence for within-canopy chemical removal of NOx

Exchange of NOx (NO+NO2) between the atmosphere and biosphere is important for air quality, climate change, and ecosystem nutrient dynamics. There are few direct ecosystem-scale measurements of the direction and rate of atmosphere–biosphere exchange of NO x. As a result, a complete description of the processes affecting NO x following emission from soils and/or plants as they transit from within the plant/forest canopy to the free atmosphere remains poorly constrained and debated. Here, we describe measurements of NO and NO2 fluxes and vertical concentration gradients made during the Biosphere Effects on AeRosols and Photochemistry EXperiment 2009. In general, during daytime we observe upward fluxes of NO and NO 2 with countergradient fluxes of NO. We find that NO x fluxes from the forest canopy are smaller than calculated using observed flux– gradient relationships for conserved tracers and also smaller than measured soil NO emissions. We interpret these differences as primarily due to chemistry converting NO x to higher nitrogen oxides within the forest canopy, which might be part of a mechanistic explanation for the “canopy reduction factor” applied to soil NOx emissions in large-scale models.


Introduction
The chemistry of nitrogen oxides is a major factor affecting the oxidative capacity of the atmosphere and the global burden of tropospheric ozone (Crutzen, 1973).Reactive nitrogen oxides are also a nutrient (Sparks, 2009;Takahashi et al., 2004Takahashi et al., , 2005a;;Teklemariam and Sparks, 2004;Lockwood et al., 2008) and interactions between available nitrogen in ecosystems and atmospheric nitrogen are many and complex, with exchange processes altering the patterns of nitrogen availability in the biosphere (Townsend et al., 1996;Vitousek and Farrington, 1997;Vitousek et al., 1997;Holland and Lamarque, 1997;Holland et al., 1997;Ollinger et al., 2002a, b;Hietz et al., 2011).Nitrogen is the limiting nutrient for plant growth in most regions outside the tropics (Hungate et al., 2003;Galloway et al., 2004;Hietz et al., 2011), thus nitrogen deposited to the surface after atmospheric transport can act as fertilizer contributing to enhanced carbon uptake (H.Morikawa et al., 2004;T. Morikawa et al., 2004;Takahashi et al., 2004Takahashi et al., , 2005a, b;, b;Sparks, 2009;Norby et al., 2010).For example, Norby et al. (2010) found that the availability of nitrogen was a major limiting factor for the CO 2 fertilization effect in the FACE (Free-Air CO 2 Enrichment) experiment.However, excess nitrogen deposition may impair ecosystem health (Hessen et al., 1997;Herman et al., 2001) by causing dehydration, chlorosis, or membrane damage from peroxy acetal nitrate (PAN) (Ordin et al., 1971; Published by Copernicus Publications on behalf of the European Geosciences Union.Oka et al., 2004), or by inducing soil acidification and eutrophication (Makarov and Kiseleva, 1995;Pawlowski, 1997;Gbondo-Tugbawa and Driscoll, 2002;Zapletal et al., 2003;Chen et al., 2004).
A comprehensive understanding of NO x exchange between the atmosphere and biosphere does not yet exist.Experimental studies have primarily focused on NO emissions from soils to the atmosphere (e.g., Butterbach-Bahl et al., 2002;Gasche and Papen, 2002;Gut et al., 2002a, b;Rummel et al., 2002;van Dijk et al., 2002;Dorsey et al., 2004;Duyzer et al., 2004;Feig et al., 2008;Bargsten et al., 2010;Yu et al., 2010) or on the leaf-level transfer of NO and NO 2 using branch enclosures (Hereid and Monson, 2001;Chaparro-Suarez et al., 2011;Breuninger et al., 2013;and references therein).Studies at the canopy-scale often assume a simple flux-gradient similarity relationship, meaning molecular movement is always along the gradient of high to low concentration, to infer the rate of exchange from vertically resolved observations (Mayer et al., 2011 and references therein).Direct measurements of the direction and rate of exchange (the flux) at the canopy scale are few (Wesely et al., 1982;Wildt et al., 1997;Horii, 2002Horii, , 2004;;Farmer et al., 2006;Neirynck et al., 2007;Li and Wang, 2009;Brummer et al., 2013).
The base conceptual model for biosphere-atmosphere exchange of NO x is shown in Fig. 1.In this model, NO is mainly emitted by soil microbial activity, is converted to NO 2 by reaction with O 3 , and is then ultimately oxidized to nitric acid (HNO 3 ), which returns back to the biosphere via wet and dry deposition.The timescale of NO to NO 2 conversion by O 3 and the photolysis of NO 2 back to NO is typically ∼ 100 s in the daytime, set by the photolytic and chemical reaction rates, and is comparable to the turbulent mixing time (τ turb ) within and out of a forest canopy.The timescale for NO x oxidation to HNO 3 via reaction of OH with NO 2 is 3-10 h, long enough that NO x changes by less than 1 % on the canopy residence timescale from this loss process.The rapid interconversion between NO and NO 2 implies that the individual fluxes of NO or NO 2 will not follow the form of fluxes of a conserved tracer, such as water, heat, or carbon dioxide (Vila-Guerau de Arellano et al., 1993), but the long lifetime of the sum implies the flux of NO x , will follow that of conserved tracers.
This model of NO x exchange is qualitatively supported by ecosystem-scale observational studies.For example, NO is observed to decrease as air is transported up through a canopy from the forest floor (e.g., Rummel et al., 2002) and the low light levels within a shaded canopy reduce NO 2 photolysis and enhance the NO 2 to NO x ratio.For these reasons, downward NO fluxes and upward NO 2 fluxes in across the canopy top at Harvard forest have been observed as expected (e.g., Horii, 2002).
Calculations of ozone by large-scale chemical transport models parameterized with measured NO soil fluxes overpredict O 3 concentrations in comparison to aircraft and tower observations (e.g., Lerdau et al., 2000).To match observations, these models invoke a canopy reduction factor of 25-80 % (Jacob and Wofsy, 1990;Yienger and Levy, 1995;Wang and Leuning 1998).This parameter removes soil NO x exclusively before it escapes the canopy, thus preventing its contribution to atmospheric ozone formation.Typical parameterizations use LAI (leaf area index) and SAI (stem area index) as parameters controlling the removal of soil NO x .To our mind these mechanisms are unphysical as they do not act on all NO x in the plant canopy -but only on soil NO x .One motivation for our experiment is to quantify NO x removal processes in order to develop a physically based model that treats all NO x identically regardless of source.
At the same time, laboratory observations at the leaf scale indicate bi-directional exchanges of NO x by plant biota, where the direction and rate of exchange is controlled by a so-called "compensation point" -a concentration above which vegetation takes up NO 2 and/or NO but below which emissions occur (Sparks et al., 2001;Raivonen et al., 2009;Chaparro-Suarez et al., 2011;Breuninger et al., 2013); however, a mechanism for the emissions remains to be discovered (Breuninger et al., 2013).Direct observations of the NO 2 compensation point are analytically challenging (Raivonen et al., 2003) and evidence suggests that compensation behavior is not fixed but rather varies by plant species, plant life cycle, and environmental conditions (Raivonen et al., 2009).That said, compensation points have been measured in the range of ambient NO 2 abundances and have been reported from 0.05 to 3 ppb (Sparks et al., 2001;Raivonen et al., 2009;Chaparro-Suarez et al., 2011;Breuninger et al., 2013).Since the NO 2 concentration in remote continental regions is typically less than 1 ppb, these observations leave open the possibility that the majority of forests on earth are a source of NO x from direct NO 2 emissions from plants, in addition to any soil NO emissions.This is in direct contradiction with the need for canopy reduction factors that remove nitrogen oxides emitted from soils prior to its exit from plant canopies (Lerdau et al., 2000).
Recent field studies suggest the existence of rapid withincanopy chemistry affecting nitrogen oxides that is not included in the conceptual model of Fig. 1.Farmer et al. (2006) observed upward exchanges of total peroxynitrates ( RO 2 NO 2 ) and HNO 3 and interpreted this as the formation of these molecules within a forest canopy.In addition, Wolfe et al. (2009) described the importance of chemical processes in speciated acyl peroxynitrate exchange, also finding observational evidence for within-canopy chemistry affecting observed fluxes.Several other experimenters have reported the occurrence of within-canopy chemistry affecting fluxes of biogenic volatile organic compounds (BVOCs) (Holzinger et al., 2005;Karl et al., 2005;Bouvier-Brown et al., 2009a, b;Park et al., 2014) and ozone (Kurpius and Goldstein, 2003).Our own recent study of peroxynitrate fluxes (Min et al., 2012) supports this idea, providing experimental evidence for upward fluxes of unidentified peroxynitrates (F XPN ) formed within the forest canopy.Taken together, these studies emphasize the importance of rapid chemistry not only for determining the magnitude but also for the direction of nitrogen exchange at the biosphere-atmosphere interface.
The Biosphere Effects on AeRosols and Photochemistry EXperiment (BEARPEX) included a component designed to provide comprehensive measurements of vertical concentration gradients and fluxes of a wide suite of nitrogen oxides -NO, NO 2 , total and speciated peroxynitrates, total and speciated alkyl ( RONO 2 ) and multifunctional nitrates, HNO 3 , and nitrous acid (HONO) -and therefore presented a direct opportunity to test our ideas about canopy-scale NO x exchange.Analyses of peroxynitrate (Wolfe et al., 2009;Min et al., 2012) and HONO (Ren et al., 2011) fluxes have been reported elsewhere.Here we present observations of vertical concentration gradients and fluxes of NO 2 and NO measured with laser-induced fluorescence and chemiluminescence, respectively.Fluxes are derived using the eddy covariance method.We describe relationships between gradients and fluxes, present and interpret evidence for chemical canopy reduction processes, and explore the significance of chemistry within the canopy to the import/export of NO x from the canopy.

Research site and instrumentations
The data used in this work were obtained as a part of the BEARPEX 2009experiment (15 June-31 July 2009).The experiment was conducted over a managed Ponderosa pine plantation on the western slope of the Sierra Nevada Mountain range, 75 km downwind of Sacramento, California and near the University of California Berkeley Blodgett Forest Research Station (UC-BFRS, 38 • 53 42.9 N, 120 • 3757.9W; 1315 m).Many of the results from BEARPEX can be found in a special issue of Atmospheric Chemistry and Physics, http://www.atmos-chem-phys-discuss.net/special_issue89.html.A brief description of the field site and of the instrumentation relevant to this paper follows.
Analysis of the local meteorology by Choi et al. (2011) andDillon et al. (2002) indicate that in the summer (May-September), winds at the BEARPEX site are characterized by daytime southwesterlies (210-240 • ) and nighttime northeasterlies (30 • ) with little variability.The major source of anthropogenic emissions in the region is the city of Sacramento and its suburbs.There is a line source of oak trees that are strong isoprene emitters aligned perpendicular to the flow between the urban center and the site.This source distribution, in combination with the regular winds, results in low concentrations of trace gases with anthropogenic or isoprene sources early in the morning and higher concentrations in air transported from the west later in day (Dillon et al. 2002;Day et al., 2003;Murphy et al., 2007;Choi et al., 2011;LaFranchi et al., 2011).The two sources arrive at distinct times; with air influenced primarily by isoprene arriving at approximately noon and the urban plume combined with the isoprene source arriving about 3-4 h later.
There were two sampling towers at the site: a 15 m walkup tower on the south side of the site (hereafter south tower) and an 18 m scaffolding tower located 10 m north of the south tower (hereafter north tower).On the south tower, temperature, relative humidity, wind speed, net radiation, photosynthetically active radiation (PAR), water vapor, carbon dioxide (CO 2 ), and O 3 were monitored at 5 heights (1.2, 3.0, 4.9, 8.75, and 12.5 m above the forest floor).At 12.5 m, fluxes of water vapor, CO 2 , and O 3 were measured.Vertical gradients of temperature, relative humidity, and wind speed were also measured on the north tower at 5 heights (1.2, 5.4, 9.2, 13.3, and 17.5 m above the forest floor).Measurements from the north tower or on an adjacent heightadjustable lift included NO, NO 2 , HONO, total peroxy nitrates ( PNs, RO 2 NO 2 ), total alkyl and multifunctional nitrates ( AN, RONO 2 ), HNO 3 , hydroxyl radical (OH), hydroxy peroxy radical (HO 2 ), OH reactivity, O 3 , several individual PNs, several individual ANs, numerous volatile organic compounds (VOCs) including many biogenic VOCs (BVOCs), formaldehyde (HCHO), glyoxal, methylglyoxal, organic peroxides, and aerosol chemical and physical properties.Needle temperature, soil moisture, soil temperature, and soil heat flux were also monitored.All measurements were made at the 17.5 m level and many were additionally recorded at one or more of the following heights: 0.5, 1.2, 5.4, 9.2, and 13.3 m.For simplicity, we refer to these measurement heights as 0.5, 1, 5, 9, 13 and 18 m.In addition, soil NO measurements were made using dynamic chambers on 2 July (09:15, 13:40 and 18:10), 12 (08:25 and 16:20) and 30 (08:50 and 16:10) to provide an observational constraint on the soil NO emission at this site: soil NO flux from four different locations (15 min sampling each) were measured to represent morning, midday and the late afternoon time window.
The upper canopy at this site was mainly Pinus ponderosa L., with a few scattered Douglas fir, white fir, and incense cedar.The understory was primarily mountain whitethorn (Ceanothus cordulatus) and manzanita (Arctostaphylos species) shrubs (see Misson et al., 2005, for a more detailed site description and history).The mean tree height was 8.8 m and the leaf area index (LAI) was 3.7 m 2 m −2 , based on a tree survey conducted on 17 July 2009.
NO was measured using a custom-built two-channel chemiluminescence NO detection system (2ch-CL) and NO 2 with two separate thermal-dissociation laser-induced fluorescence (TD-LIF) systems.The sampling inlets for NO and NO 2 were collocated at 0.5, 5, 9 and 18 m on the north tower and represent the forest floor, mid-canopy, top canopy, and above canopy, respectively.At 18 m, fluxes of NO and NO 2 were monitored along with 3-D wind and temperature from a sonic anemometer (Campbell Scientific CSAT3 3-D Sonic Anemometer).The measurements were combined to infer fluxes using an eddy covariance method (EC) (see Sect. 3).The sonic anemometer was pointing into the mean daytime wind stream with 0.02 m vertical displacement and 0.2 m horizontal displacement from the NO and NO 2 inlets.
The 2ch-CL system for the NO flux and vertical gradient measurements was based on the standard O 3 chemiluminescence method.A detailed description of the operating principle can be found elsewhere (Drummond et al., 1985 and references therein).Briefly, ambient NO is combined with an excess of O 3 generated by electric discharge in O 2 .The reaction of NO and O 3 produces excited-state NO 2 molecules, which then fluoresce.Two gold-plated detection cells were used for simultaneous flux and vertically resolved concentration measurements.The signals from photocathodes (flux channel: EMI 9658B, gradient channel: Hamamatsu H7421-50) were acquired at 5 Hz.The cell pressures were maintained at 8-8.7 Torr with pressure restricted at the inlet and a fluorinated oil-sealed rotary vane pump.During the sampling mode, 100 % of the ozone flow was added directly into the detection cell to monitor the ambient NO concentration (for 24 s).The background signal was monitored by adding 50 % of the O 3 to the sampling air prior to the detection cell to titrate ∼ 90 % NO (for 6 s).Incomplete titration of NO was employed to limit interferences from fluorescence of vibrationally excited OH molecules produced in the reaction of ozone with alkenes (Drummond et al., 1985).The other 50 % of the ozone was added directly to the cell to minimize flow changes within the reaction cell between the sampling and the background mode.Our own laboratory experiments confirm that a wide variety of terpenes react with ozone to efficiently produce vibrationally excited OH and we configured the instrument to minimize detection of this signal.
Two TD-LIF systems were used for simultaneous flux and vertical gradient measurements of NO 2 and the higher nitrogen oxide species PNs, ANs and HNO 3 .Details of LIF detection of NO 2 (Thornton et al., 2000), thermal dissociation of higher nitrogen oxides (Day et al., 2002), and appli-cation to EC flux measurement (Farmer et al., 2006) are described elsewhere.Briefly, thermal dissociation of each class of higher oxide generates NO 2 and a companion radical at the characteristic temperatures ∼ 180 • C for PNs, ∼ 350 • C for ANs, and ∼ 600 • C for HNO 3 , (Day et al., 2002).The thermal dissociation is followed by detection of NO 2 by LIF.In both TD-LIF systems, excitation at 585 nm was provided by frequency doubled Nd:YAG (Spectra Physics, average power of 2 W at 532 nm, 30 ns pulse length) pumping a custom-built tunable dye laser operating at 8 kHz.The wavelength of the dye laser beam was tuned to a specific, narrow rovibronic feature of NO2 by rotating an etalon within the dye cavity.We alternated the laser frequency between a strong NO 2 resonance (8 s) and the weak continuum adsorption (4 s) to maintain a frequency lock on the spectral feature of interest.By adapting a supersonic expansion technique, we acquired ∼ 10-fold higher sensitivity to NO 2 (Cleary et al., 2002).The fluorescence signal 700 nm long was collected and imaged onto a red sensitive photocathode (Hamamatsu H7421-50).Gated photon counting techniques (Stanford Research Systems, SRS 400) were employed to discriminate against prompt background signals.Laboratory measurements and comparison in the field showed the two TD-LIF instruments to have calibrations that were identical to within 4 % (slope: 1.0 ± 0.10, R 2 : 0.92).Allowing the intercept to vary from zero did not change the slope or R 2 .
Cell pressures in the flux system were reduced to 0.17-0.19Torr to achieve the high expansion ratios for the supersonic jet cooling by using Lysholm twin screw blowers (Whipple model 2300 superchargers) backed by an oil-sealed rotary vane pump.The jet nozzles and this pump system combined to maintain a 580-700 sccm flow through each of the four cells (total flow of 2300 sccm).To reduce high frequency damping of turbulent eddies and interference from secondary chemistry in the heated section of the inlet and sampling lines (30 m), we added a diaphragm bypass pump and maintained the total flow of a 13 000 sccm.For the gradient system, critical orifices as pressures restrictors (AirLogic, F-2815-251-B85, 0.025" orifice diameter) were placed at the end of the inlet manifold to reduce the pressure along the sampling line.
Calibrations in the field were repeated once (gradient measurement of NO and NO 2 and flux measurement of NO) or twice (flux measurement of NO 2 ) per hour.NO 2 standard gas (4.9 ± 0.2 ppm No x in N 2 , PRAXAIR) was diluted to 3-20 ppb in zero air and added to system at the inlet tip.For NO, 2.25-6.7 ppb of NO (5.4 ppm ± 5 % NO in N 2 , PRAX-AIR) was diluted with zero air and added at the inlet.Both cylinders were referenced to a library of calibration standards maintained in our laboratory.The mixing ratios were corrected (< 2 %) for quenching by water using north tower RH measurements.To evaluate the background counts due to cell scatter and photocathode dark noise, we flowed excess zero air into the inlet once/twice per hour.The diagnostics for the NO and NO 2 flux instruments (calibration and zeroing) were completed within the first 3 min of every 30 min (Fig. 2a).Flux data for both species were collected at 18 m during the first 27 min, from the 3 to the 30 min, and for the last 27 min, from the 33 to the 60 min each hour (Fig. 2a).NO and NO 2 at the lower levels were measured by switching between the 9, 5, and 0.5 m heights and sampling at each height for 2 min (Fig. 2b); the first 5 s for data of each level after the valve switching from another height were deleted to insure the measurement corresponded to the height in question.Calibrations and zeros were completed in the last 4 min from 56 to 60 min of every hour for both gradient systems (Fig. 2b).
Data affected by exhaust plumes from a nearby propane electrical generator (mostly at night) and the infrequent wafts of car exhaust were removed prior to analysis.These spikes were defined as variations in the NO or NO 2 concentration in excess of 3 times the standard deviation of the 10 min running mean.A few remaining spikes were identified through correlations with CO and removed by hand.Over the campaign, the NO 2 detection limit (defined as S/N = 2) was ∼ 45 ppt for 1 s corresponding to 1.3 ppt for a 30 min average for the flux system, and was ∼ 10 ppt 1 min (4.0 ppt for 30 min) for the gradient system.The NO detection limit for flux cell was ∼ 58 ppt for 1 s (1.6 ppt for 30 min averages) and for the gradient cell, ∼ 29 ppt for 1 min (11.5 ppt for 30 min averages) at midday (12:00-14:00, local time).

Eddy covariance calculation
The flux (F c ) of an atmospheric constituent (c) (i.e., the turbulent mass transport of c through a vertical reference layer) can be evaluated from the covariance between the concentration of c and the vertical wind (w) in a method known as eddy covariance (EC) and is represented mathematically by Eq. ( 1) (e.g., Foken, 2006;Lee et al., 2004;McMillen, 1988).
In Eq. ( 1), primes represent the deviation from the mean, subscripts i refer to individual high-time resolution measurements (NO or NO 2 ), and bars indicates the mean over the averaging interval.In this work, the flux of NO x , F NO x , is defined as the sum of the separately calculated F NO and F NO 2 .We checked this calculation by comparing with F NO x , calculated by first adding the NO and NO 2 concentrations and found the two methods agree to within 18 %, which is smaller then propagated errors of F NO x estimated (Table 1) from the sum of F NO and F NO 2 .We used 5 Hz data for the flux calculations and averaged for ∼ 30 min, a timescale that spanned the range of the major flux-carrying eddies at this site (e.g., Wolfe et al., 2009;Farmer et al., 2006).Prior to calculating fluxes, we rotated the wind measurements to ensure that the vertical winds were normal to the shear plane (Baldocchi et al., 1988;McMillen, 1988).We also de-spiked and de-trended the concentration data, where spikes were defined as data greater than 3 times the standard deviation of the 10 min running mean, and where the 10 min running mean was also used for de-trending.
To synchronize the timing of wind and concentration measurements, the lag was determined from the maximum in covariance of the deviation from the mean of vertical wind speed and concentration; the cross-correlation between scalar and vertical wind were calculated for every 30 min and then averaged with multiple time windows (i.e., morning: 10:00-12:00, midday: 12:00-14:00, afternoon: 14:00-16:00) to determine the precise lag time.We did not find any   Highest normalized correlation between wind and temperature, NO or NO 2 were observed as expected; 0 s for wind and temperature, 1.4 and 2.6 s for NO and NO 2 .change in lag time with time of day.We compared the lag times in the different periods of the experiment and found no drift over the duration of the experiment.Three periods were tested, 18-30 June, 1 -15 July and 16-30 July, and each calculated lag time agreed within 1 data point (0.2 s) for both NO and NO 2 .Figure 3a-c show the lag correlation between wind and temperature, NO, and NO 2 , respectively.The data plotted in Fig. 3 are the averaged midday (12:00-14:00) lag over the whole field campaign and are representative of lag correlation plots throughout the experiment.As expected, zero lag was observed between vertical wind speed and temperature as both quantities are synchronously measured by the same instrument, the sonic anemometer.Lag times for NO and NO 2 were measured to be 1.4 and 2.6 s, times that were consistent with transport times in the tubing (< 0.8 s) plus the time difference between sonic anemometer computer and computers for 2ch-CL and TD-LIF.
To assure that each 30 min flux was representative of the average surface exchange over the sampling period, we tested the calculated fluxes for stationarity (Farmer et al., 2006;Foken 2006;Wolfe et al., 2009).To do this, five equally divided subsets of each 30 min flux period, F sub , were averaged and compared with that of the full period, F 30 min .If F sub differed from F 30 min by more than 30 % then that measurement period was defined as non-stationary and that half-hour excluded from further analysis (Foken and Wichura, 1996).Also, the calculated flux data with a tilt angle greater than 5 • from the wind rotation (Lee et al., 2004) and with a friction velocity smaller than 0.1 m s −1 or larger than 1.5 m s −1 (Foken 2006) were excluded for further analysis.We tested this using friction velocity in the range 0.05-0.2m s −1 as filtering criteria.Changes in this range do not affect our conclusion, which is consistent with previously reported analyses of fluxes at this site (Farmer et al., 2006).
Approximately 2/3 of the daytime and half of the nighttime data remained after application of these filters.We estimate the total uncertainty in F NO x by combining the systematic and random error terms in F NO and F NO 2 flux estimations following Moore et al. (1986) and Massman (1991).Each of the individual elements is summarized in Table 1 and detailed procedures are described in Farmer et al. (2006) and Wolfe et al. (2009).
The total systematic uncertainties for F NO and F NO 2 (< 8 and < 6 %, respectively) are calculated from the root mean square of errors from instrument calibration (7 and 5 % for NO and NO 2 , respectively; see Day et al., 2002), sensor separation and inlet dampening (< 2 % for both F NO and F NO 2 ), instrument time response (< 0.2 % and < 0.7 % for daytime F NO and F NO 2 , respectively) and data acquisition sequencing (i.e., laser line-locking cycling for TD-LIF system: < 3 % or frequent background checking for 2ch-CL system: < 2 %, estimated from the sensible heat flux calculation using temperature data coincident with the NO or NO 2 data.The gaps in the data acquisition sequence are replaced by the mean concentration for 27 min.) Two different methods were used to estimate the precision (random errors) of the flux measurements: (1) estimates based on the finite precision of photon counting and (2) the variance of the flux calculation with lag determination.The precision estimates based on photon-counting statistics follow Farmer et al. (2006) and are 0.08 ppt m s −1 (20 %) and 0.14 ppt m s −1 (10 %) for F NO and F NO 2 , respectively, over half an hour.This estimate is similar to but slightly smaller than the precision of 25 % (0.10 ppt m s −1 for F NO ) and 21 % (0.29 ppt m s −1 for F NO 2 ) estimated from the flux variance over a range of lag times far from the true lag (Ruuskanen et al., 2011), indicating the presence of other sources of random error in the measurement in addition to photon counting.One common measure of the flux detection limit is 2 times the standard deviation of the covariance within the time window far from the peak covariance.We use ±230 ∼ 250 s during the daytime (09:00-18:00) (Table 1).The value is not significantly different if we use smaller time windows.For cases such as ours, where fluxes are bi-directional, an alternative approach to estimating a detection limit based on the absolute cross-correlation has been proposed by Park et al. (2013).By either metric the fluxes we report are well above the noise.
Spectral analyses of the fluxes are shown in Fig. 4 and include the cospectral density (Fig. 4a), the normalized cospectrum (Fig. 4b), and the absolute value of the normalized cumulative cospectrum (Fig. 4c) for temperature, NO, NO 2 and NO x .We show each measurement averaged for the time interval 11:00-12:00 throughout the whole field campaign.By analyzing the NO x (yellow) cospectrum, rather than the spectra of NO (green) and NO 2 (blue) separately, we are insensitive to the effects of the rapid chemical conversion between NO and NO 2 .The absolute values of the negative cospectral density were plotted as closed triangles in Fig. 4. The figures show that we capture the full range of eddies that contribute to the flux.
The spectral analysis of NO x provides additional evidence that our instruments for NO and NO 2 observe the full range of flux-carrying eddies at this site.We see in Figure 4a that the observed cospectral density of sensible heat, w T , decreases in the inertial sub-range (above 0.003 Hz) with a linear slope between that predicted by surface layer theory (−5/3) (Kaimal and Finnigan, 1994) and the slope for sensible heat observed previously at this site (−7/3) (Farmer et al., 2006;Wolfe et al., 2009).Because the cospectral density of the vertical wind speed and NO, NO 2 and NO x are parallel to that of sensible heat, we have confidence that our sampling interval and data acquisition time resolution were sufficient to capture the flux-carrying eddies.Additionally, the comparable behavior observed in the w T , w NO , w NO 2 , and w NO x cospectrum confirm that those frequencies characteristic of our instruments' sampling and operative cycles (e.g., regular patterns in on/off sampling sequencing in both NO and NO 2 ) do not interfere with the measurements of fluxes.The parallel slopes of w T , w NO , w NO 2 , and w NO x in the inertial sub-range (above 0.003 Hz), demonstrates that measured fluxes were not significantly dampened by the transport along the sampling lines.Finally, we note that we observe both positive and negative cospectral density for NO, NO 2 and NO x fluxes at different frequencies.In Fig. 4a-c triangles refer to negative and solid lines to positive cospectral density.Few studies report sign changes in a scalar cospectrum other than that of momentum flux (Wolfe et al., 2009;DiGangi et al., 2011;Park et al., 2014).Details of the cospectral analysis along with sign changes and the discussion of their underlying physical mechanisms will be presented elsewhere (Min et al., 2014).Briefly, we find that chemical reactions forming higher oxides of nitrogen from www.atmos-chem-phys.net/14/5495/2014/The normalized cospectrum, shown in Fig. 4b, indicates the fraction of the total flux at each frequency.It is calculated as the cospectra multiplied by the frequency and divided by the covariance of temperature, NO, NO 2 or NO x , with the vertical wind, which is the integrated value under the curve.Generally, the shape of the normalized cospectrum of w NO , w NO 2 , and w NO x are similar to that of w T with a maximum in the range 0.005-0.1 Hz (200-210 s), values consistent with previous observations at this site (Farmer et al., 2006;Wolfe et al., 2009;Park et al., 2014).A steeper falloff at high frequencies (> 0.01 Hz) for w NO x and w NO 2 (especially in the afternoon, not shown) than for w T was reported in previous studies of PAN at this site (Wolfe et al., 2009), in a Loblolly pine forest (Turnipseed et al., 2006), and for HCHO in a ponderosa pine forest in Colorado (DiGangi et al., 2011) as well as for a variety of BVOCs at this site (Park et al., 2014).The peaks and valleys in w NO 2 seen at frequencies of 0.002, 0.005 and 0.015-0.03Hz are associated with valleys and peaks of w NO , which may be an indication of chemical conversion between NO and NO 2 during transport.
Figure 4c shows the absolute value of the normalized cumulative distributions of the cospectra of w T , w NO , w NO 2 , and w NO x .The cumulative distribution for w T approaches a horizontal asymptote at both ends of the spec-trum, providing additional confirmation that the sampling interval and time resolution was both long enough and fast enough to capture all the important flux-carrying eddies.The patterns of w NO x as well as w NO , w NO 2 are generally comparable to that of w T , except for frequencies in the ranges 0.1-1 Hz, where cospectra signs are changing.Here, we use the the absolute magnitude frequency weighted normalized cospectra of w NO , w NO 2 , and w NO x for the cumulation noting that the frequencies of w NO , w NO 2 , and w NO x that have a negative cospectrum vary with time of day, suggesting they are not internally generated, but are rather the result of time-of-day-dependent atmospheric processes.Detailed discussion of these features will be presented elsewhere (Min et al., 2014).

Gradients and fluxes
The diurnal variations in the concentrations of O 3 , NO x , NO, and NO 2 averaged over the whole field campaign, except the short time periods during rain events on 2 and 11 July, are shown in Fig. 5 and are similar to previous observations at this site (Day et al., 2003;Farmer et al., 2006).There is no significant difference between mean and median values, indicating the data has little skew.The patterns are affected by transport from the city of Sacramento, local emission, deposition, and chemistry.O 3 , NO x , and NO 2 increase as air is transported in from the west, carrying the remnants of emissions from Sacramento, and reach a maximum after sunset between 18:00 and 21:00.Generally, NO can be thought of as controlled by the amount of soil NO x and local photochemistry.In the morning, we observe an enhancement in NO and NO 2 and a decrease in O 3 .We observe the highest NO concentrations above the canopy, decreasing NO within the forest, and increasing NO concentrations again near the forest floor -except in the late afternoon when turbulent mixing is strongest and dry soils result in NO emissions that are at their daily minimum.This pattern is consistent with our soil flux measurements, showing the lowest signal in late afternoon.At night there is some local contamination from the propane generator at the site.
For the purpose of discussion we define an enhancement factor ( X) to be the concentration difference between each height and that measured above the canopy ( X = X i − X 18m ).Positive values of X indicate concentration enhancements, and negative values indicate depleted concentrations relative to the above-canopy value.As shown in Fig. 6, NO was depleted within the canopy (a) and NO 2 was enhanced at all times of day (b).In general, we observed the least NO depletion near the soil (except at night) and the largest NO 2 enhancement at the mid-and top-canopy heights.This pattern is qualitatively explained by emissions of NO at the soil, followed by the conversion of NO to NO 2 until the steady-state ratio is established by the reaction of NO with O 3 and photolysis of NO 2 .
If soil NO emissions and the chemical cycling of NO/NO 2 were the only two process controlling NO x , then we would expect the gradient in the sum of NO and NO 2 to be a straight line connecting the enhanced concentration at forest floor with the above-canopy (boundary layer) value so long as sufficiently strong turbulent mixing exists (Vila-Guerau de Arellano et al., 1993;Gao et al., 1991;Jacob and Wofsy, 1990).This, however, is not what we observe (Fig. 6c).Rather we observe a NO x enhancement within the canopy during the day and depletion at night.In addition, the enhancement at the forest-floor, mid-canopy, and top-canopy heights changes within different time windows, indicating the existence of processes other than soil NO emission and interconversion of NO and NO 2 .For example, the within-canopy gradients at 06:00-09:00 (blue) and 14:00-18:00 (magenta) are opposite each other; the 06:00-09:00 NO x gradient indicates the existence of a NO x sink process as the height increases from mid-to top-canopy, while the 14:00-18:00 gradient indicates a NO x source.While there have been a number of indirect lines of evidence for the idea that processes other than soil NO emission and NO/NO 2 photochemical partitioning affect NO x fluxes (Jacob and Wofsy 1990;Yienger and Levy 1995;Wang and Leuning 1998;Lerdau et al., 2000;Wolfe et al., 2011;Min et al., 2012;Seok et al., 2013), to our knowledge these observations provide the first direct observational evidence.
In Fig. 7a-c, we show the eddy covariance fluxes of NO, NO 2 and NO x as well as sensible heat.In this figure and for the remainder of the analysis, we include data only when F NO , F NO 2 and F sht (flux of sensible heat) are all available.We observed upward fluxes from 09:00 to 15:00.Fluxes of NO and NO x were slightly downward from 06:00-09:00.The midday (12:00-14:00) median fluxes of NO, NO 2 and NO x are 0.32 ± 0.27, 0.67 ± 0.21 and 1.0 ± 0.43 ppt m s −1 .A comparison of the direction of the observed flux of NO 2 and NO x to the gradients in Figures 6b and c gives a picture of molecule movement consistent with standard ideas of turbulent transport moving material from a region of high to low concentration.For NO, the direction of the flux is counter to that of the concentration gradient.
The observed midday fluxes of NO x of 1.0 ± 0.43 ppt m s −1 are similar to midday soil NO emission measured at this site in the late afternoon of 0.05-0.8ppt m s −1 .Larger fluxes were observed in the morning, 2.6-5.2ppt m s −1 , and much larger fluxes were measured after rain, 21.6 ppt m s −1 .These soil NO x fluxes are at the low end of measurements in the region -which were all made much closer to urban centers where N deposition to the soils is much larger.Our own chamber measurements were in a nearby clearing and may have been in soil that was drier, and hence with lower soil NO x emission than is representative of the fetch at the site.At the Schubert Watershed at the Sierra Nevada Foothill Research and Extension Center in an oak forest, soil NO x emissions of 5.8-15 ppt m s −1 were observed during the summer (Herman et al., 2003) and typical soil NO fluxes reported for other locations in California are 2-20 ppt m s −1 (Anderson and Poth, 1989;Davidson et al., 1993).
To evaluate the contribution of soil NO emission to the NO 2 flux by the reaction with O 3 , we divide our system into a soil NO emission layer and a chemical conversion layer, where the latter layer includes the within-and abovecanopy measurement heights.The noontime (12:00-14:00) ratio of NO and NO 2 in the conversion layer (estimated as the averaged value over all measured heights) is 1 : 4.7.Thus, after reaching steady state with ozone, we expect 0.82 NO 2  molecules for every NO molecule emitted from the soil emission layer.This is a lower limit, at other times of day there can be as much as 1 NO 2 per NO.The NO emitted has a negligible effect on the O 3 flux consistent with previous analyses of ozone flux (Kurpius and Goldstein, 2003).Additional details about O 3 flux measurements at this site can be found elsewhere (Fares et al., 2010).
Using the lowest measured soil NO emission rate measured in the morning as well as the observed O 3 at this site, we calculate the conversion of NO to NO 2 induces a 2.38 ppt m s −1 NO 2 flux, a number which is 3.5 times larger than the observed NO 2 flux (0.67 ± 0.21 ppt m s −1 ), another piece of evidence supporting the existence of a canopy reduction process for NO 2 .

Analysis
For a conserved tracer, the direction and magnitude of the flux is controlled by the local concentration gradient and the rate of vertical mixing -the tracer moving from high to low abundance and the rate of movement determined by the strength of both the gradient and the turbulent mixing.This concept is known as flux-gradient similarity or Bowen ratio theory, and is often used for estimating the exchange rate of non-reactive (conserved) species from their concentration gradient (e.g., Mayer et al., 2011 and references therein).Similarity theory holds for conserved tracers when observations of fluxes are made above the roughness sublayer (Raupach and Legg, 1984) and this criterion was met at BEARPEX 2009 (18 m flux measurement height with 8.8 m mean tree height), as evidenced by a comparison of the flux data of the sensible heat and 3 biogenic VOC species, methanol, 2-methyl-3-butene-2-ol +isoprene, and monoterpene, with longer chemical lifetimes than the turbulent transport time (Park et al., 2014).
The fluxes of reactive species cannot be completely described through simple application of similarity theory.This is because reactive species will, to some extent, undergo chemical transformations faster than they will be transported by turbulent diffusion (Vila-Guerau de Arellano et al., 1993;Gao et al., 1991;Jacob and Wofsy, 1990).However, similarity theory is still a powerful tool, as quantifying the flux due to turbulence transport allows for the estimation of the effects of other within-canopy chemical processes.
A visual representation of the idea proposed in this study is shown in Fig. 8.The red line shows the gradient expected if flux-gradient similarity holds, and the blue line shows the gradient if the concentration is chemically, or otherwise, perturbed.This concept is known as localized near field (LNF) theory (Vandenhurk and Mcnaughton, 1995;Raupach, 1989).We apply LNF to conserved tracers and reactive molecules to test (1) whether the flux-gradient theory is valid within the canopy at this site and (2) whether the chemical reactivity or some other canopy processes affects the reactive chemicals.In the analysis below, we will assess the within-canopy behavior of H 2 O as conserved tracer and NO, NO 2 , and NO x through pictorial relationships analogous to Fig. 8.In Fig. 8, the green dashed line divides two layers: a within-canopy layer and an above-canopy layer.C ABOVE is the measured concentration in the above-canopy layer and C WITHIN is the measured concentration within the canopy.Using similarity theory and the measured fluxes, we calculate C CONSERVED , the concentration that would be observed for a conserved tracer.The difference between C WITHIN and C CONSERVED defines C DELTA , which represents the contribution to the concentration by non-conservative processes that act to perturb the flux gradient relationship.To quantify C CONSERVED we use flux-gradient similarity, Eq. ( 2), (Meyer, 1986) and the mixing rate, K, the so-called eddy diffusivity constant, which is inferred from the observed sensible heat flux and temperature gradient.
In an illustrative test of our approach we present our results for the conserved tracer, water.In Fig. 9, the blue circles represent the measured gradient of H 2 O and the red circles show the gradient inferred from the H 2 O flux (C CONSERVED ) in the within-canopy layer at midday (from 12:00 to 14:00).The difference between C WITHIN and C CONSERVED in the withincanopy layer, C DELTA , shown as a black arrow, is small (1 % relative to C ABOVE and within the concentration measurement uncertainty of 3 %; the estimated propagated error is 15 % based on 10 % error in sensible heat flux calculation and 10 % error in K estimation).The small difference of C DELTA,H 2 O is possibly due to the source/sink process difference with temperature and H 2 O.However, the magnitude of C DELTA is smaller than the estimated uncertainty, so we conclude that the sources/sink difference in H 2 O and temperature are not detectable and the flux-gradient similarity holds for conserved tracer even within canopy at this site.This implies there is no measurable additional source/sink process(es) for H 2 O, aside from turbulent transport, and the open canopy structure allows us to use flux-gradient similarity using within-canopy information.Similar results were obtained for CO 2 and several slowly reacting BVOCs (Park et al., 2014).
Applying the same analysis to NO and NO 2 (Fig. 10), we find C DELTA is large compared to the measurement variability.C DELTA for NO is −12.4 ± 3.3 ppt (23 % relative to C ABOVE ) and for NO 2 is 64.7 ± 4.7 ppt (44 % relative to C ABOVE ).We reach an identical conclusion, with slight numerical differences, if we reference the calculation to the canopy top height instead of the average through the canopy, finding 27 and 39 % differences for NO and NO 2 compared to the C ABOVE , respectively.The estimated uncertainty in the C DELTA calculation is 30 and 25 % (3.7 and 16.2 ppt) for NO and NO 2 based on the error propagation.The paired t test also shows statistically meaningful differences between C WITHIN and C CONSERVED for NO and NO 2 (p < 0.01).Examples of C DELTA analysis from individual days are shown in Sect.S2 of the Supplement.Based on the observed gradient of NO, standard fluxgradient similarity predicts the downward flux of NO; however, we observed an upward flux of NO (Fig. 7).This counter-gradient flux can only be explained by the formation of NO during the transport process from within the canopy layer to the above-canopy layer.Fig. 10a indicates that to explain the observed NO flux, we need to account for 12 ppt (C DELTA ) more NO molecules than were observed in the canopy layer.This is reasonable, as photolysis rates above the canopy should be faster than in the shade of the canopy.If the required NO were completely due to NO 2 photolysis it would correspond to ∼ 12 ppt removal, or 20 % of the C DELTA of NO 2 .The remaining 80 % of C DELTA in NO 2 , 52.3 ppt, must be accounted for via other mechanisms.
To evaluate the contribution of photolysis of NO 2 to the counter-gradient flux of NO, we calculate the chemical conversion rate integrated over the 100 s (τ turb ) as Eq.(3).
The photolysis rate, j NO 2 , is calculated with the Tropospheric Ultraviolet and Visible (TUV) radiation model scaled to the measured PAR.We treat RO 2 as equal to HO 2 .Using the measured concentrations of NO, NO 2 , O 3 , HO 2 , and temperature, we estimate a net loss of 22.8 ppt (over 100 s) of NO 2 , which is in excess of that needed to explain the NO countergradient flux of 12.4 (±3.3) ppt.
The large value of C DELTA for NO x (54.3 ± 5.9 ppt, 29 % relative to C ABOVE ) indicates the necessity of one or more  within-canopy loss processes.To explore the mechanism(s) controlling the C DELTA for NO x , we examine several chemical processes related to PNs, ANs, HNO 3 , and HONO using our two-layer model (concept shown in Fig. 8).The magnitudes of each of the near-field processes for NO x were inferred using Eq. ( 4) to estimate the contribution of certain processes on the ∼ 100 s timescale of turbulent mixing (τ turb ).
Here, L X (P X ) is the loss (production) rate of species X happening within the time window of turbulent air movement from within the canopy (height z 1 ) to above the canopy (height z 2 ).We chose 4.4 m, the middle level of the canopy, as a representative of z 1 , and 18 m above-canopy layer as representative of z 2 .PNs can act as either a net source or a sink of NO x through thermal dissociation (+1 NO 2 molecule) or PN formation (−1 NO 2 ).Calculating the steady-state chemical production and thermal and chemical loss of PAN (LaFranchi et al., 2009;Wolfe et al., 2009), yields 5.3 ppt of NO 2 formed in 100 s.This mechanism implies an enhancement of NO x within the canopy, as discussed in more detail in Min et al. (2014).However, we have also suggested that a local biogenic precursor drives PN formation within canopy (Min et al., 2012).This BVOC PN species, denoted XPN, exhibited an upward flux and is a candidate for NO 2 loss not included in our steady-state calculation.We estimate the flux of this XPN to be 2.3 ± 0.4 ppt m s −1 corresponding to 16.7 ppt of NO 2 loss within canopy and explaining 31 % of the NO x C DELTA .BVOC-driven AN formation from OH initiated chemistry can be calculated as where Here, α i and γ i stands for the branching ratio of AN formation from RO 2 and NO reaction, and the fraction of RO 2i from VOC i reacts with NO.Also, k isom refers to the unimolecular isomerization rate of RO 2i .We estimate the effects of MBO, monoterpenes and isoprene (important BVOCs at the BEARPEX site (Bouvier-Brown et al., 2009a;Schade et al., 2000) on AN production (for 100 s) and calculate that 3.1 ppt (5.7 %), 0.4 ppt (0.7 %) and 6.9 ppt (12.8 %) NO x is removed by AN formation, respectively.We use a 10, 11.7 and 18 % branching ratio (α i ) for MBO (Chan et al., 2009), isoprene (Paulot et al., 2009) and monoterpenes (Paulot et al., 2009).The rate constants and mechanisms for RO 2 +HO 2 , RO 2 +NO and RO 2 +RO 2 were taken from the Master Chemical Mechanism (MCM) v3.2 (Jenkin et al., 1997;Saunders et al., 2003) and isomerization rates for isoprene from Crounse et al. (2011).If we consider AN formation from ozonolysis reactions of very reactive BVOCs, such as sesquiterpenes in analogy to XPN formation through the channel described as BCSOZNO3 in MCM v3.2, we estimate an additional 16.8 ppt (31.1 %) of NO x consumed over 100 s.These calculations indicate chemical formation of nitrates is rapid enough to affect the fluxes.Formation of HNO 3 is also a potentially important pathway for the removal of NO x from the system by wet and dry deposition after formation.The gas-phase reaction of NO 2 with OH is the major source of HNO 3 formation (heterogeneous formation HNO 3 from NO 2 hydrolysis is slower by an order of magnitude than the gas-phase formation rate), although hydrolysis of tertiary organic nitrates may also be important (Darer et al., 2011;Hu et al., 2011;Browne et al., 2013).The production rate can be calculated as Using the measured OH concentration, we estimate 0.6-2 % of NO x is lost through gas-phase reaction at this site.Compared to ANs and PNs, HNO 3 formation by OH reactions is unimportant.HONO formation is another candidate for altering the flux of NO x .HONO flux measurements at this site were observed to be small (−0.11 ± 0.69 ppt m s −1 ) and slightly downward, contributing to the enhancement of NO x within canopy rather than loss.
Direct uptake through plant stomata might be responsible for the remaining NO 2 removal within the canopy.We estimate < 2.5 ppt of NO 2 is removed through plant leaves at the typical daytime NO 2 concentration and canopy conductance (< 1 cm s −1 with 3.7 m 2 m −2 ) assuming uptake rates similar to those reported in recent field and laboratory stud-ies (Chapparro-Suarez et al., 2011;Breuninger et al., 2012;Breuninger et al., 2013).However, the daytime NO 2 concentration at this site of less than few hundred ppt suggests this site is a regime where NO 2 emissions from plant biota likely dominate, consistent with the results of Breuninger et al. (2013) who estimate 0.05-0.65 ppb as NO 2 compensation point in a Norway Spruce forest.Further evidence for an NO 2 compensation point from canopy-scale observations will be presented in Min et al. (2014).
Taken together, as much as 86 % of the C DELTA for NO x (Table 2) can be explained by local chemical NO x loss mechanisms and the formation of higher nitrogen oxides.Given the uncertainties, it is reasonable to interpret this as indicative that all of the C DELTA is due to within-canopy chemistry.This leads us to suggest a conceptual model for biosphereatmosphere exchange of NO x as shown in Fig. 12.In addition to the previously suggested within-canopy process of NO x (Fig. 1), chemical pathways are added, converting NO x to higher oxides of nitrogen.These pathways are alternative mechanisms to plant uptake that have the effect of reducing the soil NO that escapes the forest canopy.The direction and magnitude of higher nitrogen oxides fluxes in this coupled mechanism are then the net resultant of upward (owing to formation within canopy) and downward (deposition from the atmosphere) fluxes of each class.

Conclusions
During the BEARPEX 2009 field experiment, we observed upward fluxes of NO and NO 2 using eddy covariance flux measurements, along with large NO 2 and NO x concentration enhancements within the canopy, and counter-gradient fluxes of NO.Applying standard flux-gradient relationships to interpret the data indicates the existence of one or more NO x loss processes within the canopy, in addition to conversion of NO to NO 2 by reaction with O 3 .We interpret these results as observational evidence for an ecosystemscale chemical canopy reduction process and suggest these reactions may be a partial mechanistic explanation for the "canopy reduction factor" that has been relied on to reconcile www.atmos-chem-phys.net/14/5495/2014/Atmos.Chem.Phys., 14, 5495-5512, 2014 discordance between leaf-level, soil-level, and atmospheric modeling studies.We investigate multiple chemical and ecophysiological processes to explain the NO x removal during vertical transport and conclude that the chemical formation of PNs and ANs (and their vertical transport before they cycle back to NO x ) are the primary mechanisms responsible -implying that the reactive nitrogen does escape the canopy and may be returned as NO x by further chemistry downwind.
The Supplement related to this article is available online at doi:10.5194/acp-14-5495-2014-supplement.

Figure 1 .
Figure 1.Schematic of the various interactions involved in the exchange of nitrogen oxides between the atmosphere and the forest canopy.Bold arrows in blue (downward) and red (upward) represent the direction of the flux of each species across the canopy surface.

Figure 3 .
Figure 3. Lag calculation of (a) w'T', (b) w'NO' and (c) w'NO2'.Highest normalized correlation between wind and temperature, NO or NO2 were observed as expected; 0 second for wind and temperature, 1.4 and 2.6 second for NO and NO2.

Figure 3 .
Figure 3. Lag calculation of (a) w T , (b) w NO and (c) w NO 2 .Highest normalized correlation between wind and temperature, NO or NO 2 were observed as expected; 0 s for wind and temperature, 1.4 and 2.6 s for NO and NO 2 .

Figure 4 . 9 Figure 4 .
Figure 4. Equally spaced logarithmic averaged (150 bins) absolute cospectral density (a), 2 frequency weighted cospectrum (b), and normalized cumulative distributions of the cospectra 3 of temperature (magenta) NO (green), NO 2 (blue) and NO X (yellow) (c) with vertical wind 4 from 11:00-12:00 through out the whole field campaign when the chemical perturbation is 5 small.Closed triangles represent the absolute value of the negative cospectral density, which 6 has the opposite sign to the general flux direction.The black dotted lines in (a) are lines with 7 slopes of −7/3 and −5/3 (see related text).8 9

Figure 5 .
Figure 5.Diurnal patterns of O3 (south tower), NOX, NO, and NO2.The data are one-hour mean values and the error bars represent the variation defined as the observed variability (± 1σ) divided by square root of the number of measurements in that time bin.Colors represent the measurement heights: above canopy (18 m) in black, top canopy (9 m) in blue, middle canopy (5 m) in green, and forest floor (1.5 m for O3 and 0.5 m for NO and NO2) in magenta.

Figure 5 .
Figure 5.Diurnal patterns of O 3 (south tower), NO x , NO, and NO 2 .The data are 1-hour mean values and the error bars represent the variation defined as the observed variability (±1σ ) divided by square root of the number of measurements in that time bin.Colors represent the measurement heights: above canopy (18 m) in black, top canopy (9 m) in blue, middle canopy (5 m) in green, and forest floor (1.5 m for O 3 and 0.5 m for NO and NO 2 ) in magenta.

Figure 6 .
Figure 6.Vertical gradients of NO, NO2, and NOX.Dot and whiskers represent means and standard errors of the mixing ratio enhancement at each height.The colors represent the enhancement at 6 different times of day through the complete diurnal cycle: early morning

Figure 7 .
Figure 7.Diurnal patterns of the NO, NO2, NOX, and sensible heat fluxes in panels a, b, c, and d, respectively.All fluxes are upward indicating molecular motion from forest to atmosphere.Median midday (12:00-14:00) fluxes are 0.32 ± 0.27, 0.67 ± 0.21, and 1.0 ± 0.43 ppt m/s for NO, NO2, and NOX and that of sensible heat flux is 0.21 ± 0.08 o C m/s. Black lines represent means and the gray areas give 25-75 % of flux data for hourly bins.

Figure 7 .
Figure 7.Diurnal patterns of the NO, NO 2 , NO x , and sensible heat fluxes in panels (a), (b), (c), and (d), respectively.All fluxes are upward, indicating molecular motion from forest to atmosphere.Median midday (12:00-14:00) fluxes are 0.32 ± 0.27, 0.67 ± 0.21, and 1.0 ± 0.43 ppt m s −1 for NO, NO 2 , and NO x and that of sensible heat flux is 0.21 ± 0.08 • C m s −1 .Black lines represent means and the gray areas give 25-75 % of flux data for hourly bins.

Figure 8 .Figure 8 .
Figure 8. Schematic of our two-layered model based on the localized-near field, LNF, concept.3 CABOVE represents the reference concentration, which for the above-canopy layer is defined as 4 the measured concentration at height 18m.CWITHIN is the measured concentration within the 5 canopy defined as the averaged concentration at 0.5, 5, and 9m.CCONSERVED is the estimated 6 concentration based on the measured eddy-covariance flux and the eddy diffusivity calculated 7 from sensible heat flux.CDELTA is the difference between CWITHIN and CCONSERVED and is a 8 measure of the importance of non-conservative processes.9 10

Figure 9 . 8 Figure 9 .
Figure 9.The estimated concentration, C CONSERVED (red) using standard flux-gradie 3 similarities of H 2 O during midday (12:00-14:00) is shown.Blue represents the measure 4 vertical gradient.Open circles and whiskers represent the mean and the variability of H 2 O 5The difference between blue and red in the within-canopy level is shown in black arro 6indicates C DELTA and is negligibly small as expected for a conserved species.7 8

Figure 10 .
Figure 10.The estimated concentration, C CONSERVED , (red) using standard flux-gradient similarities of NO and NO 2 at midday (12:00-14:00 and the measured vertical gradient (blue), giving values for C DELTA of 12 ppt (NO) and 64 ppt (NO 2 ).

Figure 11 .Figure 11 .
Figure11.The estimated concentration, C CONSERVED , (red) using standard flux-g 2 similarities of NO X at midday (12:00-14:00) and the measured vertical gradient (blue) 3 values for C DELTA of 54 ppt (NO x ).Open circles and whiskers are the mean and standar 4Figure11.The estimated concentration, C CONSERVED , (red) using standard flux-gradient similarities of NO x at midday (12:00-14:00) and the measured vertical gradient (blue) giving values for C DELTA of 54 ppt (NO x ).Open circles and whiskers are the mean and standard errors.

Figure 12 .
Figure 12.Schematic of the interactions involved in the exchange of nitrogen oxides between the atmosphere and the forest canopy as identified in this manuscript.Bold arrows in blue (downward) and red (upward) represent the direction of the flux of each species across the canopy surface.Red thin arrows within canopy indicate the NO x removal processes within the canopy in addition to plant uptake.

Table 1 .
Estimated flux analysis errors.
a Errors over half an hour.b Estimated form the ratio of covariances at true lag and several lag times far from the true lag (±230-250 s).

Table 2 .
Possible within canopy NO x consumption mechanisms.
Negative consumption indicates a source of NO x within the canopy.