AtmosphericChemistryandPhysics Observations of atmosphere-biosphere exchange of total and speciated peroxynitrates: nitrogen ﬂuxes and biogenic sources of peroxynitrates

. Peroxynitrates are responsible for global scale transport of reactive nitrogen. Recent laboratory observations suggest that they may also play an important role in delivery of nutrients to plant canopies. We measured eddy covariance ﬂuxes of total peroxynitrates ( 6 PNs) and three individual peroxynitrates (APNs ≡ PAN + PPN + MPAN) over a pon-derosa pine forest during the Biosphere Effects on AeRosols and Photochemistry EXperiment 2009 (BEARPEX


Please provide feedback
Please support the ScholarWorks@UMBC repository by emailing scholarworks-group@umbc.edu and telling us what having access to this work means to you and why it's important to you.Thank you.

Introduction
Peroxynitrates, PNs (RO 2 NO 2 ≡ acyl peroxynitrates + nonacyl peroxynitrates) represent 25 % or more of the reactive nitrogen in the troposphere.PNs are thermally unstable at characteristic surface temperatures but have very long lifetimes (days to weeks) aloft.As a result, peroxynitrates are temporary reservoirs of NO x (≡ NO + NO 2 ) affecting regional and global tropospheric ozone formation (Hudman et al., 2004;Zhang et al., 2008;Parrish et al., 2009;Fiore et al., 2011) by sequestering NO x at the source and then releasing it in downwind rural and remote regions.PNs may also have a significant influence on climate via their effects on available nitrogen and thus ecosystem productivity (Bytnerowicz and Fenn, 1996;Goulding et al., 1998;Sparks et al., 2003;Teklemariam and Sparks, 2004;Sparks, 2009) and via their phytotoxicity and thus ecosystem health (Okano et al., 1990;Goulding et al., 1998).Recent research has brought attention to new mechanisms by which peroxynitrates might participate in linking the atmospheric and biospheric N cycles (Townsend et al., 1996;Holland et al., 1997;Teklemariam and Sparks, 2004;Sparks, 2009).
Measurements at the leaf scale (Okano et al., 1990;Sparks et al., 2003) and studies assuming assume flux-gradient similarity (Doskey et al., 2004) have suggested that the direct uptake of PAN (CH 3 C(O)OONO 2 ) by plants is fast and controlled by stomatal processes.Early work focused only on PAN's phototoxicity (Taylor, 1969;Ordin et al., 1971;Okano et al., 1990;Oka et al., 2004) assuming exclusively negative impacts due to foliar uptake on ecosystem health (Bytnerowicz and Fenn, 1996).In contrast, Sparks (2009) pointed out that direct foliar uptake of atmospheric reactive nitrogen to biota could be a major source of global N input to ecosystems.Teklemariam and Sparks (2004) estimated as much as 20 % of regional dry deposition of N occurs via foliar uptake of PAN and that this addition could serve to enhance productivity (Teklemariam and Sparks, 2004).To the extent that stomatal processes dominate, PN deposition will occur according to spatial and temporal patterns that are different from HNO 3 , as HNO 3 deposition is thought to be controlled only by turbulent transport to surfaces and, as a result, PN deposition could be an even larger percentage of the total in some locations.
New measurement techniques have enabled the use of eddy-covariance methods (Turnipseed et al., 2006;Farmer et al., 2006;Wolfe et al., 2009) providing more detailed observations of the rate of exchange of PNs between the atmosphere and biosphere.Analyses of these recent experiments indicate that a complex interplay of chemistry and turbulent exchange occurs, making it difficult to isolate direct foliar uptake, but providing information about a variety of other chemical and micrometeorological processes.All of these new experiments suggest that our current understanding of the mechanisms governing ecosystem exchange of PNs is poor.Turnipseed et al. (2006) found the deposition of PAN to be four times faster than estimated based on a resistance analogy to H 2 O. Farmer et al. (2006) found that the net flux of total PNs ( PNs) were upward -contrary to all prior expectations.Wolfe et al. (2009) observed downward fluxes of speciated PNs, at the same site as Farmer et al. (2006) but three years later, and observed that three individual PNs each deposits at a different rate (Wolfe et al., 2009).
Building on the prior experiments by Farmer et al. (2006) and Wolfe et al. (2009), we measured fluxes and gradients of total and speciated peroxynitrates ( PNs and APNs, respectively) above a ponderosa pine forest during the Biosphere Effects on AeRosols and Photochemistry EXperiment 2009 (BEARPEX 2009).Our goal was to develop an observational database that could provide insights into the chemistry and ecosystem biology that affect forest-atmosphere PN exchange.These measurements provide evidence for processes that result in both the upward and downward exchange of PNs and indicate that the net is a subtle balance between these opposing processes.They also provide evidence for PN formation from one or more unidentified VOC, within the forest canopy.These molecules, hereafter denoted XPNs, are then transported to the atmosphere above.

Research site and instrumentation
The observations used in this work were collected at a research site located on the western slope of the Sierra Nevada Mountains from 15 June to 31 July in 2009 as a part of the BEARPEX field campaign.An overview of this experiment is found in Cohen et al. (in prep.) and in the articles included in this special issue of ACP.The BEARPEX research site was located 75 km downwind of Sacramento, California at a managed ponderosa pine plantation owned by Sierra Pacific Industries and near the University of California at Berkeley Blodgett Forest Research Station (38 • 53 42.9 N, 120 • 3757.9W, and 1315 m).In the summer (May to September), wind patterns at the site are controlled by the topography and are characterized by southwesterlies (210-240 • ) during daytime and northeasterlies (30 • ) at night.More detailed descriptions of the local and regional meteorology are available elsewhere (Dillon et al., 2002;Cahill et al., 2006;Choi et al., 2011;Bianco et al., 2011).The forest at this site is dominantly Pinus ponderosa L., planted in 1990, with some Douglas fir, white fir, and incense cedar.The canopy understory is largely mountain whitethorn (Ceanothus cordulatus) and manzanita (Arcostaphylos spp.) shrubbery (up to 2m in height).The average canopy heights in 2004, 2006, and 2009 were 5.1 m, 7.9 m, and 8.8 m, respectively and the leaf area indices (LAI) determined to be 2.1 m 2 m −2 , 3.2 m 2 m −2 and 3.7 m 2 m −2 by onsite surveys conducted on 8 April 2004, 4 October 2006, and 17 July 2009.At the BEARPEX site, there were two walkup towers separated by a distance of 10 m, a 15m tower (south tower) and an 18 m tower (north tower).On the south tower, meteorological parameters, including temperature, relative humidity, wind speed, net radiation, photosynthetically active radiation (PAR), and chemical species (water vapor, CO 2 , and O 3 ) were monitored at 5 heights (1.2, 3.0, 4.9, 8.75, and 12.5 m).Fluxes of water vapor, CO 2 and O 3 were also measured at 12.5 m.On the north tower, vertical gradients of meteorological parameters were also monitored, including temperature, relative humidity and wind speed.Chemical species observed on the north tower, or on an adjacent height adjustable lift, included NO, NO 2 , HONO, PNs and PAN, PPN (CH 3 CH 2 C(O)OONO 2 ), MPAN (CH 3 C(CH 2 )C(O)OONO 2 ), total alkyl nitrates ( ANs) and several speciated ANs, HNO 3 , OH, HO 2 , total OH reactivity, O 3 , numerous BVOCs, formaldehyde, glyoxal, methylglyoxal, and several organic peroxides.Aerosol chemical and physical properties were also measured.While some measurements were made only at the 17.5 m height many others were also collected at one or more of the following heights 0.5, 2.4, 5.4, 9.2 and 13.3 m.For simplicity, we refer to these measurement heights as 0.5, 2, 5, 9, 13 and 18 m in the remainder of this text.Needle temperature, soil moisture, soil temperature, and soil heat  1.The inlets for the flux measurements used in this study shared a single sonic anemometer (Campbell Scientific CSAT3 3-D Sonic Anemometer) located 0.2m horizontal and < 0.01 m vertical displacement from TD-CIMs inlet and with the same 0.2 m horizontal and 0.03 m vertical displacement from the TD-LIF inlet.
Thermal dissociation -laser induced fluorescence (TD-LIF) was used for the measurement of NOyi (NO 2 , PNs, ANs and HNO 3 ).Briefly, thermal dissociation of each of these classes of higher oxides converts them to NO 2 and a companion radical at 180 • C for PNs, 350 • C for ANs, and 600 • C for HNO 3 (Day et al., 2002).NO 2 is then detected by LIF (Thornton et al., 2000).Two autonomous homebuilt TD-LIF systems were deployed at BEARPEX 2009 for the simultaneous measurement of fluxes and gradients.In both systems, excitation of a 585 nm rovibronic feature of NO 2 was provided by a frequency doubled Nd:YAG laser (Spectra Physics, average power of 2 W at 532 nm, 30 ns pulse length) pumping a custom-built tuneable dye laser operating at 8 kHz.The fluorescence signal long of 700 nm was imaged onto a red sensitive photocathode (Hamamatsu H7421-50) and gated photon counting techniques (Stanford Research Systems, SRS 400) were employed to discriminate against prompt background signals.Laboratory measurements and in-field comparisons showed the two TD-LIF instruments to have calibrations that were identical to within 3 % (± 2.5 %, 1σ , R 2 = 0.92).
The APNs, PAN, PPN, and MPAN, were measured by thermal dissociation -chemical ionization mass spectrometry (TD-CIMS) with an iodide reagent ion source.A detailed instrument description, including a discussion of the specific configuration for the BEARPEX measurements, is presented in Wolfe et al. (2009).Briefly, thermally dissociated acyl peroxy (PA) radicals are reacted with iodide ion to form carboxylate anions, which are detected using a quadrupole mass spectrometer.Masses corresponding to PAN (m/z = 58.7),PPN (m/z = 72.7)and MPAN (m/z = 84.7)were detected at 3.3 Hz time resolution.The sampling inlets for PNs, as well as NO 2 , ANs, and HNO 3 , and APNs were co-located.The time sequence of the observations is shown in Fig. 1.Fluxes from the TD-LIF system measurements were measured at 18 m during the first 30 min from the 3rd min to the 30th min and the last 30 min from the 33rd min to the 60th min each hour (Fig. 1a).NO yi measurements at the other heights were collected by switching between the 9, 5, and 0.5m heights sampling at each height for 2 min (Fig. 1. row b).
TD-CIMS measurements had two different modes, a flux divergence mode (Fig. 1. row c) and a normal mode (Fig. 1,  row d).During the flux divergence mode (day of year 165-183), concentrations at 18 m and 13 m were measured for 15 min each, alternating between either the first or last 30 min of the hour and measuring concentrations at 18, 13, 9, 5, 2 and 0.5m for 3 min each for the other 30 min of each hour.After day 183, fluxes were only monitored at 18 m, from the 0th min to the 25th min (Fig. 1. row d).
The data used in this study are from days after 176 when the TD-CIMS measurements were more reliable.Data affected by exhaust plumes from a nearby propane power generator (mostly at night) and the occasional wafts of car exhaust were removed prior to analysis.These spikes are defined as variations in the PN or APN 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, NO, and NO 2 and removed by hand.
To ensure the time synchronous comparison of measurements by multiple instruments, 30-min data resolution was calculated as the averaged value of the 0th to the 30th min and from the 30th to the 60th min at each height and is representative of the first and the last halves of each hour.For the vertical gradients in Fig. 7, 1-h data resolution was achieved by averaging from the 0th to 60th min.
The comparison of daytime (09:00-18:00) measurements of PNs and APNs at 18 m is shown in Fig. 2. The measurements are identical to within 5 %, and a least-square linear fit, forcing the intercept through zero, yields a slope of 1.02 ± 0.026 with R 2 = 0.92.If we allow the intercept to vary we find a slope of 1.047 ± 0.038 and an 8ppt intercept, however there is no improvement in the R 2 .The quality of the comparison between PN and APN measurements is consistent with previous studies (Wooldridge et al., 2010).It should be noted that attempts to cross-calibrate with a pure PAN source in the field were unsuccessful and that each instrument was calibrated independently with respect to a different reference gas.

EC flux calculation
Detailed procedures for flux calculations are described elsewhere for PNs (Farmer et al., 2006) and APNs (Wolfe et al., 2009).Briefly, 3-D winds are rotated using the 2-step natural wind coordinate rotation, and concentrations data are de-spiked and de-trended using a 10-min running mean.The time lag between wind and concentration signals is calculated by shifting the concentration time series relative to the winds and optimizing the covariance calculated by Eq. ( 1).Since this lag should depend only on the physical setup (particularly the inlet residence time), a single average lag was applied to all data.The eddy-covariance flux, F c , of species c is calculated using the covariance between the vertical wind speed, w, and the mixing ratio of c by Eq. (1).
In this study, c corresponds to PNs, PAN, PPN, and MPAN, and w represents the vertical wind speed.The primes in Eq. ( 1) are the deviation from the mean, the subscripts i refer to individual fast time resolution measurements, and the bar indicates the mean of n data points over the course of a single measurement period (15-30 min).Several filters were applied to the derived fluxes to assure they are accurate.To assess stationarity (e.g.whether a 30min flux was representative of the average surface exchange over the sampling period), each flux measurement period (∼30 min) was divided into 5 equal periods and the averaged flux of each subset, F sub , was compared to the 30-min total flux, F 30 min .If the ratio F sub /F 30 min differed from unity by more than 30 % then the measurement period was determined to be non-stationary and that entire half hour was removed from the dataset (Foken and Wichura, 1996).Flux data with large tilt angles (>5 • ) were also rejected (Lee et al., 2004).We include only observations with frictional velocities between 0.1 m s −1 and 1.5 m s −1 (e.g.Foken and Nappo, 2008).The lower limit in this window insures that turbulence is sufficient for reliable estimates of the flux at this site (Farmer et al., 2006).Use of a lower limit of 0.05 or 0.2 m s −1 does not affect our conclusions.After the application of these criteria for reliable observations, 67 % of daytime and 61 % of nighttime data remained and were used to calculate the flux.
The flux of the sum of individual APNs, F APNs , is defined as the sum of the separately calculated PAN, PPN, and MPAN fluxes.We calculated the PN flux, F PNs , using the concentration difference between ambient (NO 2 ) and the 180 • C (NO 2 + PNs) channel.Nearly identical results are obtained if we calculated the flux of NO 2 and the flux of (NO 2 + PNs) taking the difference to be the PN flux.
We characterize the uncertainty budgets of the two instruments and the difference in concentration and flux of the two instruments as a combination of systematic and random terms.The systematic terms are ones we believe would be corrected either by application of a single multiplicative constant to the observations of each instrument or that result from imperfect correction for finite instrument response time and sensor separation.The random terms are associated with finite precision of the measurements.Many of these systematic terms are common to both instruments and thus cancel in the difference, since we show the two instruments measure nearly identical concentrations at noon.The individual elements are described in Table 2 and detailed procedure can be found elsewhere (Farmer et al., 2006;Wolfe et al., 2009;Min et al., 2012).Briefly, the systematic uncertainty in F PNs is estimated to be < 14 % calculated as the square root of the sum of squares of the uncertainties in the instrument calibration (< 10 %); sensor separation and inlet dampening (< 2 %), instrument response time (< 8 %) and line-locking cycle for the dye laser data acquisition scheme (< 3 %).The systematic uncertainty in F APNs measured by TD-CIMS is estimated to be < 25 % calculated as the square root of the sum of squares of the uncertainties in the instrument calibration (< 22 %), sensor separation and inlet dampening (< 2 %) and instrument response time (< 12 %).The concentration calibration errors do not contribute to the systematic uncertainty in the flux difference between F PNs and F APNs .We calculate this uncertainty to be ∼7 %, taking into account the common uncertainties in F PNs and F APNs and the observation that the measured concentration differences are less than 5 % (e.g Fig. 2).Fitting the measurements shown in Fig. 2 to a line and allowing the intercept to vary results in an estimate of 4.7 % difference in the concentrations observed by the two instruments.The difference is 2 % when the intercept is fixed at zero.Allowing the intercept to vary in the fit did not reduce the R 2 .We take the 5 % number to be conservative, but the 2 % number would be a reasonable choice.
The random error terms in flux analysis follow counting statistics (< 10 % for F PNs and < 20 % for F APNs over half hour average flux measurements) which drastically reduced with number data point averaging (followed by Farmer et al., 2006 andWolfe et al., 2009).The total random error is estimated from the variance of the flux calculated over a range of lag times far from the true lag (Ruuskanen et al., 2011).Here we define our flux detection limit as 2 times of the standard deviation of the covariance within the time window from ±230∼250 s during daytime (09:00-18:00) and find detection limits of 0.15 ppt m s −1 for F PNs and 0.19 ppt m s −1 for F APNs .Using this approach we find the random uncertainties in fluxes (Table 2) to be < 16 % for F PNs < 7 % for F APNs and < 19 % for F XPNs .Note that the larger fractional uncertainty in F PNs from the smaller measured flux, as the absolute variability in the covariance at ±230-250 s is smaller for the F P N measurement (Fig. 3).PNs (red crossed lines) and APNs (blue circled lines) corrected for lag time observed by TD-LIF and TD-CIMS respectively, averaged over daytime (09:00-18:00) throughout the whole campaign.

Results
As has been described previously, PN and APN abundances at the BEARPEX site reflect advection of precursors from the city of Sacramento (∼5 h upwind) and from the nearer-field upwind oak belt (∼3 h upwind) (Dillon et al., 2002;Farmer et al., 2006;Day et al., 2009;LaFranchi et al., 2009;Wolfe et al., 2009).The concentrations of all PN species increased from noon to 18:00 (PST) and high concentrations of PNs and APNs persisted past midnight (Fig. 4).Both instruments observed similar diurnal patterns at all heights Figure 5 shows the diurnal pattern of XPNs the difference between PNs and APNs (XPNs ≡ PNs -APNs) at the different heights.At 18 m, XPNs increase beginning at 18hrs to a concentration of approximately 100 ppt and are indistinguishable from zero between 9 and 18 h.In contrast, near the forest floor XPNs approach 100 ppt during the afternoon (15:00-18:00) when the above canopy values are at a minimum.In the middle of the canopy there is an apparent transition region.While it is theoretically possible that the PNs signal could be due to N 2 O 5 , we expect the contribution of N 2 O 5 to the XPN is negligible since the lifetime of NO 3 at this site characterized to reaction with BVOC is too short to sustain concentrations of N 2 O 5 that are more than a fraction of a ppt.The diurnal pattern in eddy covariance PN and APN fluxes (Fig. 6) are similar in shape and direction with peak downward fluxes around 15 h.This general pattern is similar to the APN flux observations reported by Wolfe et al. (2009) using measurements 2 yr earlier at this same study site.Both the APN and PN measurements give fluxes that are opposite in sign to the observations of PN fluxes also made at this site during summer 2004 (Farmer et al., 2006).While the sign of the two flux measurements is the same, the magnitude of PN and APN fluxes are significantly different.We observe a 30-60 % slower downward flux of PNs than of APNs with a maximum difference during the daytime (09:00-18:00).This is surprising given the concentrations at 18 m, the height where we observed fluxes, are identical to within 2-5 %.
Two pieces of information provide hints about the source of the flux difference.For convenience we define c to be the mixing ratio difference of species c from 18 m to height i, as: Positive values of c indicate enhancements in species c at height i relative to 18 m while negative values of c indicate a depletion of c at height i relative to 18 m.Figure 7 shows the vertical gradients ( c) of PNs and XPNs at five different times of day.During the daytime, PNs and XPNs are larger within the canopy than above.At night, values of both observables are much lower near the forest floor than above the canopy indicating loss rates of PNs or XPN are faster than the production rates.These profiles are suggestive represent the time of day black (21:00-3:00), blue (3:00-6:00), green (6:00-12:00), red (12:00-7 18:00) and magenta (18:00-21:00).8 Fig. 7. Diurnal pattern of vertical mixing ratio differences of PNs (left) and XPNs (right) with respect to the 18 m mixing ratios.The horizontal bars are the 1σ variance divided by square root of number of data points.Positive values of PN and XPN indicate enhanced concentrations and negative values represent depleted concentration compared to 18 m.Colors represent the time of day black (21:00-03:00), blue (03:00-06:00), green (06:00-12:00), red (12:00-18:00) and magenta (18:00-21:00). of a daytime source of XPNs within the canopy.This source would lead to a net upward flux of XPNs, consistent with the observed difference between APN and PNs fluxes.
A second piece of evidence indicating a within canopy source of XPNs is the temperature dependence of the XPN concentration at different heights.Figure 8 shows observations of the temperature dependence in the morning period (06:00-12:00), when large flux differences were observed alongside relatively shallow gradients.During this time period XPNs at 18 m decrease with increasing temperature while at the forest floor (0.5 m) XPNs increase with temperature.XPNs at 5 m and 9 m (not shown here) are intermediate between these two features.The temperature dependence of XPN at each height indicates that the production process is stronger within the canopy -due to elevated concentrations of biogenic hydrocarbons and/or soil-emitted NO x -and that XPN loss processes dominate above the canopy.Although we chose this time interval to map directly to the time window shown in Fig. 7, we arrive at the same conclusions if we use a narrower time window such as 06:00-09:00 or 09:00-12:00.

Discussion
We have also investigated a variety of possible issues with our instrumentation, finding no source of error able to explain both the flux and gradient differences.The largest of these potential sources was pointed out in a recent study by Phillips and co-workers (2012) who found that peroxyacetic acid (PAA: CH 3 C(O)OOH) is a potential interference to TD-CIMS measurements of PAN with I − detection.The suggestion they made was that a large background in the TD-CIMS instrument could contribute to the observed flux without contributing to an observed concentration difference.During BEARPEX 2009, we do observe a diurnally varying background in the TD-CIMS measurements of PAN (Fig. 9).If we assume that this signal of about 25 ppt PANequivalent at noon is entirely due to PAA and that PAA deposits at the turbulence limit of 2-4 cm s −1 , then we calculate a PAA flux of 0.25-1 ppt m s −1 .This value is 10-40 % of the flux we have attributed to an unidentified PAN-like molecule.Although a flux of this background, whatever the molecule, might be important, we emphasize that the gradient of the background and of the concentration of XPNs is the reverse of what would be expected for a species with a downward flux.Both are higher within the canopy than above.If the background has a normal flux-gradient relationship, then its flux will be upward and would have the effect of reducing the downward flux of PAN.If this is the case, then the XPN flux would be even higher than derived here.
This leads us to believe that there is an upward flux of one or more peroxynitrates that are not measured by the TD-CIMS.
Other evidence supporting this suggestion is indicated by the temperature and PAR dependence of the XPN fluxes and exchange velocities, Vex, (Fig. 10).We define the flux and Vex of XPNs as in Eq. (3): F XPNs and Vex XPNs are independent of canopy temperature below 20 • C and then increase steeply with increasing temperature.F XP N s and Vex XPNs increase with PAR to 1200 µmol/m 2 /s then decrease.These features are similar to previously reported behavior for BVOCs at this site (Lee et al., 2005;Holzinger et al., 2005;Schade and Goldstein, 2001).We can estimate source strength needed to produce the observed XPN flux by assuming steady-state XPN chemical production and XPN flux out of the canopy: Here, F P−XPNs is the net chemical production of XPNs.
Other terms in the flux budget including thermochemical loss, plant uptake, and canopy storage are calculated to be more than a factor of 10 smaller and can be neglected.The flux of XPNs, during daytime (2.3 ± 0.4 ppt m s −1 : mean ± 1σ divided by square root of number of data points) is then approximately the integral of chemical production within the canopy.To calculate the chemical production rate, we assume the canopy integrated chemical production of 2.3 ppt m s −1 occurs uniformly from the forest floor to the 18 m measurement height (Eq.5).The PA radical concentration needed to support the observed flux of XPNs is calculated by solving Eqs. ( 5) and ( 6) for XO 2 using an average daytime concentration of NO 2 of 250 ppt and a rate constant for k XO 2 +NO 2 of 1.1 × 10 −11 cm 3 molec −1 s −1 , chosen in analogy to the rate for PAN formation at 298 K and 868 mbar (Tyndall et al., 2001): Solving for XO 2 , we find a concentration of 2.3 ± 0.7 ppt.This is about 10 % of the average daytime HO 2 observed at the site (22 ± 11 ppt) (Mao et al., 2012).One plausible candidate for BVOC inducing XPN formation is sesquiterpene (SQT) oxidation.The observed sesquiterpene concentration within canopy at this site is approximately 85ppt during daytime.Using β-caryophyllene as an example, the OH and O 3 reaction rate constants are 1.97 × 10 −11 cm 3 molec −1 s −1 and 1.16 × 10 −14 cm 3 molec −1 s −1 , respectively (Shu and Atkinson, 1994;Shu and Atkinson, 1995).The typical concentrations of OH and O 3 within the canopy during BEARPEX 2009 were 0.16ppt and 54 ppb.For a 100-s canopy mixing timescale (Wolfe et al., 2009), we calculate 198 ppt of oxygenated BVOC (oBVOC) produced during SQT oxidation within the canopy.The source of XO 2 can be estimated considering the reaction of this oBVOC with OH radical, Eq. ( 7).
Solving for the rate constant to form an XO 2 radical using this estimate of oBVOC and the observed OH we find a rate constant of 2.3 × 10 −11 cm 3 molec −1 s −1 which is in the range of typical rate constants for the reaction of OH with aldehydes, confirming the plausibility of a sesquiterpene source of XPNs.
In addition to the reasonable value for the inferred rate constant there is laboratory and computational evidence supporting the idea that BVOC emitted by forest canopies can result in upward fluxes of PNs.For example, in a recent chamber study PNs built on a more complex organic framework than the three APNs observed in our study were observed during β-caryophyllene ozonolysis (Jenkin et al., 2012).Several other studies have discussed production of PNs built on a larger carbon backbone, including α-pinene (Noziere andBarnes, 1998), d-limonene (Leungsakul et al., 2005) and nonanal (Bowman et al., 2003) derived PNs.Using a canopy box model, Wolfe et al (2011) found upward fluxes of a PN derived from 2-methyl-3-buten-2-ol oxidation.This molecule likely cannot be the entire source of our upward flux as that particular product is not efficiently formed (Carrasco et al., 2006), however the calculations show that several such PN products form and are transported out of the canopy on the relevant time scales.

Implications of XPNs formation
The formation of XPNs within the canopy suggests a complex array of effects on PN fluxes and, more broadly, reactive nitrogen forest-atmosphere exchange.The first consequence is that the sign of the total PN flux depends on the relative magnitude of the upward flux of XPN and downward fluxes of PAN, PPN and MPAN.This offers one possible explanation for the upward PN flux previously observed at this site (Farmer et al., 2006).If this is the case, in the Farmer et al. measurements, the upward flux of XPNs was a consequence of vegetation acting as a source of PN precursors rather than, as is often assumed, a sink through stomatal uptake.Explaining both data sets requires the biogenic source of PN precursors at the BEARPEX site to have decreased over the last decade.The large decrease in NO x (∼67 % from 2001-2009) and NOy (∼30 %) concentration over this time interval, imply a corresponding decrease in OH (e.g.Murphy et al., 2006;La Franchi et al., 2011).Since PAN precursors are largely secondary products of OH reactions with BVOC, it is reasonable to assume they have decreased over time.In addition, high O 3 levels have also decreased, a sign of slower oxidative chemistry and reducing the role of ozone as an oxidant of alkenes (LaFranchi et al., 2011).
A second implication is the possible role of XPNs in explaining poorly understood canopy reduction factors for soil NO x emissions.XPN formation affects the total N flux, as upward fluxes of XPNs function to transfer within canopy NO 2 to the PN pool.XPN formation is a plausible within canopy NO x loss process and is consistent with the long standing discussion about missing NO x sinks within canopies, which are parameterized as a canopy reduction factors and widely invoked in global models.These models otherwise find the strength of soil NO x emissions lead to dramatic overestimates of O 3 (e.g.Jacob and Wofsy, 1990).Deposition to vegetation is the most studied NO x sink process.However, the role of vegetation with respect to the NO x flux is still controversial (Lerdau et al., 2000).Here, we suggest that XPN formation is one possible mechanism of NO x loss within the canopy.

Conclusions
We observed a significant (30-60 %) difference between the deposition rate of PNs, and that of the sum of three speciated PNs.The difference in these fluxes suggests formation within the canopy of PNs other than PAN, PPN and MPAN.The time of day and temperature dependence of the vertical gradient of XPNs also supports this idea.We show the flux of XPN must be upward, indicating the possibility that local vegetation acts indirectly as a source of PN rather than as a passive PN sink, a mechanism that can explain the upward PN flux reported by Farmer et al. (2006).In this mechanism, oxidation of biogenic VOC may affect not only the magnitude of PN fluxes, but also NO x fluxes, supporting the ideas of a strong connection between turbulent transport and chemical processes in biosphere-atmosphere exchange, as proposed by Wolfe et al. (2009and 2011) and Farmer et al. (2006), and offering at least a partial explanation for canopy reduction factors.

Fig. 1 . 8 Fig. 1 .
Fig. 1.Colors represent difference measurement heights of 18m (black), 14m (light blue), 9m 3 (blue), 5m (green), 2m (magenta) and 0.5m (red).Yellow and shaded periods are calibration 4 and zero cycles and white periods represent no data collected at that time.(a) Flux 5 measurement of ΣPNs, (b) vertical gradient measurements from ΣPNs, (c) APNs 6 measurement over the flux-divergence mode, (d) same as (c) but over the normal mode.7 8 Fig. 1.Colors represent difference measurement heights of 18 m (black), 14 m (light blue), 9 m (blue), 5 m (green), 2 m (magenta) and 0.5 m (red).Yellow and shaded periods are calibration and zero cycles and white periods represent no data collected at that time.(a) Flux measurement of PNs, (b) vertical gradient measurements from PNs, (c) APNs measurement over the flux-divergence mode, (d) same as (c) but over the normal mode.

Fig. 4 . 7 Fig. 4 .Fig. 5 . 7 Fig. 5 .
Fig. 4. Diurnal patterns of ΣPNs (solid lines) and APNs (dashed lines) at each height.The 3 vertical bars are the 1σ variance divided by square root of number of data points.The left box 4 indicates the measurement height, with green at or below the 9 m average canopy height and 5 blue above that point.Dashed lines are the specific measurement heights.6 7

Fig. 6 .Fig. 6 .
Fig. 6.Diurnal pattern of FΣPNs and FAPNs.The vertical bars are the 1σ variance divided by 3 square root of number of data points.Both species exhibit a downward flux, indicating the 4 dominance of in-canopy sinks, though PNs 30-60% less negative than FAPNs during daytime.5 Fig. 6.Diurnal pattern of F PNs and F APNs .The vertical bars are the 1σ variance divided by square root of number of data points.Both species exhibit a downward flux, indicating the dominance of in-canopy sinks, though PNs 30-60 % less negative than F APNs during daytime.

Fig. 7 .
Fig. 7. Diurnal pattern of vertical mixing ratio differences of ΣPNs (left) and XPNs (right) 3 with respect to the 18m mixing ratios.The horizontal bars are the 1σ variance divided by 4 square root of number of data points.Positive values of ∆ΣPN and ∆XPN indicate enhanced 5 concentrations and negative values represent depleted concentration compared to 18m.Colors 6

Fig. 9 .Fig. 10 .Fig. 10 .
Fig. 9.Diurnal patterns of TD-CIMS background converted to PAN-equivalent concentration.The vertical bars represent the 1σ variance divided by square root of number of data points in the observations.

Table 1 .
Vertical placement of measurements used in this study on north tower.

Table 2 .
Estimated flux analysis errors.