Photochemistry in the arctic free troposphere: NOx budget and the role of odd nitrogen reservoir recycling

,


Introduction
Nitrogen oxides (NO x =NO+NO 2 +NO 3 ) play a critical role in the oxidizing capacity of the troposphere through their impact on the production, loss and partitioning of radical species (OH, HO 2 and RO 2 ) (Levy II, 1972;Crutzen, 1974Crutzen, , 1979)).Furthermore, NO x not only affects the concentration and partitioning of radical species, but also directly participates in the reactions leading to the production and distribution of ozone (O 3 ), i.e. reactions of peroxy radicals with nitric oxide (Chameides and Walker, 1973).Sources of NO x in the free troposphere include convective transport from surface sources, lightning, stratospheric injection, aircraft emissions and recycling reactions from reactive odd nitrogen reservoir species such as nitric acid (HNO 3 ) and peroxyacetic nitric anhydride (PAN) (Roberts, 1990;Jacob et al., 1996;Levy II et al., 1999).A 3-D modeling study by Moxim et al. (1996) showed that the magnitude and seasonal cycle of the global tropospheric integral of NO x are barely affected by the inclusion of PAN chemistry; the global emissions of NO x balance formation of HNO 3 .However, as pointed out by Moxim et al. reservoir species such as PAN can regionally provide an efficient mechanism for redistributing the NO x far from source regions, and thus can perturb regional O 3 photochemistry.As an illustration, Moxim et al. estimate that PAN decomposition can increase the monthly mean NO x mixing ratios in the remote lower troposphere over the North Atlantic and North Pacific by a factor of 5.
Direct measurements of total reactive odd nitrogen (NO y ) and its suspected dominant components (NO x , HNO 3 , PAN) show good agreement at most continental sites at ppbv levels, i.e. components comprise greater than 90% of observed NO y (Parrish et al., 1993;Sandholm et al., 1994).At remote locations, NO y measurements and the NO y sum have disagreed by 30-50% (Ridley, 1991;Atlas et al., 1992;Crosley, 1996;Kondo et al., 1997); however the recent SONEX study in the North Atlantic upper troposphere showed that the sum of the speciated NO y could account for greater than 90% of measured NO y (Talbot et al., 1999).Modeled peroxynitric acid (HNO 4 ) accounted for the majority of the remaining NO y speciation.
In this study, we quantify the NO x budget in the arctic free troposphere during the Tropospheric O 3 Production about the Spring Equinox (TOPSE) 2000 campaign using a chemical box model constrained by aircraft observations.The major goal of TOPSE was to study the late winter-to-spring transition in arctic photochemistry in order to assess the importance of the springtime increase in tropospheric photochemical O 3 production with regard to the observed springtime maximum in high-latitude tropospheric O 3 .In this remote environment, it is expected that convection, lightning and aircraft sources of NO x are minimal so that recycling reactions from reservoir species dominate the NO x production and loss.Observations of the partitioning of the NO y species in the arctic free troposphere are sparse.Knowing the distribution of NO y and how recycling reactions redistribute NO y in the arctic free troposphere indicates whether the arctic free troposphere simply transports NO x as PAN or whether the arctic free troposphere is a region of permanent removal of odd nitrogen from the global atmosphere as HNO 3 .If a large fraction of the NO x accumulates as PAN in the arctic free troposphere, then subsequent episodic transport of these air masses to mid-latitudes along subsiding trajectories would provide large sources of NO x which could shift the O 3 budget to net O 3 production in impacted regions (Beine et al., 1997;Hamlin and Honrath, 2002).Here, we also explore the importance of other reactive odd nitrogen species, such as N 2 O 5 and HNO 4 , which rapidly inter-convert with NO x on intermediate time scales (on the order of a day), and thus impact how efficiently NO x is converted to either PAN or HNO 3 in arctic air masses.

Data sorting
The TOPSE campaign was composed of a series of seven round-trip deployments between 4 February 2002 and 23 May 2002 with missions generally sampling the mid-to-high latitude troposphere over North America in the corridor originating in Colorado, traversing over Manitoba and Hudson Bay before ending north of Greenland and returning back to Colorado.A C-130 aircraft was used to probe the composition of the troposphere between 0 and 8 km.We based our analysis on a merged time series of the aircraft observations where the measurements were averaged over a common time interval of 1 min.We sorted the observations based on latitude (40-50 N, 50-58 N, 58-85 N), altitude (0-1, 1-3, 3-6, 6-8 km) and month (2 February deployments, 2 March deployments, 2 April deployments, 1 May deployment) and derived statistics (mean, 1s standard deviation about the mean, median) for sorted cases.Tables 1 and 2 list statistics for many of the critical chemical species over the springtime transition in the arctic (58-85 N), separated for the 1-3 and 3-6 km altitude layers.Descriptions of the analytical measurement techniques and references can be found in a TOPSE overview paper (Atlas et al., 2003).Canister-GC/MS measurements of CH 3 CHO were performed, but mixing ratios are considered preliminary (D.Blake, personal communication) because canister sampling for CH 3 CHO has not been validated (factor of two uncertainty).Interestingly, prior informal comparisons between canister and airborne in situ CH 3 CHO measurements during PEM-Tropics-B (Singh et al., 2001) and Trace-P showed significant point-by-point variability, but overall reasonable agreement for median values.

Model approach
Our general approach in quantifying the NO x budget was to calculate 24-h average NO x production and loss rates for odd nitrogen recycling reactions for different months and altitudes by using a photochemical box model constrained by the measured medians listed in Tables 1 and 2. The photochemical box model was run for repetitive diurnal solar cycles to diel steady state with photolysis frequencies (j values) calculated using radiative transfer code (TUV Version 4; (Madronich and Flocke, 1998)).The TUV module was initialized for the various cases with measured median values for latitude, Julian day, altitude, O 3 column and albedo.Cloud correction factors were calculated by comparing the TUV clear-sky results run on individual aircraft points and individual j value measurements.Median values of the sorted cloud correction factors were then applied to all the j values calculated by TUV within the photochemical box model.The median j(NO 2 ) and jðO 1 DÞ cloud correction factor ranged between 0.82-1.06and 0.92-1.00,respectively, for the arctic free troposphere sampled during TOPSE.The TUV module was updated with cross section and quantum yields from recent evaluations for inorganic species (DeMore, 1997(DeMore, , 2000) ) and organic species (Atkinson, 1997b;Atkinson et al., 2000).The photodissociation of HNO 4 in the near-IR was incorporated into TUV based on the recent measurements of Roehl et al. (2002).The clear sky photodissociation rate for HNO 4 is on the order of 10 À5 s À1 at 240 K, largely invariant with changes in solar zenith angle.
NCAR's master mechanism was used as the chemical scheme within the model (Madronich and Calvert, 1990).The inorganic chemistry has undergone updates since then with the most recent recommendations taken from (DeMore, 1997(DeMore, , 2000)).The hydrocarbon chemistry in the master mechanism is treated explicitly and includes the photo-oxidation of partly oxygenated organic species.Alkanes up to C 8 , alkenes up to C 3 and aromatics up to C 8 were observed in the arctic free troposphere and were considered as initial hydrocarbon reagents in the gas-phase mechanism.The chemistry of the methyl peroxy radical is treated explicitly; a counter scheme is used for the other organic peroxy radicals (Madronich and Calvert, 1990).The rate coefficients for organic peroxy radical reactions were updated based on recent recommendations (Tyndall et al., 2001;Mereau et al., 2000).Alkoxy radical reactions were also updated based on the latest DeMore et al. recommendation (for CH 3 O) and the specific studies in Atkinson (1997a) and Mereau et al. for the larger alkoxy radicals.Rate coefficients for hydrocarbons reactions with OH were updated based on the latest JPL compilations (DeMore, 1997(DeMore, , 2000) ) and the Atkinson (1994) review.The OH-initiated ethene oxidation mechanism was modified to include multiple branching for the b-hydroxy ethoxy radical reaction with NO (Orlando et al., 1998).OHinitiated rate coefficients for oxygenated hydrocarbons were updated from the Atkinson (1994) compilation.The kinetics of the HO 2 self-reaction were recently measured (Christensen et al., 2002) to be lower than the current recommendation.As discussed by Stroud et al. (2003), the impact of these new data on HO 2 , H 2 O 2 and CH 3 OOH mixing ratios was not significant for TOPSE conditions.

Results and discussion
3.1.Lifetime considerations for odd nitrogen species in the arctic free troposphere In the remote troposphere, NO x is observed to be a small fraction of the total oxidized nitrogen budget due to the rapid processing of NO x to its reservoir species during transport from NO x sources.Fig. 1 illustrates the seasonal changes in lifetimes for odd nitrogen species in the arctic middle troposphere (3-6 km).Actual median latitudes and altitudes used in the TUV calculations are shown in Tables 1 and 2. NO x has a lifetime less than a week in the arctic free troposphere.The increase in photochemical activity between February and May results in the NO x lifetime decreasing from 6 to 2 days.In contrast, PAN and HNO 3 are more stable species with lifetimes on the order of a month (May) to a year (February).PAN photolysis dominates PAN thermal decomposition in the arctic middle troposphere, while OH oxidation and photolysis contribute comparably to  C. Stroud et al. / Atmospheric Environment 37 (2003) 3351-3364 HNO 3 loss.Median temperatures in the arctic midtroposphere showed little change between winter and spring (from 237 to 243 K); as a result, PAN thermal decomposition changes were small.HNO 4 and N 2 O 5 have shorter lifetimes, comparable in magnitude to NO x .HNO 4 thermal decomposition and IR photolysis are the important loss mechanisms for HNO 4 in the arctic free troposphere.Our calculations result in IR photodissociation contributing 20% and 37% to the total HNO 4 loss in February and May, respectively.Hydrolysis is the most important loss process for N 2 O 5 throughout the winter to spring transition.N 2 O 5 photolysis does make a smaller but important contribution in the winter (sensitive to February median latitude).Similarly, N 2 O 5 thermal decomposition makes an important contribution in the spring (sensitive to May median temperature).The steady-state assumption should be a reasonable approximation for HNO 4 and N 2 O 5 considering both have relatively short lifetimes and given the remote location of the observations.However, PAN and HNO 3 mixing ratios are likely impacted by chemistry as well as transport.

Observed seasonal trends in individual NO y species mixing ratios
Fig. 2 presents the observed seasonal dependence to the partitioning of the NO y budget in the lower (1-3 km, panel a) and middle (3-6 km, panel b) free troposphere.Mean mixing ratios with measurement uncertainties are shown for months between February and May with all the observations north of 58 .Table 3 lists how the measurement uncertainties were estimated for each of the odd nitrogen species (columns 1 and 2).The mean NO x mixing ratios show a weak seasonal increase from February to May in both altitude ranges (from 11 to 28 pptv for 1-3 km; from 20 to 30 pptv for 3-6 km).Surface measurements at Zeppelin mountain, Svalbard (78 N, 11 E, 474 m asl) generally ranged between 10 and 40 pptv with no apparent seasonal trend between February and May (Beine et al., 1996).Surface measurements by Beine et al. (1997) at Poker Flat, Alaska showed a weak NO x trend with daytime NO medians increasing from 5 to 15 pptv between March and May.These other surface observations are similar in magnitude and trend to the TOPSE observations.Thus, the TOPSE observations reinforce our understanding of the temporal NO x distribution near the surface while enhancing our understanding of the vertical and spatial NO x distribution.
The mixing ratios of the reservoir species, PAN and HNO 3 , show a strong upward seasonal trend, with PAN showing the largest increase from a mean value of 117 pptv in February to 319 pptv in May for the 3-6 km layer.HNO 3 increased from 39 to 76 pptv over this time period while PPN showed no significant seasonal trend  with mean mixing ratios between 32 and 35 pptv in the 3-6 km layer.The pronounced seasonal increase in PAN and HNO 3 in Fig. 2 is an indication of the springtime growth in photochemical activity.The observed increase in PAN mixing ratios is consistent with the weak seasonal increase in observed CH 3 CHO and NO x mixing ratios, the strong seasonal trend in observed OH and the persistence of cold temperatures (long PAN thermal decomposition lifetime).Changes in transport also play a role as supported by the concurrent increase in NO y .Similar trends were seen for PAN and HNO 3 in the lower free troposphere (panel 2b) with the exception of the decrease in PAN between April and May.This is likely a reflection of the significant warming trend (median T ¼ 2562266 K) between April and May in the 1-3 km layer.

ARTICLE IN PRESS
3.3.In situ NO x cycling in the arctic free troposphere

Introduction
Numerous studies have examined the role of NO y reservoir species in NO x recycling in remote environments (Schultz et al., 2000;Jacob et al., 1996;Kotchenruther et al., 2001;Hamlin and Honrath, 2002).The importance of HNO 3 and PAN as a sink/ source for NO x has been observed to vary with season and location.For example, during SUCCESS in the upper troposphere over the US, HNO 3 was the dominant NO y species and the principal sink for NO x (Jaegle et al., 2000).Similarly, over the North Atlantic during SONEX, HNO 3 composed the largest fraction of NO y (median ratio of 0.35) in the upper troposphere (Talbot et al., 1999).In urban areas, PAN is a net sink for NO x (Roberts et al., 2002); however, in remote areas PAN has been calculated to be a net source for NO x .For example, Schultz et al. (1999) showed that the NO x responsible for O 3 production within the South Pacific below 4 km can largely be explained by the decomposition of PAN transported into the region from biomass burning plumes at higher altitudes.Kotchenruther et al. (2001) showed that PAN decomposition contributed significantly to NO x mixing ratios in the eastern Pacific troposphere off the coast of the US, especially for subsiding air masses originating from Asia.Hamlin and Honrath (2002) also showed that springtime NO x in air masses transported from the arctic to the North Atlantic can largely be attributed to PAN decomposition along the subsiding trajectories.The common feature among these studies is enhanced levels of PAN in the upper troposphere (from biomass burning convection, from Asian outflow convection and arctic wintertime accumulation) followed by decomposition and release of NO x along subsiding trajectories.
From our modeling analysis, we address the in situ partitioning of NO x to its oxidized products in the arctic free troposphere.Tables 4 and 5 present the 24-h average NO x budget in the arctic free troposphere due to localized reservoir recycling in the 3-6 and 1-3 km layers.Net NO x fluxes were calculated from constrained box model simulations run to steady state where NO x ,

ARTICLE IN PRESS
A negative value indicates net NO x loss (pptv/day).Species constrained at measured medians.Calculated from the 5th day of a constrained box model simulation (fixed NO x , HNO 3 and PAN at measured medians).Model run at observed monthly median latitudes of 65 N, 74 N, 68 N, and 72 N.
a Model was constrained by estimated median CH 3 CHO measurement.
b Model calculated CH 3 CHO to steady state.
PAN and HNO 3 were fixed at measured medians.In these simulations, hydrocarbons, CH 4 , CO, O 3 , H 2 O, CH 2 O, H 2 O 2 and CH 3 OOH were also constrained at their measured medians.Two simulations were performed for Tables 4 and 5, with CH 3 CHO either fixed at its measured median or calculated.Two simulations were necessary due to significant model-measurement differences for CH 3 CHO.The following section summarizes the model results with CH 3 CHO constrained, while the remainder of the paper relies on CH 3 CHO calculated from the model photochemistry.

In situ NO x cycling with acetaldehyde constrained
Modeled steady-state CH 3 CHO in the arctic free troposphere underestimated observations by an order of magnitude for all months during the campaign.For example, in May, modeled steady-state CH 3 CHO underestimated observations by a factor of 20 (measured 160 pptv vs. model 8 pptv).Tables 4 and 5 illustrate that constraining CH 3 CHO in the model results in a net NO x loss to PAN dominating other NO x chemical losses for all months and altitude ranges.However, a closer inspection of the net NO x loss to PAN suggests the magnitudes unrealistically large.For example, in May, net NO x loss rates are calculated as large as À31 and À62 pptv/day for 3-6 and 1-3 km, respectively.Given that NO x median mixing ratios in May are 27 and 30 pptv for the 3-6 and 1-3 km, the NO x loss rates would imply significant regional sources of NO x in the arctic free troposphere.While some localized NO x sources are possible, at times, due to lightning or aircraft emissions, it does not seem feasible that for median conditions, over wide regions of the arctic free troposphere, there are NO x sources of this magnitude.Similarly, these net PAN production rates imply that in a relatively short time period air masses in the arctic may have significantly more PAN than observed.For example, a 5-day simulation using constrained CH 3 CHO resulted in an enhancement of modeled PAN by 67% from its initial median measured mixing ratio (323-539 pptv).
Several arguments may be postulated to explain these model-measurement differences for CH 3 CHO: (1) widespread surface emissions (Domine and Shepson, 2002) and efficient transport into the arctic free troposphere, (2) incomplete distribution of measured hydrocarbons and an initialization of reactive hydrocarbons at observed values instead of some representative concentration along the arrival trajectory, (3) incomplete understanding of model organic oxidation pathways, (4) CH 3 CHO artifact formation during sampling procedures or other instrumental errors.However, upon further inspection, several of these postulates should be discarded.In May, the CH 3 CHO lifetime due to photolysis and OH reaction is only 0.9 day.Despite the potential for a surface source of CH 3 CHO, this lifetime is significantly shorter than the timescale for mixing into the arctic free troposphere.Thus, mixing from surface sources should be discarded as an explanation for over an order magnitude model underestimation.Domine and Shepson (2002) also report an [HCHO]/[CH 3 CHO] ratio of 2.5.However, the TOPSE observations show CH 3 CHO mixing ratios consistently being larger than HCHO for all months and altitude ranges considered here.For example, a median [HCHO]/[CH 3 CHO] ratio of 0.34 was observed for May conditions in the 3-6 km range.This behavior of CH 3 CHO mixing ratios greater than HCHO mixing ratios does seem unusual given that typically lower molecular weight species are observed at higher mixing ratios than higher molecular weight analogs (CH 4 >C 2 H 6 , CH 3 OH>C 2 H 5 OH).Given a lifetime for CH 3 CHO of less than a day in May, the steadystate assumption for CH 3 CHO should be a reasonable approximation.A simulation was performed to calculate the equivalent ethane-mixing ratio necessary to sustain CH 3 CHO at 160 pptv for May conditions.An unrealistically large ethane-mixing ratio (40 ppbv) was necessary to photochemically sustain CH 3 CHO at 160 pptv for May conditions.Cumulatively, these arguments shed uncertainty to the measurements of CH 3 CHO during TOPSE.Interestingly, applying the ratio of [HCHO]/ [CH 3 CHO] observed by Domaine and Shepson to the HCHO observations yields CH 3 CHO mixing ratio estimates (22 pptv for May, 3-6 km) closer to the model steady-state CH 3 CHO mixing ratios (8 pptv for May, 3-6 km).Given these arguments, CH 3 CHO was calculated to steady state in the model for the remainder of the NO x budget results reported here.

In situ NO x cycling with acetaldehyde calculated
Tables 4 and 5 presents the net NO x fluxes due to PAN, HNO 3 and N 2 O 5 in situ chemistry with CH 3 CHO integrated to steady state and the odd nitrogen species constrained to their median measured mixing ratios.Since no direct measurement of HNO 4 was made during TOPSE, it is not possible to perform similar constrained calculations for HNO 4 .Later, in Section 3.5, the role played by HNO 4 chemistry is highlighted with a series of time-dependent simulations.
HNO 3 is a net sink for NO x throughout the arctic free troposphere between February and May.In the 3-6 km layer, the net NO x sink due to HNO 3 formation varied from À0.5 to À4.4 pptv/day, increasing in magnitude in spring.Similarly, in the 1-3 km layer, the net NO x sink due to HNO 3 formation varied from À0.1 to À7.0 pptv/ day.Net NO x loss rates to PAN show remarkable similarity to net NO x loss rates to HNO 3 throughout the winter to spring transition in the 3-6 km range (Tables 4  and 5).Between 1 and 3 km, net NO x loss to PAN is significantly larger than the net NO x loss to HNO 3 for February conditions.For March and April, similar net NO x loss rates were calculated for both PAN and HNO 3 .By May conditions, net NO x loss to HNO 3 dominates over loss to PAN for 1-3 km.This difference in seasonal trend between 1-3 and 3-6 km is likely a reflection of the difference in the seasonal temperature dependence for the two altitude ranges.For 1-3 km, significant warming occurred by May conditions (266 K) resulting in a shorter thermal decomposition lifetime for PAN (10 days) which, in part, explains the smaller net NO x loss rates to PAN compared to HNO 3 .Seasonal changes in the [NO 2 ]/[NO] ratio also shorten the effective lifetime of PAN for springtime conditions.
N 2 O 5 is expected to have a lifetime less than 1 day in the arctic free troposphere.At warmer temperatures (>245 K), thermal decomposition is the favored loss mechanism, while at cooler temperatures hydrolysis is expected to dominate.N 2 O 5 is a net sink for NO x in the arctic free troposphere during the campaign; however, it behaves conversely to HNO 3 by decreasing in importance as a NO x sink over the course of the winter-spring transition.N 2 O 5 hydrolysis was included in the chemical mechanism as a heterogeneous reaction.Its rate depends critically on the aerosol surface area concentration and the N 2 O 5 reactive uptake coefficient (g ¼ 0:1).Aerosol surface areas were estimated for this study from climatology generated using MOZART, a 3-D chemical transport model (Tie et al., 2003).These aerosol surface areas were recently evaluated in MOZART (Tie et al., 2001).Pseudo-first-order rate coefficients, k het ; varied between 10 À4 and 10 À5 s À1 , generally increasing with season and decreasing with altitude in the arctic free troposphere.NO x loss rates through N 2 O 5 hydrolysis peaked at À2.9 pptv/day in the 3-6 km layer in February.In the 1-3 km layer, the NO x loss rate peaked at À4.5 pptv/day in March.The increase in N 2 O 5 hydrolysis between February and March for 1-3 km may stem from the change in median latitude between February (65 N) and March (74 N).These results show that N 2 O 5 hydrolysis is the dominant loss process for NO x under wintertime arctic conditions.By May, negligible NO x loss rates due to N 2 O 5 hydrolysis were calculated throughout the free troposphere due to lower modeled NO 3 mixing ratios and, thus, slower N 2 O 5 production rates.These seasonal changes in NO x loss through N 2 O 5 are directly related to the availability of UV radiation to photolyze the intermediate NO 3 radical between February (mostly dark) to May conditions (mostly light).Since our calculation of net NO x loss to N 2 O 5 depends critically on aerosol surface area and these are non-measured parameters, two sensitivity simulations were performed for May conditions (58-85 N, 3-6 km) at five times greater/less than our estimates of aerosol surface area concentration.The results span the range from À3.4 to À0.90 pptv/day for February conditions (3-6 km).The smaller sensitivity with increases in aerosol concentration (À2.7 to À3.4 pptv/day) suggests that for arctic wintertime free troposphere conditions, air masses can saturate with regards to NO x loss and further increases in aerosol surface area.The saturation point occurs when the rate of O 3 reaction with NO 2 becomes rate limiting compared to N 2 O 5 hydrolysis.This effect has been studied in greater detail (Tie et al., 2003) and box model simulations show that for the wintertime arctic free troposphere the saturation point occurs at B10 mm 2 / cm 3 .This is of similar magnitude to the aerosol surface areas considered here, thus NO x loss estimates are sensitive to decreases in aerosol surface areas, however insensitive to further aerosol surface area increases.

NO y deficit and its seasonal dependence
Another intriguing observation from Fig. 2 is a seasonal dependence to the NO y balance (defined here as , which is a small deficit for February and systematically decreases to near balance in May for the 3-6 km layer and a small surplus in May for the 1-3 km layer.Limits for the mean NO y balance in Fig. 2 are measurement uncertainties added in quadrature.Since NO y is composed largely of PAN, then it should be noted that uncertainties in the NO y balance stem largely from the uncertainties in the NO y and PAN measurements.The NO y budget is in near balance for all months with the exception of February (3-6 km) where the NO y balance is significantly larger than zero when considering the combined measurement uncertainty (mean 60742 pptv).This is a positive result suggesting we can account for a large fraction (generally greater than 90%) of the observed NO y .This is similar agreement to the NO y balance observed for the recent SONEX campaign (Talbot et al., 1999).
Table 3 lists the median NO y balance for February conditions.These observations support missing NO y species in the range 20-100 pptv for February conditions (3-6 km).A linear fit of mean NO y deficit vs. mean day of year for the data in Fig. 2, weighted by the inverse of the squared measurement uncertainties, yields slopes of À0.8870.85pptv/day (3-6 km) and À1.0070.73pptv/ day (1-3 km).
Thus, the observations do support a marginal seasonal trend from a small deficit to small surplus.Fig. 3 also illustrates how the NO y deficit varies as a function of temperature, jðO 1 DÞ and sulfate aerosol concentration for all the data in the arctic free troposphere between 3 and 6 km.A weak negative dependence is observed for temperature, jðO 1 DÞ; and sulfate aerosol concentration, although there is large atmospheric variability.One potentially important odd nitrogen species not considered in the NO y deficit calculations above is HNO 4 .In the next section, the seasonal dependence for our modeled steady-state HNO 4 mixing ratio is calculated and compared to the seasonal trend in the observed NO y balance.

Lifetime considerations and steady-state HNO 4 mixing ratios
To calculate a steady-state HNO 4 and determine whether the steady-state assumption is reasonable, we performed four different time-dependent NO x simulations where NO x was initialized so that it decayed to its median measured mixing ratio in exactly 1-4 days (Figs. 4a and b,(3)(4)(5)(6).In these simulations, HNO 4 mixing ratios were initialized at zero and hydrocarbons, CH 4 , CO, H 2 O, and O 3 were constrained at their median measured mixing ratios.CH 3 CHO was initialized at its predetermined steady-state value and calculated during the simulation.In Fig. 4, the time axis is plotted such that model time=0 corresponds to the point where model NO x decayed to its median measured value.By comparing the HNO 4 at t ¼ 0; we can estimate how long an air mass has to remain under these environmental conditions (i.e.temperature) for HNO 4 to reach steady state.For February conditions, the time for HNO 4 to reach steady-state is 3 days with a corresponding (HNO 4 ) ss =40 pptv (panel a and b).This relatively short induction time suggests that away from immediate NO x sources (e.g.aircraft exhaust, fresh convection) HNO 4 will be in steady state with NO x .The calculated steadystate methyl peroxynitrate (CH 3 O 2 NO 2 ) mixing ratio was 5 pptv, suggesting the organic homologues play a minor role in the NO y budget compared to HNO 4 .Other modeled NO y species (HONO, N 2 O 5 and NO 3 ) were also at negligible concentrations compared to NO y ; although it should be noted that the odd nitrogen flux through these species can be significant under cold, low light conditions.For the 1-3 km layer in the arctic, steady-state model HNO 4 was o5 pptv throughout the campaign.
An interesting paradox occurs when we compare the predicted steady-state HNO 4 mixing ratios with the median NO y balance for May.For the median NO x mixing ratio of 30 pptv (May, 3-6 km, 58-85 N) and T ¼ 243 K, the model calculates a steady-state HNO 4 mixing ratio of 60 pptv.The steady-state assumption is surely valid for May conditions, as the induction time to reach steady state is less than a day.However, in comparing to the observed NO y remainder in May (3-6 km), there is no room in the budget for a 60 pptv missing species.Furthermore, the increase in HNO 4 from 40 to 60 pptv is opposite to the seasonal trend in the NO y balance.This suggests that there may be other unknown loss processes for HNO 4 in the spring.Recent laboratory studies show that HNO 4 can initiate the oxidation of S(IV) in aqueous particles (Amels et al., 1996;Warneck, 1999).HNO 4 is also known to react with HNO 2 (Logager and Sehested, 1993) and halide ions (Regimbal and Mozurkewich, 1997) in solution.Murphy et al. (2002) discuss the temperature dependence of their indirect measurement of [HNO 4 ]+[RO 2- NO 2 ] during TOPSE.In their analysis, [HNO 4 ]+ [RO 2 NO 2 ] is inferred as the difference between the measured sum of the peroxynitrates and the measured PANs, where they measured the sum of the peroxynitrates using a thermal dissociation laser-induced fluorescence technique.An even larger temperature dependence was observed for their inferred [HNO 4 ]+ [RO 2 NO 2 ] mixing ratios with estimates of B100 pptv at To240 K.These results clearly support the need for direct measurements of HNO 4 of sufficient precision and accuracy along with heterogeneous chemical modeling studies to shed further light on this apparent paradox in the NO y balance.

HNO 4 and its impact on the NO x lifetime
Since our modeling calculations and the measurements of Murphy suggest that HNO 4 can be a significant fraction of NO y in winter and there is some marginal evidence for an NO y deficit in winter, it is useful to study the role played by HNO 4 in the NO x budget in the arctic wintertime.Fig. 4c compares NO x decays for two simulations with and without HNO 4 chemistry.The solid curve is identical to the solid blue curve shown in Fig. 4a for the model simulation with a 3-day induction time and HNO 4 chemistry included.The dashed line in Fig. 4c is the corresponding NO x decay for a simulation with identical initial NO x , however no HNO 4 chemistry.The importance of HNO 4 to modeled NO x mixing ratios depends on the age of the NO x in the air mass.For 'fresh' air masses, NO x decays more quickly with the HNO 4 chemistry included due to the initial formation of HNO 4 .However, for more 'aged' air masses, the HNO 4 reservoir acts as a buffer replenishing NO x .The crossover point where the addition of HNO 4 chemistry enhances NO x mixing ratios over the case with no HNO 4 chemistry occurs between 5 and 6 days and is largely independent of initial NO x mixing ratios.It is very likely that typical air masses sampled in the arctic free troposphere have NO x ages longer than 5-6 days and thus the presence of the HNO 4 reservoir can act as a source for NO x in the arctic.Here we considered a constant temperature history for the modeled air mass at the observed median temperature (February, 3-6 km).Given that thermal decomposition is the dominant loss process for HNO 4 , the cross-over point will be a strong function of the temperature history of an air mass.
To further quantify the role played by HNO 4 in the NO x budget, we calculated diurnally averaged net NO x loss/production rates from a time-dependent NO x simulation for February in the arctic mid-troposphere (February simulation the same as in Fig. 4; analogous simulation for May also performed).If we look at the diurnally averaged net NO x rates for the first 24 h of these simulations we find significant net NO x loss rates to HNO 4 (À28 and À69 pptv/day for February and May) compared to the other odd nitrogen reservoirs (Table 6).Conversely, if we arbitrarily look at the model simulations for a more 'aged' air mass, after 4 days, we find that HNO 4 is a net source for NO x (+3.4 and +5.4 pptv/day for February and May) largely countering the net NO x losses to N 2 O 5 , PAN and HNO 3 .These results yield a consistent picture with Tables 4 and 5 for N 2 O 5 , PAN and HNO 3 and now show the additional role played by HNO 4 .For 'fresh' air masses, HNO 4 is a net sink for NO x .For more 'aged' air masses, HNO 4 is a net source for NO x , largely countering the NO x losses to N 2 O 5 in the winter and PAN and HNO 3 in the spring.

HNO 4 and its impact on O 3 production
The time evolution of modeled O 3 mixing ratios for the two scenarios discussed in the prior section (with and without HNO 4 chemistry) was also considered.The goal of these simulations was to evaluate the importance of HNO 4 chemistry on the amount of O 3 produced along a trajectory from a source region for NO x .The initial production of HNO 4 near NO x sources clearly slows the rate of O 3 production (after 5 days O 3 difference was 0.7 ppbv for February conditions).For more 'aged' air masses, the buffering impact of HNO 4 begins to narrow the difference between the two O 3 traces, however even after 20 days the modeled O 3 from the simulation with HNO 4 chemistry included still has slightly smaller O 3 mixing ratios (difference of 0.4 ppbv).Qualitatively, similar results were obtained for a simulation in May with maximum differences in O 3 of 2.3 ppbv after 5 days and 0.9 ppbv after 20 days.
Thus, these simulations suggest that HNO 4 chemistry impacts the timing of O 3 production; however, the cumulative impact on O 3 mixing ratios after a 20-day trajectory is minimal.

Conclusions
In this paper, we evaluated the in situ recycling reactions for NO x in the arctic free troposphere using a constrained box model.In the winter, N 2 O 5 hydrolysis is the largest net sink for NO x (À2.9 pptv/day in February); however, it reduces to negligible importance in spring due to the seasonal increase in the rate of NO 3 photolysis.NO x loss through N 2 O 5 is limited by the rate of the NO 2 +O 3 reaction for conditions in the wintertime arctic free troposphere.HNO 3 and PAN are also permanent net sinks for NO x in the arctic free troposphere during TOPSE.Modeled CH 3 CHO mixing ratios are an order of magnitude smaller than observations.Given our current understanding of atmospheric chemistry, constraining the model with observed CH 3 CHO yields unrealistic estimates for net NO x loss rates to PAN.Median observed CH 3 CHO mixing ratios Calculated for 0-1 day and 4-5 day of a box model simulation where NO x decreased to its median measured mixing ratio by the end of the 4th day.Other NO y species were initialized at their measured medians and not constrained.Model CH 3 CHO was initialized at model steady-state mixing ratio (from a prior determined simulation) and calculated.Model run at observed monthly median latitudes of 64 N and 73 N for February and May conditions, respectively.

ARTICLE IN PRESS
C. Stroud et al. / Atmospheric Environment 37 (2003) 3351-3364 were also larger than observed HCHO mixing ratios for arctic free troposphere cases; a rather unusual behavior.Thus, until the CH 3 CHO canister measurements are fully characterized in the laboratory and formally intercompared in the field, the TOPSE CH 3 CHO observations are considered preliminary and suspect.Using simulations where CH 3 CHO reached steady state, the net NO x loss to PAN was calculated to be of comparable magnitude to net NO x loss to HNO 3 , with net PAN formation being larger in the wintertime lower free troposphere and net HNO 3 formation being larger in the springtime lower free troposphere.Overall, NO x loss due to N 2 O 5 hydrolysis dominates in winter, whereas NO x loss due to PAN and HNO 3 formation dominates in the spring.These conclusions derived from an observationally constrained box model with explicit chemistry will be useful in future validation of NO x loss processes in 3D chemical-transport models.
The role of HNO 4 chemistry in impacting arctic free troposphere NO x mixing ratios depends on the age of the air mass.Near NO x sources, HNO 4 is a net sink for NO x ; however, HNO 4 decomposition and IR photolysis can be net sources of NO x for more 'aged' arctic air masses.Model simulations suggest the cross-over point to NO x enhancement due to the buffering impact of HNO 4 occurs after 5-6 days (dependent of temperature history of air mass).Given that the age of NO x in typical arctic free troposphere air samples is likely older than 5-6 days, HNO 4 chemistry tends to enhance remote NO x mixing ratios.These conclusions may change as our understanding of the HNO 4 heterogeneous chemistry improves in the future.While the presence of HNO 4 chemistry may locally enhance O 3 production rates for 'aged' air masses, the impact of HNO 4 over the entire air parcel trajectory is a slight decrease in O 3 mixing ratio.
The NO y budget is in near balance for all months with the exception of February (3-6 km) where a mean deficit of 60742 pptv is observed.Thus, it is possible for the observed NO y deficit to support the modeled HNO 4 mixing ratio of 40 pptv in February (3-6 km).It is intriguing that the marginal seasonal trend in the NO y balance from small deficit to small surplus is not consistent with the seasonal trend in our modeled HNO 4 mixing ratios.The modeled HNO 4 of 60 pptv in May (3-6 km) is only supported if we consider the upper limits for the combined measurement uncertainties, À18775 pptv.Thus, it is speculated that unknown HNO 4 chemistry may be important in the arctic springtime.Future measurements of HNO 4 in combination with modeling studies that incorporate heterogeneous laboratory studies are clearly warranted.
Finally, these observations and modeling results illustrate how NO x can accumulate as the PAN reservoir during the spring months in the northern free troposphere.As a result, air masses leaving the high latitude arctic free troposphere and following anti-cyclonic, descending trajectories will have high rates of PAN thermal decomposition, which can enhance O 3 production rates at impacted mid-latitude surface locations.

Fig. 1 .
Fig. 1.Seasonal trend in lifetimes of reactive odd nitrogen species in the 3-6 km layer.The plot was created from points for February and May conditions and linearly interpolated.Lifetimes calculated with median latitudes for February (64 N, day-night j profiles) and May (73 N, smoother j diurnal profiles).Photolysis rates were diurnally averaged to calculate lifetimes.N 2 O 5 chemistry is strongly dependent on median latitude.

Fig. 2 .
Fig. 2. Monthly evolution of measured mean mixing ratios for reactive odd nitrogen species in the 3-6 km layer (panel a) and 1-3 km layer (panel b) between 58 N and 85 N.Only coincident data was used to generate means.Mean RONO 2 / NO y ratio was applied to NO y observations to expand RONO 2 database.Limits represent measurement uncertainties.

Fig. 3 .
Fig. 3.The dependence of the observed NO y deficit (defined as NO y -PANs-HNO 3 -NO x ) on temperature (panel a), O 3 photolysis frequency, jðO 1 DÞ (panel b), and fine aerosol sulfate (panel c) for the 3-6 km layer and between 58 N and 85 N.

Fig. 4 .
Fig. 4. Results of a time-dependent NO x simulation where model NO x decayed to its measured median mixing ratio in 1, 2, 3 and 4 days (panels a and b).Conditions are representative of February, 3-6 km and 58-85 N. Panel c includes the results for two simulations each initialized with the same NO x mixing ratio, however one simulation run with and one without HNO 4 chemistry.Model run at observed median latitude of 64 N.
Only coincident data considered.Mean RONO 2 /NO y ratio applied to NO y observations to expand RONO 2 database.
b c A highly uncertain value due to lack of data.

Table 3
a Calculated from the median of[NO y ]-[NO x ]-[PANs]-[HNO 3 ]-[RONO 2 ] coincident points.bCalculated from the sum of uncertainties in column added in quadrature.

Table 4 24
-h average NO x budget in the arctic free troposphere due to reservoir cycling (3-6 km) negative value indicates net NO x loss (pptv/day).Species constrained at measured medians.Calculated from the 5th day of a constrained box model simulation (fixed NO x , HNO 3 and PAN at measured medians).Model run at observed monthly median latitudes of 64 N, 71 N, 72 N, and 73 N.
A a Model was constrained by estimated median CH 3 CHO measurement.bModel calculated CH 3 CHO to steady state.

Table 6 24
-h average NO x budget in the arctic free troposphere due to reservoir cycling (3À6 km) derived from a time-dependent simulation with CH 3 CHO calculated (in pptv/day)