On the role of monoterpene chemistry in the remote continental boundary layer

The formation of organic nitrates (RONO 2) represents an important NO x (NOx = NO+ NO2) sink in the remote and rural continental atmosphere, thus impacting ozone production and secondary organic aerosol (SOA) formation. In these remote and rural environments, the organic nitrates are primarily derived from biogenic volatile organic compounds (BVOCs) such as isoprene and monoterpenes. Although there are numerous studies investigating the formation of SOA from monoterpenes, there are few studies investigating monoterpene gas-phase chemistry. Using a regional chemical transport model with an extended representation of organic nitrate chemistry, we investigate the processes controlling the production and fate of monoterpene nitrates (MTNs) over the boreal forest of Canada. MTNs account for 5–12 % of total oxidized nitrogen over the boreal forest, and production via NO3 chemistry is more important than production via OH when the NO x mixing ratio is greater than 75 pptv. The regional responses are investigated for two oxidation pathways of MTNs: one that returns NO x to the atmosphere and one that converts MTNs into a nitrate that behaves like HNO3. The likely situation is in between, and these two assumptions bracket the uncertainty about this chemistry. In the case where the MTNs return NO x after oxidation, their formation represents a net chemical NO x loss that exceeds the net loss to peroxy nitrate formation. When oxidation of MTNs produces a molecule that behaves like HNO 3, HNO3 and MTNs are nearly equal chemical sinks for NO x. This uncertainty in the oxidative fate of MTNs results in changes in NOx of 8–14 %, in O3 of up to 3 %, and in OH of 3–6 % between the two model simulations.


Introduction
Measurements of concentrations and fluxes indicate that monoterpenes are emitted into the atmosphere at rates large enough to affect OH concentrations and to contribute a large fraction of the secondary organic aerosol (SOA) mass in such diverse environments as the Amazon (e.g., Karl et al., 2007), the Mediterranean (e.g., Owen et al., 1997;Davison et al., 2009;Seco et al., 2011), pine forests of western North America (e.g., Bouvier-Brown et al., 2009;Kim et al., 2010), the boreal forest (e.g., Spirig et al., 2004;Räisänen et al., 2009;Hakola et al., 2012), and in urban regions of California's Central Valley (Rollins et al., 2012).While there is a growing body of laboratory and field measurements related to monoterpene SOA (e.g., Tunved et al., 2006;Williams et al., 2007;Fry et al., 2009Fry et al., , 2011;;Hallquist et al., 2009;Rollins et al., 2009;Slowik et al., 2010), the gas-phase chemistry of monoterpenes that must precede SOA formation has not been explored in detail with chemical transport models (CTMs).In part this situation is due to the uncertainties in isoprene chemistry that, until recently, were so large that they overwhelmed any uncertainties in modeling monoterpene chemistry.With substantial progress in understanding isoprene oxidation sequences (e.g., Paulot et al., 2009a, b;Peeters and Müller, 2010), it is timely to explore the role of the gas-phase chemistry of monoterpenes in CTMs.
In the remote continental boundary layer, total RONO 2 ( ANs) formation is the primary reaction determining the lifetime of NO x (Browne and Cohen, 2012;Browne et al., 2013).Thus, in regions where monoterpene emissions are important to the total OH and NO 3 reactivity, the formation of monoterpene nitrates (MTNs) can be expected to affect the NO x lifetime, which, in turn, will affect OH concentrations and the ozone production efficiency.In Browne et al. (2013) observations of volatile organic compounds (VOCs), OH, and NO x were used to calculate that the instantaneous production of ANs is more rapid than production of HNO 3 over the Canadian boreal forest and that a significant fraction of the ANs production rate is due to the oxidative chemistry of α-and β-pinene.These are the two most abundant monoterpenes on the global scale, and thus insights from the Canadian forests are likely relevant to other locations.The analysis of observations described in Browne et al. (2013) also indicates that the lifetime of ANs is shorter than that of HNO 3 over the boreal forest, and that this short lifetime is due to a combination of oxidation of ANs resulting in the return of NO x to the atmosphere (NO x recycling), deposition, and the hydrolysis of particulate ANs (forming HNO 3 ).Each of these loss processes has different implications for the NO x budget: the first one recycles NO x , converting a reservoir species back to the active radical pool, while the latter two result in rapid removal of odd nitrogen from the atmosphere.
This paper explores the consequences of MTN chemistry for nitrogen oxides as well as the drivers of atmospheric oxidative chemistry (e.g., OH, O 3 ) using the Weather Research and Forecasting model with Chemistry (WRF-Chem) (Grell et al., 2005), a regional chemical transport model.The model is run using an extended representation of organic nitrate chemistry and the model calculations are compared to observations from the NASA Arctic Research of the Composition of the Troposphere from Aircraft and Satellites (ARC-TAS) campaign (Jacob et al., 2010) over the Canadian boreal forest.We then explore the relative contributions of OH and NO 3 chemistry to the production of MTNs and investigate how the total production impacts NO y partitioning.To understand the range of possible roles for MTNs, we examine their sinks, comparing model predictions for several atmospheric constituents in scenarios where oxidation of MTNs results in the release of NO x to predictions where the -ONO 2 functional group is retained as an organic nitrate.

Domain and emissions
WRF-Chem v3.2.1 (Grell et al., 2005) at 36 km × 36 km resolution over a 5364 km (east-west) by 3384 km (north-south) domain centered over Canada was used for this study.Data from the outer 180 km of the domain are discarded to allow for relaxation of the boundary conditions taken from MOZART-4 model (Emmons et al., 2010) using meteorology from GEOS-5 (available at http://www.acd.ucar.edu/wrf-chem/).Meteorological data are from the North American Regional Reanalysis and the model was run for six days as a spin-up prior to simulating the time period corresponding to the ARCTAS measurements over the Canadian boreal forest from 29 June to 13 July 2008.The goal was not to directly simulate the ARCTAS conditions but rather to investigate the impacts of MTNs on NO x and O 3 in the boreal forest.To that end, fire emissions of NO x and volatile organic compounds were not included in the model.Modeled anthropogenic emissions were from Reanalysis of Tropospheric composition (RETRO) and Emissions Database for Global Atmospheric Research (EDGAR) using the global emissions preprocessor for WRF-Chem (Freitas et al., 2011) and modeled biogenic emissions were from the Model of Emissions of Gases and Aerosols from Nature (MEGAN) (Guenther et al., 2006) available at http://www.acd.ucar.edu/wrf-chem/.In order to achieve approximate agreement with the ARC-TAS observations of NO x and biogenic volatile organic compounds (BVOCs), isoprene emissions were uniformly reduced by 20 % and soil NO x , lumped α-pinene, and lumped limonene emissions were each increased by a factor of two.
Aerosols are treated using the Modal Aerosol Dynamics Model for Europe (MADE) with the Secondary Organic Aerosol Model (SORGAM) (Ackermann et al., 1998;Schell et al., 2001).This mechanism is a "two-product model" approach to aerosols and consequently does not treat the partitioning of individual chemical species such as organic nitrates.Therefore, particle-phase partitioning does not impact the distribution of organic nitrates.
Although more detailed aerosol chemical mechanisms that track the individual species as they transfer between the gas and particle phases do exist (e.g., Derwent et al., 2003;Bonn et al., 2005), our focus here is on investigating the impacts of the gas-phase representation of the chemistry; a full consideration of the aerosol is beyond the scope of this work.

Chemical mechanism
The chemical mechanism is based on the RACM2 mechanism (Goliff et al., 2013), which in its original form includes three organic nitrate species: two isoprene-derived nitrates and one lumped nitrate (ONIT).This lumped nitrate is long-lived (> 100 h lifetime to oxidation by 1 × 10 6 molecules cm −3 of OH at 298 K), assumed to release NO 2 with 100 % efficiency upon oxidation, and has a long lifetime to deposition (Henry's law constant of ∼ 1 M atm −1 ).In contrast, Henry's law coefficients for hydroxy nitrates have been measured to be on the order of 10 3 -10 5 M atm −1 (Shepson et al., 1996;Treves et al., 2000).Monoterpene nitrates are likely to be multifunctional nitrates (containing a nitrate and most likely a hydroxy or peroxide group) and are thus likely to have deposition rates that are much faster than the rate assumed for ONIT.As a 10-carbon molecule, monoterpene nitrates will have a faster oxidation rate and return NO 2 to the atmosphere with less than 100 % efficiency.Furthermore, some monoterpene nitrates will be unsaturated molecules that in addition to having a faster OH oxidation rate will also undergo ozonolysis.
Due to these issues, investigating the low-NO x , high-BVOC chemistry of organic nitrates required an expansion of the organic nitrate chemistry.The changes we make to the RACM2 mechanism include the reclassification of the lumped nitrates into the new species and the addition of oxidation and photolysis reactions for the new organic nitrate species.The RACM2 mechanism includes two lumped species to represent low-reactivity (lumped into the model species α-pinene) and high-reactivity (lumped as limonene) monoterpenes.Most of the monoterpene oxidation products (e.g., pinonaldehyde, organic nitrates) are lumped into generic products (e.g., aldehydes and ketones, ONIT).
During the day, monoterpene nitrates are formed from the minor channel of the reaction of the monoterpene peroxy radical with NO: where MTO 2 is the monoterpene-derived peroxy radical formed from the reaction of monoterpenes with OH, MTO is the monoterpene-derived alkoxy radical, and α is the yield of MTNs.At night, the reaction of NO 3 with monoterpenes results in the formation of a nitrooxy peroxy radical (MTNO 2 ): This radical reacts with NO, NO 3 , HO 2 , or RO 2 with the following possible channels: where β represents the yield of MTNs from this nocturnal chemistry.The exact reactions for both monoterpene species used in RACM2 and the identity of the products can be found in the supplemental material and are discussed briefly below.
For daytime chemistry, RACM2 has an 18 % yield of organic nitrate from the reaction of the lumped α-pinene peroxy radical with NO and a 0 % yield for the reaction of the lumped limonene peroxy radical with NO.In our update, the yield from limonene was increased to 22 % as reported in Leungsakul et al. (2005) and as is assumed in the Master Chemical Mechanism v3.2 (MCM v3.2) available at http://mcm.leeds.ac.uk/MCM (Jenkin et al., 1997;Saunders et al., 2003); the yields of the other monoterpene oxidation products are reduced to account for the increase in organic nitrate yield.The exact yield of monoterpene nitrates carries some uncertainty since yields have only been directly measured for a few specific monoterpenes and these measurements are subject to biases due to loss of the nitrates to aerosol, chamber walls, and/or hydrolysis.The exact yield may be lower or higher than used here; however, any reasonable changes will only impact the exact numerical results and not the overall conclusion that MTNs are important in the NO x and NO y budgets.
RACM2 assumes that the oxidation of monoterpenes by NO 3 results in the formation of two nitrooxy peroxy radicals: one that primarily recycles NO x , and one that primarily forms a nitrate.The split between these two radicals that is in RACM2 is used here, but they are renamed such that their products form MTNs (70 % saturated and 30 % unsaturated).The split of peroxy radicals for the lumped limonene is based on Spittler et al. (2006), who report an organic nitrate yield of 67 %.Fry et al. (2011) recently reported a smaller organic nitrate yield (30 %) from the reaction of NO 3 with limonene.The reason for the discrepancies between the two studies is unclear.
Two monoterpene-derived nitrates are introduced, one unsaturated (UTONIT) and one saturated (TONIT).The lumped limonene peroxy radicals react with NO to form unsaturated nitrates, while lumped α-pinene peroxy radicals react with NO to yield both saturated and unsaturated nitrates with yields of 12 and 6 %, respectively.We note that this branching is not constrained by laboratory measurements; however changes to this branching should only be important if the NO x recycling from monoterpene nitrate oxidation is high.Oxidation rates of the MTN with OH are calculated using weighted averages of the MCM v3.2 rates of α-pinene and β-pinene (assuming a 50-50 mixture and ignoring the OH abstraction channels) for TONIT and from limonene for UTONIT.The ozonolysis rate of UTONIT is estimated using the EPA Estimation Program Interface (EPI) Suite v4.1 (available at http://www.epa.gov/opptintr/exposure/pubs/episuite.htm) (US EPA, 2011).In what we refer to as our BASE case, oxidation of both TONIT and UTONIT generates a second-generation monoterpene nitrate which is oxidized, photolyzed, and deposited identically to HNO 3 .
Assessing the impact of monoterpene chemistry on the NO x budget also requires updating the isoprene chemistry scheme to be consistent with recent laboratory and theoretical studies.Our isoprene oxidation scheme follows that of Paulot et al. (2009a, b).The formation of hydro-peroxy aldehydes (HPALD) from the isomerization reaction of isoprene  peroxy radicals is included using the rate measured by Crounse et al. (2011).The fate of HPALD follows the assumptions of Peeters and Müller (2010).The isomerization of the non-acyl methacrolein peroxy radical is included using the rate and products from Crounse et al. (2012).Peroxy radical cross reactions were determined following the protocol of Madronich and Calvert (1990) updated in some cases to Tyndall et al. (2001) and using updated reaction rates from IUPAC (Atkinson et al., 2006).Reactions of peroxy radicals with HO 2 were determined using the product yields from Hasson et al. (2004).Full details regarding the isoprene chemistry can be found in the Supplement.
Recently, Xie et al. (2013) implemented a revised isoprene chemistry scheme in SAPRC-07 based on assumptions similar to ours.Xie et al. (2013) show that this scheme, as implemented in the US EPA CMAQ (Community Multiscale Air Quality) model, results in significant improvement in the representation of chemistry over eastern US forests.They include a comparison of their results to the ANs measured from the aircraft (Perring et al., 2009) and find good agreement.
Although of minor importance in the remote boreal forest, the treatment of the organic nitrates that are primarily of anthropogenic origin also required extension.The extended organic nitrate chemistry includes four nitrates primarily of anthropogenic origin (aromatic-derived, multifunctional unsaturated, multifunctional saturated, and monofunctional saturated nitrates) that were partitioned from the lumped precursors following the procedure of Middleton et al. (1990) and using branching ratios from Arey et al. (2001) and MCM v3.2.The yields of the nitrates are generally unchanged from RACM2, with the exception of the yield from the lumped species representing short-lived alkanes, alcohols, esters, and alkynes.This yield, originally derived using the branching ratio parameterization from Carter and Atkinson (1989), was reduced (and the non-nitrate products correspondingly in-  creased) to be consistent with the more recent parameterization from Arey et al. (2001).
The elements of our extended mechanism, including oxidation, photolysis, and dry deposition rates and reactions, are listed in the Supplement.

Description of ARCTAS observations
The model is compared to ARCTAS data from the 1 min merge version 13, which is publically available at http:// www-air.larc.nasa.gov/missions/arctas/dataaccess.htm.The data used include flight numbers 17-20 and 23-24.A full summary of the flights and the measurement suite can be found in Jacob et al. (2010).A map of the flight tracks, colored by observed ANs concentration, is shown in Fig. 1.Comparisons of the vertical profiles from the surface to 600 mbar (∼ 4.2 km pressure altitude) between the BASE model and the ARCTAS measurements for key NO y and HO x species are shown in Fig. 2. In making these comparisons, only ARCTAS data over land and removed from biomass burning influences (as defined from inspection of CO, HCN, and CH 3 CN concentrations) are used.
Briefly, the measurements used to compare to the model output are from the NCAR chemiluminescence instrument for NO and O 3 (Weinheimer et al., 1994), the UC Berkeley thermal dissociation laser-induced fluorescence (TD-LIF) instrument for NO 2 , PNs, and ANs (Thornton et al., 2000;Day et al., 2002;Wooldridge et al., 2010), the Pennsylvania State University LIF for OH and HO 2 (Faloona et al., 2004), and the California Institute of Technology chemical ionization mass spectrometer for HNO 3 (Crounse et al., 2006).Biogenic species include α-and β-pinene from the UC Irvine whole-air samples (Simpson et al., 2011) and isoprene, methyl vinyl ketone (MVK), and methacrolein (MACR) from the trace organic gas analyzer (Apel et al., 2003).
The ARCTAS measurements are averaged horizontally to the WRF-Chem resolution, use the closest WRF-Chem output time (output is hourly), and sample the WRF-Chem model along the flight track.However, during one flight leg (flight #20 at 21:51 to 22:28 UTC) the aircraft sampled a plume high in biogenics that WRF-Chem predicts to be southeast of the aircraft.We use the WRF-Chem predictions from this plume and not the flight location in our comparison.

Comparison of modeled and measured concentrations
The measurements and model results are averaged into 100 mbar bins to evaluate the ability of the model to capture the regional chemistry, as shown in Figs. 2 and 3. Overall the model compares favorably to the measurements of the different components of NO y as well as OH, HO 2 , and O 3 , with the mean modeled value always within one standard deviation of the measured mean.NO, OH, and HNO 3 are, however, consistently higher in the model than in the measurements, with the mean in each bin higher than the measurement by an average of 3 pptv (27 %), 0.04 pptv (27 %), and 40 pptv (27 %), respectively.The mean values of the modeled total peroxy nitrates ( PNs) and O 3 are lower than the measurements by 35 pptv (19 %) and 6 ppbv (14 %), respectively.Some of the discrepancies are due to differences between modeled and observed clouds as reflected in comparison of measured and modeled photolysis rates of NO 2 that are 30 % higher on average in WRF (not shown).Overall, it can be concluded that within error, the model captures the regional features NO y , HO x , and O 3 despite some biases in the means of individual species.
Figure 3 shows the measurement-model comparison for the major biogenic species.As stated in Sect.2.1, emissions over the entire domain have been reduced by 20 % for isoprene and have been doubled for monoterpenes to more closely match the ARCTAS observations.Due to errors in the WRF-Chem representation of the wind field, there are still discrepancies between the measurement and model.In the 950 mbar bin the mean value of the model and measurements agree to within 4 pptv (3 %); however, since only αand β-pinene are measured during ARCTAS, this number is likely lower than the actual total monoterpene concentration.For instance the emissions of camphene and 3-carene were larger than those of α-pinene as measured by enclosure measurements of black spruce trees (Fulton et al., 1998).Black spruce trees are prevalent in the Canadian boreal forest.Summertime measurements in the Finnish boreal forest  indicate that α-and β-pinene account for just over 60 % of the monoterpene concentration (Hakola et al., 2012).
There is more isoprene, 177 pptv (69 %), in the model than in the measurements at the lowest altitude, but at higher altitudes the model has less isoprene than was measured (by 113 pptv -42 % at 850 mbar (Fig. 3b)).At 950 mbar the ratio of methacrolein (MACR) to isoprene is similar in the model (0.46) and the measurements (0.50), but the ratio of methyl vinyl ketone (MVK) to isoprene is much different (0.54 in the model compared to 0.93).This suggests the branching of MVK and MACR in either the OH oxidation or ozonolysis of isoprene in this low-NO x environment is not described properly in the chemical mechanism (e.g., Liu et al., 2013), that there are some transport issues affecting the photochemical age of the air masses, or that depositional processes are poorly represented.
Figure 4a and b show comparisons of the WRF-Chem ANs production to the calculation of AN production from the VOCs measured during ARCTAS (Browne et al., 2013).The two different calculations both show that biogenic precursors (isoprene, MVK, MACR, and monoterpenes) account for the vast majority of AN production.In WRF-Chem, monoterpenes account for only 16 % of the total production, compared to 30 % during ARCTAS.This is consistent with the lower concentration of monoterpenes in the model (Fig. 3a).
The composition of ANs as calculated by WRF-Chem is shown in Fig. 5a along with the ARCTAS observations of the contribution of small (≤ C5) alkyl nitrates measured in the whole-air samples (Simpson et al., 2011) to the total ANs measured by thermal dissociation laser-induced fluorescence (full details regarding this measurement during ARCTAS can be found in Browne et al., 2013).During ARC-TAS the small alkyl nitrates measured in whole-air samples accounted for 26 % (median -Fig.5b) of the ANs measurement, which is higher than the WRF-Chem prediction of 19 %.These small differences in the long-lived nitrates are likely associated with the boundary conditions.Due to their long lifetime, they do not affect any of the conclusions about the monoterpene chemistry.
In general, after adjustment of the emissions, the modeled and measured vertically averaged profiles of select BVOCs, NO y constituents, HO x , and O 3 agree to within one standard deviation over the domain (the boreal forest of Canada), and thus it can be concluded that the chemistry is represented well.There do exist some biases in the modeled mean values; however, the results presented below are fairly insensitive to these biases since two model runs are being compared to each other.

Monoterpene nitrate abundance, formation, and fate
The role of MTNs can be assessed using a 24 h boundary layer average from the 15 days corresponding to the ARCTAS measurements (28 June-13 July 2008) and using model output from areas located north of 49 • N and defined as forest or wooded tundra in the USGS land use categories, an area of 5.59 × 10 6 km 2 .NO x is defined as NO + NO 2 + NO 3 + 2 × N 2 O 5 .In addition to assuming that the oxidation of MTNs forms a nitrate that behaves identically to HNO 3 as in the BASE case, it is instructive to compare a simulation in which monoterpene-derived nitrates release NO 2 upon oxidation (referred to as RECYCLE).These two situations represent extremes, and the likely situation is between them.Since there are no experimental constraints on the NO x recycling from MTNs, the differences between the RECYCLE and BASE cases are useful in defining a range for the effects of MTNs.

Abundance
In the BASE case, over most of the boreal forest, firstgeneration MTNs account for ∼ 20 % of ANs and secondgeneration MTNs account for ∼ 10-15 % (not shown).Together, the MTNs account for 30-40 % of ANs over most of the domain and 60-80 % over British Columbia (not shown).However, although the fractional contributions over British Columbia are large, the absolute concentrations are very small in this region (Fig. 6a).The remaining ANs are primarily of biogenic origin.In the northern part of the domain (north of approximately 56 • N) ∼ 90-95 % of AN are biogenic.This number decreases to ∼ 80-85 % in the southern portion (which has better overlap with the ARC-TAS flight data), due to more anthropogenic influence (not shown).
Monoterpene nitrates are a significant contribution to NO y , accounting for ∼ 8-12 % of NO y over most of the domain (Fig. 6b); however in general they contribute less to NO y (∼ 6-8 %) in areas where the absolute concentration of NO y is highest (Fig. 6a).These high concentrations are in southern Canada, where the anthropogenic influence on total NO y is larger.In the RECYCLE case the contribution of MTNs to NO y has the same spatial distribution; however, the contribution to NO y is reduced to ∼ 4-8.5 % over most of the domain.

Formation
Monoterpene nitrates may be formed as a result of either OH or NO 3 oxidation.Figure 7   Production is dominated by OH chemistry for NO x concentrations < ∼ 75 pptv and is dominated by NO 3 chemistry for higher concentrations of NO x .Since these averages are weighted by mass, this split between NO 3 and OH production is not an artifact due to differences in boundary layer heights between day and night.Rather the high production rates are due to a sharp increase in NO 3 and in the NO 3 -to-OH ratio above 50 pptv NO x (not shown).There are no significant differences between the BASE and the RECYCLE (not shown) cases in the absolute or fractional production of MTNs.

Fate
Figure 8a shows the fractional loss of MTNs to OH, photolysis, and deposition for the BASE case.In the BASE case it is assumed that, upon oxidation, first-generation MTNs are converted into a more highly functionalized nitrate that behaves identically to HNO 3 .Since this oxidation step re- tains the nitrate functionality, it is not a loss of MTNs.The loss of MTNs is dominated by deposition (due to the rapid depositional loss of the second-generation MTNs), with a small contribution from the photolysis of both first-and second-generation MTNs.In contrast, for the RECYCLE case (Fig. 8b), oxidation (and thus the return of NO x to the atmosphere) is the primary contributor (48 %) to the loss of MTNs.Although oxidation by OH is more important, ozonolysis accounts for roughly one-third of the NO x loss.Because most of the deposition loss in the BASE case is due to deposition of the second-generation nitrate, which is not formed in the RECYCLE case, deposition is a smaller loss process in the RECYCLE case (Fig. 8b) than in the BASE case (Fig. 8a); however it is still important, accounting for 16 % of the loss.The calculations show that the majority (93 %) of the MTN loss removes NO x from the atmosphere by deposition in the BASE case, whereas this permanent removal drops to 16 % of the MTNs in the RECYCLE case.
The averaged MTNs loss rate is 8 % greater in the RE-CYCLE case (1.58 pptv h −1 ) compared to the BASE case (1.47 pptv h −1 ); however the averaged production rate is ∼ 7 % greater in the RECYCLE case as well.These differences are due to the changes in OH, O 3 , and NO x that are discussed in the next section.As a result of the slightly higher net MTNs loss in the RECYCLE case, the total concentration of MTNs is higher in the BASE case, with the second-generation MTNs accounting for most of this difference.In fact, the concentration of the first-generation MTNs is slightly larger (∼ 5 %) in the RECYCLE case.This larger value comes from the difference between the production of first-generation MTNs and their loss through reaction with OH and that O 3 is larger in the RECYCLE case.
5 MTNs and NO x , NO y , OH, and O 3

NO x and NO y
The lifetime of NO x and the mechanism of NO x removal are affected by MTN chemistry, as shown in Fig. 9.The net chemical loss rates (loss of NO x -production from photolysis and oxidation) are shown as a function of NO x to illustrate how, in low-NO x environments, the formation of ANs represents the dominate NO x loss as discussed in Browne and Cohen (2012).Figure 9 also shows that the net NO x loss to MTNs (red lines) is always larger than the NO x loss to PNs and in the BASE case is similar to the NO x loss to HNO 3 .Monoterpene nitrate formation thus represents a significant NO x sink.The decrease in NO x loss to MTNs in the RECYCLE case compared to the BASE case is evident when the NO x concentrations of the BASE and RECYCLE cases are compared.The oxidation of MTNs in the RECYCLE case results in a ∼ 8-14 % increase in NO x (mean of ∼ 5 pptv and max of 25 pptv) (Fig. 9a).This increase in NO x results in an increase in PNs by almost the same amount (mean 8.5 pptv and max of 30 pptv) with the same spatial distribution (not shown) in the RECYCLE versus the BASE case.

OH and O 3
The NO x differences between the BASE and RECYCLE cases in turn affect the concentrations of O 3 and OH.Changes in OH follow the same spatial pattern as NO x (Fig. 10a) and are about half as large -increases of ∼ 3-6 % (mean of 5.2 × 10 4 molecules cm −3 and max of 2.3 × 10 5 molecules cm −3 ) in the RECYCLE case over the BASE case.Ozone also increases by up to 3 % (Fig. 10b).In absolute values, ozone increases by a mean of 0.4 ppbv and a maximum of 1 ppbv.The spatial pattern of the ozone change is shifted slightly downwind from the NO x and OH changes, reflecting the slower kinetics of ozone production compared to the timescale of the kinetics affecting OH.

Mid-tropospheric effects
In addition to the boundary layer averages discussed above, we see changes between the RECYCLE and BASE at 600 mbar are ∼ 4 % for PNs and 1-2 % for O 3 over the eastern part of the domain.NO x changes by ∼ 2.5-5 % over most of the domain, with the hot spots occurring in the northwest and the east.By altering the mid-tropospheric burden, as well as the boundary layer burden of peroxy nitrates, monoterpene nitrates act to change the export of nitrogen from the continent, thus affecting ozone and OH concentrations downwind.These changes, while different in magnitude and location, should be qualitatively similar to the effects of isoprene nitrate chemistry, which have been discussed in detail in previous global modeling studies (e.g., von Kuhlmann et al., 2004;Fiore et al., 2005Fiore et al., , 2011;;Ito et al., 2007Ito et al., , 2009;;Paulot et al., 2012).

Hydrolysis of tertiary MTNs
Recent studies have shown that tertiary nitrates undergo rapid hydrolysis in aqueous solutions (Darer et al., 2011;Hu et al., 2011) and laboratory-generated SOA (Liu et al., 2012), with the assumed product being HNO 3 and thus a "permanent" loss of nitrogen from the atmosphere.Browne et al. (2013) provide evidence in the ARCTAS data that particle-phase hydrolysis of MTNs is likely a significant loss process and an important source of HNO 3 .Due to the limitations of the model representation of organic aerosol (and its water content), the calculation of this chemistry is not attempted in WRF-Chem, but rather its effect is estimated using the organic aerosol loadings observed during ARCTAS.Two distinct regions of organic aerosol loading were encountered in the boundary layer during ARCTAS (excluding air masses with recent biomass burning or anthropogenic influences): a region with low loading (average of 1 µg m −3 ) and a high loading (average of 6.6 µg m −3 ).The area of high loading was correlated with higher concentrations of acetone, an oxidation product of some monoterpenes, consistent with previous observations of high SOA yields from monoterpene oxidation (e.g., Tunved et al., 2006;Fry et al., 2009;Hallquist et al., 2009;Slowik et al., 2010;Fry et al., 2011).WRF-Chem predicts significantly less acetone (24 h average boundary layer maximum of ∼ 500 pptv) than was measured during ARCTAS (boundary layer median ∼ 2 ppbv) by proton transfer reaction mass spectrometry (Wisthaler et al., 2002), possibly indicating that monoterpene emissions and oxidation rates are underestimated or that the boundary conditions for acetone are too low.WRF-Chem does, however, predict areas of higher acetone concentration that are correlated with higher monoterpene concentrations.In these areas of high acetone (≥ 250 pptv) the organic aerosol loading is assumed to be 6.6 µg m −3 .Elsewhere we assume 1 µg m −3 of aerosol.A hydrolysis loss rate is calculated only for the firstgeneration MTNs since presumably the formation of secondgeneration MTNs would be greatly diminished if hydrolysis of MTNs results in HNO 3 .The calculation includes the assumption that there is sufficient liquid water content in the aerosol and uses the same two hydrolysis rates as in Browne et al. (2013) to represent the possible range of hydrolysis rates.As discussed in Browne et al. (2013), there is substantial variation in the hydrolysis rates that have been reported in the literature.This range is likely due to variation of the rate depending upon the molecular structure of the nitrate, although comparing bulk solution experiments (Darer et al., 2011;Hu et al., 2011) to aerosol experiments (Liu et al., 2012) may also explain part of the variation.Additionally it is assumed that 75 % of the MTNs formed from the reaction of OH with monoterpenes are tertiary, with a vapor pressure of 4 × 10 −6 torr, and that NO 3 initiated oxidation results primarily in secondary organic nitrates whose lifetime to hydrolysis is long and can be ignored.With these assumptions and conditions, the calculated loss rate of MTNs and a corresponding production rate of HNO 3 averaged over the entire domain of our calculations ranges from 0.11 pptv h −1 (hydrolysis rate of 4.63 × 10 −5 s −1 ) to 1.01 pptv h −1 (hydrolysis rate of 4.15 × 10 −4 s −1 ).For comparison, the production rate of HNO 3 via reaction of NO 2 with OH in the domain is 3.14 pptv h −1 , and thus this HNO 3 source is ∼ 3.5 to 32 % as large as the known daytime HNO 3 source.
This extra source of HNO 3 initially seems incompatible with the HNO 3 budget since WRF-Chem generally predicts larger concentrations of gas-phase HNO 3 (Fig. 2g) than indicated by the ARCTAS measurements.This measurementmodel discrepancy of HNO 3 is in part due to the higher OH concentrations predicted by WRF-Chem and the lower particle-phase NO − 3 predicted by WRF-Chem compared to the observations.This extra loss would also necessitate an increase in the source of ANs.It is reasonable to believe that the source of ANs may be larger; in particular, monoterpenes may be increased (or sesquiterpene emissions added), while still maintaining agreement between the measurements and the model.Thus, this back-of-the-envelope calculation for this loss process is compatible with the measurementmodel comparison.

Discussion
The 3-D CTM calculations presented here indicate that oxidation of monoterpenes is a major sink of NO x , a source of low-vapor-pressure MTNs, and a likely source of HNO 3 .Furthermore, given the differences in O 3 and OH observed between the BASE and RECYCLE cases here, even larger differences are likely when comparing chemical mechanisms with MTN representation to those without.
The chemistry as parameterized here represents a range that brackets some of the key uncertainties of MTNs, while remaining consistent with the measurements from ARCTAS.Although there remain important issues to resolve concerning the gas-phase and multiphase chemistry of monoterpenes and MTNs, the results of these calculations indicate that this chemistry will be locally important wherever monoterpene emissions are a large fraction of VOC reactivity.In addition to these local changes in the chemistry of the atmosphere, the depositional loss of monoterpene nitrates will have important impacts on soil nitrogen and, consequently, the carbon cycle.
To our knowledge, there has been no prior work specifically looking at how the chemistry of MTNs impacts the composition and oxidizing capacity of the troposphere.Past global modeling studies, however, have included the oxidation of monoterpenes and the formation of MTNs (Horowitz et al., 2007;Ito et al., 2007Ito et al., , 2009)), but have investigated the impact of MTNs in conjunction with isoprene-derived nitrates.Nevertheless, the results of these studies imply that modeled concentrations of NO x , OH, O 3 , and PNs are sensitive to the assumptions regarding the formation and fate of biogenic organic nitrates.Our finding that the production of MTNs from NO 3 chemistry accounts for the majority of MTNs production at higher NO x concentrations is consistent with the results of Horowitz et al. (2007) for isoprene nitrates over the eastern United States and with the results of Pye et al. (2010), who found that the aerosol produced from monoterpenes and sesquiterpenes doubled over the southeastern United States when NO 3 oxidation of these species was included in the model.Furthermore, a recent 1-D model study simulating a northern Michigan forest calculated a nighttime production rate larger than the daytime production rate for the monoterpene-derived nitrates (Pratt et al., 2012).Together, these studies in conjunction with the work presented here indicate that the chemistry of MTNs is necessary to accurately simulate SOA and tropospheric chemistry.
Although the focus here was on the boreal forest of Canada, where there were observations to test the calculations, similar results are likely wherever monoterpene emissions are abundant, including the Mediterranean (e.g., Owen et al., 1997;Davison et al., 2009;Seco et al., 2011), the Pearl River delta (e.g., Wang et al., 2011), and ponderosa pine forests (e.g., Bouvier-Brown et al., 2009;Kim et al., 2010).It is worth noting that the speciation of the monoterpenes in these environments is likely different and that a parameterization using two representative monoterpenes may not be sufficient to capture the resulting differences in chemistry.We believe that limitations in our knowledge of monoterpene emissions and their gas-phase chemistry likely overshadow any uncertainties introduced with the use of two representative species in a chemical transport model.Nonetheless, our work has demonstrated that gas-phase monoterpene chemistry is important and further improvement in emissions and representations of their chemistry appropriate for reduced chemical mechanisms is warranted.
It is also likely that emissions of sesquiterpenes (C 15 H 24 ), whose emissions are currently lumped into the monoterpenes, are important sinks of NO x and sources of SOA.The oxidative chemistry of sesquiterpenes will likely have a different impact on the oxidative capacity of the atmosphere due to their higher reactivity and lower vapor pressures.For example, a recent study using 2-methyl-1-alkenes as surrogates for terpenes found > 22 % yields of particle-phase β-hydroxy organic nitrates with a 15-carbon alkene (Matsunaga and Ziemann, 2010).This large yield suggests that sesquiterpenes (and their nitrates) should have significant local impacts on both NO y and SOA.
While the global impacts of this chemistry were not investigated here, Paulot et al. (2012), however, show that over the tropics the formation of isoprene nitrates results in a net chemical NO x sink, while over the oceans isoprene nitrates act as a net NO x source.It can be expected that MTNs have similar effects on global chemistry.MTNs will act to increase the ozone burden if the NO x is released in areas with a higher ozone production efficiency than that in which it was initially sequestered.This effect will depend on the details of the correct MTN chemistry, both directly through the reactions of MTNs and indirectly through their effect on NO y partitioning, OH, and O 3 .For example, the concentration of PNs is higher in the RECYCLE case than in the BASE case.As a result, the transport of NO x reservoirs from the continent to the oceans is larger in the RECYCLE than in the BASE case -a fact that amplifies differences in downwind O 3 .

Conclusions
Using the WRF-Chem model with an expanded organic nitrate representation, we are able to reproduce observations of a wide suite of chemicals, including ANs, observed during ARCTAS.Monoterpene nitrates were ∼ 5-12 % of NO y over the boreal forest of Canada, and NO 3 chemistry is the most important production pathway for MTNs when NO x is greater than ∼ 75 pptv.

1235
The fate of MTNs depends on the assumed oxidation products.In the BASE case, where oxidation results in a more functionalized nitrate that behaves like HNO 3 , deposition dominates the loss of MTNs.In this calculation MTN and HNO 3 formation are approximately equal sinks of NO x .In the RECYCLE case, where oxidation returns NO x to the atmosphere, MTNs are still ∼ 17 % of the net NO x removal and this sink is similar to or greater than the net chemical sink from peroxy nitrates.These calculations are likely lower bounds on the net NO x sink represented by MTNs as particlephase hydrolysis of tertiary MTNs is likely an important loss process for MTNs, as is hinted at by the ARCTAS observations (Browne et al., 2013) and as suggested by laboratory measurements (Darer et al., 2011;Hu et al., 2011;Liu et al., 2012).The particle-phase hydrolysis could be between ∼ 7 and ∼ 70 % of the MTNs loss rate calculated using the WRF-Chem results as well as an additional source of HNO 3 that ranges from a few percent to one-third as large as the source from the reaction of NO 2 with OH.
A complete representation of monoterpene chemistry will strain the capabilities of modern CTMs.However, until there is a more thorough understanding of monoterpene oxidation and the appropriate means of representing their chemistry in a condensed chemical mechanism, we recommend that models implement a basic gas-phase chemistry of monoterpenes in addition to their SOA chemistry.In parallel, evaluation of key parameters in the lab and consequences for ambient O 3 , NO x and OH should be explored both on local and global scales.

Figure 1
Figure 1 Map of ARCTAS measurements of ΣANs used in this analysis.These locations are removed from recent biomass burning influences and are at pressures greater than 600 mbar.

Fig. 1 .
Fig. 1.Map of ARCTAS measurements of ANs used in this analysis.These locations are removed from recent biomass burning influences and are at pressures greater than 600 mbar.

Fig. 3 .
Fig. 3. Same as Fig. 2 but for (a) monoterpenes (only α-and βpinene were measured during ARCTAS), (b) isoprene, (c) methyl vinyl ketone, and (d) methacrolein.The apparent maximum in the ARCTAS profiles (red) at 800-900 mbar is due to the plume high in biogenics sampled during flight #20 at 21:51 to 22:28 UTC.If this plume is removed, both the model and the measurements show a smoothly decreasing profile with increasing altitude, as expected for reactive species with surface emissions.
Figure 4 a) Instantaneous ΣANs production rate as calculated by WRF-Chem along the ARCTAS flight track at altitudes less than 2 km.b) The calculated instantaneous ΣANs production based on the volatile organic compound precursors measured during ARCTAS for altitudes less than 2 km.Details regarding this calculation can be found in Browne et al. (2013).
Figure 5 a) The ΣANs composition predicted by WRF-Chem along the flight track at altitudes 3 Fig. 5. (a) The ANs composition predicted by WRF-Chem along the flight track at altitudes less than 2 km.(b)The ANs composition during ARCTAS for the same altitudes.The ≤ C5 nitrates were measured in the whole-air samples and the total was measured using TD-LIF.Details are inBrowne et al. (2013).
Figure 6 a) 24 h, boundary layer averaged absolute concentration of MTNs in the BASE case.b) Fractional contribution of MTNs to total oxidized nitrogen in the BASE case.
Fig. 6.(a) The 24 h, boundary-layer-averaged absolute concentration of MTNs in the BASE case.(b) Fractional contribution of MTNs to total oxidized nitrogen in the BASE case.

Figure 7 Fig. 7 .
Figure 7 The 24 h, boundary layer average production rates of MTNs by OH chemistry (red line 2

Figure 9 Fig. 9 .
Figure 9 The 24 h, boundary layer averaged, net chemical NOx loss partitioned by class of 2

Figure 10 Fig. 10 .
Figure 10 Percent differences ((RECYCLE-BASE)/BASE×100) between the RECYCLE and BASE cases for the 24 h, boundary layer averaged a) NO x and b) O 3 concentrations.

2014 1232 E. C. Browne et al.: On the role of monoterpene chemistry 32 Figure
8 The partitioning of the 24 h, boundary layer averaged loss processes of MTNs in a) the BASE case and b) the RECYCLE case.The averaged total loss is 8% larger in the RECYCLE case compared to the BASE case.Loss of MTNs through oxidation by OH or O3 and by photolysis results in the return of NOx to the atmosphere.Loss of MTNs through deposition removes NOx from the atmosphere and introduces it into the biosphere.The partitioning of the 24 h, boundary-layer-averaged loss processes of MTNs in (a) the BASE case and (b) the RECYCLE case.The averaged total loss is 8 % larger in the RECYCLE case compared to the BASE case.Loss of MTNs through oxidation by OH or O 3 and by photolysis results in the return of NO x to the atmosphere.Loss of MTNs through deposition removes NO x from the atmosphere and introduces it into the biosphere.