Seasonal variability of atmospheric nitrogen oxides and non-methane hydrocarbons at the GEOSummit station , Greenland

Introduction Conclusions References


Introduction
The seasonality of ozone (O 3 ) and its precursors for photochemical production, such as NO x (NO x = NO + NO 2 ), peroxyacetyl nitrate (PAN) and non-methane hydrocarbons (NMHC), in the remote Arctic troposphere is governed by a combination of transport pathways, photochemistry and stratospheric influx (Klonecki et al., 2003;Stohl et al., 15 2006; Law and Stohl, 2007;Liang et al., 2011). Improving our knowledge on the seasonality of O 3 and its precursors and the relative importance of source regions and transport variability is essential as recent studies have suggested that tropospheric O 3 may have a large impact on radiative forcing and climate feedbacks in the Arctic region (Shindell et al., 2006;Shindell, 2007;Quinn et al., 2008). 20 Polluted air masses originating from anthropogenic and biomass burning sources in the mid-latitude regions can transport long-lived reservoir species of NO x , such as PAN, and nitric acid (HNO 3 ) to the arctic region (Wofsy et al., 1992;Wespes et al., 2012), which may reform NO x and result in enhanced levels far downwind from the emission sources (Beine et al., 1997;Walker et al., 2010). NMHC may also be trans- 25 ported in air masses from anthropogenic and biomass burning sources. The mole fractions of NMHC in the Arctic can vary greatly during the year due to seasonal variability O 3 precursors, whereas at higher altitudes, pollution plumes transported from Asia become important (e.g. Klonecki et al., 2003;Law and Stohl, 2007;Shindell et al., 2008;Walker et al., 2012;Wespes et al., 2012). A large contribution to the seasonality of O 3 and O 3 precursors in the Arctic troposphere is due to variability in the location of the Arctic polar front (Klonecki et al., 2003;Stohl, 2006). During winter in the Northern 10 Hemisphere, the polar front expands southward over North America, Europe and North Asia allowing direct transport of polluted air masses from sources within these latitudes to the Arctic. The Arctic polar front recedes in summer, reducing the impact of these pollution sources on the Arctic lower troposphere. However, it has been shown that the transport of emissions from biomass burning regions to the Arctic is possible during 15 summer (Stohl, 2006) and that they can strongly impact the atmosphere above Summit Station in Greenland . Results from a modeling study by Walker et al. (2012) using tagged emissions from the global chemical transport model GEOS-Chem show that during summer the primary emissions that impact the production of O 3 in the Arctic region were from high latitude regions, whereas, during the fall and winter 20 periods, transport of emissions from mid-latitude regions in North America and Europe is possible.
A number of studies have discussed the seasonality of surface O 3 Beine et al., 1997;Monks, 2000;Browell et al., 2003;Helmig et al., 2007b;Walker et al., 2012), nitrogen oxides (Barrie and Bottenheim, 1991;Honrath and Jaffe, 1992; production of NO x from PAN decomposition is expected to be small in this cold region. Therefore, enhanced mole fractions of nitrogen oxides are primarily a result of longrange transported pollution from anthropogenic or biomass burning sources in Europe, North America, and Asia or of downward transport from the stratosphere (e.g. Liang et al., 2011). A build-up of O 3 precursors during winter in the Arctic free troposphere 10 may have important implications for the tropospheric O 3 budget in the mid-latitudes during late spring and early summer (Gilman et al., 2010). Modelling studies have postulated that air masses originating from the Arctic region can result in the transport of NO y and NMHC to the North Atlantic and enhance tropospheric O 3 in this region due to the thermal decomposition of PAN (Honrath et al., 1996;Hamlin and Honrath, 15 2002).
This study utilizes 2 years of continuous measurements and model results to characterize the seasonally varying magnitude of O 3 and its precursors in the remote high altitude Arctic and the potential impact from transported pollution. Year-round measurements of NO x , NO y , PAN, O 3 and NMHC from the high altitude Greenland Envi- 20 ronmental Observatory at Summit (GEOSummit) station in Greenland are presented. The paper is structured as follows: in Sect. 2, the techniques to measure NO, NO 2 , NO y (total reactive nitrogen oxides NO y = NO + NO 2 + PAN + HNO 3 + HONO + others), PAN and NMHC are presented and the FLEXPART Lagrangian particle dispersion model utilized in this study is discussed. Section 3.1 discusses the seasonal cycles of O 3 pre- 25 cursors at the measurements site and the NO y speciation is investigated. In Sect. 3.2, the interannual variability and source contributions to enhanced O 3 precursors from anthropogenic emissions and biomass burning are discussed. Finally, a summary of the main findings is given in Sect. 4

GEOSummit Station
Measurements of NO x , NO y PAN and NMHC were performed at the GEOSummit Station (hereafter called Summit), Greenland (72.34 • N,38.29 • W, 3212 m.a.s.l) from July 2008 to July 2010. Inlets for the instruments were installed ∼ 7.5 m above the 5 snowpack on a meteorological tower located approximately 660 m south-west of the main camp within the "clean air" sector. Tubing and cables were routed through a heated pipe to a buried laboratory facility. 10 Measurement of NO, NO 2 and NO y were performed with an automated O 3 chemiluminescence detection system (Ridley and Grahek, 1990). The system was developed at Michigan Technological University and is based on the same design that was used in Newfoundland in 1996 (Peterson and Honrath, 1999) and subsequently installed at the Pico Mountain Site from 2002 to 2010 (Val Martín et al., 2006). NO 2 and NO y were 15 detected by chemiluminescence after reduction to NO using a photolytic NO 2 converter (Kley and Mcfarland, 1980) and a gold-catalyzed NO y converter in the presence of CO, respectively (Bollinger et al., 1983;Fahey et al., 1985). NO y is given as the sum of reactive nitrogen oxides. In the Arctic, NO y is primarily comprised of NO, NO 2 , PAN, HNO 3 , HONO and particulate nitrate (p-NO − 3 ). For the instrument used in this study, 20 a photolytic blue LED NO 2 converter (Air-Quality Design Inc., Colorado) was installed. Photolytic converters have lower conversion efficiencies than molybdenum converters, however interferences from other species photolyzing to NO, such as HONO and PAN are reduced (Pollack et al., 2011;Villena et al., 2012 (MFC) and the NO 2 and NO y converters were housed inside the inlet box on the tower to minimize the residence time of NO y species inside the PFA tubing. During each measurement cycle of 10 min, the NO and NO 2 signals were recorded as 30 s averages and NO y signals as 20 s averages, after a period of equilibration in each mode. Zero measurements of NO were performed at the start and end of each 5 measurement cycle by mixing O 3 with the sample upstream of the reaction chamber. The zero signals were measured to determine the interference signal in the reaction chamber, which was then subtracted from the measured signals. Calibration cycles, to determine the sensitivity of the instrument to NO and NO 2 converter efficiency, were performed every 12 h through the standard addition (10 cm 3 min −1 ) of ∼ 1 mmol mol The variability in the 20 and 30 s averaged data was compared to the expected value from photon counting statistics which are treated as a Poisson distribution. Measurements with variability greater than 3 times the Poisson value were then removed from the final dataset (∼ 4 % were removed with this filter). Evaluation of these peri-20 ods shows that they typically occur when the wind direction was from the main camp, confirming that local pollution is the main source of the variability. Additional filtering processes were implemented to remove bad data caused by (1) spikes from electronic noise or intermittent instrument malfunctions; (2) high variability due to periods when the skiway was groomed or periods not captured in the Poisson statistics filter and (3)  The NO, NO 2 , NO x and NO y data used in this work were further averaged over a 30 min period. NO x was determined as the sum of the NO and NO 2 measurements during each 10 min cycle. The overall uncertainty for the 30 min data is calculated from the root sum of the squares of the measurement accuracy, artifact uncertainty and precision. Maximum uncertainties for NO,NO 2 , NO x at 50 pmol mol −1 are 10 %, 17 % 5 and 19 %. For NO y , the uncertainty at 200 pmol mol −1 is < 20 % and typically 9 %.

Nitrogen oxides
Detection limits for the 30 min averages were determined from the 2σ precision of the instrument and error in the artifact. Detection limits for NO, NO 2 , NO x and NO y are 4 pmol mol −1 , 8 pmol mol −1 , 9 pmol mol −1 and 7 pmol mol −1 , respectively. Measurements below the detection limit were included in all averaging calculations to ensure 10 the final values were not biased. Further details on the calibrations performed and the precision and accuracy of the measurements are given in the Supplement.

Peroxy-acetyl nitrate
A commercial PAN gas chromatography analyzer (PAN-GC, Metcon, Inc., Boulder, CO) was installed alongside the NO xy instrument to determine PAN mole fractions. The PAN 15 instrument is based on gas chromatography with electron capture detection (GC-ECD). The instrument was equipped with a preconcentration unit to improve the detection limit whilst allowing for PAN sampling every 10 min. The preconcentration unit traps PAN and carbon tetrachloride (CCl 4 ) on a peltier cooled (5 • C) capillary column prior to injection onto the main GC column which was set to a temperture of 13 • C to reduce the thermal 20 decomposition of PAN. Ultra-pure nitrogen gas (99.9999 % purity) was used as the carrier gas for the PAN-GC. The instrument was calibrated approximately every week using a known amount of PAN, which was photochemically produced from the same NO-calibration gas used for the NO xy instrument described in Sect. calibration gas was sent to the inlet on the tower and then sampled by the GC-ECD. The conversion efficiency of NO to PAN was determined at the beginning and end of the measurement period through the standard addition of NO/NO 2 to the NO xy instrument. The conversion efficiency remained relatively constant throughout the measurement period at 96 ± 1 %. 5 The sensitivities determined from the weekly PAN calibrations were interpolated to the measurements to take into account any drifting. CCl 4 was also used as an internal reference during periods when calibrations were not taken (Karbiwnyk et al., 2003). The atmospheric concentration of CCl 4 should be relatively constant; therefore any changes in the CCl 4 peak area would be caused by changes in the instrument sensitiv-10 ity. During a period between 28 February 2009 and 17 May 2009 there was a gap in the calibrations caused by a blockage in the tubing that delivered the PAN calibration gas to the inlet. During this period the relationship between the CCl 4 peak area and PAN sensitivity from the previous calibrations was used to obtain the PAN sensitivity. Over the duration of the measurement period the detector became dirty resulting in drifting 15 and a noisy baseline. Due to this issue no data after 28 April 2010 were included in the analyses here.
Similarly to the NO x and NO y data, the PAN measurements were averaged over 30 min. The total uncertainty for the 30 min averaged PAN mole fractions was determined from the precision of the instrument (estimated as 2σN 0.5 , where N is the num-20 ber of points averaged in 30 min (N = 3)) and the uncertainty in the calibration standard. Uncertainty in the PAN calibration standard is associated with uncertainties in (a) the calculation of the NO addition, (b) the conversion of NO to PAN from the calibration unit and (c) variability in the PAN sensitivity between calibrations. The total uncertainty was estimated to be 16 % during normal operation. This value increased to 22 % during the 25 period in spring 2009 when there were no calibrations. The limit of detection (LOD) of the instrument was estimated from the peak to baseline noise ratio. The LOD is defined as the mole fraction giving a signal to noise (S/N) ratio of 3. The baseline noise was determined from a region just after the PAN peak for Introduction

Non-methane hydrocarbons
NMHC were continuously sampled using a fully automated and remotely controlled GC system that was specifically designed for this study. Details of the setup at Summit are given in Helmig et al. (2014a). The GC is a further development of the instrument oper-10 ated at the Pico Mountain Observatory and described in detail by (Tanner et al., 2006). The instrument provided ∼ 6000 ambient measurements of C 2 -C 6 hydrocarbons, in addition to ∼ 1000 blank and standard runs from June 2008 to July 2010. The inlet for the GC instrument was installed on the same tower as the PAN, NO y and NO x inlets. The instrument relies on a cryogen-free sample enrichment and injection system. All . Blanks and standard samples were injected regularly from the manifold. The gravimetric and whole air standards that were used were cross-referenced against our laboratory scale for volatile organic compounds, which has been cross-referenced against national and international Introduction 3-5 %, yielding a combined uncertainty estimate of ∼ 5 %. The instrument achieves low single digit pmol mol −1 detection limits.

Ozone
Surface O 3 was measured by an O 3 analyzer located in the Temporary Atmospheric Weather Observatory (TAWO) building a few hundred meters from camp by the Na-5 tional Oceanic and Atmospheric Administration (NOAA) as part of the core atmospheric measurements that began in 2000 (Petropavlovskikh and Oltmans, 2012

FLEXPART
The Lagrangian particle dispersion model FLEXPART was utilized to identify potential periods when polluted air masses impacted the measurement site. FLEXPART simulates atmospheric transport using wind fields from global forecast models to determine source to receptor pathways of air masses (Stohl et al., 2005). The model was 15 driven with meteorological analysis data from the European Centre for Medium Range Weather Forecasts (ECMWF) and run backward in time in so-called "retroplume" mode (Stohl et al., 2003). Every 3 h 40 000 particles were released from the measurement site location and followed backwards in time for 20 days. Sensitivities to anthropogenic and fire emissions were determined during the backwards simulations and are propor-20 tional to the particle residence time over the source areas. In this work a black carbon tracer was used to simulate both anthropogenic (BC anthro ) and biomass burning emissions (BC fire ). The BC tracer was susceptible to both wet and dry deposition during transport. The wet deposition process is simplified in the simulations (no ageing of BC with conversion from hydrophobic to hydrophilic properties) and may result in an  . However, for this study the tracer was only used to identify events; therefore absolute BC values were not required.  Mountain, Svalbard (Beine et al., 1997;Solberg et al., 1997;Beine and Krognes, 2000) and Alert, Northwest Territories, Canada (Worthy et al., 1994;Dassau et al., 2004). There is a rapid transition towards lower levels of PAN in the summer, with a minimum mean monthly average of 66 ± 29 pmol mol −1 in July. NO y mole fractions do not decrease as quickly from spring to summer as PAN and reach a minimum monthly 10 average of 100 pmol mol −1 in September. We find that the PAN and NO y summer mole fractions observed here are comparable to previous measurements performed at the same site in 1998 and 1999, when observed PAN levels were typically 20-150 pmol mol −1 and NO y levels ranged between 100-300 pmol mol −1 (Honrath et al., 1999;Ford et al., 2002). The slower decrease in NO y from spring to summer, compared 15 to PAN, is a result of the presence of NO x and odd NO y (odd NO y = NO y − PAN − NO x ) over the summer months and is discussed further below. The seasonal cycle of PAN is governed by the rate of thermal decomposition and transport patterns. The warmer summer temperatures result in the decomposition of PAN during long range transport, additionally, during the summer months the polar 20 front recedes north, thus reducing the potential for anthropogenic emissions to reach the measurement site (Beine and Krognes, 2000;Stohl, 2006). Measurements have shown that PAN is typically the largest contributor to NO y in the Arctic, due to the rapid formation of PAN near the source region and a long lifetime in the free troposphere (Solberg et al., 1997;Munger et al., 1999;Ford et al., 2002;Alvarado et al., 2010;25 Singh et al., 2010;Liang et al., 2011). However, there have been very few studies on the seasonal variability of the NO y speciation in the Arctic due to limited measurements over winter months. The full annual cycle of NO y contributions from PAN and NO x during this study provides some information on the NO y speciation, year round at Summit. Introduction The results plotted in Fig. 2 show that PAN is the dominant form of NO y all year round, with monthly average [PAN]/[NO y ] ratios reaching a maximum of 78 % in April and a minimum of 49 % in July. Over the summer, NO x contributes approximately 10-13 % to the total NO y . In winter this decreases to 4 % and often NO x levels were below the dectection limit of the instrument. What is particularly striking about the NO y speciation 5 shown in Fig. 2 is that odd NO y levels can be significant, particularly over winter, when they reach a maximum monthly mean of 95 ± 36 pmol mol −1 (mean ± SD) in February. Odd NO y decreases to approximately 30-50 pmol mol −1 in the summer months; however, this still accounts for ∼ 20-38 % of the total NO y during this period. Snowfall rates increase during the summer months over Summit (Dibb and Fahne-10 stock, 2004) therefore an increase in deposition of water-soluble species such as HNO 3 to the snowpack may result in the depletion of ambient odd NO y in the summer. The increase in solar radiation may also play an important role in the reduction of odd NO y species in the summer. Solberg et al. (1997) observed a decrease in odd NO y with increasing solar UV radiation in Spitsbergen, Norway. The authors suggested that 15 species such as HONO, HNO 3 , NO 3 , N 2 O 5 , HO 2 NO 2 , and alkyl nitrates may contribute to NO y over the winter with the impact reducing in spring due to an increase in photolysis. A study on the seasonal variability of alkyl nitrates at Summit in 1998-1999 found that the light C 1 -C 4 alkyl nitrates peak through late winter until April with total mole fractions between 30 and 42 pmol mol −1 (Swanson et al., 2003). Therefore, alkyl 20 nitrates could account for a large portion of the odd NO y observed during the winter months at Summit. However, there still remains a large fraction of NO y unaccounted for over winter and further measurements are required to determine both the species and sources of this odd NO y . The seasonal cycle for NO x does not follow PAN and NO y at Summit. As shown in 25 of PAN is a possible source of NO x during spring and summer months, however, the contribution is expected to be very small in this high latitude region as temperatures during the measurement period were always below 0 • C (Beine et al., 1997). Thus, the increase of NO x with radiation in spring suggests a possible photochemical source. The role of snowpack emissions on NO y species within the arctic boundary layer is 5 still uncertain, however, studies have suggested that photochemical reactions within the snowpack may result in the release of NO x and HONO to the overlying atmosphere (e.g., Honrath et al., 1999Honrath et al., , 2000aHonrath et al., , b, 2002Munger et al., 1999;Beine et al., 2002;Dibb et al., 2002;Dominé and Shepson, 2002;Beine et al., 2003;Grannas et al., 2007). During late spring to summer, odd NO y species can contribute over twice as much 10 as NO x to the total NO y . To investigate the source of these species and the possible impact from snowpack photochemistry we have analyzed the diurnal variability of NO x , NO y and odd NO y at Summit. Our measurements of NO x and NO y mole fractions at a height of ∼ 7.5 m above the snowpack display clear diurnal cycles from April-June ( Fig. 3a, b). It is observed that the amplitude of the NO y diurnal cycle is greater than for 15 NO x , with average diurnal amplitudes of 33 pmol mol −1 and 14 pmol mol −1 for NO y and NO x respectively. It has been hypothesized that photochemically produced odd NO y species such as HNO 3 and HONO may account for some of the NO y diurnal variability at Summit (Ford et al., 2002). An analysis of the diurnal cycle for odd NO y (Fig. 3c) averaged over April-June from 2008-2010 shows that the odd NO y peaks just after solar 20 noon in our measurements, suggesting a photochemically produced odd NO y species may be present. The diurnal variability of ambient NO y species above the snowpack is further complicated by vertical mixing and boundary layer dynamics, which may vary with season. For example, the downward transportation of pollution from aloft due to a growing boundary layer may result in a daytime maximum in NO x and NO y , which 25 then decreases at night due to surface uptake. There is also a possible contribution to odd NO y in the summer from long range transport of reactive nitrogen species such as HNO 3 and alkyl nitrates as these species have previously been observed in anthropogenic and biomass burning plumes in the Arctic (Liang et al., , 2012). It should also be noted that some of the variability in the NO y speciation discussed here may be influenced by uncertainties in the PAN measurements which increased in spring 2009. The year-round measurements obtained from Summit in 2008-2010 provide new insight on the relative role of photochemistry and boundary layer stability on diurnal cycles of nitrogen oxides and is the subject of future investiga-5 tion. Figure 4 shows the results for the C 2 -C 5 alkane NMHC measured during 2008-2010 at Summit. The NMHC show a strong seasonal cycle with maximum mole fractions during the winter and early spring period and a rapid decline towards the summer 10 due to an increase in photochemical processing. The monthly mean averages for the C 2 -C 5 NMHCs are given in Table 2. The phase of each NMHC is shifted due to the rate of reaction with OH. The lightest of the NMHC shown in Fig. 4 is ethane (C 2 H 6 ), which peaks in March with a monthly mean of 1974 ± 209 pmol mol −1 (mean ±1σ) and reaches a minimum of 633 ± 65 pmol mol −1 in August. Heavier NMHC have lower mole 15 fractions and peak earlier in the year and reach a minimum earlier in summer as their rate of reaction with OH is much faster. The seasonal cycle of NMHC at Summit including NMHC firn air measurements from 2008 to 2010 have previously been presented in detail (Swanson et al., 2003;Helmig et al., 2014a, b). The seasonality of NMHC can provide some insight into the potential for the photo-20 chemical production of O 3 in the Artic troposphere. The accumulation of O 3 precursors, such as nitrogen oxides and NMHC over winter has been suggested as a potential in situ source of O 3 that may contribute to the tropospheric O 3 peak observed in spring in the Arctic (e.g., Penkett et al., 1993;Honrath et al., 1996;Monks, 2000;Blake et al., 2003;Wang et al., 2003 this study show similar results for NMHC, with the sum of the C 2 -C 6 NMHC decreasing by ∼ 4.4 ppbC from February to May. The magnitude of the O 3 increase, at ∼ 8 ppbv, is smaller than observed during TOPSE, however, the photochemical processing of NMHC in spring may contribute to the spring time peak of O 3 over Greenland.

burning emissions
In this section the interannual and short term variability in the measured species at Summit from 2008-2010 due to variability in transport pathways and the relative source contributions of pollutants from North America, Europe and Asia are investigated. Anthropogenic and biomass burning emissions are considered separately, over different 10 seasons. Figure 4 shows that there is considerable variability in the NMHC mole fractions superimposed on the seasonal cycle; in particular during the winter months, suggesting polluted air masses were transported to the measurement site during this period. 15 Mean ±1σ Table 3); 25 therefore, we focus the analysis on anthropogenic emissions between these months. To investigate the source of the observed variability and the impact on ozone precursor levels at Summit, the anthropogenic tracer from FLEXPART retroplume simulations (BC anthro ) and NMHC emissions ratios were used to determine changes in the transport pathways and relative source contributions of anthropogenic emissions from different continents.

5
An event with pollution transport was defined as identified when the BC anthro tracer was greater than 75th percentile of the total BC anthro during the 2 year measurement period (corresponding to BC anthro > 0.0082 pmol mol −1 ) for a minimum of 12 h. The FLEX-PART temporal resolution for backward simulations is 3 h, so identifying events when the BC anthro was enhanced for at least 12 h ensured that significant polluted air masses 10 impacted the site and also allowed for some temporal mismatches in simulated and observed plume arrivals. larly interesting is that during some of these events, low O 3 mole fractions (∆O 3 < 0) were measured. Analyses of ∆O 3 and FLEXPART BC tracer masses from July 2008 to July 2010 show that decreases in O 3 below the background level (when ∆O 3 was negative for at least 12 h and reached a minimum ∆O 3 level below −2 nmol mol −1 ) were observed throughout the year. During April to September 45 events with low ozone 15 were observed, however, these events were typically associated with low levels of pollution with only 10 out of the 45 events classified as polluted (as indicated by either high FLEXPART BC anthro (> 75th percentile) or BC fire (> 90th percentile) tracers). For the remaining 35 events enhancements in NO y , PAN and ethane were either low or negative, suggesting pollution levels were minimal in the sampled air masses. 20 Decreases in O 3 (where the minimum ∆O 3 < −2 nmol mol −1 ) that coincided with anthropogenic pollution events were observed during 28 periods from July 2008 to July 2010, with 21 of these events occurring between October and March, when sunlight is at a minimum. Thus, it is possible that the decrease in O 3 observed during winter/early spring is due to titration of O 3 by NO within the sampled air mass soon 25 after emission (Eneroth et al., 2007;Hirdman et al., 2010). For example, from 23 to 31 January 2010 (indicated by the shaded area in Fig. 7), O 3 decreased rapidly during two periods, coinciding with increases in O 3 precursors. Using the vertically integrated emission sensitivity (also called the total column sensitivity, measured in nanosecond Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | meters per kg) simulated by FLEXPART, the overall pathways of the air masses during these events can be determined. As shown in Fig. 9a, the air mass on 23 January originated from Northern Europe and was transported to Summit in only a few days. The mean weighted age of the plumes estimated from FLEXPART during this period decreased from ∼ 15 to 8 days (Fig. 7). 5 Studies have shown that NMHC ratios can provide an indication of the photochemical aging of the air mass as the rate of reaction of different NMHC, and hence the ratio, is dependent on the amount of photochemical processing that occurs during transport (Parrish et al., 2004;Helmig et al., 2008;Honrath et al., 2008). High photochemical processing results in a decrease in the ln([propane]/[ethane]) ratio as propane reacts 10 more readily with OH than ethane. The two low ozone events between 23 and 31 January 2010, coincided with enhancements in the ln([propane]/[ethane]) ratio suggesting low photochemical processing, and the impact of fresher air masses at the site. However, care must be taken when interpreting this result as dilution of the measured species during transport will also have an impact on the mole fractions measured at 15 the site. PAN and ethane reached peaks of 188 pmol mol −1 and 3 nmol mol −1 , respectively, during the event on 23 January, supporting the FLEXPART analyses which indicated a polluted air mass originating from Europe was sampled at the site (note there were no NO y or NO x data available during this period). Ozone, however, decreased by 6.2 nmol mol −1 below the monthly background level (O 3(bkg) = 41.6 nmol mol −1 ) to reach 20 a minimum level of 35.4 nmol mol −1 . The FLEXPART retroplume on this day shows that air mass resided in the lower ∼ 2 km during transport, until 1 day upwind when the air mass ascended over the surface of Greenland to the measurement site (Fig. 9a), thereby reducing the potential to mix with high O 3 from stratospheric origin. In contrast, a few hours later the transport patterns quickly changed, and the air masses sampled at 25 Summit originated from high altitudes over North Canada, as shown in Fig. 9b. As a result, O 3 levels increased and PAN and ethane decreased back toward their background levels (PAN (bkg) = 72 pmol mol −1 , C 2 H 6(bkg) = 1.55 nmol mol −1 ). Air originating from the high Arctic region was sampled at the site until 29 January, when the retroplume 13837 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | moved southward and air masses residing in the lower troposphere over North America transported polluted air to Summit (Fig. 9c). From 29-30 January, ethane, NO y , PAN and NO x all increased again by 1.3 nmol mol −1 , 267 pmol mol −1 , 146 pmol mol −1 and 32 pmol mol −1 respectively, from the calculated monthly background levels, and O 3 decreased by 2 nmol mol −1 .

5
Analyses of all the pollution events over winter and early spring indicate that low O 3 events from December to March typically coincide when sampling air masses originating from either Europe or North America, which have resided in the lower troposphere until ascending over Greenland to the measurement site (examples of FLEX-PART retroplumes are shown in Fig. S1a-e in the Supplement). In contrast, periods 10 identifed during winter as pollution events with positive O 3 enhancement values occurred when the air masses resided in the mid-troposphere during transport to the site ( Fig. S2a and b in the Supplement), thus allowing for greater mixing with air from high tropospheric or stratospheric origin.
In this study the impact from anthropogenic emissions, as identified through FLEX-

15
PART retroplume analyses, were the primary focus. However, enhancements in the measured species were also observed during periods which are not correlated with pollution events simulated by FLEXPART. For example, between 12 February and 1 March 2010, Fig. 7 shows two periods when enhancements in PAN, NO y , and ethane that do not coincide with high FLEXPART BC anthro tracer are observed. Analysis of the 20 BC fire tracer from FLEXPART also indicated no pollution plume from biomass burning origin. The retroplume analysis shows that air masses from these two events were transported over the far north region of Canada and remained in the arctic for many days before arriving at Summit. It is unlikely that the air sampled was from a stratospheric origin, as ethane levels were high and ozone decreased during these events. 25 It is possible that there may be an error in the simulated retroplume by FLEXPART or that the pollution originated prior to the 20 day simulation. Further investigations are necessary to determine the cause of these events and enhancements in the O 3 precursors. Introduction

Summer impact from biomass burning and anthropogenic events
The extended whiskers shown on the plots in Fig. 1 indicate a large amount of variability in the O 3 precursors during the summer months. Radiation, surface emissions, boundary layer height and changes in air mass sampling may all contribute to the variability observed, which is typically in the range of hours to days. Anthropogenic emis-5 sion impacts tend to be lower in the summer due to reduced transport from source regions, however pollution from anthropogenic and, especially, BB emissions can still impact the center of Greenland (Stohl, 2006), resulting in elevated mole fractions for short periods . Studies based on aircraft measurements and models during the ARCTAS campaigns in both spring and summer 2008 show that biomass 10 and anthropogenic plumes can result in elevated NO x , NO y , PAN and hydrocarbons in the Arctic (e.g., Alvarado et al., 2010;Singh et al., 2010;Hornbrook et al., 2011;Liang et al., 2011). The BC fire tracer from FLEXPART was used to identify periods at Summit that were potentially impacted by BB emissions. FLEXPART has been used to identify long 15 range transport of biomass burning emissions in many studies (e.g., Brioude et al., 2007;Stohl et al., 2007;Lapina et al., 2008;Quennehen et al., 2011Quennehen et al., , 2012Schmale et al., 2011;Cristofanelli et al., 2013). However, due to potential inaccuracies with the FLEXPART simulation of transport pathways, fire identification and tracer emission uncertainties, BB events identified may be under or overestimated. Biomass burn-20 ing events were characterized as having a FLEXPART BC fire tracer > 90th percentile ( 7 pmol mol −1 ). In total, 13 events were observed between July 2008 and July 2010 ranging in duration between 12 and 252 h. Details regarding the start date, duration, mean plume age, FLEXPART tracer levels and trace gas levels for each event are shown in Table 3. A more conservative threshold was applied here than for the anthro- 25 pogenic emissions in Sect. 3.2.1, so the events identified had significant BB but small anthropogenic signatures. Of these 13 events, 2 were identified as having potentially high anthropogenic signatures (BC anthro > 75th percentile) and were most likely plumes Introduction of mixed anthropogenic and biomass burning emissions, 5 events were identified as having medium anthropogenic signatures (75th percentile > BC anthro > 50th percentile) and the remaining 6 events were classified as having low anthropogenic signatures. The analysis of the source contribution from FLEXPART shows that the majority of the BB events (9 out of 13) originated in North America, with BB events originating 5 in Europe all occurring in March 2009. To ensure that the results were not biased through only identifying events during the measurement period, a statistical analysis of all the FLEXPART data from 2008, 2009 and 2010 was performed. For these 3 years, 23 BB events lasting longer than 12 h (both with and without anthropogenic mixing) were identified. Of these 23 events, ∼ 67 % of the BC fire tracer originated from North 10 America, 15 % from Europe and 19 % from Asia, confirming that North America is the primary source of BB emissions impacting Summit.
Analyses of O 3 and its precursors at Summit show that the mean enhancements for PAN, NO x , NO y and C 2 H 6 during the BB events identified by FLEXPART are highly variable (Table 3) for ∆NO y and −22.9 to 338.0 pmol mol −1 for ∆C 2 H 6 , for the 11 BB events characterized as having low/medium anthropogenic signatures. When considering only those events with low anthropogenic signatures, the results show that air masses sampled with potential BB contributions have ∆O 3 ranging between −4.4 and 10.8 nmol mol −1 , 20 with positive enhancements during 4 out of 5 events. These enhancements are comparable to those by Thomas et al. (2013) who estimated ozone production of up to 3 nmol mol −1 in aged BB plumes in the mid to upper-troposphere (peaking at 7 km) over Greenland. Our results suggest that in the lower troposphere the enhancement may even be greater. However, the enhancement values presented here can only be 25 considered as best estimates based on the FLEXPART transport model simulations.
Additionally, the long transport times from source region to the measurement site suggest significantly aged BB plumes, with mean weighted plume ages for the events ranging between 9-18 days (median 14 days). These aged plumes will be well mixed Introduction with the background air, therefore, separating the pollution impacts from background levels is challenging. For example, in 2008 FLEXPART indicated that a BB event impacted the measurement site from 25-26 July (event 1). The total column sensitivity from the FLEXPART retroplume (see Fig. 10) shows that the air masses arriving at the site during this event in July originated from a region with BB sources over Canada and 5 Alaska and was transported in the lower troposphere over the Arctic Ocean. The air resided over the Arctic Ocean for ∼ 4 days before ascending to Summit; therefore, it is likely that the polluted air containing BB emissions mixed with cleaner air from this region. As a result, measurements during this event show small positive enhancements for NO x and C 2 H 6 , and negative values for ∆O 3 , ∆PAN and ∆NO y .

10
The largest BB event identified by FLEXPART was observed in August 2008, when the BC fire tracer indicated BB plumes impacted the site continuously from 3 August to 14 August, peaking at ∼ 91 pmol mol −1 , as a result of large wildfires in Canada. We find O 3 and its precursors were all positive during this period, with mean enhancements of 56.5, 19.4 and 141.1 pmol mol −1 for ∆PAN, ∆NO x and ∆NO y respectively, 15 and 10.5 nmol mol −1 for ∆O 3 . A closer analysis of the measurements in Fig. 11 shows that O 3 was consistently high during the event. Analysis of the FLEXPART total column sensitivity indicates that during this event air masses were often transported at high altitudes in the free troposphere, enhancing the probability of mixing with high ozone from stratospheric origin, which may contribute to the elevated O 3 levels that were observed 20 (Alvarado et al., 2010;Roiger et al., 2011). Anthropogenic events during the summer months (April-September) were identified using the same threshold for winter/spring as in Sect. 3.2.1. During the measurement period, 28 events were identified with mean enhancement values up to 16 nmol mol −1 , 28 pmol mol −1 , 237 pmol mol −1 , 205 pmol mol −1 , and 1.0 nmol mol −1 for ∆O 3 , ∆PAN, 25 ∆NO x , ∆NO y and ∆C 2 H 6 , respectively. The air masses of anthropogenic origin primary originated from Europe with a mean plume age ranging between 7 to 15 days. The maximum enhancements during anthropogenic events are much larger than those from Introduction the BB events, suggesting that air masses containing anthropogenic emissions may have a larger impact on levels of O 3 and precursors at Summit during the summer.

Summary
These anlayses of NO y , NO x , PAN, NMHC and O 3 from the high altitude GEOSummit Station in Greenland show that PAN is the dominant species of NO y at the site, year 5 round, ranging from 49 % in the summer months to 78 % in spring. However, the NO y seasonal cycle does not follow that of PAN, due to significant contributions from NO x in the summer and odd NO y species during both summer and winter. We hypothesize that alkyl nitrates may account for a large portion of the odd NO y observed in winter and that photochemically produced species such as NO x and HONO within the snowpack 10 may impact the NO y budget during summer. However, these hypotheses cannot be confirmed without coincident measurements of indivdual NO y species and alkly nitrates above the snowpack.
Analyses of the C 2 -C 6 alkane NMHC data show that there is a large build up of NMHC during the winter period in the atmosphere above Summit which peaks between 15 January and March. The increase in photochemical processing after polar sunrise coincides with a decrease in NMHC levels during subsequent months. Between February and May total C 2 -C 6 NMHC decreased by approximately 4.4 ppbC. The decrease in NMHC may contribute partly to the spring time peak in O 3 observed over the same period. Further analyses using a photochemical model, constrained by the measure-20 ments, is needed to evaluate the springtime O 3 photochemical production rate at the measurement site and during subsequent transport downwind (Hamlin and Honrath, 2002).
Rapid changes in the origin of sampled air masses, from regions in Europe, North America and the high latitude Arctic, result in a large variability in the O 3 levels at Summit are typically higher than those observed at lower elevation Arctic sites due to a stronger influence of transport from the stratosphere and a reduction in 10 ozone depletion events from halogens (Helmig et al., 2007a, b;Hirdman et al., 2010). Short periods with reduced ozone are observed throughout the year. During the summer months, these low ozone events tend to occur when the sampled air masses contain low levels of pollution. In contrast, during the winter, low ozone coincided with the occurrence of polluted air masses that have been transported in the lower troposphere 15 to the site, possibly due to the occurrence of O 3 titration and reduced mixing with high background O 3 .
FLEXPART tracer simulations indicated that biomass burning emissions transported to Summit during the summer in 2008-2010 primarily originated from North America. Plumes originating from BB events in Europe were only present during a short 20 period in March 2009. The analyses focused on 11 BB events during the measurement period which did not have large anthropogenic signatures. During these events O 3 and precursor levels were typically enhanced within the BB plumes with ∆O 3 levels up to 12.4 nmol mol −1 and ∆PAN, ∆NO y and ∆C 2 H 6 levels enhanced by up to 90.4, 141.1 pmol mol −1 and 338 pmol mol −1 , respectively. However, we cannot say with confi-25 dence here whether the enhanced levels observed were directly as a result of biomass burning emissions or whether they occurred as a result of the plumes mixing with background air at high altitudes. In fact, it was found during the summer months that enhancements in all the measured species were greater when sampling air masses from anthropogenic origin rather than BB plumes. High NO y levels observed above the background during the events discussed here may have an impact on snow photochemistry and the subsequent release of NO x , due to the uptake of NO y species such as HNO 3 and HONO to the snow pack (Grannas et al., 2007, and references therein). Due to the stability of the Arctic free troposphere the region is an effective reservoir for O 3 pre-5 cursors. Therefore, the high O 3 precursor mole fractions above background levels in the summer may have important implications for NO x and O 3 in the mid-latitudes during southerly flow of air masses (Hamlin and Honrath, 2002). However, we find there is a need for future studies to constrain the speciation of NO y above the snowpack, through year-round coincident measurements of NO x , PAN, NO y HONO and HNO 3 10 determine the sources of odd NO y in the winter and summer. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Arctic using statistical analysis of measurement data and particle dispersion model output, Atmos. Chem. Phys., 10, 669-693, doi:10.5194/acp-10-669-2010, 2010. 13836, 13843 Honrath, R. E. and The seasonal cycle of nitrogen oxides in the Arctic troposphere at Barrow, Alaska, J. Geophys. Res., 97, 20615, doi:10.1029/92JD02081, 1992. 13821 Honrath, R., Hamlin, A., andMerrill, J.: Transport of ozone precursors from the Arctic tropo-5 sphere to the North Atlantic region, J. Geophys. Res., 101, 29335-29351, 1996. 13822, 13833 Honrath, R., Peterson, M., Guo, S., Dibb, J. E., Shepson, P. B., and Campbell, B.: Evidence of NO x production within or upon ice particles in the Greenland snowpack, Geophys. Res. Lett., 26, 695-698, 1999. 13830