Interactive comment on “ Changes to the chemical state of the northern hemisphere atmosphere during the second half of the twentieth century ”

General comments. This manuscript is an ambitious work that derives alkyl nitrate trends from firn air measurements and combines them with similarly-derived (already published) parent alkane data to characterize possible scenarios for large changes in the chemical state (especially the [NO]/[HO2] ratio) of the northern hemisphere during the 1970s to late 1990s. The manuscript explores the possibility that the data could indicate significant changes in NOx sources and sinks (driven by trends in sulfate aerosol precursors). Other potential explanations (such as changes in OH and transport) are also considered. The manuscript makes some very wide-ranging/large scale conjectures using this relatively limited new data. However, I consider that the authors are careful to back up their suggestions where possible and provide appropriate caveats


Introduction
The NO x (NO + NO 2 ) and HO x (OH + HO 2 ) budgets of the troposphere act to control the concentrations of oxidants such as OH, ozone and NO 3 (Fig. 1).These in turn control removal of pollutants from the atmosphere.Emissions of NO x in the Northern Hemisphere are mainly anthropogenic, with roughly equal proportions from power generation and transport (Olivier and Berdowski et al., 2001;Olivier et al., 2001).NO x and HO x are linked through ozone production, which is positively correlated with NO x concentrations in the background atmosphere through the photolysis of NO 2 (Reactions R1-R2).The photolysis of ozone in the presence of water vapour then leads to the production of OH (Reactions R3-R4).Other processes, such as alkene ozonolysis (Johnson and Marston, 2008) and photolysis of HONO (formed from heterogeneous reactions of NO 2 , Stone et al., 2012) may also be important primary sources of HO x , particularly in winter (e.g.Heard et al., 2004).
Removal of NO x from the atmosphere is controlled by the reaction of NO 2 with OH during the daytime (Reaction R5).This forms nitric acid, HNO 3 , which is lost from the atmosphere by wet deposition.At night and during the winter, the heterogeneous reaction of the NO x reservoir species N 2 O 5 , formed from the reaction of NO 2 with NO 3 (Reactions R6-R7), with H 2 O on aerosol becomes an important NO x sink (Reaction R8).OH and HO 2 rapidly interconvert through the reactions of OH with CO and hydrocarbons, such as alkanes, and the reaction of HO 2 with NO (Reaction R9).The reaction of NO with peroxy radicals (HO 2 and RO 2 -Reactions R9-R10) recycles the NO back to NO 2 .The main removal process for HO x in urban regions is the reaction of OH with NO 2 (Reaction R5) (Stone et al., 2012), while HO 2 selfreaction and reaction with RO 2 (in particular CH 3 O 2 ) (Reactions R11-R12) dominate in low NO x environments (Stone et al., 2012).HO However, changes to the atmospheric concentrations of both HO x and NO x during the previous century are poorly constrained.This is because all HO x and NO x species are short-lived, present at low concentrations (0.01-10 ppt), and have a high spatial and temporal variability (e.g.Stone et al., 2012).This makes them difficult to measure and trends are difficult to identify (based on spatially and temporally variable data sets).Furthermore, a range of state-of-the-art atmospheric chemistry transport models give no consensus of even the sign of OH change between 1850 and 2000 (Naik et al., 2013).However, the models do agree that between 1980 and 2000 there was an increase in Northern Hemisphere OH concentrations, with the best estimate of the increase being 4.6 (±1.9) %.This modelled increase is driven by increases in the NO x burden and in the water vapour concentration.
To attempt to study historical trends in HO x and NO x we have examined trends in longer lived species which are affected by changes to HO x and NO x in the atmosphere.
In this paper we report long-term atmospheric trends of three alkyl nitrates (2-butyl nitrate, 2+3-pentyl nitrate, 3methyl-2-butyl nitrate) derived from Arctic firn air.These are chemically produced in the atmosphere from the oxidation of alkanes and subsequent reaction of the peroxy radical formed with NO.The alkyl nitrate records are combined with previously reported trends of their parent alkanes from the same Arctic firn site.These records provide a proxy from which we can learn about the chemical state of the atmosphere at the time they were formed.

Alkanes
Emissions of butanes (C 4 H 10 ) and pentanes (C 5 H 12 ) to the atmosphere are almost entirely anthropogenic (> 98 % globally, Pozzer et al., 2010) and associated with fugitive emissions during oil and natural gas extraction and transmission, and evaporation and combustion of fossil fuels, such as in road vehicles (Pozzer et al., 2010;Pétron et al., 2012;Helmig et al., 2014a).Butane and pentane emissions from vehicles will be dependent on fuel composition, with evapo-rative emissions also dependent on temperature.Many areas in North America are part of "ozone attainment areas", and during summer months (June-15 September) they have been required by law since 1990 to provide a gasoline blend with a low Reid vapour pressure (RVP) to reduce the ozone production potential (www.epa.gov).This reduction in RVP is generally achieved by reducing the fuel's butane content relative to wintertime fuel (e.g.Gentner et al., 2009).
Measurements in firn air from Greenland (Aydin et al., 2011;Worton et al., 2012;Helmig et al., 2014b) suggest Northern Hemisphere C2-C5 alkane mixing ratios increased through the middle of the past century to a peak in ∼ 1980 (∼ 1970 for ethane) and then declined to roughly 1960 levels by 2000.In situ measurements from the urban areas of London (1993London ( -2008) ) (Dollard et al., 2007;von Schneidemesser et al., 2010) and Los Angeles (1960Angeles ( -2010) ) (Warneke et al., 2012) show steadily decreasing alkane mixing ratios, as do measurements at the semi-rural site of Hohenpeissenberg, Germany (von Schneidemesser et al., 2010).Emission estimates from the ACCMIP global emission inventory (Lamarque et al., 2010) (available at http://eccad.sedoo.fr)show butane and pentane emissions in Europe and North America increasing steadily between 1950 and 1980 before falling again to roughly 1965 levels by 2000.
The primary removal mechanism of alkanes from the atmosphere is reaction with the hydroxyl radical, OH (minor sinks include reaction with atomic chlorine, Cl, and the nitrate radical, NO 3 ).
Atmospheric mixing ratios of butanes and pentanes display a large seasonal cycle in mid-high latitudes (e.g.Swanson et al., 2003;Helmig et al., 2009) due to changes in their chemical lifetimes (∼ 1 month in the winter and 4-5 days in the summer) driven by the seasonal cycle in OH concentration.

Alkyl nitrates
Alkyl nitrates (RONO 2 ) are secondary oxidation products of alkanes (RH).Their atmospheric lifetimes are of the order of months in winter and 10 days in summer (Clemitshaw et al., 1997).Consequently, they display a strong seasonal cycle in the Arctic, with peaks in the late winter/early spring and minima in the summer (Swanson et al., 2003), similar to the alkanes.Alkyl nitrates are formed when alkanes react with OH to form a peroxy radical, RO 2 (Reaction R13), which subsequently reacts with NO to form an alkyl nitrate (Reaction 14b) (e.g.Talukdar et al., 1997).This is a minor channel of the RO 2 + NO reaction (Reaction R14a) which generally leads to ozone production via recycling of NO to NO 2 and the subsequent photolysis of NO 2 .RO 2 can also react with HO 2 (the hydroperoxyl radical) (Reaction R11) to form a peroxide (ROOH).The probability of RO 2 reacting with NO (leading to alkyl nitrate production) is thus governed by the ratio ).The 2008 EU hole was sampled using the firn air system of the University of Bern (Schwander et al., 1993), and the US hole, sampled using the US firn air system (Battle et al., 1996)

Firn analysis
The firn air samples from the EU hole at NEEM were analysed for alkyl nitrates at the University of East Anglia (UEA) using a gas chromatography mass spectrometer (GC-MS) in negative ion chemical ionisation mode (GC-NICI-MS) (e.g.Worton et al., 2008).2-pentyl nitrate and 3-pentyl nitrate are presented together as 2+3-pentyl nitrate because the two are not baseline separated in the chromatogram.
The NEEM samples were analysed using the UEA calibration scale.This was converted to the NCAR scale (against which the North GRIP 2-butyl nitrate and 2+3-pentyl nitrate samples are calibrated) for direct comparison with the North GRIP atmospheric histories from Worton et al. (2012) and with the in situ measurements at Summit Station by the University of California, Irvine (UCI) (Swanson et al., 2003;Dibb et al., 2007) as the North GRIP measurements were made on the UEA scale.
Firn air samples from the EU hole at NEEM were analysed for alkanes at the Max Planck Institute Laboratory (MPI) by gas chromatography with flame ionisation detection (GC-FID) (see Baker et al., 2010 for further details).At the Institute of Arctic and Alpine Research (INSTAAR) firn air samples were analysed from both the EU and US holes at NEEM for alkanes by GC-FID (see Pollmann et al., 2008 andHelmig et al., 2014b for further details).

Firn modelling
The air sampled from any given depth in the firn column is representative of a range of ages because of the interconnected nature of the firn.Firn models can be used to derive the atmospheric history of a gas from measurements of air trapped in the firn.The extent and rate at which the gas diffuses through the firn depends on the diffusivity profile of the firn, the diffusivity coefficient of the gas and on the gravitational fractionation (caused by the molecular weight) of the gas.The diffusion profile is different for every firn site.
To determine the atmospheric history of a gas from firn air measurements, the firn diffusion profiles must first be constrained.This is done using a series of reference gases with well-known atmospheric histories.At NEEM the reference gases used were CO 2 , CH 4 , SF 6 , HFC-134a, CFC-11, CFC-12, CFC-113, and CH 3 CCl 3 , as well as 14 CO 2 (Witrant et al., 2012).
Each gas also has a different diffusion rate through the firn based on its molecular structure; this is called the diffusion coefficient.The diffusion coefficient is calculated relative to a reference gas, generally CO 2 .Different methods have been reported for the calculation of these diffusion coefficients (e.g.Chen and Othmer, 1962;Fuller et al., 1966).The diffusion coefficients of the alkyl nitrates were calculated using the method of Fuller et al. (1966) based on the sum of the Le Bas molar volumes of the molecule.Model runs were also performed using diffusion coefficients for the alkyl nitrates calculated using the Chen and Othmer method.These coefficients are ∼ 10 % lower than those calculated using the Fuller method.However, the atmospheric scenarios derived from the modelling are very similar, well within the 2σ uncertainty envelopes presented in Fig. 2. The diffusion coefficients used for the firn modelling for each molecule within this work are given in Table 1.
The inverse model used for the atmospheric history reconstructions was the most recent version of the LGGE-GIPSA atmospheric trend reconstruction model described in Witrant and Martinerie (2013).
The atmospheric mole fraction derived from the firn reconstructions represents an annual mean.The alkanes and alkyl nitrates examined in this work have a strong atmospheric seasonality due to changes in their chemical lifetimes driven by seasonal variability in OH concentration in the air Table 1.Diffusion coefficients used in the firn modelling, calculated from Le Bas molecular volumes using the method of Fuller et al. (1966).

Compound
Diffusion coefficient relative to CO 2 2-butyl nitrate 0.467 2+3-pentyl nitrates 0.428 3-methyl-2-butyl nitrate 0.428 masses in which they are transported to the Arctic.Thus the seasonal cycle of both species follows a roughly sinusoidal curve with a peak in the late winter (March) and a trough in midsummer (July-August) (Swanson et al., 2003).Consequently, changes to the firn-derived mole fractions are likely to be dominated by changes to wintertime atmospheric concentrations.This model cannot take into account the seasonality in the signal that is preserved in the upper part of a firn profile.Therefore, measurements above a certain depth must be excluded from the model input.It is noted that the latter part (post-1995) of the model-derived scenarios for 2+3-pentyl nitrate is rather sensitive to the inclusion or exclusion of the measurement at 34.72 m (the shallowest measurement used).The scenarios presented in this work are based on including this measurement.

Alkyl nitrate trends
Atmospheric histories of the three alkyl nitrates 2-butyl nitrate, 2+3-pentyl nitrate, and 3-methyl-2-butyl nitrate (formed from n-butane, n-pentane, and iso-pentane respectively) derived from firn air measurements at NEEM are shown in Fig. 2. The records of all three alkyl nitrates show similar features (as would be expected from the similar sources and sinks).All show a steep increase in mixing ratio from the 1970s to the 1990s with increases of a factor of 3-4.The peak in the 1990s is followed by a steep decline to the sampling date of 2008.
Figure 2 also shows the atmospheric histories of the same three alkyl nitrates derived from firn air from North GRIP, Greenland, up to 2001, presented in Worton et al. (2012) (pink shaded area).There is very good agreement between the derived trends at the two sites.Differences can be attributed to the limited number of measurements at both sites, possible drift in the calibration standard used, and uncertainties in the firn modelling.Both sites show the same large increase in mixing ratios from the 1970s to the 1990s.Importantly, the NEEM records show that the turnover and subsequent decline in mixing ratios, the beginnings of which were evident in the North GRIP records, appears to continue through the 2000s.However, as noted in Sect.2.3, the derived atmospheric history of 2+3-pentyl nitrate is sensitive to the inclusion of the measurement at 34.72 m.A scenario that did not include this measurement was almost flat from 1995 to 2008 rather than declining as in Fig. 2.
There are very limited in situ measurements of alkyl nitrates in the Arctic and even fewer that cover a whole seasonal cycle.Swanson et al. (2003) reported the seasonal cycle of 2-butyl nitrate at Summit Station, Greenland (72.34 • N, 38.29 • W, 3250 m a.s.l), from June 1997 to June 1998 based on samples collected roughly every 2 days.Dibb et al. (2007) reported monthly mean measurements of 2-butyl nitrate for the period June 2000 through to August 2002 based on samples taken roughly weekly also from Summit Station.In order to compare these in situ measurements to output derived from the firn measurements the annual mean is taken.This is because the firn smoothes out the seasonality and represents the annual mean of mixing ratios.Calculating the 2-butyl nitrate annual mean for the three periods 1997-1998 (6.8 ppt), 2000-2001 (5.3 ppt), and 2001-2002 (5.0 ppt) gives values that can be compared to the output from the firn model for 2-butyl nitrate.These agree with the firn model output in terms of absolute mixing ratios of 2-butyl nitrate during this period (5-8 ppt) (Fig. 2).They also show a declining trend through the period, in agreement with the firn model output, though this is not statistically significant within the uncertainties.
Considering Reactions (R13)-(R16), the trends in the alkyl nitrate mixing ratios (Fig. 2) could be caused by the following: i. changes to the atmospheric mixing ratios of the parent alkanes; ii. changes to [OH]t, where t is time since emission of the alkane.i.e. the amount of photochemical processing that the air mass in which the alkyl nitrates are being formed undergoes before reaching the Arctic; iii. changes to the production efficiency of the alkyl nitrates, i.e. whether the peroxy radical reacts with NO (Reaction R14) or with HO 2 (Reaction R11); iv.changes to the alkyl nitrate sinks, i.e. changes in [OH] or radiation.
Concerning point (i), the peak in alkyl nitrate mixing ratios in the 1990s is not contemporaneous with that of the parent alkanes (∼ 1980 -Fig. 3).This suggests that the changes to the alkyl nitrate mixing ratios are not being primarily driven by changes to the parent alkane.By considering the ratio of the alkyl nitrate to its parent hydrocarbon, using the firnderived alkane trends from NEEM presented in Helmig et al. (2014b), we can effectively remove the effect of changes to the parent hydrocarbon from the alkyl nitrate signal.This is done in Sect. 4. Concerning point (iv), there is evidence for global dimming (i.e. a decrease in surface solar radiation) of about 5 % between 1960 and 1990 in the Northern Hemisphere.However, this began to turn around during the mid-1980s and there was a brightening trend between 1985and 2000(Wild et al., 2005)).This minor change to the alkyl nitrate sink is unlikely to have had a noticeable effect on mixing ratios.
Points (ii), (iii), and (iv) are discussed further in the following sections.
4 Ratios of alkyl nitrate to parent alkane Bertman et al. (1995) presented a mathematical equation to describe the production of alkyl nitrates in a NO x -rich environment (Eq. 1 assumes an initial zero mixing ratio for alkyl nitrates).
We extend Eq. ( 2) to include the possibility of alkyl nitrate production at less than 100 % efficiency, in a non-NO xrich environment, i.e. that the peroxy radical, RO 2 , formed may react with something other than NO.This is achieved by the inclusion of the term k 14 [NO] / (k 14 [NO]+other RO 2 sinks).In high-NO x environments, this value is ∼ =1.However, in lower NO x environments, other sinks for the peroxy radical, RO 2 , will compete with NO.In reality the term k 11 [HO 2 ] is likely to dominate the "other RO 2 sinks" term in a background environment.The only other species likely to be present at high enough concentrations to compete with the RO 2 + HO 2 reaction is the methyl peroxy radical (CH 3 O 2 ), which may be present at similar concentrations to HO 2 , but the reaction rate of CH 3 O 2 with other alkyl peroxy radicals larger than CH 3 O 2 is ≤ 2 × 10 −13 cm −3 s −1 (IUPAC), 2 orders of magnitude slower than the reaction with HO 2 (IU-PAC).Hence in Eq. ( 3) we extend Eq. ( 2) by including the term k 14 [NO] / (k 14 [NO]+k 11 [HO 2 ]).
For the purposes of our calculations, k 14 [NO] / (k 14 [NO]+k 11 [HO 2 ]) is assumed (in the same way as [OH]) to represent an integrated value for this ratio during transport of the air mass from the source region to the Arctic, i.e. 1/t • k 14 [NO] / (k 14 [NO]+k 11 [HO 2 ]) dt.We denote this term γ (Eq.4). in the firn are considerably larger than this, of the order of a factor of 3-5.Hence we consider the assumption of γ as a constant to be a reasonable assumption for the sake of making the problem tractable and that the changes to γ that we calculate in the paper are not an artefact of this assumption.Since γ is treated as a constant, integration of Eq. ( 3) gives an equation the same as Eq. ( 1) from Bertman et al. (1995) except with the addition of the term γ (Eq.5).
Atmospheric histories of the three parent alkanes of the alkyl nitrates presented in Fig. 2 were presented in Helmig et al. (2014b) -Fig.7 (n-butane, n-pentane, iso-pentane).These are used here, in conjunction with the alkyl nitrate histories in Fig. 2, to determine trends of the ratio [RONO 2 ] / [RH] for each alkyl nitrate-alkane pair.By rearranging Eq. ( 5), we can then probe two of the possible causes for the observed alkyl nitrate trends.Firstly, that the mean production efficiency (i.e.γ ) has changed over the time period of the firn record.Secondly, that the processing of the air mass, i.e. mean OH concentration, [OH], multiplied by the transport time from source regions to the Arctic, t, has changed.

Changes to the production efficiency of the alkyl nitrates
In an urban environment, daytime [NO] can range from 10 to a few hundred ppb.In this case, the production efficiency of the alkyl nitrates ≈ 1, i.e. all of the alkyl peroxy radicals formed in Reaction (R13) go on to form alkyl nitrates at a yield determined by the branching ratio α 13 .However, in rural and more remote regions of the atmosphere, daytime [NO] ranges from 1 to 100 ppt.At these mixing ratios k 14 [NO] / (k 14 [NO]+k 11 [HO 2 ]) would be expected to vary between around 0.3 and 1, assuming a daytime [HO 2 ] = 2 × 10 7 molecules cm −3 (wintertime -the alkyl nitrate and alkane signals in the firn are dominated by wintertime concentrations).Changes to [NO] or [HO 2 ] in these remote environments will affect the production efficiency of the alkyl nitrates.Since the term γ is an average across the whole transport time it reflects both the urban and remote environments.Equation ( 6) is a rearrangement of Eq. ( 5) from which historic changes to γ can be calculated using the measured changes to the [RONO 2 ] / [RH] ratio (assuming that the photochemical processing, [OH]t, has remained constant through time).All rate constants and branching ratios used in the calculations are taken from MCMv3.3.1 (mcm.leeds.ac. uk;Jenkin et al., 1997) (see Table S2, Supplement) assuming a temperature of 273 K.The magnitude of the diurnal photolysis sink of the alkyl nitrates, j 16 , will vary with emission region and during transport.j 16 is included in the term λ, which represents the ratio j 16 /k 15 [OH].λ is assumed to be 1 in Fig. 4, i.e. j 16 = k 15 [OH].The sensitivity to the magnitude of this sink is discussed further in the Supplement.
Figure 4 shows the historical trend in mean alkyl nitrate production efficiency, γ , relative to 1970, calculated using Eq. ( 6) if [OH]t is assumed to have remained constant during this period.A value of 5 × 10 11 molecules cm −3 s is used for the constant [OH]t.This is based on a mean transport time of air masses from Europe (from where the majority of wintertime pollutants are transported to the Arctic -see Sect.6.1) to the Arctic in the winter of 10 days (Stohl, 2006), and a mean wintertime [OH] of ∼ 6×10 5 cm −3 (in reasonable agreement with that derived by Derwent et al. (2012) for the North Atlantic in wintertime).However, it is noted that the relative change in γ shown in Fig. 4 is independent of the value used for [OH]t.
The trend in the mean production efficiency of the alkyl nitrates, γ , relative to 1970 values, shows similar features to those of the alkyl nitrate trends.The ratio increases by a factor of between 2.5 (2-butyl nitrate) and 4 (3-methyl-2butyl nitrate) from the 1970s to the late 1990s and then subsequently declines.
This analysis suggests that the observed changes to the [RONO 2 ] / [RH] ratio in the firn could be explained by changes to γ .This is driven by changes to the [NO] / [HO 2 ] ratio experienced by air masses in transport to the Arctic.We now investigate whether trends in processes that could drive this ratio are consistent with this scenario, i.e. how NO x concentrations may have changed.The atmospheric NO x concentration is determined by the relative magnitudes of the sources and sinks.The main sources of NO x in the Northern Hemisphere are anthropogenic emissions from fossil fuel use, power stations and transport (Olivier and Berdowski et al., 2001;Olivier et al., 2001).
Figure 5 shows how NO x emissions from North America, Europe, and Russia have varied between 1970 and 2008, taken from the bottom-up estimates of the EDGAR database (EDGAR v4.2 (EC-JRC/PBL, 2011), http://edgar.jrc.ec.europa.eu).Emissions were fairly constant between 1970 and 1990 and then fell by about 25 % from 1990 to 2008.This is in good agreement with the NO x emission trends for OECD Europe presented in Vestreng et al. (2009).
Assuming that these bottom-up emissions estimates are correct in the timing of the NO x emissions changes, it seems unlikely that an increase in the alkyl nitrate production efficiency during the period 1970 to the late 1990s could have been driven primarily by changing NO x emissions.
A decline in the alkyl nitrate production efficiency after the late 1990s, on the other hand, may well have been driven by decreasing NO x emissions.Measurements at a range of UK sites showed a decrease in NO x concentrations from 1996 (the beginning of the reported measurements) to 2002-2004, of 1-3.5 % per year, depending on the site (Carslaw et al., 2011).During the same period, the fraction of the NO x that is NO 2 (f -NO 2 ) roughly doubled, suggesting that NO has decreased by more than NO x concentrations.Recent trends at many European sites show similar trends with small decreases in NO x between 1999 and 2007 (the period for which measurements are available) but level or increasing NO 2 over the same period (Carslaw et al., 2011;Gilge et al., 2010).
Declining NO x emissions have been used to explain these trends in measured concentrations.However, the decline in these ambient concentrations is not as large as would be expected using current emission inventories (Carslaw et al., 2011).

NO x sinks
At mid-high latitudes in the daytime during the summer, the main sink for NO x is the reaction of NO 2 with OH.This reaction produces nitric acid (HNO 3 ), much of which is then removed from the atmosphere by wet deposition.However, at night and during the winter months, when daily mean [OH] is more than an order of magnitude lower than during the summer (e.g.Derwent et al., 2012), the dominant NO x sink is the conversion of dinitrogen pentoxide (N 2 O 5 ) to HNO 3 .This only occurs when photolysis is low, allowing NO 3 (formed from the reaction of NO 2 with O 3 ) to build up.This NO 3 reacts with NO 2 to form N 2 O 5 .While the reaction of N 2 O 5 with H 2 O is slow in the gas phase (Tuazon et al., 1983), it occurs rapidly in aerosol.
A modelling study by Dentener and Crutzen (1993) predicted that changes to the loss of NO x via sulfate aerosol could have a significant effect on Northern Hemisphere NO x concentrations and that these changes would also affect O 3 and OH concentrations.Subsequent modelling studies, though often focussing on remaining uncertainties in the uptake coefficients of N 2 O 5 to aerosol, have broadly agreed with the magnitude of the NO x changes suggested by Dentener and Crutzen (Brown and Stutz, 2012).
There has been a large decrease in sulfate aerosol observed in Europe and the United States since 1980 (Berglen et al., 2007;Turnock et al., 2015).Figure 6 shows the measured trend in wintertime (DJF) sulfate mass concentration presented in Turnock et al. (2015), with decreases of about 75 % from 1979 to 2005.This decreasing trend has been driven by a ∼ 70 % decrease in SO 2 emissions (Smith et al., 2011) from these regions (Fig. 6).It is noted that while global SO 2 emissions have only decreased about 15 % from the peak in the 1970s due to rapidly increasing emissions in eastern Asia in recent decades, sulfate aerosol has a lifetime of about 5 days in the troposphere (and SO 2 of about 1 day) (Stevenson et al., 2003) and so aerosol concentrations will be largely driven by regional SO 2 emissions.
These large decreases in sulfate aerosol in Europe and the US (the main source regions for air masses arriving in the Arctic in the winter) may be expected to have led to a decrease in NO x removal by N 2 O 5 hydrolysis, and hence to an increase in the NO x lifetime and atmospheric [NO x ], either through long-term changes to the total sulfate aerosol burden (Turnock et al., 2015) or long-term changes to particle acidity driven by reductions in sulfate (e.g.Murphy et al., 2017).The time period of decreasing SO 2 emissions and sulfate aerosol is broadly in line with the derived steep increase in the alkyl nitrate production efficiency.2011), and mean European sulfate mass concentration (µg S m −3 ) in winter (DJF) from Turnock et al. (2015).SO 2 emissions: blue is Europe, green is North America (US and Canada), gold is the former Soviet Union (Russia, Ukraine, and others), pink is eastern Asia (China, Japan, South Korea, and others).Brown dashed line is the mean European sulfate mass concentration in winter (DJF).
However, work remains ongoing to determine the exact effect of a number of parameters (e.g.relative humidity, particulate organic / sulfate ratio, particle acidity) on the N 2 O 5 uptake coefficient, and thus the efficacy of the reaction, and the extent to which changing sulfate content and abundance of aerosol would be expected to affect the uptake coefficient and thus oxidant concentrations (Brown and Stutz, 2012).
6 Changes to photochemical oxidation An alternative explanation for the observed alkyl nitrate trends is that the amount of photochemical processing of the air mass changed during the period of study.An increase in processing could be caused by a change in either the concentration of the OH radical (assuming photolysis to have remained constant) or by an increase in the transport time of the air mass from the source region to the Arctic.Equation ( 7) is a rearrangement of Eq. ( 6) from which historic changes to the photochemical processing, [OH]t, can be calculated using the measured changes to the [RONO 2 ] / [RH] ratio, assuming that γ has remained constant. [OH]t Figure 7 shows the trends in [OH]t derived from the alkyl nitrate-alkane pairs if a constant value for γ is assumed.The value used for the constant γ for each alkyl nitrate was the mean value derived in Fig. 4 for the period 1960-2007 (0.31 for 2-butyl nitrate, 0.34 for 2+3-pentyl nitrate, 0.17 for 3methyl-2-butyl nitrate).Equation ( 7) also has [OH] terms on the right-hand side of the equation, incorporated in λ.The results in Fig. 7 are determined through an iterative process of fitting a polynomial to the trend calculated using an a priori assumption that λ = 1 for the whole time period.The calculation is then repeated with a temporally varying value for λ using this fit to determine the changes.This process converges towards the unique solution presented in Fig. 7.For example, in Fig. 7 for 2-butyl nitrate, in 1970 the assumed value of λ is 1, at the peak of [OH]t in 1997, when [OH]t = 2.1, the value of λ is 0.46 (1/2.1).
Figure 7 shows that the observed [RONO 2 ] / [RH] ratios between around 1970 and the late 1990s could be explained by a relative change in [OH]t of a factor of between 2.1 (2butyl nitrate) and 2.6 (3-methyl-2-butyl nitrate).The sensitivity of these calculated values to the assumed value for γ and for j 16 in 1970 is discussed in the Supplement.

Air mass transport time to the arctic
The transport time, t, of pollutants to the Arctic from the source is dependent on (i) the atmospheric transport patterns, and (ii) the source regions of the pollutants.
Concerning (i), Kahl et al. (1999) have suggested that there is a decadal-scale (4-14 years) variability in transport patterns of pollutants from the NH to the Arctic but note no long-term trend.Hirdman et al. (2010) note that while changes to transport patterns can explain much of the interannual variability of Arctic concentrations of black carbon and sulfate aerosol (pollutants with similar source regions to the alkanes), they played only a minor role in long-term changes.Eckhardt et al. (2003) have shown that transport of pollutants to the Arctic from European and US source regions is more rapid during positive phases of the weather pattern, the North Atlantic Oscillation (NAO).During the period www.atmos-chem-phys.net/17/8269/2017/Atmos.Chem.Phys., 17, 8269-8283, 2017Phys., 17, 8269-8283, 1960Phys., 17, 8269-8283, -1980 the NAO was predominantly in a negative phase in winter, between 1980 and 2000 it was predominantly in a positive phase, and since 2000 neither phase has been dominant (Hurrell and Deser, 2010).This suggests that there was more rapid transport of pollutants to the Arctic during the period 1980-2000 compared to the preceding and succeeding periods.This would mean a shorter processing time for the air masses and hence less alkyl nitrate production and lower alkyl nitrate to alkane ratios.This is the opposite to what we observe in the firn records, suggesting that changes to the NAO are unlikely to be responsible for the observed alkyl nitrate trends.Concerning (ii), changes to the relative distribution of the major source regions of the alkanes could have occurred for a number of reasons.Fuel composition has changed through time as a response to technological development of vehicles.Clean air legislation has led to the development of cars with progressively lower evaporative and tailpipe emissions (e.g.Wallington et al., 2006), through developments such as catalytic converters.In addition emissions may have changed simply due to a change in vehicle usage.If such changes were to have occurred in more northerly regions significantly earlier than in more southerly regions, this could have increased the mean transport time of air masses to the Arctic.
For many areas in North America, the Reid vapour pressure of fuel is regulated in the summer season (1 June-15 September) (epa.gov.uk),leading to sale of a different fuel mix in summer compared to winter.This is generally achieved by producers reducing the butane content of the fuel (Gentner et al., 2006).This legislation was introduced in 1990.However, the observed alkane and alkyl nitrate signals in Greenland are almost entirely wintertime signals (e.g.Swanson et al., 2003), and so such seasonal variation in fuel composition would not be expected to affect the firn measurements.
The main source of anthropogenic emissions to the Arctic of gases with lifetimes of the order of a few weeks, particularly during the winter, has been identified as northern Eurasia (e.g.Shindell et al., 2008;Stohl et al., 2007;Klonecki et al., 2003).Emissions from Europe and North America have followed a similar declining trend in recent years (Lamarque et al., 2010;von Schneidemesser et al., 2010;Warneke et al., 2012); thus the relative contribution from each region is not expected to have changed dramatically.

The hydroxyl radical, OH
The alternative explanation for an increase in photochemical processing is an increase in the mean [OH] to which the air mass is exposed.This would represent a regional trend in [OH] that is representative of regions from and through which air masses are transported to the Arctic, and would relate primarily to the winter (since the alkyl nitrate and alkane signals in the firn are dominated by wintertime concentrations).It is noted that the increased chemical processing ob-served could also result from an increase in an oxidant other than OH, e.g.atomic chlorine, as suggested in Helmig et al. (2014b).
Studies using changes to atmospheric mixing ratios of methyl chloroform (CH 3 CCl 3 ) have suggested that global mean OH concentrations are "well buffered" (e.g.Montzka et al., 2011).Since the main sink of CH 3 CCl 3 in the atmosphere is reaction with OH, and the emission sources and other sinks are thought to be well constrained, the variation in its observed mixing ratios at a number of remote sites can be used to infer variations in global [OH].Global mean [OH] has been inferred in this way in a number of studies (Prinn et al., 1995(Prinn et al., , 2001(Prinn et al., , 2005;;Rigby et al., 2008;Montzka et al., 2011).The most recent of these (Montzka et al., 2011) reported little interannual variability in mean global atmospheric [OH], estimating roughly 5 % variation from the mean value during the period 1997-2008, but this does not cover the period of interest here (1970 to late 1990s).Similarly, Kai et al. (2011) inferred a low variability in global [OH] between 1998 and 2005 based on a relatively constant δ-D-CH 4 .Earlier work using methyl chloroform (Prinn et al., 2001) reported an increase in NH [OH] of roughly 40 % between 1979 and 1991 but this increase has been questioned in more recent work (e.g.Montzka et al., 2011).
However, there are a growing number of observational data sets of trace gases in the NH which show trends since 1980 that could be explained, at least in part, by changes to the concentration of the OH sink.E.g. decreasing Arctic alkane mixing ratios (Helmig et al., 2014b;Aydin et al., 2011), decreasing Arctic CO mixing ratios (Petrenko et al., 2013), increasing d 13 C of methane (Monteil et al., 2011;Sapart et al., 2013), and decreasing dC 16 O of Arctic CO (Wang et al., 2012).
A recent multi-model intercomparison exercise of seventeen global chemical transport models, showed agreement for a small increasing trend in global mean [OH] of 3.5 (±2.2) % between 1980 and 2000 and a slightly larger [OH] increase in the Northern Hemisphere of 4.6 (±1.9) % (Naik et al., 2013).Dalsøren et al. (2016) determined an increase in global mean [OH] of about 10 % between 1970 and 2006 from modelled increases of the methane lifetime.

Discussion
The alkyl nitrate trends presented herein suggest a profound change to the chemical state of the Northern Hemisphere mid-high latitude atmosphere in winter between the 1970s and the late 1990s and then again between the late 1990s and the mid-2000s.
A key species of the tropospheric chemistry cycle, tightly linked to NO x and HO x , is ozone (Fig. 1).Ozone mixing ratios increased at background sites across the NH during the second half of the twentieth century, roughly doubling since 1960 (Parrish et al., 2012).Ozone production is positively linked to [NO x ] (at low NO x concentrations such as the background atmosphere).Hence, an increase in the [NO] / [HO 2 ] ratio from around 1970 to the mid-1990s is consistent with long-term trends in ozone in the background atmosphere.
Furthermore, while the alkyl nitrate measurements represent changes to the wintertime atmosphere, the ozone trends are seen in both summer and winter.If these are being driven by increases to [NO x ] in the background atmosphere, then this suggests that the chemical changes to the atmosphere may have been present throughout the year and are not just a wintertime phenomenon.
This work also implies that there may have been a change in [OH].Indeed due to the connected nature of the chemistry of NO x , ozone and OH (Fig. 1) it seems unlikely, given the implied increases in NO x suggested here and the recorded increases in ozone (Parrish et al., 2012), that there was not a commensurate increase in OH production during this period.A major primary production route of OH is via photolysis of ozone (Eq.8 - Smith et al., 2006).
where P (OH) is primary production of OH, and f is the fraction of O( 1 D) that reacts with water vapour.Ozone has increased at background sites between 1960 and 2000 (Parrish et al., 2012) and water vapour has also increased slightly (Hartmann et al., 2013).It therefore seems that the primary production of OH in the background atmosphere from this source must have increased through the final decades of the past century.Another primary OH source is via ozonolysis of alkenes (Johnson and Marston, 2008).
A third source of OH that may be important is photolysis of HONO (e.g.Stone et al., 2012).There is still considerable uncertainty about the sources of HONO, with formation from heterogeneous conversion of NO 2 via a range of postulated processes appearing to dominate over the HO x / NO x recycling reaction OH + NO (e.g.Michoud et al., 2014).This again would be a primary source of OH which would be expected to correlate positively with NO x concentrations.The primary sink of OH in the background atmosphere, CO, has decreased by about 15 % since 1980 (Petrenko et al., 2013), with the secondary sink, CH 4 , having increased between 1980 and 2000 by about 15 %.
The global growth rate of methane in the atmosphere continually declined throughout the period of the 1970s to 2000, culminating in the "methane pause" between 1999and 2006(Dlugockencky et al., 2009)).A possible cause of this change in the growth rate of methane could be an increase in OH concentration (e.g.Dalsøren et al., 2016).This is consistent with an increase in [OH] also being the cause of the trend in the ratio of alkyl nitrate to parent alkane seen in this work.It should be noted, though, that the majority of the OH oxidation of methane occurs in the tropics (e.g.Bloss et al., 2005), while any increase in OH suggested by the work herein must be viewed as representative only of the mid-high latitude Northern Hemisphere and the wintertime.However, the processes suggested herein, such as changes to the N 2 O 5 sink, have been shown to be effective at a hemispheric scale.
In a recent intermodel comparison project, ACCMIP, it was shown that models failed to capture the measured magnitude of the increase in ozone over recent decades (Parrish et al., 2014), in particular the steep increase seen between 1980 and 2000.This failure to capture measured changes to ozone may suggest that models are likely to underestimate changes to OH production, from ozone photolysis or reactions of ozone with alkenes, over the same period (i.e. they may be larger than the 4.6 (±1.9) % reported in Naik et al., 2013 for the NH).
Including alkyl nitrate chemistry and using the alkyl nitrate measurements presented herein could provide a valuable constraint for global chemical transport models modelling changes to NO x and HO x back to the middle of the twentieth century.

Conclusions
Time series such as those presented here are fundamental to improving our understanding of trends in atmospheric composition during the twentieth century.The long-term trends of alkyl nitrates presented herein suggest major changes to the chemical state of the atmosphere during the past five decades.The observed large increase in the [RONO 2 ] / [RH] ratio between the 1970s and late 1990s could be explained by a 2-4 fold increase in the mean production efficiency of the alkyl nitrates, driven by an increase in the [NO] / [HO 2 ] ratio in the background atmosphere.This is not consistent with reported changes to Northern Hemisphere NO x emissions, but may have been driven by a reduction in the NO x sink.The recent decreases (since the late 1990s) in alkyl nitrate concentrations are in qualitative agreement with recent decreases in NO x emissions and in measured NO x concentrations.Alternatively, the observed increase in the [RONO 2 ] / [RH] ratio between the 1970s and late 1990s could be explained by an increase in the amount of photochemical processing [OH]t of air masses reaching the Arctic by a factor of 2-3.This could be driven by an increase in concentrations of the hydroxyl radical (OH), or to the transport time (t) of air masses from source regions to the Arctic.If the observed trends are driven by changes to the chemical state of the atmosphere, then it is likely that they represent a combination of changes to the concentrations of both NO and OH.
Data availability.The firn measurements and the firn model output are available as part of the Supplement as an Excel spreadsheet.

Figure 1 .
Figure 1.Schematic of the NO x -O 3 -OH relationship in the background troposphere.

Figure 2 .
Figure 2. Concentration-depth profiles in the firn and the model-derived atmospheric histories.(b) The concentration-depth profiles measured in the firn (ppt): red open circles are measured mixing ratios (ppt) with error bars indicating the 2σ uncertainty; solid blue line are best fit of the firn model, dashed blue lines indicate the 2σ combined analytical and model uncertainties.(a) Atmospheric histories of the alkyl nitrates derived from the firn air measurements using the inverse modelling technique described within (solid blue lines).Dashed lines represent the 2σ confidence margins of the model calculations, combining the analytical and model uncertainties.Pink shaded area represents atmospheric histories presented in Worton et al. (2012) derived from firn air measurements at North GRIP, Greenland.Red open circles show the annual average of in situ measurements at Summit Station, Greenland (see text for details) with 1σ uncertainty.
γ = k 14 [NO] k 14 [NO] + k 11 [HO 2 ] = mean alkyl nitrate production efficiency (4) k 14 [NO] / (k 14 [NO]+k 11 [HO 2 ]) would not be expected to be constant in reality since [NO] is likely to change by orders of magnitude during transport, with values of the order of 2.5 × 10 11 cm −3 close to the emissions source, and falling to ∼ 1 × 10 8 cm −3 further from the source.Changes to the ratio k 14 [NO] / (k 14 [NO]+k 11 [HO 2 ]) at different times along the air mass trajectory will affect d[RONO 2 ]/dt at that time differently because d[RONO 2 ]/dt is also driven by [RH], which is a function of time.However, the assumption of k14[NO] = (k14[NO]+k11[HO2]) as an integrated value through time, i.e. γ , leads to uncertainties in the calculated [RONO 2 ] / [RH], at time t = 10 days, of the order of 5 % (see Supplement).The observed changes in [RONO 2 ] / [RH]

Figure 4 .
Figure 4.The trend in the mean alkyl nitrate production efficiency, γ , of the air masses in which the alkyl nitrates were formed, calculated using Equation E5 for each of three alkyl nitratealkane pairs, relative to 1970 values.This assumes that the amount of photochemical processing, [OH]t, remained constant at 5 × 10 11 molecules cm −3 s.All rate constants and branching ratios used in the calculations are taken from MCMv3.3.1 (mcm.leeds.ac.uk) assuming a temperature of 273 K.The mean diurnally averaged photolysis sink j 16 was assumed to be equal to the mean OH sink, k 15 [OH].

5. 1
Changes to atmospheric NO x concentrations 5.1.1NO x sources

Figure 6 .
Figure 6.SO 2 emissions (Tg yr −1 ) 1950-2005 from Smith et al. (2011), and mean European sulfate mass concentration (µg S m −3 ) in winter (DJF) from Turnock et al. (2015).SO 2 emissions: blue is Europe, green is North America (US and Canada), gold is the former Soviet Union (Russia, Ukraine, and others), pink is eastern Asia (China, Japan, South Korea, and others).Brown dashed line is the mean European sulfate mass concentration in winter (DJF).

Figure 7 .
Figure7.The trend in [OH]t calculated using Eq.(6) for each of three alkyl nitrate-alkane pairs assuming a constant value for γ .