Arctic tropospheric ozone: assessment of current knowledge and model performance

. As the third most important greenhouse gas (GHG) after carbon dioxide (CO 2 ) and methane (CH 4 ), tropospheric ozone (O 3 ) is also an air pollutant causing damage to human health and ecosystems. This study brings together recent research on observations and modeling of tropospheric O 3 in the Arctic, a rapidly warming and sensitive environment. At different locations in the Arctic, the observed surface O 3 seasonal cycles are quite different. Coastal Arctic locations, for example, have a minimum in the springtime due to O 3 depletion events resulting from surface bromine chemistry. In contrast, other Arctic locations have a maximum in the spring. The 12 state-of-the-art models used in this study lack the surface halogen chemistry needed to simulate coastal Arctic surface O 3 depletion in the springtime; however, the multi-model median (MMM) has accurate seasonal cycles at non-coastal Arctic locations. There is a large amount of variability among models, which has been previously reported, and we show that there continues to be no convergence among models or improved accuracy in simulating tropospheric O 3 and its precursor species. The MMM underestimates Arctic surface O 3 by 5 % to 15 % depending on the location. The vertical distribution of tropospheric O 3 is studied from recent ozonesonde measurements and the models. The models are highly variable, simulating free-tropospheric O 3 within a range of ± 50 % depending on the model and the altitude. The MMM performs best, within ± 8 % for most locations and seasons. However, nearly all models overestimate O 3 near the tropopause ( ∼ 300 hPa or ∼ 8 km), likely due to ongoing issues with underestimating the altitude of the tropopause and excessive downward transport of stratospheric O 3 at high latitudes. For example, the MMM is biased high by about 20 % at Eureka. Observed and simulated O 3 precursors (CO, NO x , and reservoir PAN) are evaluated throughout the troposphere. Models underestimate wintertime CO everywhere, likely due to a combination of underestimating CO emissions and possibly overestimating OH. Throughout the vertical proﬁle (compared to aircraft measurements), the MMM underestimates both CO and NO x but overestimates PAN. Perhaps as a result of competing deﬁciencies, the MMM O 3 matches the observed O 3 reasonably well. Our ﬁndings suggest that despite model updates over the last decade, model results are as highly variable as ever and have not increased in accuracy for representing Arctic tropospheric O 3 .


Introduction
Tropospheric ozone (O 3 ) is the third most important greenhouse gas (GHG) after CO 2 and methane (IPCC, 2021), and it is an air pollutant causing damage to human health (World Health Organization (WHO), 2021).It also causes damage to vegetation following dry deposition to the surface (U.S. EPA, 2013).However, our knowledge about the sources and sinks of tropospheric O 3 is still uncertain (AMAP, 2015(AMAP, , 2022;;Gaudel et al., 2018), in particular in regions where fewer observations exist and where our understanding of key processes is still evolving.The Arctic is one such region where few long-term measurements of O 3 exist and measurements of compounds that are important for producing and destroying O 3 in the atmosphere are scarce at the surface and even more so in the free troposphere.Progress has been made recently in terms of our understanding of certain processes, and a picture is emerging about the distribution of Arctic tropospheric O 3 as well as seasonal cycles and trends at different locations (e.g., Young et al., 2018;Tarasick et al., 2019b).In particular, the connection between surface O 3 depletion episodes and halogens is now well-established (e.g., Simpson et al., 2007;Abbatt et al., 2012).
However, the role of natural cycles in the Arctic O 3 budget relative to O 3 produced from anthropogenic emissions and how that relationship is changing in response to rapid warming in the Arctic are still uncertain.Arctic warming and associated development in the Arctic are also driving changes in local anthropogenic emissions, which could already be leading to changes in the relative contributions of O 3 produced due to long-range transport of midlatitude anthropogenic emissions and O 3 produced from within or near Arctic anthropogenic emissions.Increases in emissions, such as from shipping (Gong et al., 2018) or boreal fires, can affect Arctic air quality (Schmale et al., 2018).
Ozone radiative forcing resulting from changes in tropospheric O 3 in the Arctic is highly sensitive to altitude.The sensitivity of the Arctic O 3 vertical profile and resultant forcing from particular anthropogenic emission sources and regions vary substantially with altitude (Rap et al., 2015).Arctic surface O 3 may be most sensitive to European or local sources (Sand et al., 2015;AMAP, 2015AMAP, , 2022)), whereas emissions from North American and Asian sources are more important in the middle and upper troposphere (Monks et al., 2015;Wespes et al., 2012).Therefore, a combination of varied source sensitivities in the vertical profile and the increased efficacy of longwave O 3 forcing with altitude in the troposphere leads to a complex picture in terms of drivers of climate forcing by Arctic O 3 .The presence of temperature inversions in the Arctic lower troposphere may result in negative local forcing (Rap et al., 2015;Flanner et al., 2018), in particular for local sources such as shipping (Marelle et al., 2018).Hence, to improve the quantification of O 3 radiative effects in the Arctic there is a need first to assess model performance in terms of seasonal cycles and vertical distributions.The annual mean vertical distributions of O 3 and CO were examined in AMAP (2022) and Whaley et al. (2022) compared to the Tropospheric Emission Spectrometer (TES) and Measurement of Pollution in the Troposphere (MOPITT) satellite retrievals.Those studies showed good agreement between models and satellite measurements for O 3 in the free troposphere, where it is a strong GHG.
This paper assesses the current state of knowledge about the dynamics of Arctic tropospheric O 3 and the ability of a suite of current chemistry-transport and chemistry-climate models to simulate seasonal cycles of O 3 and selected precursors.We first review our current understanding of sources and sinks of Arctic tropospheric O 3 in Sect. 2. We summarize the models used in this study in Sect. 3 and the recent findings from satellite observations in Sect. 4. We then examine the extent to which our understanding of Arctic tropospheric O 3 can explain observed seasonal cycles at different surface sites in the Arctic and assess the ability of models to simulate observed distributions (Sect.5).We also examine vertical distributions of O 3 and its precursors and the extent to which models are able to capture observed seasonal variations (Sect.6).Finally, conclusions are presented in Sect.7. Trends in Arctic tropospheric O 3 over the last 20-30 years and possible changes in seasonal cycles will be presented in a follow-on study and compared to results from a subset of these models.

Arctic O 3 : sources and sinks
This section reviews tropospheric O 3 sources and sinks that are particularly relevant to the Arctic region, and many of these processes are shown in the schematic in Fig. 1.

Ozone sources
Tropospheric O 3 is a secondary air pollutant, which is not directly emitted but produced from the photochemical reactions of anthropogenic and natural precursor emissions of volatile organic compounds (VOCs), CO, and CH 4 in the presence of NO x .Besides significant anthropogenic sources of these O 3 precursors, there are also important natural sources for these species, such as boreal fires, lightning, vegetation, and transport of O 3 from the stratosphere (Fig. 1), which show marked seasonal and interannual variations.Away from the surface and in remote environments the tropospheric O 3 lifetime is around 20 d or more (Young et al., 2013), which facilitates the long-range transport of O 3 in the troposphere.Production of O 3 from lower-latitude emission sources and its subsequent transport to the Arctic constitute a substantial source of Arctic tropospheric O 3 (Hirdman et al., 2010), where the dry Arctic conditions and stably stratified atmosphere further prolong the O 3 lifetime.In addition, the stratosphere-troposphere exchange of O 3 makes a substantial contribution to the Arctic O 3 budget.The weak in situ O 3 formation in the Arctic relative to lower latitudes increases the relative importance of this exchange.Downward transport of O 3 from the stratosphere is an important source of O 3 in the Arctic troposphere and may be key in driving seasonality in Arctic tropospheric O 3 (Shapiro et al., 1987;Hess and Zbinden, 2013;Ancellet et al., 2016).Based on modeling, Liang et al. (2009) show that in spring (March and April), most of the O 3 in the Arctic upper troposphere originates from stratospheric injection (78 %) and that 20 %-25 % of surface O 3 originates from direct injection of O 3 or the injection of NO y and secondary O 3 formation.Analysis of observations by Tarasick et al. (2019a) is consistent with this picture.Global model simulations conducted as part of the Coupled Model Intercomparison Project Phase 6 suggest an increase in near-surface O 3 over the Arctic during the 21st century, driven by increased stratospheric O 3 import into the troposphere, particularly in winter (Zanis et al., 2022).In contrast, during summer, in situ production in the Arctic contributes a significant fraction, with a model study estimating a contribution of more than 50 % of O 3 in the Arctic boundary layer and 30 %-40 % in the free troposphere for the month of July (Walker et al., 2012).Methane (CH 4 ) is a key precursor for tropospheric O 3 via its oxidation in the presence of sufficient NO x .Increases in anthropogenic CH 4 emissions are estimated to be responsible for 44 ± 12 % of the global tropospheric ozone radiative forcing from the pre-industrial era to present day (Stevenson et al., 2013).Fiore et al. (2008) estimated that anthropogenic CH 4 emissions contribute 15 % to the annual average total global O 3 burden (including natural and anthropogenic sources  , 2015).Thawing permafrost and release from organic deposits in shallow Arctic Ocean waters in a warmer climate represent a new source of methane (Isaksen et al., 2014).Import of O 3 and its precursors from lower latitudes associated with episodes of long-range transport of anthropogenic or biomass burning pollution leads to enhancements in Arctic tropospheric O 3 (Wespes et al., 2012;Monks et al., 2015;Ancellet et al., 2016).Whilst very low levels of NO x within the Arctic, away from local sources, often limit local O 3 production, the release of NO x from thermal decomposition of peroxy-acetyl nitrate (PAN) (an important NO x reservoir) imported from lower latitudes can lead to in situ production of O 3 , particularly in the warmer Arctic summer lower troposphere (Wespes et al., 2012;Walker et al., 2012;Arnold et al., 2015).Investigation of long-range transport of O 3 precursors has shown efficient export of PAN from East Asia to the North Pacific, with relative contributions to longrange O 3 transport of 35 % in spring and 25 % in summer (Jiang et al., 2016).Ship observations over the Arctic Ocean and Bering Sea also identified events of long-range pollution transport with enhancements in O 3 (Kanaya et al., 2019).
Recently, there has been progress in improving knowledge of local O 3 precursor sources.Surface O 3 in summer is influenced by shipping NO x emissions along the northern Norwegian coast (Marelle et al., 2016;Marelle et al., 2018) and the Northwest Passage (Aliabadi et al., 2015).Additionally, Tuccella et al. (2017) showed that background O 3 is influenced by emissions downwind of oil and gas extraction platforms in the southern Norwegian Sea.Natural sources of Arctic tropospheric O 3 precursors include lightning NO x , emissions of NO x and reactive VOCs from the snowpack (Honrath et al., 1999;Guimbaud et al., 2002;Hornbrook et al., 2016;Pernov et al., 2021), natural emissions of VOCs from high-latitude vegetation (Holst et al., 2010;Ghirardo et al., 2020), and the sea surface microlayer (Mungall et al., 2017).Evidence from both observations and models suggests that boreal fires are also an important source of O 3 precursors and NO x reservoir species like PAN in spring and summer, with impacts on Arctic O 3 (Thomas et al., 2013;Arnold et al., 2015;Viatte et al., 2015;Ancellet et al., 2016).

Ozone sinks
Photochemical loss of O 3 is mainly via photolysis in the presence of water vapor or direct reaction of O 3 with hydroperoxyl (HO2) or hydroxyl radicals (OH).Photochem- ical destruction involving HO 2 may be particularly important in the Arctic where water vapor abundances are low (Arnold et al., 2015).Where local emission sources give rise to high NO x concentrations in urban regions or regions of shipping activity, O 3 loss via titration with NO can be dominant (Thorp et al., 2021;Raut et al., 2022).Dry deposition of O 3 and its precursors to ice and ocean surfaces is slower than to vegetated terrestrial surfaces (Fig. 1).Van Dam et al. (2016) reported O 3 dry deposition velocities that were 5 times higher over Arctic snow-free tundra in the summer months at Toolik Lake (northern Alaska) compared to the snow-covered ground.Dry deposition, combined with possible chemical loss (e.g., involving biogenic volatile organic compounds, BVOCs) producing lower O 3 concentrations during stable (lower light) night conditions may explain the different diurnal cycle observed at this tundra site compared to Arctic coastal locations.Interestingly, gradient studies at the NOAA Barrow Observatory near Utqiaġvik and at Zeppelin showed a positive gradient with height during O 3 depletion events (ODEs) and atmospheric mercury depletion events (AMDEs), suggesting that O 3 was removed at the surface due to fast photochemical reactions at or close to snow surfaces initiated by the release of halogen species (Skov et al., 2006;Solberg et al., 1996;Berg et al., 2003;Eneroth et al., 2007).During ODEs at Arctic sites in the Canadian archipelago (Alert, Resolute, and Eureka), vertical profiles show that ozone is typically uniformly depleted in the boundary layer, whereas a positive gradient is observed above the boundary layer (Tarasick and Bottenheim, 2002).
During Arctic spring, photochemical cycling of halogens in so-called "bromine explosion" events leads to rapid depletion of surface O 3 to low or near-zero concentrations (Barrie et al., 1988;Skov et al., 2004;Helmig et al., 2007;Simp-son et al., 2007).These phenomena are observed at Arctic coastal locations and in the Arctic Ocean (Bottenheim et al., 2009;Jacobi et al., 2010) in March-April and attributed to bromine (halogen) sources linked to Arctic sea ice, coupled with stable surface temperature inversions (e.g., Fig. 1; Hermann et al., 2019).Some model studies were able to explain major depletion events in simulations by introducing the wind-induced release of bromine from the snowpack and have shown that both blowing snow and the snowpack are important sources of bromine during the spring (e.g., Yang et al., 2010;Toyota et al., 2011;Yang et al., 2020;Huang et al., 2020;Swanson et al., 2022).Figure 2 shows the vertical extent of low O 3 episodes observed by lidar at Eureka in northern Canada.On 7 May, low O 3 concentrations were observed, and back trajectories showed that air masses came in from the ice-covered Arctic Ocean and had been in contact with the surface multiple times during the previous 6 d, whereas the concentrations were high on 9 May, when air came down from the mountains located to the south (Seabrook and Whiteway, 2016).Peterson et al. (2018) showed that active halogen chemistry and related O 3 depletion can also occur up to 200 km inland over snow-covered tundra in Alaska.Simpson et al. (2018) reported high levels of bromine oxide (BrO) at Utqiaġvik occurring earlier in February in air masses originating from the Arctic Ocean polar night.Their findings suggest a dark wintertime source of reactive bromine (halogens) that could feed halogen photochemistry at lower latitudes as the sun returns.This dark mechanism was also observed over sea ice in Antarctica by Nerentorp Mastromonaco et al. (2016).
In addition, whilst earlier studies proposed indirect evidence that O 3 and gaseous elemental mercury (Hg 0 ) are removed by reaction with Br atoms (e.g., Skov et al., 2004;Skov et al., 2020;Dastoor et al., 2008), Wang et al. (2019) showed, for the first time, a direct connection between O 3 and Hg 0 with atomic bromine (Br) during O 3 and Hg 0 depletion episodes at Utqiaġvik on the northern coast of Alaska (see Fig. 3) where O 3 and Hg 0 are removed in competing reactions with Br.Here, the Br/BrO ratio anti-correlates with O 3 concentrations, and box modeling confirms that O 3 is removed by Br.
This result is significant since the main source of halogens in the Arctic is the release from refreezing sea ice, blowing snow over sea ice, heterogeneous reactions of aerosol particles, and snowpack recycling (Petersen et al., 2016;Peterson et al., 2017, Wang et al., 2017;Yang et al., 2020).Burd et al. (2017) found a strong relationship between the end of the reactive bromine season and snowmelt timing.In the future, continued decreases in Arctic sea ice extent or the relative distributions of multi-year and seasonal sea ice cover (Peterson et al., 2019), coupled with increases in the length of the snow-free season over land, could influence the magnitude and seasonality of O 3 sinks via changes in halogen fluxes or dry deposition fluxes to tundra and ocean rather than snow and ice surfaces.It has also recently been shown that substantial O 3 depletion can occur due to reactions with iodine (Benavent et al., 2022).That study, which was based on ship measurements during the MOSAiC expedition in March to October 2020, found that iodine contributed more to O 3 loss than bromine, thus highlighting how the dynamics of high Arctic O 3 depletion are still not fully elucidated.

AMAP models and simulations
To evaluate our process understanding of controls on the Arctic tropospheric O 3 budget and distribution, we evaluate a subset of the same model simulations that were used in AMAP (2022) and by Whaley et al. (2022).A total of 12 atmospheric models participated in this study: seven chemical transport models (DEHM, EMEP MSC-W, GEOS-Chem, MATCH, MATCH-SALSA, OsloCTM, WRF-Chem) and five chemistry-climate models (CESM, CMAM, GISS-E2.1,MRI-ESM2, and UKESM1), with simulations of the years 2014-2015 for comparisons to observations.All models used the same set of anthropogenic emissions called ECLIPSEv6b (AMAP 2022), though they had different sources for fire, biogenic emissions, and meteorology (see Table S1  CH 4 emissions and various assumptions on natural CH 4 emissions (Olivié et al., 2021;Prather et al., 2012).Models then allow CH 4 to take part in photochemical processes.
The participating models have varying degrees of spatial resolution and chemical complexity; air-quality-focused models, such as DEHM, EMEP MSC-W, GEOS-Chem, MATCH, and WRF-Chem, have detailed HO x -NO x -hydrocarbon O 3 chemistry, with speciated VOCs and secondary aerosol formation, and they tend to run at higher resolution.The Earth system models GISS-E2.1,MRI-ESM2, and UKESM1 also contain this level of tropospheric chemistry, though they run globally at coarser resolution, whereas climate-focused models like CMAM run at a coarse resolution and have simplified tropospheric chemistry in order to be able to run for long periods.For example, CMAM's tropospheric chemistry consists only of CH 4 -NO x -O 3 chemistry, with no VOCs.
As mentioned above, Arctic tropospheric O 3 is heavily influenced by imports from the stratosphere.The models vary, too, in their representation of the stratosphere.Only a subset of participating models has a fully simulated stratosphere.CMAM, MRI-ESM2, GISS-E2.1,OsloCTM, and UKESM1 contain relatively complete stratospheric O 3 chemistry (NO x , NO x , Cl x , Br x chemistry that controls stratospheric O 3 ).Other models have a simplified stratosphere, such as GEOS-Chem which has a linearized stratospheric chemistry scheme (LINOZ; McLinden et al., 2000) and WRF-Chem which specifies stratospheric concentrations from climatologies.Finally, several models have no stratosphere or stratospheric chemistry at all (e.g., DEHM and EMEP MSC-W).Most atmospheric models, including all of the models in this study, do not yet contain Arctic tropospheric bromine chemistry and thus cannot simulate the surface-level bromine-driven O 3 depletion events that occur during spring.However, there are research versions of some models which are starting to contain this chemistry (e.g., Parrella et al., 2012;Falk and Sinnhuber, 2018;Badia et al., 2021) These same 12 model simulations were also evaluated against a different set of measurements in AMAP (2022) and Whaley et al. (2022).Those studies focused on many SLCF species over the Northern Hemisphere and generally reported model biases for the annual mean concentrations.They found that all models overestimated surface O 3 concentrations at midlatitudes but that there were both overestimation and underestimation in the Arctic.Particularly, models overestimated surface O 3 in the western Arctic (e.g., Alaska), particularly in the summertime, but were better able to simulate the surface O 3 seasonal cycle in the eastern Arctic (e.g., northern Europe).They also found that model biases were small throughout the free troposphere when compared to remote measurements from the TES satellite instrument.
In the next sections, these models are compared with observations of O 3 (at measurement sites located in Fig. 4) and its precursors either individually or as the multi-model median (MMM) -whereby the median of all 12 atmospheric models at the measurement locations is shown unless otherwise noted.The model output was selected from the model grid box that contains the latitude and longitude of the observation location without any spatial interpolation.

Arctic-wide tropospheric distributions from satellite data
Despite the potential limitations of some satellite data products at high latitudes, several studies have exploited satellite observations to investigate tropospheric O 3 and precursor distributions as well as trends relevant to the Arctic.Satellite observations are also useful in evaluating the sources and export of O 3 precursors from midlatitude source regions and their subsequent transport to the Arctic.Tropospheric NO 2 columns measured from the Ozone Monitoring Instrument (OMI) have been used to detect enhancements and trends in NO x emissions due to gas flaring in highlatitude (up to 67 • N) areas of Russia and North America (Li et al., 2016).Assessment of a suite of chemical transport models using OMI tropospheric NO 2 columns for summer 2008 showed a potential overestimate in NO 2 over biomass burning regions in eastern Siberia, with lower biases over European and North American source regions and underestimates over China (Emmons et al., 2015).A comparison of regional model-simulated tropospheric NO 2 columns with observations from OMI suggests potential underestimates in anthropogenic NO 2 emissions over high-latitude Siberia and the Russian Arctic (Thorp et al., 2021).Monks et al. (2015) exploited limited profile information from MOPITT CO retrievals to evaluate relationships between CO seasonal cycles in the lower and upper troposphere over the Arctic and midlatitude source regions.Atmospheric Infrared Sounder (AIRS) CO retrievals from 2007 to 2018 have been used to characterize atmospheric circulation patterns coincident with pollution enhancements during Arctic spring (Thomas et al., 2021), and IASI CO column measurements have been used to analyze transport pathways for Asian anthropogenic pollution to the Arctic (Ikeda et al., 2021).Osman et al. (2016) constructed three-dimensional (5 • × 5 • , 1 km) gridded climatologies of CO via a domain-filling trajectory mapping technique based on MOZAIC-IAGOS in situ measurements of commercial aircraft flights.These climatologies agreed well using forward and backward trajectories (< 10 % difference for most cases) and against vertical measurements from MOZAIC-IAGOS not included in the climatologies.These climatologies were compared with CO retrievals from MOPITT; small biases were found in the lower troposphere, while differences of ∼ 20 % were found between 500 and 300 hPa, which declined throughout the study (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012).Interannual variability in PAN retrieved by TES over eastern Siberia for April 2006-2008 was documented by Zhu et al. (2015), and it was shown to be largely controlled by boreal fire emissions at this time of year.More recently, PAN data from the TES instrument were used to help characterize Asian influence on exported PAN and downwind O 3 production (Jiang et al., 2016).A temperature-dependent high bias in TES PAN was found at cold temperatures over high latitudes.
In both Chapter 7 of the 2022 AMAP SLCF report (AMAP, 2022) and Whaley et al. (2022), data from satellite instruments, TES, the Atmosphere Chemistry Experiment (ACE) Fourier Transform Spectrometer (FTS) (ACE-FTS), and MOPITT are used to evaluate modeled O 3 , CH 4 , and CO in the Northern Hemisphere.They showed that model biases for CH 4 were small, though they tended to be negative in the Arctic due to a lack of north-south gradient in the prescribed global distribution.Model biases were also negative for free-tropospheric O 3 ; however, it was by approximately the same amount that TES O 3 retrievals have been shown to be biased high by Verstraeten et al. (2013).The ACE-FTS comparison for O 3 showed good agreement but had higher model biases around 300-100 hPa in Whaley et al. (2022) and AMAP (2022).The MOPITT CO comparisons in AMAP (2022) showed that all models' CO values are biased low over land in the midlatitudes but biased high over the oceans at lower latitudes.Monks et al. (2015) discussed the fact that models had high biases in the outflow from Asia and low biases north of there due to lack of transport.The

Surface ozone
Seasonal differences in the Arctic are important because of differences between the local meteorological conditions, as well as atmospheric transport, in the warm and the cold seasons and seasonal variations in O 3 sources and sinks.Surface O 3 at remote midlatitude sites with limited influence from local and regional anthropogenic O 3 precursor emissions has been found to frequently exhibit a characteristic seasonal cycle with peak values during spring and a minimum in the summer, while sites with high exposure to O 3 from anthropogenic precursors have summertime O 3 maxima (Monks, 2000;Parrish et al., 2013Parrish et al., , 2019;;Gaudel et al., 2018).The spring maxima have been explained by stratospheric intrusions as well as enhanced photochemical formation during this period of the year.The summer minima, e.g., observed at the Mace Head site (Derwent et al., 1998(Derwent et al., , 2013(Derwent et al., , 2020)), which is strongly influenced by marine air, appear to be explained by photochemical destruction in the absence of anthropogenic precursors.Seasonal cycles at Arctic stations have been discussed in the literature, and it is evident that the halogen chemistry discussed above, which is most frequently observed at high Arctic coastal stations, leads to a significant reduction during the springtime (e.g., Oltman and Komhyr, 1986;Tarasick et al., 1995;Monks et al., 2015, Helmig et al., 2007).Anderson et al. (2017) found that monthly mean observed near-surface O 3 concentrations at background sites in Sweden from 1990 to 2013 had a maxima in spring, with the most northerly stations experiencing their maximum in April and the southerly (non-Arctic) ones in May.
In order to get an overview of the annual O 3 cycles at different types of Arctic surface measurement sites, we have calculated the monthly medians and interquartile range for the period 2003-2019 for a series of sites.A map of the sta-tions as well as their coordinates and elevation can be seen in Fig. 4. Figure 5 illustrates the range of seasonal cycle behavior observed in the Arctic at different measurement sites and shows different seasonal cycles depending on location.

High Arctic sites
Figure 5a shows that the seasonalities in O 3 at Villum, Barrow (Utqiaġvik), Alert, Tiksi, and Eureka are similar: they have a local minimum in spring due to the occurrence of ODEs, a slight increase or recovery in June, and a second minimum in July due to surface removal and photochemical degradation of O 3 .These stations are located at high-latitude coastal sites close to sea level.During winter, O 3 reaches a maximum; due to an absence of photochemical degradation of O 3 , vertical mixing is suppressed during polar night since the Arctic boundary layer is often highly stratified, thus hampering removal by dry deposition (Esau and Sorokina, 2016).

Sites near the Arctic Circle
The characteristic seasonal variations of surface O 3 measured at stations close to the Arctic Circle are shown in Fig. 5b.The stations are Karasjok and Tustervatn in Norway (Aas et al., 2021), Esrange in Sweden, Pallas in Finland, and Denali in Alaska (note that regular O 3 monitoring at Karasjok ended in February 2010).The sites in Fig. 5b, which are not influenced by ODEs, exhibit a yearly cycle that is more similar to lower-latitude European stations at remote locations.Here, surface O 3 exhibits a late spring maximum which is attributed to photochemical production and transport of O 3 from the stratosphere (Monks, 2000).The largest differences between the stations are mainly found during the summer months, most likely due to differences in the influence of local sources on photochemical O 3 production (e.g., shipping; Marelle et al., 2016) and differences in the distance to pollution sources (Anderson et al., 2017).

Inland, high-elevation sites
Summit (located in the free troposphere on the Greenland Ice Sheet) is much less affected by bromine chemistry originating from sea ice or other low-altitude processes than the coastal high Arctic sites (Huang et al., 2017).Consequently, the seasonal variation is different with a clear maximum in May and a minimum in September; the higher concentrations compared to other surface stations can be explained by the high sensitivity to stratospheric O 3 -enriched air (Monks et al., 2015) at this high-elevation (3211 m a.s.l.) site.Short episodes of depletion have been reported (Brooks et al., 2011), but they do not appear to substantially affect the monthly mean values as shown in Fig. 5c.
Zeppelin, although it is a high Arctic site, is located on a mountain ridge at 474 m a.s.l. and thus experiences freetropospheric air masses more often compared to sea level sites.For this reason, it is less influenced by ODEs and consequently does not have an O 3 minimum in spring like the other high Arctic coastal stations (Fig. 5c).That said, ODEs have been reported there by Solberg et al (1996), Lehrer et al. (1997), Berg et al. (2003), Eneroth et al. (2007), and Steffen et al ( 2008), for example.ODEs have also been observed at the foot of the mountain, at the coastal station Gruvebadet, Ny-Ålesund (40 m a.s.l.), by Ianniello et al. (2021).
We also note that surface O 3 can be influenced by local anthropogenic emissions such as shipping (e.g., Marelle et al., 2016;Aliabadi et al., 2015;Eckhardt et al., 2013)

Surface O 3 model evaluation
It has been found that halogen chemistry, stable boundary layers, and dry deposition explained differences between measured and modeled O 3 concentrations, as demonstrated by Kanaya et al. (2019), who performed measurements of CO and O 3 during several ship cruises in the Bering Sea and the Arctic Ocean in September (2012 to 2017).None of the models in our study contain surface halogen chemistry, but they also display highly variable agreement in their surface O 3 seasonal cycles.Figure 6 shows the seasonal cycle from the models and observations averaged for 2014-2015 at several Arctic observation locations.Since the models do not contain surface-level bromine chemistry, at locations like Alert and Barrow (Utqiaġvik), they do not capture the springtime minimum in O 3 .Some models (e.g., UKESM1) greatly underestimate wintertime O 3 .This may be related to ter with observations and each other on summertime surface O 3 abundance at all locations and on the full seasonal cycle at Summit, the high-elevation background location.The large range of modeled surface O 3 is similar to previous model studies (Shindell et al., 2008;Monks et al., 2015;Gaudel et al., 2018).Despite the large range in model performance, the overall average negative O 3 bias and the seasonality in model bias at Barrow (Utqiaġvik) and Summit are consistent with these previous studies.The comparisons highlight little change in the skill of models in simulating Arctic surface O 3 over the past decade.These particular model simulations have been evaluated in Whaley et al. (2022), who grouped all western Arctic (defined as lat > 60 • N, and long < 0 • ) and eastern Arctic (lat > 60 • N, long > 0 • ) O 3 measurements together and showed the range in modeled and measured seasonal cycles for those two regions.That analysis included additional locations at lower latitudes, and thus their results emphasized that some models overestimated summertime O 3 in the western Arctic.Otherwise, the results from that study are consistent with what we report here.

Ozone precursors
NO x monitors have been used at several Arctic sites, but in a study at Zeppelin, it was shown that most of the NO x was in the form of the reservoir species PAN (Beine et al., 1997;Beine and Krognes, 2000).We evaluate and discuss PAN in Sect.6.3 from aircraft measurements.There are only limited sources for NO x in the Arctic and the lifetime of NO x is on the order of a day.Whaley et al. (2022) evaluated surface NO x volume mixing ratios and found that these models underestimated surface NO 2 by −59 % at low Arctic latitudes that were mostly around 60 • N.
The dominant source for NO x is long-range transport of dominantly PAN (Beine and Krognes;2000) and particulatebound HNO 3 , followed by reactivation in the Arctic by thermal decomposition and photoreduction processes, respectively.Kramer et al. (2015) determined at Summit from July 2008 to July 2010 that PAN accounted for 295 ppt and NO x for 88 ppt.In a more recent study, Huang et al. (2017) found in the period July 2008-June 2010 that PAN and NO x were maximum in spring at about 250 ppt and 25 ppt, respectively, and in summer 75 and 20 ppt, respectively.Beine and Krognes (2000) measured PAN at Zeppelin Mountain between 1994and 1996.They found that 3-month seasonal mean values were lowest in summer at 89.4 ppt and highest in spring at 222.6 ppt.HNO 3 in the gas phase is in general very low (Wespes et al., 2012).Particulate-bound nitratepotentially a significant source of NO x in the atmosphere and snowpack -is close to the detection limit in summer and up to 124.7 ng N m −3 in winter at Villum (Nguyen et al., 2013).
Pernov et al. ( 2021) measured organic O 3 precursors online with a proton-transfer-reaction time-of-flight mass spectrometer (PTR-ToF-MS) at Villum from April to October 2018.Sources were apportioned with positive matrix factorization.During the late spring, the Arctic haze factor was a source of oxygenated VOCs (OVOCs) arising from longrange transport of anthropogenic emissions, whilst during summer, OVOCs, namely organic acids, and dimethyl sulfide (DMS) originated from the marine cryosphere factor, with source regions in the Greenland Sea.During autumn, the biomass burning factor peaked in importance and was dominated by acetonitrile.The most abundant compound during the campaign was acetone, with a mean mixing ratio of 0.6 ppbv, as well as 0.027 ppbv for benzene and 0.046 ppbv for DMS.In the future, local NMVOC emissions might increase from both natural and anthropogenic sources due to retreating sea ice, with more biological activity, more industrial activity, and shipping affecting future levels of O 3 .The long-term VOC measurements at Zeppelin and Pallas (Platt et al., 2022;Hellén et al., 2015) provide valuable datasets for better understanding tropospheric O 3 at those locations.However, in this study, models did not provide much VOC output and, when they did so, only as monthly means of a few species (e.g., ethane C 2 H 6 ).Therefore, we did not evaluate modeled VOCs in this study other than CO.
Figure 7 shows the observed and simulated seasonal cycle of CO at Zeppelin and Barrow (Utqiaġvik).Simulated CO ranges about 50 ppbv across models, and all models underestimate surface CO at these sites.The low model biases are dominated by the winter and spring months.The 2014-2015 annual multi-model median (MMM) bias is −11 % and −16 % at Zeppelin and Barrow (Utqiaġvik), respectively.Figure 7 shows that for the first 6 months of the year, the MMM is 20 %-30 % too low, but that in the summer, the MMM is much closer to observations.These CO results are very similar to those found in previous multi-model studies (Shindell et al., 2008;Monks et al., 2015;Whaley et al., 2022).Similar to O 3 , these results imply little change in the skill of models in simulating Arctic surface CO over the past decade.The modeled CO underestimations are well-reported in the literature and attributed to either a lack of CO from combustion sources in the emission inventories (e.g., Kasibhatla et al., 2002;Pétron et al., 2002;Jiang et al., 2015) https://doi.org/10.5194/acp-23-637-2023Atmos.Chem.Phys., 23, 637-661, 2023 or to errors in OH, which impact the lifetime of CO (e.g., Monks et al., 2015;Quennehen et al., 2016).Indeed, both may be at cause here, as the anthropogenic CO emissions from ECLIPSEv6b are lower than those in the CMIP6 emission inventory, neither of which have taken into account the reported discrepancies from top-down emissions studies (Kasibhatla et al., 2002;Pétron et al., 2002;Jiang et al., 2015, Miyazaki et al., 2020).Monks et al. (2015) showed that models with lower global mean OH concentrations produced smaller underestimates in Arctic surface CO and that models with larger underestimates in CO over midlatitude source regions also had larger underestimates in Arctic CO.Emmons et al. (2015) showed that the models with larger tropospheric OH also had higher photolysis rates of O 3 to O( 1 D) and that there was also some relationship between higher photolysis rates and lower cloud cover fraction in some models.Previous multi-model results have also shown that variability in model water vapor abundance in the Arctic appeared to be the leading driver of model variability in OH, despite being much less important at lower latitudes (Monks et al., 2015).
Evaluating OH and water vapor is unfortunately beyond the scope of our study.
The models of this study prescribed CH 4 concentrations, including their increasing trend, and they were found to have a small bias of ∼ 2 % in Whaley et al. (2022) compared to surface and satellite measurements.Going forward, models are starting to simulate CH 4 explicitly from emissions, and this will be important for simulating future changes in Arctic tropospheric chemistry.

Vertical distributions of O 3 and precursors in the Arctic
Observations and models have both demonstrated extensive layering of pollution signatures in the Arctic troposphere vertical profile, associated with varying air mass origins with altitude (Zheng et al., 2021;Willis et al., 2019).Largescale isentropic transport pathways result in air masses from warmer more southerly latitudes being imported into the Arctic upper troposphere, while emissions from cooler northerly latitudes enter the Arctic near the surface and in the lower troposphere (Stohl, 2006).The presence of the Arctic dome during winter essentially shuts off access to the Arctic surface to air mass import from southerly midlatitudes, while it facilitates efficient low-level transport of emissions from northern Eurasia and Russia to the Arctic surface, giving rise to the well-known Arctic haze (Shaw, 1995).In practice, this large-scale dynamical control on long-range transport to the Arctic gives rise to a well-characterized vertical dependence of source region sensitivities for O 3 and precursors through the Arctic troposphere, where emissions from South and East Asia have the most influence in the Arctic upper troposphere, emissions from North America have the most influence in the Arctic mid-troposphere, and northern Eurasian and Russian emissions dominate at the surface (in addition to local influ-ences) (Wespes et al., 2012;Monks et al., 2015).As mentioned in Sect. 1, this vertical layering and changes in the efficacy of O 3 radiative forcing with altitude have implications for the sensitivity of Arctic tropospheric O 3 forcing to regional emission perturbations.Despite evidence for extensive vertical layering in the Arctic troposphere and the potential for highly varying source contributions with altitude, aside from a limited set of regular O 3 sonde profiles, there is a severe lack of observations available on the vertical distribution of O 3 , and particularly its precursors, in the Arctic troposphere.There is an especially poor constraint on seasonal and interannual variability in O 3 precursor profiles.In this section, we make use of available vertical profile measurements of O 3 and its precursors to document our understanding of Arctic tropospheric O 3 profiles and to evaluate model-simulated vertical profiles of O 3 and precursors.

Ozonesondes
Ozone soundings provide a long-term record of Arctic O 3 through the depth of the troposphere.Since 1966, weekly soundings have been available at Resolute, and since the 1980s regular soundings, typically once a week, have been available from six stations north of 60 • N (Fig. 4, Table S2).All of these stations are located in the Canadian and European sectors, meaning that regular soundings are lacking in a large sector of the Arctic (e.g., Russia and Alaska).The measurements are conducted using the balloon-borne electrochemical concentration cell (ECC) ozonesondes, typically reaching an altitude of about 30 km.Random uncertainties in tropospheric measurements are about 5 %, and biases reported from field and laboratory comparisons to UV reference photometers are 1.0 ± 4.4 % in the lower troposphere and 5.3 ± 4.4 % in the upper troposphere (Tarasick et al., 2019b).Mean observed concentrations have a minimum close to the surface, gradually increase throughout the troposphere by about 50 %, and then increase sharply going into the upper troposphere and lower stratosphere (Figs. 8 and S1-2 in the Supplement).Observed seasonal cycles in the Arctic troposphere generally show a maximum in spring and summer and a minimum in fall and winter.For example, Christiansen et al. (2017) examined long-term ozonesonde records at nine Arctic stations reporting consistent seasonal cycles as a function of altitude between sites with later maxima in the mid-troposphere compared to the surface layers and upper troposphere.

Model evaluation against ozonesondes
Figure 8 shows a comparison of the ozonesonde measurements at Eureka to the simulations from the 12 participating models for the annual and seasonal averages for the years 2014-2015.In the Supplement (Fig. S2), modelmeasurement comparisons at other Arctic locations are shown.Generally, the models are highly variable, ranging ±50 % of the measured O 3 profiles for most seasons and locations.The MMM performs well and is within ±8 % throughout most of the troposphere.However, all models, except UKESM1, have a bulge with a high model bias around 300-400 hPa, which is at or near the tropopause, implying that most models simulate the tropopause height too low (having larger stratospheric O 3 concentrations appearing too low in altitude).This results in a positive bias of about 20 % for the MMM around the tropopause.This feature in models was also reported in AMAP (2015), where model biases were particularly large at Ny-Ålesund and Summit.They associated those with differences in the transport of air masses from the stratosphere.This issue will have an impact on estimating the tropospheric O 3 burden, which is a common climate diagnostic (Griffiths et al., 2021).
At Alert, there are both surface and ozonesonde measurements, and we find that the results in the lowest levels of the Alert ozonesonde comparisons (Fig. S1) are consistent with the model biases found in Fig. 8 in that both show the models underestimating winter and fall O 3 , overestimating spring, and matching observations well in the summer at this location.
Note that the models' monthly average O 3 values were used in this comparison, which does not match the time of day and day of the week of the ozonesonde measurements.However, when a careful time matching to 3-hourly model output is carried out, the general features of the model biases remain the same (Fig. S2), likely because of the lack of a strong diurnal cycle in Arctic O 3 and its relatively long lifetime in the free troposphere.
The results of this model evaluation of the Arctic O 3 vertical profiles are consistent with Whaley et al. (2022), who compared the same model simulations to TES O 3 retrievals throughout the troposphere at lower Arctic locations (∼ 60-70 • N).They found models to be biased low (around −10 %), though the TES measurements have been shown to be biased high by about the same amount (+13 % bias in TES measurements reported in Verstraeten et al., 2013).They also saw a small positive shift in the model bias profile around 300 hPa.Finally, the Whaley et al. (2022) study included O 3 , NO x , CH 4 , and CO comparisons to the Atmosphere Chemistry Experiment (ACE) Fourier Transform Spectrometer (FTS) satellite instrument, and those results also independently implied that the modeled tropopause heights are too low.

Vertical distribution of O 3 precursors
Intensive field measurement campaigns using aircraft provide the most detailed observational constraint on vertical profiles of tropospheric O 3 precursors in the Arctic.While these datasets tend to provide excellent spatial and temporal resolution measurements on a wide range of species, they are episodic in nature and often cover only a period of a few days to several weeks, flying in specific regions of the https://doi.org/10.5194/acp-23-637-2023Atmos.Chem.Phys., 23, 637-661, 2023 The airborne NASA ATom (Atmospheric Tomography) mission (Wofsy et al., 2018;Thompson et al., 2022) has undertaken extensive surveying of the global troposphere.This includes repeated vertical profile measurements between 60 and 90 • N providing useful insights into the variation of O 3 (Bourgeois et al., 2020) and its precursors through the depth of the Arctic troposphere at different times of the year.Figure 9 shows these mean results and their standard deviation in the left-side panels, while the equivalent MMM results are in the right-side panels.The models' monthly mean results went into the MMM calculation, and the standard deviation from the models is shown.
The results show that near-surface NO 2 is greatly enhanced during winter, associated with a longer NO 2 lifetime and accumulation of pollution in the Arctic haze.The MMM simulates the surface NO 2 increase and the seasonality of the NO 2 profiles reasonably well.However, generally, the modeled NO 2 is biased low in the tropospheric profile, having average values of about 15 pptv in the 2-6 km range, whereas the measurements are about 25 pptv on average.This underestimate is consistent with that found at the surface by Whaley et al. (2022).PAN is also enhanced at the surface in the winter and can thermally decompose in the spring and summer to release NO x .The MMM generally overestimates PAN (Fig. 9c-d) and does not simulate the same shape in vertical profiles.For example, models are not able to simulate the wintertime surface-level increase in PAN, and they have the inverse shape of the observed profile in April-May.The best agreement is in summertime PAN (July-August), when the MMM vertical profile better matches that of the observations.The underestimate of NO x and the lack of winter surface increases in PAN by the models may be a reason why the wintertime surface O 3 concentrations in Sect.5.2 and Fig. 5 were underestimated.
In line with ozonesonde data and previous airborne campaigns (AMAP, 2015), ATom profiles also demonstrate a springtime enhancement in O 3 extending through the troposphere, with evidence of stratospheric influence in the upper troposphere and lower O 3 in the summertime lower troposphere.The models capture that springtime O 3 enhancement as well.Summer enhancements in O 3 precursors, such as CO and PAN in the mid-troposphere, were also observed associated with the import of forest fire and anthropogenic emissions from lower latitudes, as also seen during POLARCAT in 2008.The models capture this feature for PAN, but less so for CO.Indeed, most models underestimate CO.The annual mean MMM bias for surface CO in the Northern Hemisphere has been reported to be −30 % (Whaley et al., 2022).Figure 9 shows that below the tropopause, modeled O 3 is actually close to observed O 3 , despite the significant MMM biases for CO, NO x , and PAN.Around the tropopause, the aircraft data show the same issue that the ozonesonde data did -that models significantly overestimate O 3 near the tropopause.

Conclusions
Recent research on Arctic tropospheric O 3 has resulted in improvements to our understanding of this pollutant and GHG in the rapidly changing and sensitive Arctic environment.We have shown in this study that Arctic surface O 3 seasonal cycles are different depending on whether sites are near the coast, inland, or at high elevation.Coastal sites have springtime minima due to halogen chemistry causing ODEs and show a maximum during the winter.The inland locations near the Arctic Circle have quite consistent seasonal cycles, with maxima in April and minima in August.While the highelevation sites that are less influenced by halogen chemistry than coastal locations are more variable, Summit has a later maximum (May) and minimum (September); Zeppelin has an earlier maximum (March) and minimum (July).
Despite model development that has occurred since the 2015 AMAP assessment report on ozone (AMAP, 2015) to add processes, improve parameterizations, and increase resolution, among others, the resulting performance of the models remains more or less the same in terms of model variability and biases compared to measured O 3 and O 3 precursor species in the Arctic.Model results for CO would improve if CO emissions from combustion were increased, as suggested in the literature.It would also be useful to compare modeled OH and VOCs in the Arctic, but that was beyond the scope of this study.However, as Arctic O 3 is limited by NO x availability, improvements to CO and VOCs may not have a large effect on O 3 .Improvements to modeled PAN and NO x are needed; however, sensitivity studies to determine the cause of the model biases will be required to improve model performance for those species.For surface O 3 distributions in the Arctic, models simulate background levels reasonably well (e.g., at the high-elevation location of Summit), but surface bromine-halogen chemistry needs to be included to simulate springtime surface O 3 properly in the Arctic.Except near the tropopause, models simulate O 3 throughout the vertical profile well, with the MMM performing best at ±8 % depending on the location and altitude in the troposphere.Attention to improving the height of the modeled tropopause and/or the stratosphere-tropospheric exchange is still required since downward transport of high stratospheric O 3 concentrations is causing model biases around 6 to 8 km (400 to 300 hPa) to be significantly large (> 20 %).
While they are logistically challenging, additional O 3 measurements in the Arctic, such as O 3 deposition measurements, observations of stratospheric-tropospheric exchange, and O 3 concentrations in the Siberian Arctic, together with long-term measurements of O 3 precursors (such as those performed at Zeppelin and Pallas), would be particularly helpful to improve our understanding and modeling capabilities.This is particularly important as climate change alters the chemistry and dynamics of tropospheric O 3 in the future.as Jeffrey Seabrook and James Whiteway for their figure originally published in the Journal of Geophysical Research.We thank the ATom team for the original aircraft measurements.The technicians and logistical support staff at the different stations are gratefully acknowledged for their work, particularly Doug Worthy for Alert data, Karin Sjöberg for the Esrange data, and Karri Saarnio for the Pallas data.Review statement.This paper was edited by Radovan Krejci and reviewed by two anonymous referees.

Figure 2 .
Figure 2. Ozone lidar measurements from Eureka in the spring of 2008 showing effects of large-scale meteorology including low O 3 in the lower troposphere when air masses originate from the north over the Arctic Ocean and enhanced O 3 during downward transport into the Arctic boundary layer when the airflow was from the south over mountains.From Fig. 3 in Seabrook and Whiteway (2016).
in the Supplement).The years 2014-2015 were chosen for model validation as it was the most recent time period that the ECLIPSE v6b historical emissions were available when the model simulations were being set up.All participating models prescribe CH 4 concentrations based on box model results, which are, in turn, based on the ECLIPSEv6b anthropogenic https://doi.org/10.5194/acp-23-637-2023Atmos.Chem.Phys., 23, 637-661, 2023

Figure 4 .
Figure 4. Map of the surface (red) and ozonesonde (blue) sites cited in the present study, with coordinates and elevation.Eureka and Alert are both surface and sounding sites.Utqiaġvik was formerly called Barrow.The Arctic Circle at 66.55 • N is also shown in the figure for reference.
Quennehen et al. (2016) study also suggested that summertime CO transport out of Asia is zonal.This could explain some of the underestimations in the Arctic CO in the midtroposphere.https://doi.org/10.5194/acp-23-637-2023Atmos.Chem.Phys., 23, 637-661, 2023 5 Arctic surface O 3 and precursors: seasonal cycles In the high Arctic (> 70 • N), there is very little diurnal variation in surface O 3 because the local and regional photochemistry is of limited importance most of the time and due to the 24 h daylight during Arctic spring, summer, and autumn as well as the polar night during winter.The lack of diurnal cycle is also because there is inefficient O 3 deposition to the ice, snow, and water surfaces in the Arctic and a sparsity of vegetation.Therefore, with less deposition and limited photochemical production, there is a very limited diurnal cycle.For high Arctic sites, the seasonal dynamics of O 3 can be explained mostly by long-range transport, particularly in the winter and springtime, and intrusion from aloft (Hirdman et al., 2010); see Figs. 1 and 5a.Moving southwards to the Polar Circle a clearer diurnal pattern is evident caused by the seasonal behavior of vertical mixing, deposition, transport, and local chemistry (Andersson et al., 2017; Aas et al., 2021; AMAP 2022) like the stations on the Scandinavian peninsula and Denali, central Alaska.

Figure 5 .
Figure 5. Seasonal behavior of surface O 3 at selected Arctic stations that are representative of sites in the (a) coastal high Arctic, (b) near the Arctic Circle, and (c) at high elevation.Monthly medians are calculated for the period 2003 to 2018.Data were not available from 2003 to 2006 for Villum or 2004 and 2013-2015 for Alert.Data from Tiksi were available for the period 2013-2018, and at Karasjok the measurements stopped in 2010.The error bars show upper (75 %) and lower (25 %) quartiles.
or oil field emissions(McNamara et al., 2019).McNamara et al. (2019) discussed potentially important interactions between local anthropogenic NO x emissions from the Barrow (Utqiaġvik) settlement or the Prudhoe Bay oil extraction facilities in northern Alaska and snowpack (chlorine) chemistry leading to elevated concentrations of nitrogen-containing compounds (e.g., N 2 O 5 , HO 2 NO 2 ), with implications for Arctic tropospheric O 3 .Therefore, while none of the Arctic sites currently exhibit summertime surface maxima due to photochemical production, as often observed in polluted locations further south, this may change in the future with increasing local anthropogenic emissions (e.g.,Granier et al., 2006;  Law et al., 2014;Marelle et al., 2018).He et al. (2016) measured O 3 and black carbon on a ship cruise to the Arctic Ocean (31.1 to 87.7 • N and 9.3-90 • E to 168.4 • W) from June to September 2012.Comparing the observed O 3 concentrations to those measured at Barrow (Utqiaġvik) showed no statistically significant differences; the authors suggest that coastal stations between July and September may be representative of the entire Arctic, but this hypothesis requires further investigation.Indeed, our results show significant differences in the O 3 seasonal cycles at different Arctic locations depending on whether they were coastal, inland, or high-elevation.https://doi.org/10.5194/acp-23-637-2023Atmos.Chem.Phys., 23, 637-661, 2023

Figure 6 .
Figure 6.Arctic surface O 3 by month; seasonal cycle model comparisons.Top row: coastal high Arctic sites; middle row: sites near the Arctic Circle; bottom row: high-elevation sites.The solid black line is the observed O 3 monthly means, and the dashed black line is the multi-model median.Bottom row: sub-panels show the MMM percent difference [(MMM − measurements)/measurements × 100].Note that model results are from the 2014-2015 mean.When available, the same years are used for the observations.However, Alert did not have data for 2014-2015, so its most recent years were used: 2010-2013.Summit had 2014 but only 1 month in 2015, so its 2013-2015 data were used.
deficiencies in boundary layer mixing or an overly shallow boundary layer depth, resulting in the overly active titration of O 3 by NO near NO x emission sources and subsequent underestimation of Arctic surface O 3 .However, other model deficiencies could also play a role, including dry deposition and NO x lifetime.Indeed,Barten et al. (2021) found that overestimation of oceanic O 3 deposition can explain some differences between modeled and measured surface O 3 in the high Arctic.Some models in Fig.6do not agree on the timing of the springtime peak, with CMAM, DEHM, and GISS-E2.1 peaking in April and EMEP MSC-W and MRI-ESM2 peaking in May-June.The same groupings of models display different O 3 behavior at the end of the year (October-December), with CMAM, DEHM, and GISS-E2.1 all correctly simulating an increase in O 3 and EMEP-MSC-W and MRI-ESM2 having a decrease.All models agree bet-

Figure 7 .
Figure 7. Arctic surface CO by month; seasonal cycle model comparisons.The solid black line represents the observed CO monthly means, and the dashed black line is the multi-model median (MMM).Bottom panels show the MMM percent difference [(MMM − measurements)/measurements × 100].Note that model results are from the 2014-2015 mean.When available, the same years are used for the observations.However, for Zeppelin observations are the mean of 2013-2014.

Figure 8 .
Figure 8.Comparison between observed (thick black line in the left panels) and AMAP models' (colored lines) O 3 seasonal averages for 2014-2015 at Eureka, NV, Canada.These use monthly mean model output.In each right panel, the dotted black line is the MMM, and the dashed black line shows zero bias for reference.See Fig. S1 for the rest of the ozonesonde locations and a sample comparison done with 3-hourly model output (Fig. S2).

Figure 9 .
Figure 9. Mean vertical profiles of O 3 , CO, PAN, and NO 2 (left) measured in Alaska and Greenland from the NASA ATom missions during summer 2016, winter 2017, autumn 2017, and spring 2018 (horizontal lines indicate 1 standard deviation spread around mean values at each altitude).The (right) MMM for the years 2014-2015 (with the MMM standard deviation as horizontal lines).The observations appear as dashed lines in the right panels for ease of comparing to the MMM.https://doi.org/10.5194/acp-23-637-2023Atmos.Chem.Phys., 23, 637-661, 2023

Financial support .
Makoto Deushi and Naga Oshima were supported by the Japan Society for the Promotion of Science KAK-ENHI (grant numbers: JP18H03363, JP18H05292, JP19K12312, JP20K04070 and JP21H03582), the Environment Research and Technology Development Fund (JPMEERF20202003 and JP-MEERF20205001) of the Environmental Restoration and Conservation Agency of Japan, the Arctic Challenge for Sustainability II (ArCS II) under program grant number JPMXD1420318865, and a grant for the Global Environmental Research Coordination System from the Ministry of the Environment, Japan (MLIT1753 and MLIT2253).Joakim Langner and Manu A. Thomas were supported by the Swedish Environmental Protection Agency through contract NV-03174-20 and the Swedish Clean Air and Climate research program.Svetlana Tsyro and Michael Gauss have received support from the AMAP Secretariat and the EMEP Trust Fund.Ulas Im received support from the Aarhus University Interdisciplinary Centre for Climate Change (iClimate) OH fund (no.2020-0162731), the FREYA project funded by the Nordic Council of Ministers (grant agreement nos.MST-227-00036 and MFVM-2019-13476), and the EVAM-SLCF funded by the Danish Environmental Agency (grant agreement no.MST-112-00298).Henrik Skov received funding from the Danish Ministry for Energy, Climate and Utilities (grant agreement no.2018-3767), the Danish Environmental Agency (grant agreement no.MST-113-00140), and AMA.Kostas Tsigaridis and Gregory Faluvegi received support from the NASA Modeling, Analysis and Prediction Program (MAP).Steven T. Turnock received financial support from the Arctic Monitoring and Assessment Programme.Kathy S. Law, Jean-Christophe Raut, Louis Marelle, and Tatsuo Onishi (LATMOS) received support from the EU iCUPE (Integrating and Comprehensive Understanding on Polar Environments) project (grant agreement no.689443) under the European Network for Observing our Changing Planet (ERA-Planet) and from access to IDRIS HPC resources (GENCI allocation A009017141) as well as the IPSL mesoscale computing center (CICLAD: Calcul Intensif pour le CLimat, l'Atmosphère et la Dynamique) for model simulations.Jesper Christensen (DEHM model) received support from the Danish Environmental Protection Agency and Danish Energy Agency (DANCEA funds for Environmental Support to the Arctic Region project: grant no.2019-7975, grant no.MST-112-00298, grant no.TAS 4005-0153).Stephen R. Arnold and Steven T. Turnock both received financial support from the Arctic Monitoring and Assessment Programme.Stephen R. Arnold also received support from the UK Natural Environment Research Council and Belmont Forum via the ACRoBEAR project (grant NE/T013672/1).Joshua Fu received funding from the Oak Ridge Leadership Computing Facility at the Oak Ridge National Laboratory, which is supported by the Office of Science of the U.S. Department of Energy under contract no.DE-AC05-00OR22725.The research by Irina Petropavlovskikh and Peter Effertz was supported by NOAA cooperative agreements NA17OAR4320101 and NA22OAR4320151.