Sources of surface O 3 in the UK: tagging O 3 within WRF-Chem

. Tropospheric ozone (O 3 ) concentrations depend on a combination of hemispheric, regional, and local-scale processes. Estimates of how much O 3 is produced locally vs. transported from further aﬁeld are essential in air quality management and regulatory policies. Here, a tagged-ozone mechanism within the Weather Research and Forecasting


Introduction
Tropospheric ozone (O 3 ) is a pollutant of concern for policymakers because of its detrimental effects on human health, agriculture, and ecosystems (Fuhrer, 2009;WHO, 2016). Near ground level, O 3 has a typical atmospheric lifetime of a few hours. In the free troposphere, however, the lifetime can be up to several weeks (Stevenson et al., 2006), and O 3 can be transported from its point of production downwind over long distances crossing countries and continents (Wild et al., 2004;HTAP, 2010). The concentration of O 3 at a given location is therefore dictated by a combination of hemispheric, regional, and local-scale factors (Jenkin, 2008). Examples of this are long-range transport of O 3 and its precursors, including stratospheric intrusions, and photochemical reactions happening on a local and regional scale (e.g. Monks, 2000;HTAP, 2010).
The production of O 3 in the troposphere is highly nonlinear. It depends on the abundance of nitrogen oxides (NO x = NO 2 + NO) and peroxy radicals (HO 2 ) produced after the oxidation of volatile organic compounds (VOCs) by hydroxyl radicals (OH) (Monks, 2005). The reaction of NO with HO 2 and the subsequent photolysis of NO 2 generating O 3 constitute the primary known mechanism of O 3 production (Atkinson, 2000;Monks, 2005). NO x concentrations determine whether O 3 is produced or chemically removed (Monks, 2005). In the rural areas of most industrialized countries, where NO x is available at moderate levels, the rate of O 3 formation increases with increasing NO x concentrations (NO x -limited regime). In more polluted areas, by contrast, high NO x concentrations inhibit O 3 formation as this begins being depleted by NO (NO x titration effect). Subsequent formation of nitric acid (HNO 3 ) from the reaction of NO 2 with OH constitutes a major endpoint for O 3 in such environments (Monks, 2005). However, elevated inputs of non-methane VOCs (NMVOCs) can increase the production of O 3 as the reaction of VOCs with OH radicals become more significant (NO x -saturated regime). Furthermore, O 3 concentrations also depend on O 3 deposition; uptake by vegetation; and meteorological variables such as temperature, winds (direction and speed), solar radiation intensity, and precipitation (e.g. Sillman, 1999;Coyle et al., 2002). For instance, high-O 3 -concentration episodes in the UK have been associated with heatwave periods (Finch and Palmer, 2020). The contribution of each process varies with location. Remote sites are largely controlled by hemispheric background O 3 (Pilling et al., 2009). Photochemical pollution episodes, on the other hand, are more severe in the south and east of the UK, and O 3 titration by NO x is higher in urban areas (Jenkin, 2008).
In the UK, tighter UK and European precursor emission controls in the last 30 years have led to a substantial decrease in the concentration of O 3 primary precursors and successfully reduced the severity of the high-O 3 -concentration episodes (Pilling et al., 2009;Derwent et al., 2018;Finch and Palmer, 2020). Even so, exposure to surface O 3 continues to cause considerable damage to human health in Europe and the UK, leading to an estimated 17 000 premature deaths in 2015 (EEA, 2017). Evidence suggests that annual mean O 3 concentrations in the UK have been increasing in urban and suburban areas and to a lesser extent in rural areas (Jenkin, 2008;Pilling et al., 2009;Munir et al., 2013;Finch and Palmer, 2020). Reductions in NO x emissions, mainly by road traffic, have led to reductions in the O 3 scavenging in urban areas so that O 3 concentrations have generally increased (Finch and Palmer, 2020). The increase in rural areas, on the other hand, has been largely driven by rising hemispheric O 3 levels, up to +0.31 ppbv a year over the 20-year 1987-2007 period (Derwent et al., 2007) and +0.25 ppbv a year over the 25-year period (Derwent et al., 2013). Accordingly, increasing emissions of precursors in Asia and North America influence O 3 concentrations entering Europe from the North Atlantic, offsetting the effects of European regional emission reductions on O 3 (HTAP, 2010;Derwent et al., 2018). Therefore, efficient emission control policies aimed at reducing O 3 concentration in a given region require a holistic assessment of both O 3 transport from outside the region and in situ O 3 production. Such quantitative estimations can be made by applying source-receptor methods (S-R) within chemical transport models (CTMs).
S-R studies often compare model simulations that include all anthropogenic emissions with those obtained after modifying emissions from a region of interest (the so-called perturbation approach). However, as O 3 chemistry is highly nonlinear, this approach can lead to unrealistic attribution estimates; e.g. Emmons et al. (2012) underestimated the O 3 contribution by up to a factor of 4 when perturbing NO emissions by 20 %. Tagged-ozone methods, on the other hand, use additional diagnostics to follow the reaction of different emissions to the formation of O 3 , making the approach suited to investigating the contribution of different precursors Grewe et al., 2012;Butler et al., 2018).
Several studies have investigated the contribution of intercontinental transport to O 3 in Europe, in particular from North America and Asia, using different modelling techniques (e.g. Li, 2002;Derwent et al., 2004;Auvray and Bey, 2005;Sudo and Akimoto, 2007;Derwent et al., 2008;Emmons et al., 2012;Derwent et al., 2015;Mertens et al., 2018;Butler et al., 2018;Lupaşcu and Butler, 2019;Butler et al., 2020). However, these studies do not provide a quantitative estimate of the contribution of the different source regions to the total amount of O 3 over the UK at a regional scale but rather an estimate at a national scale or at individual locations across Europe or for the European region as a whole.
The present study quantifies the contributions to surface O 3 in 12 receptor regions in the UK from anthropogenic NO x emissions from inside and outside the UK using the taggedozone method developed in Lupaşcu and Butler (2019). Dividing the UK into several regions serves to separate meteorological features and chemical environments that are known to impact the spatial distribution and temporal variation in air pollutants such as O 3 (Coyle et al., 2002;Jenkin, 2008). We focus on summer 2015, which saw several heatwaves causing elevated O 3 values in central and western Europe that exceeded the European Union (EU) information threshold of a 1 h average mixing ratio of 90 ppbv (Tarrason et al., 2016). We also look at the impact of O 3 on human exposure, crops, and vegetation using two well-known O 3 metrics, the MDA8 and the AOT40. The MDA8 is a health-related metric commonly used to assess the impacts of O 3 exposure on the population (e.g. Reidmiller et al., 2009;Stock et al., 2013;. The metric is defined as the maximum daily 8 h average (MDA8) O 3 values (in ppbv) and is strongly influenced by photochemical episodes (Pilling et al., 2009). The AOT40 (accumulated O 3 above a threshold of 40 ppbv) is commonly used to assess the effects of O 3 on crops and vegetation and is based on exposure over 40 ppbv using only the 1 h values measured during daylight hours.
The Weather Research and Forecasting model coupled with chemistry (WRF-Chem) settings, including an introduction to the tagging approach and a summary of the model evaluation for NO, NO 2 , and O 3 , are presented in Sect. 2. Model evaluation is discussed in detail in the Supplement. Results for the contributions of UK and European precursor emissions, along with transport across the lateral model boundaries to surface O 3 in the UK, are presented and discussed in Sect. 3. Section 4 summarizes our findings.

Methods
We used the Weather Research and Forecasting model (WRF) version 3.7.1 (Powers et al., 2017) coupled with chemistry (WRF-Chem) (Grell et al., 2005). The model domain was centred at 53 • N, 3 • E, covering most of Europe, as shown in Fig. 1a. The spatial resolution was set to 27 km × 27 km, with 35 vertical levels starting from the surface up to 10 hPa.
The initial and boundary conditions (ICs and BCs, respectively) for meteorology were obtained from the ERA-Interim reanalysis dataset (Dee et al., 2011), which has a spatial grid resolution of 0.75 • × 0.75 • and 6 h temporal resolution. ICs and BCs for the chemistry fields were extracted from global simulations produced by the Model for Ozone and Related chemical Tracers (MOZART-4) Goddard Earth Observing System-5 (GEOS-5) (Emmons et al., 2010). BCs were ingested into the model every 3 h. The schemes used to parameterize the atmospheric processes are listed in Table 1. These are the same schemes deployed in Mar et al. (2016) to evaluate meteorology, O 3 , and NO x fields in a European domain using the MOZART-4 chemical mechanism.
Simulations were conducted for the period between 24 April and 31 August 2015 for gas-phase chemistry using a tagged-ozone mechanism based on the MOZART-4 chemical scheme. Note that omission of heterogeneous chemistry can lead to overestimation of NO 2 due to the absence of aerosol nitrate formation through the reaction of OH + NO 2 as well as N 2 O 5 hydrolysis, which represents an important sink for NO 2 (Badia and Jorba, 2015;Archer-Nicholls et al., 2014;Stone et al., 2014). The first week of output was treated as model spin-up and hence discarded. The meteorology was not nudged but re-started every 3 d as in the methodology adopted in the second phase of the Air Quality Model Evaluation International Initiative (AQMEII) (e.g. Im et al., 2014). This decision was made after a test analysis showed that nudging of winds above the planetary boundary layer (PBL) and temperature at all layers, as done in , leads to a representation of hourly NO 2 and O 3 mixing ratios in the East Anglia region (in the east of the UK) that is inconsistent with observations. The nudging simulation predicted shallower boundary layers compare with that obtained using the re-starting method, particularly over the Norfolk coast, leading to high concentrations of NO 2 , especially at nighttime, and larger O 3 lost due to increased dry deposition. Anthropogenic emissions of carbon monoxide (CO), NO x , sulfur dioxide (SO 2 ), and total NMVOCs for the European domain, including shipping lanes, were taken from the TNO-MACC-III European inventory  for the year 2011. The emissions were provided as yearly totals (kg yr −1 ) by source sector following the SNAP (selected nomenclature for sources of air pollution) convention at a 0.125 • × 0.0625 • longitude-latitude resolution. For the UK domain, emissions were taken from the UK National Atmospheric Emissions Inventory (NAEI) for the year 2014, http://naei.beis.gov.uk/ (last access: 20 January 2022), which has a spatial resolution of 1 km × 1 km. Biogenic emissions were calculated online using the Model of Emissions of Gases and Aerosols from Nature (MEGAN) version 2.04 (Guenther et al., 2012).

O 3 tagging mechanism
The contribution of hemispheric O 3 and domestic and European anthropogenic emissions to tropospheric O 3 in the UK is studied using the O 3 tagging technique developed in Lupaşcu and Butler (2019), in which O 3 molecules are labelled according to the identity of their source regions. This is achieved by tagging NO x emissions at selected source regions and tracking them through the formation of O 3 , including the recycling of NO x via the production of odd nitrogen species (e.g. peroxyacetyl nitrate (PAN), nitric acid (HNO 3 ), and organic nitrates).
To implement the tagging method, a new chemical mechanism was created, "mozart_tag_kpp (chemopt = 113)", containing the original chemical reactions in the MOZART-4 mechanism plus a duplicated set of reactions with additional tracers accounting for the source regions of interest. See Appendix A in Lupaşcu and Butler (2019) for a list of the model's edits to accommodate the new mechanism in WRF-Chem. The present study uses different sources and recep-   (Lin et al., 1983) Radiation (short wave) RRTMG (Iacono et al., 2008) Radiation (long wave) Goddard shortwave scheme (Chou and Suarez, 1994) Boundary layer physics Yonsei University scheme (Hong et al., 2006) Surface layer MM5 similarity based on Monin-Obukhov scheme (Beljaars, 1995) Land surface processes Noah land surface model (Chen and Dudhia, 2001)

Cumulus convection
Grell 3-D scheme (Grell and Dévényi, 2002) tor regions. Furthermore, it does not attribute the contributions from the lateral boundary to any specific geographical location (e.g. the Task Force on Hemispheric Transport of Air Pollution, TF HTAP regions) as in Lupaşcu and Butler (2019). Instead, the lateral boundary is tagged as a single source. O 3 formation requires both NO x and peroxy radicals from VOCs. Several tagging methods exist that can take different approaches to estimate the attribution of O 3 to these two chemically distinct precursors (Butler et al., 2018). Butler et al. (2020) demonstrate that anthropogenic NMVOC emissions play a marginal role in regional-scale O 3 production, with methane and biogenic VOCs being the most relevant chemical species for O 3 production. As the present study primarily focuses on the anthropogenic influence on O 3 , the use of NO x tagging for O 3 source attribution is considered appropriate. Table 2 lists the tagged sources and receptor regions also highlighted in Fig. 1a and b, respectively. The chemical lateral boundary is defined as an independent source region (LB) and is provided by the MOZART-4 GEOS-5 model. Note that the LB-tagged region also includes O 3 contributions of stratospheric origin. Moreover, all O 3 that enters the model domain through the lateral boundaries is tagged as LB, and there is no feedback between the global model and WRF-Chem.

Receptors and source regions
To generate the receptor regions, the UK domain is divided into its 12 administrative regions as used in previous air quality studies such as Heal et al. (2013), as shown in Fig. 1b: East Anglia, South East, London, South West, Wales, West Midlands, East Midlands, Yorkshire and Humberside, North East, North West, Northern Ireland, and Scotland.
The UK is well known for the regional variability in its weather. Generally, places in the east and south tend to be drier, warmer, sunnier, and less windy than those in the west  (Jenkin, 2008). Thus, dividing the UK into several regions also serves to separate relevant meteorological features such as temperature, sunshine, precipitation, and wind, as well as emissions within each region that are known to have an impact on the spatial distribution and temporal variation in air pollutants such as O 3 (Coyle et al., 2002;Jenkin, 2008).

O 3 metrics for source contribution assessment
Current European and national air quality standards to mitigate the effects of O 3 on human health are expressed as 8 h averages. The regulatory framework establishes that the maximum 8 h mean O 3 concentration (MDA8) should not exceed 120 µg m −3 (∼ 60 ppbv) in the EU, and 100 µg m −3 (∼ 50 ppbv) in the UK. Here, the contributions to these health metrics were estimated by computing an 8 h moving mean of O 3 for each receptor region and selecting the days when the MDA8 exceeded 50 and 60 ppbv between May and August 2015. Once these were identified, tagged O 3 concentrations were extracted for the same periods and used in the analysis. The AOT40 is defined as the accumulated excess of hourly O 3 concentrations above 40 ppbv measured during daylight hours (between 08:00 and 20:00) central European time (CET) over a typical 3-month growing season May-July. Here, the contribution of tagged O 3 to the cumulative metric AOT40 was calculated as the sum of the difference between hourly mixing ratios when O 3 exceeded the 40 ppbv threshold and 40 ppbv between 08:00 and 20:00 central European time (CET) from May-July over the most relevant arable farming areas in the UK, East Anglia, and the South East; see Eq.
(1). The target value in the EU and UK is 9000 ppbv h −1 (∼ 18000 µg m −3 h −1 ) over a typical 3-month growing season (May-July) averaged over 5 years. Observational data were taken from the UK's Met Office Integrated Data Archive System and the European Monitoring and Evaluation Programme (EMEP). The EMEP air quality monitoring network records hourly measurements at regional background sites, mostly in farmland and rural areas (Tørseth et al., 2012). The choice to only analyse background representative stations is based on the need to provide an evaluation with spatial scales consistent with the model resolution. Kuik et al. (2016), for example, have shown that a 15 km resolution is too coarse to resolve the differences between urban and rural atmospheric chemical composition. The resolution of the domain considered here is even coarser. The EMEP network was therefore selected as it provides surface measurements at sites intended to represent regional background pollution.
Model evaluation is detailed in the Supplement. Table S2 in the Supplement summarizes the domain-wide statistical performance for NO, NO 2 , and O 3 . The predicted temporal correlation coefficient (r) for NO and NO 2 is fairly low (0.3), which is a feature also exhibited in other regional studies in Europe using WRF-Chem, e.g. Tuccella et al. (2012), Pirovano et al. (2012), and Lupaşcu et al. (2022). The model underestimates NO mixing ratios in most analysed sites with a domain-wide mean bias (MB) of −0.3 ppbv. NO 2 mixing ratios, on the other hand, are generally overestimated with a domain-wide MB of 0.31 ppbv and no specific patterns distinguished in the bias distribution. This is consistent with the negative NO and positive NO 2 biases obtained across Europe using MOZART-4 chemistry reported in Mar et al. (2016).
The model's temporal variation in hourly O 3 concentrations at most sites is well represented, with an average r value of 0.6. The model tends to underestimate concentrations in most locations, with a domain-wide mean bias of −3.7 µg m −3 . Correlation values above 0.5 are obtained in most sites, particularly in the UK; see Fig. S4a in the Supplement. In contrast, low r values (∼ 0.4) are concentrated in high-altitude sites, which might indicate difficulties in the model representing O 3 transport. This is in line with previous studies using MOZART-4 chemistry, such as , showing low production of peroxyacetyl nitrates (PANs), which are an essential reservoir for NO 2 and a key player in remote O 3 production. Correlation values are consistent with summertime O 3 values below 0.40 reported in the WRF-Chem model evaluation over a European domain in Mar et al. (2016) using MOZART-4 chemistry. Figure 2 shows that the day-to-day variation in hourly O 3 mixing ratios is well represented by the model, except for large under-predictions during 1-3 July and 22-24 August, particularly at stations on the east coast, e.g. Weybourne.
Note that the observed maximum hourly O 3 at this site is larger than those seen inland, e.g. Lullington Heath and Harwell (2015). This may indicate inflow of O 3 and precursors from nearby large metropolitan areas within the UK (e.g. London) or longer-range transport from continental Europe. Thus, underestimation of O 3 during those days may be caused by uncertainties in O 3 transport. This feature has also been identified in other source apportionment studies such as Lupaşcu and Butler (2019).

Results and discussion
3.1 Contributions from tagged sources 3.1.1 Spatial distribution and temporal variation  Table 2 for abbreviation definitions) accounts for nearly 16 % of the simulated monthly mean O 3 . The largest Eu super-region contributions are observed in the UK locations that are closer to continental Europe and that together contain about 40 % of the UK population (East Anglia, the London area, the South East, and Yorkshire and Humberside). The smallest Eu super-region contributions are observed over Scotland (May and June) and Ireland (July and August). Emissions from UK sources, on the other hand, accounted for about 13 % of the simulated monthly O 3 . The domestic contributions tend to increase in June and decrease again in August. This monthly variation in the O 3 contributions is mainly caused by larger photochemical activity taking place during the summer months (e.g. Monks, 2005). Under these conditions, O 3 is formed by reactions involving the oxidation of NMVOCs in the presence of NO x and under the influence of solar radiation.
The spatial distribution of the monthly O 3 concentrations and the absolute contribution of the source regions (UK, LB, and Eu super-region) to surface O 3 in the UK are shown in Fig. 4. The first column shows that the monthly surface O 3 concentrations are higher in May than during the summer months, in particular over Ireland, most of the Atlantic Ocean, north of the UK, Scandinavia, and the northern North Sea. This is consistent with the Northern Hemisphere midlatitude spring maximum (Monks, 2000), which is characteristic of remote locations and attributed to both transport of O 3 from the stratosphere to the troposphere (Monks, 2000) and transport of O 3 produced from anthropogenically emitted precursors (NO x and VOCs) (Monks, 2000;Butler et al., 2018). By contrast, the south-east of the UK, the southern North Sea, and continental Europe exhibit sustained high O 3 mixing ratios throughout the entire analysed period (May to August), reflecting a spring-summer maximum that is frequently attributed to photochemical O 3 production (Monks, 2000).
A marked latitudinal gradient is observed in the monthly O 3 mixing ratios, in particular during June, July, and August. Over the ocean areas, O 3 concentrations tend to be higher at the lower latitudes and in the North Sea; see Fig. 4e, i, and m. Mean O 3 mixing ratios as low as 22 ppbv (Fig. 4i) are observed in most of the UK and Scandinavia in July, while mixing ratios as high as 32 ppbv (Fig. 4e) are calculated for southern locations in the UK and western Europe in June. This is consistent with previous estimates such as those in Butler et al. (2018). Part of the latitudinal gradient in surface O 3 over land can be attributed to the changing O 3 mixing ratios arriving from the Atlantic Ocean. Moreover, high mixing ratios in the south-east UK during summertime are generally associated with photochemical production of O 3 (Monks et al., 2005), in particular from anthropogenic NO x and biogenic NMVOC emissions (Atkinson, 2000;Butler et al., 2018) as well as transport of O 3 rich air masses from continental Europe during anticyclonic conditions (Jenkin et al., 2002;Lee et al., 2006;Francis et al., 2011;Romero-Alvarez et al., 2022). Low mean O 3 mixing ratios (as low as 20 ppbv), on the other hand, are observed in the vicinity of the main urban centres, e.g. Greater Manchester, the Midlands, and the London area (first column, Fig. 4). This is because strong titration by excessive local NO x emissions takes place over the main urban centres, whereas high O 3 production rates are expected in the outskirts following the progressive reduction in NO x concentration relative to that of NMVOCs (Jenkin, 2008). Note that the latitudinal gradient across the UK is not evident in high-altitude areas in Wales and northern England, which have relatively high concentrations of O 3 . This is because much of the time high-altitude areas are above the shallow boundary layers that form over the lower-lying land and experience therefore larger exposure to O 3 .
A suitable way to identify the areas influenced by fresh NO x emissions is comparing the mixing ratios of O 3 and O x (i.e. O 3 + NO 2 ). O x is considered a conservative quan-  tity as it is, to a large extent, free from the titration effect of NO + O 3 → NO 2 (Kley et al., 1994). The effect of titration for July is evident in Fig. 5. When O x is considered, the mixing ratios tend to increase over the main urban centres such as London, Birmingham, Nottingham, Sheffield, and Greater Manchester and to a lesser extent over Edinburgh and Glas-gow. The NO titration effect is also observed within the main urban centres in continental Europe and along the shipping lanes over the North Sea and the English Channel due to the high NO content of ship emissions compared with those from NMVOCs (Aulinger et al., 2016).   The decrease in the monthly average O 3 mixing ratios towards the summer months over the Atlantic Ocean and most of the British Isles coincides with a progressive reduction in the contribution from LB O 3 over the same areas (second column in Fig. 4). Over remote marine areas, it is likely that the decrease in total O 3 is due to an increase in the photochemical activity and concentration of water vapour during the summer months. O 3 concentrations over land, on the other hand, are likely to be altered by both the changing background contribution from over the ocean and processes occurring at the regional and local scale (Jenkin, 2008). Such processes include O 3 scavenging near emission sources, changes in meteorology (wind direction influencing transport and temperature and radiation influencing photochemical production of O 3 ), and planetary boundary layer stability (influencing vertical mixing and deposition) (Pilling et al., 2009). In addition, LB O 3 can be chemically lost near emission sources, e.g. the Midlands and London area, shipping lanes, and over an extended area on the southern part of the Atlantic Ocean, as shown in the map of the net midday (11:00-14:00 UTC) near-surface LB O 3 chemical production rate in Fig. 6. The figure also shows how the absolute contribution from the LB decreases southwards and eastwards. Over the Atlantic, part of this can be attributed to a greater chemical O 3 sink due to the increase in photolysis of O 3 and subsequent production of OH radicals from water vapour (Johnson et al., 1999). Transport of O 3 from the stratosphere might also influence the spatial gradient in the contributions.
A marked reduction in LB O 3 is observed over the UK in column 2, Fig. 4 (e.g. a decrease of ∼ 10 ppbv between LB O 3 over the ocean and the UK). Depletion of surface O 3 by dry deposition and chemical loss processes within the UK, such as the reaction of O 3 with NO, may help explain the observed spatial gradient. Reductions in LB O 3 due to the effect of local O 3 scavenging by reaction with NO in the urban centres as illustrated in Fig. 7 might be an additional cause.
Whereas the absolute contribution from the LB over the UK tends to decline with distance towards the south-east, the absolute contribution of UK anthropogenic NO x emissions to the mean surface O 3 over the UK (Fig. 4, third column) decreases from the south-east to the north and west. The mean contribution of the UK-to-UK surface O 3 concentrations is marginal. Over most of the north and the west, the UK contributions ranged from 1-3 ppbv. By contrast, maximum UK contributions can reach up to 7 ppbv in the east and the Midlands during the summer months. These areas tend to be drier, warmer, and sunnier than those regions further west and north, features that are conducive to photochemical O 3 formation (e.g. Coyle et al., 2002;Jenkin, 2008). Furthermore, these regions contain some of the UK's largest cities (e.g. London, Birmingham, Nottingham, Manchester, and Leeds), which can lead to net O 3 formation downwind of the emission sources where the NO x titration effect is reduced. Indeed, overall, the south and east of the UK exhibit the highest midday (11:00-14:00) O 3 chemical production from UK anthropogenic sources; see Fig. 8a. Lower O 3 chemical production is instead observed in the west and the north beyond Yorkshire and Humberside, as shown in Fig. 8a. On the other hand, the UK makes a positive contribution to O 3 , of around 4-8 ppbv, downwind over continental Europe.
The contribution from European NO x emissions to the mean surface O 3 over the UK (Fig. 4, fourth column) is comparable to that observed from the UK contribution and tends to be higher along much of the eastern, southern, and southwest borders, reaching up to 10 ppbv in East Anglia during July. This reflects the effective transport of continental O 3 by south-easterly winds during high-O 3 -pollution events. The European contributions then decrease towards the northern and western areas of the UK, with a minimum (1-3 ppbv) over Scotland and Ireland. Figure 8b demonstrates that surface O 3 from anthropogenic sources from the Eu superregion is mainly produced outside the UK. This indicates that the contribution from EU emissions to UK surface O 3 is predominantly due to transport of O 3 rather than its NO y precursors. Also, O 3 from EU sources is chemically lost near the largest cities in the UK (e.g. the London area, Birmingham, Nottingham, Manchester, and Leeds) and in the English Channel and North Sea, as shown in the plot of the net midday surface chemical O 3 production rate from European anthropogenic NO x emissions in Fig. 8b. Chemical production is generally concentrated over central Europe and the Baltic Sea. By contrast, chemical loss happens within the main urban centres; near point sources; and along the shipping routes  around western Europe, the North Sea, and the English Channel, e.g. sites previously identified to be influenced by NO x titration.

Regional dependence
The modelled contributions of the different source regions to the UK receptor regions for May are presented in Fig. 9. The figure contains 12 nested pie charts, each one associated with a receptor region in the UK that shows the absolute and relative contributions to O 3 mixing ratios in the UK from all anthropogenic NO x sources, including ship emissions. Note that the contributions from the Rest_Eu source need to be carefully interpreted since these include emissions from the Republic of Ireland, Iberian Peninsula, southern EU, southeastern EU, eastern EU, and northern EU and ship emissions from the Atlantic, Baltic Sea, and Mediterranean. The LB source is the principal contributor to the modelled mean O 3 mixing ratios in every receptor region. The contributions peak in May (mean absolute contribution 25 ppbv), reflecting the seasonal cycling in the northern hemispheric background O 3 (e.g. Monks, 2000;Pilling et al., 2009). Contributions from this source are more prominent in the regions located in the north, east, and north-west of the UK, e.g. Scotland (30 ppbv), Northern Ireland (28 ppbv), North East (27 ppbv), the North West (26 ppbv), and Wales (26 ppbv). These regions contain about 20 % of the UK population and are primarily impacted by westerly flows and associated hemispheric O 3 background due to their geographical posi-tion (Pilling et al., 2009). Also, they generally experience less than 10 d with O 3 concentrations above the EU limit of 120 µg m −3 (DEFRA, 2020) because of low NO x emissions locally. The contributions from the LB source in the South East, East Anglia, and East Midlands, on the other hand, can be up to 8 ppbv smaller than in the east of the UK, particularly during summertime; see Figs. S10-S11 in the Sect. S2 in the Supplement.
The UK contributions are generally more significant in the east, south-east, and Midlands, showing a maximum value in June and July in every receptor area (Figs. S10 and S11 in the Supplement). The source region provides up to 20 % of the surface O 3 mixing ratios in East Anglia, 18 % in the London area and East Midlands, and 16 % in Yorkshire and Humberside and the South East, making it the second-biggest source of O 3 in these locations after the LB. This area incorporates about 50 % of the UK population and often experiences more than 10 d with O 3 concentrations above the EU and UK O 3 threshold (concentration >120 and 100 µg m −3 , respectively) (DEFRA, 2020). The Eu super-region, on the other hand, is the second-largest source region in the northern and western UK, with contributions in summertime reaching up to 10 % in Scotland and 16 % in the South West, England, and 14 % in Wales. Regardless, this source region still significantly impacts the South East, East Anglia, and London, where the relative contributions can increase from 13 %, 12 %, and 13 % in May to 16 %, 15 %, and 16 %, respectively, in July.
The contributions from ship emissions from the North Sea and English Channel are significantly lower than those from UK sources and the Eu super-region (3 %-4 % of the total contribution in the South East and East Anglia in May and up to 6 % of the total surface O 3 during the summer months). The impact is also less important in the west than in the east and south of the UK due to the proximity with the region. As for the relative contributions from the different Eu subregions (inner circle in Fig. 9), these are largely influenced by the geographical situation of the receptors and the predominant wind direction. In every receptor, the principal contributor from the Eu super-region is the Rest_Eu source region, providing between 60 %-70 % of the Eu super-region O 3 in May and up to 83 % of the O 3 during summertime. The relative contributions of the Rest_Eu region are larger in the northern and western locations, in particular during the summer months when there is a marked difference in the distribution of the contributions across the UK regions. The summer months see an increase in the input from France, Germany, and the Benelux region, in particular during anticyclonic weather conditions and over the receptor regions located in the south and east of the UK (e.g. the South East, East Anglia, the London area, and the East Midlands); see Figs. S10-S11 in the Supplement. This is consistent with results of studies on extreme O 3 in the EU and the UK reporting an increase in surface O 3 concentrations under anticyclonic conditions (e.g. Pope et al., 2016;Ordóñez et al., 2017;Romero-Alvarez et al., 2022). Romero-Alvarez et al. (2022), in particular, have shown that a wide area of high pressure centred over the Netherlands coast affected most of England during the first days of July 2015. During the same period, regions such as East Anglia reported increases in O 3 mixing ratios of up to 16.6 ppbv h −1 that overlapped with wind direction changes from south-south-west to south-south-east.
Depending on the predominance of the wind direction (south-south-east and south-south-west), O 3 from anthropogenic sources within France can impact both the west and the east of the UK. The contribution is greater in the southern UK due to the proximity to the source region. The contributions from the Benelux region and Germany are more significant in the east of the UK due to the proximity with the continent and association with easterly flows (east and south-east); about 14 % and 6 % of the Eu super-region in East Anglia during the summer months come from these two source regions, respectively.
Notably, anticyclonic conditions and easterly winds in the UK have been associated with enhanced O 3 concentrations, whereas cyclonic conditions and westerly winds have been linked to O 3 transport from the UK mainland and cleaner air from the North Atlantic (Jenkin et al., 2002;Pope et al., 2016;Romero-Alvarez et al., 2022). The contribution patterns described above may thus serve as predictors of future O 3 source apportionment over the UK regions.

MDA8 O 3 exceeding 50 ppbv
The mean contribution from each source region for the hours when the MDA8 O 3 exceeded 50 ppbv at each receptor area from May to August is presented in Fig. 10. The figure shows large contributions from source regions that were not seen as dominant sources. France, for example, becomes a major source, particularly in receptors in densely populated areas such as the south and east of the UK. The absolute mean contributions at the sites sometimes exceed the input from the LB O 3 (mean value ranging between 10 and 15 ppbv and maximum reaching up to 35 ppbv in the London area). The impact of UK NO x on O 3 varies across the sites, but in general, its share increased from the south-east to the north. In the Midlands, the North East, the North West, Scotland, and Yorkshire and Humberside, O 3 from UK sources becomes dominant, surpassing the LB mean input in most receptors. In the remaining locations, the UK source is the thirdlargest input for surface O 3 except for the South West, where most of the O 3 comes from France (mean ∼ 18 ppbv), the LB (mean ∼ 14 ppbv), NOS (mean ∼ 6 ppbv), and Rest_Eu (mean ∼ 8 ppbv). The impact from the shipping component (NOS) also becomes important in all receptor regions with an estimated mean of 4-7 ppbv. O 3 from central Eu, Germany, the Netherlands, Belgium, and Luxembourg, on the other hand, is almost negligible in the west of the UK (mean less than 1 ppbv). However, its impact increases towards the east and north with mean values ranging about 1-6 ppbv (e.g. in the East Midlands, the North East, Yorkshire and Humberside, and Scotland), reflecting the efficient transport of pollutant-loaded air masses under anticyclonic conditions. Figure 11 shows the mean contribution at each receptor area to hourly surface O 3 when the MDA8 O 3 exceeded the 60 ppbv thresholds. There were two occasions when the modelled MDA8 exceeded 60 ppbv, the main occasion being the episode on 1 July. Romero-Alvarez et al. (2022) have shown that MDA8 O 3 above 50 ppbv in the South East and East Anglia regions in July 2015 coincided with days when easterly winds prevailed (east-south-east flows). In contrast, MDA8 O 3 above 60 ppbv coincided with a shift in the wind direction from east-south-east to south-south-east and south and a sharp rise in the surface temperature.

MDA8 O 3 exceeding 60 ppbv
France was the most significant contributor to the buildup of O 3 when the mixing ratios exceeded the EU threshold in the South East (mean ∼ 18 ppbv), East Anglia (mean ∼ 21 ppbv), and the London area (mean ∼ 26 ppbv) because convergence of westerly and south-easterly winds in the west of the UK diverted the contributions of domestic sources from these regions, as reported in Romero-Alvarez et al. (2022). O 3 from UK NO x emissions, on the other hand, has a greater impact on the East Midlands (mean ∼ 16 ppbv) and Yorkshire and Humberside (mean ∼ 15 ppbv). In the South East and the London area, the contributions from Rest_Eu equal those from UK O 3 , while the influence is comparable to that from the west and central Europe in the rest of the regions. As in the contributions to the MDA8 O 3 threshold of 50 ppbv above, the lateral-boundary component remained nearly constant in all receptor areas with a mean contribution of about 12 ppbv. This is because most of the UK's weather was dominated by anticyclonic conditions. The impacts from the North Sea and the English Channel are also important in all receptor regions, with a mean between 4-7 ppbv. Results suggest that ship emissions along these routes affect the air quality of the UK, particularly over the east and south-east. However, the current model configuration does not consider the chemical evolution of the different emitted species (chemical loss and production rates) during the dispersion of the ship plume. In fact, once species are emitted, they are instantaneously mixed in each model grid cell (27 km × 27 km). In the case of chemically reactive species such as NO x , this can lead to overestimations of both NO x and O 3 concentrations due to the non-linearity of the chemical processes involving NO x and O 3 evolution during the dispersion of the ship plume (e.g. Huszar et al., 2010;Van Der Werf et al., 2010).

AOT40 index
The average simulated AOT40 for 2015 in the two most relevant arable areas in the UK -East Anglia and the South East -is 3674 and 1833 µg m −3 h −1 , respectively (the target value for the EU and UK is ∼ 18 000 µg m −3 h −1 ). Figure 12 shows the source contributions to the surface O 3 when the mixing ratio exceeded 40 ppbv during the daytime hours (08:00 and 20:00) central European time from May to July in two receptor regions in the UK. When exceedances to the hourly surface O 3 mixing ratios above 40 ppbv are considered, the LB component becomes the dominant source in both receptor regions (estimated mean concentration between 21-24 ppbv) as its threshold is close to the tropospheric baseline ozone level associated with maritime North Atlantic air masses. The second-largest contributor is the UK with a higher impact in the East Anglia region (estimated mean concentration 10 ppbv) than in the South East (estimated mean concentration 6 ppbv). The third, fourth, and fifth contributions in East Anglia come from the Rest_Eu and Rest_CEu regions and France, while in the South East the contributions come from the North Sea and English Channel, Rest_Eu, and Germany. The contributions from the Netherlands, Belgium, and Luxembourg become almost negligible.
Note that the AOT40 metric assesses the impacts of O 3 on the vegetation by considering an O 3 threshold (e.g. concentrations above 40 ppbv) during the months when plant growth is most likely to be affected and when daytime O 3 concentrations are at their highest. However, research experiments have shown that the response of plants to O 3 exposure is nonlinear due to a mismatch between the peak daytime O 3 concentrations and stomatal opening (Heath et al., 2009). This means that the effective amount of O 3 taken up by plants is not always correlated to the ambient O 3 concentrations. The AOT40 index does not account for the plant's physiological control of stomatal opening, which limits the potential of the index to accurately assess the impacts of O 3 on the vegetation. Future work should consider flux-based metrics, which have proven to be more suitable for ozone risk assessment on plants as they take into account the ambient concentration of O 3 , the physiological control on stomatal openings, and the efficiency of the leaf antioxidant system (Fares et al., 2010) as well as improved the O 3 deposition routines in WRF-Chem so that they can take into account factors such as stomatal opening cycles.

Conclusions
An O 3 tagging technique within the WRF-Chem model was used to investigate the origin of surface O 3 from May to August 2015 and the contribution of different source regions to O 3 regulatory metrics in the UK. Evaluation against observations presented in the Supplement has shown that the model setup gives a good representation of O 3 in the European domain. Domain-wide examination demonstrates that the hemispheric O 3 , here represented by lateral-boundary O 3 , has the largest impact on the concentrations of O 3 in the UK, with an estimated 71 % of the modelled monthly mean surface O 3 coming from this source region. About 16 % of modelled surface O 3 is produced from anthropogenic NO x emissions within the EU that contain lumped NO x emissions from continental Europe; the Republic of Ireland; and ship emissions in the Atlantic, North Sea, Baltic Sea, and Mediterranean. UK emissions (England, Scotland, Wales, and Northern Ireland) contributed 13 %. Assessment of the contributions to different receptor regions in the UK revealed that the UK relative contribution to UK surface O 3 tends to be higher in June and July with a marked spatial gradient, with high mixing ratios obtained in the south-east and lower values in the north and west. In fact, UK NO x emissions are the second-largest contributor to surface O 3 in the East Midlands, the West Midlands, Yorkshire and Humberside, East Anglia, the South East, and the London area after the lateral-boundary source region. The monthly and spatial variation in the contribution of UK NO x emissions to UK surface O 3 is primarily caused by larger photochemical activity taking place during the summer months in the south and downwind of emission sources. Similarly, the absolute contribution from European sources to UK surface O 3 tends to be higher in June and July and along much of the eastern, southern, and south-west borders, reflecting the effective transport of continental O 3 by southeasterly winds during O 3 pollution events. The tagging technique also shows that O 3 from this region is generally transported to the UK rather than produced in situ.
O 3 tagging has also made it possible to demonstrate that more stringent emission controls would be required in different source regions for compliance with UK and EU O 3 standards, e.g. MDA8 O 3 of 50 and 60 ppbv. Emission controls in France, in particular, would significantly reduce O 3 concentrations in densely populated areas such as the South East, the South West, and East Anglia, while domestic emission controls are more relevant for the Midlands and the north of the UK. Exposure thresholds, such as those considered in the AOT40 O 3 metric, are instead most affected by lateral-boundary components in the first place followed Figure 11. Hourly contributions (in ppbv) to surface O 3 at 5 UK receptor regions from 10 source regions (UK, background (LB), the Netherlands, Luxembourg, Belgium, France, Germany, the rest of central Europe (Rest_CEu), the North Sea and English Channel (NOS), and the rest of Eu) during days when the MDA8 is above 60 ppbv between May and August. The lower and upper ends of the boxes indicate the 25th and 75th percentiles, the dashed black line the median, the white boxes the mean, and the whiskers the minima and maxima.

Figure 12.
Hourly contributions (in ppbv) to surface O 3 at 2 UK receptor regions from 10 source regions (UK, background (LB), the Netherlands, Luxembourg, Belgium, France, Germany, the rest of central Europe, the North Sea and English Channel (NOS), and the rest of Eu) for the AOT40 between May and August. The lower and upper ends of the boxes indicate the 25th and 75th percentiles, the dashed black line the median, the white boxes the mean, and the whiskers the minima and maxima. by UK NO x emissions. Emission controls in regions such as the East Midlands, the West Midlands, Yorkshire and Humberside, East Anglia, the South East, and the London area will aid in the mitigation of the impacts on crops. Nonetheless, emission controls will also be necessary over the larger Northern Hemisphere area.
The results from model simulation should be interpreted in the context of the observed bias in O 3 (domain mean bias (MB) = −3.71 µg m −3 ) and the underestimation of the number of days with MDA8 O 3 above 50 and 60 ppbv and the AOT40 metric. Also, controlling emissions of NO would not necessarily translate into a reduction in O 3 concentration in the UK. In fact, for the AOT40 measure in rural regions it seems that reducing UK emissions might well still help improve the situation, whereas for urban regions, reducing NO x will increase O 3 concentrations due to a reduction in the titration effect. In this regard, future work should consider extending the tagging mechanism to include the competing NO x -VOC interactions in O 3 production. Emission perturbation studies might also complement the investigation by adding an understanding of response of O 3 to different emission control scenarios.
Author contributions. JRA designed the study, conducted the numerical simulations, and analysed the data. The original tagging method was developed by AL and TB and adapted by JRA for this study. DL and SAN developed the emissions pre-processor. JRA wrote the paper under the supervision of CER, SD, and AB, with contribution from all authors.