. Investigating the background and local contribution of the oxidants in London and Bangkok. Faraday

The contribution of NO x emissions and background O 3 to the sources and partitioning of the oxidants [OX ( ¼ O 3 + NO 2 )] at the Marylebone Road site in London during the 2000s and 2010s has been investigated to see the impact of the control measures or technology changes inline with the London Mayor ’ s Air Quality Strategy. The abatement of the pollution emissions has an impact on the trends of local and background oxidants, [OX] L and [OX] B , decreasing by 1.4% per year and 0.4% per year, respectively from 2000 to 2019. We also extend our study to three roadside sites (Din Daeng, Thonburi and Chokchai) in another megacity, Bangkok, to compare [OX] L and [OX] B and their behavioural changes with respect to the Marylebone Road site. [OX] L and [OX] B at the Marylebone Road site (0.21[NO x ] and 32 ppbv) are comparable with the roadside sites of Thailand (0.12[NO x ] to 0.26[NO x ] and 29 to 32 ppbv). The seasonal variation of [OX] B levels displays a spring maximum for London, which is due to the higher northern hemispheric ozone baseline, but a maximum during the dry season is found for Bangkok which is likely due to regional-scale long-range transport from the Asian continent. The diurnal variations of [OX] L for both London and Bangkok roadside sites con ﬁ rm the dominance of the oxidants from road transport emissions, which are found to be higher throughout the daytime. WRF-Chem-CRI model simulations of the


Introduction
Nitrogen dioxide (NO 2 ) and ozone (O 3 ) are key urban air pollutants with welldocumented public health impacts. 1 It is well established that the chemistry of O 3 , NO 2 and nitric oxide (NO) is strongly coupled by the following null cycle, typically on the timescale of a few minutes under most ambient conditions. 2,3 + O 3 / NO 2 + O 2 (1) Because of this rapid interconversion, the term "oxidant" (hereaer denoted OX) has sometimes been used as a collective term for NO 2 and O 3 , 4,5 similarly to NO x being the well-established collective term for NO and NO 2 .Reactions (1) and (2) therefore partition NO x between its component forms of NO and NO 2 , and OX between its component forms of O 3 and NO 2 , but conserves both NO x and OX.
Reported analyses of ambient data have shown that the concentration of OX at a given location, [OX], can be described in terms of the combination of a background (NO x -independent) source and a local (NO x -dependent) source, denoted here as [OX] B and [OX] L , respectively, 5,6 as also apparent from the example data shown in Fig. 1.
[OX] L is believed to be mainly derived from primary emissions of NO 2 , particularly at roadside and kerbside locations, such that the slope of the [OX] vs.
[NO x ] relationship provides an estimate of the volumetric fraction of NO x emitted as NO 2 . 6,7[OX] B provides a quantication of the background [O 3 ] which would exist at the given location in the notional absence of NO x .It can therefore be regarded as the global (hemispheric) baseline O 3 level, modied by regional-scale processes (i.e.deposition and chemistry) that can either remove or produce OX.Analyses of this type have been carried out in a number of studies, 5,6,8,9 with values of [OX] B and [OX] L reported for a variety of locations and time periods.
4][15] The combined health effect of NO 2 and O 3 has been reported to be elevated compared with their individual effects, which has led to interest in the processes controlling OX levels in the urban areas. 16,17A recent study showed that the OX can enhance the chronic health risks of ne particulate matter (PM 2.5 ), 18 thus the reduction of OX levels could be benecial to reduce the public health impacts of PM 2. 5 .
Currently there are 33 megacities (cities with population of more than 10 million) declared by United Nations in 2018 and the number is likely to increase to 43 by 2030. 19London is one of the megacities in the world where air quality is an issue of increasing public concern.For example, levels of NO 2 at roadside and kerbside locations in London have exceeded the EU annual mean limit value of 40 mg m À3 on many occasions. 202][23] We therefore investigate the time dependence of these oxidant components at the London Marylebone Road kerbside site, over the time period 2000-2019, to examine the effects of the control measures.
Thailand is one of the rapid economically developing and urbanised countries, which has led to the emergence of a megacity (e.g.Bangkok) and concomitant problems of poor air quality.The transportation, electricity generation and manufacturing industries are considered to be the major sources of air pollutants in Bangkok. 24A recent study showed that NO x levels in Bangkok are mostly below the National Ambient Air Quality Standards (NAAQS), but, O 3 levels in Bangkok frequently exceed the NAAQS. 25Uttamang et al. 25 analysed the Bangkok metropolitan region pollutant data for 2010-2014 and found that both local and background OX contributions are responsible for increasing the concentration of O 3 .We update the work of Uttamang et al. 25 using more calendar year-coverage data (2005-2018), highlighting roadside and kerbside locations in Bangkok and compare the results with those of the Marylebone Road site in London to investigate any systematic differences in the magnitudes and time dependences of the oxidant components for the two different scenarios (midaltitude vs. tropics).
Megacities are immense sources of air pollutants, with large impacts on regional and global atmospheric chemistry. 26In addition to local sources, ozone chemiluminescence is used for NO x measurements) as described by DEFRA. 30owever, it is important to note that these analysers are subject to potential interferences from several other oxidised nitrogen species, which can be erroneously reported as NO 2 . 31,32Despite this, these species are only likely to have a signicant impact at rural or remote locations where NO x concentrations are reduced and so-called NO z components abound, due to their formation from NO x processing (e.g.PANs).Thus, [NO z ]/[NO x ] increases with chemical processing time, so the interferences are not important to consider at roadside or kerbside locations.The uncertainty (expressed at a 95% condence level) of the measured data for NO x and O 3 was around 15%. 33 Measurements for these species are reported at hourly resolution and are openly available from the U.K. Air Information Resource (http://uk-air.defra.gov.uk/data/data_selector).This study considers the hourly mean OX and NO x data for the years 2000-2019 inclusive where data has been divided by hour of the day for each month of the year.The data coverage at this site is reasonably consistent except 2011 with the majority of months showing well over the 80% coverage required for inclusion in this study.
For the Bangkok roadside sites, NO and NO 2 measurements were performed using chemiluminescence detection (Thermo Scientic 42i).The working range of the instrument is 0-500 ppb to 0-20 ppm with a minimum detection limit <0.5 ppb and the precision of the instrument is 0.5 ppb (<1%).O 3 was analysed by using UV absorption photometry detection spectroscopy (Thermo Scientic 49i).The working range of the instrument is 0-500 ppb to 0-10 ppm and the instrument has a detection limit of <0.6 ppb and with a precision of <1%.The single point calibration for detectors was performed every 15 days and the multi-point calibration with 3 span levels (20%, 40% and 80%) was performed every 90 days.The data was accepted with the span dri better than AE10% of full scale for the NO 2 detector and AE10% of full scale for the O 3 detector and zero dri better

Background and local OX calculations
Using the approach of Jenkin, 6,7 the 'background' and 'local' contributions to the total oxidant, [OX] B and [OX] L , are estimated from a linear t of [OX] against [NO x ].The annual NO x mixing ratios vary over a wide range of values with diurnal and seasonal variations.Thus, we used the hourly data of NO x and OX for each measurement site in London and Bangkok to generate the [NO x ] vs. [OX] plots on a month-by-month basis.From here, statistics for the linear relationship were calculated using the 'least squares' method to give exact values from gradient (local NO x -dependent contribution), intercept (background NO x -independent contribution) and standard errors for both values.This data was compiled and averaged accordingly to give insights into the monthly and hourly dependencies of [OX] B and [OX] L and how their contributions and trends vary over time.

Modelling
We used a mesoscale non-hydrostatic 3-D meteorological model, WRF-Chem-CRI, which is an online fully coupled chemistry transport model. 34The chemistry and aerosol components, along with the prognostic meteorological variables, are integrated over the same timestep as the transport processes, using the same advection and physical parameterisations. 34The meteorological, physical and chemical parameterizations of the model are described in detail elsewhere. 28,35he meteorology was driven by the European Centre for Medium Range Weather Forecasts (ECMWF) ERA-Interim reanalysis data. 36The chemical boundary conditions were provided from MOZART-4 global model. 37Biogenic emissions were calculated online by the Model of Emissions of Gases and Aerosols from Nature (MEGAN) (https://www2.acom.ucar.edu/modeling/model-emissionsgases-and-aerosols-nature-megan). 38The MEGAN model calculates the biogenic emissions of the species from terrestrial ecosystems with a resolution of 1 km Â 1 km and is driven by land cover by vegetation, environmental factors (e.g.temperature, humidity, solar intensity, soil moisture) and atmospheric chemical composition.The anthropogenic emissions for the London study have been sourced from a combination of the UK National Atmospheric Emissions Inventory (NAEI) (https://naei.beis.gov.uk) and the European The Netherlands Organisation for Applied Scientic Research, Monitoring Atmospheric Composition and Climate's emission inventory (TNO-MACC-IIIT). 39The NAEI data for the year 2012 with a resolution of 1 km Â 1 km was used in this study.The NAEI includes seven direct greenhouse gases addressed at the Kyoto summit, as well as other trace gases believed to be of atmospheric signicance: NO x , CO, non-methane volatile organic compounds (NMVOCs) and SO 2 .Where data from the NAEI was not available, the TNO was employed to ll the gaps.The TNO data with a resolution of 0.125 Â 0.0625 for the year 2011 was used in the study.Scaling factors based on those built for the EMEP model are applied to the anthropogenic emission inventories in order to account for daily, weekly and monthly differences. 40The anthropogenic emissions for the Bangkok study were extracted from the global Emission Database for Global Atmospheric Research with Task Force on Hemispheric Transport of Air Pollution (EDGAR v4.3.2) emission inventories for 2010 at 0.1 Â 0.1 resolution. 41,42These datasets are provided as monthly means giving the seasonal cycles.Daily activity cycles, and speciation of NMVOC emissions (see ESI, Table S1 †), have been imposed (see details in Chen et al. 43 and Lowe). 44The chemistry scheme used is CRI (Common Representatives Intermediate Mechanism), a chemical mechanism describing gas-phase transformations of methane and 25 NMVOCs, via 232 chemical species participating in 638 simulated reactions. 28he model domain for the London study covers North-West Europe with a 15 km horizontal resolution with a size of 134 (E-W) by 146 (N-S) grid cells and 41 vertical levels with enhanced resolution within the planetary boundary layer.Two scenarios were run using this domain from 00:00 UTC on 30 July 2012 to 00:00 UTC 24 August 2012 and from 00:00 UTC on 8 January 2012 to 00:00 UTC on 12 February 2012 covering summer and winter months, respectively.Because of the long time period in the scenarios, the re-initialisation of the model meteorology was performed every 3 days using ECMWF meteorology to minimize the errors in the modelled meteorology (bias from model to actual).The Bangkok study region included Thailand, Vietnam, Laos and Cambodia (8 N-22 N and 94 E-108 E, see ESI Fig. S2 †) centred at 15 N and 101 E with 15 km horizontal resolution and 41 vertical levels.Three scenarios were run using this domain for the periods of 1-7 January 2017 (winter season), 1-7 May 2017 (summer season) and 1-7 September 2017 (rainy season) with 2 additional days for spin-up.oxidants is lower (e.g.14.5 AE 5.0 ppb (31% to the total OX, Fig. 3b) for Din Daeng; 12.5 AE 6.0 ppb (29% to the total OX, Fig. 3d) for Chokchai and 14.1 AE 6.0 ppb (32% to the total OX, Fig. 3f) for Thonburi) than the Marylebone Road.The aspect ratio (ratio of building height to street width, H/W) comparison (Marylebone Road $1.0, 46 Bangkok roadside sites $0.4 (ref.47)) suggests that street canyons are much deeper in the Marylebone Road than that in the roadside sites of Bangkok.Thus the street canyons in Marylebone Road are not as well ventilated as the comparatively open locations in the Bangkok roadside sites, which is a possible reason for increased [OX] L in the Marylebone Road. 48,49However, the pollution levels dened by the fraction of NO x emitted as NO 2 in Bangkok roadside sites are highly variable with lower in Din Daeng compared with Thonburi and Chokchai roadside sites (ESI Fig. S4b †).This might be explained by the mobile sources (e.g.vehicles) in Bangkok metropolitan region, 25 which is highly dependent on traffic density and traffic ow.The NO x emissions from vehicles are dependent on fuel type and speed of the vehicles, 50 which can be responsible for the variation of the [OX] L for different roadside sites in Bangkok.

Results and discussion
For the Marylebone Road, decreasing trends of both [OX] L and [OX] B levels, by 0.33 ppb per year (1.4% per year, p < 0.001) and 0.11 ppb per year (0.4% per year, p < 0.001), respectively, have been found for the period of 2000 to 2019 (Fig. 2a).The

Paper Faraday Discussions
This journal is © The Royal Society of Chemistry 2020 Faraday Discuss.
local D[OX] has been found to have an upward trend (2.12 ppb per year, 8.9% per year, p < 0.001) for the 2000s, but a downward trend (1.17 ppb per year, 5.4% per year, p < 0.001) for the 2010s (see ESI Fig. S5 †).An increasing trend of [OX] L from 2002-2005 is seen, which can be explained by the absolute increase in the percentage of NO x emitted from vehicles as NO 2 (Fig. S4a †).In response to the Euro III control in the U.K., we found an increase of 5  21,52 On-road emissions from Euro IV heavy vehicles and buses showed a reduction in the primary NO 2 fraction resulting in NO 2 emissions from 2010 to 2014 in London. 53,54Furthermore, it has also been shown that the emissions from aged Euro 4 and 5 diesel cars reduced the ratio of NO 2 /NO x . 53These control strategies could be the main reasons to see a decrease of 1.5% per year NO x emitted as NO 2 from 2010 to 2019.This is consistent with the study of Font et al. 55 and this NO 2 decrease is likely to contribute to the decrement of [OX] L in the 2010s.It can be concluded that the adopted emission control in recent years has been successful in the abatement of pollutant emissions, which reduced [OX] L levels at a rate of 5.4% per year from 27.2 ppb (2000s) to 15.6 ppb (2010s).The [OX] B in London is strongly dependent on the mixing ratios of O 3 on the northern hemispheric ozone baseline. 7,56We found very similar decreasing trends of [OX] B in the 2010s as shown for northern hemispheric mid-latitude baseline O 3 by Derwent et al. 57 The relatively constant northern hemispheric ozone baseline during the 2000s is reported in the Derwent et al. 57 study, but we found a higher downward trend of the D[OX] B with 0.55 ppb per year (1.6%, p < 0.001) in the 2000s compared with 0.08 ppb per year (0.3%, p ¼ 0.05) in the 2010s.The O 3 can be formed from regional-scale photochemical processing of VOCs and NO x over north-west Europe during the anti-cyclonic heat-wave period 58 and which can then be transported to the south and east of the U.K. making an additional contribution of [OX] B at the Marylebone Road.The signicant reductions of the European anthropogenic emissions of NO x , VOCs in response to precursor emissions controls (e.g.1999 Gothenburg Protocol) from 2000 to 2009 can be responsible for reducing regional photochemical ozone production. 59This is likely to be the main reason for the decreasing [OX] B trend during the 2000s.
For Bangkok roadside sites, similar downward trends for [OX] L levels are found in Din Daeng with 0.44 ppb per year (2.8% per year, p ¼ 0.001), in Thonburi with 0.69 ppb per year (4.6%, p < 0.001) and in Chokchai with 0.26 ppb per year (2.0%, p ¼ 0.004) (Fig. 3a, c and e).From 2007-2015, the registered vehicle numbers in Bangkok metropolitan area increased from 6.2 million to approximately 10 million, but the annual local pollutants (e.g.NO x , CO, NMVOCs) emissions from vehicles have signicantly decreased due to the policy implemented on road transport activities. 60Thailand has subscribed to the Euro systems of light duty vehicle emissions standard since 1997, but have adopted the more intensive enforcement of fuel and engine standard levels for light duty vehicles from the Euro 2 standard level in 1999 to the Euro 4 standard level in 2013 (with a stricter limit on NO x and VOCs emissions from gasoline engines and on CO, NO x , VOCs + NO x and PM emissions from diesel engines adopted). 61These policies have had some impact on decreasing NO x emitted as NO 2 by 1.8% per year (Din Daeng; 2005-2014), 0.6% per year (Chokchai; 2005-2014), and 1.0% per year (Thonburi; 2005-2018) (see Fig. S4b †).
For [OX] B levels, the upward trends are found in Din Daeng (0.04 ppb per year, 0.1% per year, p ¼ 0.18) and in Thonburi (0.97 ppb per year, 2.9%, p ¼ 0.19), but the changes are not signicant in the 95% condence interval (Fig. 3a and e).However, a downward trend of [OX] B is seen in Chokchai (0.77 ppb per year, 2.6% per year, p < 0.001) (Fig. 3c).Although the NO x emissions have decreased since 2011, the O 3 concentration increases in the Asian continental air are still significant due to the continued increase in VOC emissions. 62Thus, the decreasing local NO x emissions are not the only effective way to reduce the oxidant levels due to the regional air pollution transport, which can increase the [OX] B levels in Bangkok.The different [OX] B trends for London and Bangkok roadside sites can be explained by meteorological variability and hemispheric transport of pollution. 63

Seasonal variation of local and background OX
For the Marylebone Road, the [OX] B has an annual cycle with a maximum in May (41 ppb), inuenced by the spring maximum northern hemispheric ozone baseline and a minimum in November (27 ppb) (Fig. 4a).More generally, high NO x is associated with low boundary layer heights due to weak convective activities and turbulent mixing in the winter, which can increase the loss of O 3 and NO 2 by deposition to the surface, thereby reducing [OX] B in winter months.The previous study at the Marylebone Road showed the seasonal variation of [OX] B with a maximum during spring (43 ppb in April) and a minimum during autumn (30 ppb in October). 7From the 2000s to the 2010s, the spring [OX] B is reduced by $1 ppb but the summer [OX] B is reduced by $3 ppb (see Fig. 4a) which could be explained by the reduced photochemical O 3 production as a result of reduced European regional O 3 precursor emissions.The number of extreme heat waves in Europe is fewer in the 2010s than in the 2000s. 64In summer, heat waves are

Paper Faraday Discussions
This journal is © The Royal Society of Chemistry 2020 Faraday Discuss.
common and [OX] B levels are expected to increase due to an increase in regionalscale photochemical events. 56However, the signicant control of O 3 precursor emissions in recent years decreased the [OX] B formation from European polluted air mass arriving at Marylebone Road.This conrms the success of the ozone precursor emission control strategies in the abatement of oxidant emissions/ production.For [OX] L , the seasonal variability will depend on the level of the primary pollutants, NO 2 and the prevalent local photochemistry, thus a peak of [OX] L is found during summer months (Fig. 4b).The contribution of [OX] L is found to be higher during the 2000s compared with the 2010s which can be explained by the extra emissions of NO 2 due to the introduction of EURO III light duty diesel vehicles (tted with oxidation catalysts) into the vehicle eet and the tting of catalytically regenerative particle traps to bus eets. 30Signicantly high [OX] L is seen in the 2000s during winter (when a shallow inversion layer is formed) which is most likely due to the combination of extra emissions of NO 2 and stagnant air.
For Bangkok, the seasonal variations of [OX] B and [OX] L (Fig. 5) show that high mixing ratios were found during dry months (November to April) with low mixing ratios during wet months (June to September).The differences in the meteorological conditions (solar radiation, temperature, cloud cover) between wet and dry seasons affect the photochemical production of OX.The dry season is linked to a more stable atmosphere and less intensive wet removal of oxidants 65 resulting in higher [OX] L and [OX] B levels.The long-range transport from the Asian continent can contribute to [OX] B concentrations in Thailand, 66 which plays an important role in seasonal uctuations of OX in Bangkok. 67There is a slight increase of [OX] B during April-May (Fig. 5a), which could be due to the transport of ozone precursors from the burning region. 68

Diurnal variation of local and background OX
The diurnal cycles of [OX] B in London and Bangkok roadside sites are found to reach a maximum during the aernoon (Fig. 6 and 7) which is attributed to a deepening boundary layer causing entrainment of ozone-rich air from alo.The diurnal cycle is very consistent with that in similar studies for London 7 and Bangkok 25 roadside sites.
The diurnal variation of [OX] L is related to transport emissions of NO 2 .Thus, the levels of [OX] L at the Marylebone Road and Bangkok roadside sites are high throughout the daytime (Fig. 6 and 7).The diesel NO x emissions peak during the

Faraday Discussions Paper
Faraday Discuss.This journal is © The Royal Society of Chemistry 2020 middle of the day. 69The NO 2 /NO x emission ratio is high for diesel engines which can increase even further with lower vehicle speeds. 30The high [OX] L in the Marylebone Road during the 2000s can be explained by the introduction of EURO III light duty diesel vehicles (tted with oxidation catalysts) into the vehicle eet.
Thonburi is an industrial area, so the frequent movement of diesel engine heavy vehicles in this area can emit a signicant amount of NO x as NO 2 resulting in higher [OX] L levels compared with Din Daeng and Chokchai.The traffic at both Din Daeng and Chokchai throughout the day is busy, but, due to the close proximity of a bus stop at the Chokchai monitoring site, buses regularly stop and idle which could be responsible for enhanced [OX] L levels compared with that in Din Daeng.Aer sunset, the formation of [OX] L at both London and Bangkok roadside sites is inhibited and the planetary boundary layer becomes more stable resulting in [OX] L reduction through chemical reaction (e.g.O 3 + NO 2 / NO 3 + O 2 ) and physical processes (e.g.dry deposition). 70Since NO 3 is building up in the night-time and rapidly photolysed during the day, this process regenerates local oxidants in early daylight hours.

Modelling of OX
We simulated model [OX ¼ (NO 2 + O 3 )] mixing ratios and compared with the measurements at the ve ground sites in the centre of London.The results showed a general trend in peak [OX] which is well represented during summer months (Fig. 8).The correlation coefficients for the Westminster, Harlington, Eltham and Bloomsbury sites are high (r $ 0.8) suggesting that the model captures the [OX] trend very well.The model-measurement agreement is also reasonably good as the biases and root mean square errors (RMSE) for all these  sites are within $10 ppb (Table 2).The Westminster and Bloomsbury sites are strongly inuenced by local traffic, which cause increased biases and RMSEs compared with the Eltham and Harlington sites.The small correlation coefficient (r ¼ 0.4) and high bias (À36 ppb) and RMSE (40 ppb) for the Marylebone Road suggests that the model agreement with measurement for this site is relatively poor compared with other London sites.This is likely to be a consequence of the monitoring station being in close proximity to a busy six-lane road in a street canyon, which regularly experiences high traffic volumes and prolonged periods of congestion.The site is subject to a well-characterized canyon vortex leading to concentrations dominated by wind direction.Although running at a relatively ne resolution of 15 km by 15 km, WRF-Chem-CRI is unable to pick up the local sources and sub grid-scale chemical dynamics, which drive the air quality at street level sites.Dispersion models such as ADMS take time-varying local sources (such as traffic at street level) into account, as well as incorporating features such as building, complex terrain, which alter the dynamics and thus composition of the surrounding atmosphere.Overall, the model is generally able to reproduce the [OX] levels in London to a reasonable extent and provide an accurate representation of air quality in complex urban environments.
There are many discrepancies in the [OX] peaks during winter months.The NO x emissions from vehicles are higher in winter months compared with summer months, thus the increased emissions fraction of NO x in the form of NO 2 from vehicles can have a direct local-scale impact on the levels of [OX].Except the Eltham site (a suburban site which experiences very low traffic volumes compared with other sites), the model-measurement agreements for all other sites have poor correlation coefficients, biases and RMSEs during winter months compared with summer months.This is likely to be a result of the resolution of the model, which is not enough to resolve the local effects (e.g.fraction of NO x emitted as NO 2 from vehicle).In the model, the local sources are spread out across the grid cell, resulting in diffusion of the emissions and leading to lower simulated mixing ratios of [OX].There have also been studies which have found signicant underestimations (30-40%) of traffic-related NO 2 emissions in the inventory data, NAEI, used by WRF-Chem-CRI. 71As traffic (diesel vehicles in particular) is a signicant source of NO 2 in urban environments such as London, the observed underestimation in modelled [OX] mixing ratios seen at London sites is therefore likely to be a consequence of the underestimation of traffic-related emissions in the inventory data used in the model.WRF-Chem-CRI is able to reproduce the diurnal variations of background [OX] at London sites, with good agreement between mixing ratios in the modelled and measured data (biases/RMSE < 10 ppb and correlation coefficient > 0.5) (Fig. 9 and Table 3).In general, the [OX] B is underestimated by the model, which is likely due to the under-prediction of the diesel related VOCs used in the model. 72Furthermore, the VOCs used in the model were not constrained using the AURN dataset which can partly contribute the imbalances between model-measurement.However, there are times when peak [OX] B is overestimated by the model (e.g.summer time in Westminster, Bloomsbury and Eltham), which could be due to the regional photochemical events occurring below the resolution of the model.Westminster and Bloomsbury are urban background sites, but they are very close (within 2 km) to the traffic sites (e.g.Marylebone Road) and Eltham is a suburban site in London, which is also not far from the Marylebone Road ($14 km).These sites lay on a large concentration gradient, thus the WRF-Chem model with 15 km Â 15 km resolution was unable to distinguish between points beyond the areas.This can also explain the large model-measurement [OX] B deviation for the Marylebone Road site compared with the other London sites.
The model-measurement comparison plots (Fig. 10) for different months (January, May and September) show that the model matched the changes in [OX] levels with measurement very well for the Bangkok roadside site Thonburi with  observed peaks and troughs at nearly similar times (correlation coefficients, r > 0.5).However, there is a large overprediction of model [OX] values with biases and RMSEs of $10 and $15 ppb for January, $17 and $27 ppb for May, and $21 and $26 ppb for September (Fig. 10 and Table 4) which is most likely arising from inaccurate NO x emissions (traffic, industrial) in the inventory data.Models are used to predict and forecast the effects of changes in emissions.They therefore require accurate emission inventory data and complete representations of physical and chemical atmospheric processes in order to do this as well as possible.
There have been noted issues with emissions inventories (RETRO, INTEX-B, MACCity, SEAC 4 RS) most recently which cause an over-prediction of O 3 mixing ratios in Thailand for March and December. 73The spatial distribution of the regional emissions, with hot spots seen in Bangkok, implemented in EDGAR is based on population, which does not reect the real emission scenarios of Thailand.Although Bangkok has a large population, the stationary sources of emissions are likely to be overestimated in the EDGAR emission inventory.To improve the model-measurement deviation, the development of an alternative regional emission inventory based on emission control technology for stationary and mobile sources is necessary.
AE5 ppb for both NO 2 and O 3 detectors.The hourly data of NO x and O 3 measurements over the period from January 2005 to December 2014 for Din Daeng and Chokchai and January 2005 to December 2018 for Thonburi was retrieved from offline source of PCD (http://www.pcd.go.th).The monthly data of NO x and O 3 have been computed for [OX] calculation if more than 70% of the data was captured over the individual month (see ESI Fig. S1 † for valid data coverage of each site).
Local and background oxidants and their contribution and trend The comparison of the average 'local' and 'background' contribution to the OX levels, [OX] L and [OX] B , in the Marylebone Road (2000-2019), Din Daeng (2005-2014), Chokchai (2005-2014) and Thonburi (2005-2018) with the other studies is shown in Table 1.The variability of [OX] L for London and Bangkok for different studies could be due to the control strategy response (e.g.changing the eet technology by introducing Euro-classes and alternative-fuelled vehicles and/or adopting a range of regulations) for reducing vehicle emissions over time.The [OX] L and [OX] B levels in the Marylebone Road found in the study are comparable with the study of Jenkin 6 who analysed the [OX] L and [OX] B in London rural, urban and kerbside sites over 1992-2001.The large difference of [OX] L levels in the study from the Jenkin 6 study can be explained by the larger fraction of NO x emitted as NO 2 in the roadside site than that in the rural, urban and kerbside sites.For Bangkok roadside sites, we found average [OX] B levels of 32 ppb, 31 ppb, and 29 ppb for Din Daeng, Chokchai and Thonburi, respectively, but Uttamang et al. 25 reported signicantly higher average [OX] B of 48 ppb and 95 ppb for non-episodes ([O 3 ] < 100 ppb) and episodes ([O 3 ] > 100 ppb), respectively.This is to be expected

for
Fig. 2 (a) The trends and (b) the percent contribution of oxidant levels in the Marylebone Road for the period of 2000-2019.Statistical significance is based on a p < 0.001 and the trends are reported with 95% confidence intervals.

Fig. 3
Fig. 3 The trends (a, c and e) and the percent contribution (b, d and f) of oxidant levels in Din Daeng (2005-2014), Chokchai (2005-2014) and Thonburi (2005-2018), respectively.Statistical significance is based on a p < 0.001 and the trends are reported with 95% confidence intervals.

Fig. 4
Fig. 4 The seasonal cycles of (a) background OX and (b) local OX contributions in the Marylebone Road (2000-2019).

Fig. 6
Fig. 6 The diurnal cycles of (a) background and (b) local oxidant levels in the Marylebone Road (2000-2019).

Fig. 8
Fig. 8 The model-measurement comparison of [OX] levels in London sites for the period of 30 July to 24 August 2012 (left side figures) and 8 January to 11 February 2012 (right side figures).The red and blue lines represent measurement and modelled data, respectively.

Fig. 9
Fig. 9 The model-measurement comparison of background [OX] levels in London sites for the period of (a) 30 July to 24 August 2012 and (b) 8 January to 11 February 2012 (right side figure).The red, blue, black, yellow, green lines represent Marylebone Road, Bloomsbury, Eltham, Westminster and Harlington, respectively.The solid lines are measurement data and broken lines are modelled data.

Fig. 10
Fig. 10 The model-measurement comparison of [OX] levels in the Bangkok roadside site Thonburi for January, May, and September 2017.The red and blue lines represent measurement and modelled data, respectively.

Table 1
The comparison of the local and background OX levels in London and Bangkok sites derived from fitted linear regression analysis .2% per year NO x emitted as NO 2 in the Marylebone Road from 2001 to 2005.The increased introduction of Euro-III diesel vehicles tted with oxidation catalysts or the tting of catalytically regenerative particle traps to buses can produce this extra amount of NO 2 , 51 which could be responsible for increasing [OX] L levels from 2001 to 2005 by 9.8% per year.In response to the Euro IV-VI and Euro 4-6 legislations, a large number of policy initiatives were introduced in line with the London Mayor's Air Quality Strategy (e.g. the roll out of new hybrid and electric vehicles or zero-emission vehicles), thus changing the composition of the vehicle eet, which is a direct consequence of the Low and Ultra Low Emission Zone legislature and inclusion of the Congestion Charge Zone across London to improve air quality by reducing NO x and particulate matter (PM) levels.

Table 2
Correlation coefficients, biases and root mean square errors (RMSE) between model-measurement [OX] levels for different London sites during summer and winter months This journal is © The Royal Society of Chemistry 2020Faraday Discuss.

Table 3
Correlation coefficients, biases and root mean square errors (RMSE) between model-measurement [OX] B levels for different London sites during summer and winter months Faraday Discuss.This journal is © The Royal Society of Chemistry 2020

Table 4
Correlation coefficients, biases and root mean square errors (RMSE) between model-measurement [OX] levels for the Bangkok roadside site Thonburi during January, May and September 2017 by the Upper Atmosphere Research and Tropospheric Chemistry Programs.© 2020 all rights reserved.
This journal is © The Royal Society of Chemistry 2020Faraday Discuss.supported