Modelling the impacts of emission changes on O 3 sensitivity, atmospheric oxidation capacity and pollution transport over the Catalonia region

. Tropospheric ozone (O 3 ) is an important surface pollutant in urban areas, and it has complex formation mechanisms that depend on the atmospheric chemistry and meteorological factors. The severe reductions observed in anthropogenic emissions during the COVID-19 pandemic can further our understanding of the photochemical mechanisms leading to O 3 formation and provide guidance for policies aimed at reducing air pollution. In this study, we use the air quality model WRF-Chem coupled with the urban canopy model BEP-BEM to investigate changes in the ozone chemistry over the Metropolitan Area of 5 Barcelona (AMB) and its atmospheric plume moving northwards, which is responsible for the highest number of hourly O 3 exceedances in Spain. The trajectories of the air masses from the AMB to the Pyrenees are studied with the Lagrangian particle dispersion model FLEXPART-WRF. The aim is to investigate the response of ozone chemistry to reduction of precursos emissions (NO x , VOC). The results show that with the reduction in emissions: 1) the ozone chemistry tends to enter the NOx-limited or transition regimes; however, highly polluted urban areas are still in the VOC-limited regime, 2) the reduced O 3 production

Ozone is photochemically produced through nonlinear chemical processes, involving mainly reactions of nitrogen oxides (NO x = NO 2 +NO) and volatile organic compounds (VOCs); this results from both anthropogenic and biogenic sources in the presence of sunlight (Monks et al., 2015;Crutzen, 1974;Derwent et al., 1996).This chemistry occurs in two photochemical regimes, NOx-sensitive and VOC-sensitive (Sillman et al., 1990;Sillman, 1999).In the NOx-sensitive regime (low NO x and high VOC), ozone production is controlled (or limited) by the concentration of NO x , therefore, the O 3 levels increase with increasing NO x , and there are only small changes with increases in the VOC levels.In this regime, O 3 reacts mainly with hydrogenated species to form hydrogen peroxide (H 2 O 2 ), which is removed by wet and dry deposition.In the VOC-sensitive regime (high NO x ), ozone levels increase with increased VOC concentrations and decrease with increased NO x concentrations.This last regime is typical of urban areas where a reduction in NO x emissions enhances ozone levels locally due to the higher levels of oxidants (mainly hydroxyl radicals, OH) and reactions with VOCs.Peak concentrations of ozone usually occur during the midday hours, when the sunlight is most intense.However, during the afternoon and evening, high ozone concentrations are observed in remote areas due to higher biogenic VOC emissions, less titration by NO, and transport of O 3 and its precursors from their sources.At night and next to a source with high emissions of NO (e.g., power plants), ozone is lost through the process of NO x titration and forms NO 2 , which is subsequently converted to nitric acid (HNO 3 ) and removed from the atmosphere by wet and dry deposition (Monks et al., 2015).In summary, O 3 levels can be reduced only if there are reductions in the amounts of the precursors NO x , VOC and carbon monoxide (CO).Reductions in VOC emissions would be an effective pathway to reducing ozone in a high NO x area (VOC-sensitive).On the other hand, reductions in NO x emissions would be effective in reducing O 3 if NO x -sensitive chemistry dominates.In urban areas, this photochemistry illustrates the difficulties involved in developing policies to reduce O 3 in polluted regions (Sillman, 2003).Thus, a more profound understanding of the sensitivity of local ozone formation to changes in NO x and VOC levels is essential for developing effective air quality policies.
Other important processes for removal of O 3 are reactions of halogens species (Badia et al., 2021a), which remove 30-35 Tg (11-15 %) of tropospheric ozone, and dry deposition, which accounts for about 20% of the O 3 lost from the troposphere (Wild, 2007).
In addition to photochemical reactions, the concentration of ozone is sensitive to meteorological variables such as solar radiation and wind speed and direction (Neiburger, 1969).Ozone production is intensified on warm, sunny days when the air is stagnant.Therefore, the increase in frequency, severity, and duration of heatwaves during recent decades increases the need to understand the influence of meteorological drivers and anthropogenic factors on ground-level ozone pollution.This is becoming more important for urban cities in the Mediterranean area with high summer temperatures, and heatwaves are projected to become more severe in the future due to anthropogenic climate change (Pyrgou et al., 2018;Zittis et al., 2015).Indeed, local ozone levels depend not only on local production and loss mechanisms that are sensitive to meteorological factors but also on the transport of ozone and its precursors.Previous studies have shown that ambient ozone concentrations are strongly influenced by transport of regional ozone and its precursors, while local precursor emissions play limited roles in ozone formation (Romero-Alvarez et al., 2022;Kleanthous et al., 2014).Cristofanelli and Bonasoni (2009) showed that the background tropospheric ozone concentration in the Mediterranean area and southern Europe is affected mainly by three transport processes: 1) regional and long-range transport of pollutants, 2) downwards transport from the stratosphere, and 3) transport of dust from the Sahara Desert.
The lockdown period provided a significant reduction in ozone precursors, and it represents an excellent opportunity to further our understanding of the photochemical reactions involved in ozone chemistry.Estimated average emission reductions in Spain during the most severe lockdown period were reported, with road and air traffic reductions reaching 80-90% Guevara et al. (2021).During this period, Guevara et al. (2021) estimated the average emission reductions at the EU-30 level to be -33% for NO x .Consequently, restrictive mobility measures that included important reductions in traffic had many positive environmental impacts, 70 and improvements in air quality were reported globally (Liu et al., 2020;Venter et al., 2020a;Sharma et al., 2020).
Previous modelling studies analysed the changes in air quality, emissions, and chemical regimes seen on global and regional scales during the COVID-19 lockdown (Miyazaki et al., 2021;von Schneidemesser et al., 2021;Roozitalab et al., 2022;Sicard et al., 2020;Badia et al., 2021b).A global modelling study by Miyazaki et al. (2021), showed that the global total tropospheric ozone burden declined by 6 Tg (∼2%) during May-June 2020, mainly due to emission reductions in Asia and the Americas.
The modelling study of Venter et al. (2020b) found that, after taking into account the meteorological variations, lockdown measures have reduced the levels of NO 2 , mainly due to the reduction in transportation, and PM levels by approximately 60% and 31% in 34 countries, with a general increase in O 3 of 4% (-2 to 10%).Sicard et al. (2020) described ozone increases in cities (17% in Europe, 36% in Wuhan) resulting from lower titration of O 3 by NO due to the strong reduction in NO x emissions from road transport.Most of this literature was focused only on the lockdown period, and the de-escalation period, which had different ozone chemistry, was not analysed.In addition, these studies did not discuss changes in the chemical regimes arising for different land uses and the transport of pollutants due to lockdown measures from cities to rural areas.
In recent decades, EU emission mitigation policies have been successful in decreasing emissions of key air pollutants such as SO 2 , NO x , NMVOCs, and PM (Sicard et al., 2021;and European Environment Agency et al., 2019).However, the current levels for the secondary air pollutant O 3 in cities continue to exceed the EU standards and WHO air quality guidelines (Guerreiro et al., 2014;and European Environment Agency et al., 2019).Indeed, local ozone pollution mitigation efforts are generally inefficient, mainly because 1) ozone formation depends on nonlinear chemical interactions, 2) with a lifetime of several weeks, ozone levels are strongly influenced by long-distance transport, which is associated with specific weather conditions and the hemispheric background, and 3) its precursors are emitted mostly far from the sites of ozone exceedances.Sicard et al. (2013) found a decrease in annual O 3 averages (-0.4% yr-1) at rural sites and an increase at urban and suburban stations (0.6% and 0.4% , respectively) during 2000-2010 for the Mediterranean area.These changes resulted from the mitigation policies for NO x and VOC emissions in the EU that led to an increase in O 3 levels in urban areas due to a reduction in titration by NO.Therefore, there is an urgent need to expand our knowledge of ozone chemistry to help decision-makers choose better mitigation strategies.
In the present study, we use the air quality model WRF-Chem coupled with the urban canopy model BEP-BEM to investigate the response of NO x -VOCs-O 3 chemistry to changes in precursor emissions in the Metropolitan Area of Barcelona (AMB); this furthers our understanding of ozone formation mechanisms and transport to rural areas, and it enables the design of effective air quality mitigation strategies.We compare the ozone levels for two periods with very different anthropogenic and biogenic activity levels: March and April 2020, when ozone precursors were at their lowest levels because of repeated lockdowns due to COVID, and 2) May 2020, when the AMB was in a de-escalation :::::::: relaxation plan but the ozone levels were typically highest because of the bloom of biogenic emissions under the intense sunlight conditions.In addition, ozone production regimes for different land uses of the AMB are analysed to determine changes in the NOx/VOC ratio in different areas of the city that are affected by different biogeochemical effects (biogenic VOC emissions, dry deposition, anthropogenic emissions of ozone precursors).Here, we also discuss changes in the AOC ::::::: oxidants and propose that the AOC should be considered when designing air quality control policies.Changes in ozone circulation from the AMB to the Pyrenees mountains are also discussed for specific days characterized by high ozone levels.
The case study is described in Sect. 2. The air quality model, including the model setup, validation, and simulations, is described in Sect.3, followed by the results (Sect.4.1).A discussion of the ozone chemistry that includes ozone sensitivity, AOC and transport from the city to rural areas is presented in Sec. 5.

Case study
The Metropolitan Area of Barcelona (AMB) serves as our case study.It is located in Catalonia (Spain) in the northeastern part of the Iberian Peninsula (Fig. 1).This region is characterized by a Mediterranean climate, with dry and hot summers and clear skies.Due to a complex orographic territory with different altitudes and several peripheral mountain ranges and depressions, it is not easy to generalize the climatic features of the Catalan lands for the whole territory (Martín-Vide et al., 2010).The AMB, with more than 3 million people, is the most populated urban area on the Mediterranean coast.
The city of Barcelona annually reports some of the highest air pollution levels in Europe, and the most problematic pollutants are NO 2 , PM2.5, and PM10 (Rivas et al., 2014).In particular, in 2019, the NO 2 annual mean levels in the high traffic urban air pollution ground monitoring stations (Eixample and Gràcia-Sant Gervasi) exceeded the WHO guideline (40 µg m − 3) (Rico et al., 2019).In the same year, the mean values for PM2.5 and PM10 exceeded the WHO guideline (20 and 10 µg m − 3, respectively) at all urban stations in the city (Rico et al., 2019).Exceeding these air quality reference levels is associated with significant risks to public health (Organization, 2021;Rivas et al., 2014).The year 2020 was the first year in which the NO 2 values in Barcelona remained within the WHO limits (Rico et al., 2020) due to the significant reduction in traffic emissions that resulted from the Spanish government's emergency rule and its lockdown restrictions (see Supplement Fig. S1).Note that 2020 had more rainfall than previous years, which has important implications for the removal of pollutants from the air.
Another air quality problem was found for the Vic plain (see Fig. 1), which records the highest number of exceedances for hourly O 3 levels (180 µg m − 3) when the sea breeze transports the ozone precursors from AMB inland to this rural plane (Querol et al., 2017;Massagué et al., 2019;Jaén et al., 2021).During the late spring and summer seasons, the combination of daily upslope winds and sea breezes may cause the intrusion of polluted air masses up to 160 km inland.Thus, the air massed from a polluted area (such as the AMB) can be transported northwards and injected at high altitudes (2000-3000 masl) by the Pyrenean mountain ranges (Querol et al., 2017;Massagué et al., 2019).
Ozone levels were higher compared to previous years in Ciutadella, the background station in Barcelona, during the period March-June 2020 (see Supplement Fig. S2).In Tona, a rural station located in the Vic Plain and situated 45-70 km north of Barcelona and surrounded by high mountains, and in Pardines, also a rural site situated in the Pyrenees mountains (see Fig. 1), high ozone concentrations were registered and clearly exceeded the WHO limit of 60 µg m − 3 for peak seasons.Ozone levels were high in these two rural stations during March-June with higher values for 2015-2019 compared to 2020 due to the reduced emissions in the AMB.Note that for some weeks, the O 3 levels were higher in 2020 than in 2015-2019 due to meteorological conditions that increased the levels of ozone precursors (see Supplemental Figs.S1 and S2).
The first period ::: (30 :::::::: March-12 :::::: April) was characterized by meteorological dynamism in the Iberian Peninsula.The period began with a high-pressure centre in the northern Atlantic Ocean, which generated a cold and dry continental NE flux over Europe.General precipitation was registered in Catalonia during the first days (March 30 -April 4), caused by a low-pressure centre that moved from the Atlantic to the Mediterranean.These conditions were optimal for generating precipitation in the Mediterranean area of the peninsula.The next days were defined by high pressures over Europe, which generated a more stable meteorology with isolated precipitation and a warm air mass crossing Europe from the south.During this period, no important fluxes from any directions were detected.At the surface level, the Azores anticyclone became stronger and reached pressures above 1030 hPa over the Atlantic.The second period ::::: (18-30 ::::: May) had more stability due to a strong anticyclonic ridge covering the SW and centre of Europe, resulting in typical summer weather.Winds from the N and NW were detected NE of Catalonia and in the Ebro Valley, respectively, due to the Pyrenees natural barrier, which modifies the trajectories of superficial winds.
Low pressures were found over Italy, which intensified the winds from the north over the east of the peninsula.The last days of this period (May 24-30) registered weak precipitation in the Pyrenees and the area NE of Catalonia, with weak winds from different directions (turning from north to south).This was caused by a strong anticyclone located in northern Europe (over 1030 hPa), which generated an undetermined situation with no effects of high or low pressures in the Peninsula.See Table 1) for a summary of the meteorological conditions for these periods.
In addition, we select two days in the lockdown period (the 3rd and 6th of April) and two days in the relaxation period (the 22nd and 26th of May), during which high ozone concentrations were registered (see Table S1 in the Supplement) ::: and  S2 of the Supplement.According to the meteorological data from the Servei Meteorològic de Catalunya Servei Meteorològic de Catalunya (SMC), there was no precipitation during these four days ::: for ::: the :::: days :::::::: selected, except in the Pyrenees (2.9 mm) on the 22nd of May, the wind intensity was low and the surface temperatures were significantly high for this time of the year during the two days in May.
To supplement this information, Figures 9 and 10 show the trajectories of the air masses arriving at the monitoring stations on the selected days; which were modelled with the Lagrangian particle dispersion model FLEXPART-WRF (Brioude et al., 2013) .This version of the Lagrangian model works with the Weather Research Forecasting (WRF) mesoscale meteorological model, with the same parametrization as the WRF-Chem model (see section 3.1).The transport model has been run in backwards mode, which means that what is represented in each plot is the residence time, at each grid cell of the map, for the air masses arriving at each site.Twenty-four-hour back trajectories were calculated for each day at a release time of 16 h and with a grid cell size of 0.03 × 0.03 degrees.The air masses on the 3rd of April and 22 of May were transported from the AMB to rural areas such Montseny and the Vic Plain, and we can see an influence from the bottom layers (0-300 m) and the upper layers (300-2000 m) at the different sites.The air masses on the 6th of April were channelled from the AMB northwards to Montseny, the Vic Plain and the Pyrenees.The air masses on the 26th of May were also transported from the AMB northwards to Montseny, the Vic Plain and the Pyrenees, but the air masses that arrived at the surfaces of these locations had strong local components and a larger influences from the upper layers.

Air quality model
We used the regional chemistry transport model WRF-Chem (Grell et al., 2005) version 4.1, a highly flexible community model for atmospheric research in which aerosol-radiation-cloud feedback processes are considered.The WRF-Chem model is widely used for simulations of air pollution episodes (Georgiou et al., 2018;Yegorova et al., 2011) and, in particular, the air quality over the AMB has been analysed in Badia et al. (2021b).

Model set-up
The WRF-Chem model is configured with two domains covering the Iberian Peninsula (D1: 9 km×9 km) and the Catalonia region (D2: 3 km×3 km) with 45 vertical layers up to 100 hPa (Fig. 2).The meteorological and chemical initial and lateral boundary conditions (IC/BCs) were determined using the ERA5 global model data (Hersbach et al., 2020) and WACCM (Gettelman et al., 2019), respectively.The HERMESv3 preprocessor tool (Guevara et al., 2019) was used to create the anthropogenic emissions files from the CAMS-REG-APv3.1 database (Granier et al., 2019).This emission inventory is based on data from 2016.Biogenic emissions are computed online from the Model of Emissions of Gases and Aerosols from Nature v2 (MEGAN; Guenther et al. (2012)).For the gas-phase chemical scheme, we used the Regional Acid Deposition Model (RADM2, Stockwell et al. (1990)), which accounts for 63 chemical species, 21 photolytic reactions and 136 gas-phase reactions.NMVOC oxidation in RADM2 only explicitly treats ethane, ethene, and isoprene species, and all other NMVOCs are classified as grouped species based on OH reactivity and molecular weight.Thus, the RADM2 gas-phase chemical mechanism grouped the VOCs into 14 species, such as alkane, alkene, aromatic, and formaldehyde.In WRF-Chem, RADM2 is coupled to the MADE/SORGAM aerosol module (Ackermann et al., 1998;Schell et al., 2001).RADM2 has been broadly used in studies of the air quality over Europe (Im et al., 2015;Tuccella et al., 2011;Badia et al., 2021b) .Here, we used a multilayer layer urban canopy scheme, the building effect parameterization (BEP) coupled with the building energy model (BEP+BEM, (Salamanca et al., 2011)) to represent the urban areas in our domain; this takes into account the energy consumed by buildings and the anthropogenic heat, which has been previously validated for the area under study (Ribeiro et al., 2021;Segura et al., 2021).The local climate zone (LCZ) classification (Stewart and Oke, 2012) is used for the AMB, which associates specific values of the thermal, radiative and geometric parameters of the buildings and ground into 11 urban classes, which are used by the BEP+BEM urban canopy scheme to compute the heat and momentum fluxes in the urban areas (see Segura et al. (2021) for more details on the use of LCZ and urban morphology).We performed a spin-up of 1 month.
Table 2 describes the main configuration of the model.

Description of the simulation cases
To better understand the impacts of emission reduction measures on air quality, the WRF-Chem model was utilized to calculate the changes in O 3 chemistry during the COVID lockdown period.We ran two simulations: 1) Business As Usual (BAU) and COVID for the period of March-June 2020 (see Table 1).The COVID run used the emissions changes (see Fig. 3) provided by the Barcelona Supercomputing Center (Guevara et al., 2021), which were previously used in other studies (von Schneidemesser et al., 2021;Brancher, 2021).These emission changes varied per day, country and sector. Figure 3 displays the emission changes used in this study, and the highest changes were found for the road transport (up to 80%) and aviation (up to 90%) sectors.
Inputs for the other emissions (biogenic, dust, sea-spray) and meteorology used in WRF-Chem were set to be consistent.As a result, the differences in pollutant concentrations calculated by WRF-Chem were attributed to changes in the anthropogenic emissions.

Model validation
Several meteorological and air quality stations were used herein to evaluate the model (COVID simulation) for the lockdown (30 March to 12 April 2020) and relaxation (18 to 30 May 2020) periods.The same model configuration has been evaluated previously over the AMB for the meteorology (Ribeiro et al., 2021;Segura et al., 2021) and the chemistry, without any reduction in anthropogenic emissions (Badia et al., 2021b), for different periods.

Meteorology
The meteorological data used to validate our model were from the Xarxa d'Estacions Meteorològiques Automàtiques (XEMA).
Stations within this network are classified as urban and rural according to the land use of the model.Herein, we used data for the wind speed (WS), temperature (T) and relative humidity (RH).Tables S3 and S4 in the Supplement present statistical evaluations of hourly data for the Metropolitan Area of Barcelona (AMB) and Catalonia (CAT) region for the lockdown ( 30March to 12 April) and relaxation (18 to 30 May) periods, respectively.
The validations of the WRF-Chem simulation (COVID run) revealed that the model generally reproduced the air temperatures of the two simulation periods well, but the performance was poorer in representing the relative humidity and the wind speed.For the first period, the simulated air temperatures for the urban stations showed low positive biases of 0.4 °C and 0.3 °C for the AMB and CAT, respectively, and an RMSE of 1.4 °C.The rural stations presented a higher bias inside the AMB (0.9 °C), which resulted from erroneous descriptions of the land use at the model resolution level (3 km).Loose performance was found for the relative humidity and the wind speed, with average RMSEs of 12.2% and 2.5 m/s, respectively.The model underestimated the relative humidity in the urban and rural areas of the AMB by 2.0% and 5.2%, respectively, while it overestimated the humidity in the rural areas of the CAT (1.4%).In the case of the wind speed, the model overestimated the wind flow over the entire domain (1.5 m/s on average in CAT), especially in the rural areas of CAT (1.9 m/s).For the second study period, a similar performance was obtained for the air temperatures inside the AMB, with a slight increase in the RMSE (1.5 °C and 1.6 °C for urban and rural areas) and a decrease in the correlations between modelled and observed data (0.90 and 0.92, respectively).Unlike the first period, the model overestimated the relative humidities at all stations in the second period, except for the rural stations in the AMB (-3.8%).The model provided lower overestimates for the wind speeds during the second period (1.0 m/s on average in CAT), although the correlations decreased in the second period.

Air quality
Air quality data from the monitoring stations Xarxa de Vigilància i Previsió de la Contaminació Atmosfèrica (XVPCA) were used here.Stations in this network were classified into different groups: urban background, urban traffic, suburban background, and rural.Here, we used the data for O 3 and NO 2 .Tables S5 and S6 in the Supplement present statistical evaluations of hourly data for the AMB and Catalonia during the lockdown (30 March to 12 April) and relaxation (18 to 30 May) periods, respectively.The modelled concentrations were converted to units of µg m − 3 by using the temperatures and pressures from the model.
Overall, the model (COVID simulation) showed reasonable agreement with the observations for NO 2 and O 3 concentrations during both periods.The best performance in the lockdown period was observed over the urban background (R between 0.43 and 0.45 for NO 2 and between 0.70 and 0.73 for O 3 ), while low R values were found over the rural areas (0.32 for NO 2 and 0.42 for O 3 ).The performance for the relaxation period was not as good as that for the lockdown period, with R values between 0.24-0.40for NO 2 and 0.42-0.62 for O 3 .However, there were negative and positive biases in both periods for NO 2 (NMB between -0.15 and -0.66) and O 3 (NMB between 0.13 and 0.28), respectively.Similar biases were seen in another study (von Schneidemesser et al., 2021).Part of our model bias was attributed to the 1) boundary conditions used for this study (WACCM model) that added a bias to the O 3 background levels (Giordano et al., 2015) and 2) the current emission inventory was too coarse to accurately represent the spatial distributions and temporal variations in NO x emissions, e.g., from road transport.
Low values for the modelled NO x levels underestimated ozone loss via NO titration, which resulted in high nighttime surface ozone concentrations.The lifetime of surface NO x (few hours) is shorter than that of O 3 (days or weeks); thus, the surface NO x concentrations are very sensitive to emissions.We should also mention that there might be large uncertainties for the calculations of emissions factors, as discussed in Doumbia et al. (2021); underestimates of traffic NO x emissions over Europe have been mentioned previously in several air quality modelling studies (von Schneidemesser et al., 2021;Karl et al., 2017).and :::::::: relaxation ::::::: periods ::: for ::: the ::: year ::::: 2020 :: in ::: the ::: city ::: of ::::::::: Barcelona.
However, despite the cuts in emissions, most of the grids close to highly polluted areas (compact urban, industrial, and water) were still in the VOC-sensitive or transitional regimes all day, especially in the evenings (high traffic emissions), for which we found the highest ozone increases (2.4-5%).Similar results were found when we compared both runs (BAU and COVID) for the period in May in terms of the changes in the chemical for each land use.However, we found that during May, the maximum ozone levels decreased in the COVID run during the afternoons (up to 1.6% in green areas), which was attributed to reductions in the anthropogenic emissions that decreased the ozone precursor levels (24 to 40% for NO x and 24 to 40% for VOCs) and consequently ozone production.For green areas far from anthropogenic sources (forests), the ozone levels were also reduced in the mornings and evenings during this period.
The lockdown measures inhibited NO titration of the O 3 , mainly due to changes in the local NO x emissions resulting from road transport.This resulted in an increase in the O 3 levels during the evening hours, where there was no photolytic reaction with NO 2 , in urban areas with high population densities.We found that air quality policies based solely on transport reduction (as illustrated by the COVID lockdowns, which reduced NO x levels) actually intensified O 3 levels over urban areas, indicating the need for a protocol with strident control measures to reduce NO x emissions without significantly reducing anthropogenic VOCs to control O 3 levels.However, high ozone production during May was reduced due to reduced levels of the precursors, and consequently, there were reductions in the maximum ozone levels for that period.

Impacts on the atmospheric oxidation capacity
In addition to understanding changes in the levels of O 3 precursors, it is important to determine how emission changes affect the atmospheric oxidation capacity (AOC) because this plays an important role in the loss and production rates of primary and secondary pollutants.OH is the dominant tropospheric daytime oxidant, and it increases considerably (up to 0.12 ppt, +45%) because of significant reductions in NO 2 levels, since NO 2 is the primary OH sink (Elshorbany et al., 2009).The rises in ::::::: increase :: of : these free radical levels :::::::::::: concentrations could be the leading cause for the diurnal O 3 increases given their strong link with O 3 production ( ::: (see :::: Fig. :: S5 ::: in ::: the ::::::::::: Supplement) :::: given :::: that : VOC and CO oxidation by OH are the initial reactions for ozone formation).In addition, the NO 3 radical, which is a primary night-time oxidant, also increases in areas close to the airport and harbour (4 ppt, 210%).This increase can be explained by reductions in the VOC and NO 2 levels, which are important sinks for NO 3 radicals (Elshorbany et al., 2009;Saiz-Lopez et al., 2017).

Pollution transport from urban to rural
In addition to studying the mechanisms for ozone formation in the AMB, we also explored how ozone is transported to rural areas to determine the influence of urban pollution.Rural areas far from the city, such as the Vic Plain and the Pyrenees mountain range, are frequently affected by the atmospheric plume transported northwards from the AMB (Massagué et al., 2019).Indeed, ozone exceedances over these places occur when there are high levels of NO 2 (mainly due to road traffic) over the AMB (see section 2).The urban plume is driven inland by southeast and southern combined sea-valley-mountain breeze winds, channelled by north-south valleys, and crosses the coastal and precoastal Catalan Ranges to an intramountain plain.
We found significant decreases in ozone levels in the COVID simulation (up to 3 ppb, from the surface up to 2000 m) from the AMB north to the mountain area of Montseny on both days in May (22nd and 26th,Figs. 13 and 14,respectively), especially during the afternoons when the PBLH was the highest and solar radiation led to enhancement of the sea breeze front, which provided favourable conditions for regional transport (Massagué et al., 2019).The decreases in ozone precursor emissions (COVID simulation) resulted in less ozone production :::: from ::: the ::::: AMB ::::: plume :: as :::: well ::: as ::::::::: production :: of :::: new ::: O 3 , and consequently, the ozone concentrations decreased (discussed in sections 5.1 and 5.2).This decrease was also seen for the evening hours.Note that at night, when ozone accumulated on the surface following the decrease in the PBLH, there was a slight increase in its levels due to limited titration by NO.However, there were still reductions in the ozone levels from 500 m to 2000 m at night.
A comparison of these two periods, April-March and May, showed that the mitigation strategies designed to reduce the high ozone levels were more efficient in May, when ozone formation was high (high biogenic emissions coinciding with anticyclonic conditions).Thus, given the importance of meteorology in air pollution events occurring over urban and rural areas, new mitigation strategies are needed to improve the air quality and would result in significant O 3 reductions; the local O 3 coming from the AMB plume would be reduced, as would the recirculated O 3 and thus the intensity of surface O 3 fumigation from high O 3 reservoir layers in other areas.

Conclusions
Improving air quality is a top priority in urban areas and requires a better understanding of how the O 3 levels respond to changes in the emission levels of the precursors, as well as the ozone formation regimes and the atmospheric oxidation capacity and associated O 3 formation.Furthermore, urban emissions affect the O 3 levels in rural areas outside the cities.In this study, we used the air quality model WRF-Chem to analyse the air quality changes occurring over the Metropolitan Area of Barcelona and other rural areas affected by transport of the atmospheric plume from the AMB during mobility restrictions.
The large reduction in NO x levels (up to 60%) seen during the lockdown period combined with a slight change in VOC levels (up to 10%) led to increased O 3 concentrations (up to 20% in the evening).The significant increase found in the evening was mainly due to reduced O 3 titration by NO, which prevailed over the lower O 3 production level caused by decreases in the levels of the O 3 precursors.The lockdown occurred during April-March when ozone photochemical production was still not at the highest level.In addition, our results showed a significant increase in the atmospheric oxidation capacity (AOC) indicated by the enhanced oxidant (OH and NO 3 ) levels, which was consistent with the slight increases seen in the maximum O 3 concentrations during the lockdown.We also found that for several days, these increases were seen further north in rural areas such as the Vic Plain, which produced the most annual exceedances in Spain.Large enhancements over these areas were the result of 1) a higher regional O 3 background level, 2) vertical recirculation of the air masses that transport high concentrations of O 3 from the upper levels to the lower levels, and 3) the contributions of the AMB pollution plume travelling along the S-N valley connecting the AMB and the Vic Plain and the Pyrenees.High ozone levels seriously affect human health and the environment.In addition, the consistent differences seen in O x (NO 2 + O 3 ) concentrations during the period April- March have important policy implications, i.e., that effective mitigation strategies designed to reduce air pollutants and their health effects should include reductions in both O x and VOC levels to avoid increases in ozone levels.

Figure 1 .
Figure 1.a) Location and b) main topographic features of the study area.Base maps in Panel a were taken from Google Earth.The locations of air pollution monitoring stations (Xarxa de Vigilància i Previsió de la Contaminació Atmosfèrica, XVPCA) along the S-N axis (Barcelona-Vic Plain-Pyrenean range) are shown in Panel a (right).

Figure 3 .
Figure 3. Emissions reduction percentage (%) for each sector.Note that the other stationary combustion sector has a different reduction level for each pollutant.The periods analysed here are written at the bottom of the figure.

Figure 11 .Figure 12 .
Figure 11.O3 changes between the COVID and BAU simulations along the atmospheric plume from the AMB to the Pyrenees for the 3rd of April.The modelled PBLH is shown with a green line.

Figure 14 .
Figure 14.Same as Figure 11 for the 26th of May.

Table 2 .
Model details and experiment configuration . :