University of Birmingham Phenomenology of high-ozone episodes in NE Spain

Ground-level and vertical measurements (performed using tethered and non-tethered balloons), coupled with modelling, of ozone (O3), other gaseous pollutants (NO, NO2, CO, SO2) and aerosols were carried out in the plains (Vic Plain) and valleys of the northern region of the Barcelona metropolitan area (BMA) in July 2015, an area typically recording the highest O3 episodes in Spain. Our results suggest that these very high O3 episodes were originated by three main contributions: (i) the surface fumigation from high O3 reservoir layers located at 1500–3000 m a.g.l. (according to modelling and non-tethered balloon measurements), and originated during the previous day(s) injections of polluted air masses at high altitude; (ii) local/regional photochemical production and transport (at lower heights) from the BMA and the surrounding coastal settlements, into the inland valleys; and (iii) external (to the study area) contributions of both O3 and precursors. These processes gave rise to maximal O3 levels in the inland plains and valleys northwards from the BMA when compared to the higher mountain sites. Thus, a maximum O3 concentration was observed within the lower tropospheric layer, characterised by an upward increase of O3 and black carbon (BC) up to around 100–200 m a.g.l. (reaching up to 300 μg m−3 of O3 as a 10 s average), followed by a decrease of both pollutants at higher altitudes, where BC and O3 concentrations alternate in layers with parallel variations, probably as a consequence of the atmospheric transport from the BMA and the return flows (to the sea) of strata injected at certain heights the previous day(s). At the highest altitudes reached in this study with the tethered balloons (900–1000 m a.g.l.) during the campaign, BC and O3 were often anti-correlated or unrelated, possibly due to a prevailing regional or even hemispheric contribution Published by Copernicus Publications on behalf of the European Geosciences Union. 2818 X. Querol et al.: Phenomenology of high-ozone episodes in NE Spain of O3 at those altitudes. In the central hours of the days a homogeneous O3 distribution was evidenced for the lowest 1 km of the atmosphere, although probably important variations could be expected at higher levels, where the high O3 return strata are injected according to the modelling results and non-tethered balloon data. Relatively low concentrations of ultrafine particles (UFPs) were found during the study, and nucleation episodes were only detected in the boundary layer. Two types of O3 episodes were identified: type A with major exceedances of the O3 information threshold (180 μg m−3 on an hourly basis) caused by a clear daily concatenation of local/regional production with accumulation (at upper levels), fumigation and direct transport from the BMA (closed circulation); and type B with regional O3 production without major recirculation (or fumigation) of the polluted BMA/regional air masses (open circulation), and relatively lower O3 levels, but still exceeding the 8 h averaged health target. To implement potential O3 control and abatement strategies two major key tasks are proposed: (i) meteorological forecasting, from June to August, to predict recirculation episodes so that NOx and VOC abatement measures can be applied before these episodes start; (ii) sensitivity analysis with high-resolution modelling to evaluate the effectiveness of these potential abatement measures of precursors for O3 reduction.

Abstract. Ground-level and vertical measurements (performed using tethered and non-tethered balloons), coupled with modelling, of ozone (O 3 ), other gaseous pollutants (NO, NO 2 , CO, SO 2 ) and aerosols were carried out in the plains (Vic Plain) and valleys of the northern region of the Barcelona metropolitan area (BMA) in July 2015, an area typically recording the highest O 3 episodes in Spain. Our results suggest that these very high O 3 episodes were originated by three main contributions: (i) the surface fumigation from high O 3 reservoir layers located at 1500-3000 m a.g.l.
(according to modelling and non-tethered balloon measurements), and originated during the previous day(s) injections of polluted air masses at high altitude; (ii) local/regional photochemical production and transport (at lower heights) from the BMA and the surrounding coastal settlements, into the inland valleys; and (iii) external (to the study area) contribu-tions of both O 3 and precursors. These processes gave rise to maximal O 3 levels in the inland plains and valleys northwards from the BMA when compared to the higher mountain sites. Thus, a maximum O 3 concentration was observed within the lower tropospheric layer, characterised by an upward increase of O 3 and black carbon (BC) up to around 100-200 m a.g.l. (reaching up to 300 µg m −3 of O 3 as a 10 s average), followed by a decrease of both pollutants at higher altitudes, where BC and O 3 concentrations alternate in layers with parallel variations, probably as a consequence of the atmospheric transport from the BMA and the return flows (to the sea) of strata injected at certain heights the previous day(s). At the highest altitudes reached in this study with the tethered balloons (900-1000 m a.g.l.) during the campaign, BC and O 3 were often anti-correlated or unrelated, possibly due to a prevailing regional or even hemispheric contribution X. Querol et al.: Phenomenology of high-ozone episodes in NE Spain of O 3 at those altitudes. In the central hours of the days a homogeneous O 3 distribution was evidenced for the lowest 1 km of the atmosphere, although probably important variations could be expected at higher levels, where the high O 3 return strata are injected according to the modelling results and non-tethered balloon data.
Relatively low concentrations of ultrafine particles (UFPs) were found during the study, and nucleation episodes were only detected in the boundary layer.
Two types of O 3 episodes were identified: type A with major exceedances of the O 3 information threshold (180 µg m −3 on an hourly basis) caused by a clear daily concatenation of local/regional production with accumulation (at upper levels), fumigation and direct transport from the BMA (closed circulation); and type B with regional O 3 production without major recirculation (or fumigation) of the polluted BMA/regional air masses (open circulation), and relatively lower O 3 levels, but still exceeding the 8 h averaged health target.
To implement potential O 3 control and abatement strategies two major key tasks are proposed: (i) meteorological forecasting, from June to August, to predict recirculation episodes so that NO x and VOC abatement measures can be applied before these episodes start; (ii) sensitivity analysis with high-resolution modelling to evaluate the effectiveness of these potential abatement measures of precursors for O 3 reduction.

Introduction
Ozone (O 3 ) is an airborne secondary pollutant that is produced through the photo oxidation of volatile organic compounds (VOCs) in the presence of nitrogen oxides (NO x = NO + NO 2 ), with more intensive production in high insolation regions. It is well known that its formation processes are very complex and that the reaction and production rates are not linear (Monks et al., 2015, and references therein). According to EEA (2015), 97 % of the European population is exposed to O 3 concentrations that exceed the WHO guideline (see below) for the protection of human health. The complexity of this pollutant is also reflected in its air quality targets; thus, the European air quality directive 2008/50/EC establishes a number of O 3 target values (which are not legally binding, as opposed to the limit values set for the majority of pollutants): -A human health target value fixed at 120 µg m −3 as 8 h maximum in a day that should not be exceeded more than 25 days in a year as a 3-year mean. This target value was (arbitrarily) increased from the recommended 100 µg m −3 in the WHO air quality guidelines (where no specific number of exceedances is recommended).
-A population information hourly threshold of 180 µg m −3 .
-A vegetation protection target, AOT40 (expressed in µg m −3 h), as the sum of the excess of hourly concentrations above 80 µg m −3 in a given period using hourly values measured between 08:00 and 20:00, Central Europe Time (CET), for every day. Hourly AOT40 from May to July should not exceed 18 000 µg m −3 h O 3 as a 5-year mean. NO x has a catalytic effect in O 3 generation, and is only removed from the system by either deposition or oxidation to nitric acid (HNO 3 ) and reaction with VOCs to yield secondary aerosols, such as inorganic and organic nitrates, which may sequester a significant fraction of NO x . Consequently, O 3 generation involves not only local and regional air masses but also long-range transport. Thus, as a general observation, long-range transport of O 3 and its precursors influence markedly the background O 3 levels in Europe (UN-ECE, 2010;Doherty et al., 2013). However, this situation might be very different when considering the high summer O 3 episodes of southern Europe (e.g. Millán et al., 1997Millán et al., , 2000Palacios et al., 2002;Castell et al., 2008aCastell et al., , b, 2012Stein et al., 2005;Escudero et al., 2014;Pay et al., 2014;Querol et al., 2016).
In the western Mediterranean Basin the problem of tropospheric O 3 has been intensively studied since the early 1980s (Millán et al., 1991(Millán et al., , 1996a(Millán et al., , b, c, 2000Millán, 2002;Millán and Sanz, 1999;Mantilla et al., 1997;Salvador et al., 1997Salvador et al., , 1999Gangoiti et al., 2001;Stein et al., 2004Stein et al., , 2005Dieguez et al., 2009Dieguez et al., , 2014Doval et al., 2012;Castell et al., 2008aCastell et al., , b, 2012Escudero et al., 2014). Results have evidenced that (i) the meteorology driving O 3 fluctuation in this region is markedly influenced by a very complex orography, with high mountain chains surrounding the basin; (ii) in summer, the lack of a marked synoptic advection caused by the presence of the Azores anticyclone and the Iberian and North African thermal lows, together with the sea and land breezes, give rise to air mass recirculation episodes (lasting for several days); and (iii) during these summer vertical and horizontal recirculations of air masses loaded with O 3 precursors and coinciding with high insolation and elevated biogenic VOCs (BVOCs) emissions (Seco et al., 2011), high O 3 concentrations may be recorded.
Millán's team results demonstrated that western Mediterranean Basin dynamics are very different from those in central Europe. The latter are dominated by neutral-cloudy conditions, where O 3 episodes are usually associated with advection, and transformation takes place within large displacements of air masses. There, morning fumigation from a high O 3 residual (and stratified) boundary layer (BL) formed over the previous days, in addition to local formation in the sunny midday period may give rise to peak O 3 episodes if conditions persist after several days. In contrast, vertical recirculations developed over all western Mediterranean coastal areas, determine a very different O 3 dynamics. Air masses travel all the way from the sea to the continental divide, or to the top of the Apennines in the case of Italy. These air mass circulations create layers over the sea at various altitudes, with accumulated pollutants/precursors in several stages of transformation. These processes can occur during a few consecutive days (e.g. 10 days; Millán et al., 1997). The layers already over the basin descend from 1000-3500 m a.s.l. during the day and can reach the lower levels, providing a high background O 3 to coastal cities when the sea and upslope breezes build up (Millán et al., 2000). Layers, by definition, are stratified and decoupled from each other so they can move in different directions and speeds at their own heights. Rodriguez et al. (2002) and Cusack et al. (2013) showed that these high background O 3 episodes are characterised also by high particulate matter (PM) concentrations, mostly due to the formation of secondary organic and inorganic aerosols. Such episodes are very common from June to August, and are usually limited by the occurrence of episodic Atlantic or African advective conditions that help (especially the first) to ventilate the western Mediterranean Basin. Minguillón et al. (2015) demonstrated also the occurrence of very intense aerosol nucleation episodes (into the boundary layer) under high-insolation scenarios in the vertical column (from 200 to 1000 m a.s.l.) over the city of Barcelona. As the surface air ascends, aerosols are diluted and levels of O 3 are expected to increase.
The Barcelona metropolitan area (BMA) is a highly industrialised and dense urban agglomeration extending over the Mediterranean side of Northeast Spain. High anthropogenic NO x and VOC emissions arise both from road (and shipping) traffic and power generation, which, combined with BVOCs emissions, very often cause severe O 3 episodes in the northern plains and valleys (Toll and Baldasano, 2000;Barros et al., 2003;Gonçalves et al., 2009;Seco et al., 2011;Valverde et al., 2016;Querol et al., 2016). The urban plume is transported inland by sea breezes, heading north channelled by N-S valleys that cross the coastal and pre-coastal Catalan ranges to an intra-mountain plain (the Vic Plain) where the cities of Manlleu, Vic and Tona lie, 40-65 km north of Barcelona (Fig. 1). A mean of 15 annual exceedances of the hourly O 3 information threshold/site are recorded at the urban background monitoring sites of these cities (Querol et al., 2016). In 2015, 96 out of the 115 h exceeding the O 3 information threshold in the whole Catalonia air quality monitoring network were recorded in the area within 40-90 km north of Barcelona (towards the Pyrenees), and 82 in the Vic Plain itself (http://www.gencat.cat/mediamb/qaire/ciozo.htm).
This work focuses on an intensive campaign on O 3 and particulate pollutants performed in and around the BMA during July 2015, when high O 3 episodes were recorded, with the aim of investigating the origin of the most intense O 3 events in northeastern Spain. To this end, regional air quality monitoring network data, passive dosimeters at ground level, vertical profile measurements of O 3 and ultrafine particles (UFPs) in the Vic Plain, and modelling tools were employed.

Study area
This study is set in central Catalonia (Fig. 1), in northeastern Spain. The mountain ranges surrounding the area (Pyrenees and Catalan coastal ranges) protect the area from the advection of Atlantic and continental air masses but hamper dispersion of pollutants. The typical winds in the region are the tramontana (northern winds), the mistral or cierzo (northwestern winds channelled by the Ebro Valley) and the sea breezes in the coastal region. In summer, daytime upslope winds combined with sea breezes may result in air masses penetrating 120-160 km inland that are injected aloft the top of the mountains, and follow the return night flows towards the sea (Millán, 2014). This scenario of air mass regional recirculation during periods of several days prevails in summer (Millán et al., , 2000. Hence, summer pollution events are characterised by (i) the absence of large-scale forcing and the predominance of mesoscale circulations; (ii) the formation of a thermal low at a peninsular level (forcing the convergence of surface winds from the coastal areas towards the central plateau with strong levels of subsidence over the western Mediterranean Basin); and (iii) combined breeze dynamics, resulting in the recirculation and accumulation of pollutants over the whole western Mediterranean Basin, including the eastern Iberian Peninsula (Millán, 2014;Millán et al., 1997Millán et al., , 2000. The region is characterised by important atmospheric pollutant emissions from road traffic, industries, biomass burning, livestock, and airport and shipping activities, which coupled with high solar radiation turns into a high rate of secondary PM and O 3 formation (Rodriguez et al., 2002). Industrial activities are mostly concentrated in the Barcelona and Tarragona provinces, and include 19 combustion/energy plants, 84 metallurgy plants and 70 mineral industries. Road traffic, airport and shipping emissions are concentrated in the Barcelona area with > 3.5 × 10 6 vehicles (0.6 per inhabitant, with high diesel and motorbike proportions; DGT, 2014), > 45 × 10 6 tons of shipping transportation, and > 37 × 10 6 aircraft passengers in 2014 (Ajuntament de Barcelona, 2015).
Agriculture, livestock and biomass burning emissions are spread over the rural areas but concentrated in the core study area, the Vic Plain: a 30 km long depression in the northsouth direction located 60 km to the north of Barcelona. The area is surrounded by mountains and it is affected by thermal inversions during the night. The summer atmospheric dynamics dominated by sea breezes from the southern sector, channelled through the valleys formed by the coastal ranges, giving rise to the transport of pollutants from the BMA and the numerous surrounding highways.

Main valley
A-A ' Figure 1. Top: study area and location of monitoring sites (regional air quality monitoring sites XVPCA, dosimeters, meteorological stations and vertical measurements). BMA: Barcelona metropolitan area. Bottom: topographic profiles across the study area, red arrows point to the valleys connecting BMA with the Vic Plain and the pre-Pyrenean regions.

Online measurements of gaseous pollutants
Measurements of gaseous pollutants were performed at 48 sites belonging to the regional air quality network (Fig. 1, XVPCA; http://dtes.gencat.cat/icqa, Table S1 in the Supplement) from 1 to 31 July 2015. Continuous measurements of O 3 , NO, NO 2 , CO and SO 2 were carried out using MCV 48AV UV photometry analysers, Thermo Scientific chemiluminescence analysers (42i-TL), Teledyne 300 EU gas filter correlation analysers and Teledyne 100 EU UV fluorescence analysers, respectively.

Measurements of gaseous pollutants with passive dosimeters
Diffusion tubes for NO 2 and O 3 sampling (Gradko Environmental) were deployed at 17 locations between the cities of Barcelona and Ripoll ( Fig. 1) covering strategic areas not monitored by the regional air quality network. The dosimeters were positioned along two main river basins in the study area (Besòs/Congost and Tordera), from the BMA to the Vic Plain. Sampling points were selected avoiding the direct influence of vehicular emissions and located at a height of approximately 2.5 m above ground level. One sample per site and sampling period (1-14 and 14-29 July 2015) were collected. After exposure, samples were stored at 4 • C until analysis. Replicas were placed in nine locations, showing good reproducibility of the results (relative errors of 5 % ± 6 % for O 3 and 4 % ± 7 % for NO 2 ). Dosimeters were collocated also at some XVPCA sites for comparison with reference measurements (6 samples for O 3 at Vic (VIC), Montcada (MON) and Montseny (MSY); and 10 samples for NO 2 at Santa María de Palau Tordera (SMPT), Palau Reial (PLR), Tona (TON), Manlleu (MAN), and Granollers (GRA)). Correction factors were obtained from the comparison between dosimeter and reference data (Dosimeter O 3 = 1.01 · Reference O 3 + 17.43, R 2 = 0.97; and Dosimeter NO 2 = 1.27 · Reference NO 2 + 1.14, R 2 = 0.90, in all cases in µg m −3 ) and then applied to the whole dosimeter dataset (Fig. S1 in the Supplement).
NH 3 passive samplers (CEH ALPHA; Tang et al., 2001) were also used in 30 specific points.

O x concentrations
O x values (NO 2 + O 3 ) were calculated to complement the interpretation of O 3 concentrations. The concept of O x was initially proposed by Kley and Geiss (1994) to analyse the O 3 spatial and time variability by diminishing the effect of titration of O 3 by NO (NO + O 3 → NO 2 + O 2 , with the subsequent consumption of O 3 ) in highly polluted areas with high NO concentrations.

Laboratory van in Vic
A laboratory van was deployed next to the Vic air quality station during the period 10-17 July 2015. Eight-hour PM 2.5 samples were collected three times per day (00:00-08:00, 08:00-16:00, and 16:00-00:00 UTC) by means of Digitel DH80 high-volume samplers (30 m 3 h −1 ) on Pallflex quartz fiber filters (QAT 150 UP). Filters were conditioned at 20-25 • C and 50 % relative humidity over at least 24 h before and after sampling to determine gravimetrically the PM 2.5 concentration. Subsequently a detailed chemical analysis following the procedure described by Querol et al. (2001) was carried out.
Hourly equivalent black carbon (BC) in PM 2.5 concentrations were determined by a multi-angle absorption photometer (MAAP, model 5012, Thermo). PM 1 , PM 2.5 and PM 10 hourly concentrations were determined by an optical particle counter (GRIMM 1107).

Vertical profiles
During the period 14-17 July 2015 several vertical profiles up to 1550 m a.s.l. (1000 m a.g.l.) were performed by means of a tethered balloon (details can be found in Table S2) in the city of Vic (Fig. 1), at less than 200 m from the laboratory van and the Vic air quality station. The tethered balloon of 27 m 3 filled with helium was equipped with an instrumentation pack attached 30 m below the balloon. This setting has been successfully used in previous studies (Minguillón et al., 2015), hence the lack of a fixed support for the instrumen-tation pack is not expected to hinder the quality of measurements.
The instruments included in the pack were as follows: -A miniaturised condensation particle counter (Hy-CPC) measuring particle number concentration larger than 3 nm with a time resolution of 1 s and a flow rate of 0.125 L min −1 , using isopropyl alcohol as the working fluid (Lee et al., 2014). The particle number concentration measured by the Hy-CPC will be referred to as N 3 . Prior studies have demonstrated the agreement of the Hy-CPC and conventional TSI CPCs (Minguillón et al., 2015).
-A miniaturised nano-particle sizer for the determination of the particle number size distribution (Hy-SMPS, -A microaethalometer (MicroAeth AE51), which provided BC concentrations derived from absorption measurements on a 5 min basis with a flow rate of 0.15 L min −1 .
-A portable O 3 monitor that measures concentrations every 10 s based on UV absorption (POM ™ 2B Technologies, Fig. S2). The personal POM O 3 monitor was compared with the O 3 concentrations from the nearby reference station, yielding good results (n = 34 min data; POM O 3 = 0.85 · Reference O 3 + 0.56, R 2 = 0.93) (Fig. S1). The measured vertical O 3 concentrations reported in this study were normalised to standard temperature and pressure conditions (25 • C, 1013.2 hPa).
-Temperature, relative humidity, pressure, wind direction and wind speed sensors.
A further sounding was carried out on the 16 July 2015 at 11:00 UTC. A free balloon was used with an instrumentation pack equipped with a Hy-CPC, a GPS, a temperature sensor and a relative humidity sensor. The pack was placed in an insulated box (Fig. S4).
X. Querol et al.: Phenomenology of high-ozone episodes in NE Spain

Meteorological parameters
The 30 min meteorological data from 11 sites located throughout the study area in the proximity of air quality monitoring sites were provided by Meteocat (Meteorological Office of Catalonia) ( Fig. 1 and Table S3). Hourly average wind components were calculated and used in polar plots with hourly O 3 and O x concentrations, by means of the OpenAir software (Carslaw, 2012). These are bivariate polar plots where concentrations are shown to vary by wind speed and wind direction as a continuous surface.

Modelling system for O 3
The ambient O 3 concentrations were modelled using the ARAMIS (A Regional Air Quality Modelling Integrated System) high-resolution modelling system that integrates the Weather Research and Forecasting model (WRF-ARW) version 3.1.1 (Skamarock et al., 2008) as a meteorological model, the High Resolution Emission Model (HIREM) (Soler et al., 2004(Soler et al., , 2011(Soler et al., , 2015 and the Models-3 Community Multiscale Air Quality Modelling System (Models-3/CMAQ) (Byun and Ching, 1999) as a photochemical model. The modelling system was configured using four nested domains: D1, D2, D3 and D4 with horizontal grids of 27, 9, 3 and 1 km, respectively. For the coarser domain, initial and boundary conditions for the meteorological model were taken for the European Centre for Medium-Range Weather Forecast (ECMWF) global model with a 0.5 • × 0.5 • resolution using nudging and the boundary conditions are forced every 6 h, whilst for the photochemical model initial and boundary conditions model came from a vertical profile supplied by CMAQ itself. Domains are run in one-way nesting and 24 h spin-up was performed to minimise the effects of initial conditions for the inner domains. The output data are saved every hour. ARAMIS is continuously updated, and has been extensively evaluated (Soler et al., 2015) to simulate air quality over regional and local scales. In the present study the domain D3 was used, which covers the area of interest.

Meteorological background and diurnal cycles of pollutants
Two types of episodes that will be discussed in the following sections were identified concerning the meteorological patterns and the O 3 concentrations recorded. Type A episode. Under "usual summer conditions", with the Azores High located west of Iberia, and a ridge of high pressures extending into southern France, air masses in the western Mediterranean Basin rotate clockwise (anticyclonic) during the day, following the combined sea breezes and upslope flows at eastern Iberia and a simultaneous generalised compensatory sinking is observed in the basin. During night-time, drainage flows into the sea develop at the coastal strip, subsidence over the basin weakens and the wind over the sea is observed moving southward, transporting the coastal emissions almost parallel to the shoreline (Gangoiti et al., 2001). At the same time, Atlantic gap winds (through the Ebro and Carcassonne valleys), weaken during daytime due to inland sea breezes and become strengthened during nighttime Gangoiti et al., 2001Gangoiti et al., , 2006Millán, 2014). In such conditions, the air layers over the sea in front of Barcelona tend to move within the southwesterlies during the day, following the clockwise rotation, i.e. towards southern France and the Gulf of Genoa, and within the northerlies (towards Valencia) during the night. Thus, direct transport of O 3 and precursors from the Fos-Berre-Marseille-Piombino (Livorno) area towards the BMA is weak or null. However, indirect transport is more likely, first into the sea during nighttime conditions, and then following the daytime southwesterlies for the combined coastal sea-breeze and anticyclonic gyre at the coastal strip of Catalonia, which could bring a fraction of the referred O 3 and precursors originated in southern France, together with those emitted at the eastern coast of Iberia.
Type B episode. When the anticyclone establishes over central Europe with relative low pressure to the west over the Atlantic, the flow pattern over the western Mediterranean changes: southerly winds blow at height over eastern Iberia, while at ground level, gap winds may weaken or stop the Mistral and the tramontana winds in the Gulf of Lion, and Barcelona could then be directly affected by O 3 and precursors, coming with the easterlies blowing at the marine boundary layer (emissions from Corsica, Sardinia and Italy). However, under these atmospheric conditions O 3 levels did not reach the observed values found during type A episodes, and the O 3 daily records did not show the classical pattern of accumulation from one day to the next, characteristic of the highest O 3 episodes in the western Mediterranean (Millán et al., , 2000Castell et al., 2008a).
Under the above "usual summer conditions", Millán et al. (1991Millán et al. ( , 1996aMillán et al. ( , b, c, 1997Millán et al. ( , 2000Millán et al. ( , 2002, Gangoiti et al. (2001) and Castell et al. (2008a) demonstrated the vertical recirculation of O 3 -rich masses in the western Mediterranean, with O 3 being formed from precursors transported inland by the combined upslope and sea-breeze winds. O 3 loaded air masses, elevated by topography and sea-mountain breezes will be transported back to the coastal area at a certain altitude during the day and accumulates in elevated stably stratified layers at the coastal areas during the late evening and night. During nighttime and at ground level O 3 depletion dominates mainly in urban and industrial centres, driven by reaction with new emissions, which at the coastal area are transported offshore within the stable surface drainage flows.
The synoptic atmospheric situation in July 2015 was characterised by an intense high-pressure system over central and southern Europe during almost the whole month (Fig. 2). Type A and B scenarios alternated, transporting warm air masses from North Africa towards higher latitudes by the anticyclonic dynamic and reaching extremely high temperatures in Europe. The stagnation of air masses induced a regional meteorological scenario in the area under study, characterised by local/regional recirculations and sea-land breezes, both channelled by the complex topography. The flow pattern, together with the observed stably stratified layers developed up to a height of 2000-2500 m a.s.l. (Fig. 3) associated with subsidence, enhanced the accumulation of pollutants and caused several pollution episodes in the northeastern Iberian Peninsula. Coastal and pre-coastal locations (Barcelona) were mainly affected by daily sea breezes, starting blowing from the east (around 08:00 UTC) and turning progressively to the south and southwest. The sea breezes were channelled through the valleys, which are mainly located following a north-to-south axis, and arrived at the monitoring stations predominantly from a southerly direction. However, during the night atmospheric conditions were much more stable with flow patterns dominated by land breezes from NNW.
The VIC site was characterised by stagnant conditions during the day, reaching the maximum wind speed (4 m s −1 on average) at around 15:00 UTC, when sea-breeze inten-sity was at the highest (Fig. 4). During the night very light winds blew from the north (Fig. 4). During the periods 14-20 July 2015 (episode type A) and 3-6 July 2016 (episode type B) the sea breeze blew from 10:00 to 18:00 and 10:00 to 21:00 UTC, respectively, but in the first episode the wind speed was higher (maximum of 2.7 m s −1 as an average for the period) and maximal at 14:00 UTC, whereas in the second wind speed was lower, with a maximum mean value for the period of 2.4 m s −1 at 17:00 UTC, but only 1.5 m s −1 at midday.
Averaged ground O 3 concentrations during the type A episode recorded at VIC were clearly influenced by these wind patterns, showing a typical midday peak, followed by a higher peak at 13:00-14:00 UTC probably caused by the transport of BMA air masses by the breeze (Fig. 4). Mean O 3 levels during this A episode reached 195 µg m −3 at 13:00 UTC. During the type B episode average O 3 levels were also very high (142 µg m −3 at 14:00 UTC) but clearly lower than during the A episode (Fig. 4).
Intensive surface measurements were only available for 10-17 July 2015 (when the mobile laboratory was working at VIC). Average SO 2 levels for this period (included in the type A episode) showed a similar daily pattern to that of O 3 (Fig. 4) Figure 3. Data from the non-tethered balloon measurements (at VIC) of temperature, relative humidity and particle number concentrations performed from 10:00 to 11:30 UTC on 16 July 2015. Red lines identify the limit between different atmospheric layers.
process (Hewson, 1964;Geiger et al., 1992); when midday convective flows that abate the surface a SO 2 -rich layer accumulated in the limit of the boundary layer. Ground-level concentrations of BC, NO 2 and PM x showed a similar daily pattern driven by stagnation and traffic rush hour, with maximum concentrations around 06:00 UTC (08:00 local time, Fig. 4). Finally, extremely high concentrations of NH 3 (this is one of the most intensive farming regions of Spain and mean values of the Vic Plain dosimeters reached 30 µg m −3 NH 3 for 1-31 July 2015) followed the typical midday maximum due to evaporative emissions from fertilisers, but the rapid increase of the wind speed and dilution by the growth of the planetary boundary layer (PBL) thickness (see vertical profiles of temperature, aerosols and O 3 at VIC in following sections) probably account for a relative reduction of groundlevel NH 3 concentrations during the central hours of the day (Fig. 4).
The varying diurnal and nocturnal air mass patterns in the Vic Plain are also shown by the PM 2.5 chemical composition. Figure S5 shows the 8 h concentration patterns of selected components during the week period of 10-17 July 2015, including several days (14 to 17 July 2016) of the type A episode defined above, affected by polluted air masses from the BMA.
In addition to the regionally transported O 3 , concentrations of elemental carbon (EC) and traffic and industryrelated metals (including Zn, Cu, Pb, Sn and Sb) were notably enhanced at the end of the week, and were attributed to local sources. This enhancement was most obvious during the 00:00-08:00 UTC period (Fig. S5), under calm or northerly low wind (drainage slope winds) carrying metallic pollutants from the Cu smelter located 13 km to the north of Vic, and leading to high Cu, Zn, Sn, W, Pb and Sb concentrations on the nights of 15 and 16 July 2015 (Fig. S5). The increase in EC was related to local traffic emissions during the morning rush hour as deduced from the peaking MAAP BC concentrations during 05:00-08:00 UTC (07:00-10:00 local time), up to 5 µg m −3 hourly BC, when compared to 3 µg m −3 recorded as maximum traffic rush hour concentrations in the preceding days (data not shown). In contrast, the rise of organic carbon (OC) concentrations observed during 08:00-16:00 UTC is attributed to the formation of secondary organic aerosols.
Sulfate concentrations did not show any trend, as expected from secondary inorganic components present in relatively homogeneous concentrations on a regional scale, whereas nitrate (and partially of ammonium) concentrations increased during the 00:00-08:00 UTC periods as a result of gas to particle partitioning (Fig. S5) due to the thermal instability of ammonium nitrate (Hertel et al., 2012), under typical high daytime temperatures reached in July 2015. Interestingly, the stronger southerly winds during the daytime in the second part of the week (see below) appear to have brought polluted air from the BMA as signalled by slightly higher V concentrations (tracer of fuel oil combustion), but the fumigation from high strata (polluted air masses that were injected the previous day or days) might also account for these SO 2 and V increases.
The concentrations of mineral matter and all its components (Al, Fe, Mg, Li, Ti, Rb, Sr, Ti) were constant during the week, with relatively higher concentrations in the 08:00-16:00 UTC samples (Fig. S5), indicating a higher resuspension caused by stronger afternoon winds. The increment on the 15 July 2015 (08:00-16:00 UTC) was attributed to resuspension of local dust, given that the occurrence of African dust outbreaks was not observed during this period.
The free sounding measurements carried out at 11:00 UTC on 16 July 2015 revealed stratified air masses up to 3000 m a.g.l. (Fig. 3). The vertical profiles of potential temperature, water vapour and aerosol concentration distributions can be used for the identification of atmospheric layers presenting different properties: a lower layer up to about 1100 m a.g.l. characterised by a relatively high aerosol concentration, well mixed and with a relatively high and uniform water vapour content. A clear discontinuity between 1100 and 1500 m a.g.l. limits the mentioned lower layer and a series of stably stratified layers up to a height of 3000 m a.g.l. This layering of pollutants is probably related to the development of regional and mesoscale convective cells driven by the combined upslope and sea-breeze flows developed the day before . Figure S6 shows the average O 3 and O x ground-level concentrations recorded in July 2015 in the study area at the XVPCA air quality monitoring network and with the passive dosimeters. O x maximum concentrations were recorded at the Vic Plain area and in the coastal sites northeast of Barcelona. This may be due to the high O 3 concentrations in Vic and to a higher proportion of primary NO 2 (emitted mainly from diesel engines, and not formed in the atmosphere from NO titration by O 3 ) in the coastal cities, respectively.

O 3 and O x episodes
In July 2015, the O 3 hourly information threshold was exceeded a total of 74 times at the XVPCA stations of Catalonia, 57 taking place in the Vic Plain stations (TON, VIC and MAN), and 69 in the surrounding areas (pre-Pyrenees, High Llobregat River and Montseny).  (Fig. 5a) show relatively narrow diurnal variations and multiday episodes, with low or enhanced concentrations, according to meteorological fluctuations (accumulation and air mass renovation cycles of 3 to 12 days cause a wider O 3 and O x concentrations range than the typical daily cycles evidenced in most of the other sites). O 3 variations at the coastal BEG are opposed, in periods such as 1-3, 10-12 and 26 July 2015 and several periods from 14-20 July 2015, to those at the inland MSC. As shown by the polar plots from Fig. 6, relatively low O 3 concentrations (but still high in absolute terms) were recorded at the BEG coastal site (easternmost site in this figure) when the wind blows from the sea, whereas polluted air masses are transported towards the inland remote MSC (westernmost location in the figure) site under the same meteorological conditions. Conversely, when westerly winds blow, the inland remote MSC site received relatively clean air masses with low O 3 (Fig. 6), which are progressively loaded with regional pollution as these are transported towards the coastal BEG site.
Data from two urban background sites of Barcelona (PLR and CTL, 81 and 5 m a.s.l., grey and black in Fig. 5b) show evidence of a high nocturnal O 3 consumption, with differences due to local NO x traffic emissions. Following the trans-   Fig. 5d; 45, 55 and 62 km from BMA in a NNE direction, respectively) as well as at the remote eastern pre-Pyrenean site of Pardines (PAR, brown, 1226 m a.s.l., 102 km from BMA in a NNE direction), where O 3 formation and fumigation seems to have already reached its maximum, and similar O 3 concentrations were recorded at all sites during the midday increase. This suggests that the intensity of O 3 formation and fumigation was clearly reduced in the Vic Plain-Pyrenees transect with respect to the Barcelona-Vic Plain (an intermediate production place would be MSY (720 m a.s.l.; green in Fig. 7, 39 km from BMA in a NE direction). Polar plots of GRA, TON, MSY, VIC, MAN show clearly that the highest O 3 levels were recorded with wind blowing from the direction where BMA is located (Fig. 6).
As can be observed in Figs. 5 and 7, during two periods (1-2 and 7-20 July 2015) O 3 concentrations increased progressively from Barcelona city towards the northern BMA (GRA and MON), the intermediate MSY regional background area and towards the northern Vic Plain sites; and from there it slightly decreased towards the eastern pre-Pyrenees (PAR) following the midday-afternoon combined breeze transport (Figs. 5 and 7). During these days, no exceedance of the information threshold was produced in the urban environment; only sporadic peak concentrations above the human protection target value were recorded in the close surroundings of Barcelona. However, frequent exceedances of both thresholds were recorded in a regional transport context towards the north of the BMA.
While differences in O 3 concentrations between TON, GRA, MSY, BEG and CTL were observed during the period 3-6 July 2015 (type B episode), O x concentrations show a very similar behaviour along the Vic Plain, both qualitatively and quantitatively (Fig. 7, O x is not reported at BEG due to the lack of NO 2 measurements). Conversely, in the period 7-20 July 2015 (which includes the type A episode), characterised by a change in the synoptic conditions, differences in daily maximum O x values resemble the same behaviour of O 3 alone, with a positive and marked inland gradient. O 3 concentrations at BEG, a coastal site far in the northeast, were higher during the former period and showed low intra-day variation, indicating probable long-range transport of polluted air masses (Fig. 7).
O 3 and O x concentrations at the regional background site (MSY, 720 m a.s.l., green in Fig. 7) depict also the meteorologically influenced patterns (in the sense previously described for BEG and MSC), but with a clear overlapped and pronounced daily fluctuation, with marked higher concentrations indicating O 3 generation from a regional origin and fumigation from upper layers" especially on 1-2 and 7-20 July 2015 (Fig. 7).
Diurnal O 3 concentrations in the Vic Plain (around 460-620 m a.s.l.) were markedly higher than at the coastal (CTL, PLR) sites, and slightly higher than at mountain sites (MSY, PAR and MSC, from 720, 1226 and 1570 m a.s.l.) during the 1-2 and 7-20 July 2015 periods. The O 3 hourly information threshold of 180 µg m −3 was exceeded 55 times in the Vic Plain (three sites), with 50 of these exceedances taking place during 1 July 2015 and 14-20 July 2015. For these exceedances, an hourly contribution of up to 150 µg m −3 of O x (mostly O 3 ) both from fumigation of recirculated return layers (injected at an altitude of 1500-3000 m a.g.l. in the prior day(s)), and from transport and photochemical generation of O 3 of the BMA plume, might be estimated based on the differences of the O x early afternoon maxima recorded at the coastal BMA sites (CTL, PLR) and the ones in the Vic Plain (TON, MON, VIC). Thus, as shown in Fig. 7, the 14-18 July 2016 midday maxima recorded at CTL (into BMA) range between 38 and 62 ppb O x , on an hourly basis, whereas at TON (in the Vic Plain) they reach 102-115 ppb. Accordingly, differences of 50-73 ppb O x (close to 100-150 µg m −3 O x ) between CTL and TON can be estimated for these days. Furthermore, different intensity and duration of fumigation from upper high O 3 layers might also contribute to the higher inland O 3 surface concentrations since the PBL depth is much higher at the inland Vic Plain than at the coastal site, where sea breeze flows favour a lower growth of the PBL.

Type A episode (14-20 July 2015)
During this episode, a progressive time shift of the daily hourly O 3 and O x maxima was observed from the Barcelona  (Fig. 7), with maximum average O 3 hourly levels of around 200 µg m −3 . These O x differences are mostly due to O 3 differences (Fig. 8). Accordingly, during these intense O 3 pollution episodes, more than 50 % of the O x and O 3 hourly maxima concentrations are attributable to (i) O 3 contributions from the previously mentioned surface fumigation of recirculated strata (over the VIC-MAN-TON area) containing the polluted air masses injected the day before by complex topographically induced circulations, and to (ii) the local O 3 generation and surface transport of the BMA plume into inland valleys. Attributing these O 3 exceedances to local/regional causes is also supported by the spatial distribution of the hourly O 3 maxima, the number of hourly exceedances of the information threshold, the time shift of the exceedances at the different sites (as moving towards the north) (Fig. 9), and the polar plots of hourly O 3 concentrations pointing towards BMA as the main source region (Fig. 6). It is important to note that, as previ-ously stated, in the coastal sites the PBL height is markedly reduced when compared with the inland regions and then the capture of these high-altitude O 3 -rich layers by the PBL growth and the consequent fumigation on the surface is less probable in the coastal areas than in the inland ones.
Thus, during the A episode, O 3 has mostly a major local/regional origin (with O x maximum hourly levels progressively increasing from 166 to 246 µg m −3 from the BMA to the Vic Plain). The concatenation of daily cycles of regional/long-range recirculation of air masses and regional/local O 3 production in the A episode accounted for the accumulation of O x and the consequent exceedance of the hourly information threshold. Castell et al. (2008a) have already reported a correlation between their "recirculation factor" (a meteorological parameter devised to increase with the concatenation of days with regional vertical recirculation of air masses) with the occurrence of O 3 episodes in 2003. The relevance of these recirculations in originating these high O 3 episodes in southern Europe has been highlighted already, not only by scientific papers by the CEAM team but also assumed by the European Commission (EC, 2004). Figures 10 and 11 show results for the vertical profiles (0-1100 m a.g.l.) of O 3 concentrations, particle number concentrations for particles > 3 nm (red), 0.3-0.5 µm (blue), 0.5-1.0 µm (brown), ambient temperature, relative humidity and wind direction, obtained at the beginning of the type A episode (from 14 to 17 July 2015). In the profiles from 07:06 to 08:21 UTC on 14 July 2015, a boundary layer (150 to 250 m thick) with relatively high levels of N 3 (0.8 to 2.0 × 10 4 cm −3 ) was differentiated from the free troposphere (0.2 to 0.8 × 10 4 cm −3 ) (Fig. 10). However, in the profile obtained from 09:42 to 10:52 UTC on 17 July 2015, the growth by convective turbulence accounts for a homogeneous boundary layer and profile of N 0.3−0.5 below 1000 m a.g.l. (Fig. 10). Inside the PBL nucleation occurred (yellow to red areas in Fig. 12 for 16 July 2015) regionally driven by photochemical processes. Minguillón et al. (2015) showed the occurrence of these nucleation events into the PBL as convective transport elevates and dilutes air masses from polluted areas under high insolation in Barcelona. During the period 14-16 July 2015 nucleation episodes occurred occasionally, but only inside the PBL. On 17 July 2015 at 09:42-10:52 UTC new particle formation occurred probably at relatively high altitudes, also inside the PBL, as deduced from the high N 3 levels measured from 400 to 1000 m a.g.l., with concentrations reaching 1 × 10 4 cm −3 , while simultaneously low concentrations (< 0.3 × 10 4 cm −3 ) were measured at ground level (Fig. 10). This vertical gradient is not observed for the coarser particles (N 0.3−0.5 and N 0.5−1 ) and O 3 (Figs. 10 and 11, for which relatively constant levels were measured inside the PBL), suggesting new particle formation. On 14 July 2015 at 07:06-08:21 UTC a well-stratified atmosphere ( Fig. 11) with both thermal and O 3 layers is observed, with a general upward increasing trend for O 3 from 40 µg m −3 at ground level to much higher levels in different strata, one reaching 150 µg m −3 in strata at 500 m a.g.l. and others of 140, 100 and 40 µg m −3 , at 300, 800-1000 or 400 m a.g.l. respectively, reflecting, in addition to stratification of O 3 concentrations in altitude, the effect of surface depletion by NO titration and by deposition during the night (see in Fig. 11 the progressive O 3 depletion from 150 µg m −3 at 500 m a.g.l. to 40 µg m −3 at surface levels). From 13:49 to 15:03 UTC on 14 July 2015 (Fig. 11) the vertical profile changed substantially, with an already unstable atmosphere near the ground, showing very high surface O 3 concentrations of 217 µg m −3 that increase up to 330 µg m −3 in a layer around 100 m a.g.l., decreasing again through an upper layer with values of 240 µg m −3 until 300 m a.g.l. (where measurements were not available due to instrumental problems). This 100-200 m a.g.l. high O 3 layer agrees with the modelled O 3 concentrations for the study area (Toll and Baldasano, 2000;Barros et al., 2003;Gonçalves et al., 2009) and reflects elevated O 3 concentrations due to local production and transport of O 3 , that decrease from 100 m a.g.l. to the surface due to its titration, consumption and deposition. On 15 and 16 July 2015, a similar upward increasing O 3 gradient was observed in the early morning (Fig. 12). On 17 July 2015 at 07:39-08:40 UTC O 3 concentrations were relatively constant, but also showed a strongly stratified profile, in the range of 100-165 µg m −3 in the lower 500 m. In the last profile, from 09:42 to 10:52 UTC, O 3 concentrations increased from 140 to 200 µg m −3 for 200 to 1000 m a.g.l., but again a maximum close to 200 µg m −3 was observed at the same height around 100 m a.g.l. (Fig. 11).
Thus, vertical profiles of the type A episode are characterised in the early morning by a strong stratification, showing low ground-level O 3 concentrations, due to low production (low insolation) and/or consumption (titration and deposition), and increasing concentrations with altitude. This variation is related to prevailing meteorological conditions enhancing local recirculation or larger-scale transport with high O 3 masses injected (the day before) at certain altitudes by vertical recirculations into the residual layer, above the nocturnal surface stably stratified boundary layer. Nevertheless, during specific days, homogeneous O 3 vertical profiles up to 1000 m a.g.l. (the maximum height reached with captive sounding) were also evidenced, but probably not maintained at higher levels (where we were not able to measure with our system). Thus, as shown by the 4500 m profile measured with the free sounding on 16 July 2016 (Fig. 3), high PM (and probably O 3 ) strata are present between 1500 and 3000 m a.g.l., these being probably the polluted air masses injected the day before in the northern mountain ranges and recirculated to the coast at certain altitudes (see modelling outputs below). On the other hand, with constant southerly winds (from the coastal area to the Vic Plain) Interesting results are also obtained by comparing the vertical profiles of BC and O 3 (Fig. 13). BC is a tracer of local primary pollution at ground level, and of the potential transport and stratification of regional/local primary pollutants (together or not with regional O 3 ) when present at high altitude. On 14 July 2015 07:06-08:21 UTC, at 350 m a.g.l.
(and similarly for 15-17 July 2015 but at varying heights, 100-350 m a.g.l.) a clear discontinuity is evidenced with sudden and simultaneous decreases of BC and O 3 above these heights. The relatively high BC levels within the lower layer suggest nocturnal accumulation, while O 3 appears in strata Figure 12. Time variation of altitude, temperature, relative humidity, N 3 , particle number size distributions and O 3 concentrations during the tethered balloon measurements on 16 July 2015. 1-3 illustrate the nucleation episode recorded at surface level with particle number size distributions, and 4 the typical regional background N size distribution at about 300 m above ground.
(with low values near the ground due to titration and deposition) and with a high concentration just above that level (350 m), now with low BC concentrations. There is a further upward decrease of BC and an increase of O 3 up to the limit of the sounding (870 m).
The occurrence of an O 3 maximum layer around 100-200 m a.g.l., on top of the nocturnal stably stratified PBL, reinforces the idea of an important local production contributing to an upward increase of O 3 inside the layer. Finally, at the highest altitudes reached in this study (900-1000 m a.g.l.), BC and O 3 concentrations were often anticorrelated or unrelated, possibly more related with aged air masses recirculated within the whole region and with a mixed origin: including local-to-regional sources, more distant over the W-Mediterranean or even from hemispheric transport of air masses as reported by UNECE (2010). Figure 14 shows mean O 3 hourly concentrations recorded at VIC for the episodes A and B, as well as mean wind speed and direction. Mean hourly concentrations are characterised by an increase until 10:00-11:00 UTC, followed by an inflexion point and a more marked increase, with a maximum between 13:00 and 14:00 UTC, and then a progressive decrease, more marked in episode A. As stated above, processes contributing to increased levels were attributable to fumigation, photochemical production and transport of high O 3 air masses, all controlled by insolation. Millán et al. (2000)  cal production from precursors. On the other hand, the second O 3 concentration "hump" is coincident with maximum wind speed and probably corresponds to a more intensified sea-breeze transport compared with local photochemical formation and fumigation. Figure 14 shows that the two O 3 increases (and consequently the contributions from the three above processes) are more pronounced in the type A compared to the type B episode, and that the second maximum (more associated with inland surface transport by sea breeze) is wider, coinciding with a shift of the maximum wind speed towards the late afternoon, in the B episode.
Modelling outputs for the A episode point to light winds from the south, transporting pollutants from the BMA towards northern areas (including the Vic Plain), and triggering the hourly O 3 exceedances under the effect of the seaand land-breeze transport. Thus, Figs. 15 and 16 show the horizontal wind vector at 10 m a.g.l. and NO 2 and O 3 concentrations both at ground level and at a height above the surface layer, at different hours for two representative days of the type A and B episodes, respectively. During the type A episode day (15 July 2015), the effect of the land-breeze transport accumulates NO 2 over the sea during the night, starting intense O 3 production when sun rises and sea breezes start the inland transport. Maximum concentrations of O 3 , exceeding 180 µg m −3 were calculated by the model and measured at the stations located in the Vic Plain (TON, VIC and MAN, Fig. 16), although the model overestimated maximum O 3 concentrations in TON and VIC and delayed the hourly maximum value in all stations. The vertical distribution shows an important accumulation of around 110 µg m −3 trapped in a reservoir layer at around 1500 m a.s.l. during the   night (Fig. 17), which will fumigate downwards into the new developing mixing layer during the following hours. Local O 3 production from fresh precursors accumulated during the night in the stably stratified surface layer and then progressed inland along the midday hours. This results in an O 3 -enriched plume within a layer of 1000-1500 m depth in the late afternoon, following the model results (Fig. 17). This mixing layer also incorporates O 3 from upper reservoir layers after fumigation during the inland travel. The O 3 located at upper levels can recirculate back into the sea and will be potentially available to be transported inland , to start a new cycle the following day.

Type B episode (3-6 July 2015)
As opposed to the type A episode, during the type B episode and on 22-31 July 2015, despite the high O 3 and O x concentrations, the concentrations were very similar in the urban and remote coastal sites and all along the northern sites, including the Vic Plain. Hence, the averaged O x hourly concentrations of all the study sites were close to those at the coastal urban site in Barcelona CTL (and in the case of the O 3 close to the remote coastal site of BEG) compared with the large differences reported for the A episode (Fig. 8). The high O x peak measured at the urban site during the mornings of the B period (Fig. 8) and from 08:00 to 10:00 UTC in the average hourly patterns ( Fig. 9) is probably due to the contribution of primary NO 2 . According to Carslaw et al. (2016) the Euro 1 to Euro 2 diesel engines in Europe (early 1990s) emitted 5-10 % of primary NO 2 and 90-95 % of NO, whereas the Euro 4 to Euro 5 equivalent engines (2004 and 2009 onwards) emit 16-29 % of primary NO 2 and 71-84 % of NO.
Also as opposed to the A episode, during the B episode, O x levels varied in a very narrow range from east (coastal) to west (mountains, MSC site) and from south (BMA) to north (Vic Plain) and at different heights (from Barcelona and BEG at sea level to MSC at 1570 m a.s.l.). Following the results of the model in Figs. 15 and 17, O 3 does not recirculate around the region in this period. There is no accumulation from one day to the next in reservoir layers located over the region. Southerly winds blow at height during the whole period and the combined sea-breeze and upslope winds developed at lower layers during daytime, after coupling with the southerlies, and vent out the O 3 production and the rest of the pollutants to the north. The circulation is open, as opposed to the type A episode, which show a closed circulation (it is never completely closed) (Millán et al., , 2000. Unfortunately, vertical profiles of O 3 , UFP, PM and BC profiles were not obtained for this episode. Model outputs also evidence a net night and early morning transport of O 3 at lower layers from east and northeast during the B episode, supporting the hypothesis of a regional transport from southern France, advecting aged air masses to the whole region, while O 3 and its precursors from the BMA were transported during the morning to the southwest regions ( Fig. 15) giving rise to hourly O 3 exceedances in some stations situated in this area. Figure 15 also shows that during this episode (3 July 2015) the combined sea breeze and upslope wind transported O 3 and precursors to the western pre-Pyrenees area, and values lower than 180 µg m −3 were measured and modelled in all monitoring stations. The vertical distribution of O 3 also shows relatively low concentrations (as compared to 15 July) over most of the domain (Fig. 17).

Conclusions and implications for air quality
Very high levels of O 3 were recorded in the plains and valleys of the northern regions surrounding the BMA during July 2015, where 69 out of the 74 exceedances of the hourly O 3 information threshold measured in the entire air quality monitoring network of Catalonia were recorded. This represents a major environmental problem for which air quality managers must implement European and national legislation.
Both experimental measurements and modelling exercises suggest that these very intense O 3 episodes were originated by the concatenation of daily cycles of vertical recirculation of air masses that accumulated photochemically generated pollutants (Millán et al., , 2000Gangoiti et al., 2001;Castell et al., 2008a;Valverde et al., 2016), favoured by the high BVOCs and anthropogenic NO x emissions in the BMA region. The lower 1000 m a.g.l. were highly enriched in O 3 by fumigation from precursors and O 3 located at upper levels (1500-3000 m a.g.l). Additionally, local contributions of O 3 to these episodes were also demonstrated by soundings of the lower layers (0-1000 m a.g.l.). Thus, slightly higher concentrations of O 3 were measured in stations located at the plains and inland valleys than at higher altitudes (up to 30-40 µg m −3 exceeding the 180 µg m −3 maximum hourly concentration reached in the mountain sites), due to the local photochemical production from fresh precursors emitted during the night and early morning, and their channelling within the combined upslope and sea-breeze circulation that transports O 3 and precursors from the BMA. Thus vertical profiles identified a high O 3 layer at 100-200 m a.g.l., produced by these local processes and also by the high deposition and titration of O 3 at the lower 100 m depth layer. In our (mostly rural) study low concentrations of UFPs were recorded at this high O 3 100-200 m a.g.l. layer and nucleation episodes were only detected into the PBL, most of the days at the lower atmospheric levels.
Two types of high O 3 episodes (A and B) were identified in the area: Type A episodes are characterised by major local/regional O 3 recirculation, fumigation, production and transport, superimposed on the typical regional/long-range transport, and by the occurrence of major exceedances of the O 3 information threshold (14-20 July 2015). Surface fumigation from high O 3 return (to the sea) layers injected the day(s) before at altitudes of 1500-3000 m a.g.l., and recirculated over the VIC-TON-MAN area, as well as direct surface transport and formation of local/regional polluted air masses (with O 3 and precursors) from the BMA towards the north, decisively contributed to these exceedances. Thus, this atmospheric scenario is governed by poor ventilation under local breeze circulations and vertical recirculation of air masses over the study area.
Type B episodes are characterised by a major regional transport O 3 contribution, yielding similar O x levels at all monitoring sites of the study area, with the arrival of aged air masses from the east/northeast (high O 3 levels entering through the coast), but without major transport from BMA to the Vic Plain (3-6 July 2015), and without vertical recirculation of air masses over the study area. When sunlight activated atmospheric photochemistry in the early morning, the northern regions were loaded with air masses with lower content of O 3 precursors, since air masses were transported from the BMA to the southwest (parallel to the coast) from 00:00 to 09:00 UTC. The combined breeze at midday favoured transport towards the northwest, rather than to the north, as described for the type A episode. In addition, the aged air masses are not vertically recirculated and leave the region towards the northeast (to France). Thus, O 3 concentrations were still relatively high (exceeding 120 µg m −3 but below 180 µg m −3 ) due to local production from fresh precursors and background O 3 contributions from the western Mediterranean, but not enough to exceed the information threshold.
From the perspective of possible precursor abatement strategies, direct mitigation measures at the BMA would have had a minor effect on O 3 concentrations at the Vic Plain area during the type B episode. However, during the type A episode, a reduction of NO x and/or VOCs emissions in the BMA, some days before and during the episode, might have an effect on O 3 concentrations recorded in the Vic Plain. Nonetheless, due to the non-linearity of the O 3 generation processes, sensitivity analysis with high-resolution modelling is necessary to evaluate the possible effects in terms of absolute concentrations.
The use of O x data from strategically selected monitoring sites in the east coast, western and central mountain areas, urban background sites of the BMA and sites in the Vic Plain, tracking the natural routes of pollutant transport, is a useful tool to assess the different regimes leading to high O 3 concentrations, and to differentiate between type A and type B episodes, with important implications in the design of potential abatement strategies.
We are aware that we only analysed the most intense O 3 episodes occurring in July 2015, and that there might be other scenarios, different to type A and B, yielding high O 3 events, such as the transport of aged air masses from other regions of Europe or the transport of the BMA emissions in meteorological scenarios different to those described here. However, in a recent study (Querol et al., 2016) we demonstrated with the analysis of the 2000-2015 O 3 data series that the Vic Plain (40-50 km north of Barcelona) is the area in Spain recording the highest number of annual exceedances of the O 3 information threshold, orders of magnitude higher than the surrounding areas of the axis BMA-Vic Plain-pre-Pyrenean ranges. Thus it is clear that the BMA emissions and the vertical recirculations caused by the local complex orography have an important role in the occurrence and development of intensive O 3 episodes in the region.
To implement potential (and difficult) abatement strategies two major key tasks are proposed: 1. Meteorological forecasting from June to August to predict recirculation episodes in order to apply abatement measures for O 3 precursors before a recirculation episode starts. As stated above, the relevance of these recirculations in originating these high O 3 episodes in southern Europe has been assumed already by the European Commission (EC, 2004).
2. Sensitivity analysis with high-resolution modelling to evaluate the effectiveness of NO x and VOCs abatement measures on O 3 reduction.

Data availability
Hourly measurements of gaseous pollutants at the 48 monitoring sites belonging to the regional air quality network used in this paper can be downloaded from the regional air quality monitoring network (XVPCA) at http://dtes.gencat.cat/icqa. Table S1 in the Supplement gives details of these monitoring sites.
The meteorological data from the stations used can be requested free of charge from smc@meteo.cat. Table S3 in the Supplement gives details of these monitoring sites.
In an Excel file included in the Supplement, data from diffusion tubes for NO 2 and O 3 are reported. Finally, data from measurements with tethered balloons can be requested by mail from the corresponding authors of this paper.
The Supplement related to this article is available online at doi:10.5194/acp-17-2817-2017-supplement.