Introduction

Tropospheric ozone (O3) is a key air pollutant that harms human health and the environment, and it is an important greenhouse gas (Fowler et al. 2009; Myhre et al. 2014; GBD 2016; IPCC 2021; WHO 2021, 2013a, b). Additionally, it is a secondary atmospheric pollutant, 90% of which, globally, arises from photochemical reactions in precursors, particularly nitrogen oxides (NOx), volatile organic compounds (VOCs), including methane (CH4) and carbon monoxide (CO), and the rest from stratospheric contributions (McLinden et al. 2000; Olson et al. 2001; Stevenson et al. 2006; Young et al. 2013). Emissions of the two main O3 precursors, NOx and VOCs, may have different impacts on ground-level O3, depending on the local conditions and their relative mixing ratios (Sillman 1999).

NOx-limited conditions tend to occur at locations with low NOx emissions (i.e., rural environments or locations downwind of urban plumes and large sources) during periods of high photochemical activity, promoting enhanced O3 formation. Consequently, NOx-limited conditions are often associated with times and locations of high O3 levels, with increased NOx emissions leading to O3 increases, while increased VOC emissions may have limited effects (Sillman 1999; Sillman and He 2002; Sillman et al. 2003).

VOC-limited conditions typically occur in areas with high NOx emissions (e.g., urban centers or large pollution plumes) and under conditions of reduced photochemical activity. Such conditions are often associated with lower O3 levels (i.e., due to NO titration near high NOx emissions and/or low photochemical activity). In VOC-limited conditions, increased NOx emissions can result in localized O3 decreases, whereas increases in VOC emissions cause increases in O3 (Jacob et al. 1995; Sillman 1999; Sillman and He 2002; Sillman et al. 2003).

The application of emissions controls on O3 precursors in Europe, ranging from urban to the national and European scales, has resulted in a varied impact on ground-level O3. This is a consequence of the spatiotemporal variability of emissions changes in recent decades and the variability of O3 chemical formation regimes (Monks et al. 2015, Fleming et al. 2018). A review of studies on O3 trends by Sicard (2021) pointed to a general decrease in O3 levels in rural European areas from the early 2000s due to effective O3 precursor emissions control policies. However, these reductions in the precursors were insufficient to transition from VOC-limited to NOx-limited conditions in the cities. Consequently, there has been a general increase in O3 concentrations in urban environments, underscoring the necessity to apply effective control strategies to VOC emissions. Thus, decreased NOx emissions may to lead to both a reduced number of high O3 peaks and an increase in the minimum O3 values. This, in consequence, can lead to a narrowing of the O3 distribution (Simon et al. 2015), and ultimately drive a convergence between urban and rural O3 pollution in the long term (Paoletti et al. 2014; Yan et al. 2019). Moreover, at regional background-O3-monitoring stations, concentrations have increased gradually over the last few decades in the Northern Hemisphere (Sicard 2021). This can likely be attributed to a combination of factors, including the influence of climate change, increased stratospheric O3 intrusions, increased CH4 emissions, decreased NO titration due to reduced NOx emissions in production areas, and enhanced hemispheric transport (Monks et al. 2015; Sicard 2021, and related references).

Apart from changes in anthropogenic precursor emissions at the local, regional or global level, O3 trends are influenced by various interconnected factors, including changes in meteorology (radiation, temperature, transport patterns, etc.), shifts in biogenic emissions, the impacts of biomass burning (regionally and globally), and changes in stratosphere-troposphere exchange, among others (Monks et al. 2015; von Schneidemesser et al. 2015).

The Southern European regions, and especially the Western Mediterranean Basin, are the most exposed to O3 pollution in Europe, where Spain and several other countries systematically exceed the legal thresholds for the protection of health and vegetation (EEA 2021a).

In a companion study, Massagué et al. (2023) examined the trends in several relevant ground-level O3 metrics in Spain covering 2008 − 2019. The selected period was framed between two major events that had a substantial impact on global emissions––the 2008 financial crisis and the COVID-19 pandemic in 2020 (Peters et al. 2011; Castellanos and Boersma 2012; Sokhi et al. 2021). The study aimed to classify the atmospheric regions in Spain based on their O3 pollution patterns and identify the main O3 hotspots. These included the Madrid region, the interior of the Valencian Community, the northern (downwind) areas of Barcelona, the Guadalquivir Valley, which includes the urban area of Sevilla, and the closed basin of Puertollano.

The O3 trends in these areas exhibited marked differences: (i) the Madrid region showed the highest number of upward trends across all metrics, often with the highest increasing rates, indicating increasing O3 levels associated with both chronic and episodic exposure; (ii) the Valencian Community had a mixed variation pattern, depending on the metric considered; (iii) areas situated downwind of Barcelona, the Puertollano Basin, and the Guadalquivir Valley generally did not show O3 trends, but (iv) the urban area of Sevilla stood out as the only major city in Spain with generalized decreasing O3 trends.

Building upon this work, in the framework of the development of a Spanish Ozone Mitigation Plan, this study was aimed at exploring possible causes for these contrasting O3 trends through the analysis of changes in meteorological parameters relevant to O3 and available information on O3 precursors.

Methodology

Study area

This study was focused on mainland Spain and the Balearic Islands (Fig. 1a). Figure S1.1 provides general information on the land cover and use, demographics, and main climatic characteristics of the study area. Madrid, Barcelona, Valencia, and Sevilla are the most populated metropolitan areas in Spain. The northern and north-western regions in Spain have the lowest O3 levels due to meteorological conditions that are not favorable for O3 production (Gangoiti et al. 2002, 2006; Saavedra et al. 2012) (see Figure S1.1c–f). Conversely, in the central, southern, and Mediterranean areas, O3 levels tend to be higher due to the occurrence of large anthropogenic and biogenic precursor emissions, the prevailing meteorological conditions during the warm seasons, and the characteristic orography that favors the production and accumulation of O3 (Millán et al. 1997, 2000; Gangoiti et al. 2001).

Fig. 1
figure 1

a O3 hotspots or areas relevant to this study––yellow (fully described in Querol et al. 2016), autonomous regions––white, and important cities mentioned in this study––red. b Summary of the main O3 trends for the main hotspots for 2008–2019 (from Massagué et al. 2023). The size of the arrows provides an indication of the magnitude and number of the trends observed. Upward arrow––O3 increasing, downward arrow––O3 decreasing, and gray block––no O3 variation

Full size image

The areas frequently impacted by the most intense episodes of O3 in Spain––or O3 hotspots (Fig. 1a)––have been described in Diéguez et al. 2009), Querol et al. (2016), and later updated by Massagué et al. (2023). The O3 phenomenology in these areas has been studied previously––the Madrid metropolitan area (e.g., Plaza et al. 1997; Querol et al. 2018; Reche et al. 2018; Escudero et al. 2019), northern Barcelona (e.g., Toll and Baldasano 2000; Querol et al. 2017; Massagué et al. 2019), the Guadalquivir Valley (e.g., Notario et al. 2012; In ’t Veld et al. 2021; Massagué et al. 2021), the Valencian Community, generally including the O3 dynamics affecting the Western Mediterranean Basin (WMB) (e.g., Millán et al. 1997, 2000; Gangoiti et al. 2001), and the Puertollano Basin (e.g., Saiz-Lopez et al. 2009; Notario et al. 2013).

Figure 1b schematically illustrates the main O3 trends of the various hotspots during the period 2008 − 2019, as described in the “Introduction” section.

Data and methods

We utilized multiple sources of data, encompassing: (i) concentration data on the available O3 precursors, including ground-level measurements of NO2 and NO, satellite observations of NO2 and formaldehyde (HCHO) tropospheric columns; (ii) emissions data on the main O3 precursors, including NOx, VOCs, CO, and CH4, obtained from official inventories; and (iii) reanalyzed data of several meteorological parameters relevant to O3. For all data except the emissions data, we considered annual aggregates for the period April to September (i.e., O3 season), which corresponds to the main vegetation growing season and is when the O3 levels are typically at their maximum in Europe (EC 2008; Langner et al. 2012; Gaudel et al. 2018; Mills et al. 2018). Additional information can be found in Section S2.

Ground-level concentrations of pollutants

We used NO2 and NO hourly data series from all the air-quality monitoring stations (AQMSs) in Spain that reported to the European Council based on Decision 2011/850/UE and were active during at least one of the last two years in the study period (2018 or 2019).

Following the same criteria as for the analyses in Massagué et al. (2023), for the assessment of the present-day distribution (2015–2019), each AQMS was required to have had at least three years’ worth of valid data for that period, and for the trend estimates (2008–2019), at least 10 years’ worth of valid data (Section S2.1). To robustly detect and estimate the trends, we used the non-parametric Mann–Kendall test, along with the Theil–Sen statistical estimator (hereafter, MK–TS) (Theil 1992; Sen 1968). To this end, we used the R package Openair (Carslaw and Ropkins 2012; R Core Team 2021) to obtain the regression parameters for the trends (slope, uncertainty, and p-value), estimated via bootstrap resampling, and considered them to be statistically significant if p < 0.05.

Based on the data availability constraints, we used data from 414 and 319 AQMSs for the present-day and trend assessments, respectively (see detailed information in Figure S1.2 and Tables S1.1 and S1.2).

Remote sensing

To characterize the spatiotemporal variability of the tropospheric NO2 (used as a proxy for NOx emissions, Liu et al. 2016) and HCHO (used as a proxy for the total VOC reactivity, Martin et al. 2004), we employed spaceborne observations provided by NASA’s ozone monitoring instrument (OMI) aboard the Aura satellite (OMI Team 2012). The OMI follows a sun-synchronous orbit, with a daily overpass at approximately 13:45 h local solar time, and has a resolution of 13 × 24 km2. The area assessed included the entire Iberian Peninsula, southern France, and northern Africa. We used official NO2 and HCHO monthly data obtained from the Quality Assurance for Essential Climate Variables website (QA4ECV) (http://www.qa4ecv.eu, refer to Section S2.2), which was developed within the framework of the EU FP7 project (Boersma et al. 2017; De Smedt et al. 2018). Data is stored in uniform global grids with a 0.25° × 0.25° spatial resolution.

With these monthly data, we calculated April–September averages of tropospheric NO2, HCHO and the HCHO/NO2 tropospheric column ratio, used sometimes as an indicator of the O3 sensitivity regime (Li et al. 2021), for every year for each pixel. Then, we calculated the present-day spatial distribution and estimated the trends.

Emissions of O3 precursors

We used data on national emissions of relevant O3 precursors: (i) NOx, VOCs (non-CH4), and CO contained in the EU emissions inventory report 1990–2020 (EEA 2022) under the UNECE Convention on Long-range Transboundary Air Pollution; and (ii) for CH4, from the national greenhouse gas inventories submissions to the UNFCCC and the EU Greenhouse Gas Monitoring Mechanism (EU Member States). More information can be found in Section S2.3.

We estimated the trends in the emissions using the Openair MK–TS for 2008–2019. Firstly, we evaluated the Spanish emissions in a European context and, secondly, at a national level with data disaggregated by main emissions sector (using the EEA sector classification; see EEA 2022).

Meteorological parameters

To complement the analysis with other potential factors that might have had an effect on the surface O3 trends, we used the fifth generation of the European Centre for Medium Range Weather Forecasts (ECMWF) global reanalysis (ERA5). We evaluated monthly averaged meteorological data from the ERA5 reanalysis dataset, which provides a continuous dataset in time and space with a regular 0.25° × 0.25° grid (Hersbach et al. 2019). We selected nine meteorological parameters that might have impacted surface O3 (Jacob and Winner 2009; von Schneidemesser et al. 2015; Coates et al. 2016; Otero et al. 2016; Lefohn et al. 2018; Wei et al. 2022). These include temperature, downward surface solar radiation, downward UV radiation at the surface, cloud cover, boundary layer height, evaporation, surface wind speed, surface pressure, and total precipitation. Refer to Section S2.4 for information on calculations.

For each grid pixel, we calculated the annual April–September averages in order to determine the present-day (2015–2019) spatial variation, and to estimate the 2008–2019 trends. The trends and level of significance were obtained by adapting the script to calculate the trends for the raster images from Abdi et al. (2019) to be used with the Openair MK–TS. It should be noted that a 12-year dataset is presumably of limited duration for the detection of climatological trends because, traditionally, at least 30 years’ worth of records are needed to identify a genuine trend in climate. Nevertheless, we considered that the statistically significant changes detected, even within this relatively short timeframe, could provide valuable supplementary insights.

Results and discussion

Meteorological parameters

Section S3 presents the results of the analyses concerning meteorological parameters that could have influenced O3 concentrations during the warm season. Most of the variations observed within the national territory for the period 2008–2019 did not exhibit statistical significance (Figure S3.1), probably due to the relatively short duration of the dataset when viewed from a meteorological/climatological standpoint. Thus, we cannot disregard the importance of exploring other potentially significant trends that could involve examining additional meteorological parameters, employing different and/or less strict statistical approaches, considering alternative temporal aggregations, or extending the analysis over longer timeframes.

Firstly, we briefly discuss the (non-statistically significant) variations in meteorological parameters (see central panels in Fig. 2 and Figure S3.1). A temperature increase was observed practically all over the country (Fig. 2b), consistent with the findings of Borge et al. (2019) for 1993–2017. Moreover, there was a general increase in solar radiation (Figure S3.1b and e), in line with Pfeifroth et al. (2018), and with greater intensity in areas of the south-west, the Mediterranean coast, and northern parts of the country, in logical consonance with the decrease in cloud cover observed over these areas (Figure S3.1 h).

Fig. 2
figure 2

April–September surface temperatures at 2 m altitude. a Present-day (2015–2019); b variation (2008–2019); and c statistically significant trends (calculated from ERA5 monthly averaged data). See Section S3 for the rest of the meteorological parameters

Full size image

The thickness of the planetary boundary layer (PBL) increased, especially in the interior of the country (Figure S3.1 k), which may have had an effect on the amount of stratospheric and free tropospheric O3 influencing the ground-level O3 concentrations. An increase in evaporation was also observed over much of the interior (Figure S3.1q), especially in Castilla-La Mancha and eastern Andalucia. Precipitation showed a decrease throughout the country, in agreement with Borge et al. (2019) for 1993–2017, except in dispersed areas of the north and north-east and inland (Figure S3.1w). The surface wind showed no clear spatial patterns, with regions of the Mediterranean Sea exhibiting decreases, but with no variation, or with slight increases, in inland areas (Figure S3.1n). Azorin-Molina et al. (2016) found wind speed decreases for 1961–2011 in Spain, in line with Garrido-Perez et al. (2018, 2019), that indicated that stagnation, which is an important O3 driver, increased in Southern Europe in 1998–2015.

Most of the observed variations in the meteorological parameters for the interior and central areas may have been favorable for O3 production (Jacob and Winner 2009; von Schneidemesser et al. 2015; Otero et al. 2016; Coates et al. 2016; Lefohn et al. 2018; Wei et al. 2022). This may be consistent with the general increase in O3 concentrations in central Spain found in Massagué et al. (2023). The extent to which such non-statistically significant variations (or others) could have influenced the O3 concentrations during the study period requires further analysis.

Although the overall temperature increase across the country was not statistically significant, it should be noted that this parameter exhibited a statistically significant increase in most parts of the Madrid region, as well as in some large areas of western CastillaLa Mancha and other dispersed regions (Fig. 2c). This is consistent with the results of Borge et al. (2019) who reported a temperature increase across the country (1993–2017) during the warm season, with the central regions experiencing the greatest rate of increase.

Temperature is the meteorological parameter with the greatest potential impact on ground-level O3, as supported by studies conducted by Jacob and Winner (2009), Pusede et al. (2015), and Wei et al. (2022). Thus, ground-level O3 is strongly correlated with surface temperature, particularly in highly polluted areas (Porter and Heald 2019). Increasing temperatures can enhance O3 concentrations by accelerating chemical reactions, thereby increasing emissions of biogenic VOCs (BVOCs) (Sillman and Samson 1995; Coates et al. 2016) or through other mechanisms (see Porter and Heald 2019 and reference therein). It should also be considered that, although anthropogenic VOCs (AVOCs) are usually not temperature dependent, evaporative emissions of certain AVOCs can increase with rising temperatures (Rubin et al. 2006). Furthermore, Borge et al. (2019) also determined that weather changes consistent with our findings, including temperature increase, caused increments in O3 (and other pollutants) in Spain.

O3 precursors

Present-day O3 precursors

The present-day distribution of ground-level NO and NO2 in April–September is shown in Figure S1.3. As expected, the highest concentrations were found in high-traffic environments, especially in large metropolitan areas, such as Barcelona and Madrid, and, to a lesser extent, in northern industrial locations. The lowest concentrations were found in rural, and especially regional, background sites (Figure S1.3a, b, e, and f). Although these concentrations were evaluated during April–September—naturally lower than during colder months—the NO2 concentrations at many stations were still above the annual NO2 limit of both the Directive (40 µg·m−3) (EC 2008) and, obviously, the latest, stricter annual WHO Guideline (10 µg·m−3) (WHO 2021), with 65% of the AQMSs recording exceedances of this.

The present-day NO2 tropospheric column spatial distribution (OMI data), often used as a proxy for NOx emissions (Liu et al. 2016), is shown in Fig. 3a. The highest concentrations were recorded in large urban areas, such as Madrid, followed by Barcelona (up to approximately 4 × 1015 molecules·cm−2), and, with levels below half, Valencia–Castellón and Sevilla. Significant tropospheric NO2 levels were also recorded in the Gibraltar Strait and the Alborán Sea, probably caused by heavy maritime traffic (Nunes et al. 2020), a large maritime area in the south of Marseille, and in the northern industrial areas of Oviedo–Gijón, Coruña, and Bilbao.

Fig. 3
figure 3

Spatial distribution of NO2 and HCHO tropospheric columns for April–September measured by OMI-NASA and HCHO/NO2 ratio. Each pixel covers an area of 13 × 24 km2. a, d, g Present-day (2015–2019) spatial variability; b, e, h absolute trends for 2008–2019; and c, f, i absolute trends for 2008–2019, showing only pixels with statistically significant (p < 0.05) trends. Gray pixels––p ≥ 0.05. Relative variations (in %) can be found in Figure S1.4 in the Supplemental Information. Map 3b shows the locations of the power plants operational in 2018 (from Byers et al. 2021)

Full size image

To partly overcome the lack of detailed in situ VOC measurements, we here briefly discuss the variability in the HCHO tropospheric columns obtained from the OMI (Fig. 3d). HCHO is a frequent by-product of the oxidation of VOCs, and is thus sometimes used as a proxy for the total VOC reactivity (Martin et al. 2004). A common issue of current space-based HCHO observations lies in their relatively weak signal-to-noise ratio, which greatly limits their interpretation, except in regions of the world with strong HCHO levels (e.g., the tropics, eastern USA, and Asia). However, when focusing on April–September (the seasonal maximum for HCHO), some spatial variability can still be observed over the Iberian Peninsula, with tropospheric columns ranging between 5 and 10 × 1015 molecules·cm−2. The lowest values were observed over the central Pyrenees and in specific rural areas, whereas highest values occurred predominantly over northeastern Spain and Portugal. The highest HCHO in the northeast was probably a combination of high isoprene emissions and the transport of aged high-O3 air masses from the coast. Overall, no clear HCHO hotspots were observed over the main Spanish urban areas, which suggests a predominantly biogenic origin (including, for instance, the oxidation of biogenic isoprene), although some smaller and/or weaker hotspots could be hidden by the aforementioned signal-to-noise ratio (and the still-coarse spatial resolution of the OMI observations).

Given the weaker spatial variability of HCHO compared to NO2, the HCHO/NO2 tropospheric column ratio tended to follow a similar spatial distribution as 1/NO2 tropospheric columns (Fig. 3g). Although this ratio is sometimes used as an indicator of the O3 sensitivity regime (NOx-limited versus VOC-limited) (Li et al. 2021), the threshold distinguishing both regimes can vary substantially in time and space, and undefined regimes can prevail over a rather large range of HCHO/NO2 values, which limits its interpretability (Souri et al. 2020, 2022). Nonetheless, the present-day mean April–September HCHO/NO2 values ranged between 2 and 11, suggesting a predominantly NOx-limited regime over Spain, with NOx most strongly limited over the Spain–Portugal and Spain–France (Pyrenees) border regions. Lower values of this ratio were logically found over large NO2 hotspots. It is worth noting that these results were obtained in the early afternoon (the OMI overpass being at around 13:45 h local solar time), thus not during the morning rush hours when BVOC emissions tend to be lower and NOx emissions higher, and thereby causing the HCHO/NO2 values to significantly change.

Trends of O3 precursors

The assessment of ground-level NO and NO2 (April–September) trends is shown in Figure S1.3. The results showed that a relevant proportion (39%) of the AQMSs recorded downward NO trends across the country in 2008–2019 (Figure S1.3c), with the largest absolute decreases mostly occurring in northern locations and a few specific sites in Barcelona, especially in the urban and high-traffic environments (Figure S1.3d). In the Madrid area and in eastern and southern areas (including Sevilla and Valencia), the absolute average NO decreases were weak. However, a detailed comparison between Madrid and Barcelona (Table S1.3) showed that, on average, the NO absolute decreases were similar (− 0.5 µg·m−3·year−1), although, in relative terms, the NO decreases were more intense in Madrid (− 5.0%·year−1) than in Barcelona (− 3.5%·year−1), considering all types of AQMSs, and also only traffic. Similarly to the NO, 40% of the stations recorded downward NO2 trends (Figure S1.3 g), most of them located in the northern half of the country. The largest absolute decreases were also mostly recorded in the northern areas, including Barcelona. In Madrid, a large proportion of AQMSs did not record significant trends. Again, in absolute terms, the average decreases were similar in Madrid and Barcelona (− 0.8 µg·m−3·year−1) (Table S1.3), although in relative terms, the NO2 decreases were more intense in Madrid (− 3.0%·year−1) than in Barcelona (− 2.4%·year−1). Throughout the Mediterranean coastal area (except in Catalonia), practically no NO2 downward trends were recorded. In Sevilla and along the Guadalquivir Valley, several AQMSs recorded NO2 downward trends, but these were of very low intensity. As for NO, the strongest NO2 decreases were mainly driven by changes in traffic (either quantity or type), as these were observed especially at traffic stations (Figure S1.3 h and Table S1.3).

Figure 3b shows the variation in the April–September tropospheric NO2 for 2008–2019, while Fig. 3c provides an estimation of the statistically significant trends (p < 0.05) (see Figure S1.4 for relative variations). The Oviedo–Gijón and León areas in northwestern Spain have been historically highly industrialized, with the highest density of coal-powered plants in the country (indicated in Fig. 3b). These plants were associated with high tropospheric NO2 levels (Cuevas et al. 2014). In the last decade, many of these plants have reduced/stopped their operations or have applied measures to reduce their atmospheric emissions (e.g., Directive 2010/75/EU) (REE 2021), potentially contributing to two outcomes: firstly, the clear decrease as shown in Fig. 3b in the northwest (similar to findings by Castellanos and Boersma 2012), probably statistically insignificant because of sudden changes in NOx emissions (e.g., from the closure of large power plants). Secondly, the explanation for why these areas do not appear as NO2 hotspots in the present-day (2015–2019) (Fig. 3a). Additionally, decreases in the NO2 tropospheric columns were recorded in highly industrialized and/or high-traffic areas, such as Barcelona–Tarragona, the entire Ebro Valley, País Vasco–Santander, and the eastern coastal strip, likely as a result of implemented control measures. According to the European Environmental Agency’s Emission Inventory Report (EEA 2022), NOx declining emissions in Spain were driven by three key sectors: (i) the energy sector through the adoption of renewable energy, abatement techniques in thermal power plants, and transitioning to combined cycle gas plants; (ii) road transport, due to Euro standards implementations; and (iii) the industry sector, driven by abatement techniques and a shift to natural gas, particularly in non-metallic minerals. For detailed information on NOx and other O3 precursor emissions in Spain, see MITERD (2023), and refer to Section S4 for relevant emissions abatement policies.

Most of the downward tropospheric NO2 trends were statistically significant. The most marked declines were mainly found in/around Madrid, both in absolute (Fig. 3c) and relative (Figure S1.4 a, b) terms. To a lesser extent, similar declines were observed in/around Barcelona, with a maximum absolute decreasing rate approximately 30% lower than that in Madrid. This pattern was followed by industrial/urban areas like Oviedo, País Vasco and Cantabria, and part of the Ebro Valley (that includes a major coal-fired plant), with maximum absolute decreasing rates roughly 60% lower than in Madrid. Statistically significant downward trends on tropospheric NO2 were also detected in Sevilla and Puertollano, in this latter area despite the reduced area of the closed basin and the relatively coarse spatial resolution of OMI. It is relevant that, in Barcelona city, the decreases found were weaker than in other areas with similar present-day levels of tropospheric NO2, such as Madrid (Fig. 3a).

At first glance, the variations in satellite-based and ground-level NO2 show similarities. For example, in qualitative terms, the tropospheric NO2 decreases in large areas of the north, the Ebro Valley, and the large urban areas of Madrid and Barcelona are consistent with those from the ground-level measurements. However, over large parts of the Mediterranean coast, where tropospheric column NO2 levels also clearly declined (Figs. 3c, S1.4b), the AQMSs mostly did not detect any NO2 trends (Figure S1.3 g). The measurements obtained using these different tools are not directly comparable here, but are complementary (Bechle et al. 2013), since their spatial and temporal representativeness are different. For instance, ground-based observations provide concentrations at specific locations that can be highly influenced by nearby sources, whereas satellite-based observations measure average concentrations integrated over specific land areas (several square kilometers) and in the vertical. Similarly, satellite-based observations have a daily frequency, typically in the afternoon (subject to cloud cover limitations and associated biases), while ground-based observations are daily averages derived from 24-h data. In any case, to support the interpretation of O3 trends, a wider representation of the NO2 load in a large air mass is desirable.

Over the same period, the statistically significant trends affecting the HCHO tropospheric columns were mostly positive, albeit scarce and dispersed over the peninsula, with no clear spatial patterns discernable (Fig. 3f). The aforementioned noise affecting these measurements added uncertainties, making it difficult to detect significant trends over the relatively short time period of 12 years. The highest positive HCHO trends were observed in northwestern Spain. Similarly, the significant HCHO/NO2 ratio trends were dispersed and mostly positive (Fig. 3i). In large cities like Madrid or Barcelona, this increase might have been partially driven by the reduction in NO2, whereas in northwestern Spain, it is likely to have been driven by the increase in HCHO. Overall, under the assumption that HCHO tropospheric columns are a reliable proxy for total VOC reactivity, these results suggest that, at least in the early afternoon when the OMI observations are available, NOx tends to play an increasingly important, albeit perhaps not entirely limiting, role in the formation of O3 over Iberia.

The mostly increasing HCHO trends are qualitatively consistent with those of Opacka et al. (2021), who found positive isoprene and HCHO trends over Iberia (2001–2016), attributing them to changes in the land cover. However, as background O3 concentrations have increased (Sicard 2021), and HCHO can be generated from VOCs (including isoprene) (Wolfe et al. 2016) by O3 oxidation, it is plausible that this global O3 increase contributed to the HCHO increase.

Trends in emissions of the main O3 precursors

Although inventories only consider emissions of anthropogenic origin, it should be noted that biogenic emissions of precursors can have a large impact on O3 (Monks et al. 2015). For instance, among the most important O3 precursors, and at a global scale, 90% of atmospheric VOCs are biogenic, isoprene, and monoterpenes being the greatest contributors (Guenther et al. 1995), or 15% of global NOx emissions originate from soils, which can be a significant source of the NOx budget outside the cities (Weng et al. 2020). However, at local/regional scales, these contributions can change dramatically (Sartelet et al. 2012), for instance, in situations where O3 pollution episodes occurr in Spain, mostly in and around highly populated and/or industrialized conurbations (Querol et al. 2016).

The inventories showed that, in 2008–2019, emissions of the main anthropogenic precursors of O3 evolved differently in the EU-28 and Spain (Fig. 4a, b). In the EU-28, all compounds declined (− 1.3 to − 2.9%·year−1). However in Spain, NOx steadily declined (− 3.4%·year−1), but AVOCs, CO, and CH4 increased from 2014, probably indicating a rebound of emissions after the financial crisis (Pacca et al. 2020), although they slightly declined overall in 2008–2019, following no statistically significant trends (only p < 0.1 for CH4). The AVOC/NOx ratio in Spain significantly increased during this period, by 1.6 times from 2007 to 2019, which is in general agreement with the trends observed for the HCHO/NO2 ratio (Fig. 3i). A breakdown of the variations in emissions by region is unavailable at present, but information on the emissions sectors at the national level is accessible (Fig. 4c–f).

Fig. 4
figure 4

a, b Emissions of the main O3 precursors (NOx, AVOCs, and CO in Gg, CH4 in ktCO2 eq.) from the EU-28 and Spain, 2000–2019. In the top right table, estimation of 2008–2019 trends by region and compound. Trend magnitudes are significant at p < 0.001 (***), p < 0.01 (**), p < 0.05 (*), and p < 0.1 ( +); blank––no statistical significance (p ≥ 0.1). c–f Emissions of the main O3 precursors by sector in Spain for 2000–2019. Only the top five sectors are shown. RCI––residential/commercial/institutional. f For CH4, the sectors have a different classification because they came from a different database (EEA 2021c). Note that the secondary (right) y-axis shows the emissions sectors (dotted lines––industry, residential, and energy supply) plotted one order of magnitude below agriculture and waste on the primary y-axis. In the bottom right table, estimation of 2008–2019 trends by compound and sector in Spain. Only the top five sectors are plotted in the charts. In the table, these are highlighted in gray and in bold, while the others are retained for reference. NOx, AVOCs, and CO data from EEA (2021b)

Full size image

In 2008–2019, three of the top five NOx-emitting sectors in Spain (industry, energy, and road transport) significantly decreased their NOx emissions (− 4.3 to − 5.2%·year−1). Emissions from agriculture slightly decreased (− 0.8%·year−1), and those from non-road transport showed no significant trend, albeit they increased from 2014 (Fig. 4c).

The most relevant AVOCs-emitting sectors are industry (including emissions from solvents), followed by agriculture (representing less than half the industrial emissions), and residential/commercial/institutional (RCI), road transport, and waste, all representing less than 15% of the industrial emissions. The AVOCs emissions from the industrial sector did not follow a trend, although, after 2013, they clearly increased. Agricultural emissions weakly increased (+ 1.0%·year−1) along the period, while RCI and waste AVOCs emissions did not follow any trends. Finally, road transport emissions of AVOCs strongly decreased (− 7.6%·year−1) (Fig. 4d).

The top five CO-emitting sectors are waste, then RCI, industry, road transport, and agriculture. Road transport CO strongly decreased (− 6.1%·year−1) and RCI emissions also decreased, but only slightly (− 0.7%·year−1). Conversely, the CO from industry did not change during the studied period, whereas that from the agricultural and waste sectors clearly increased (+ 3.5 and + 3.9%·year−1, respectively) (Fig. 4e).

The most relevant CH4-emitting sector is agriculture, followed by waste management, then, more than an order of magnitude below, by industry, RCI, and energy. The agricultural sector showed no trend for 2008–2019, but did steadily increase from 2013, whereas waste management (representing about half the agricultural sector emissions) decreased by − 1.4%·year−1 (Fig. 4f).

Interpretation of O3 trends in specific areas

Massagué et al. (2023) found statistically significant O3 trends in the Spanish O3 hotspots for 2008–2019. However, these trends were observed in only a relatively small proportion of the AQMSs, in line with the findings of other studies in which similar time periods were analyzed (Fleming et al. 2018; Mills et al. 2018; EEA 2020). This is partly attributed to the meteorological sensitivity of O3, which varies from year to year, thereby complicating the detection of trends on relatively short timescales (Colette et al. 2016; Fleming et al. 2018). Moreover, depending on the part of the O3 distribution under consideration, an individual time series may show opposite trends, potentially leading to differing conclusions in the evaluation of emissions control strategies (Lefohn et al. 2017, 2018; Yan et al. 2019). For this reason, Massagué et al. (2023) employed a variety of metrics associated with different O3 concentration levels. Moderate O3 concentrations were evaluated using the annual mean O3 concentrations (O3YR), while mid to high O3 concentrations, by means of the well-established SOMO35, AOT40, and 93.2 percentile (P93.2) of the maximum daily 8-h average (MDA8). Furthermore, peak O3 concentrations were evaluated using the fourth highest MDA8 within a year (4MDA8) and the number of exceedances of Europe’s hourly information threshold (IT) (180 µg·m−3). This approach allowed for the assessment of a wide range of the distribution of O3 concentrations.

The main hotspots followed contrasting O3 trends (Fig. 1b). The Madrid area had mostly upward O3 trends for all the metrics and, in many cases, the highest increasing rates. The interior of the Valencian Community had a mixed variation pattern, depending on the specific O3 metric considered. The area north (downwind) of Barcelona, the Puertollano Basin, and the Guadalquivir Valley revealed almost no variations, and Sevilla was the only major Spanish city where O3 mostly decreased and, in many cases, with the highest decreasing rates. Multiple causes might explain these contrasting O3 trend behaviors in Spain, some of which are discussed below.

Increasing O3 trends in the Madrid area

This O3 hotspot (Fig. 1a), located on a continental plateau (700 m a.s.l.), is frequently affected by high O3 episodes caused by intense NO2 emissions, mainly from traffic coupled with high BVOCs. During the typical summer anticyclonic conditions, the surrounding mountain ranges enhance stagnant conditions and trigger recirculations, which lead to acute O3 episodes (see Querol et al. 2016 and references therein).

As previously mentioned, the Madrid area experienced a statistically significant upward trend in temperature during the study period, which might have enhanced the O3 levels, albeit to a limited extent, through various mechanisms (Porter and Heald 2019 and references therein). These mechanisms include an increase in BVOC emissions (Sillman and Samson 1995) and, to a minor degree, emissions from specific AVOCs (Rubin et al. 2006). The variations in other meteorological parameters observed in the interior areas, which can be conducive to O3 production, such as increasing solar radiation, an increase in PBL thickness, and a decrease in cloud cover, although lacking statistical significance, might still have contributed to the observed increase in O3 levels (see the “Meteorological parameters” section). Furthermore, as depicted in Figure S3.1j, the Madrid area typically reaches high PBL thicknesses due to the intensive convective circulations that give rise to the usual Iberian thermal low (Millán et al. 1997, 2002). These high convective flows may account for a greater influence of stratospheric and free tropospheric O3 (which is globally increasing, depending on the location) (Monks et al. 2015) via mixing down to the surface and impacting ground-level concentrations.

In terms of emissions of O3 precursors, the road transport sector in Madrid stands as the largest contributor to the share of NOx emissions (Borge et al. 2014). Additionally, it contributes significantly more to NOx emissions in Madrid than in Barcelona (Valverde et al. 2016), where other sectors account for significantly higher proportions (Guevara et al. 2014; Soret et al. 2014) (see Figure S1.5a). Assuming that the national 50% reduction in NOx emissions from road transport (Fig. 4c) applies similarly to both metropolitan areas for 2008–2019 (or potentially even higher in Madrid, as suggested by the ground-level measurements, Table S1.3), this would explain the substantial decrease in tropospheric NO2 observed by the OMI over the Madrid area, while Barcelona experienced a less pronounced decline. Because local O3 production depends on the VOC/NOx ratio, in urban or industrialized areas with high NOx concentrations, such as Madrid (and Barcelona), the VOC/NOx ratio is low, and O3 formation tends to be VOC-limited (Sillman et al. 2003), as also suggested by the lowest national HCHO/NO2 ratios reported in Fig. 3, especially for the Madrid area. Under these (VOC-limited) conditions, a reduction in NOx emissions might cause an increase in O3 (Monks et al. 2015). In this context, the marked NOx decreases over Madrid, as observed by the OMI and mainly related to road transport, partially might explain the generalized upward trends in O3 in this area for 2008–2019. Additionally, as Saiz-Lopez et al. (2017) also proposed for the same area during 2007–2014, it is likely that a portion of the observed increasing O3 trends, particularly in the case of moderate O3 metrics like O3YR or SOMO35 (Lefohn et al. 2017; Fleming et al. 2018), can be attributed to a reduction in the NO titration effect (Sicard et al. 2013, 2016; Simon et al. 2015). This is supported by the fact that these O3 trends were detected in urban and/or traffic stations, which also recorded large NO decreases (as documented in Table S1.3).

The clear increase in industrial AVOC emissions since 2013, with this sector being the largest contributor to AVOC emissions in Spain (Fig. 4) and bearing the dominant responsibility for the largest share of AVOC emissions in Madrid (Figure S1.5b), coupled with significant increases in emissions from other sectors, may also have contributed to the general upward trend in O3 levels in the Madrid area. Increases in VOC emissions, which are known to cause increases in O3 under VOC-limited conditions, may have been particularly influential in this regard, according to previous research (Jacob et al. 1995; Sillman 1999; Sillman and He 2002; Sillman et al. 2003).

Finally, it is worth mentioning the case of two AQMSs located in Castilla La Mancha, 40 km north-east of central Madrid, in an area likely prone to be in NOx-limited conditions. These stations, which are affected by the pollution plume from the Madrid area and typically experience high levels of O3, recorded statistically significant decreasing trends for certain O3 metrics, contrasting with the generalized increasing trends observed nearer to the urban area (Massagué et al. 2023). This situation illustrates how NOx reductions can impact O3 trends, depending on different O3 formation regimes. 

Absence of O3 trends in the northern Barcelona area

The Vic Plain (located downwind of Barcelona) (Fig. 1a) is an important O3 hotspot in Spain, partly due to the high precursor emissions transported from the Barcelona metropolitan area (Diéguez et al. 2009; Querol et al. 2016, 2017). This region is affected by the large-scale WMB recirculations (see Querol et al. 2016 and references therein). Its proximity to the sea causes a thinner PBL in this area compared to the interior areas, such as the Madrid region (Figure S3.1 k), which likely means Barcelona is less affected by air masses from the free troposphere.

Massagué et al. (2023) found no statistically significant O3 trends in this hotspot, thus the intense reduction in NOx emissions in the Barcelona area, as evidenced by both satellite-based (Fig. 3c) and ground-level (Figure S1.3) measurements, likely had a limited or no impact on the O3 levels at this location. The decrease in tropospheric NO2 in the Barcelona area was less intense than in Madrid, probably because NOx emissions from sectors other than road transport (industry, harbor, energy, among others), which contribute more to the NOx share in Barcelona than in Madrid (Guevara et al. 2014; Soret et al. 2014), decreased less than those from road transport, or even increased (Fig. 4c).

The absence of significant meteorological variations likely also contributed to the lack of change in O3 levels in this area, in contrast to central Spain and, therefore, the Madrid area. However, a few rural background sites not located in the Vic Plain, but also affected by the Barcelona area’s pollution plume, recorded slight downward trends for the high and peak O3 metrics (P93.2 and 4MDA8), consistent with the findings of Lefohn et al. (2018) and Yan et al. (2019) who indicated that, at most stations in Europe, O3 in the upper end of the distribution decreased due to effective NOx abatement measures.

In Barcelona city, only some urban traffic sites recorded increasing trends for several O3 metrics in 2008–2019 (Massagué et al. 2023), pointing to the effect of lower titration by NO (Simon et al. 2015; Sicard et al. 2013, 2016), which is supported by the marked ground-level NO decreases recorded during that period at those same sites (Figure S1.3).

Mixed O3 trends in the Valencian Community

The O3 phenomenology in the Valencian Community (Fig. 1a) resembles that of Barcelona–Vic Plain because this region lies in the WMB and is also affected by large-scale recirculations, where O3 episodes occur inland due to high urban (Alicante and Valencia) and urban-industrial (Castellón) emissions that are transported inland by meso-meteorological circulations (e.g., Millán et al. 1997, 2000).

According to Massagué et al. (2023), the Valencian Community was the O3 hotspot with the second highest number of positive trends after Madrid. Trends in the metrics based on moderate and mid to high concentrations were mostly positive, found in coastal locations, and, to a lesser extent, in the interior. In addition, a few rural and regional AQMSs recorded downward trends for peak O3 in the north-west.

Levels of tropospheric column NO2 decreased only in Valencia and Alicante (Fig. 3c), where road transport prevails as a main NOx source. However, no statistically significant trends were found for Castellón, where high industrial emissions, mainly from the ceramics cluster, a petrochemical plant, and a power plant (Minguillón et al. 2013), likely moderated the NOx reductions from road transport. Although there is no data available on the share of O3 precursor emissions in the Valencian Community, it is known that the industrial and agricultural economic sectors are (relatively) more important there than in Madrid (INE 2022). Therefore, the relative weight of the emissions from these (and other) sectors might have been expected to be greater in the Valencian Community, and these emissions showed minor decreases, stability or even increases in the period studied (Fig. 4c, d, e).

The absence of evident variations in meteorological parameters may indicate that the dominant increasing O3 trends in the region (mostly occurring in the coastal cities) are probably due to the decrease in NOx emissions in a VOC-limited regime and/or by a lower NO titration effect (Simon et al. 2015; Sicard et al. 2013, 2016). However, the decreasing O3 trends for peak O3 (4MDA8) in regional background sites located north of the Valencian Community (probably in NOx-limited conditions) might be attributable to the significant reduction in nearby NOx emissions (observed in the tropospheric column) (Figs. 3c and S1.4b). This reduction in NOx emissions, is due to the gradual reduction in the operational activities of one of Spain’s largest coal-fired power plants (1050 MW), which formerly had a notable impact on the air quality of this region. The plant’s activity began to diminish in 2010 and eventually ceased operations in 2020.

Absence of O3 trends in Puertollano

In Puertollano, the coexistence of large industrial sources with specific orographic features and atmospheric conditions, such as thermal inversion and stagnation, have resulted in this location having one of the poorest dispersion characteristics among the inland areas of Spain (Diéguez et al. 2009, 2014). The large industrial plant in Puertollano is the main source of NOx and AVOC emissions in the region (Querol et al. 2016 and references therein), and concentrations of hydrocarbons there have been found to be higher than in other large industrial facilities (Diéguez et al. 2009), thus the O3 formation regime should tend to be NOx-limited. Episodes of O3 in this closed basin are characterized by their acute nature, where exceedances of Europe’s IT are recorded multiple times every year, but only at two closely located AQMSs. However, when considering O3 metrics indicative of chronic pollution, the AQMSs in Puertollano tend to register lower values than other AQMSs in nearby areas (Massagué et al. 2023).

Satellite observations have shown a slight decrease in NO2 levels in Puertollano, possibly due to a reduction in industrial NOx emissions (Fig. 3c). However, unlike other cities that experienced a significant reduction in ground-level NOx due to decreased road traffic, Puertollano did not show any decrease (Figure S1.3). This difference may have caused the absence of positive trends in moderate O3 metrics in Puertollano, as reported in Massagué et al. (2023), which were observed in most cities and are likely due to the effect of reduced NO titration.

According to the inventories, AVOC emissions from the industrial sector did not show a clear trend for the whole period (Fig. 4d), but have clearly increased since 2013–2014, likely due to the rebound effect after the 2008 financial crisis (Pacca et al. 2020). These changes in AVOC emissions may have had a limited effect on O3 levels in NOx-limited conditions (Sillman 1999; Sillman and He 2002; Sillman et al. 2003).

The temperature changes conducive to O3 production, together with other non-statistically significant meteorological changes observed in Puertollano in the studied period, are similar to those observed in the Madrid region (Figs. 2c and S3.1). However, it appears that temperature and other meteorological parameters may have a comparatively smaller influence on O3 levels in the Puertollano area than in other O3 hotspots because O3 episodes in Puertollano can occur throughout the year, including winter, rather than being limited to the typical O3 season (Diéguez et al. 2009).

Decreasing O3 trends in Sevilla and absence of trends in the Guadalquivir Valley

The urban area of Sevilla, located in the Guadalquivir Valley and downwind of the large petrochemical area of Huelva (Fig. 1a), frequently experiences acute O3 episodes (Diéguez et al. 2009; Querol et al. 2016), with 1-h concentration-based thresholds being exceeded. However, in inland areas of the Guadalquivir Valley, located downwind of Sevilla and Huelva, O3 pollution exhibits chronic behavior, where, instead of high hourly threshold exceedances, MDA8 concentration-based thresholds are exceeded for several days during summer (Massagué et al. 2021, 2023).

No AQMSs in this basin recorded any O3 trends except for in the city of Sevilla, which showed clear and widespread downward trends for most of the O3 metrics (Massagué et al. 2023). The reason behind the distinctive decreasing O3 trends in Sevilla city may be attributable to the relatively high urban NOx concentrations that interact with the high AVOC concentrations emitted from the large industrial petrochemical area in Huelva, located upwind (although not evident in Fig. 3d, as no HCHO hotspots were detected). This unique scenario in Spain might have transformed the typical urban VOC-limited regime into one that tends to be NOx-limited (Diéguez et al. 2009). Therefore, changes in the emissions in Sevilla (decrease in NOx, as observed by the OMI) (Fig. 3c) and/or Huelva (which were not observed in this study) may have played a part in causing the overall O3 decrease in this city.

Importantly, only the low O3 concentration metrics (in this case annual O3) increased in this city, which supports the possibility of a reduced NO titration effect (Simon et al. 2015; Sicard et al. 2013, 2016) being the primary cause, similarly to most of the cities studied here, as evidenced by the decrease in ground-level NO concentrations (Figure S1.3c).

The highest national temperatures recorded during the O3 season were observed along the Guadalquivir Valley (Fig. 2a), which may trigger BVOC emissions from the large forested and planted areas. However, no significant trends in temperature or other evaluated meteorological parameters were found in the analysis. The national inventories showed an increase in agricultural AVOCs, which likely had relevance in Andalucia (and therefore in the Guadalquivir Valley, where most crops are located, as shown in Figure S1.1b), since this sector plays an important role in the local economy (INE 2022). Although increases in HCHO were observed in the Guadalquivir Valley, which would be consistent with the increase in AVOCs from agriculture, these are not clearly distinguishable likely due to the aforementioned signal-to-noise ratio. However, it should be noted that in NOx-limited environments—likely most of the Guadalquivir Valley (also suggested by the HCHO/NO2 ratio) and also Sevilla (Diéguez et al. 2009)—an increase in VOC emissions has a limited effect on O3 (Sillman 1999; Sillman and He 2002; Sillman et al. 2003), although it contributes to maintaining the area in a NOx-limited regime.

Other considerations for O3 trends

Other considerations need to be taken into account when interpreting O3 trends:

  • Concentrations of CH4 have increased globally in the last decade (Saunois et al. 2020), causing increases in the background O3 (Archibald et al. 2020) and counterbalancing, to some degree, the decline in European emissions of O3 precursors (EEA 2020). Thus, non-European sources of CH4 (and other O3 precursors) may play a relevant role in ground-level O3 levels in Europe. However their influence tends to be more pronounced on the annual mean O3 levels compared to other metrics like SOMO35, which mainly depends on high summer O3 concentrations and is more influenced by European precursor emissions (Jonson et al. 2018; Turnock et al. 2018; EEA 2020). At a local/regional level, CH4 emissions also need to be considered because their contribution to regional O3 formation is nowadays considered to be higher than previously reported (Van Dingenen et al. 2018; IPCC 2021), although in the urban atmosphere, its contribution can be relatively small, where reactive organic compounds and CO dominate O3 production (Archibald et al. 2020). Despite the lack of data on the relative by-region CH4 emissions weights in Spain, these might have had an influence on the O3 trends because, for example, during this period, national emissions from agriculture and waste management (one order of magnitude larger than the other CH4 emissions sectors considered) increased from 2013, or slightly decreased (Fig. 4f). Furthermore, very high CH4 emissions have been recently detected seeping from Madrid landfills (ESA 2021; Tu et al. 2022) that have probably been underestimated in the inventories, and these might have had a positive influence on the O3 levels in and around Madrid, as well as other areas during the period.

  • It is possible that other emissions relevant to O3, such as those from agricultural and biomass burning for power generation (e.g., In ’t Veld et al. 2021), might not have been considered or may have been underestimated in the inventories used here. Shipping emissions also need to be considered because they may increase O3 along the Spanish coastal regions (Jonson et al. 2020; Nunes et al. 2020). For example, emissions in the area between continental Spain and the Balearic Islands—probably relevant to the O3 in Cataluña and the Valencian Community—might have increased, as port traffic (and other shipping metrics) strongly increased during the study period (Figure S1.6). However, the NO2 tropospheric columns did not show a clear pattern, with increases recorded in the area around the Balearic Islands, but decreases along the Catalan coast (Fig. 3). In this context, to illustrate the significance of these emissions, the abrupt reductions in shipping traffic, particularly from cruises and ferries, (and air traffic emissions) caused by the COVID-19 restrictions in 2020 and 2021 (Guevara et al. 2022; Puertos del Estado 2022), likely played a role in preventing the European Union’s established O3 thresholds from being exceeded for the first time in the Spanish Mediterranean coastal areas (Querol et al. 2021; Targa et al. 2022). This is because the relatively low reductions in road transport emissions observed in summer, when O3 levels are at their highest, may not fully account for the large improvement in O3 levels (Oliveira et al. 2023).

  • It is important to acknowledge that other factors operating at different spatial and temporal scales than those considered in this study may also have contributed to the observed trends in O3. These factors, among others, include the following: (i) the long-range transport of O3 and its precursors, although the two stations that best represent the regional O3 behavior in Spain (located on the western Atlantic coast and in the Balearic Islands) (Diéguez et al. 2009) did not capture any O3 trends during the period (Massagué et al. 2023); (ii) the exchange of O3 between the stratosphere and troposphere; and (iii) climate variability at both high and low frequencies. For instance, climate phenomena, such as the El Niño–Southern Oscillation or the North Atlantic Oscillation, as well as global warming, can modulate the tropospheric O3 burden. These factors are interconnected with each other as well as with the other factors considered in this study. To fully understand their impact on surface O3 trends, complex modeling tools are necessary.

Conclusions

A related study (Massagué et al. 2023) revealed contrasting surface ozone (O3) trends among the O3 hotspots in Spain from 2008 to 2019. Specifically: (i) the Madrid metropolitan area showed generalized upward trends, often with the highest increasing rates; (ii) the interior of the Valencian Community area had a mixed variation pattern; (iii) the area downwind of Barcelona, the Puertollano Basin, and the Guadalquivir Valley showed no variations; and (iv) Sevilla was the only major urban area in Spain that recorded general decreasing trends.

While understanding the drivers behind these contrasting O3 trends requires the use of highly complex models, this study aimed to make progress toward this goal by employing a simplified methodology and observational data. Its ultimate purpose is to assist in the planning of future advanced studies for designing a mitigation plan for O3 in Spain. To this end, relevant O3 precursors, such as nitrogen oxides (NOx) and volatile organic compounds (VOCs), among other species, along with meteorological parameters associated with O3, were analyzed from different data sources, including ground-level measurements and satellite observations, emissions inventory estimates, and meteorological reanalysis datasets.

The O3 hotspots with the most opposing behaviors regarding O3 trends were, at both ends, the Madrid area and the city of Sevilla. These act as both emissions sources of O3 precursors and receptor areas. Madrid is impacted mainly by its own emissions, whereas Sevilla is affected by a large petrochemical area located downwind, while also contributes to the emission of pollutants that impact the interior of the Guadalquivir Valley (Querol et al. 2016 and references therein). The results suggest that, in the Madrid area, the generalized increasing O3 trends may be mostly driven by major NOx decreases attributed to the road transport sector in this urban VOC-limited environment. Additionally, and to a lesser extent, an increase in surface temperature and other variations in meteorological parameters conducive to O3 production may be at play. Conversely, the generalized decreasing O3 trends in Sevilla city probably resulted from declining NOx emissions in a peculiar urban NOx-limited regime caused by AVOC contributions from a large petrochemical area located upwind (Diéguez et al. 2009), and increases in BVOCs associated with increased agricultural emissions and favorable meteorological conditions.

The other O3 hotspots, including the interior of the Guadalquivir Valley, areas north of Barcelona, the closed Puertollano Basin, and the interior of the Valencian Community are receptor areas (located downwind of major emissions sources). Notably none of these areas exhibited O3 trends, except for the interior of the Valencian Community, which displayed a less evident pattern. All of these hotspots are partly or entirely impacted by industrial, or other sector, emissions, apart from those from the road transport sector, as opposed to the Madrid area, where this is the primary contributor. The absence of clear trends in emissions from sectors other than road transport, and in some cases, increased emissions from specific sectors and precursors (possibly linked to the economic recovery following the 2008 financial crisis), coupled with the absence of significant meteorological variations conducive to O3 production, seem to have contributed to the overall stability of O3 concentrations in these hotspots.

Furthermore, in all the urban areas analyzed here, with the exception of Puertollano, surface NOx decreases attributed to the road transport sector and increases in O3 (especially in the lower part of the O3 distribution) were detected for the study period, pointing to a decrease in the titration effect as the main cause of these O3 variations.

The findings from this study show that the parameters influencing O3 vary distinctively across the different Spanish atmospheric regions—or O3 hotspots—and that the causes underlying the contrasting O3 trends differ significantly. This emphasizes the necessity of adopting an independent regional/local approach in planning O3 mitigation measures.

In this study, we relied on satellite-based observations of formaldehyde (HCHO) as a proxy for VOCs emissions. However, the utility of these measurements in relation to O3 is limited due to the presence of noise, low data frequency, and the fact that HCHO can be either primary or secondary. To improve on the available information, VOCs measurements with high temporal resolution, NOx/VOC sensitivity studies using both experimental and modeling methods, and higher disaggregation and temporal resolution in the VOCs and NOx inventories are required. We believe that future studies should pursue these directions to advance our understanding of the complex relationships between O3 and its precursors.