European air quality in view of the WHO 2021 guideline levels: Effect of emission reductions on air pollution exposure

Although anthropogenic emissions of air pollutants have decreased over the last 20 years (Kuenen et al., 2022), air pollution remains a hazardous issue in Europe and around the world. It can cause adverse health effects but also negatively affects climate change and vegetation. In Europe, about 400,000 premature deaths are attributed to polluted air every year (European Environment Agency [EEA] et al., 2018; Im et al., 2018), making air pollution the single biggest environmental threat to humans (World Health Organization [WHO], 2021). Among air pollutants, PM_2.5 (particulate matter with an aerodynamic diameter not larger than 2.5 µm) concentrations have the largest effect on human health (Global Burden of Disease [GBD] 2019 Risk Factors Collaborators, 2020) as small particles are able to penetrate deep into the lungs and into the blood cycle, causing acute or chronic respiratory or cardiac symptoms, and may damage the body’s tissue through their toxicity (Schraufnagel et al., 2019).

Further, air pollutants are linked to climate change by both directly changing the radiative budget of the earth (Szopa et al., 2021) and influencing the abundance of climate forcers like carbon dioxide (CO₂) and methane. In addition, the changes in the radiative budget of the earth alter the atmospheric (thermo-) dynamics leading to changes in the transport, chemical transformation, natural emissions, and deposition of air pollutants (e.g., Doherty et al., 2017; Szopa et al., 2021). Air pollutants and greenhouse gases are mainly emitted simultaneously by anthropogenic sources; thus, variations in the emission patterns of air pollutants (e.g., through regulations) directly influence the mass of emitted greenhouse gases. Finally, the enhanced concentrations of air pollutants impact crops and other vegetation (Mills et al., 2011 and references therein), which reduces the harvest yield (Xu, 2021; Chaudhary and Rathore, 2022). Fertilization increases the emissions of nitrogen oxides and ammonia (NH₃) and the nitrogen concentration within the ground, which can cause a reduction of biodiversity (EEA et al., 2018). These processes directly change the atmospheric composition by altering the biogenic non-methane volatile organic compounds (NMVOC) emissions and enhancing the emissions of nitrogen oxides and NH₃ from soils, especially in agriculturally active regions.

Air quality in Europe is routinely evaluated by European and national environment agencies (e.g., Minkos et al., 2017; EEA et al., 2018) through air quality monitoring efforts on the European level (https://www.eea.europa.eu//publications/status-of-air-quality-in-Europe-2022, accessed: March 20, 2024) and daily forecasts and analyses of air pollutant concentrations by, for example, the Copernicus Atmosphere Monitoring Service (CAMS; Marécal et al., 2015). Sicard et al. (2021) provide a detailed literature review on air quality trends and population exposure in the European Union (EU) member states (including Great Britain). They summarize a significant trend in reducing annual mean concentrations of air pollutants throughout the EU member states (e.g., −0.42 µgm⁻³ year⁻¹ for PM_2.5 between 2000 and 2017), with the notable exception of ozone (O₃) annual mean concentrations. Nevertheless, the maximum O₃ daily 8-h mean values (DM8) have decreased (−0.75 ppb year⁻¹, which corresponds to −1.5 µgm⁻³ year⁻¹ using a temperature of 20°C and a pressure of 1,013 hPa) in the analyzed period from 2000 to 2017, resulting in reduced peak O₃ concentrations. Guerreiro et al. (2014) show similar trends and conclude that PM and O₃ are the most problematic air pollutants in Europe with 91%–96% (97%–98%) of the population exposed to PM_2.5 annual mean (O₃ DM8) concentrations above the WHO guideline levels of 2008.

Different air quality guidelines and directives exist to support efforts in mitigating air pollution. In fall 2021, the WHO released updated air quality guideline levels (WHO_GL in the following) for the air pollutants PM₁₀, PM_2.5, O₃, nitrogen dioxide (NO₂), sulfur dioxide (SO₂), and carbon monoxide (CO). These guideline levels, which define “the lowest levels of exposure for which there is evidence of adverse health effects,” are compiled to provide “health-based” and “evidence-informed” recommendations (WHO, 2021) to assist local air quality management. The WHO_GL are valid for outdoor and indoor air quality and defined for long-term (i.e., annual mean level) and short-term exposure (i.e., daily mean level). For O₃, a peak season guideline level for the DM8 is defined, where the peak season is defined as the 6 consecutive months with the highest averaged O₃ concentrations. Table 1 summarizes the WHO_GL for reference.

Meeting the updated WHO_GL is a challenging task, as commented by several experts (e.g., Burki, 2021; Ouyang et al., 2022). However, studies have shown that premature deaths will decrease when air pollution levels meet the WHO_GL. For example, Khomenko et al. (2021) estimated that by meeting the WHO_GL for PM_2.5 (NO₂), more than 100,000 (50,000) premature deaths could be prevented annually, based on a study for 969 European cities. Yang et al. (2022) analyzed the reduction of disease burden in 315 Chinese cities by following the WHO_GL for short-term exposure. They found a total of 134,025 avoidable deaths for these cities in 2019, where more than 58,000 deaths were avoidable by reducing NO₂ concentrations toward the WHO_GL. Bowdalo et al. (2022) compared the WHO_GL with ground station observations across Europe. While they found a limited guideline excess for CO and SO₂ daily averaged values (1% and 16% of the analyzed station, respectively), the guideline levels are widely exceeded for the other air pollutants (up to 98% of stations exceeding the annual PM_2.5 guideline level of 5 µgm⁻³). In addition, they analyzed the median relative distance of the concentration distribution to the WHO_GL, which is largest for NO₂ and PM_2.5 (120% and 160%, respectively). Here, the median relative distance is calculated as the median of the percentage difference between the observation and the air quality guideline level, averaged over all stations exceeding the guideline level.

While most of the studies investigate the observed concentrations of air pollutants, this study uses the analyzed correction factors for anthropogenic emissions, which are applied to the EURopean Air pollution Dispersion—Inverse Model (EURAD-IM; Elbern et al., 2007) to investigate European air quality in 2016. With this approach, the effect of emission reductions on air pollutant concentrations is addressed in detail, which has not been published for Europe before. The study aims to evaluate the compliance of European air pollution with the WHO_GL considering different emission scenarios. Further, the potential to reach the guideline levels with a focus on PM_2.5 is conducted. Finally, the effect of emission changes on mortality is analyzed.

This article is organized as follows: First, the experimental setup, model configuration, and used data are described followed by a comparison of air pollution concentrations with the WHO_GL. The modeled chemical composition of PM_2.5 is evaluated afterward, and the effects of emission reductions on PM_2.5 concentrations and population exposure are analyzed. The results of this study are discussed before conclusions are given.

European air quality is analyzed for the year 2016. For this, the analyzed emission correction factors from a full-year reanalysis using the four-dimensional variational (4D-var) data assimilation method implemented in the EURAD-IM (Elbern et al., 2007) are applied. Within the reanalysis, ground-based in situ observations of NO₂, O₃, CO, SO₂, PM_2.5, and PM₁₀ provided by the EEA, operational aircraft observations of CO and O₃, and satellite observations of CO and NO₂ columns were assimilated to retrieve daily emission correction factors for each grid box of a 15 km × 15 km model domain covering Europe. Due to computational constraints, the reanalysis has been started for each season individually with a spin up of 5 days. As the emission correction factors showed an adjustment period of approximately 2 weeks to changes in emissions, the emission factors of observed species have been averaged for each month. The emission correction factors for unobserved species (NMVOC, NH₃) showed a longer adjustment phase. Therefore, these factors have been averaged for full seasons. More information about the analysis is given in Lange et al. (2023). Although the reanalysis is designed to evaluate the European emissions of trace gases and aerosols, it is well suited to investigate the horizontal distribution of air pollution and the effect of emission reductions on air pollutant concentrations in Europe.

The EURAD-IM is a state-of-the-art chemistry transport model predicting trace gas and aerosol concentrations. It has been applied in several case studies (e.g., Gama et al., 2019; De Souza Fernandes Duarte et al., 2021; Vogel and Elbern, 2021; Franke et al., 2022). It is part of the regional CAMS ensemble (Marécal et al., 2015) and, hence, routinely evaluated against other chemistry transport models and observations (https://atmosphere.copernicus.eu/regional-services, accessed: March 20, 2024). Table 2 summarizes the main model components and references used in this analysis. Due to the complexity of atmospheric processes happening on multiple timescales, the EURAD-IM makes use of the operator-splitting approach (McRae et al., 1982), in which the transport is split before and after the chemistry and aerosol transformation equations are solved. The EURAD-IM comes with an advanced 4D-var data assimilation scheme, which enables the joint optimization of initial values and emission correction factors for gas phase and aerosol species (Elbern et al., 2007). The ordinary differential equations of chemical reactions are solved using the Rosenbrock method implemented in the Kinetic PreProcessor (Sandu and Sander, 2006). The aerosol module MADE (Ackermann et al., 1998) simulates the aerosol processes in the atmosphere, including nucleation, accumulation, coagulation, and chemical transformation. Further, the EURAD-IM comprises routines for calculating the chemical generation and destruction of secondary organic and inorganic aerosols (Nieradzik, 2005; Li et al., 2013).

In order to validate the model results, the simulated ground-level concentrations of NO₂, O₃, PM_2.5, and PM₁₀ are compared with ground-based in situ observations retrieved from the EEA. Only observations at rural, suburban, and urban background sites are used for validation as regional models are not representative for inner-city and road traffic observations (e.g., Weger et al., 2022). The scope of the study is to evaluate the compliance of European air quality with the WHO_GL and the effect of emission reductions on European air quality. Thus, the comparison of model results with observations is restricted to annual mean concentrations. Further, population data from the GBD 2019 Risk Factors Collaborators (2020) for the year 2016, representing approximately 578.6 million European inhabitants (Figure S1), are used to evaluate population exposure to air pollution (see Supplement Material Text S1 for further information). These population data, along with additional population distribution data from the NASA Socioeconomic Data and Applications Center, are used in the calculation of mortality due to air pollution (Center for International Earth Science Information Network, 2016).

For the evaluation of European air quality and the effect of emission modulations, the simulations of the year 2016 are chosen as it is representative for current climate conditions in Europe and shows no extreme air pollution events affecting Central Europe (e.g., no enhanced forest fire activity, volcanic eruptions, nor pronounced O₃, and aerosol episodes; Lange et al., 2023). More recent years show either strong anomalies in air pollution, for example, low PM concentrations in 2017 (Lange et al., 2023), strong O₃ episodes in 2018 and 2019 (figure 12 in Minkos et al., 2020), or air pollution was affected by the COVID-19 pandemic as in 2020 (Gkatzelis et al., 2021) limiting the representativity of these years. The model simulations are performed with 15 × 15 km² horizontal resolution covering the full European continent (see also Figure S1 for the outline of the model domain) and on 30 vertical layers up to 100 hPa; 21 model layers are located below 2 km with the lowermost layer located around 7 m height, such that the vertical distribution of air pollutants within the planetary boundary layer is well captured.

Four full-year simulation experiments are performed using the EURAD-IM on the JURECA-DC supercomputer (Jülich Research on Exascale Cluster Architectures—Data Centric; Thörnig, 2021) that differ in the applied anthropogenic emission reductions (see Figure 1 for a schematic description of the simulations). Central to all simulations is the use of emission inventory data from the German Environment Agency (submission 2019, version 2) for Germany and CAMS-REG-AP, version 3.1, for Europe (Kuenen et al., 2014). While monthly correction factors are applied to the emissions of CO, SO₂, NO_x, PM_2.5, and PM₁₀, seasonal correction factors are applied to NH₃ and NMVOC emissions. These correction factors have been derived from the full-year reanalysis for the year 2016 using the 4D-var method within EURAD-IM (in preparation). In this reanalysis, the SORGAM module was deactivated as its adjoint is not available due to the complexity and nonlinearity of the SORGAM forward model. However, it is used for the simulations presented in this study for an analysis of the aerosol composition.

The Base simulation comprises the inventory emission data with applied correction factors (i.e., without emission reduction). Committed emission reductions by the EU (European Parliament and Council of the European Union, 2016) were used to reduce the emissions in the second sensitivity simulation (EU_emis, see Supplemental Material Text S2, Figure S2, and Table S1 for more details on the emission reductions). In the third sensitivity simulation, an 85% reduction of primary PM_2.5 emissions and a 30% reduction of emissions from power generation, industry, other stationary combustion, road transport, and agriculture are performed (EU₃₀) in addition to the reduction applied in the EU_emis scenario. This simulation is designed to especially mitigate the highest simulated PM_2.5 concentrations. Therefore, the aggressive emission reductions have been applied to the main sources of anthropogenic PM_2.5. Although not realistic in design, the scenario provides suitable information for further emission reduction plans and illustrates possible air pollution mitigation pathways. Finally, all anthropogenic emissions are removed in the last sensitivity simulation (Zero_emis).

The analysis of daily guideline level excess (not shown) reveals similar conclusions as the analysis of annual guideline levels. Thus, the evaluation solely focuses on the latter. Figure S3 compares modeled and observed annual mean concentrations at European observation sites for PM_2.5, PM₁₀, NO₂, and O₃ (seasonal mean) for the Base simulation in 2016. The model shows a systematic underestimation of NO₂ and PM₁₀ annual mean concentrations at suburban and urban sites as well as for PM_2.5 annual mean concentrations above 15 µgm⁻³ at urban sites. Seasonal O₃ concentrations show an overestimation below 70 µgm⁻³ (suburban and urban sites). The bias of the annual mean concentrations across all station types is −6.8 µgm⁻³ (NO₂; −44.0% relative bias), −6.0 µgm⁻³ (PM₁₀; −36.4%), −1.2 µgm⁻³ (PM_2.5; −9.6%), and 6.8 µgm⁻³ (O₃ seasonal mean; 7.7%). This illustrates the reduced representativity of the coarse model resolution of 15 × 15 km² for urban and suburban stations (red and yellow colors in Figure S3). While the NO₂ modeled concentrations differ from the observations by more than a factor of 2 at 32.8% of the stations, this value drops to 2.5% and 0.4% for PM_2.5 and O₃, respectively. The PM_2.5 and PM₁₀ concentrations above 15 µgm⁻³ that are underestimated by the Base simulation occur mainly in Eastern and Southeastern Europe (not shown). As the comparison of PM_2.5 and PM₁₀ shows a similar spatial distribution of aerosol concentrations (not shown), only results for PM_2.5 are evaluated in the following.

Figure 2 shows the modeled annual mean concentrations for PM_2.5, NO₂, and O₃ (seasonal mean) for the Base simulation in 2016. Annual mean PM_2.5 concentrations in Europe exceed the WHO_GL of 5 µgm⁻³ except for Northern Scandinavia. The highest PM_2.5 concentrations of more than 15 µgm⁻³ are found in Central, Eastern, and Southeastern Europe with maximum concentrations reaching 28 µgm⁻³ in the Po Valley, Northern Italy, and Southern Poland.

Although the simulation underestimates the NO₂ annual mean concentrations in 2016, especially at urban sites, Figure 2 shows the expected correlation between urbanized areas (close to the main NO_x emitters) and high annual mean NO₂ concentrations. In these regions, namely, Central Europe, the Po Valley, and metropolitan areas, such as Madrid, Paris, and Moscow, the annual mean NO₂ concentrations range from 10 to 30 µgm⁻³. In the Po Valley, the modeled results for NO₂ annual mean concentrations reach values up to 52 µgm⁻³, which is 5.2 times the annual WHO_GL (10 µgm⁻³).

Seasonal O₃ WHO_GL (60 µgm⁻³) are exceeded throughout Europe. A meridional gradient for seasonal mean O₃ ranging from 65 µgm⁻³ in Northern Scandinavia to 118 µgm⁻³ in the Mediterranean area is simulated, thus, reaching up to a factor of 2 times the WHO_GL in Southern Europe. Further, the seasonal mean O₃ concentrations show a gradient of 10–20 µgm⁻³ at coastlines, which is in accordance with the findings of other studies (e.g., Pleijel et al., 2013; Finardi et al., 2018). Millán et al. (2002) and Adame et al. (2010) identified a sea breeze system over the Iberian Peninsula as a main driver for the accumulation and formation of high O₃ concentrations over the ocean by recirculating and building of O₃ reservoir layers aloft. In addition, high O₃ concentrations are reinforced by shipping emissions of O₃ precursors (Gencarelli et al., 2014). However, Figure 2 also illustrates the nonlinear effect of anthropogenic emissions on O₃ production (see, e.g., Xing et al., 2011). While in the Strait of Gibraltar, local ship emissions lead to reduced O₃ seasonal mean concentrations (approximately 20 µgm⁻³ below the surrounding concentrations), local NO_x emissions in the Po valley lead to an increase in seasonal mean O₃ concentrations (approximately 20 µgm⁻³ above the surrounding concentrations).

Although the analysis reveals that annual mean concentrations in the Base simulation underpredict PM_2.5 and NO₂ annual mean and overpredict O₃ seasonal mean concentrations, the spatial distribution and the strength of WHO_GL excess are captured by the model. This is deducible from the low relative bias for PM_2.5 and O₃ and the homogeneous spatial distribution of the bias across Europe (not shown). The model simulations are therefore considered to err on the positive side, meaning that the guideline excess and the population exposure are expected to be higher than analyzed, especially close to local sources like urban areas and industries.

The annual mean concentrations of PM_2.5, NO₂, and O₃ (seasonal mean) exceed the WHO_GL in 73% of European area, which is persistent throughout the year (season, not shown). The effects of different emission scenarios aiming to reduce air pollution (cf. Figure 1) are discussed in the following. As one of the “most problematic” air pollutants (Guerreiro et al., 2014), the evaluation is focused on PM_2.5 and its components.

Figure 3 compares the annual mean PM_2.5 concentrations for all emission reduction scenarios (cf. Figure 2 for the annual mean PM_2.5 concentrations in the Base simulation). The WHO_GL is exceeded in 61% and 39% of the European area in the EU_emis and EU₃₀ scenarios, respectively. In the EU_emis emission scenario, the annual mean PM_2.5 concentrations are reduced throughout Europe to about 10 µgm⁻³ and below, except for the Po Valley, Southern Poland, and Russian cities, where concentrations up to 17 µgm⁻³ are found. Mountainous regions in Romania, France, and Spain show annual mean PM_2.5 concentrations below 5 µgm⁻³. The additional emission reduction in the EU₃₀ scenario lowers annual mean PM_2.5 concentrations to values below 8 µgm⁻³ except for the Po Valley, where annual mean PM_2.5 concentrations show the values of up to 12 µgm⁻³. In the Zero_emis emission scenario, annual mean PM_2.5 concentrations drop below the WHO_GL of 5 µgm⁻³ throughout Europe, where the lowest concentrations are simulated in Central and Northern Europe (approximately 2 µgm⁻³) far away from the Southern, Eastern, and Western domain boundaries. Note that the emission scenario simulations only consider emission reductions within the simulated model domain, not affecting the lateral boundary values.

PM_2.5 comprises primary aerosols and secondary aerosols (see Table 3 for a description of the model aerosol species defined in the EURAD-IM). Figure 4 shows the relative share of the different aerosol species to total PM_2.5 annual mean concentrations in 2016, averaged over all European countries (see Figure S1). The relative share is given for the Base simulation and the EU_emis and EU₃₀ emission scenarios.

For the Base simulation (Figure 4, left panel), the smallest fraction of total PM_2.5 are anthropogenic secondary organic aerosols (SOA) with a 1% relative share. Biogenic SOA is significant (up to 17% relative share) for annual mean PM_2.5 concentrations below 6 µgm⁻³, with these concentrations being found mainly in Northern Europe, where the population density and anthropogenic emissions are lower than in Central and Southern Europe. SO₄²⁻ and sea salt/dust show the largest share of more than 30% each for annual mean PM_2.5 concentrations below 4 µgm⁻³, which are found only in the North of Scandinavia and Iceland. These values decrease below 10% and 3% for the highest concentrations of 28 µgm⁻³ for SO₄²⁻ and sea salt/dust, respectively. Contrarily, the relative share of primary anthropogenic aerosols (ANT + pORG = pANT) increases with increasing annual mean PM_2.5 concentrations. For annual mean concentrations of 28 µgm⁻³, pANT is found to be 61% of the total PM_2.5 mass. This indicates the mainly anthropogenic origin of the highest simulated PM_2.5 annual mean concentrations. NH₄⁺ and NO₃⁻ show an almost constant fraction of the total PM_2.5 concentration of approximately 13% and approximately 32%, respectively, which is about 50% larger than described in Pozzer et al. (2017) using the global EMAC model for the year 2010. Exceptions are seen for concentrations below 5 µgm⁻³ and above 25 µgm⁻³, where the relative share of NH₄⁺ and NO₃⁻ to PM_2.5 is reduced (7% and 17%, respectively).

Although the maximum annual mean PM_2.5 concentration in the EU_emis emission scenario is reduced to 18 µgm⁻³, the composition of aerosols is approximately the same as in the Base simulation (Figure 4, center panel). Primary anthropogenic aerosols (ANT and pORG, cf. Table 3) show the largest share at concentrations above 10 µgm⁻³ with approximately 22% and approximately 14%, respectively. Concentrations below 3 µgm⁻³ are mainly composed of biogenic aerosols (BSOA, sea salt/dust) and SO₄²⁻. NH₄⁺ and NO₃⁻ show a relative constant share of the annual mean PM_2.5 concentrations of approximately 12% and approximately 32%, respectively.

The EU₃₀ emission scenario aims primarily to reduce the highest simulated PM_2.5 annual mean concentrations while keeping emission reductions implementable in the future. Therefore, large emission reductions of 85% are applied to primary anthropogenic aerosol emissions, while other emission reductions (30%) act on those sectors that provide the largest share of NO_x and NH₃. These are the primary sources of NO₃⁻ and NH₄⁺, respectively. As intended by the emission scenario definition, the EU₃₀ scenario shows a decrease in the relative share of pANT to PM_2.5 toward 4%–9% (Figure 4, right panel). However, NO₃⁻ remains the largest share of PM_2.5 in this scenario with a monotonically increasing share of up to 42% for 10 µgm⁻³ to 12 µgm⁻³ annual mean PM_2.5 concentrations.

Although the emission scenarios primarily aim at reducing annual mean PM_2.5 concentrations, the reduction of emissions influences the concentrations of all air pollutants. Figure 5 compares the annual mean concentrations of NO₂ and the seasonal mean O₃ concentrations for all 3 emission scenarios. In comparison to the Base simulation (see Figure 2), annual mean NO₂ concentrations in the EU_emis emission scenario decrease in regions with annual mean NO₂ concentrations above 20 µgm⁻³ in Central Europe as well as the larger cities in Eastern Europe. Here, maximum annual mean NO₂ concentrations of 15 µgm⁻³ are found. Only the Po Valley shows enhanced annual mean NO₂ concentrations of up to 35 µgm⁻³. While annual mean NO₂ concentrations vanish in the Zero_emis scenario, they remain above 10 µgm⁻³ in the most polluted areas in Central Europe in the EU₃₀ scenario. However, all emission scenarios reduce NO₂ annual mean concentrations throughout Europe.

Contrarily, seasonal mean O₃ concentrations remain above the WHO_GL in Europe for all 3 emission scenarios. While the meridional gradient of seasonal mean O₃ concentrations persists (although reduced in strength) in the EU_emis and EU₃₀ scenarios, it turns to a Northeast–Southwest gradient in the Zero_emis emission scenario caused by a stronger decrease of seasonal mean O₃ concentrations in East and Southeast Europe compared to Western Europe, illustrating the larger impact of regional emissions in the East of Europe. In this scenario, the seasonal mean O₃ concentrations are between 60 and 82 µgm⁻³ throughout the continent. The lowest O₃ concentrations are found in the Northeast (60 µgm⁻³), where population density is low and only a limited amount of O₃ and O₃ precursors from outside of Europe are expected. In the EU_emis and EU₃₀ emission scenarios, the seasonal mean O₃ concentrations are decreased by 10–20 µgm⁻³ compared to the Base simulation in most parts of Europe, with concentrations below 80 µgm⁻³ in the West of France and North/Northeast of Europe, where comparable seasonal mean O₃ concentrations in the Base simulation are modeled.

Figure 6 summarizes the distribution of the European population exposed to annual mean concentrations of PM_2.5, NO₂, and O₃ (seasonal mean) for the Base simulation and all 3 emission scenarios in 2016. In the Base simulation, 99% of the European population are exposed to annual mean PM_2.5 concentrations exceeding the WHO_GL (574 million). Here, the distribution shows a peak value of 89 million inhabitants exposed to annual mean concentrations near 14 µgm⁻³, which exceeds the WHO_GL by a factor of 2.8. For the EU_emis simulation, the distribution peaks at 8 µgm⁻³ (150 million inhabitants), whereas the EU₃₀ simulation shows the highest exposure at 6 µgm⁻³ (246 million inhabitants). Besides, by enhancing the emission reductions, the distribution becomes narrower, which leads to a decrease in population exposed to annual mean PM_2.5 concentrations above the WHO_GL by 8% in the EU₃₀ scenario compared to the Base simulation. Thus, the emission reduction is most effective in mitigating the highest simulated PM_2.5 annual mean concentrations above 10 µgm⁻³.

The distribution of population exposure to annual mean NO₂ concentrations in the Base scenario peaks at 4 µgm⁻³, which is below the annual WHO_GL of 10 µgm⁻³. However, because the annual mean concentrations above the WHO_GL are located mainly in Central Europe and larger cities (10% of the European area, compare Figure 2), 323 million inhabitants (56% of the European population) are affected. In addition, about 4.7 million people are exposed to NO₂ annual mean concentrations above 40 µgm⁻³, which is 4 times the WHO_GL. The EU_emis emission scenario reduces the population exposed to annual mean NO₂ concentrations above the WHO_GL to 149 million inhabitants (a 46% reduction compared to the Base simulation). The EU₃₀ emission scenario shows an additional reduction of 54 million inhabitants exposed to annual mean NO₂ concentrations above 10 µgm⁻³. However, as described before (cf. Figure S3), annual mean NO₂ concentrations are underestimated by the model, especially in urban areas, with a mean bias of −6.8 µgm⁻³. Thus, the actual number of inhabitants exposed to air pollution above the WHO_GL is expected to be higher for all emission scenarios.

For O₃, the whole European population is exposed to seasonal mean concentrations above the WHO_GL of 60 µgm⁻³ for all applied emission scenarios. The distribution of population exposure to O₃ in the model results from the Base simulation shows a wide range between 70 and 115 µgm⁻³ and peaks around 95 µgm⁻³, with 30 million exposed inhabitants. While the EU_emis and EU₃₀ emission scenarios decrease the population exposed to seasonal mean O₃ concentrations above 90 µgm⁻³, they show no effect on the population exposed to concentrations below 90 µgm⁻³. However, a full elimination of emissions (Zero_emis scenario) decreases seasonal mean concentrations to below 80 µgm⁻³. In this extreme case of omitting all anthropogenic emissions, the population is exposed to a median seasonal mean O₃ concentration of 68 µgm⁻³.

The change in mortality due to 6 diseases, namely lung cancer, diabetes, chronic obstructive pulmonary disease, stroke, ischemic heart disease, and lower respiratory infection, attributed to air pollution (PM_2.5 and O₃) compared to the Base simulation is given in Figure 7 for the emission reduction scenarios EU_emis, EU₃₀, and Zero_emis. A detailed description of the calculation of mortality is given in the Supplemental Material Text S3. The change in emissions leads to substantial reductions in mortality. On average, the mortality attributed to air pollution reduces by 47% in the EU_emis scenario and 72% in the EU₃₀ scenario (see Table S2). For the Zero_emis scenario, mortality due to air pollution is almost eliminated entirely, with a 99% reduction compared to the Base scenario.

This study evaluates the horizontal distribution of the WHO air quality guideline level excesses and their magnitude in Europe for PM_2.5, NO₂, and O₃ for the year 2016 as simulated by the EURAD-IM. Further, the effect of different emission reduction scenarios on air pollutant concentrations is analyzed, along with resulting changes in mortality attributable to air pollution.

The study uses emission correction factors estimated by a full-year reanalysis of the year 2016. Although a comparison on the resulting air pollutant concentrations with ground-based observations showed that simulated PM_2.5 annual mean, O₃ seasonal mean, and to a lesser extent NO₂ annual mean concentrations differ by less than a factor of 2 from the observations at 97.5%, 99.6%, and 67.2% of sites, respectively, the emission correction factors may mask other modeling errors, such as boundary values or reaction rates.

Although the EU routinely evaluates European air quality, less is published about the exceedances of the WHO_GL. For PM_2.5, NO₂, and O₃, different regions in Europe have been identified, where air quality exceeds the WHO_GL. These regions are Eastern and Southeastern Europe for PM_2.5, Central Europe, and metropolitan areas for NO₂ and Southern Europe for O₃. A few regions, for example, the Po Valley, one of the most polluted areas in Europe, show guideline level excess by up to a factor of 5.6 and 5.2 for PM_2.5 and NO₂, respectively, and by a factor of 2 for O₃. Robotto et al. (2022) attribute the air pollution in the Po Valley to meteorological effects induced by the characteristic orography with mountain ridges on 3 sides of the valley. They concluded that this orographic effect inhibits the air pollution mitigation efforts in the Po Valley making air quality improvements even more challenging. The different emission scenarios support this finding as the EU_emis and EU₃₀ emission scenarios show the highest PM_2.5 annual mean concentrations of 17 and 12 µgm⁻³ in the Po Valley, respectively, which is larger than the PM_2.5 annual mean concentrations of surrounding regions.

The presented evaluation of European air quality illustrates the local effect of air pollution with highest concentrations close to primary sources (except for O₃). However, coordinated efforts need to be performed to reduce both the air quality and climate impact of anthropogenic emissions to avoid mutually negative effects (Maione et al., 2016). For example, this negative effect was reported for vehicular emissions in Europe, where diesel engines have been subsidized over gasoline engines, which effectively decreased CO₂ emissions at the cost of increased PM and NO_x emissions (EEA et al., 2018).

As O₃ mean concentrations are above the seasonal WHO_GL throughout Europe for all emission scenarios that were explored here, adverse effects on crops and vegetation and therefore on harvest yields persist (although to a lesser degree) even after applying emission reductions. Due to the coarser model resolution, simulated NO₂ and PM_2.5 annual mean concentrations are locally expected to be underestimated, especially in valleys in mountainous regions. Further, the investigation revealed that NO₂ annual mean concentrations underestimate observations in urbanized areas due to the low representativity of the 15 × 15 km² model resolution for these areas. Hence, the analysis is considered to provide a lower bound of guideline level excess.

A unique real-life experiment on the effect of emission reduction on air quality was realized by the COVID-19 lockdowns worldwide. Gkatzelis et al. (2021) compared more than 200 studies investigating the effect of COVID-19 lockdowns on air pollution throughout the world. They showed that especially NO_x emission reductions presented positive feedback on NO₂ concentrations during the lockdowns. However, the response of O₃ and PM_2.5 to emission reductions was found to be less pronounced because of the nonlinear chemistry of O₃ and complex aerosol formation processes (Gkatzelis et al., 2021). Overall, most studies related to COVID-19 showed an increase in O₃ and a slight decrease in PM_2.5 concentrations during the lockdown. This was confirmed by Bowdalo et al. (2022) for reductions of NO₂ and PM_2.5 concentrations during the lockdown at European measurement stations. However, they found that daily WHO_GL were still exceeded at more than 50% of the stations for both air pollutants, although the lockdown had been relatively stringent. These trends have been confirmed by the emission scenarios calculated in this study, although the emission scenarios used a stronger emission reduction than that which occurred during the lockdowns (cf. Guevara et al., 2021). Thus, planned mitigation strategies, such as the banning of combustion motor vehicles as main NO_x emitters by the EU may not be sufficient to push NO₂ concentrations (and other air pollutants) below the WHO_GL.

The study reveals the nonlinearity of the O₃ chemistry. While elevated NO₂ annual mean concentrations are associated with reduced O₃ seasonal mean concentrations in most parts of Western and Central Europe (e.g., in Great Britain), the opposite is true in Southern Europe (e.g., in Italy along the Central West Coast), where annual mean NO₂ concentrations above 30 µgm⁻³ and O₃ concentrations above 110 µgm⁻³ (both 20 µgm⁻³ above the surrounding concentrations) coincide. For the simulations conducted in this study, O₃ production is mainly VOC-limited making the O₃ concentrations insensitive to NO_x emission reductions. This becomes evident by comparing the effects of the EU_emis and EU₃₀ emission scenarios on NO₂ and O₃ concentrations. While NO₂ annual mean concentrations are reduced on average by 13% in the EU₃₀ scenario compared to the EU_emis scenario, the O₃ seasonal mean concentrations are reduced on average by 1% (not shown). Thus, the strategy of emission reductions in the emission scenarios is limited for mitigating seasonal mean O₃ concentrations. A simultaneous reduction of anthropogenic VOC emissions benefits the O₃ reduction, as shown by the Zero_emis scenario. However, the full omission of anthropogenic emissions does not lead to seasonal mean O₃ concentrations below 60–82 µgm⁻³. This supports the results of Pay et al. (2019) and Lupaşcu and Butler (2019) that the bulk of ground-level O₃ in Europe originates from long-distance transport. Besides, the meteorological conditions in Southern Europe are more favorable to produce O₃, where seasonal mean concentrations remain 10 µgm⁻³ higher than in Northeastern Europe.

The study reveals that the largest fraction of PM_2.5 is NO₃⁻ (approximately 30%), especially in Central Europe, whereas PM_2.5 concentrations below 4 µgm⁻³ (above 25 µgm⁻³) are mainly governed by SO₄²⁻ (pANT). Furthermore, the study shows that SOA is not important in view of the annual WHO_GL excess in Europe. Thus, the study illustrates that air pollution mitigation strategies focusing on NO₃⁻ and pANT have the largest impact, while SO₄²⁻ is important to push the PM_2.5 annual mean concentration below the annual WHO_GL of 5 µgm⁻³. This finding is the basis for the definition of the EU₃₀ emission scenario. The results of this scenario confirm that the emission reductions, especially the reduction of primary anthropogenic aerosols of 85%, lead to an 8 µgm⁻³ decrease of the highest simulated PM_2.5 annual mean concentrations. Although NO₃⁻ is the largest fraction of PM_2.5 annual mean concentrations, it is likely overestimated in numerical models with coarse grid resolution due to the mixing of NO_x-rich air with the background atmosphere (Zakoura and Pandis, 2018), which cannot be sufficiently quantified with limited NO₃⁻ observations available.

The study shows a limited fraction of 1% of total PM_2.5 concentrations from anthropogenic SOA. However, the effect of anthropogenic emissions on SOA is not agreed upon in the literature. Zheng et al. (2015) showed almost no response of SOA concentrations due to changes in NO_x emissions, which they attribute to the lack of change in the chemical regime due to the emission reduction. In contrast, Xu et al. (2015) highlight the role of anthropogenic SO₂ and NO_x emissions in the mediation of SOA formation. Further, a current study reveals that the SOA production in EURAD-IM is potentially too low as analyzed in sensitivity studies in comparison with in situ observations (Lu Liu, personal communication, September 8, 2023). Thus, the relative importance of emitted species on the reduction of aerosol concentrations may change as different chemical regimes become more important with changing aerosol and atmospheric chemical composition.

The targeted emission reductions by the EU, which are used in the EU_emis and EU₃₀ emission scenarios, focus on the emissions of selected species only, namely, NO_x, SO₂, NMVOC, NH₃, and PM_2.5. No requirements on sectoral emission reductions are given in the directive (European Parliament and Council of the European Union, 2016). Therefore, an opportunity arises to determine the optimal sectoral emission reductions out of the EU target values. Table 4 summarizes a possible sectoral emission reduction plan to achieve the EU target values. These sectoral emission reductions are computed using the bounded least squares method and European averaged emission values for all species as shown in Table S1 in the Supplemental Material. It is noted that the emission reductions in the EU_emis and EU₃₀ emission scenarios are applied to the aggregated bulk emissions per species and not for each sector individually. The example in Table 4 intends to show that the emissions for each sector need to be reduced largely, in this example by 41%–79% on European average to achieve the target values, whereas locally different emission reduction plans may be more appropriate. The EU_emis emission scenario reveals that the annual mean PM_2.5 concentrations remain up to a factor of 2–3 above the WHO_GL in large parts of Europe (especially the Po Valley and Southeastern Europe) even after applying the emission reductions targeted by the EU.

This study also shows the effect of emission reductions on mortality due to air pollution. The meta-regression–Bayesian regularized trimmed exposure model, defined by GBD 2019 Risk Factors Collaborators (2020) and used by the GBD, has been used here. The estimated mortality attributable to air pollution should be interpreted as a lower bound, as cause-specific illnesses have been included, while a larger health burden is to be expected if all noncommunicable diseases were included (see Burnett et al., 2018). As of the coarse model resolution, both population data and air pollutant concentrations (especially close to local emitters) are expected to show a heterogeneous subgrid-scale distribution. Thus, it is assumed that the calculated exposure varies within the uncertainty range given in Table S3. The European Commission (2021) released the “zero pollution” action plan as an addition to the European Green Deal that aims, among others, at reducing premature deaths by air pollution by 55% in 2030 based on 2005 air pollution. The present study shows that targeted emission reductions by the European Commission, which are used in the EU_emis scenario, reduce premature deaths by 47% based on 2016 air pollution. As the target values by the European Commission are based on air pollution in 2005, a direct comparison of this study’s results with the target values is not possible. However, as the premature deaths have declined by about 30% in 2016 compared to 2005 (cf. figure 10.1 in EEA et al., 2018), an additional reduction by 47% is in line with the target value by the European Commission is achievable.

The study evaluates European air quality as simulated by the EURAD-IM using updated emissions for the year 2016. The focus is placed on annual mean concentrations of PM_2.5, NO₂, and seasonal mean DM8 O₃ concentrations in view of the WHO_GL released in 2021. Further, the study presents the effect of potential emission reductions on the conformity of European air quality with the guideline levels.

Air quality is a persistent problem throughout Europe, with some regions such as the Po Valley in Northern Italy being heavily exposed to elevated concentrations of all 3 air pollutants. In these areas, air pollutant concentrations exceed the WHO_GL by a large amount (e.g., a factor of 5.6 and 5.2 for PM_2.5 and NO₂ in the Po Valley, respectively), which makes air pollution mitigation more challenging.

The study reveals that emission reductions of approximately 40% and more throughout Europe are required to achieve a substantial reduction in mortality due to the 6 primary diseases understood to be exacerbated by air pollution. The targeted emission reductions by the EU are shown to achieve a mortality reduction of 47% compared to 2016.

Overall, air pollution in Europe exceeds the WHO_GL. EU’s emission reduction plans are not sufficient to decrease annual mean concentrations to meet the guideline levels. Thus, new approaches involving artificial air pollution sinks need to be designed and implemented to reduce the exposure of the European population and avoid adverse health effects through air pollution.

Model output data of daily mean concentrations for species discussed in this study, mortality calculations, and population data are available at https://doi.org/10.5281/zenodo.10046058 (Franke et al., 2023).

The supplemental files for this article can be found as follows:

WHO_Guidelines_Supplement_final (Doc file).

This file contains the following supplemental material:

Text S1. Description of population data used to calculate population exposure for the European states.

Text S2. Description and derivation of the emission reductions applied in the EU_emis and EU₃₀ emission scenarios.

Text S3. Derivation of mortality calculations.

Figure S1. Population density across Europe for 2016.

Figure S2. Emission reductions applied in the sensitivity simulations EU_emis and EU₃₀.

Figure S3. Comparison of modeled annual mean concentrations with observations for PM_2.5, NO₂, O₃, and PM₁₀.

Table S1. Committed EU emission reductions.

Table S2. Global mortality for each emission scenario.

Table S3. Mortality by European country, for each emission scenario.

The authors gratefully acknowledge the computing time granted by the JARA Vergabegremium and provided on the JARA Partition part of the supercomputer JURECA at Forschungszentrum Jülich.

The study was partially funded by the German Environment Agency (Umweltbundesamt) with research ID FK3717512520.

AP is an associate editor of Elementa but does not interfere with the review process of this manuscript, as this is ensured by the publisher. Besides this, the authors declare to have no competing interests.

Contributed to conception and design: PF, ACL, AKS.

Contributed to acquisition of data: PF, ACL.

Contributed to analysis and interpretation of data: PF, ACL, AKS, AW, BS, AP.

Drafted and/or revised the article: PF, ACL, AW, BS, AP.

Approved the submitted version for publication: PF, ACL, AW, BS, AP.

How to cite this article: Franke, P, Lange, AC, Steffens, B, Pozzer, A, Wahner, A, Kiendler-Scharr, A. 2024. European air quality in view of the WHO 2021 guideline levels: Effect of emission reductions on air pollution exposure. Elementa: Science of the Anthropocene 12(1). DOI: https://doi.org/10.1525/elementa.2023.00127

Domain Editor-in-Chief: Detlev Helmig, Boulder AIR LLC, Boulder, CO, USA

Guest Editor: Frank Flocke, National Center for Atmospheric Research, Boulder, CO, USA

Knowledge Domain: Atmospheric Science