During boreal springtime of 2020, worldwide measures for curbing the spread of the COVID-19 virus have led to unprecedented and abrupt lockdowns in transportation (road, aeroplanes, and ships) and industry. These strong limitations drastically reduced the emissions of anthropogenic pollutants, inducing significant changes in atmospheric composition and air quality from local to worldwide scales, and particularly in regions such as China (e.g. Le et al., 2020) and Europe (e.g. Menut et al., 2020). Gkatzelis et al. (2021) provided an exhaustive overview of the current understanding of the influence of emission reductions due to the lockdown throughout the world on atmospheric pollutant concentrations, where air quality is described in terms of nitrogen dioxide (NO2), particulate matter (PM2.5), ozone (O3), ammonia, sulfur dioxide (SO2), black carbon, volatile organic compounds (VOCs), and carbon monoxide (CO).

Numerous efforts have been made to document and quantify the reduction in the atmospheric abundance of primary air pollutants, such as nitrogen oxides NOx (NO+NO2) and sulfur oxides SOx (SO2 + H2SO4), which are directly emitted to the atmosphere. Satellite measurements derived from the TROPOspheric Monitoring Instrument (TROPOMI) and Ozone Monitoring Instrument (OMI) have shown reductions in NO2 total columns of about 30 %–40 % over Chinese, European, and North American source regions (e.g. Bauwens et al., 2020; Muhammad et al., 2020; Le et al., 2020; Potts et al., 2021) during the pandemic lockdown. Surface in situ measurements have also revealed clear reductions in the same order of magnitude of NO2 (e.g. Ordóñez et al., 2020; Barré et al., 2021) and SO2 (e.g. Le et al., 2020).

Modelling approaches based on the construction of emission inventories accounting for the changes in anthropogenic activities have also shown consistent reduction of these primary pollutants (NOx and SOx) but also different regional regimes of either decreases or increases in the amounts of secondary pollutants, which are produced by photochemical reactions in the atmosphere (e.g. Le et al., 2020; Menut et al., 2020; Giani et al., 2020). This more complex picture is shown for secondary aerosols (organic or inorganic species) and tropospheric ozone.

An enhancement of the production of secondary particles has been linked to the alteration of atmospheric oxidizing capacity during the pandemic lockdowns (Le et al., 2020). However, the total amount of aerosols, which partly have primary origin, has shown an overall reduction in terms of PM2.5 (particle matter with aerodynamic diameter less than 2.5 µm) by roughly 5 %–10 % over Europe and 10 %–20 % over China, during the 2020 springtime lockdown period.

On the other hand, ozone pollution either decreased or increased due to the reduction in the emissions of primary pollutants, depending on the photochemical regime. This has been shown by modelling and in situ observational studies, both in China and Europe (e.g. Le et al., 2020; Shi and Brasseur, 2020; Giani et al., 2020; Souri et al., 2021; Mertens et al., 2021; Nussbaumer et al., 2021). Ozone decreased in rural areas, where photochemical production of ozone is controlled by the abundance of nitrogen oxides (thus a NOx-limited regime), the concentrations of which clearly decreased during the pandemic lockdown. Ozone pollution was clearly enhanced in urban areas (Ordóñez et al., 2020) and particularly in megacities (Sicard et al., 2020), as the reduction of NO2 amounts strongly inhibited night-time titration. The reduction of this dominant sink led to accumulation and therefore enrichment of ozone, which prevailed over the reduction of ozone production associated with the lack of precursors. Although these processes have been observed in previous works, the quantification of them is complex and it remains a challenge, as well as the precise areas over which different regimes have prevailed. An additional major factor affecting the abundance of ozone, which adds complexity to its analysis, is the meteorological conditions. Using in situ surface observations and a predictive model of ozone variation against meteorological conditions, Ordóñez et al. (2020) suggested that clear-sky conditions and relatively high surface temperature in April 2020 as compared to previous years induced intense ozone enhancements (in terms of daily 8 h maximum) over northern Europe, comparable and even larger than the enhancement associated with the pollutant emission changes linked to the pandemic lockdowns. Based on observation and meteorological analysis over the last 7 years Deroubaix et al. (2021) suggested that the total oxidant concentrations (Ox + O3 + NO2) decreased in southwestern Europe while they remained unchanged in northern Europe. Using a chemistry-transport model with emission adjusted with respect to satellite measurements of NO2 and formaldehyde, Souri et al. (2021) suggest a significant role of meteorological conditions but smaller than that associated with anthropogenic emission changes during lockdowns (58 % of the ozone enhancement, where it is 42 % for meteorology change between 2020 and 2019). Moreover, a larger-scale study based on ozone measurements from in situ sondes and lidars observed a reduction of free troposphere ozone across the Northern Hemisphere of about 7 % (4 ppb) due to worldwide activity reduction during spring and summer 2020 (Steinbrecht et al., 2021). It is worth noting that none of these studies have used satellite measurements of ozone concentrations themselves as a complementary source of information.

The present work presents the first study integrating satellite ozone observations of lowermost tropospheric (LMT) ozone, surface in situ measurements, and chemistry-transport modelling for analysing the reductions and enhancements of ozone pollution over Europe associated with the COVID-19 lockdown of spring 2020. We use satellite measurements derived from the IASI+GOME2 (Infrared Atmospheric Sounding Interferometer + Global Ozone Monitoring Experiment 2) multispectral synergism, which provides currently unmatched sensitivity to near-surface ozone (Cuesta et al., 2018). The sensitivity of this satellite retrieval shows a relative maximum for ozone around 2.2 km of altitude over land (while it is typically above 3 km for other standard satellite approaches; see details in Cuesta et al., 2013, 2018). It offers a full cloud-free spatial coverage, which is complementary to in situ measurements. We also use the chemistry-transport model CHIMERE (Menut et al., 2013; Mailler et al., 2017) for assessing the consistency of this model with respect to the observations and to analyse the influence of meteorological conditions in 2020 as compared to 2019 (the year used as baseline or standard conditions in this study). The consistency of satellite data and simulations is analysed with respect to the Air Quality e-Reporting database providing near-real-time air-quality measurements in Europe. Section 2 of the paper describes these datasets and the strategy used for their analysis. Section 3 presents the results, first in terms of total differences between 2020 and 2019, secondly the influence of meteorological conditions, and finally the impact of the COVID-19 lockdown on ozone pollution. Additional discussions on anthropogenic emissions and meteorological conditions are also given in Sect. 4. A sensitivity study shown in the Supplement of the current paper considers a different setup of the CHIMERE model, in terms of emission inventories and meteorological models, for analysing possible uncertainties in the simulations. A last Sect. 5 provides the conclusions of this work.

Being a secondary and reactive pollutant, the analysis of tropospheric ozone is complex since it is influenced by many competing factors. Tropospheric ozone sources and sinks are affected by multiple photochemical processes and non-linear effects, which vary in time and space according to the availability of different precursors, either NOx or VOC (originating themselves from multiple anthropogenic and natural sources) and meteorological conditions. An integrated approach including multiple datasets of different kinds is therefore valuable for better understanding the evolution of this pollutant. In the current study, we use both satellite and in situ surface data to quantify the difference between the ozone pollution observed over Europe in spring 2020 with respect to that in spring 2019, which is affected by both the lockdown reduction of pollutant emissions and meteorological conditions. First, we verify the consistency of these observations with respect to simulations from CHIMERE. This model simulates the 2 years, with a standard or business-as-usual inventory for 2019 and a modified version in 2020 accounting for the lockdown conditions, hereafter referred to as STD and COVID, respectively. Second, we use the chemistry-transport model with the same inventory (standard) for both years, 2020 and 2019, to derive a first-order adjustment for accounting for the influence on ozone pollution of the change in meteorological conditions between the 2 years. This allows us to obtain an approximate estimate of the effects of the COVID-19 lockdown conditions on ozone pollution from the two observational datasets (in situ and satellite). This is expressed by the following equation: 1ΔO3obs&modcovid≈O3obs2020-O3obs2019-O3modSTD2020-O3modSTD2019, where modSTD corresponds to simulated ozone with the standard or business-as-usual inventory. This adjustment does not rely on any estimation of the variations in anthropogenic emissions during the lockdown. For simulations, the superscripts 2020 and 2019 refer to the year of meteorological conditions that have been used. For observations, these superscripts indicate the year when they are performed. The same equation stands for the adjustment of both surface measurements and satellite retrievals (which are adjusted independently) but gridded at the location or horizontal resolution of each observation. The adjustments for surface measurements and satellite retrievals are different, as they use simulated ozone concentrations, respectively, at the surface and at the lowermost troposphere (LMT, below 3 km of altitude), while these last ones are smoothed by the satellite averaging kernels. The accuracy of this first-order adjustment or correction depends on the performance of the chemistry-transport model to simulate ozone concentrations in business-as-usual conditions as a function of meteorological conditions. It implicitly accounts for changes in biogenic emissions between 2020 and 2019 which are directly linked to meteorological conditions. However, this adjustment cannot account for the changes in the chemical regimes (either NOx-limited or VOC-limited) due to the changes in the abundance of ozone precursors nor other complex chemical regimes (as modelled for parts of the Po Valley, Thunis et al., 2021).

We compare this synergetic observational and model estimate with the one derived from models only, based on the difference of ozone simulations with inventories COVID and STD and the meteorological conditions of 2020, as follows: 2ΔO3modcovid=O3modCOVID2020-O3modSTD2020.

In the current work, we mainly focus on the period from 1 to 15 April 2020 (and use 1–15 April 2019 as standard period), which corresponds to the 15 d period of most perturbed emissions by the lockdown according to the COVID emission reduction factors from CAMS (https://atmosphere.copernicus.eu/covid-data-download, last access: 1 April 2022). This work reports for example a reduction of -73 % of gasoline road transport for the five most populated countries in Europe (Germany, France, Great Britain, Italy, and Spain, which are also the largest contributors to the total European emission decreases), whereas it is -64 % and -66 % for the 15 d before and after (and then -54 % and -38 % for the two following fortnights). We also show some of the results for the whole month of April (2020 and 2019) to directly compare our estimates with those from previous works (Ordóñez et al., 2020; Souri et al., 2021) provided as monthly averages.

The following paragraphs describe briefly the datasets used by this approach.

For better understanding the information provided by observations and simulations, the current multi-data analysis is presented in several steps. First, we focus on the total changes of ozone pollution between 2020 and 2019 (Sect. 3.1) that are directly observed by in situ sensors and from space, and which we compare with the corresponding amount simulated by the model. Then, we analyse the model-derived changes between these 2 years but only associated with meteorological conditions (Sect. 3.2). Finally, we compare changes of ozone pollution only linked with the pandemic lockdown conditions, estimated from models and observations (Sect. 3.3). The originality of these results resides in the use of in situ and satellite observational estimates of the changes in ozone pollution associated the lockdown conditions ΔO3obs&modcovid, which are adjusted for subtracting meteorology effects.

A fair qualitative consistency is observed between observations (in situ and satellite) and CHIMERE datasets with respect to the regions of ozone enhancements and reductions associated with the pandemic lockdown conditions. Ozone increased over more urbanized regions with large NOx emissions, such as central northern Europe and the Po Valley, and mainly decreased far from NOx hotspots where night-time titration plays a less important role in the surface ozone daily balance. The significant model/observation difference over eastern Germany and Poland might be linked to the difficulty to accurately simulate neutral chemical regimes. There, an accurate simulation of competing terms of ozone production and sinks requires a larger precision of the emission inventories and other factors affecting the abundance of active chemical species.

Significant differences are seen in the amplitudes of the positive and negative anomalies of ozone pollution in spring 2020. On the one hand, in situ and satellite observations agree in terms of the total range of changes of ozone (both between 2020 and 2019 and adjusted for meteorological effects) of about ∼ 25 ppb between maximum positive and minimum negative anomalies over land, with satellite data shifted by ∼ 8 ppb towards smaller values (which can partly be linked to the altitude of sensitivity of these retrievals). On the other hand, CHIMERE underestimates the amplitude of these changes in ozone pollution (with the largest range for variability of ∼ 12 ppb), with respect to observations. This limited variability of ozone with respect to observations is also found for the chemistry-transport model of Souri et al. (2021) that uses adjusted European emissions of NOx and VOC with respect to satellite data. Particularly, we notice underestimations of the absolute amplitude of negative anomalies which are seen over land and over ocean with respect to satellite data. This may be partly explained by the fact that the model only accounts for the changes in ozone concentrations and its precursors associated with the pandemic lockdown over Europe, but other changes at a global scale are not considered. Nevertheless, the observational study of Steinbrecht et al. (2021) suggests that these last ones correspond to a significant ozone reduction in the free troposphere across the whole Northern Hemisphere in spring and summer 2020 (-7 % or -4 ppb over the northern extratropics). We likely expect this large-scale ozone reduction to be more pronounced at the beginning of this period (e.g. beginning of April 2020) as it is a period with more generalized and severe lockdowns over Europe and North America, as well as the last part of the lockdown over China and some restrictive measures over South Korea and Japan. Moreover, the ozone reduction in spring/summer 2020 is more noticeable near the North Sea (Lerwick) than for the south of France (Haute Provence) near the Mediterranean, as observed with the IASI+GOME2 retrievals over these two oceanic regions (Figs. 1a and 7a).

The amplitude of the ozone anomalies associated with the pandemic lockdown simulated by CHIMERE is likely related to the abundance of ozone precursors (Fig. 6). In 2019, NO2 simulated concentrations are lower than those measured by in situ measurements over a large heterogenous area such as the Benelux to northern Italy band and for a megacity such as Paris (see Table 6). For all in situ stations, the mean biases and root-mean-squared differences of coincident NO2 concentrations simulated in 2019 are quite significant (respectively around -5 and 6 ppb). When comparing with in situ measurements of ΔNO22020–2019, the measured reduction over a horizontally homogenous hotspot such as Paris clearly matches CHIMERE. Inversely, it is overestimated (by a factor ∼ 4) in absolute values over a heterogenous large area (such as the Benelux to northern Italy band). Therefore, the agreement between simulated and measured ΔNO22020–2019 clearly depends on the criteria and the horizontal homogeneity of the abundances over the area considered in the comparison.

The uncertainties in the simulated NO2 concentrations are partly linked to the inventories used by the model. The CHIMERE model used here is based on emission inventories estimated for 2010 (by HTAPv2.2, on a monthly basis). A sustained negative trend for NO2 concentrations observed over Europe between 2010 and 2017 (e.g. Pazminio et al., 2021) suggests a positive bias for this inventory. This trend is ∼ 30 % in France, slightly higher in Italy or Belgium, and smaller for other countries such as Germany and Poland (∼ 13 %; see EMEP database, https://www.ceip.at/webdab-emission-database/emissions-as-used-in-emep-models, last access: 1 April 2022).

Other factors significantly affecting simulated concentrations of ozone and its precursors are clearly linked to the meteorological fields used by the model. This is shown in terms of changes on 2020 with respect to 2019 of ozone photolysis rates, surface temperatures and winds, and mixing boundary layer heights used by CHIMERE (Fig. 7). Two distinct behaviours are clearly observed over the continent north of 44∘ N and over the Iberian Peninsula. North of 44∘ N, anticyclonic conditions prevailing in 2020 induced clearer sky conditions (thus enhancements of ozone photolysis rates), higher surface temperatures, and lower wind speeds, which clearly favour photochemical production of ozone. This explains the frank positive anomaly of surface ozone over this region visibly simulated by CHIMERE, accounting (Fig. 3a) or not (Fig. 4a) for the emission changes during the lockdown. Over the Iberian Peninsula, reduced ozone photolysis rates (Fig. 6a) associated with enhanced cloudiness in 2020 are likely at the origin of the meteorology-associated decrease in ozone concentrations (Fig. 3b). However, other meteorological conditions likely produce the opposite effect: enhanced surface temperatures and lower wind speeds in 2020 are expected to favour ozone production and shallower mixing boundary layers to inhibit turbulent vertical dilution of ozone, thus inducing a relative enhancement of surface ozone concentrations in 2020. These effects are expected to compensate between them, explaining the moderate reduction of ozone simulated by CHIMERE over this region (-2.4 % for the southwestern region in Table 5).

Furthermore, the variability of ozone at the free troposphere may also be a significant factor influencing near-surface ozone, depending on vertical mixing. The enhanced anticyclonic conditions in 2020 with respect to 2019 are particularly seen north of 44∘ N by increased geopotential heights and lower wind speeds at 850 hPa (Fig. 8a). This situation favours subsidence and thus vertical advection of air masses from the free troposphere down to the atmospheric boundary layer. This is less clearly noted over the Iberian Peninsula and the eastern Mediterranean, where a transition between lower and higher geopotential heights is seen (Fig. 8a). Ozone anomalies at the upper troposphere are depicted by IASI+GOME2 retrievals between 6 and 12 km in Fig. 8c. They mainly reveal an overall reduction of ozone concentrations in 2020 with respect to 2019, particularly over the North Sea and the central Mediterranean. This is probably related with the large-scale reduction of free tropospheric ozone in 2020 observed by Steinbrecht et al. (2021), mainly related with the lockdown-associated drop of precursor emissions over the Northern Hemisphere. Downward mixing of these ozone-poorer air masses probably contributes to the large-scale reduction of ozone observed at the LMT by IASI+GOME2 and its negative shift with respect to surface concentrations (Figs. 1 and 2a). Indeed, the only geographically coincident patterns observed both at the LMT and the upper troposphere are the ozone reductions of ozone over the Mediterranean and the North Sea.

At the upper troposphere, a near-zero variation is observed over northeastern Europe and an ozone enhancement over the western Iberian Peninsula (Fig. 8c). This last one is probably associated with coincident lower tropopause heights (Fig. 8b) and thus with a relatively larger contribution of stratospheric ozone. Over the North Sea, the reduction of upper tropospheric ozone at 6–12 km of altitude is strengthened by a depletion of stratospheric ozone occurring in 2020 (see in Fig. 8d as ozone anomalies with respect to 2019).

We have presented a comprehensive analysis using in situ and satellite measurements of ozone as well as chemistry-transport modelling tools of the changes in ozone pollution over Europe associated with the COVID-19 pandemic lockdown of springtime 2020. To the authors' knowledge, this is the first time that satellite direct observations of lowermost tropospheric ozone are used in such a multi-data analysis. While satellite observations of ozone show a fairly good agreement with respect to in situ surface measurements of ozone across Europe in the regions where both are available, only satellite data provide full horizontal coverage both over land and ocean (at 0.25∘ × 0.25∘ horizontal resolution). The observations quantify the changes of ozone pollution in 1–15 April 2020 with respect to the same period the previous year, thus the 15 d period when the lockdown measures modified the most anthropogenic activities over the continent.

An additional original aspect in the present analysis is the adjustment of both in situ and satellite observations for estimating the impact of the pandemic lockdown on ozone pollution, using measurements from 2020 and 2019. This adjustment is derived from the chemistry-transport model CHIMERE using the meteorological conditions of each year but the same standard (business-as-usual) anthropogenic emissions. This method relies on the model accuracy in standard conditions, and it neglects possible feedbacks between meteorological conditions and photochemical regimes. The influence of biogenic emissions is accounted for by the method, as these sources are derived according to the meteorological conditions. The adjustment can be directly applied to both in situ and satellite data, with good consistency with respect to other independent approaches used for the same kind of adjustment of surface in situ data (Ordóñez et al., 2020) or using integrated process rates derived from a chemical transport model (Souri et al., 2021).

The satellite and in situ observational estimates of the changes in ozone pollution associated with the pandemic lockdown show a significant enhancement of ozone in the VOC-limited regions of central and northern Europe and the Po Valley, as pointed out previously by models and in situ surface data (e.g. Menut et al., 2020; Ordóñez et al., 2020; Souri et al., 2021). This effect overlaps with a large-scale reduction of ozone over the whole continent (seen in the free troposphere by ozone sondes, Steinbrecht et al., 2021), which is clearly evidenced over the ocean (Atlantic, North and Mediterranean seas).

We compare these observation-based estimations with model-only estimates derived by changing emissions according to the reductions in anthropogenic activities estimated for the lockdown period. The model shows similar regional patterns of ozone enhancements and reductions linked to the lockdown, except for regions with neutral photochemical regimes (such as eastern Germany and Poland). However, it underestimates the amplitude of the positive and negative anomalies. Furthermore, the model does not simulate the ozone decrease observed at a large hemispheric scale or the stratospheric influence, as the simulation domain covers Europe and the troposphere. Sensitivity studies changing the emission inventory and meteorological conditions of the model highlight a particular sensitivity to vertical mixing within the lower atmospheric layers as a key factor influencing the amplitude of the ozone anomalies, as well as the emission inventories used for standard and lockdown conditions. The differences between simulations and observations highlight the complexity to simulate the effect of the changes in ozone concentrations due to changes in anthropogenic emissions, as occurred during the lockdown.