Ozone Anomalies in the Free Troposphere During the COVID-19 Pandemic

Using the CAM-chem Model, we simulate the response of chemical species in the free troposphere to scenarios of primary pollutant emission reductions during the COVID-19 pandemic. Zonally averaged ozone in the free troposphere during Northern Hemisphere spring and summer is found to be 5%–15% lower than 19-yr climatological values, in good agreement with observations. About one third of this anomaly is attributed to the reduction scenario of air traffic during the pandemic, another third to the reduction scenario of surface emissions, the remainder

Most of the early data analyses about the effect of the pandemic on air quality have focused on chemical species anomalies at the Earth's surface and were based on measurements from monitoring stations Shi & Brasseur, 2020) and, for a limited number of species (e.g., nitrogen dioxide), on information deduced from satellite observations (e.g., the Tropospheric Monitoring Instrument; Bauwens et al., 2020). Little information on the effects of the chemical perturbations during the COVID-19 pandemic in the free troposphere is currently available. A recent study (Steinbrecht et al., 2021) based on ozone measurements by balloon-borne ozone sondes as well as ground-based FTIR and LIDAR systems during the period 2000-2020 at latitudes 82.5°N-54.5°S reported changes of free tropospheric ozone related to the COVID-19 disruptions. It shows that, from April to August 2020 and from 1 to 8 km altitude, the average concentration of ozone was 7% lower than the climatological mean values across most of the Northern Hemisphere.
To help interpret the reduced ozone concentrations, we use the global Community Atmosphere Model with chemistry (CAM-chem) and, assuming a few emission reduction scenarios, quantify the relative importance of the different processes that have contributed to the observed ozone anomalies. Unlike the situation in the boundary layer where the lifetime of ozone is of the order of a few days (Goldberg et al., 2015), the timescales associated with the temporal evolution of odd oxygen (O x = O 3 + NO 2 ) in the free troposphere are of the order of several weeks (Stevenson et al., 2006), or even several months (Bates & Jacob, 2020) if one includes hydrogenated compounds (HO x and its chemical reservoirs) in the definition of O x . The behavior of ozone in the free troposphere therefore depends both on photochemical processes and on the effect of transport due to the atmospheric circulation.
During the year 2020, several events potentially affected ozone in the free troposphere: (a) the intense world-wide disruption of the surface emissions of primary pollutants in response to the COVID-19 pandemic; (b) the related reduction in air traffic with a reduced injection of NO x , SO 2 , and black carbon (BC) into the upper troposphere; (c) the particularly intense depletion of ozone in the lower Arctic stratosphere due to the abnormally stable and vigorous polar vortex during the first months of 2020 (Inness et al., 2020;Manney et al., 2020;Wilka et al., 2021;Wohltmann et al., 2020), (d) the interannual variability associated with meteorology, lightning and fires. The perturbation in air traffic during the COVID-19 pandemic has also modified the density of aircraft-generated contrails and cirrus clouds (Schumann et al., 2021).
Here, we estimate the response of free tropospheric ozone to the aforementioned potential causes of the 2020 ozone anomaly by performing several sensitivity simulations in which the different sources of disturbances are taken into account. We compare the simulated overall responses with observed ozone anomalies from Steinbrecht et al. (2021).

Model Description and Overview of Simulations
We use the Community Earth System Model (CESM) version 2.2 (Danabasoglu et al., 2020;Gaubert, Emmons, et al., 2020;Gettelman et al., 2019;Tilmes et al., 2020), described and evaluated in Gaubert et al. (2021). We adopt the MOZART Troposphere Stratosphere (TS1) chemistry mechanism , which includes 221 gas phase and aerosol species and 528 chemical and photochemical reactions. Aerosol concentrations and size distribution are derived from the four-mode Modal Aerosol Model (MAM4, Liu et al., 2016;Mills et al., 2016). In order to realistically represent meteorological conditions for the period under consideration, the wind velocity components and the temperature are nudged toward the MERRA-2 meteorological analysis (Gelaro et al., 2017). Figure S1 shows the calculated zonally mean of NO x and ozone concentrations averaged over the month of 2020 (baseline case = control).
Baseline anthropogenic surface emissions rely on the Copernicus Atmosphere Monitoring Service (CAMS)-GLOB-ANT_v4.2-R1.1 global inventory (Elguindi et al., 2020;Granier et al., 2019). Three-dimensional aircraft emissions are based on Hoesly et al. (2018). Biogenic emissions are calculated online from the Model of Emissions of Gases and Aerosols from Nature (Guenther et al., 2012). Daily biomass burning emissions are based the Quick-Fire Emissions Dataset (Darmenov & da Silva, 2014) include, for example, the 2019/2020 large fires in California, Colorado, and Australia. Deposition of gases and aerosols are calculated through an active coupling between the atmosphere and the Community Land Model version 5 (Lawrence et al., 2019). To account for the effect of the COVID-19 lockdowns, we follow the emission reductions estimated in Doumbia et al. (2021). Anthropogenic emissions are reduced for each economic activity sector (industrial, mobility, residential, energy, shipping, and aviation) and geographical region according to the central best estimate of the CONFORM dataset , see also Figure S2). The emission reductions adopted in other studies (Lamboll et al., 2021;Mertens et al., 2021;Miyazaki et al., 2021;Weber et al., 2020) can be rather different, which highlights the substantial uncertainties in these estimates. The uncertainties estimated by Doumbia et al. (2021) for different sectors in different regions of the world are provided in Table S1. Globally, the uncertainties range from ±12% to ±25%. Nevertheless, the different model simulations performed for the present study (summarized in Table S2), as well as results similar to ours obtained in the studies of Weber et al. (2020), Miyazaki et al. (2021), and Mertens et al. (2021), indicate that the responses in the free troposphere depend more or less linearly on the chosen reduction scenarios and their magnitude. Two specific uncertainties, which have a large influence on the ozone response in the middle to the upper troposphere are assessed explicitly here: (a) the calculated springtime ozone depletion in the Arctic, which is a strong function of the adopted denitrification rate inside the polar vortex and (b) the reduction in air traffic during the pandemic.
We first present the calculated anomalies in the concentration of chemical species in the troposphere during year 2020 relative to a baseline case in which the COVID-related changes in the emissions are ignored. These numerical experiments only quantify changes due to the anthropogenic emissions following lockdowns across the world. We consider three scenarios: changes only in the surface emissions during the pandemic (Case 1 = COVID-surf − Control); changes only in the air traffic emissions (Case 2 = COV-ID-airc − Control) and the combined effects (Case 3 = COVID-ALL − Control). We focus on the monthly mean changes in the global distribution of NO x and ozone in a global domain extending from the surface to the lower stratosphere and from pole to pole. We then assess the contribution of meteorological inter-annual atmospheric variability, including the influence of the exceptionally high ozone depletion inside the 2020 Arctic vortex, by comparing the baseline 2020 results (no COVID related effects included) with 2001-2019 climatology (Case 4 = Control − CLIMO). Finally, we perform a comparison similar to Case 4, but with the year 2020 simulation accounting also for the reduced anthropogenic emissions during the pandemic (Case 5 = COVID-ALL − CLIMO). This last case can be compared with the results of Steinbrecht et al. (2021), in which observed ozone concentrations in 2020 are contrasted to the observed ozone climatology. In Case 6 (=COVID-ALL-W − CLIMO), we assess how higher denitrification rate in the polar vortex, which affects the intensity of the springtime Arctic stratospheric ozone depletion, also increases the calculated tropospheric ozone anomaly in 2020. Case 7 (=COVID-ALL2 − Control) assesses the sensitivity of the calculated ozone perturbation to the magnitude of the reduction in aircraft emissions. This case is similar to Case 3 (=COVID-ALL − Control), but with the COVID-19 adjustment in aircraft emissions reduced by one-third. Figure 1 shows the response of the zonally and monthly averaged ozone concentration to the perturbed emissions (Case 3 = COVID-ALL − Control) relative to the baseline case in which no lockdown effect is applied to the emissions. We note the gradually larger reduction in the ozone concentration as time proceeds and photochemical activity increases; the relative anomaly does not exceed 2% in March, but reaches 7% in May and June before it slightly decreases in July (panels a-f of Figure 1). While the lockdown measures were stricter in the Northern Hemisphere winter and spring 2020, the photochemical response of ozone was largest in summer. The relative changes in the concentration are more pronounced in the lower to middle troposphere (800-300 hPa, or 2-9 km altitude), but the absolute changes (up to 8 ppbv in June, see Figure 1, panels g-i) are largest at higher altitudes (between 300 and 200 hPa or 9 and 12 km, respectively) in the extratropics of the Northern Hemisphere. For Case 7, in which the reduction in air traffic emissions is one third smaller than in Case 3, the maximum absolute ozone depletion (ppbv) located near 300 hPa in the northern extratropics is smaller by about 1-2 ppbv (roughly 30% less) during the May-July 2020 period (panels j-l of Figure 1). When examining the relative changes, we also note that the location of the maximum response evolves with latitude following the mean solar radiation. The largest response occurs first in the tropics (March and April) with a gradual displacement toward the northern polar region (May-July).

Results
We now investigate the contribution of the different forcing factors to the calculated ozone anomaly. We focus on June 2020 during which the ozone reduction is largest. In Figure 2 . Panels a-f shows relative changes from February to July 2020 (%). Panels g-i shows similar results but in absolute terms (ppbv) for the period of May-July 2020. Panels j-l shows the same, but with the COVID-19 related adjustment in air traffic reduced by one third (Case 7 = COVID_ALL2 − Control). and monthly mean concentrations of NO x (panels a-c), ozone (panels d-f) and particulate matter (PM 2.5 , panels g-i) to the changes in surface emissions (middle panels, Case 1 = COVID-surf − Control) and aircraft emissions (right panels, Case 2 = COVID-airc − Control), and to the combined changes (left panels, Case 3 = COVID-ALL − Control). In the case of NO x , the response to the reduced surface emissions (Case 1 = COVID-surf − Control) is generally largest in the planetary boundary layer (larger than 10%), except BOUARAR ET AL.  in the tropics where NO x -depleted near-surface air masses are lifted to the upper troposphere by convective transport resulting in 5%-8% reductions in the concentrations. The effect of tropical convection is also visible in the case of ozone (reduction of 3%-4%) and PM 2.5 (reduction of 5%-8%).
Large concentration changes resulting from the dramatic reduction in air traffic during the pandemic (Case 2 = COVID-airc − Control) are derived by the model. Between 300 and 200 hPa (9 and 12 km), the zonal and monthly mean NO x concentration is reduced by more than 20% north of 30°N, while that of ozone is reduced by 4%-5% north of 60°N. Because of the increase with altitude of the background ozone concentration, the maximum ozone depletion in relative terms is located near 400 hPa (7 km), while in absolute terms (reduction of 7 ppbv), it is located higher in the atmosphere near 250 hPa (10 km). A secondary maximum decrease in the NO x concentration of 7% is found near 30°S. The reduction in PM 2.5 associated with reduced air traffic reaches 15% near 300 hPa and results from a reduction of similar magnitude in the concentration of sulfate and BC particles.
The zonally averaged perturbations in June, resulting from the combined changes in surface and aircraft emissions during the pandemic relative to the baseline simulation (Case 3 = COVID-ALL − Control), are shown by the left panels of Figure 2. Note that combined changes (in the left panels) are comparable to the sum of the individual changes (sum of middle and right panel). For ozone and NO x , the overall response in the free troposphere depends fairly linearly on the emission reduction scenarios. In the specific case of NO x (panel a), for the chosen scenario, the relative reduction in the concentration is higher than 10% in the boundary layer at several latitudes and in the upper troposphere north of 30°N and between 40°S and 15°S. In the case of ozone (panel d), the calculated reduction in June reaches 6%-7% north of 30°N between 800 and 300 hPa (2 and 9 km). The reduction is close to 5% in the tropics (30°S-30°N) and extends up to the tropopause. Vertical profiles of the changes in the monthly and zonally mean ozone reductions (in ppbv) relative to the baseline simulation and calculated poleward of 65°S in the tropics and poleward of 65°N are shown in Figure S3.

Effects of 2020 Meteorological Conditions and Comparison With Observations
The interannual variations in the strength of the stratospheric circulation and dynamical variability impact the tropospheric ozone burden . Specifically, deep intrusions of stratospheric ozone frequently reach the middle and even lower troposphere at midlatitudes during winter and spring, and can contribute significantly to ozone variability in the troposphere (Terao et al., 2008). This meteorologically induced variability (Case 4 = Control − CLIMO) needs to be accounted for, for example, when comparing our simulations with observed changes (Case 5 = COVID-ALL − CLIMO). Particularly in 2020, early and persistent cold conditions led to an exceptionally stable polar vortex and to record-low ozone in the Arctic, as highlighted by MLS observations , ozone sondes measurements (Wohltmann et al., 2020), and chemical reanalyzes by the CAMS (Inness et al., 2020). The minimum ozone column occurred in the first half of March, with March and April 2020 corresponding to the lowest ozone recorded for the period 1979-2020.
The anomalies in the June monthly mean NO x concentrations relative to climatology ( Figure S5), resulting from interannual variations in atmospheric circulation and temperature, lightning-related NO x formation and wildfire-related emissions (Case 4 = Control − CLIMO), reaches up to 25%. This magnitude is comparable to, or even higher than, the effect generated by the COVID-19 related emission reductions (up to −20%; see Case 3 = COVID-ALL − Control). Based on the "meteorological" model estimates (panel a), Case 4 = Control − CLIMO), NO x should have been abnormally abundant in the free troposphere during 2020, particularly in the northern hemisphere. However, the perturbations in emissions due to the pandemic substantially reduced the NO x level in northern hemisphere and tropics (panel b), Case 5 = COVID-ALL − CLIMO).
Poleward of 45°N, the anomaly in the zonally and monthly mean free tropospheric ozone concentration relative to the 19-yr climatology is influenced substantially by the pronounced springtime Arctic ozone depletion in the first months of 2020 (Case 4 = Control − CLIMO, panel c of Figure S5). This anomaly persisted between 400 and 20 hPa, poleward of 60°N, as late as June, although with a considerably lower amplitude. The ozone concentration anomaly resulting from the perturbed emissions during the COVID-19 pandemic combined with the interannual variability ranged from 5% to 15% north of 30°N (Case 5 = COV-ID-ALL − CLIMO, panel d of Figure S5). Averaged vertical profiles of the anomalies are provided in Figure S6 (polar latitudes) and Figure S7 (hemispheric and tropical averages).
It is interesting to note that meteorologically induced positive ozone anomalies everywhere south of 30°N in 2020 (panel c of Figure S5, Case 4 = Control − CLIMO) appear to have masked the COVID-19 related ozone reductions in this region (see Figures 1 and 2, Case 3 = COVID-ALL − Control). The net ozone anomaly in 2020 was therefore small south of 30°N (panel d of Figure S5, Case 5 = COVID-ALL − CLIMO). This is consistent with the lack of large negative anomalies derived from the observations in the tropics and in the Southern Hemisphere (Steinbrecht et al., 2021;Figures S8, S10, and S11).
For the northern hemisphere, the comparison between our simulations (Case 5 = COVID-ALL − CLIMO) and the observations of Steinbrecht et al. (2021) is shown in Figure 3. Ozone monthly mean anomalies in the year 2020 were observed at about 45 locations worldwide (see Figure S8 for a map of the locations), and are averaged here over all stations north of 15°N. Figure 3 also shows corresponding anomalies simulated BOUARAR ET AL.  by the CAMS. The CAMS simulations account for 2020 meteorological conditions (including the large stratospheric ozone depletion in the Arctic), but do not include effects of the reduced emissions due to the COVID-19 pandemic in 2020 (Case 4). Panel a of Figure 3 shows the resulting annual courses of ozone anomalies at 6 km altitude (∼420 hPa), averaged over all northern extratropical stations (stations north of 15°N). All data sets show increasingly negative anomalies from January to April 2020, largely due to 2020 meteorological conditions and Arctic stratospheric ozone depletion in 2020 (compare Figure S5). Observations and CAMS show similar decline from January to April; the CAM-chem simulations (Case 5 = COVID-ALL − CLIMO) give less of a decline. From April onwards, photochemical ozone production becomes increasingly important, and the reduced emissions of 2020 play a major role (compare Figures 1  and S7). Consequently, observations and CAM-chem simulations show persistent negative anomalies (Case 5) of −5% to −10%. Panels b-d in Figure 3 shows vertical profiles of the ozone anomaly, averaged over the 4 months from April to August 2020, and for the single months April and June.
To assess uncertainties for the influence of the large Arctic spring-time stratospheric ozone depletion, a sensitivity test was performed with the CAM-chem model (Case 6): According to the suggestion of Wilka et al. (2021), larger number densities of nitric acid trihydrate particles (10 −5 cm −3 ) were assumed in the Arctic lower stratosphere (Case 6, scenario CAM-chemW). This increases the denitrification rate in these layers and hence increases the catalytic destruction of ozone by active chlorine. When they adopted these conditions, Wilka et al. (2021) found better agreement between the calculated and observed nitric acid and ozone concentrations. This scenario provides an upper bound for the stratospheric ozone reduction (Case 6, CAM-Chem-W, dark blue line) and gives about 1% more ozone reduction in the troposphere from May to August. It is generally in better agreement with the observations (red lines and shaded area, see also panels b-d)). In contrast to observations and both CAM-chem simulations (Cases 5 and 6 = COVID-ALL(2) − CLIMO), CAMS (no emission reductions, Case 4, gray lines and shading) simulates increasing ozone from May onwards. By July, CAMS simulates anomalies near or above zero. The good agreement between observations and CAM-chem simulations (cases 5 and 6 = COVID-ALL(2) − CLIMO) from April to August, and their difference with respect to CAMS (Case 4, no emission reductions), further confirms that the negative ozone anomaly of −5% to −10% in late spring and summer 2020 was caused largely by reduced emissions, with some influence from the 2020 Arctic spring-time depletion of stratospheric ozone.

Uncertainties
As stated above, the model simulations have been performed by adopting the central values (best estimates) of the CONFORM adjustment factors . These factors are subject to uncertainties ranging typically from 10% to 25% depending on the economic sector and the region of the world (see Table S1). These uncertainties translate into errors of the same order of magnitude on the calculated changes in the concentration of primary species. As the photochemistry of secondary species such as ozone is nonlinear, the response of these species to reduced emissions is expected to vary with the local photochemical environment, and should be quantified by an ensemble of model simulations that considers the uncertainties in the emissions for each economic sector. For example, when considering the uncertainty associated only with the reduction in aviation activity, the error in the calculated ozone reduction near 250 hPa is close to 1 ppbv. A complete estimate of the error in the calculated response of atmospheric constituents should also account for the differences in model formulations, which can also lead to substantial differences in the calculated responses.

Summary
The ozone abundance in the extratropical northern hemisphere free troposphere during spring and summer of 2020 has been 5%-15% lower than climatology. Assuming realistic scenarios, the ozone response to decreased emissions of primary pollutants associated with the reduction in economic activity including air-traffic during the COVID-19 pandemic is estimated to be 4%-8%. Reduced worldwide aircraft operations had the highest impact in the middle and upper troposphere of the northern hemisphere during the summer months. The impact of 2020 meteorological conditions and the abnormally high ozone depletion in the Arctic lower stratosphere during the spring and summer of 2020 also produces a noticeable ozone reduction of 3%-10% in the northern extratropical free troposphere. This effect is noticeable until late spring and reaches a maximum in June. Below 400 hPa, however, the influence of the stratosphere remains small, compared to the effect of the COVID-19-related reduction. For regions south of 30°N, the tropics and the southern hemisphere, the simulations indicate that a 4%-6% reduction of ozone due to COVID-19 related emission reductions did take place in 2020, but was largely compensated by ozone (and nitrogen oxides) increases caused by the specific meteorological conditions of 2020.
Our study has estimated the response of free tropospheric ozone to an unprecedented real reduction in global anthropogenic emissions. The model simulations successfully reproduce the observed ozone anomalies in the free troposphere during the 6 months that followed the COVID-19 outbreak. They provide a quantitative estimate of the different factors that contributed to the observed ozone anomalies. We also tested the sensitivity of our results to enhanced spring-time ozone destruction in the Arctic stratosphere, and to a more moderate reduction of aircraft emissions during the COVID-19 crisis. While these changes have some effect, they do not fundamentally alter our conclusions. Overall, our different tested scenarios, together with similar ozone reductions obtained in other simulations (Mertens et al., 2021;Miyazaki et al., 2021;Weber et al., 2020) indicate that the uncertainty of the calculated ozone anomaly in the free troposphere is of the order of 1-2 ppbv (2%-4%), and is associated with the adopted model formulation as well as the assumed magnitude of the emission reductions. As more accurate information on actual emission reductions is becomes available, future simulations should reduce the current uncertainties. Clearly, global and regional air quality forecast and reanalysis models are now starting to account better for the disturbances, which occurred in the atmospheric chemical system due to reduced emissions under the COVID-19 pandemic after January 2020.