The Effects of the COVID-19 Lockdowns on the Composition of the Troposphere as Seen by IAGOS

The European Research Infrastructure IAGOS (In-service Aircraft for a Global Observing System) equips commercial aircraft with a system for measuring atmospheric composition. A range of essential climate variables and air quality parameters are measured throughout the flight, from take-off to landing, giving high resolution information in the vertical in the vicinity of international airports, and in the upper-troposphere/lower-stratosphere during the cruise phase of the flight. Six airlines are currently involved in the programme, achieving a quasi-global coverage under normal circumstances. During 5 the COVID-19 crisis, many airlines were forced to ground their fleets due to a fall in passenger numbers and imposed travel restrictions. Deutsche Lufthansa, a partner in IAGOS since 1994 was able to operate a IAGOS-equipped aircraft during the COVID-19 lockdown, providing regular measurements of ozone and carbon monoxide at Frankfurt airport. The data form a snapshot of an unprecedented time in the 27 year time-series. In May 2020, we see a 39% increase in ozone near the surface with respect to the 26 year climatology, a magnitude similar to that of the 2003 heatwave. The anomaly in May is driven by an 10 increase in ozone at nighttime which might be linked to the reduction of NO during the COVID-19 lockdowns. The anomaly diminishes with altitude becoming a slightly negative anomaly in the free troposphere. The ozone precursor carbon monoxide shows an 11% reduction in MAM near the surface. There is only a small reduction of CO in the free troposphere due to the impact of long-range transport on the CO from emissions in regions outside Europe. This is confirmed by IASI-SOFRID CO retrievals which display a clear drop of CO at 800 hPa over Europe in March but otherwise show little change to the abundance 15 of CO in the free troposphere. 1 https://doi.org/10.5194/acp-2021-479 Preprint. Discussion started: 22 June 2021 c © Author(s) 2021. CC BY 4.0 License.


Introduction
The World Health Organization declared the global COVID-19 pandemic in March 2020 (WHO, 2020). The serious threat to public health led countries to adopt lockdowns and other coordinated restrictive measures aimed at slowing the spread of the 20 virus. Such measures had an important effect on economic activity and by consequence on the emissions of primary pollutants from industrial and transport sectors. Much discussed, is the extent to which these lockdowns have had a significant effect on local air quality and more widely on atmospheric composition and climate (Le Quéré et al., 2020).
Many studies have focused on primary pollutants such as NO 2 , decreases of which were almost immediately apparent in satellite imagery from the TROPOspheric Monitoring Instrument (TROPOMI) on the Sentinel-5 Precursor satellite (Veefkind 25 et al., 2012) over China in January/February, and later over Europe (Bauwens et al., 2020). Emissions of NO 2 are strongly linked to economic activity. Instruments such as TROPOMI have registered weekly cycles of NO 2 and drops in NO 2 related to behavioural patterns of work and holiday periods (Beirle et al., 2003;Tan et al., 2009). Thus, the large reductions in NO 2 consequent impact on the number of IAGOS aircraft flying, and the amount of data collected. However, one of the Lufthansa aircraft was converted to carry cargo, and operated throughout the lockdown period. The aircraft made regular flights from Frankfurt to Asia carrying important medical supplies. Frankfurt airport has the longest, densest and most homogeneous time-90 series of all the airports visited by IAGOS. Thus, the climatology calculated there is the most robust (Petetin et al., 2016b) with ozone being measured since 1994 and CO since the end of 2001.
In this article, we present the observed anomalies of both ozone and CO seen over Frankfurt and benefit from the fine 30m vertical resolution throughout the troposphere, to distinguish the surface anomalies from the observations in the free troposphere. This offers a valuable check on satellite data, and adds unique and valuable vertical information which is not 95 offered by surface sites. We judge the significance of the ozone anomalies against the 26 year climatology  at Frankfurt, putting the observed anomalies in context with other important events such as the heatwave in 2003. To complement the IAGOS data at Frankfurt we use IASI-SOFRID CO retrieval which give an idea of the extent of any regional changes over Europe.

100
The research infrastructure IAGOS, is described in detail in (Petzold et al., 2015). Using commercial aircraft as a platform, IAGOS instruments make routine measurements of ozone, and carbon monoxide along with water vapour, cloud particles and meteorological parameters including temperature and winds. A full description of the instruments that measure ozone and CO used here, can be found in (Nédélec et al., 2015). The ozone instrument, a dual-beam ultraviolet absorption monitor has a response time of 4 s, and an accuracy estimated at about 2 ppbv (Thouret et al., 1998). This 4 second response time corresponds 105 to a vertical distance of about 30 m. In the horizontal, the aircraft covers a distance of about 80km during the first 5km of ascent (Petetin et al., 2018a). Therefore during the ascent and descent phases of the flight, IAGOS provides fine-scale quasi-vertical profiles. Carbon monoxide is measured with an infrared analyser with a time resolution of 30 s (7.5 km at cruise speed of 900 km h-1) and a precision estimated at 5 ppbv (Nédélec et al., 2003) IAGOS began life in 1994 under the name MOZAIC (Measurement of Ozone and Water Vapour on Airbus in-service 110 Aircraft) (Marenco et al., 1998) and as such IAGOS has provided a long time-series of ozone data over 27 (1994-present) years, and of CO for almost 20 years (2001-present). The homogeneity of the time-series since 1994 has been demonstrated by (Blot et al., 2021), giving confidence that IAGOS data can be used for a robust climatology and for the study of long-term trends. As mentioned above, this gives IAGOS some important advantages over more short-lived data-sets such as those from satellites, and allows us to put any anomalies into context within the same reference observations.

115
For the IAGOS measurements, a number of auxiliary, diagnostic fields are delivered with the data as standard level 4 products. These include potential vorticity, geopotential height and boundary layer height which we will use in this article. The boundary layer height which is defined as the boundary layer thickness (zPBL) + orography is calculated by interpolating the European Centre for Medium-Range Weather Forecast's (ECMWF) operational boundary layer heights to the position and time of the IAGOS aircraft. The ECMWF fields were 1 • horizontal resolution and 3 hour time resolution with 60, 90 or 137 levels in the vertical depending on the time period used (http://www.iagos-data.fr/#L4Place:).
In order to determine the geographic origin and source of the CO measured by IAGOS, a tool known as SOFT-IO (Sauvage et al., 2017a, b) has been developed, that uses FLEXPART (Stohl et al., 2005;Forster et al., 2007) to link the IAGOS measurements with emissions databases via 20-day back trajectories. For the entire IAGOS flight track, SOFT-IO v1.0 (Sauvage et al., 2017a(Sauvage et al., , 2018 estimates the source region of the CO contribution from 14 different world regions of emissions from the GFAS 125 v1.2. The source regions are as defined by the Global Fire Emissions Database (GFED), although the emissions inventories are GFAS. It can also estimate the contributions from anthropogenic or wildfires. As for the auxiliary diagnostic fields mentioned above, the meteorological data for FLEXPART come from the 1°by 1°ECMWF operational analyses and forecasts with a 6 hour and 3 hour time resolution respectively (Sauvage et al., 2017b).
To set the IAGOS measurements at Frankfurt airport into a regional context, we use CO satellite retrievals from the Infrared 130 Atmospheric Sounding Interferometer (IASI) on the MetOp meteorological platforms (Clerbaux et al., 2009). These retrievals are performed with the SOftware for a Fast Retrieval of IASI Data (SOFRID) described in Barret et al. (2011);De Wachter et al. (2012). This software is based on the RTTOV (Radiative Transfer for TIROS Operational Vertical Sounder) operational radiative transfer code (Saunders et al., 1999;Matricardi et al., 2004) combined with the 1D-Var software (Pavelin et al., 2008).
For CO the SOFRID retrievals provide a maximum of two pieces of information about the vertical profiles from the surface to

Anomalies of ozone in early 2020
In this first section, we look at the anomalies of ozone which were strongly evident in spring 2020. Figure 1, shows the seasonally averaged profile of ozone measured at Frankfurt for March-April-May 2020. The data were acquired by one of the 140 IAGOS-equipped Lufthansa passenger aircraft which was based at Frankfurt. It was converted to cargo operations and was kept flying throughout the lockdown period making a total of 84 flights from March-May 2020.
In  The period MAM 2020 corresponded to the period with the most stringent COVID-19 lockdowns across western Europe, but each country had its own date of onset, duration, and different levels of severity. Measures of European mobility (Grange et al. (2021), based on Google mobility data) reveal that the depths of lockdown were in early April, showing a very slight recovery 155 throughout May. At Frankfurt airport, there was 50% less traffic in March 2020 compared with March 2019, with nearly 80% less traffic in April and May (source, FRAPORT https://www.fraport.com/en/investors/traffic-figures.html, last accessed 18 December 2020). According to Grange et al. (2021) , the restriction measures began in Germany on 22nd March 2020 and had a "Stringency index" defined as "a measure of the strictness of 'lockdown style' policies", that remained relatively high until the end of May, but that was by no means the strictest in Europe.

160
In Fig. 2, we present the anomalies in ozone for March and May 2020. We did not have any ozone data in April 2020. It was during the month of May, after the lockdown had been in place for several weeks, when the ozone anomaly in the surface layer (pressure P > 950hPa) was most pronounced (Fig. 2). We require there to be 7 days to make the monthly average otherwise the month is excluded. Excluded months are marked with a cross. In May, ozone was recorded at 13.9ppbv (39%) higher than the reference average  and was the largest anomaly for the month of May since the time-series began in 1994.

165
The anomaly is apparent in the first 800m of the atmosphere (Fig. 1). A positive anomaly was also observed in March 2020 (15%) with a smaller value compared with May. Positive anomalies in ozone have not been unusual in recent years (see Fig. 3) suggesting that the lockdowns are not the only explanation.
To set the magnitude of these anomalies into context with other periods, the time-series for each month for the surface layer over Frankfurt, is shown in the top panel of Fig. 3. In addition to the monthly averages, we also show the number of profiles 170 per month as the solid grey bars. During the lockdown period, there were fewer flights than normal and we need to be aware of any sampling bias that this may introduce. The bottom panel of the plot shows a time-series of the monthly anomalies in the surface layer from 1994-2020. There were a number of occasions when the ozone anomalies were comparable to that of An increase of ozone near the surface can result from increased production of ozone, or reduced sinks of ozone, depending on the conditions and the time of day. Positive anomalies of ozone may be due to an increase in the precursors of ozone, or a prevalence of certain meteorological conditions including increased UV radiation, stagnant airmasses or lower boundary layer heights which trap the pollutants near the surface. Otherwise, there can be a decrease in the sinks of ozone, such as a decrease in 180 the rate of dry deposition or a decrease in titration by NO due to a reduction in the reservoir of NO. During the 2003 heatwave, IAGOS data showed that there were positive anomalies at Frankfurt in both ozone and the precursor carbon monoxide in the low troposphere, with the ozone anomalies up to 2.5km deep and with the magnitude of the anomalies increasing towards the surface (Tressol et al., 2008). Tressol et al. (2008) found that near the surface, ozone was almost twice the normal amount, and CO was more than 20% higher. The increased CO was due to the transport of plumes from wildfires over Portugal exacerbated The monthly anomalies calculated with respect to the reference average 1994-2019 in the surface layer.
by the dry conditions created by the heatwave. Thus, during the 2003 heatwave, the increased ozone was caused by an increase in precursors and the favorable meteorological conditions. During lockdown, the chemical environment was quite different. The positive ozone anomaly was accompanied by a drop in the amount of NO as evidenced by the TROPOMI satellite (Bauwens et al., 2020), and there is some evidence from IAGOS measurements that levels of the precursor carbon monoxide also fell (see section 2.2). The ozone anomaly in the 190 surface layer was most likely due to the combination of increased ozone production due to the exceptionally sunny con- changes in the emissions of precursors (Ordóñez et al., 2020;Petetin et al., 2020;Lee et al., 2020). All found that there were important and differing impacts of meteorology, but that the photochemical effects from NOx were dominant.
The magnitude of any anomaly may be significantly influenced by the sampling times within the diurnal cycle. Petetin et al. (2016a), described the typical diurnal cycle of ozone at Frankfurt airport observed with IAGOS data at different altitudes.
They noted that the mixing ratios of ozone are minimum at nighttime due to dry deposition and titration by NO in the shallow nocturnal boundary layer and reach a maximum in the afternoon, due to photochemistry and mixing with ozone-rich layers above the boundary layer. Petetin et al. (2016a) showed the diurnal cycle of ozone at Frankfurt to be maximum between 12:00 and 18:00 UTC in MAM in the layers below 900hPa. The amplitude is maximum at the surface and decreases with altitude, becoming almost insignificant at altitudes above 900hPa. We consider the anomaly observed in May with respect to the diurnal 205 cycle. More measurements in the afternoon would lead to an oversampling of the maximum and a positive ozone anomaly, and conversely, more measurements at nighttime would be an oversampling of the minimum and a negative anomaly. In Fig.   4, we can see the hourly distribution of the IAGOS profiles for the month of May in 2020 compared with the same month for the reference period 1994-2019. In the climatology, there is a bias towards early morning measurements and in 2020, a bias towards measurements in early afternoon. This refects the different flight operations carried out during the COVID-19 period.

210
To account for this bias, we calculate the anomaly for May for the daytime (10:00-18:59 UTC) and the nighttime (00:00-09:00/19:00-23:59 UTC) applying this to both the climatology and 2020. These are shown in Fig. 5 (a) and (b) respectively. In The positive ozone anomaly observed in IAGOS data for the surface layer is in agreement with the other studies cited based on the surface networks and as reviewed by (Gkatzelis et al., 2021). The IAGOS data for the remainder of 2020 (Fig. 3), show smaller positive anomalies which were not significant within the time-series, suggesting that the anomaly in MAM, was short-lived. We now explore the vertical extent of the ozone anomaly, using the unique perspective that IAGOS offers.

Anomalies of ozone in the free troposphere (850-350hPa)
In contrast to the positive anomaly in the surface layer up to 800m, the anomaly in the free troposphere above 2000m is negative ( Fig. 1) lying just on the edge of the range of interannual variability based on the 26 year time-series shown in Fig. 6. The grey bars in the top panel of Fig. 6 represent the number of available daily profiles in each month (where there were more than 7 days available in the month). There was a -6.7 ppbv or -12% drop in ozone in March (Fig. 7). This negative anomaly 225 is the largest for March since 1997 for the IAGOS observations in the free troposphere. It is too early to have resulted from the European lockdowns, and not easy to link with the Asian lockdowns. There was only a 2.0 ppbv (3%) reduction in ozone in the free troposphere over Frankfurt in May 2020 which might be linked to regional European lockdowns (Fig. 7). We had no data for April 2020. The IAGOS data show that ozone levels remained lower than usual for several months after the main lockdown period ended, with an -11% anomaly being observed in July (the largest anomaly recorded for July in the 27 year

Carbon Monoxide in Spring 2020
As mentioned in the introduction, some studies have demonstrated a fall in the ozone precursors during MAM 2020. The reductions in CO in near-surface air masses due to COVID-19 reported by (Gkatzelis et al., 2021) ranged from 20% to 50% for CO. Due to the long (weeks to months) lifetime of CO in the atmosphere, the causes of these decreases in CO are difficult 240 to attribute.

Anomalies of carbon monoxide in the surface layer (>950hPa)
In the surface layer, we see a downward trend in monthly values of CO on the seasonal and annual scale (see Fig. 9) in 260 agreement with Petetin et al., (2016b). Also shown in Fig. 9, is the number of flights per month as the solid grey bars which reveals a reduction in flights due to the reduction in global travel during the first phase of the pandemic. As for ozone, only    The observations of CO near the surface from IAGOS are less impacted by the local emissions at airports that might be thought. Petetin et al (2018a), compared IAGOS with monitoring stations from the local Air Quality monitoring network (AQN) and more distant regional surface stations from the Global Atmospheric Watch (GAW) network. They found that the mixing ratios of CO and O 3 close to the surface do not appear to be strongly impacted by local emissions related to airport activities and are not significantly different from those mixing ratios measured at surrounding urban background stations. It 295 is therefore unlikely that the reduction in airport activity during COVID-19 was a big contributor to the negative anomaly observed at Frankfurt in the surface layer. In the free troposphere, the local emissions have even less effect and the mixing ratios tend to background concentrations as typically measured by the GAW regional stations.
To examine more closely the source regions of the CO at Frankfurt, we have used the SOFT-IO tool which routinely connects emissions databases to each IAGOS measurement via FLEXPART trajectory calculations (Sauvage et al., 2017b(Sauvage et al., , 2018. How-300 ever, due to the anthropogenic emissions database (MACC-city) not being updated to take account of the COVID-19 period, we will ignore the indicated contribution from fire or anthropogenic sources given by the link to the emissions databases. We present just the geographic origin of the CO at Frankfurt as determined by the trajectory calculations. The source regions are  defined as in Fig. 12. We compare the source regions in 2020 with those in our reference period 2016-2019. In agreement with Petetin et al. (2018b), our analysis shows that the largest contribution to CO measured at Frankfurt is from the European region.

305
Usually in this period the biomass burning emissions are low, so we deduce that most of the emissions were anthropogenic.
In Fig. 13 arriving in Europe. This analysis shows that it is primarily local emissions across Europe that are reflected in the CO recorded at Frankfurt in the surface layer, and therefore we can suppose that the lockdown measures played a significant role. In the free troposphere, which we discuss in the following section, we shall see that inter-continental transport has a more important contribution.

Anomalies of Carbon Monoxide in the free troposphere (850-350hPa)
In the free troposphere, the anomalies of CO (Fig. 14) were negative in March (-6.9ppbv, -5%) and May (-1.8 ppbv, 1%) but 315 much smaller in magnitude than in the surface layer (see section 2.2.1). The time-series of CO in the free troposphere from 2001-2020 is included for reference in Fig. B2. In April, the anomaly was positive (8.2ppbv, 6%). Since the free troposphere is more representative of the background concentrations due to mixing and transport, it is instructive to relate the IAGOS data over Frankfurt to the larger geographical context. Satellite fields of CO in the tropospheric column are presented for Europe in Fig. 15. Figure 15 represents the percentage change in the tropospheric CO at 795 hPa with respect to the 2016-2019 average as 320 retrieved from IASI for the months of March, April, and May. In March, the IASI-SOFRID data confirm the negative anomaly in CO present at Frankfurt, which is generalised over large parts of Europe. In April and May, the IASI-SOFRID data showed little anomaly at Frankfurt and a mixed picture over Europe. Thus, the IAGOS and IASI data show some reduction in CO during the lockdown period which is not unexpected given the trend towards decreased CO. It is difficult to link this anomaly to the lockdown measures due to other factors such as the increased boundary layer height, the long-range global transport of 325 CO, and interannual variability.
Using the same trajectory analysis as for Fig. 13, Fig. 16 shows that in MAM 2020, there was a much lower amount of CO from sources in Europe than in MAM 2016-2019, and an increase in the contribution from North America (TENA) and from Central Asia (CEAS). We suggest that the increase in airmasses from outside Europe negated the effects of the cut in emissions during the European lockdown resulting in a smaller than anticipated negative anomaly observed by both IAGOS 330 and IASI-SOFRID. This result is similar to that of Field et al. (2020), who noted only a 2% change in background abundances of CO over Eastern China despite the cut in industrial emissions during the Chinese lockdown. In the case of Field et al. (2020) it was the cross-boundary transport from areas with active biomass burning that off-set the drop in anthropogenic emissions.
In summary for this section on CO, we conclude that the drop in surface CO is largely the result of the drop in emissions during European lockdown, with higher than usual boundary layer heights further diluting the surface concentrations. In the 335 free troposphere, where the negative anomalies were small, the influence of long range transport is more apparent, and offsets the impact of the reduction in CO emissions across Europe.
In this article, we use the IAGOS dataset of in situ observations of ozone and carbon monoxide collected during landing and take-off at Frankfurt airport. The atmosphere is sampled from the surface to the upper troposphere forming a quasi vertical 340 profile. The data form part of a time-series which extends back for 27 years for ozone and 20 years for CO. During the springtime (MAM) of 2020, we noted a 27% increase in ozone in the surface layer (> 950hPa, 600m). The month of May saw the largest anomaly since the time-series began 1994. The magnitude of this anomaly is comparable to the European heatwave in August 2003. As this anomaly occurred during the first lockdown period of the COVID-19 outbreak, we thus considered if the anomaly was related to the changes in emissions that resulted from the decreased traffic and industrial activity. The daytime 345 increase was significant (19%) but the increase at nighttime (41%) was double and has not been seen before in the 26 year climatology. Despite the fall in the abundance of NOx over Europe (as seen by satellite data), there were still enough available precursors to produce ozone under the meteorological conditions that were very favorable, especially enhanced solar radiation, that prevailed at the time (van Heerwaarden et al., 2021). Our larger increase at nighttime suggests that less ozone was lost through titration with NO due to the reduction in the NO reservoir during the lockdown period, signifying a reduction in the 350 "sink" of ozone.
In the free troposphere, ozone abundances fell slightly. For the period MAM 2020 ozone was 5% lower than the mean 1994-2019 based on the same months. The IAGOS time-series shows that these free tropospheric abundances of ozone were the lowest since 1997 which is probably more reflective of the widespread reduction of emissions over Europe and beyond, with less impact of local meteorology and chemistry. The results from IAGOS are also consistent with those from the sonde 355 and balloons reported by Steinbrecht et al. (2021) who noted a 7% drop in tropospheric ozone compared with the 2000-2020 climatological mean. They also attributed this to the reduction in pollution during the COVID-19 lockdowns.
A reduction of CO was seen at Frankfurt, with an 11% reduction found in the surface layer but no anomaly in the free troposphere as averaged over MAM. Our trajectory analysis shows that the CO is largely of European origin and therefore we suggest that the anomalies in the surface layer are directly linked to the drop in emissions across Europe due to the lockdown 360 measures. The increase in boundary layer height explains the onset of the anomaly before the onset of lockdown and probably contributed to a further dilution of the CO at the surface.
In the free troposphere there is a small reduction of CO in March (-6.9 ppbv, -5%) and May (2 ppbv, 1%) which is smaller than at the surface over Frankfurt. IASI-SOFRID fields of CO show a clear decrease of CO during March over all of Europe. In April and May small positive and negative anomalies are detected by IASI-SOFRID over northern Europe and mostly negative 365 anomalies over the Iberian Peninsula. In the free troposphere, there is an important role of transport of CO from distant biomass and anthropogenic sources. In particular in MAM 2020, there was a greater contribution from CO originating in North America and Asia which off-set some of the reduction in regional European emissions, with the result that the lockdown measures did not have a big impact on CO in the free tropopshere.
The lockdowns provided a unique experiment to assess the impact of a reduction of economic activities on atmospheric 370 composition and climate. The IAGOS data complement other in-situ data from the ozonesonde network, with the added value of having ozone precursors measured simultaneously. This study highlights the importance of long and continuous time-series in setting this brief period in context since there are many competing factors and it is difficult to attribute a single cause. We look forward to future model sensitivity studies to separate these factors and to provide a more realistic magnitude of the impact of lockdown on the environment.