3.2.1 A_PublicPower and B_Industry

Eurostat data (https://ec.europa.eu/eurostat/databrowser/view/sts_inpr_m/default/bar?lang=en, last access: 16 September 2021) were used to account for changes in the sectors A_PublicPower and B_Industry. The energy data provided there comprise monthly information on the volume index of production for electricity, gas, steam and air conditioning supply. They are available for 35 countries in Europe. The industry data comprise monthly information on the volume index of production for mining and quarrying; manufacturing; electricity, gas, steam and air conditioning supply; and construction and are available for 20 countries in Europe. The indices are based on an index value of 2015. However, since we want to use them to evaluate the lockdown period, we normalized the changes based on the January 2020 value. The data are given in a monthly resolution; however, for many countries in Europe the lockdown started in the middle of March. Therefore, a piecewise cubic spline interpolation procedure was applied to derive daily lockdown adjustment factors while still maintaining the monthly values. Examples are given for both sectors in Germany in Fig. 2.

Figure 2Examples for monthly values and interpolated functions for lockdown adjustment factors (in %) for the sectors A_PublicPower and B_Industry in Germany.

Download

3.2.2 F_RoadTransport

Google mobility reports (https://www.google.com/covid19/mobility/, last access: 16 September 2021) deliver daily percentage change of visits in different areas (e.g. residential, transit, recreation, work places). The reference value is the median of the corresponding weekday between 3 January and 6 February 2020. We use Google mobility reports for transit on a national level to account for the changes in road traffic emissions. Through this method, reduced traffic on national holidays, e.g. around Easter and 1 May, is considered as well; however, vehicle types cannot be distinguished.

3.2.3 G_Shipping

To derive scaling factors that account for ship traffic and emission reductions in this sector, bottom-up ship emission inventories were created with the MOSES ship emission model (Schwarzkopf et al., 2021) using automatic identification system (AIS) data for 2019 and 2020 covering the German Bight and the western Baltic Sea. The data were recorded in Bremerhaven and Kiel by the German Federal Maritime and Hydrographic Agency (BSH). A 7 d rolling mean filter was applied to the calculated CO₂ emission ratios (Fig. 3). On average, the data revealed a slight reduction of ship traffic in the North Sea area by approx. 10 %. For the Baltic Sea traffic reductions were clearly visible with a downward trend from March until the middle of June that could be mainly attributed to roll-on/roll-off (RoRo) ships and passenger ships. For the first 75 d of the year until 15 March 2020 no reductions were applied; afterwards daily LAFs were used similar to the approach for road traffic. LAFs for the North Sea were also applied for the Mediterranean Sea; those for the Baltic Sea were also applied to inland shipping. The reasoning behind this is that shipping in the Mediterranean is mostly international cargo transport, similar to the North Sea, and inland navigation is connected to short-range transport, similar to the Baltic Sea. As can be seen in Fig. 3, relative increases in shipping emissions might also occur during limited time.

Figure 3Lockdown adjustment factors created from the 7 d rolling mean ratios of CO₂ emissions from shipping in 2020 relative to 2019. Until day 75 (15 March) no changes and a LAF of 1 was assumed.

Download

3.2.4 H_Aviation

Airport traffic total arrivals and departures data from Eurocontrol (https://ansperformance.eu/data, last access: 16 September 2021) were used to account for emission changes in the aviation sector. We applied a reduction based on a weekday mean from 3 January until 6 February 2020, similar to Google mobility data. Daily values for 42 European countries are available. The relative reductions in this sector were most pronounced, reaching −90 % in March and April and a slower recovery than the other sectors.

3.2.5 Sector comparison

LAFs for Germany, France, the UK and Sweden are exemplarily shown in Fig. 4. Huge emission reductions in road traffic and air traffic between 10 and 20 March (day of the year (DOY) 70–80) can clearly be seen. Public power and industry, on the other hand, show much smaller reductions (10 %–30 %) and almost reach previous-year levels until the end of June. At the same time in France and Germany, road traffic was back to 90 % of the previous year; however, in the UK and in Sweden 20 %–40 % reductions were still visible in the activity data. Comparisons of country-specific LAFs for the sectors F_RoadTransport and H_Aviation are given in Appendix A (Figs. A1 and A2).

Figure 4LAFs for Germany (a), France (b), the United Kingdom (c) and Sweden (d) for the sectors A_PublicPower, B_Industry, F_ RoadTransport and H_Aviation.

Download

Figure 5Daily average values for sector separated NO_x emissions summarized over the entire central European model domain for the noCOV and the COV case (with LAF).

Download

Figure 5 presents total daily NO_x emissions in the entire central European domain (see Fig. 1) for the time period from 1 January to 30 June 2020 for the COV and the noCOV case separated by GNFR sectors. Road transport is the most important emission sector with approx. 20 % to 30 %, followed by ocean shipping, other stationary combustion, industry and public power, which all have similar contributions of approx. 10 %. Combustion shows a clear decline towards the summer months due to the fact that domestic heating is mainly necessary in winter.

Reductions caused by the lockdown stem mostly from the road transport sector, with a strong drop in emissions starting around DOY 75 (15 March). The aviation sector, which experienced the strongest relative drop in emissions during the lockdown, does not play a major role for the overall emission of NO_x. However, it might be important near airports and in the upper troposphere. Overall, NO_x emissions in central Europe dropped by around 25 000 mol s⁻¹ (approx. 4 kt h⁻¹, when given as NO₂) during the strictest lockdown period in late March and early April. This corresponds to a relative drop of around 30 % (Fig. 5).

4.1.1 Pre-lockdown period

In February 2020, an unusually wet period occurred due to strong cyclonic activity in central Europe. Westerly and north-westerly cyclonic regimes were observed on 76 % of the days, whereas high-pressure-type regimes were observed on only 24 % of the days. Thus, the shortwave downwelling irradiance in February 2020 is one of the lowest measured at the weather station Wettermast Hamburg (53^∘31^′09^′′ N and 10^∘06^′10^′′ E) (https://wettermast.uni-hamburg.de, last access: 5 July 2021) (Brümmer and Schultze, 2015) during the last 25 years (Fig. A4), being representative for north-western Europe. The accumulated precipitation for February at this weather station with an amount of more than 120 mm was exceptionally high compared to the last decades (Fig. A4).

4.1.2 Main lockdown period

For the meteorological characterization of the main lockdown period between the middle of March and the end of April we rely in addition to the GWL analysis on maps of the 500 hPa geopotential height and the surface pressure distribution. The underlying data were extracted from simulations with the COSMO–MERRA system, the same meteorological fields which have been used for the chemistry transport calculations with CMAQ displayed and discussed in the following chapters. In Fig. 6 a subset of those maps for three selected time periods is shown; the complete set of maps generated can be found in Appendix A (Fig. A5). To characterize and quantify horizontal advection, wind roses derived from observations at the Wettermast Hamburg are displayed in Fig. 7. The wind data in each plot cover a time period of about 15 d. Measurements at an altitude of 110 m were chosen to better represent a larger area and eliminate parts of the surface influences on the wind.

Figure 6500 hPa geopotential heights (in geopotential decametres, gpdm) and surface pressure (in hPa) for selected time segments in March and April 2020 according to the COSMO simulations. The geopotential heights are averaged over 4 d (21–24 March, 6–9 April and 21–24 April from left to right, respectively). Displayed surface pressure distributions are representative snapshots within those time segments.

Figure 7Wind roses derived from measurements of the weather station Wettermast Hamburg at an altitude of 110 m. Results for three periods covering about 15 d each are shown: 16–31 March 2020, 1–15 April 2020 and 16–30 April 2020, from left to right.

Download

In the middle of March, the synoptic regime substantially changed over Europe. High-pressure-type GWLs became dominant; i.e. high ridges over central Europe and high-pressure systems led to a typical atmospheric blocking of cyclones. The weather situation shows first a varying blocking in northern and central Europe followed by a high-pressure ridge reaching from the Azores to Scandinavia (Fig. 6, left), which changed to a high-pressure ridge stretching from Iceland into Russia. In northern Germany the wind regime was dominated by a flow with mainly easterly components, which were relatively high wind speeds (Fig. 7, left). In southern Europe the situation, which was similar at the beginning of the period to that one in the north, with changes starting on about 23 March, an isolated trough formed, leading to low-pressure system activity. For 28 and 29 March dust transport from Asia and Northern Africa to the Po Valley was reported (Collivignarelli et al., 2020).

In the first half of April the weather in the north-eastern part of central Europe was again quite variable, and in southern Europe the cut-off from the northern regime could still be recognized. In the western part of central Europe a ridge has established, which stretched towards the UK. Accordingly, winds in northern Germany blew predominantly from westerly/north westerly directions. Later on, a ridge over all of central Europe dominated the weather in the study domain (Fig. 6, middle), only the eastern Mediterranean was still influenced by a cut-off trough. In the Po Valley, according to measurements around Milan, the weather during the second half of March to 10 April was dry and very sunny with low to medium wind speeds (Collivignarelli et al., 2020). Towards the middle of April a high-pressure bridge was established reaching from Iceland into eastern Europe.

In the second half of April a high-pressure system established over the British Isles attached to a ridge located over central Europe leading to dry and sunny weather all over Europe. This condition was basically stable until 25 April, when a cyclonic flow took over, leading to more westerly winds over central Europe, a situation which lasted until the first days of May. Winds in northern Germany switched over from easterly to more westerly directions this time (Fig. 7, right).

Overall, an exceptionally dry period occurred which started in the early lockdown period and continued until the end of April. The weather was characterized by very low cloud cover and record-breaking large amounts of solar irradiance (see the record at the Wettermast Hamburg in Fig. A4) and little precipitation. This exceptional weather period is also discussed by van Heerwaarden et al. (2021), who reported record breaking solar irradiation for the Netherlands.

4.1.3 Lockdown transition

In May 2020, atmospheric conditions were very different in central Europe compared to the previous months. For instance, Germany was dominated by large amounts of rain in the south, sunny conditions in the west, and dry but cloudy conditions in the east and north. Observed sunshine duration and solar irradiance corresponds approximately to average climatic conditions. In contrast, large parts of western Europe (the Netherlands, Belgium, West Germany, the UK) experienced sunny and dry weather throughout the entire month of May (van Heerwaarden et al., 2021). Finally, the large-scale conditions in June turned out to favour long-lasting periods with dry and sunny weather conditions in northern Germany due to blocking conditions caused by high-pressure systems located over Scandinavia. However, the more southerly regions were rather too wet in a climatological sense.

5.1.1 NO₂ concentrations

Figure 10 shows maps of the modelled average NO₂ concentrations in central Europe between 16 and 31 March for the case without lockdown measures (noCOV) together with the absolute and relative concentration reductions caused by the lockdown. The NO₂ concentrations for the noCOV case in central Europe show the typical pattern with highest concentrations in densely populated areas like England, Belgium, the Netherlands, and western Germany as well as northern Italy (Fig. 10a). Average concentrations range between 5 and 10 µg m⁻³. Reductions in NO₂ concentrations caused by the lockdown are highest in the same regions, also reaching several µg m⁻³. Relative reductions are highest in France, Belgium, Italy and Austria, reaching more than 40 % on average. Germany, the Netherlands, the UK, southern Sweden and the Czech Republic show lower reductions between 15 % and 30 %. In the following weeks, NO₂ concentrations stayed more or less on the same level in most parts of Europe, but the lockdown effects decreased slightly as it could be expected from the emission changes. Overall, relative concentration reductions were most significant in England, France, Belgium and Italy, as it was seen for the second half of March. Maps for relative reductions due to the lockdown for six half-month periods between 1 March and 31 May 2020 are given in Appendix A (Fig. A6).

Figure 10CMAQ results for NO₂ concentrations in central Europe between 16 and 31 March 2020. (a) Concentrations without lockdown measures (noCOV run). (b) Absolute concentration reductions due to lockdown measures (noCOV – COV run). (c) Relative concentration reductions due to lockdown measures (noCOV – COV run). Positive values for absolute and relative differences denote high reductions.

5.1.2 O₃ concentrations

It can be expected that reduced NO_x emissions are also reflected in modified O₃ concentrations with lower values in all regions that are NO_x-limited. However, for the second half of March increased O₃ concentrations between 1 and 8 µg m⁻³ were modelled in the COV case for northern central Europe and the Po Valley (Fig. 11). Because these are the regions with the highest NO_x emissions in Europe, they were most likely VOC-limited during this first lockdown period, and O₃ titration with NO was reduced when NO_x emissions were reduced. Most of the southern parts of the modelling domain exhibited a decrease in ozone of 1–2 µg m⁻³ on average caused by the lockdown and the reduced NO_x emissions. In the following weeks, areas with increased ozone turned smaller week by week and were limited to large cities and the most densely populated areas; see Fig. 12 for the first half of April and the first half of May. Most regions in Europe turned into NO_x-limited areas in spring 2020, resulting in lower ozone concentrations of 1–2 µg m⁻³ (about 2 %–4 % change) caused by the emission changes during the lockdown (Fig. A7).

Figure 11CMAQ results for O₃ concentrations in central Europe between 16 and 31 March 2020. (a) Concentrations without lockdown measures (noCOV run). (b) Absolute concentration reductions due to lockdown measures (noCOV – COV run); positive values denote high reductions. (c) Relative concentration reductions due to lockdown measures (noCOV – COV run); positive values denote reductions, and negative values denote increases.

Figure 12CMAQ results for changes in O₃ concentrations due to lockdown measures in central Europe between 1 and 15 April 2020 (a) and 16–31 May 2020 (b). Positive values denote concentration reductions, and negative values denote concentration increases.

5.1.3 PM_2.5 concentrations

Simulated PM_2.5 concentrations in the second half of March 2020 for the noCOV case show relatively high concentrations between 12 and 15 µg m⁻³ in large parts of central Europe and the Po Valley, while the UK, Denmark and northern Germany exhibited concentrations below 10 µg m⁻³ (see Fig. 13, top). The lockdown emission reductions lead to concentration reductions between 1 and 3 µg m⁻³ in those regions with higher concentrations and values below 1 µg m⁻³ in the north-western part of the domain. Relative concentration decreases were most significant in France and northern Italy with values up to 20 %, while in the rest of the domain 6 %–10 % lower PM_2.5 was simulated. In the following weeks, PM_2.5 concentrations were typically reduced by 10 %–20 % because of the lockdown measures in most parts of central Europe. Somewhat lower values were found in the northern and southern parts of the domain. The reduction in PM_2.5 concentrations decreased to 6 %–12 % in the second half of May (see Fig. A8).

Figure 13CMAQ results for PM_2.5 concentrations in central Europe between 16 and 31 March 2020. (a) Concentrations without lockdown measures (noCOV run). (b) Absolute concentration reductions due to lockdown measures (noCOV – COV run); positive values denote reductions. (c) Relative concentration reductions due to lockdown measures (noCOV – COV run); positive values denote reductions.

An investigation of the chemical components of the modelled PM_2.5 concentrations for the noCOV case reveals that about two-thirds consists of the inorganic components nitrate (NO${}_{\mathrm{3}}^{-}$), sulfate (SO${}_{\mathrm{4}}^{-}$) and ammonium (NH${}_{\mathrm{4}}^{+}$). The lockdown measures caused large reductions in NO_x emissions. Consequently, nitrate was reduced by more than 24 % in large parts of France and northern Italy between the middle of March and the end of April; see Fig. A9 in Appendix A. The reduction was usually somewhat lower in other parts of the domain. Particulate nitrate is mostly bound to ammonium; however, the model results show a lower relative reduction of the ammonium concentrations compared to nitrate. It is only on the order of 8 %–20 % at maximum (Fig. A10). This is because ammonium is preferably bound to sulfate in atmospheric aerosols, and sulfate concentrations even increased by a few percent as a consequence of the lockdown measures (Fig. A11). This can be explained by the large reduction in the formation of particulate nitrate in the COV case. Less nitrate means less ammonium which is then available as gaseous ammonia. This may lead to the formation of additional ammonium sulfate in areas where gaseous sulfuric acid is available.

5.1.4 Temporal development of concentration changes

The detailed temporal development of the effect of lockdown emission reductions on atmospheric concentrations of NO₂, O₃ and PM_2.5 is followed at selected measurement stations. Figure 14 shows the modelled differences between the noCOV and the COV model runs at Waldhof, Vredepeel and San Rocco. Lockdown emission reductions lead to reduced concentrations of NO₂ and PM_2.5 at all stations; however, the amount varies considerably in time and by station. At Waldhof, only very small changes are simulated. At Vredepeel, NO₂ is significantly reduced (by more than 10 µg m⁻³ on individual days), and PM_2.5 shows only small reductions. At San Rocco, both NO₂ and PM_2.5 are reduced by several µg m⁻³ until the end of April. In May and June, lockdown effects on the concentrations get much smaller, also at Vredepeel and San Rocco.

Figure 14Temporal development of the differences in the simulated concentrations of O₃ (red), NO₂ (blue) and PM_2.5 (black) in Waldhof (a), Vredepeel (b) and San Rocco (c) between 1 January and 30 June 2020.

Download

O₃ shows higher values despite the emission reductions until the middle of April at Vredepeel and San Rocco. This is because these stations are in VOC-limited areas at that time, where NO_x emission reductions lead to decreased O₃ titration. This pattern changes towards the end of April, and in the following O₃ is decreased on most of the days at all stations as a consequence of lower NO_x emissions. This effect remains variable at Vredepeel, a station close to the region with highest NO_x emissions in Europe. At Waldhof, O₃ reductions are observed between the beginning of April and the end of June. On average, between 16 March and 30 June, O₃ is only decreased by 0.6 µg m⁻³ (< 1 %) at Vredepeel. At Waldhof and San Rocco, the reductions are 1.2 µg m⁻³ (1.6 %) and 1.5 µg m⁻³ (1.9 %), respectively.

6.2.1 NO₂ concentrations

During the 6 weeks of the most stringent lockdown measures in central Europe (16 March to 30 April), emission reductions caused NO₂ concentration reductions between 15 % and more than 50 %. This is in good agreement with other studies (Velders et al., 2021; Menut et al., 2020; Gaubert et al., 2021) and also close to what was estimated from satellite observations. Bauwens et al. (2020) report columnar NO₂ reductions of approx. 20 % around Hamburg, Frankfurt and Brussels; 28 % for the area around Paris; and 33 %–38 % for northern Italy. These reductions are almost independent of the meteorological situation, as can be seen in Fig. 16 (top row). Differences in modelled NO₂ concentrations between 2020 and 2016 or 2018 show variations of more than 30 %, but they are fluctuating in both directions on small spatial scales (see Fig. 15, top row). Larger areas with systematic differences are mainly found over sea and in areas with relatively low average concentrations, like in the western UK. It can be concluded that the NO₂ concentration reductions during the lockdown were dominated by the emission reductions and not very much by the meteorological situation. This is in agreement with the fact that NO₂ concentrations are spatially closely connected to the emission sources. NO₂ is quickly formed from NO after the latter was emitted into the atmosphere. It will then react further to form O₃ at daytime. Compared to O₃ and secondary PM, NO₂ is a rather short-lived gas with high spatial gradients and a clear annual cycle. However, as the situation in February 2020 shows, very unusual meteorological conditions can also cause large deviations from expected concentrations.

6.2.2 O₃ concentrations

Ozone concentrations depend more strongly on weather conditions and on emissions of other precursors like VOCs. Therefore, meteorological variations from year to year might have a much stronger influence on average concentrations than the emission reductions during the lockdown. The 6-week-average ozone concentrations vary by ±15 % between 2020 and 2016 or 2018 (Fig. 15, middle row), while the lockdown effects are mostly in the range of ±5 % (Fig. 16, middle row), except in densely populated areas. Weather conditions between 16 March and 30 April 2020 favoured relatively lower ozone concentrations in most parts of central Europe when compared to 2016 and 2018. In the simulations, only areas in the western Alpine region show higher ozone in 2020 (Fig. 15, middle row). First of all, this is surprising because 2020 was comparably sunny and dry, which should favour ozone formation. The latter was also stated by Deroubaix et al. (2021) and Gaubert et al. (2021) in their studies about the COVID19 lockdown effects on air quality. However, advection of relatively clean air from Scandinavia into the north-eastern part of the model domain led to lower ozone concentrations particularly in the second half of April. A comparison of the meteorological effects on NO₂ and O₃ in Fig. 15 also shows that NO₂ was relatively high and O₃ relatively low in 2020 in the English Channel, in the south-western UK and Belgium. The high-pressure situation with relatively low wind speeds in 2020 resulted in efficient ozone destruction at night in areas with high NO emissions.

Lockdown effects on ozone might differ in sign under different meteorological conditions, as can be seen in Fig. 16. The emission reductions caused relative ozone increases in urban areas and throughout the northern part of the model domain, because these areas are VOC-limited regions. This was also reported by Menut et al. (2020) and Mertens et al. (2021). The effect is most pronounced in the second half of March and then decreases over time when VOC emissions, in particular from natural sources, increase (Fig. A7). In northern central Europe the small effects on ozone are connected with advection of clean air from the north-east. For most parts of central Europe, O₃ concentrations were decreased by lockdown measures. About 2 %–4 % O₃ concentration reductions in most parts of central Europe could have been expected with 2018 meteorological fields, when solar radiation was lower but more southerly winds prevailed in northern central Europe. On the other hand, with 2016 meteorological conditions ozone changes would show similar patterns to 2020. Ozone chemistry depends on radiation, precipitation, atmospheric mixing and the availability of precursors in a complex way. The response of ozone concentrations to emission changes is therefore not straightforward to predict. Also long-range transport, which was neglected here, may play role (see also Deroubaix et al., 2021, and Mertens et al., 2021).

6.2.3 PM_2.5 concentrations

PM_2.5 is another secondary pollutant that depends strongly on weather conditions, but emission reductions will primarily lead to concentration reductions (see Figs. 13 and 14). However, the strength of this effect might also vary considerably with meteorological conditions. Figure 15 (bottom row) shows that the main lockdown period in 2020 was favourable for PM_2.5 formation in most parts of central Europe, with often 20 % to 50 % higher PM_2.5 concentrations compared to other meteorological situations. An exception is the north-eastern part of the model domain, where the meteorological situation in 2020 led to much lower PM_2.5 concentrations compared to 2018 (more than 50 % lower) and 2016 (20 %–40 % lower). Similar to the situation for ozone, this is connected to the easterly and north-easterly winds and the advection of clean air. Consequently, lockdown emission reductions had only very minor effects on PM_2.5 concentrations in 2020 in southern Sweden, Denmark, Poland and northern Germany.

Among the PM_2.5 components, particle-bound nitrate is reduced the most (Figs. A9–A11). Sulfate might even increase in some areas where ammonia becomes available when ammonium nitrate aerosol concentrations are reduced. Small amounts of additional ammonium sulfate can then be formed. Reduced VOC emissions are likely to also cause a decrease in secondary organic aerosol (SOA) formation, as proposed by Gaubert et al. (2021). Given the uncertainties in SOA formation mechanisms in regional CTMs (Bessagnet et al., 2016), lockdown effects on SOA were not investigated in this study.

Higher PM_2.5 reductions would have been observed in most parts of Europe with 2016 and 2018 meteorological conditions. This can be interpreted in such a way that the main lockdown period in 2020 was favourable for PM_2.5 formation in large parts of Europe, leading to smaller relative PM_2.5 concentration reductions, given that the emission changes are the same.

In summary, it can be said that the effects of lockdown emission reductions depend strongly on the meteorological situation and that concentration changes because of weather conditions might be stronger than those of large emission changes during a 6-week period in spring. However, this mainly holds for the secondary pollutants O₃ and PM_2.5, while the effects on NO₂ concentrations are less pronounced. In particular, changes in O₃ concentrations are difficult to predict because of the complex emission–chemistry–meteorology interactions.