Impact of COVID-19 pandemic related to lockdown measures on tropospheric NO2 columns over Île-de-France

The evolution of NO2, considered as proxy for air pollution, was analyz ed to evaluate the impact of 1 st lockdown 10 (March 17 – May 1


Introduction 25
Megacities can be considered as a hot spot of anthropogenic pollution due to the concentration of population and human activities. People living in urban areas are exposed to air quality levels that often exceed the World Health Organization (WHO) recommended limits (WHO, 2006). In 2020, the emergence of a novel coronavirus that causes the COVID-19 disease in many countries around the world has prompted the governments of the affected states to apply restrictive regulations. Most countries implemented lockdown measures (restrictions on people movements) to limit the progression of 30 https://doi.org/10.5194/acp-2021-456 Preprint. Discussion started: 18 June 2021 c Author(s) 2021. CC BY 4.0 License. the COVID-19 pandemic. As a result, urban areas have become interesting "laboratories" for analyzing the impact of these measures on air quality. Atmospheric concentrations of air pollutants in megacities were expected to decrease as a direct impact of air and road traffic activity drop during the lockdown period. Observations of TROPOMI instrument onboard the Copernicus Sentinel 5-Precursor (S5P) satellite (Veefkind et al., 2012) were the earliest ones to be presented by the media to show the significant decrease of tropospheric NO 2 columns in the Hubei province in China (20-50% in urban areas, Ding et 35 al., 2020), the first region affected by the COVID-19 in December 2019. Indeed, tropospheric NO 2 is considered as a good proxy for NO x (NO x =NO+NO 2 ) concentrations since NO is rapidly converted into NO 2 by the photochemical cycle involving tropospheric ozone. The NO x are directly linked to human activities especially with respect to the combustion of fossil fuel (∼50%) and the biomass burning (∼20%) at global scale (Delmas et al., 1997).
Many studies have focused on NO 2 reductions due to lockdowns in 2020 at specific cities in China (Ding et al., 2020, 40 Griffith et al., 2020, and in other affected countries (Bauwens et al., 2020, Prunet et al., 2020 using only satellite observations (Bauwens et al., 2020, Koukouli et al., 2020, Liu et al., 2020 or additionally ground-based instruments (Biswal et al., 2020, Prunet et al., 2020. Other studies analyzed the lockdown period using in situ monitoring networks in the cities (Baldasano, 2020, Biswal et al., 2020, Krecl et al., 2020. Model simulations were also analyzed to assess the respective NO 2 decreases (Koukouli et al., 2020, Liu et al., 2020, Menut et al., 2020. 45 The objective of this study is to quantify the effect of NO 2 decreases due to lockdown considering long-term variability and meteorological conditions over Ile-de-France region during the last decade. Two complementary sites are used, one in the center of Paris and the other one in the peripheral zone. Ground-based and satellite measurements of tropospheric NO 2 columns are analyzed to characterize these sites. In addition, measurements of surface NO 2 concentration are used to complement the analysis at the lowest altitude layers near pollution sources. The study relies not only on a single reference 50 year before the COVID-19 pandemic, but on a long decadal data set, in order to account for variability as well as long-term trends. Specific data filtering using wind speed and direction is applied in order to isolate data which are affected by local pollution in the Greater Paris area, and to consider the changes in meteorological conditions for the different years. This paper is organized as follows. Observations of tropospheric and surface amounts of NO 2 by ground-based and satellite measurements are presented in Sect. 2 as well as the wind data from European Reanalysis. The description of the method 55 used to discriminate specific data to calculate NO 2 decrease in 2020 taking into account similar meteorological conditions is presented in Sect. 3. The results of NO 2 decreases in 2020 due to lockdown are shown in Sect. 4 for the different datasets. The results of NO 2 level reductions in respect to literature findings are discussed in Sect. 5. Conclusions are finally presented in Sect. 6.

NO Data 60
Tropospheric NO 2 columns measured by two ground-based SAOZ instruments were analyzed to trace and intercompare the evolution of NO 2 in the urban and suburban regions of Ile-de-France. The analysis was supplemented by a study of NO 2 https://doi.org/10.5194/acp-2021-456 Preprint. Discussion started: 18 June 2021 c Author(s) 2021. CC BY 4.0 License. column satellite measurements using the TROPOMI instrument. In addition, the in-situ measurements of NO 2 surface concentrations from the AIRPARIF air quality network were also considered. In this work, the ten years period 2011-2020, with the first year corresponding to the start of SAOZ measurements at the suburban site of Guyancourt was considered. 65 Table 1 shows the ground-based stations, type of instrument and geographical coordinates.

SAOZ data
The NO 2 tropospheric columns at Ile-de-France region are measured by two ground-based SAOZ (Système d'Analyse par Observation Zénithale) instruments (Pommereau and Goutail, 1988) that are part of French research infrastructure ACTRIS.
The first one was installed in 2005 at the observation platform QUALAIR (http://qualair.aero.jussieu.fr/) of Sorbonne University in Paris (urban station) and the second one is operational at LATMOS laboratory in Guyancourt (South-West 75 suburban station) since 2011. SAOZ is a UV-Visible spectrometer primary designed for monitoring stratospheric ozone and NO 2 during twilight observations in the frame of NDACC (Network for the Detection of Atmospheric Composition Change) (see Hendrick et al., 2011 for a description of retrieval). The long-term data series of SAOZ instruments were compared with data from most satellite missions to validate or monitor their performance. For example, SAOZ instruments participated in the validation of the latest satellite mission Sentinel 5 Precursor launched on October 2017 for the measurements of ozone 80 (Garane et al., 2019) and stratospheric NO 2 (Verhoelst et al., 2021) columns.
During the day, SAOZ observations are sensitive to increased tropospheric NO 2 amounts in polluted regions (Tack et al., 2015). Every ~2 minutes, the sunlight backscattered by the atmosphere in the zenith direction of SAOZ is acquired and the DOAS (Differential Optical Absorption Spectroscopy) method (Platt and Stutz, 2008) is applied in the NO 2 absorptions bands to obtain the respective slant column densities. The stratospheric NO 2 columns are removed from slant columns to 85 retrieve the tropospheric NO 2 for Solar Zenith Angles (SZA) lower than 80°, see Dieudonné et al. (2013) for a detailed description of the SAOZ tropospheric NO 2 retrieval. The SAOZ dataset of tropospheric NO 2 measurements at Paris was used in different studies to relate NO 2 concentrations at the surface with integrated NO 2 column in the boundary layer (Dieudonné https://doi.org/10.5194/acp-2021-456 Preprint. Discussion started: 18 June 2021 c Author(s) 2021. CC BY 4.0 License. et al., 2013), to interpret ozone measurements (Klein et al., 2017) and the seasonal cycle of ozone gradient (Ancellet et al., 2020). 90 SAOZ tropospheric NO 2 columns are available at the SAOZ webpage (http://saoz.obs.uvsq.fr/SAOZ_tropo_Paris.html and /SAOZ_tropo_Guyancourt.html, last access on 1 January 2021). These data were daily averaged between 6 and 18 UT and between 11 and 14 UT for comparison with satellite observations.

TROPOMI data
Tropospheric NO 2 columns retrieved by TROPOspheric Monitoring Instrument (TROPOMI) aboard Sentinel 5 Precursor 95 (S5P) satellite (Veefkind et al., 2012) launched in October 2017 were also used to discriminate air masses above SAOZ instruments benefiting from the high spatial resolution of this instrument (3.5 × 7 km 2 and 3.5 × 5.5 km 2 since August 2019).
TROPOMI is a passive-sensing hyperspectral nadir-viewing imager, aboard a near-polar sun synchronous orbit satellite at an altitude of 817 km, with an overpass at 13:30 local time and practically daily global coverage.
Retrieval applied on TROPOMI data allows distinction between tropospheric, stratospheric and total NO 2 columns. The 100 algorithm was adapted from DOMINO/TEMIS approach for OMI (Boersma et al., 2007(Boersma et al., , 2011 based on Differential Optical Absorption Spectroscopy (DOAS) method to obtain slant column densities (SCD) of NO 2 that are assimilated to the TM5-MP Chemical Transport Model (CTM) to separate the SCD. The CTM runs using 0-12 h forecast meteorological data from European Centre for Medium-Range Weather Forecasts (ECMWF) correspond to the OFF-line product. Finally, each slant column is converted to vertical column using pre-calculated Air Mass Factor (AMF) look-up-tables. Detailed description can 105 be found at TROPOMI webpage (http://www.tropomi.eu/data-products/nitrogen-dioxide). Verhoelst et al. (2021) compared NO 2 total, tropospheric and stratospheric columns with the data of ground-based instruments Pandora, Multi-Axis Differential Optical Absorption Spectroscopy (MAX-DOAS) and Zenith-Scattered-Light DOAS (ZSL-DOAS or SAOZ) distributed around the world. Observations from MAX-DOAS were used for tropospheric comparisons since they are sensitive to absorbers in the lowest few kilometers of the atmosphere (Hönninger et al., 2004). A 110 negative bias of 23 to 37% is observed in the cases of clean to slightly polluted conditions. In the case of highly polluted areas, the bias can reach -51%.
TROPOMI tropospheric NO 2 columns have been widely used to estimate the reduction of NO 2 amounts linked to the lockdown in 2020 established in different countries to prevent the spread of COVID19 (e.g. Bawens et al., 2020, Biswal et al., 20202020, Koukouli et al.,2020, Lieu et al., 2020. 115 TROPOMI tropospheric NO 2 columns are available at Copernicus webpage (https://s5phub.copernicus.eu). The data have been filtered using the quality assurance value higher than 0.5.

Surfaces concentrations
AIRPARIF is a network of standard in situ sensors to monitor air quality over Ile-de-France region. One of the key variables measured by AIRPARIF is NO 2 . Hourly NO 2 concentrations are measured at most of the stations. The concentrations are 120 https://doi.org/10.5194/acp-2021-456 Preprint. Discussion started: 18 June 2021 c Author(s) 2021. CC BY 4.0 License. measured by chemiluminescence (Fontijn et al., 1970) where the NO 2 amount is obtained after reduction to NO on a heated molybdenum converter. This kind of in situ sensor can overestimate ambient NO 2 concentrations due to interferences with non-NO x fraction of reactive nitrogen (NO z ). As an example, for urban sites in Mexico-city, Dunlea et al., 2007 found an average NO 2 overestimation for this type of sensor by 22%. AIRPARIF network is formed by the 1) so-called "traffic" stations located at the edge of major traffic axes, 2) urban 125 background stations, located in the city but not in the immediate vicinity of emission sources, 3) suburban and rural stations, and finally, a station installed on the top of the Eiffel Tower at an altitude of 300 m.
In this study, two AIRPARIF sites near the SAOZ of Paris were used, one considered as "traffic" site (Quai de Célestins) and the other as "urban" (Paris 13). AIRPARIF data of Versailles, nearest station to the SAOZ of Guyancourt was used to represent suburban site. Finally, two more stations at the foot (Paris 7) and on top of Eiffel Tower were considered to 130 compare evolution of NO 2 concentration at different altitudes in the boundary layer. Data were obtained from Airparif webpage (https://www.airparif.asso.fr/telechargement/telechargement-station, last access on 22 January 2021). Daily average data between 0 and 24h UT are used in this study.

ERA5 is the last reanalysis of the ECMWF (European Centre for Medium-Range Weather Forecasts) generated by 135
Copernicus Climate Change Service. ERA5 is produced by the Integrated Forecast System (IFS) CY41r2 version released in 2016 with a ten-member 4-D-Var assimilation each 12 hours. The horizontal grid resolution is ~31 km with 137 hybrid vertical levels up to 0.01 hPa (Hersbach et al., 2020). In addition to the significant increase of horizontal and vertical resolution of ERA5, as well as the 10 years' experience of model forecast and assimilation, new and reprocessed observational data records were considered. Further information can be found in online documents at ECMWF webpage 140 (https://confluence.ecmwf.int/display/CKB/ERA5).
In this study, wind speed and direction at 950 hPa (mid-altitude of the convective boundary layer) were extracted from 0.25° horizontal resolution in latitude and longitude data over the [48.75N, 49.00N], [2.00E, 2.50E] region at noon. The available quality-checked final product was considered for January 1 st 2011 to October 31 th 2020 and a provisional product for November-December 2020, the latter is expected to rarely differ from the final product (Hersbach et al., 2020). 145

Methodology
The evaluation of lockdown effects on atmospheric NO 2 amounts is performed by selecting air masses moving from the Parisian agglomeration to the suburban region. The objective is to consider only days when air masses for both sampling sites have a long enough residence time over the Paris area and have been influenced by local pollution. Combined wind speed and direction are considered in this study to identify such days. This procedure aims at selecting datasets with similar 150 meteorological conditions for different years, so reducing the impact of interannual weather variability. The evolution of https://doi.org/10.5194/acp-2021-456 Preprint. Discussion started: 18 June 2021 c Author(s) 2021. CC BY 4.0 License. NO 2 concentrations and tropospheric columns at AIRPARIF and SAOZ stations (Table 1) are considered. The data of NO 2 concentration measurements by in situ instruments and NO 2 tropospheric column measurements by SAOZ were daily averaged between 6 and 18 UT. The measurements data are filtered using wind speed and direction of ERA5 analysis at noon to select weather conditions in which the Guyancourt site receives air masses that have passed the Paris agglomeration. 155 Equation 1 represents the estimated residential time t of air masses coming from the center of Paris to Guyancourt.
where ν era5 and θ era5 correspond to speed and direction of wind at 12 UT and 950 hPa (altitude level in the middle of the convective boundary layer), dir_g is the direction between Guyancourt and Paris (290°) and D is the approximate diameter of agglomeration (9.5 km) if we consider it as a circle. 160 Using this parameter t, three types of days were distinguished and for each class a linear fit between urban versus suburban observations was calculated: 1. Air masses of Parisian agglomeration not influencing Guyancourt or Versailles (t<0) 2. Air masses of Parisian agglomeration influencing Guyancourt or Versailles (t>0) 3. Air masses of Parisian agglomeration in a condition of weak wind influencing Guyancourt or Versailles, a subclass of the 165 precedent one (t>30 min). by grey points and case 3 by dark grey points. Linear robust fit with zero intercept was applied for the three cases to highlight the relationship between urban and suburban stations for the different conditions of wind speed and direction. For 170 each case, higher NO 2 amounts are observed at Paris, and the air masses at the surface present lower linear regression slopes than tropospheric columns. Case 1 presents the largest slopes, 2.19±0.04 (2σ standard error) for SAOZ measurements and 1.54±0.02 for AIRPARIF highlighting the importance of wind direction. In this case, air masses pass over Guyancourt without having "touched" the agglomeration. Those air masses arriving in Paris center have crossed part of the agglomeration and then show larger NO 2 columns. Case 2 and 3 correspond to air masses generally crossing first the Parisian 175 agglomeration and then south-west suburban region. They show slopes closer to unity. In case of SAOZ, the slopes of 1.24±0.02 and 1.17±0.04 were obtained for case 2 and 3, and the slopes of 1.07±0.01 and 1.04±0.02 in case of AIRPARIF, respectively. For our study, the classification of days with air masses associated to t>30 minutes will be considered, because in this case air masses pass over both stations with weak wind allowing for pollutant accumulation over the Paris agglomeration. 180 https://doi.org/10.5194/acp-2021-456 Preprint. Discussion started: 18 June 2021 c Author(s) 2021. CC BY 4.0 License.

NO 2 evolution in 2020
The period preceding the lockdown represents meteorological conditions over Ile-the-France mainly characterized by high occurrence of oceanic air masses (see Fig. S3 of Petit et al., 2021) and fairly strong south-westerly winds (Fig. 2, left wind rose) preventing pollution events over this region. Changes in weather conditions three days after the implementation of lockdown on March 17 th 2020 (middle wind rose on Fig. 2) were mostly anticyclonic contributed to the stagnation of 190 pollutants in air masses advected from Paris to Guyancourt. Low wind speeds (<6m/s) are predominantly north-easterly in the mid-March-to mid-May period. The period after the end of lockdown (Fig. 2, right wind rose) shows winds from southwesterly and north-easterly directions in the mid May to July period.    During period 1 (before the lockdown) only two particular events with high NO 2 values above both stations are detected at the same time (t>0 min) by SAOZ instruments (Jan. 19-25 and Feb. 5-6). These events are also highlighted in AIRPARIF data. Only one day with t>30 min is observed on Feb. 5 th . Frequent occurrence of oceanic air masses with high precipitation 210 and wind speed leads to advection of clean air masses above the Île-de-France region before lockdown period (Viatte et al., 2021) and low NO 2 values are observed, lower than observed during period 2 (lockdown) for suburban stations (Guyancourt and Versailles). A NO 2 peak is observed on March 17 th coincident to the start of the lockdown period, which could be linked

Tropospheric NO 2 columns
TROPOMI tropospheric NO 2 measurements in 2020 were widely used to show a decrease of NO 2 amounts in different 225 countries, which was attributed to policies restricting human activities by comparing lockdown and pre-lockdown period or same period in 2019 (e.g. Ding et al., 2020;Koukouli et al., 2020;Prunet et al., 2020;Siddiqui et al., 2020). SAOZ measurements between 11 and 14 hUT were averaged to match overpass time of TROPOMI above the stations. Figure 4 shows the evolution of the monthly mean and two standard error (2σ) of tropospheric NO 2 columns above Paris and Guyancourt stations since January 2019 observed by SAOZ and TROPOMI (left panels). Similar inter-monthly evolution is 230 observed by both instruments with a generally good agreement within ±2σ and a correlation of 0.80 at Paris and 0.70 at Guyancourt. TROPOMI presents generally lower NO 2 values than SAOZ but within the 2σ uncertainty level. This is not the case in May 2020 (month 17 on Fig. 4) during which TROPOMI NO 2 amounts are significantly larger at 2σ level than SAOZ. Monthly mean values present a seasonal variation reaching values above 10 Pmolec cm -2 in winter at Paris while they vary between 4 to 7 Pmolec cm -2 at Guyancourt. The first months of 2020 present lower values compared to 2019, 235 mostly due to weather conditions while March-May NO 2 decrease (month 15-17) is coincident with the lockdown period. A histogram of the differences between TROPOMI and SAOZ is also shown in Figure 4 (right panels). A mean and median difference of -0.2 Pmolec cm -2 and +0.12 Pmolec cm -2 respectively is obtained at Paris station and of -0.6 Pmolec cm -2 and -0.7 Pmolec cm -2 respectively at Guyancourt. It corresponds to a median relative difference of 2% at Paris and -22% at https://doi.org/10.5194/acp-2021-456 Preprint. Discussion started: 18 June 2021 c Author(s) 2021. CC BY 4.0 License.
Guyancourt stations. Dispersion of the difference represented by the half of the 68% interpercentile (IP68/2) is 2.9 and 240 1.6 Pmolec cm -2 respectively at Paris and Guyancourt.

median, mean and dispersion by the half of the 68% interpercentile (IP68/2).
TROPOMI and SAOZ data selected for days with t > 30 min were averaged between 11h and 14h UT for the period of the 2020 lockdown in France (March 17 th to May 10) and median values were computed from the SAOZ and TROPOMI data for the 2011-2020 annual range ( Figure 5). TROPOMI NO 2 decrease in 2020 compared to 2019 is 35±12% for Paris and 250 22±27% for Guyancourt. Bauwens et al. (2020) have found a decrease of 28% during the 21 st days of lockdown over 50 km region centered at Paris using TROPOMI and OMI data compared to same period in 2019. A larger tropospheric NO 2 decrease of about 47% is found from SAOZ observations between 2019 and 2020 at both studied stations (see Figure 5). Prunet et al. (2020) found an even large decrease of NO 2 values varying from 52% to 86% during the lockdown in a 120 km region around Paris using yearly 2019-2020 TROPOMI data and the city-scale NO 2 plume mass method. 255 It should be noted that the SAOZ data sets show a long-term negative trend since 2011.  To better account for traffic-related pollution events in the daily averaged NO 2 columns the full daytime data of tropospheric NO 2 measurements by SAOZ (SZA<80°) of the corresponding day were considered. The median value of daily columns with t>30 minutes was computed for each year during period 2 and 3 above Paris and Guyancourt. Period 1 and 4 were not 270 considered since only one day with t >30 min was observed above the stations during these periods in 2020. Period 3 was restricted to May 11 th -July 15 th (period 3') to avoid the effect of NO 2 seasonal variation in the final median value. A robust regression fit (reweighted least squares with the bi-square weighting function) was applied to period 2 and 3' to compute the trend for the 2011-2019 period. We will focus only on the period of lockdown since important NO 2 interannual variability in the period 3' does not present a 2σ significant slope value neither at Paris, nor at Guyancourt. Only the lockdown period 275 presents a significant negative slope of -1.51±0.48(1σ) Pmolec cm -2 yr -1 at Paris and -1.42±0.14(1σ) Pmolec cm -2 yr -1 at Guyancourt as shown in Figure 6. These values correspond to a negative trend of -5.86±1.92 % yr -1 at Paris and -6.79±0.66 different results, e.g. slightly lower value at Paris (55±10.7%) and even higher at Guyancourt (58.9±12.5%) when using year 2018 as a reference ( Figure 6). Moreover, choosing earlier years as a reference would pose the problem of NO 2 variability factors associated with both the lockdown and the long-term NO 2 reductions. This confirms the advantage of our method that calculates the reference from a decadal data base and corrects for the long-term trend. It should be noted that the data 290 filtering procedure based on meteorological conditions (wind speed and direction) significantly changes the result of the NO 2 reduction estimate in Guyancourt, making it statistically insignificant (9.7±41.6%) if filtering is not applied; at the same time the estimate for Paris has not changed much (58.3±20.9%). Table 2 presents a summary of the NO2 reductions in 2020 using different datasets described previously in the text. This indicates that results at Paris site located in the center of the agglomeration are not dependent in 2020 on meteorological conditions. On the contrary, for the Guyancourt site at the edge 295 of the agglomeration selecting the days when the site is impacted by emissions within the agglomeration is crucial.

Surface NO 2 concentrations
The annual median NO 2 concentration at AIRPARIF stations since 2011 (Table 1) were computed from daily available hourly data during the lockdown period filtered for the wind speed and direction as it has been done for the tropospheric NO 2 column (t>30 minutes). Figure 7 presents the interannual variability of NO 2 concentration at the five AIRPARIF stations. In addition, the calculated robust fit for the decadal evolution at each station is shown. The background or urban stations (Paris  Table 1).
The five AIRPARIF stations present negative trends from -3 to -1.3 µg m -3 yr -1 equivalent to -4.6 to 2.4 % yr -1 (Table 3). Font et al. (2019) found similar negative trend varying from -3.4 to -2.4 % yr -1 for roadside stations at Paris for the 2010-2016 period. These trends appear to be less negative than those obtained from column measurements. Possible reasons for 320 this are an increase of the NO 2 to NO x emission ratio, and a limitation by the available amount of O 3 for the NO to NO 2 conversion. Both factors affect more strongly the surface concentration than the tropospheric column, which could lead then to the different trend estimates. That is, at the surface, the system is to some extent buffered against changes in NO x emissions. In 2020 significant decrease compared to the extrapolated value using the above calculated linear trends is observed at all stations and reach similar median values, slightly higher for the traffic station and slightly lower for Eiffel 325 Tower observation station. The relative values of NO 2 reductions are shown in Table 3. Comparable values at 1σ are observed for traffic and urban stations in Paris, with lower values at Paris 13 where standard error is higher. Nevertheless, https://doi.org/10.5194/acp-2021-456 Preprint. Discussion started: 18 June 2021 c Author(s) 2021. CC BY 4.0 License. the reduction of NO 2 concentration observed in absolute values is more important at traffic stations (as CELES) compared to urban station (as Paris 7). The observation station installed at 300 m of Eiffel Tower presents 53% of reduction identical to Paris 7, station located at the foot of the tower. The suburban station of Versailles presents the lowest reduction of 28.5%, 330 significantly different to other stations at 1σ except for Paris 13. It should be noted that both stations show an almost twice larger standard deviation of 14%. Reasons for these lower values are not clear. It can be speculated that at this suburban site the relative contribution of residential heating to NO x sources is stronger than at Paris sites, and probably these sources have increased during the lockdown due to the presence of people at home (Menut et al., 2020). for traffic stations. However, when considering similar meteorological conditions with respect to rainfall, temperature and wind speed, the authors found a reduction of 51.5 % corresponding to traffic stations and approximately 45% for background ones, similar to values obtained in this study.

Discussion
Various studies have been conducted to assess the impact of recent lockdowns on air quality in many countries around the 345 world due to COVID-19 pandemic. In a number of works, the observed NO 2 contents were compared with respective levels for the same period of previous years using ground-based and/or satellite measurements. Shi and Brasseur (2020) found a decrease of NO 2 concentrations in China by 50% compared to 2019 during the same period of the lockdown and by 60% compared to 2018, highlighting the interannual variability of NO 2 reductions that could depend on meteorological conditions or long-term variability. Others authors compared NO 2 amounts before and during lockdown. For example, Siddiqui et al. 350 (2020) observed 46% reduction of NO 2 tropospheric columns in India using satellite data, Liu et al. (2020) estimated 48% of reduction in China before and during the Lunar New Year, which is 21% more than in previous years 2015-2019 (given that a NO 2 reduction has been observed over the past years even without COVID); Bauwens et al. (2020)  conditions in their data. In the case of Paris, 45-52% reduction of NO 2 concentration was estimated by Collivignarelli et al. 355 (2021) using equivalent temperature and wind speed days, ~50% by Barré et al. (2020) using a Gradient Boosting Machine Learning (GBML) technique. In case of tropospheric NO 2 columns measured by satellite instruments, Prunet et al. (2020) estimated a 2 weeks averaged reduction of NO 2 varying between 52 and 86% using the city-scale NO 2 plume mass method for March 16 th -April 26 th . In the present study, the long-term evolution was considered from one decade of measurements combined to air masses filtering based on slow wind speed and long residence time. The calculated reductions in the 360 tropospheric NO 2 column and surface concentration are comparable in magnitude to the results of previous studies in Western Europe: 46-56% and 28-54%, respectively. Menut et al. (2020) compared the results of two special model calculations performed for the March 2020 lockdown period in Western Europe. They used the WRF-CHIMERE model for two simulations: one using Business As Usual (BAU) scenario with classical emissions and the other one using realistic scenario taking into account an estimate of lockdown 365 measures on NO 2 in 2020. The authors found a maximum reduction of 43% of average NO 2 concentration over France. This simulation was based on a reduction in emissions of about 80% in the transport sector and 40% reduction in the industrial sector, but an increase for residential emissions during the second half of March, reducing emissions of NO x probably by more than 50% (taking into account the distribution of NO x emissions as given by CITEPA (https://www.citepa.org/fr/2020nox/). Thus, NO 2 concentration reductions are slightly lower than NO x emissions changes in these simulations, probably due 370 to an increase in the NO 2 /NO ratio for lower NO x concentrations. This suggests that, at least when spatially averaged, NO x emission reductions due to lockdown are similar to those of NO 2 surface concentrations.

Conclusions
To assess the impact of France's policy decision to limit the spread of the SARVS-CoV-2 virus by establishing a restrictive lockdown between March 17 and May 10, 2020, NO 2 surface concentrations and tropospheric columns over Île-de-France 375 were analyzed, more specifically in Paris and suburban areas in the south-west of the agglomeration. Possible factors that can influence NO 2 changes other than NO x emissions reduction due to lockdown were considered. The data sets were partitioned to select the conditions of light winds moving air masses from Paris to a suburban area in the southwest. In addition, the known long-term reduction of NO 2 is also considered using the measurements in the previous decade. The tropospheric NO 2 reduction obtained from the SAOZ data is about 50% (56% at Paris site and 46% at the southwest 380 suburban site). These values are close to the literature data found for Europe within the estimated error bars (Barré et al., 2020;Prunet et al., 2020). This work highlights the ability of satellite TROPOMI measurements to distinguish between urban and suburban sites tropospheric columns, showing higher mean values at an urban station compared to a suburban one.
The latter is also confirmed by the ground-based SAOZ measurement data. The agreement between the evolution of NO 2 in the troposphere observed at urban and suburban sites improves when selecting similar meteorological conditions. Surface 385 NO 2 concentrations inside Paris are highly influenced by local pollution and differences between the data of traffic and https://doi.org/10.5194/acp-2021-456 Preprint. Discussion started: 18 June 2021 c Author(s) 2021. CC BY 4.0 License.
background urban sites are observed as expected. Surface concentrations were reduced by ~50% at all stations (similar at ±1σ), except the site of Paris 13 in the Choisy Park that shows a lower reduction. The suburban station of Versailles presents NO 2 concentrations similar to Paris 13 and the reduction in 2020 was 10% lower, within the error bars.
The reductions at Paris sites during the lockdown are important using or not a filter to remove the effect of different 390 meteorological conditions. On the contrary, selecting data according to air mass residence time over the agglomeration, strongly changes the estimates of NO 2 reductions at the suburban sites. As expected, if filtering is not applied, lower NO 2 reductions are found for suburban sites, since the datasets include also measurements that are less affected by the agglomeration and closer to background conditions. If the long-term evolution is not considered, the computed reductions highly depend on the year of reference. In this study, a negative tropospheric NO 2 trend of -1.5 Pmolec cm -2 yr -1 (equivalent 395 to ~6.3 % yr -1 ) is observed. Surface NO 2 concentrations also show negative trends with a mean value of -2.2 µg m -3 yr -1 (~3.6 % yr -1 ).
In conclusion, the negative trend estimated during the last decade, indicates the long-term benefits of the environmental measures taken to reduce NO x emissions. The magnitude of the NO 2 supplementary reduction in 2020, which we calculate to be around 50%, is consistent with the reduction in emissions associated with the lockdown in France, as suggested in a 400 recent modelling study (Menut, 2020).
Author contributions. AP, FG contributed to the processing, analysis and availability of SAOZ data. AB and DI processed the TROPOMI data. AH provided ERA5 data above Paris. MB developed the filter method to account for meteorological 410 conditions. AP and SGB performed the statistical analysis. AP wrote the paper with the assistance from all authors.
Competing interests. The authors declare that they have no conflict of interest.