Ammonia and PM 2.5 Air Pollution in Paris during the 2020 COVID Lockdown

: During the COVID-19 pandemic, the lockdown reduced anthropogenic emissions of NO 2 in Paris. NO 2 concentrations recorded in 2020 were the lowest they have been in the past 5 years. Despite these low-NO 2 levels, Paris experienced PM 2.5 pollution episodes, which were investigated here based on multi-species and multi-platform measurements. Ammonia (NH 3 ) measurements over Paris, derived from a mini-DOAS (differential optical absorption spectroscopy) instrument and the Infrared Atmospheric Sounding Interferometer (IASI) satellite, revealed simultaneous enhancements during the spring PM 2.5 pollution episodes. Using the IASI maps and the FLEXPART model, we show that long-range transport had a statistically signiﬁcant inﬂuence on the degradation of air quality in Paris. In addition, concentrations of ammonium (NH 4+ ) and PM 2.5 were strongly correlated for all episodes observed in springtime 2020, suggesting that transport of NH 3 drove a large component of the PM 2.5 pollution over Paris. We found that NH 3 was not the limiting factor for the formation of ammonium nitrate (NH 4 NO 3 ), and we suggest that the conversion of ammonia to ammonium may have been the essential driver.


Introduction
Lockdowns imposed to decelerate the spread of the 2020 COVID-19 pandemic have led to substantial reductions of anthropogenic air pollutant emissions. Unprecedented temporal reduction of carbon dioxide (CO 2 ) emissions has been recorded at the global scale [1]. Satellite observations have enabled monitoring the decrease in concentrations of several other gaseous species (https://iasi-ft.eu/covid-19/), such as nitrogen dioxide (NO 2 ) that is mainly emitted by combustion activities (e.g., traffic, power plants, industries) over China, Western Europe, and the United States [2]. Emission abatements also influenced surface measurements of NO 2 worldwide, in particular in urban environments, such as in Barcelona (Spain, [3]), several cities in China [4], India [5,6], and Morocco [7].
In addition to primary pollutants, air quality can be affected by complex interactions between diverse air pollutants, transformation processes, and meteorological conditions. The decrease in NO x concentrations, for instance, caused an increase in surface ozone detector, the integrated concentration of NH 3 along the optical path can be determined. By dividing the integrated concentration by the length of the absorption path, the average NH 3 concentration (in µg/m 3 ) is obtained over the optical path [54]. The mini-DOAS system was evaluated during dedicated field campaigns, which demonstrated its reliability for measuring NH 3 concentrations over an urban area [55] and NH 3 vertical profiles [56] as well as its adequacy for long-term monitoring of NH 3 in an air quality network [57].
The mini-DOAS used in this study was installed at the Qualair facility in December 2019 and has operated since then. This new dataset presented the only current continuous (day and night) NH 3 measurements at high temporal frequency in Paris. Over the period of study (January-June 2020), the mini-DOAS recorded mean hourly NH 3 concentrations up to 22.6 µg/m 3 with a mean concentration of 2 µg/m 3 . An error analysis allowed estimating an average noise error on the hourly measurements of 1.6% and a detection limit of 0.5 µg/m 3 [50].

IASI
The Infrared Atmospheric Sounding Interferometer (IASI) is a spectrometer launched on board the MetOp polar orbiting satellites A, B, and C in October 2006, September 2012, and November 2018, respectively. It is operated by EUMETSAT (European Organisation for the Exploitation of Meteorological Satellites) to measure the thermal infrared radiation of the Earth in the spectral range from 645 to 2760 cm −1 with a spectral resolution of 0.5 cm −1 (apodized). The three IASI instruments (onboard MetOp A, B, and C) have a nadir-viewing geometry and a field of view composed of 2 × 2 pixels of 12 km diameter each at nadir [52]. Measurements are performed at 09:30 and 21:30 local mean solar time for the descending and ascending orbits, respectively.
The IASI NH 3 total column concentrations used in this study were derived by means of the Artificial Neural Network for IASI algorithm [58,59]. Morning observations from the version 3 of the near-real time dataset were used (ANNI-NH 3 -v3, see [60]). The number of available NH 3 observations fluctuated with the state of the atmosphere being remotely sensed (e.g., thermal contrast and cloud cover).
In this work, the IASI satellite observations were used to study the transport of NH 3 over Paris to identify the source regions of NH 3 during the pollution episodes and to evaluate the local NH 3 concentrations derived from the mini-DOAS. Therefore, we considered all measurements from IASI on board MetOp A, B, and C jointly to maximize the number of observations within a 50-km radius circle centered in Paris city center ( Figure S1 in Supplementary Materials). Hourly PM 2.5 and NO 2 concentrations were measured using automated measurement systems [61] and chemiluminescence analyzers [62], respectively. Data were obtained via the open access database provided by Airparif, the association responsible for air quality monitoring in the Paris region (https://www.airparif.asso.fr/en/telechargement/ telechargement-polluant). Over the studied period (and the investigated pollution events), PM 2.5 and NO 2 averaged concentrations were 12 (28) and 29 (35) µg/m 3 , respectively. The spatial variability of hourly concentrations, calculated as the standard deviation (SD) of the datasets available for 13 stations ( Figure S1), led to an average value of 4 (7) and 22 (17) µg/m 3 , respectively. Considering such limited variabilities, mean PM 2.5 and NO 2 concentrations representative of regional conditions are estimated here as the average of measurements at all of these stations.

Chemical Speciation of Submicron Aerosols
The major submicron chemical components (with an aerodynamic diameter less than 1 µm: PM 1 ) have been continuously characterized from 2012 onward at the SIRTA facility located about 20 km southwest of Paris city center ( [63]; see Figure S1) and integrated in  [64]). The ACSM provides concentrations of Organic Matter (OM), nitrate, sulfate, ammonium, and chloride collected at 30-min time intervals. Details of its onsite deployment and quality control and quality assurance procedures can be found elsewhere [18,65,66]. Measurements of these chemical species at SIRTA are assessed as representative of air quality conditions in the Paris region [65,67]. Hourly PM 2.5 and PM 1 concentrations measured at SIRTA were strongly correlated with the average of measurements at all the Airparif stations with R = 0.94 and R = 0.92 (p-value < 0.05), respectively, for the period of this study (see Figure S2).

The Lagrangian Particle Dispersion Model FLEXPART
The Flexible Particle (FLEXPART) model [68,69] has been used to compute airmasses' back-trajectories in order to investigate the geographical sources of pollution transported over Paris. FLEXPART is a Lagrangian tracer dispersion model, which compares well with other widely used models such as HYSPLIT (Hybrid Single Particle Lagrangian Integrated Trajectory) or STILT (Stochastic Time-Inverted Lagrangian Transport), for instance [70,71].
To determine the emission sensitivity response functions to various regions (i.e., influence of a particular geographical region to the measurements made in Paris), meteorological data from the National Centers for Environmental Prediction (NCEP) Climate Forecast System (CFS v.2) 6-h product [72] were used. The model runs involved releasing an ensemble of 160,000 tracer particles in a 0.5 • × 0.5 • grid box centered around the Paris city-center at altitudes ranging from 200 to 300 m.

NO 2 and PM 2.5 Concentrations during the 2020 Lockdown
The lockdown imposed in France from 17 March 2020 resulted in a significant reduction of emission of air pollutants. Over Paris, the usual 15.5 million daily motorized transport trips were highly limited, halving the daily NO 2 emissions for the first 3 weeks after lockdown [73]. Consequently, concentrations of NO 2 diminished by 24% (averaged over 17 March-1 June, compared to 1 January-17 March 2020, see top right panel of Figure 1). This value was lower than that for emission reductions (24% compared to 50%) because of complexity of the atmospheric chemical system [74] and also because of the drier meteorological conditions, i.e., unusual precipitation rate in March and April ( Figure S3), which favored accumulation of pollutants in the atmosphere. Before lockdown, NO 2 and PM 2.5 concentrations were also much lower than usual (−28 and −39%, respectively) due to enhanced dispersion conditions associated with a high occurrence of oceanic air masses, along with precipitation and high wind speed ( Figure S3). Nevertheless, NO 2 concentrations recorded during the 2020 lockdown were at their lowest than in the last 5 years.
On the other end, concentrations of PM 2.5 appeared to have been barely affected by the lockdown and, contrasting with previous years at the same time periods, a 16% increase was observed after the start of lockdown (averaged over 17 March-1 June compared to 1 January-17 March 2020, see bottom right panel of Figure 1). The high interannual variability of PM 2.5 concentrations for the first five months of the year (box plot's length in lower panels of Figure 1) seemed to indicate no significant patterns (reduction or increase) compared to the mean of the five previous years. Meteorological conditions highly influence PM variability in Paris [75].

Identification of PM 2.5 Pollution Episodes
Despite the exceptional lockdown situation that significantly reduced NO 2 concentrations, the Paris megacity experienced several fine particulate pollution episodes in 2020. Here we chose to use a threshold of 25 µg/m 3 on hourly concentrations to identify and investigate all five PM 2.5 pollution episodes occurring in Paris. We note that four of the five episodes were observed during the lockdown period. Table 1 summarizes the dates of these episodes, along with the associated mean and 1-σ (standard deviation) of PM 2.5 and NH 3 concentrations measured in Paris by Airparif and at the Qualair facility, respectively. Averaged PM 2.5 and NH 3 concentrations over the period of study (Jan-June 2020) were 12 µg/m 3 and 2 µg/m 3 . Pollution episodes differed in duration: 165 h, 96 h, 41 h, 13 h, and 37 h for event 1, 2, 3, 4, and 5, respectively, and composition: event 1 with high PM 2.5 (31 µg/m 3 ) but low NH 3 concentrations (below detection limit: 0.5 µg/m 3 ) whereas event 3 exhibited both high PM 2.5 (36 µg/m 3 ) and NH 3 (11 µg/m 3 ) concentrations. On the other end, concentrations of PM2.5 appeared to have been barely affected by the lockdown and, contrasting with previous years at the same time periods, a 16% increase was observed after the start of lockdown (averaged over 17 March-1 June compared to 1 January-17 March 2020, see bottom right panel of Figure 1). The high interannual variability of PM2.5 concentrations for the first five months of the year (box plot's length in lower panels of Figure 1) seemed to indicate no significant patterns (reduction or increase) compared to the mean of the five previous years. Meteorological conditions highly influence PM variability in Paris [75].

Identification of PM2.5 Pollution Episodes
Despite the exceptional lockdown situation that significantly reduced NO2 concentrations, the Paris megacity experienced several fine particulate pollution episodes in 2020. Here we chose to use a threshold of 25 µg/m 3 on hourly concentrations to identify and investigate all five PM2.5 pollution episodes occurring in Paris. We note that four of the five episodes were observed during the lockdown period. Table 1 summarizes the dates of these episodes, along with the associated mean and 1-σ (standard deviation) of PM2.5 and NH3 concentrations measured in Paris by Airparif and at the Qualair facility, respectively. Averaged PM2.5 and NH3 concentrations over the period of study (Jan-June 2020) were 12 µg/m 3 and 2 µg/m 3 . Pollution episodes differed in duration: 165 h, 96 h, 41 h, 13 h, and 37 h for event 1, 2, 3, 4, and 5, respectively, and composition: event 1 with  NH 3 concentrations derived from both the mini-DOAS and IASI, as well as the speciation of aerosols, are shown in Figure 2 (for Jan-Jun 2020) and Figure S4 (pollution episodes). NH3 concentrations derived from both the mini-DOAS and IASI, as well as the speciation of aerosols, are shown in Figure 2 (for Jan-Jun 2020) and Figure S4 (pollution episodes). NH3 total column (molecules/cm 2 ) measured in a 50 km radius circle from Paris city center (d), and mean relative contributions of major PM1 chemical species (OM: organic matter, SO4 2− : sulfate, NO3 − : nitrate, and NH4 + : ammonium) measured at SIRTA from January to June 2020, and for the 5 investigated periods (e).
Simultaneous increases of PM2.5 and NH3 hourly concentrations were observed during most of pollution episodes ( Figure 2). The aerosol composition differed, however, between episodes, with varying dominating contribution of organic matter (OM), sulfate (SO4 2− ), nitrate (NO3 − ), or ammonium (NH4 + ). The IASI total columns averaged within a Simultaneous increases of PM 2.5 and NH 3 hourly concentrations were observed during most of pollution episodes ( Figure 2). The aerosol composition differed, however, between episodes, with varying dominating contribution of organic matter (OM), sulfate (SO 4 2− ), nitrate (NO 3 − ), or ammonium (NH 4 + ). The IASI total columns averaged within a 50-km circle radius centered on the Paris city center were also higher during the pollution episodes. In addition, the mini-DOAS NH 3 measurements were in strong agreement with IASI data, with Pearson's correlation coefficient R = 0.75 (p-value < 1 × 10 −16 ) over 85 days of coincident observations (see Figure S5), illustrating the high quality of the new mini-DOAS dataset and its representativeness of the Paris area.

Sources and Transport of NH 3 during Pollution Episodes
Complex equilibrium between contributions of long-range transport, local emission and dispersion, and chemical transformations in the atmosphere control the air quality in megacities [76] such as Greater Paris. To attribute the transport input to pollution episodes detected over Paris in winter and spring 2020, the FLEXPART Lagrangian particle dispersion model was used. Figure 3 shows the sensitivity (i.e., influence) to various regions for air parcels reaching the Paris region during the five pollution episodes. Complex equilibrium between contributions of long-range transport, local emission and dispersion, and chemical transformations in the atmosphere control the air quality in megacities [76] such as Greater Paris. To attribute the transport input to pollution episodes detected over Paris in winter and spring 2020, the FLEXPART Lagrangian particle dispersion model was used. Figure 3 shows the sensitivity (i.e., influence) to various regions for air parcels reaching the Paris region during the five pollution episodes. The NH3 source regions that could have affected the air quality over Paris during pollution episodes were identified with the IASI NH3 maps (see Figure 3). NH3 hotspots averaged over the whole period of study ( Figure 3a) were very similar to those detected using the 10-year average of IASI observations [25]. These included the French Champagne-Ardennes region as well as the Netherlands and the north of Belgium and also Piedmont and Lombardy areas in the northern Italy.
Concerning event 1, no significant NH3 enhancement was observed from the IASI and mini-DOAS observations ( Figure 2). During this event, the pollution was transported from three directions (east, south, and west) according to the FLEXPART map ( Figure 4).
For episodes 2, 3, 4 and 5, FLEXPART maps show that the pollution plumes were transported from the northeast with a higher sensitivity (red and yellow colors, Figure 4) to the nearby surrounding region on the east-southeast of Paris, over which major NH3 enhancements were observed by IASI. High NH3 concentrations for episodes 2 and 5 were located over Central Europe, mainly over Germany and the northeast of France The NH 3 source regions that could have affected the air quality over Paris during pollution episodes were identified with the IASI NH 3 maps (see Figure 3). NH 3 hotspots averaged over the whole period of study (Figure 3a) were very similar to those detected using the 10-year average of IASI observations [25]. These included the French Champagne-Ardennes region as well as the Netherlands and the north of Belgium and also Piedmont and Lombardy areas in the northern Italy.
Concerning event 1, no significant NH 3 enhancement was observed from the IASI and mini-DOAS observations ( Figure 2). During this event, the pollution was transported from three directions (east, south, and west) according to the FLEXPART map ( Figure 4).
For episodes 2, 3, 4 and 5, FLEXPART maps show that the pollution plumes were transported from the northeast with a higher sensitivity (red and yellow colors, Figure 4) to the nearby surrounding region on the east-southeast of Paris, over which major NH 3 enhancements were observed by IASI. High NH 3 concentrations for episodes 2 and 5 were located over Central Europe, mainly over Germany and the northeast of France (Figure 3c,f). In contrast, high NH 3 concentrations were located over France, the Netherlands, and south of United Kingdom during event 3 (Figure 3d). During event 4, which lasted only thirteen hours over Paris, high NH 3 concentrations were located across Europe (east-west) and over the Atlantic Ocean (Figure 3e), and fast transport of the NH 3 plume from the east-northeast was supported by the transport of air parcels seen in the FLEXPART map.
Overall, the highest sensitivity from FLEXPART corresponded to NH 3 hotspots derived from the IASI maps for each pollution event except event 1 (Figures 3 and 4). This analysis confirms the predominant transboundary transport of pollutants from the east and northeast of France over Paris, which was already indicated by previous studies [20,77]. This highlights the substantial influence of long-range transport on fine particulate matter pollution over Paris. ( Figure 3c,f). In contrast, high NH3 concentrations were located over France, the Netherlands, and south of United Kingdom during event 3 (Figure 3d). During event 4, which lasted only thirteen hours over Paris, high NH3 concentrations were located across Europe (east-west) and over the Atlantic Ocean (Figure 3e), and fast transport of the NH3 plume from the east-northeast was supported by the transport of air parcels seen in the FLEXPART map. Overall, the highest sensitivity from FLEXPART corresponded to NH3 hotspots derived from the IASI maps for each pollution event except event 1 (Figures 3 and 4). This analysis confirms the predominant transboundary transport of pollutants from the east and northeast of France over Paris, which was already indicated by previous studies [20,77]. This highlights the substantial influence of long-range transport on fine particulate matter pollution over Paris.

Correlations between Hourly PM2.5, NH3, and PM1 Major Components
To better understand the origin of the particles and their composition over Paris during the period studied, the Spearman's rank correlation coefficients between hourly surface concentration of particulate matter (PM2.5), ammonia (NH3), and PM1 major components (OM: organic matter, SO4 2− : sulfate, NO3 − : nitrate, and NH4 + : ammonium) were calculated for the period of study (January-June 2020) and for the five pollution episodes ( Table 2).

Correlations between Hourly PM 2.5 , NH 3 , and PM 1 Major Components
To better understand the origin of the particles and their composition over Paris during the period studied, the Spearman's rank correlation coefficients between hourly surface concentration of particulate matter (PM 2.5 ), ammonia (NH 3 ), and PM 1 major components (OM: organic matter, SO 4 2− : sulfate, NO 3 − : nitrate, and NH 4 + : ammonium) were calculated for the period of study (January-June 2020) and for the five pollution episodes ( Table 2).
The identified PM 2.5 pollution episodes occurring over Paris seem to have different origins (transport/chemistry). Indeed, during event 3, the Spearman's rank correlation coefficients were higher than 0.71 for all the species (PM 2.5 , NH 3 , and PM 1 major components) concentrations, which suggests the build-up of pollution downwind of Paris induced by permissible meteorological conditions, e.g., lack of wind/rain that prevent the venting/raining out of pollutants (see Figure S3). The different origins of the PM 2.5 pollution episodes were also related to differences in the chemical composition of particles. During event 1, PM 2.5 and OM concentrations, which was mainly emitted within Paris by wood burning, were strongly correlated (r s = 0.90) suggesting a more typical winter event driven by local emission of pollutants. For the other episodes, the highest correlation was not found with OM but between PM 2.5 and NO 3 − or NH 4 + concentrations. Moreover, OM dominated the PM 1 composition over the whole time period and during event 1, whereas NO 3 − and NH 4 + accounted for more than one-half of PM 1 composition during the four spring pollution episodes (see Figure 2d). Thus, the sources, including transformation processes of particulate pollution, were different, which confirms the findings obtained from observations made between mid-2011 and mid-2013 in Paris [65]. ammonium) for the period of study (January-June 2020, and the five pollution events). r s > 0.85 are in bold. NH 3 concentrations derived from the mini-DOAS (differential optical absorption spectroscopy) below detection limits were discarded. Although NH 3 and PM 2.5 concentrations were moderately well correlated over the whole time period (R = 0.61), important variability existed within the correlation values (see Figure S6). The correlations between concentrations of NH 3 and PM 2.5 were not expected to be close to unity since NH 3 can react in the atmosphere, especially over cities when the availability of acidic species (mainly sulfuric and nitric acid) is sufficient to form ammonium salts. Nevertheless, because the atmospheric lifetimes of NH 3 and ammonium salts are on the order of one to a few days [78], high correlation between NH 3 and PM 2.5 concentrations can be found when a portion of NH 3 has condensed onto acids downwind of a city and been transported in a plume containing both NH 3 and ammonium salts. NH 4 + concentrations were strongly correlated with PM 2.5 concentrations for all episodes occurring in spring (r s > 0.84 compared to r s = 0.68 during non-events period) when NH 3 concentrations were high (episodes 2, 3, 4 and 5, Table 1). This suggests that NH 4 + drove a large part of PM 2.5 pollution over Paris in spring. In addition, NO 3 − and NH 4 + concentrations presented similar correlation values with PM 2.5 concentrations and were almost perfectly correlated with each other (r s ≥ 0.98 and 0.89 during pollution events and non-events period, respectively), to a larger extent than between SO 4 2− and NH 4 + , confirming the important abundance of ammonium nitrate (NO 3 NH 4 ) over Paris [20,65,77].

January-June
Finally, the contributions of SNA (Sulfate: SO 4 2− , Nitrate: NO 3 − , and Ammonium: NH 4 + ) in Paris were correlated with PM 2.5 concentrations over the period of study (r s = 0.71-0.78) and for episodes 2, 3, 4 and 5, with r s ranging from 0.59 to 0.97, showing the importance of the inorganic secondary aerosol, especially in spring pollution episodes. SNA are key species that can make up more than 50% of PM 2.5 mass loadings [79][80][81].
These findings confirm the persistent role of ammonia and ammonium nitrate in particulate pollution episodes over Paris in springtime [23,25]. The response of SNA to ammonia concentrations is detailed in the Section 4.

Discussion
In this section, we discuss the formation processes of SNA in Paris during COVID−19 lockdown despite the low NO 2 loadings. Previous studies have reported the need for better understanding of the processes involved in driving ammonium nitrate formation over Paris [65,82]. Despite the usual high NO x emissions in the Parisian city, the local formation of ammonium nitrate has been found to be minor overall [23], although it can be significant for specific episodes [24]. Here we aim to evaluate the potential formation of SNA under the exceptional lockdown situation that significantly reduced concentrations of ammonium nitrate precursors (e.g., NO x = NO 2 + NO). Indeed, it has been shown that based on observations in California, a reduction of NO x levels can affect the acidic content of the atmosphere and change the ammonium nitrate chemistry [83].
In the troposphere, NH 3 neutralizes acidic sulfate before reacting with nitric acid to condense onto aerosols as ammonium nitrate [84]. Thus, NH 3 f, the free total ammonia concentrations, defined as NH 3 f = NH 3 + NH 4 + − 2 × SO 4 2− [85], is an indicator of the excess NH 3 available to form ammonium nitrate (NH 4 NO 3 ) if all SO 4 2− has been neutralized (ratio NH 4 + /SO 4 2− higher than 2); in other words, if NH 3 f is positive and NH 4 + /SO 4 2− higher than 2, NH 3 concentrations are sufficient to form NH 4 NO 3 . Over the time period of study, molar ratios of concentrations of NH 4 + to SO 4 2− showed considerable variability with a mean of 2.09 ± 230.90 ( Figure S7, upper panel). During spring pollution episodes, this ratio was significantly higher (7.96 ± 0.04), indicating a complete neutralization of SO 4 2− by NH 4 + . Over the time period January-March 2020, the mean hourly NH 3 f was 0.15 ± 0.21 µg/m 3 ( Figure S7, lower panel). During event 1, NH 3 f was close to zero (0.08 ± 0.04 µmole/m 3 ), whereas for all the other episodes the NH 3 f were significantly positive (event 2: 0.61 ± 0.34 µmole/m 3 ; event 3: 0.88 ± 0.38 µmole/m 3 ; event 4: 0.53 ± 0.23 µmole/m 3 ; event 5: 0.51 ± 0.14 µmole/m 3 ). This indicates that NH 4 NO 3 could not be formed through NH 3 reaction in the January pollution episode, whereas NH 3 was not the limiting species for NH 4 NO 3 formation during spring pollution episodes. These observation-based results verify the findings based on numerical simulations that aerosol formation is not immediately limited by the availability of ammonia over Europe [33], at least in spring.
In addition, complete neutralization of SO 4 2− by NH 4 + was observed in Paris from January to June since the ratio of molar concentrations of 2xSO 4 2− with NH 4 + was relatively correlated (r 2 = 0.68) and significantly lower than 1.0 (0.19). During spring pollution events (black marker edges in Figure 5), the ratio of molar concentrations of 2xSO 4 2− with NH 4 + Atmosphere 2021, 12, 160 11 of 18 was even lower (0.14), suggesting that NH 4 + was mainly condensed in the (NH 4 ) 2 SO 4 in the aerosols. Moreover, the regression yielded a strong linear correlation between 2xSO 4 2− + NO 3 − with NH 4 + (r 2 = 0.99 and a slope of 0.94 for the time period of study and 0.95 during pollution events), with the sum of 2xSO 4 2− +NO 3 − sometimes exceeding the NH 4 + concentrations, implying that NH 4 + might have been present in forms other than NH 4 NO 3 over Paris.
In addition, complete neutralization of SO4 2− by NH4 + was observed in Paris from January to June since the ratio of molar concentrations of 2xSO4 2− with NH4 + was relatively correlated (r 2 = 0.68) and significantly lower than 1.0 (0.19). During spring pollution events (black marker edges in Figure 5), the ratio of molar concentrations of 2xSO4 2− with NH4 + was even lower (0.14), suggesting that NH4 + was mainly condensed in the (NH4)2SO4 in the aerosols. Moreover, the regression yielded a strong linear correlation between 2xSO4 2− + NO3 − with NH4 + (r 2 = 0.99 and a slope of 0.94 for the time period of study and 0.95 during pollution events), with the sum of 2xSO4 2− +NO3 − sometimes exceeding the NH4 + concentrations, implying that NH4 + might have been present in forms other than NH4NO3 over Paris. To further investigate the relationship between atmospheric SNA formation and NH3 concentrations during the pollution episodes occurring in spring over Paris (episodes 2, 3, 4 and 5), the conversion rate of NH3 to NH4 + , which is described by the ratio of To further investigate the relationship between atmospheric SNA formation and NH 3 concentrations during the pollution episodes occurring in spring over Paris (episodes 2, 3, 4 and 5), the conversion rate of NH 3 [42,55]. A value of AGF higher than 0.5 means that the total ammonia NH x is dominated by gaseous NH 3 rather than by the aerosol load [86]. In our study, the AGF derived from hourly measurements were highly variable with a mean of 1.01 ± 7.25, whereas during spring pollution episodes the mean AGF value was 0.71 ± 0.12, suggesting that NH 3 remained predominantly in the gas phase rather than the particle phase during spring particulate episodes in 2020 over Paris.
On the other hand, the relationships between the conversion rate of ammonia to ammonium, SNA, and NH 3 concentrations were nonlinear since higher NH 3 concentrations (red dots, Figure S8) can correspond to very low SNA concentrations. Thus, this suggests that high concentrations of precursors (NH 3 ) do not necessarily result in high secondary aerosols (SNA) formation, which is supported by the AGF values within spring pollution episodes. Therefore, we conclude that the conversion rate of ammonia to ammonium, rather than ambient NH 3 concentrations, played an important role in the SNA formation in Paris during spring pollution events in 2020.
The conversion rate of ammonia to ammonium has been reported to be strongly affected mainly by meteorological conditions (temperature and humidity) and concentrations of primary acid gas (total nitrate NO 3 (HNO 3 + NO 3 ) and total sulfate (SO 2 + SO 4 )) [79,87,88]. Low temperature and high relative humidity conditions are shown by thermodynamic equilibrium models such as ISORROPIA II [89] and observations in several cities (Beijing [90]; Shanghai [55]; Paris [25,91] to facilitate multi-phase reactions for aerosol formation in urban areas. A recent study has attributed urban PM 2.5 pollution during COVID-19 lockdown to enhanced efficiency of nitrate aerosol formation promoted by high relative humidity despite low atmospheric NO 2 levels in China [14]. However, Chang et al. (2020) argued that regional long-range transport, rather than meteorological conditions, mainly drove the enhanced nitrate formation in Shanghai during the 2020 lockdown. In Paris, relative humidity levels in 2020 were similar to, if not lower than, normal conditions, with a change of −8 ± 24% in 2020 compared to the 2015-2019 average (see Figure S9). Our results therefore tend to corroborate more the explanation of Chang et al. (2020) and indicate that relative humidity did not enhance ammonium nitrate formation over Paris in 2020.
The high SNA concentrations in spring could also be due to the nonlinear production chemistry and titration of tropospheric ozone. Indeed, the anomalously elevated ozone levels found in urban areas of France as a consequence of reduced NO x concentrations during lockdown [11] could have enhanced atmospheric oxidizing capacity (by the OH radical) and further facilitated SNA formation in Paris in 2020.

Conclusions
The COVID-19 lockdown measures reduced anthropogenic emissions and affected atmospheric composition in the urban area of Paris. Consequently, NO 2 concentrations recorded in the first half of 2020 were the lowest they have been in the last 5 years. Despite a change of NO 2 concentrations by −24% (averaged over 17 March-1 June compared to 1 January-17 March 2020), fine particles (PM 2.5 ) concentrations increased by about 16% over the same period. Five PM 2.5 pollution events, identified here when PM 2.5 concentrations were higher than 25 µg/m 3 , over the Paris region, occurring in January, March, and April 2020, were investigated in this study.
Analysis of NH 3 and PM 2.5 concentrations as well as PM 1 chemical composition were performed to better understand the origin of these episodes. Multi-platform NH 3 observations were derived from a new in situ (mini-DOAS) instrument located in Paris city center and from the IASI satellite instrument.
Atmospheric composition of the five pollution episodes was different, with a large contribution from organic matter in winter, in contrast to nitrate and ammonium in spring.
During lockdown, simultaneous enhancements of PM 2.5 , NH 3 , and SNA surface concentrations were monitored. Hence, we confirm the role of NH 3 and inorganic secondary aerosols in spring particulate episodes in Paris.
The NH 3 hotspots revealed by IASI outside of the city were found to be consistent with the highest sensitivity regions obtained with the FLEXPART model. This suggests the contribution from regional sources rather than local pollution from Paris.
Moreover, we have shown that sulfate was completely associated with ammonium and that NH 3 was not the limiting species for the formation of ammonium nitrate in 2020 over Paris. We have further demonstrated that the conversion rate of ammonia to ammonium, which mainly depends on atmospheric temperature and humidity, might be the main driver, rather than NH 3 concentrations, for SNA formation over Paris.
Finally, our results clearly illustrate the need for a better assessment of the origins of springtime PM 2.5 pollution episodes in the Paris region, which could eventually be considered by policy makers with relevance to mitigation strategies. Future investigations will focus on the impact of the exceptionally low NO 2 levels in 2020 on the relationships between NH 3 , NH 4 + , and SNA over Paris using state-of-the-art and optimized CTMs.  Figure S4: Timeseries of NH 3 , PM 2.5 and major submicron chemical species concentrations over Paris during the five periods investigated. Figure S5: IASI and mini-DOAS NH 3 comparison. Figure S6: Correlation plot (using Spearman's rank correlation coefficient) between hourly PM 2.5 and NH 3 concentrations over Paris during the period of study (January-June 2020 in grey), color-coded according to the five periods investigated. NH 3 concentrations derived from the mini-DOAS below detection limits were discarded. Figure S7: Hourly (blue) and daily (black) NH 4 + to SO 4 2− molar concentrations ratio (upper panel), and hourly (green) and daily (black) free total ammonia concentrations (µmole/m 3 ) defined as NH 3 f = NH 3 + NH 4 + − 2 × SO 4 2− (lower panel) as function of time. Figure S8: Relationship between the conversion rate of ammonia to ammonium (NH 4 + /NH x ) and SNA (sulfate, nitrate, and ammonium (µmole/m 3 )) during pollution episodes occurring in spring 2020. Figure