Formation of secondary organic aerosol in the Paris pollution plume and its impact on surrounding regions

Secondary pollutants such as ozone, secondary inorganic aerosol, and secondary organic aerosol formed in the plumes of megacities can affect regional air quality. In the framework of the FP7/EU MEGAPOLI (Megacities: Emissions, urban, regional and Global Atmospheric POLlution and climate effects, and Integrated tools for assessment and mitigation) project, an intensive campaign was launched in the greater Paris region in July 2009. The major objective was to quantify different sources of organic aerosol (OA) within a megacity and in its plume. In this study, we use airborne measurements aboard the French ATR-42 aircraft to evaluate the regional chemistry-transport model CHIMERE within and downwind of the Paris region. Two mechanisms of secondary OA (SOA) formation are used, both including SOA formation from oxidation and chemical aging of primary semivolatile and intermediate volatility organic compounds (SI-SOA) in the volatility basis set (VBS) framework. As for SOA formed from traditional VOC (volatile organic compound) precursors (traditional SOA), one applies chemical aging in the VBS framework adopting different SOA yields for highand low-NOx environments, while another applies a single-step oxidation scheme without chemical aging. Two emission inventories are used for discussion of emission uncertainties. The slopes of the airborne OA levels versus Ox (i.e., O3+NO2) show SOA formation normalized with respect to photochemical activity and are used for specific evaluation of the OA scheme in the model. The simulated slopes were overestimated slightly by factors of 1.1, 1.7 and 1.3 with respect to those observed for the three airborne measurements, when the most realistic “high-NOx” yields for traditional SOA formation in the VBS scheme are used in the model. In addition, these slopes are relatively stable from one day to another, which suggests that they are characteristic for the given megacity plume environment. The configuration with increased primary organic aerosol (POA) emissions and with a single-step oxidation scheme of traditional SOA also agrees with the OA / Ox slopes (about ±50 % with respect to the observed ones); however, it underestimates the background. Both configurations are coherent with observed OA plume buildup, but they show very different SI-SOA and traditional anthropogenic SOA (ASOA) contributions. It is hence concluded that available theoretical knowledge and available data in this study are not sufficient to discern the relative contributions of different types of anthropogenic SOA in the Paris pollution plume, while its sum is correctly simulated. Based on these simulations, for specific plumes, the anthropogenic OA buildup can reach between 8 and 10 μgm. For the average of the month of July 2009, maximum OA increases due to emissions from the Paris agglomeration are noticed close to the agglomeration at various length scales: several tens (for primary OA) to hundreds (for SI-SOA and ASOA) of kilometers from the Paris agglomeration. In addition, BSOA (SOA formed from biogenic VOC precursors) Published by Copernicus Publications on behalf of the European Geosciences Union. 13974 Q. J. Zhang et al.: Formation of SOA in the Paris pollution plume Figure 1. Ratios of OOA vs. Ox from studies in Mexico City, Houston, Los Angeles, Tokyo and Paris. Ratios for Houston, Los Angeles and Tokyo are derived from ground-based measurements during typically 1 month and located in the metropolitan area. For Houston, the ratio derived during influences from a combination of urban and petrochemical emissions, typically 0.03 μgm−3 ppb−1 (Wood et al., 2010), is presented. Ratios for Paris and Mexico City are derived from three and two individual flights, respectively, performed at about 100–150 km downwind from the agglomeration. is an important contributor to regional OA levels (inside and outside the Paris plume).

Photochemical ozone formation is related to precursor molecules: nitrogen oxides (NO x ), and volatile organic carbon species (VOC), emitted mainly from human activities, such as traffic, industrial production, solvent use, in addition to biogenic emissions. In large European agglomerations, a VOC limited chemical regime is in general realized (Beekmann and Vautard, 2010), in which ozone production is directly related to that of 10 VOC precursors. Secondary aerosol formation is induced by formation of condensable or semi-volatile species from precursors like NO x , SO 2 , NH 3 and VOC (Seinfeld and Pandis, 2006). Due to the large number of chemical reactions occurring in different phases, secondary organic aerosol (SOA) formation pathways are still uncertain (e.g. Hallquist et al., 2009), its formation is still difficult to estimate quantitatively (e.g. Hodzic 15 et al., 2010; Zhang et al., 2013), and the contribution of anthropogenic vs. biogenic sources are still under debate (e.g. Hallquist et al., 2009;Beekmann et al., 2014).
Field data provide strong constraints on SOA related processes. In particular, the relation between SOA and O x (O 3 + NO 2 ) has been used to express SOA formation as a function of oxidant formation (Herndon et al., 2008;Wood et al., 2010;Hayes 20 et al., 2013;Morino et al., 2014). Indeed, in a "VOC-limited regime" in an urban area, such as Paris, VOC oxidation by OH, ozone or NO 3 is the rate limiting step for both SOA and ozone or O x production. The ratio or slope of SOA vs. O x represents the ratio of the photochemical production of SOA to the photochemical production of O x , both from VOC oxidation, that is, the SOA yield normalized by current photochemical Introduction The main objective of the MEGAPOLI Paris campaign in summer 2009 was to de-5 termine organic aerosol sources in a post-industrial megacity and in its plume. In this work, we apply the regional chemistry transport model (CTM) CHIMERE (Menut et al., 2013) during the MEGAPOLI summer campaign in order to evaluate the model performance against airborne measurements especially for organic aerosol and to assess the impact of Paris agglomeration emissions on OA formation in surrounding regions. 10 Different configurations of the SOA formation schemes are implemented into CHIMERE, in particular the Volatility-Basis-Set (VBS) approach (Robinson et al., 2007;Donahue et al., 2006;Murphy and Pandis, 2009;Lane et al., 2008a). Based on ground level evaluation with data from the MEGAPOLI summer campaign, Zhang et al. (2013) show a better agreement with OA measurements, when taking into account the volatility 15 of primary organic aerosol (POA), the existence of additional intermediate VOC (IVOC), as well as the chemical aging of the semi-volatile VOC from anthropogenic and biogenic origin. However, SOA was overestimated during long range transport episodes. This result is consistent with previous studies for the Mexico City plume (Dzepina et al., 2011) and for Los Angeles plume (Hayes et al., 2014). 20 In the framework of the MEGAPOLI project, airborne measurements were performed with the French ATR-42 aircraft operated by the SAFIRE (a CNRS-MétéoFrance-CNES headed unit) in order to document the evolution of pollutants within the Paris agglomeration pollution plume during the MEGAPOLI summer campaign (Freney et al., 2014). The advantage of the airborne measurements over the ground based ones is to follow 25 the evolution of the city plume over time and space up to 200 km downwind of the emissions. Data from these flights will be used to extend the model evaluation performed in Zhang et al. (2013) for urban and suburban sites in the Paris agglomeration to plume conditions. Focus is again to monitor the build-up of secondary organic aerosol within the plume in relation with tracers of primary emissions, and photochemical activity. In particular, the OA/O x ratio will be analyzed and used for model evaluation.
For megacities, sources of organic aerosol are still under debate and need to be quantified (e.g. Molina et al., 2010). While in Beekmann et al. (2014), the local vs. advected and the fossil vs. non fossil nature of OA sources within the agglomeration is 5 analyzed, here we focus on additional OA build-up in the agglomeration plume, and on its impact on aerosol concentrations in the surrounding of Paris.
The paper is organized as follows. In Sect. 2, the airborne measurements during the MEGAPOLI summer campaign are described. Model configuration and simulations set-up for the VBS approach to model POA and SOA are introduced in Sect. 3. The 10 evaluation of model performance for plume simulations is presented in Sect. 4, and the impact on regional air quality is described in Sect. 5.

Airborne measurement during the MEGAPOLI summer campaign
Flight patterns flown during the MEGAPOLI campaign ( Fig. 2) consisted of several transects of the pollution plume at increasing distances from the urban area (Freney 15 et al., 2014). Flight legs perpendicular to the plume time were chosen long enough (50-100 km) to sample rural background conditions at the lateral plume edges. Taking into account the aircraft autonomy of about 3.5 h, this allowed flying four legs across the plume. The maximal distance for a flight was about 200 km from the Paris agglomeration center. The flight level was chosen to lie well inside of the well-developed afternoon conditions favor the accumulation of primary pollutants and photochemical processes leading to the formation of secondary pollutants like O 3 and SOA.
An extensive set of gas phase pollutants and aerosol species and properties were measured on each flight (Freney et al., 2014). For this work, for each flight, a complete measurement set of primary pollutants, BC and NO x (NO and NO 2 ), and of secondary pollutants, O 3 and OA, is available and analyzed. Measurement frequencies of all instruments, including the aerosol chemical composition, are rapid enough (< 40 s) to have a relatively good spatial resolution. All measurements during the flights are corrected to temperature (22 • C) and pressure (950 hPa) of the plane (Freney et al., 2014). Thus compared to other values given in this paper and taken at standard conditions, 15 our values are about 5 % lower. Table 1 summarizes the deployed instruments and the measured concentration levels for these pollutants during these flights. Only measurements at a stable flight altitude are used for this study.
The 30 percentile concentrations on the flight legs downwind of Paris represent the background levels of pollutants. For NO x and BC, they are 1.11, 1.03 and 1.14 ppb, 20 and 0.33, 0.49 and 0.38 µg m −3 on 16, 21 and 29, respectively ( The maximum plume ozone concentrations of are 62.0, 79.0 and 62.4 ppb during these flights, respectively, as compared to the 30th percentile (i.e. background) concentrations of 49.0, 58.0 and 50.0 ppb ( Table 1). The largest O 3 values are observed at the flight leg most distant from the agglomeration, allowing for the longest photochemical processing (Fig. 4). For the 16th, the transects across the plume show a double 5 maximum and a relative central minimum due to ozone titration by NO.
The background concentrations of OA are 3.87, 6.47 and 4.13 µg m −3 , respectively during these three flights ( in the three outer flight legs. OA plumes are wider and less homogeneous than primary pollutant ones, which could be due to a secondary organic aerosol production from additional biogenic sources, in addition to formation from emissions in the Paris agglomeration. OA vs. O x (O x = O 3 + NO 2 ) plots from measurements on these flights are used to 15 study the ratio of the photochemical productivity of OA and O x build-up in the plume from Paris emissions following an approach first proposed by Herndon et al. (2008). In this study, OA is used instead of SOA, because contrary to SOA, it is directly measured. Among OA factors derived from Positive Matrix Factorization of AMS measurements, LV-OOA (Low volatility oxygenated) and SV-OOA (Semi volatile oxygenated OA) are 20 commonly attributed to SOA (Hallquist et al., 2009). These factors made up for about 70 % of resolved OA factors during MEGAPOLI flights (Freney et al., 2014). HOA (hydrocarbon like OA) make up for the remaining 30 % (and always below 39 %). While this factor is generally attributed to POA, it might partly also correspond to oxidized POA, considered as SOA (Aumont et al., 2012;Cappa and Wilson, 2012 The OA/O x slopes for these flights are 0.14-0.15 µg m −3 ppb −1 . This result is close to the one obtained from a previous flight study of urban air mass in Mexico City (0.14-0.15 µg m −3 ppb −1 , Wood et al., 2010). It is also close to ground-based study, downwind urban emissions from ground-based measurements in Mexico City (median OOA vs.

Simulation configurations
Here, a brief summary on the two distinct simulation configurations with different emis-15 sions and SOA yields in the inner domain is given.
The VBS-LNOX simulation in which high SOA yields under low-NO x conditions are used (the same as the so-called VBS-MPOLI simulation in Zhang et al., 2013), assuming that most of OA is advected to the Paris agglomeration from outside (Petetin et al., 2014a) and probably formed under low-NO x conditions ( Table 2). The emissions in-20 ventory on the MEG3 domain is from the emission inventory of the MEGAPOLI project prepared by TNO for both gas-phase and particulate phase in which the refined Paris emissions from Airparif with a resolution of 1 km are integrated (Timmermans et al., 2013).
The VBS-HNOX simulation in which lower SOA yields under high-NO x conditions 25 (Murphy and Pandis, 2009;Lane et al., 2008,) are used for the MEG3 domain (see Table 2). This is more realistic for SOA formation in the Paris agglomeration and in its plume. All other model settings are equal to the VBS-LNOX configuration. For each of the 4 to 5 transects through the pollution plume of a flight, the simulated 10 and observed maximum concentrations are depicted and averaged over all transects of a flight. The same procedure is applied for P30 percentiles over each transect, considered as representative for background conditions outside of the plume.

Individual species model to observation comparisons
The visual inspection of simulated and observed BC plumes shows that the plume di-15 rection is correctly simulated on the 21 and 29, while a difference of about 20 • occurs on 16 (Fig. 2). This is still acceptable due to the rather circular structure of the agglomeration. Both in the modeled fields and in observations, most important concentrations appear close to the Paris agglomeration during these three flights.  (Table 3). As for BC, the modeled NO x maximum concentrations are located close to the Paris agglomeration ( Table 3). The modeled O 3 concentrations are slightly overestimated compared to the measured O 3 concentrations, by 7.5 (12 %), 4.3 (5 %) and 8.3 ppb (13 %) for the maximum concentrations, and 4.3 (9 %), 11.3 (20 %) and 3.0 ppb (6 %) for the background concentrations during the three flights, respectively. Note that for O x , the concentra-10 tions can be slightly less overestimated, by respectively 8.5 (13 %), 3.6 (4 %), 8.0 ppb (12 %), for the maximum concentrations, and 3.0 (6 %), 11.0 (19 %) and 1.8 ppb (4 %) for the background concentrations due to the opposite sign in O 3 and NO x differences with measurements. Similar to the observations, the modeled maximum O 3 levels are located at farthest distances from the agglomeration. The measured OA plume is corre- The plume O x production, calculated again from the difference between the maximum concentrations and the background concentrations are 12.9, 21.8 and 12.6 ppb 20 from the measurements and 18.4, 14.4 and 18.8 ppb in the simulations (again for the three flight days respectively). As for OA, we encounter an overestimation of plume O x for the 16 and 29, and an underestimation for the 21. This suggests that the representation of photochemical conditions might be partially responsible for differences observed for OA, and thus that the given data set could not be used directly for evaluation of the 25 OA scheme in the model. To overcome these problems, we analyze here OA vs. O x plots. As explained in the introduction, the slopes of these plots can represent in plume OA build-up, normalized with respect to the availability of VOC precursors and oxidant agents (OH, O 3 and NO 3 ). This holds under the ideal hypothesis of a constant mix of VOC precursors and oxidant agents for the considered data points of a flight. In Sect. 2, we presented corre-5 lations of about 0.7 (R) between OA and O x measured on the flight legs for a given day. Modeled OA and O x on these flight legs show even higher correlation of more than 0.95 ( Table 4). These good correlations suggest that we are close enough to the "constant mix" hypothesis to make the OA vs. emissions are more diffuse. Even if some differences are made evident, such a good agreement in OA vs. O x slopes between simulations and measurements represents a valuable validation of the VBS scheme for the conditions of the Paris plume.

Impact of Paris plume on surrounding regions
In this section, we analyse the contribution of OA build-up from emissions in or near 5 the Paris agglomeration to regional OA levels. This analysis is based on simulations with the VBS-HNOX configuration. We will first analyse the individual build-up of OA species for the three flight days (Sect. 5.1), then we will study the time evolution of a pollution plume on the 16 July (Sect. 5.2), and finally, we present average results for July 2009.

Plume build-up of individual OA species
The slopes of modeled SI-SOA (SOA from oxidation of primary semi-volatile and intermediate volatile VOCs, previously referred as OPOA in Zhang et al. (2013), ASOA (anthropogenic SOA) and BSOA (biogenic SOA) vs. O x are well correlated, generally with R > 0.7 (Table 4). They are used here to analyze the plume production of individ-15 ual OA species. SI-SOA is formed by functionalization and condensation of evaporated POA and IVOC species (Robinson et al., 2007)  tion can both be due to fresh BVOC emissions from mainly isoprene emitting forests north of Paris or from condensation of biogenic SVOC when temperatures decrease in the later afternoon. Figure 9 gives a typical picture of the OA species evolution in the Paris plume (at sur-10 face). On 16 July at 07:00 UTC, a morning peak of OA is formed inside the Paris agglomeration as a result of POA emissions, low wind speeds, and a low boundary layer height and is transported northeast. It disappears in the later morning (10:00 UTC) due to an increase of the PBL height and stronger winds. In the early afternoon (13:00 UTC), an OA plume is formed at about 50 km from the agglomeration center 15 due to photochemical SOA production. At 16:00 UTC, the plume travels further northeast. Largest OA values occur between 49.5 • N and 50 • N, about 100 km north of Paris, in agreement with measurements. Major contributors ASOA and SI-SOA add more than 5 and 2 µg m −3 of OA to the plume maximum (Fig. 10). The ASOA and SI-SOA plumes are clearly cut from the Paris agglomeration, (i) because of the time needed for 20 processing of precursor emissions, and (ii) because of the largest accumulation of precursor emissions in the early morning hours when wind speeds over the agglomeration were very low (also seen in the POA peak at the same location). BSOA contributes to the regional background and is little affected by anthropogenic Paris agglomeration emissions (Fig. 10). The highest OA concentrations of about 10 µg m −3 occurs in the evening at 19:00 UTC in northern France (at ∼ 150 km distance from the from the agglomeration center) due to continuous photochemical SVOC production and aging, This phenomenon of continuing SOA formation which is detached from the original rush hour emission area due to transport is very similar to that observed for Los Angeles in the CalNex study (Hayes et al., 2013). 5

Average July 2009 urban OA contribution to the surroundings of Paris
Here, we analyze the regional scale OA build-up from the Paris emissions for the average of July 2009 from the VBS-HNOX simulation (Figs. 11 and 12). Average OA concentrations around the Paris agglomeration do not show distinctive pollution plumes, but instead a strong W-E gradient near the agglomeration, presumably due to averaging different plume directions, and due to differences in background conditions. OA values also show strong decreasing gradients at about 100 to 150 km in the N-NE of Paris. This behavior can be analyzed by considering specifically the contributions to OA.
Average POA from Paris emissions is only about 0.15 µg m −3 over Paris and the 15 area of enhanced values is extending to E/NE because of the largest climatological frequency of south-westerly to westerly winds in July. The areas of enhancements of POA occur on a length scale of some tenths of kilometers around the agglomeration. ASOA is enhanced within the agglomeration and within the SW and NNE plume, up to 100 to 150 km downwind the agglomeration respectively. The maximum concentrations 20 in these plumes are 0.4 and 0.35 µg m −3 , respectively (always for the July 2009 average). In the NNE direction, enhanced values originate from pollution events under SW flow such those studied in this work (see Sect. 5.2). The enhanced values in SW originate from a pronounced pollution plume occurring in the beginning of July, for which no measurements were available. SI-SOA is most enhanced in the NNE direction where its 25 maximum concentration is about 0.35 µg m −3 , thus somewhat smaller than the ASOA concentration. It is noted that these increases in ASOA and SI-SOA concentrations are much larger when analyzing individual events than when looking at averages, due 8088 Introduction Predicted maximum OA is found on the flight leg most distant from the agglomeration (at about 150 km), as for observations, indicating secondary anthropogenic SOA formation from Paris emissions over all the distances and during several hours. On a monthly average, OA from Paris emissions contributes to the OA regional build-up at different length scales, from several tenths for POA to several hundreds of kilometers 5 for ASOA and SI-SOA. Clearly, ASOA build-up from precursor emissions in the Paris agglomeration affects atmospheric composition at regional scale. Simulating this buildup has been possible only after an original model evaluation showing good agreement between simulated and observed OA to O x slopes. These slopes are an interesting parameter to measure the SOA build-up efficiency of a given environment. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Moukhtar, S., Kolmonen, P., Stohl, A., Eckhardt, S., Borbon, A., Gros, V., Marchand, N., Jaffrezo, J. L., Schwarzenboeck, A., Colomb, A., Wiedensohler, A., Borrmann, S., Lawrence, M., Baklanov, A., and Baltensperger, U.: Regional emissions control fine particulate matter levels in the Paris megacity, Atmos. Chem. Phys. Discuss., submitted" 2014. Bessagnet, B., Menut, L., Curci, G., Hodzic, A., Guillaume, B., Liousse, C., Moukhtar, S., 5 Pun, B., Seigneur, C., and Schulz, M.: Regional modeling of carbonaceous aerosols over Europe -focus on secondary organic aerosols, J. Atmos. Chem., 61, 175-202, 2009. Cappa, C. D. and Wilson, K. R.: Multi-generation gas-phase oxidation, equilibrium partitioning, and the formation and evolution of secondary organic aerosol, Atmos. Chem. Phys., 12, 9505-9528, doi:10.5194/acp-12-9505-2012, 2012. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Murphy, B. N., Pandis, S. N.: Simulating the formation of semivolatile primary and secondary organic aerosol in a regional chemical transport model, Environ. Sci. Technol., 43, 4722-4728, 2009. Nenes, A., Pilinis, C., and Pandis, S.: ISORROPIA: a new thermodynamic model for inorganic multicomponent atmospheric aerosols, Aquat. Geochem., 4,1998.