Evaluation of Aerosol Chemical Composition Simulations by the WRF-Chem Model at the Puy de Dôme Station ( France )

The high altitude Puy de Dôme (PUY) research station, located at 1465 m a.s.l. in central France, hosts many instruments allowing continuous measurements and intensive campaigns to measure meteorological parameters, gas-phase species, aerosol and cloud properties. Aerosol chemical composition measurements provided by a compact Time-of-Flight Aerosol Mass Spectrometer (cToF-AMS) are used to evaluate the WRF-Chem model and more precisely its ability to simulate organic aerosol (OA) for a particular event in summer 2010. Using the Volatility Basis Set approach (VBS), dedicated to the formation of secondary organic aerosol (SOA), the WRF-Chem model strongly underestimates the high concentration levels of OA observed at the PUY station: 12.5 μg m were observed and only 2.6 μg m were simulated. By means of several measurements of gas-phase volatile organic compounds (VOCs), the robustness of both emissions and SOA formation processes in the WRF-Chem model was tested. The underestimation of the OA mass concentration appears to be mainly due to a misrepresentation of the oxidation rate of the organic condensable vapours (OCVs) and the SOA yields of both anthropogenic and biogenic VOCs in the VBS parameterization.


INTRODUCTION
In the atmosphere, aerosol particles have major impacts on climate, air quality and human health (IPCC, 2013).Organic aerosol (OA) represents 20 to 90% of the submicron particulate mass (Zhang et al., 2007) dependent on location, season and long-range transport conditions.OA is traditionally divided into two categories: primary organic aerosol (POA) directly emitted into the atmosphere and secondary organic aerosol (SOA) produced through a gasto-particle conversion involving Volatile Organic Compounds (VOCs) with their various volatilities.SOA represents the predominant fraction of OA (from 63 to 95%; Zhang et al., 2011).Despite the abundance of OA in the atmosphere, there are still large uncertainties in understanding their formation pathways (Hallquist et al., 2009).
Fields measurements indicate that the gaseous precursors and the primary particles emitted by biogenic sources are the main sources of OA over the European continent in summer (Gelencsér et al., 2007).Analysis of OA mass spectra provided by an Aerosol Mass Spectrometer (AMS) demonstrate that the contribution of oxygenated organic aerosol (OOA), produced through gas to particle conversion, is predominant during summer (Lanz et al., 2010).Some aircraft and ground-based measurements showed that the OA contribution to the total particulate matter is lower in the boundary layer than in the free troposphere (Crumeyrolle et al., 2013).According to the air masses transport pathways, the mass concentrations of the OA vary: the highest concentrations are measured for air masses transported over polluted areas (Freutel et al., 2013).To interpret such observations and describe the OA variations, chemistrytransport models are appropriate tools as they allow studying simultaneously the emissions, the transport and the physicalchemical transformations of gas phase species and aerosol particles.However, whatever scale is considered (regional or global), large gaps still remain between modelled and measured SOA concentrations and more particularly for summer periods (Knote et al., 2011).The underestimation of the organic aerosol concentrations is observed in most modelling studies (de Gouw, 2005;Heald et al., 2005;Jimenez et al., 2009;Hodzic et al., 2010;Yu, 2011).Several recent regional modeling studies are accounting for the semivolatile nature and chemical aging of organic compounds both in the gas phase (OCVs which are the first generation VOC oxidation products) and/or in the aerosol phase demonstrating improvements in reproducing the organic aerosol budget and its chemical resolution (Murphy and Pandis, 2009;Tsimpidi et al., 2010;Shrivastava et al., 2011;Ahmadov et al., 2012;Bergström et al., 2012;Athanasopoulou et al., 2013;Zhang et al., 2013;Fountoukis et al., 2014).Langmann et al. (2014) found that the OA mass concentrations are in the range of the measurements performed at the Puy de Dôme station by increasing the BVOCs emissions by a factor of 5 and considering that 10% of the secondary organic compounds reach a saturation vapour pressure low enough to move into the next lower volatility category.
In this paper, the Weather Research and Forecasting -Chemistry model (WRF-Chem; Grell et al., 2005) and the 1D-Volatity Basis Set (1D-VBS; Donahue et al., 2006) approach dedicated to the SOA formation and aging are used to re-evaluate measurements performed at the Puy de Dôme for the same 2010 summer period as Langmann et al. (2014).Moreover, in this study, measurements of the gaseous precursors (anthropogenic and biogenic VOCs) of SOA are available to complement AMS measurements and allow testing the uncertain parameters involved in the SOA formation in the WRF-Chem model such as emissions and SOA formation processes.

WRF-Chem Model Configuration
The version 3.4.1 of the online WRF-Chem model (http://ruc.noaa.gov/wrf/WG11/) is used in this study.The main physical and chemical parameterizations adopted in the WRF-Chem model for this study are listed in Table 1.The simulations, which last one week, are performed over two nested domains: a coarse domain located over Europe and an inner domain located over France, with spatial resolutions of 30 km and 10 km respectively.To better represent the lower layers of the atmosphere and the variations of the boundary layer height, an enhanced resolution in the lower atmosphere is used for vertical levels.
The initial and boundary meteorological conditions (Dee et al., 2011) are provided by the European Centre for Medium-range Weather Forecast (ECMWF) with a temporal resolution of 6 hours and a horizontal resolution of 0.5°.The initial and boundary conditions for gaseous species and aerosol particles concentrations come from the global model MOZART-4 (Emmons et al., 2010).They are available every 6 hours with a horizontal resolution in latitude and longitude of 1.9° per 2.5°.
Anthropogenic emissions are derived from the MACCity inventory (MACC/CityZEN EU projects; Granier et al., 2011).MACCity provides emission fluxes for gaseous (CO, SO 2 , NH 3 , NO x , VOCs) and aerosol (organic carbon (OC) and elemental carbon (EC)) species with a monthly temporal resolution and a spatial resolution of 0.5° per 0.5°.The VOCs emissions are disaggregated into the Regional Atmospheric Chemistry Mechanism (RACM; Stockwell et al., 1997;Geiger et al., 2003) model species following the aggregation procedure proposed by Middleton et al. (1990).To convert the emissions of OC into POA, a conversion factor of 1.6 is used as proposed by Bessagnet et al. (2008).Then, 20% and 80% of POA and EC are allocated respectively to the Aïtken and the accumulation modes of the aerosol lognormal distribution (Ackermann et al., 1998).
Biogenic emissions are calculated online using the Model of Emissions of Gases and Aerosol from Nature (MEGAN; Guenther et al., 2006) that uses a 1 × 1 km resolution to quantify the net emissions of isoprene and estimate the emissions of monoterpenes, oxygenated compounds, sesquiterpenes and nitrogen oxide.
Biomass burning emissions are derived from the Fire Inventory from NCAR version 1.0 (FINNv1; Wiedinmyer et al., 2011) that provides global estimates of gas (CO 2 , CO, CH 4 , VOCs, H 2 , NO x , NH 3 , SO 2 ) and aerosol particles (OC, BC, PM 2.5 , PM 10 ) emissions with a daily temporal resolution and a spatial resolution close to 1 km.

Measurements
The PUY research station, located in the middle of France (45.46°N, 2.57°E) at 1465 m a.s.l, is part of the Aerosols, Clouds, and Trace gases Research InfraStructure (ACTRIS) and Global Atmosphere Watch (GAW) networks.This station is equipped with a wide range of instruments providing information on meteorological parameters, gas-phase species concentrations and aerosol properties.It also hosts intensive  (Ahmadov et al., 2012) measurements campaigns during which further instruments are deployed at the station.Among them, the cToF-AMS (Canagaratna et al., 2007) provides detailed information on the chemical composition and mass concentration of the nonrefractory submicron particulate matter (NR-PM 1 ).VOCs species measurements performed by gas chromatographymass spectrometry (GC/MS), which is an analytical technique that combines the performance of gas chromatography, for separating compounds of a sample, and mass spectrometry for the detection and identification of compounds based on their mass/load ratio, are also available at the PUY station since 2010.
As part of a EMEP intensive measurements campaign, detailed investigations were performed at the PUY station to characterize chemical and microphysical aerosol properties for several periods by Freney et al. (2011).In this paper, the summer 2010 case (from the 23 rd to the 27 th of June 2010) has been selected because it is the longest event with very high levels in aerosol concentrations within the same polluted continental air mass.

Boundary Layer Height and Volatile Organic Compounds Concentrations
A recent study, based upon LIDAR measurements follows the seasonal variations of the boundary layer height (BLH) at the PUY station (Hervo et al., 2014).This study shows that the summit is mainly located in the free troposphere during winter and at night-time while in daytime summer it is merged into the boundary layer (BL).Further comparisons between the BLH retrieved from LIDAR profiles and the time series of BL tracers were conducted in Rose et al. (2014), indicating that there were little forced convection effects from the PUY on the BL dynamics, likely due to its relatively isolated peak topographic configuration.
WRF-Chem model is able to reproduce the general diurnal evolution of the BLH observed by Hervo et al. (2014) (red line in Fig. 1).At the Puy de Dôme the surface elevation calculated by the model is 799 m (dotted black line in Fig. 1) instead of 1465 m (the real altitude of the Puy de Dôme (black line in Fig. 1)).This is due to the resolution of the topography in the model.For this reason, according to the model, the summit would be always located in the BL independently of the season.Instead of discussing the results for the modeled altitude of the Puy de Dôme, results are discussed for the real altitude of the station, at 1465 m.It is worth mentioning that at 1465 m, meteorological parameters such as temperature and humidity reflect best the measurements performed at the station (Fig. S1).
During the 2010 campaign, three biogenic VOCs (BVOCs) were measured at the PUY station: α-pinene, β-pinene and limonene.In the model, "API" species groups together αpinene with -pinene and other cyclic terpenes with one double bond, and "LIM" species groups together limonene with d-limonene and other cyclic diene-terpenes.After their oxidation, these BVOCs can contribute to the formation of SOA.Their SOA yields range from 20% to 60% according to their initial concentrations, the NO x concentrations, the type of oxidants, the temperature, the relative humidity and the radiation.
Fig. 1 shows the temporal evolution of BVOCs and AVOCs simulated and observed but also the variations of the BLH simulated by the WRF-Chem model.Despite an underestimation of the concentration levels (by factors 2 and 10 respectively for pinenes and limonene), the model well captures the diurnal variation of pinene (Fig. 1(A)) and limonene (Fig. 1(B)) with lower concentrations levels during nighttime, when the PUY station is located in the free troposphere, than during daytime, when the PUY station is located in the boundary layer.For the AVOCs (alkanes: Fig. 1(C), xylene: Fig. 1(D) and toluene: Fig. 1(E)), the underestimation of the concentration levels is strongest than for BVOCs.The evolution of the concentrations of AVOCs seems to be correlated with the time evolution of the BLH.Indeed, the concentrations of these species increase as soon as the Puy de Dôme enters in the boundary layer and then decrease when the summit leaves the boundary layer.
There is some model underestimation for nitrate and ammonium, probably due to some residual sources caused by agricultural activities that are not well considered in the model.Analysis of aerosol chemistry measurements in a number of stations all over Europe has shown that organic nitrates can represent 34% to 44% of the measured aerosol nitrate and are found at all urban and rural sites (Kiendler-Scharr et al., 2015).This significant contribution to particulate nitrate is not considered in the model and could explain why nitrate is underestimated.

Sensitivity Tests on Parameters Controlling OA Formation and Evolution
Several parameters are sources of uncertainties in the representation of OA formation in the model: emissions of gas-phase precursors of SOA but also parameters such as SOA yields and oxidation rates of OCVs that control the SOA formation.
As already mentioned, the model underestimates the concentrations of BVOCs observed by at least a factor 2. The first test consists of doubling the BVOCs emissions that leads, as shown in Table 2, to an increase of the OA mass concentration by 18% (from 2.0 µg m -3 for the baseline simulation to 2.4 µg m -3 for the sensitivity test).The comparisons between AVOCs measured and simulation results highlighted a strong underestimation of HC8, TOL and XYL species by the model.As observations are not available for the others AVOCs, it is difficult to determinate an accurate correction factor.So, the emissions of the AVOCs have been increased by a factor of 2, as has been done for the BVOCs emissions.The average OA mass concentration slightly increases by doubling the amount of AVOCs emissions: from 2.0 to 2.5 µg m -3 (+ 25%).So, the impact of increasing the AVOCs emissions is equivalent to the one of increasing the BVOCs emissions.
In the model, the SOA yields defined in the VBS parameterization (Ahmadov et al., 2012) and the oxidation rate of OCVs by HO • are the two main parameters that drive the OA formation.The SOA yields, i.e., the mass of SOA produced per mass of reacted precursor, depend on both the organic precursor molecules and on the oxidants initiating the degradation.SOA yields in the model (Table S1) are based on smog chamber studies (Murphy and Pandis, 2009), where several single VOCs are oxidized under specific conditions (temperature, relative humidity, NO x concentrations, …).The individual yields of each VOC are summed to model SOA from multiple species.However, due to the myriad of VOCs present in the atmosphere, it is not realistic to provide SOA yields for all these species through experimental studies and the extrapolation of chamber-derived SOA yields to atmospheric conditions is an important source of error (Hao et al., 2011).Specifically, Zhang et al. (2014) demonstrated that vapor wall loss could lead to an underestimation of SOA yields by as much as a factor of 4. Due to these potential underestimations on the SOA yields and in order to favour SOA formation, the yields of both AVOCs and BVOCs are increased by a factor 2 (however, yields values are at most equal to 100%).When doubling the SOA yields of AVOCs, the OA mass concentration is increased from 2.0 to 3.3 µg m -3 (+ 62%) while it is only increased from 2.0 to 2.8 µg m -3 (+ 41%) by modifying the SOA yields of BVOCs (Table 2).
In the WRF-Chem model, to implement the oxidation reaction between OCVs and the hydroxyl radical HO  , an oxidation rate of 1.0 × 10 -11 cm 3 molecule -1 s -1 is used (Ahmadov et al., 2012).However, the most widely used oxidation rate with the VBS approach is 4.0 × 10 -11 cm 3 molecule -1 s -1 (Bergström et al., 2012).Applying this last oxidation rate in the WRF-Chem model to anthropogenic and biogenic OCVs improves the results: the OA mass concentration is increased by around 30% (from 2.0 to 2.6 µg m -3 ) in both cases (Table 2).
Thus, the SOA yields of the VOCs and the oxidation rate of the OCVs are very sensitive parameters that drive the transfer of the VOCs from the gas to the particulate phase.
To look into more details, Fig. 3 shows the time evolution of OA mass concentrations over the time period where VOCs measurements were available.This plot can be directly compared to the results obtained by Langmann et al. (2014) with their regional REMOTE model.Precipitation was not measured in the Puy de Dôme after 23 June 2010, whereas models determine amount of precipitation on 26 June for the WRF-Chem model and for 24 and 28/29 June for REMOTE model.Similar to the study of Langmann et al. (2014), a reduction of modelled OA mass concentration is observed after June 26 th in connection with a convective precipitation event.Underestimation of OA mass concentration in the model is probably due to fast convective transport that washed away chemical compounds.Both studies come to the same conclusions underlining the importance of aging processes in the formation of OA even if very different approaches for SOA formation (mixed thermodynamickinetic approach vs. 1D-VBS approach) are used in REMOTE and WRF-Chem model respectively.They also show that emission data on SOA precursors are missing in inventories.However, we did not increase the biogenic emissions as much as in Langmann et al. (2014), since we use a factor 2 increase instead of 5 and apply it to both biogenic and anthropogenic precursors of SOA.In our case, measured anthropogenic precursors were much more underestimated by the model than the biogenic ones.Only a factor of 2 difference was observed for pinenes between measurements and model results.Also, WRF-Chem model is able to reproduce the observed diurnal cycle for OA mass concentrations while this was not clearly reproduced by Langmann et al. (2014) with their global model.This brings two remarks, first the tests on oxidation rates of OCVs mainly by OH radical seem to stress contrasts between night and day, then using a finer resolution model allows for better comparisons with observations.

CONCLUSIONS
This paper presents a comparison between AMS measurements performed at the PUY station and WRF-Chem model outputs for a one week period in summer 2010.During this period, several measurements of VOC precursors of SOA were available allowing us to examine the sources of the measured SOA (anthropogenic vs. biogenic ones) and provided us with a good opportunity to target individual parameters implicated in the SOA formation and aging with a 10 km model resolution.The differences between modeled and measured OA mass concentrations are less important when adjusting the VOCs emissions in the model with some available measurements of VOCs.The model underestimates the unexpected high OA concentration levels observed (11.1 µg m -3 ) and the VOCs concentrations are much below those observed, especially for AVOCs such as alkanes and aromatic compounds.The model underestimation appears to be mainly due to a lack in SOA formation processes.Indeed, the OA mass concentration increases by 6.6 µg m -3 by doubling the SOA yields of both anthropogenic and biogenic VOCs and by modifying the traditional oxidation rate of OCVs from 1.0 × 10 -11 to 4.0 × 10 -11 cm 3 molecule -1 s -1 whereas an increase of only 0.8 µg m -3 is noticed by doubling VOCs emissions.So, the SOA yields and aging processes of OCVs are highly important in the SOA formation as already noticed by Langmann et al. (2014) for the same set of observations in June 2010 at the Puy de Dôme.Contrary to these authors, WRF-Chem model with a higher resolution (10 km instead of 55 km) was able to reproduce the observed diurnal cycle of OA.More VOCs and OA concentrations measurements in altitude sites are needed to improve simulations of gases and aerosols in the boundary layer and in the free troposphere.

Fig. 2 .
Fig. 2. Temporal evolution of the aerosol mass concentrations (µg m -3 ) measured at the PUY station (a) and simulated by the WRF-Chem model for the baseline simulation (b).

Fig. 3 .
Fig. 3. Time evolution of OA mass concentration observed (in green) and simulated by the WRF-Chem model at the PUY for the baseline simulation (in red) and for sensitivity tests (with blue lines): test modifying the emissions, the SOA yields of VOCs/oxidation rate of OCVs and test with all effects cumulated.

Table 2 .
OA mass concentration measured at the PUY station and modeled by the WRF-Chem model for the baseline simulation and sensitivity tests: emissions, SOA yields of VOCs/oxidation rate of OCVs and all effects.Percentages indicate the OA mass increase between baseline and test simulations.