Supplement of Impact of halogen chemistry on air quality in coastal and continental Europe : application of CMAQ model and implication for regulation

Supplement of Impact of halogen chemistry on air quality in coastal and continental Europe: application of CMAQ model and implication for regulation Qinyi Li1, Rafael Borge2, Golam Sarwar3, David de la Paz2, Brett Gantt4, Jessica Domingo2, Carlos A. Cuevas1, and Alfonso Saiz-Lopez1* 5 1 Department of Atmospheric Chemistry and Climate, Institute of Physical Chemistry Rocasolano, CSIC, Madrid 28006, Spain 2 Environmental Modelling Laboratory, Department of Chemical & Environmental Engineering, Universidad Politécnica de Madrid (UPM), Madrid, Spain 3 National Exposure Research Laboratory, Environmental Protection Agency, Research Triangle Park, NC 27711, 10 United States 4 Office of Air Quality Planning and Standards, Environmental Protection Agency, Research Triangle Park, NC 27711, United States


Introduction
Halogen (Cl, Br, and I) species and related processes have been known to deplete stratospheric 35 ozone (O3) for several decades (Molina and Rowland, 1974;Farman et al., 1985).In the troposphere, it has only been recognized recently that halogen species affect the concentration of air pollutants, e.g., directly destroying O3 (R1), influencing the NO/NO2 ratio (R2) and the HO2/OH ratio (R3 and R4) (Saiz-Lopez and von Glasow, 2012;Simpson et al., 2015).The budgets of NOx (NO+NO2) and HOx (OH+HO2) also affect the formation of O3 (e.g., Sillman,40 1999; Li et al., 2018).Chlorine radical (Cl) initiates the oxidation of hydrocarbons (methane, CH4, and non-methane volatile organic compounds, NMVOC, R5) in a similar way to OH radical, reducing the lifetime of CH4 and NMVOC and leading to the formation of O3 in the presence of NOx (Thornton et al., 2010).
RH + Cl → HCl + RO2 (R5) The combined effect of halogen chemistry on air quality, therefore, is complicated and depends heavily on local conditions, e.g., atmospheric compositions, oxidative capacity, etc. (Sherwen et al., 2016;Muñiz-Unamunzaga et al., 2018).Evaluation of the complex role of halogen chemistry in air quality requires the employment of advanced, high-resolution chemical transport models.

55
A number of modeling studies have been conducted to investigate the impact of individual halogen species on air quality.The chemistry of chlorine, mainly that of ClNO2, has been reported to increase the oxidation capacity and the formation of O3 in recent studies (Sarwar et al., 2012(Sarwar et al., , 2014;;Li et al., 2016).Bromine and iodine (Br and I) chemistry are reported to decrease the concentration of O3 over the oceanic and terrestrial regions (Fernandez et al., 2014; 60 Saiz-Lopez et al., 2014).Only a few regional modeling studies have explored the combined influence of the halogen chemistry on air quality.The first modeling study with combined halogen (Cl, Br, and I) chemistry was conducted by Sarwar et al. (2015) who used a hemispheric version of the Community Multiscale Air Quality (CMAQ) model (Ching and Byun, 1999;Byun and Schere, 65 2006; Mathur et al., 2017) to explore the effect of bromine and iodine chemistry on tropospheric O3 over the Northern Hemisphere.Gantt et al. (2017) then utilized the CMAQ model to explore the role of halogen chemistry at a regional scale over the continental United States (US).While these studies focused on the hemispheric impact or over the continental US, Muñiz-Unamunzaga et al. ( 2018) applied the full-halogen chemistry version of CMAQ to examine the effect of the 70 halogen sources on air quality at a city scale (4 km resolution) in Los Angeles, California, US.
The regulation of air quality and the control of air pollutants emission in Europe started in the early 1970s and over forty years of effort has successfully improved air quality throughout Europe (http://ec.europa.eu/environment/air/index_en.htm).Nonetheless, poor air quality persist in major cities like Madrid, Paris, and London (EEA, 2018a).To our best knowledge, the only 75 modeling study including halogen chemistry in Europe was conducted by Sherwen et al. (2017) who used a global model, GEOS-Chem, in a regional configuration (with a grid size of 0.25°× 0.315°, ~25km) to examine the effect of halogens on air quality.Considering that the grid size has a noticeable impact on air quality model predictions (Gantt et al., 2017;Sherwen et al., 2017), it is important to conduct high-resolution simulations using regional models to examine 80 the overall effect of halogen species on air pollution over Europe and to assess potential air quality policy implications.
In this study, we use a state-of-the-art regional chemical transport model (CMAQ) with 12 km horizontal resolution, instrumented with comprehensive halogen sources and chemistry (Sarwar et al., 2015), to simulate the levels of halogen species over Europe, examine the effect on the 85 oxidation capacity and the concentration of air pollutants, and explore the potential implications for air quality policy related to NO2 and O3.

Data
The meteorological inputs for the CMAQ model were obtained from the Weather Research and 90 Forecasting model (WRF 3.7.1)(Skamarock and Klemp, 2008;Borge et al., 2008a (van Loon et al., 2007).
Biogenic emissions were estimated using the Model of Emissions of Gases and Aerosols from Nature (MEGANv2.10)(Guenther et al., 2012).All emissions were gridded to our model domain, temporally allocated and chemically speciated using the Sparse Matrix Operator Kernel 105 Emissions (SMOKE) model, version 3.6.5 (UNC, 2015;Borge et al., 2008b).
In addition, we used measurement data of NO2 and O3 from 465 background stations (traffic and industrial stations are not included) across Europe from database AirBase (public air quality database system of the European Environment Agency, 2018) to compare the results of our simulation with observations (Fig. 1).Among these stations, 340 are located in inland areas (223 110 for NO2 and 315 for O3), and 123 are located in the coastal areas (80 for NO2 and 101 for O3).The CMAQ model is widely used and includes comprehensive representations of many essential atmospheric processes.The skill of the model in reproducing observed air quality has been demonstrated in many previous studies (Foley et al., 2010;Appel et al., 2013Appel et al., , 2017;;Mathur et al., 2017), including applications over Europe (Borge et al., 2008a;Appel et al., 2012;Solazzo et al., 2017).CMAQ version 5.2 (www.epa.gov/cmaq;doi:10.5281/zenodo.1167892)containing 120 the Carbon Bond chemical mechanism with halogen chemistry was used in this study (Appel et al., 2017).The chlorine chemistry includes 26 gas-phase chemical reactions (Sarwar et al., 2012).
In addition, the heterogeneous hydrolysis of dinitrogen pentoxide (N2O5) can produce nitryl chloride (ClNO2) and nitric acid (HNO3) in the presence of particulate chloride.In the absence of particulate chloride, heterogeneous hydrolysis of N2O5 produces only HNO3.The bromine 125 chemistry contains 39 gas-phase chemical reactions and one heterogeneous reaction while the iodine chemistry contains 53 gas-phase chemical reactions (Sarwar et al., 2015).

Simulation setup
A detailed description of physics and other model options can be found in (de la Paz et al., 2016) 130 (Table S1).The CMAQ modeling domain covers the entirety of Europe (Fig. 1 diffusion processes, and the multiscale method to describe horizontal diffusion processes.Gasphase chemistry, aqueous chemistry, aerosol processes, and dry and wet deposition were also included.The Rosenbrock solver was used for gas-phase chemistry. The study was completed for the month of July 2016 with a spin-up period of 7 days.We performed three simulations to isolate the effect of halogen chemistry on air quality (in brackets 140 the name of the scenario used hereafter): (1) Base model without halogen chemistry (BASE), (2) BASE and chlorine chemistry (CHL), and (3) CHL and Br and I chemistry (HAL).
The BASE model simulation includes the Carbon Bond chemical mechanism but does not Therefore, the difference between CHL and BASE simulations represents the impact of the chlorine chemistry on air quality and the difference between HAL and BASE simulations represents the effect of halogen chemistry on air quality.

Results and Discussions
165

Evaluation of model performance
The performance of the CMAQ model in simulating air quality over Europe is evaluated using observation data collected from 465 measurement stations.We separate the stations into coastal (within 24 km from the coast) and continental stations (Fig. 1).Table 1 presents the statistics of the model performance for O3 and NO2 for BASE and HAL simulations.under-predicts O3 compared to observations both at coastal and continental stations (Table 1).
The BASE simulation under-predicts NO2 compared to observations both at coastal and continental stations (Table 1).Such an under-estimation of NO2 can occur for many reasons (Table 1).
Overall, the evaluation of the CMAQ model over Europe demonstrates that the model is capable of reproducing the levels of atmospheric chemical species and can be used to investigate the 185 impact of halogen chemistry on air quality over Europe.It also suggests that the incorporation of halogen chemistry improves the model performance for O3 concentrations by a small margin while deteriorating the model performance for NO2 by a smaller margin.Direct measurements of halogen species are very scarce and not available for the period covered in the present study (July 2016).Since a direct comparison is not possible, here we present a comparison of the simulated concentrations with observations from previous studies (Table 3), to provide an approximate assessment of the representation of halogen species in the HAL 220 simulation of the CMAQ model over Europe.Numerous ClNO2 measurements have been reported around the globe which show that ClNO2 is ubiquitous in the boundary layer with maximum values ranging from hundreds to thousands pptv in polluted coastal (Osthoff et al., 2008;Wang et al., 2016) and continental regions (Tham observed over the Dead Sea (Matveev et al., 2001;Holla et al., 2015).CMAQ is not able to reproduce such a high level of BrO due to the lower bromide content in typical ocean water (which was used in the present study for the Dead Sea) compared to the exceptionally high 245 bromide content in Dead Sea (Tas et al., 2006;Sarwar et al., 2015).
Global measurements of IO show that the IO levels observed by ground-based campaigns were generally between 0.2 and 2.4 pptv while those by ship measurement were ~3.5 pptv (Saiz-Lopez and von Glasow, 2012).Observations of IO have also been conducted in Europe.
Maximum IO levels of 4.0~50.0pptv were measured at Mace Head (Allan et al., 2000; 250 Commane et al., 2011).CMAQ predicts a value of 3.1 pptv at Mace Head while GEOS-Chem predicted a value of 0.6 pptv (Sherwen et al., 2017).In Brittany, up to 7.7~30.0pptv of IO were observed by Bitter et al. (2005) and Furneaux et al. (2010).CMAQ predicts 1.7 pptv of IO at Brittany and Sherwen et al. (2017) predicted 0.07 pptv.A maximum IO concentration of 2.0 pptv was reported in Dagebull (Peters et al., 2005), and CMAQ predicts 4.1 pptv at that site, 255 while GEOS-Chem predicted 1.8 pptv (Sherwen et al., 2017).In the BASE simulation, the Cl concentration was negligible because there was no relevant chlorine source incorporated in the CMAQ model.The CHL simulation contains the production of ClNO2 and its subsequent photolysis which increases the Cl concentration of as high as 9.3 × 310 10 -4 pptv.The HAL simulation predicted a very similar magnitude and spatial distribution of chlorine concentration.Sherwen et al. (2017) reported Cl concentrations less than 1.4 × 10 4 atom cm -3 (~5.6 × 10 -4 pptv) over Europe, which is lower than but comparable to our prediction.Hossaini et al. (2016) reported over 1.0 × 10 4 atom cm -3 (~4.0 × 10 -4 pptv) of chlorine over Asia, Europe and North America, with a maximum of 8.5 × 10 4 atom cm -3 (~3.4 × 10 -3 pptv), using a 315 global chemical transport model (TOMCAT) that incorporated chlorine sources from sea salt dechlorination, coal and biomass burning, oxidation of natural and anthropogenic chlorocarbon, and heterogeneous reactions on sea salt and sulfate aerosol.
The current study and the previous works simulated a broad range of the surface Cl concentrations although they were all within the scope of the reported observed (observation-320 based calculation) values of 10 3 to 10 5 atom cm -3 (~4.0 × 10 -5 to 4.0 × 10 -3 pptv) according to the review of Saiz-Lopez and von Glasow (2012).In light of the considerable variation of observed and model predicted Cl level, further study may be needed to comprehensively evaluate the significant role of Cl in the troposphere.The monthly average modeled NO2 and O3, two major gaseous air pollutants in Europe, and the effect of chlorine and halogen chemistry on the two regulated gaseous species were shown in Europe with an average reduction of -13.5 ppbv in the domain and a maximum of -28.9 ppbv in some locations.

Implications for policy assessment
The  (Jerrett et al., 2009;Malley et al., 2017), O3 is also known to have a negative impact on vegetation (Mills et al., 2011)  We find that halogen chemistry strongly affects ambient O3 concentration and may need to be considered in the formulation of plans and strategies for O3 non-attainment areas.We see differences between BASE and HAL simulations (over land in July 2016) as high as 12% and 36% for the number of days with daily maximum 8 h O3 over 120 µg•m -3 and the AOT40 420 respectively (Fig. S3 and Fig. S4).Furthermore, we notice strong regional differences, mainly between coastal and inland areas.The considerable effect of halogen chemistry on air quality implies the need to improve the robustness and accuracy of modeling tools to design customized policies to control O3.
In Section 3.3 and 3.4, we have also discussed the effect of halogen chemistry on the This study also demonstrates that chlorine chemistry enhances the formation of O3.The current policy is only designed to control the long-lived chlorinated species (Hossaini et al., 2015), but not the reactive chlorine species, e.g., HCl, chloride, and short-lived chlorocarbons, from the 435 coal burning, biomass burning, and industrial activities.The coal-fired power plants in EU (EEA, 2018b;Kuklinska et al., 2015) can potentially provide chlorine sources, making the implications of halogen chemistry even more relevant.

Conclusion 440
We applied the CMAQ model with comprehensive halogen chemistry (Cl, Br and I) to conduct high-resolution simulations for examining the impact of halogen chemistry on air quality over Europe.
The comparison of model results with observations from 465 monitoring sites indicates that the CMAQ model is capable of reproducing the concentrations and temporal variations of air 445 pollutants over Europe and can be employed to study the impact of halogen chemistry in Europe.
The comparison of predicted halogen species concentrations with measurements suggests that CMAQ model is able to predict observed levels of chlorine and iodine species although it underestimates bromine species.
The chlorine chemistry enhances the atmospheric oxidation capacity by significantly increasing Although the incorporation of the halogen chemistry may improve the capabilities of 3D 460 Eulerian chemical transport models, we acknowledge that large uncertainties still exist in the assessment of halogen chemistry impact due to emission inventories, model configuration (e.g., grid size), chemical mechanism, etc.Further field, laboratory, and theoretical studies are needed to constraint modeling studies for evaluating the impacts of halogen chemistry on air quality and for assessing air quality policy implications. 465
145 contain any halogen chemistry, and only the HNO3 is produced from the heterogeneous hydrolysis of N2O5.The CHL simulation contains the Carbon Bond chemical mechanism with chlorine chemistry and considers ClNO2 and HNO3 production from the heterogeneous uptake of N2O5 on the aerosol surface.The HAL simulation contains the Carbon Bond chemical mechanism with full halogen chemistry and produces ClNO2 and HNO3 from the heterogeneous 150 uptake of N2O5 on the aerosol surface.Boundary conditions for the model were derived from the hemispheric CMAQ simulations.Two simulations were conducted using the hemispheric CMAQ simulations: the first simulation used the Carbon Bond chemical mechanism and the chlorine chemistry, while the second simulation used the Carbon Bond chemical mechanism and the full halogen chemistry.Results from the 155 Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2019-171Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 23 April 2019 c Author(s) 2019.CC BY 4.0 License.hemispheric CMAQ simulation using the Carbon Bond chemical mechanism and the chlorine chemistry were used to generate boundary conditions for the BASE and CHL simulations, while results from the hemispheric CMAQ simulation using the Carbon Bond chemical mechanism and the full-halogen chemistry were used to generate boundary conditions for the HAL simulation.160

170
The BASE and HAL simulations generally reproduce the concentration levels and the temporal variations of O3 and NO2 both at coastal and continental stations.The correlation coefficients between simulations and observations(Fig S1 in supplement)  show that CMAQ satisfactorily reproduces the variation of O3 and NO2 over most of Europe especially the coastal regions (> 0.7 for O3 and > 0.5 for NO2).The BASE simulation over-predicts O3 while the HAL simulation 175 including (1) positive artifacts of NO2 monitors, (2) under-estimation of NOx in the emission inventory, and (3) rapid transformation of NO2 into HNO3 in the model compared to the real 180 atmosphere.However, model performance is reasonable as the NO2 underestimation is relatively small.The HAL simulation deteriorates NO2 comparison with observations by a small margin Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2019-171Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 23 April 2019 c Author(s) 2019.CC BY 4.0 License.

Figure 2 .
Figure 2. Monthly average ClNO2, HCl, BrO, and IO concentration in the HAL simulation.The spatial distributions of key halogen species are shown in Fig 2. The HAL simulation with full halogen chemistry simulates generally higher ClNO2 levels (with the highest average value 225et al., 2016;Thornton et al., 2010).Two campaigns have been conducted in Europe.Phillips et al. (2012) reported a maximum of 800 pptv ClNO2 in Hessen, Germany where CMAQ predicts a Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2019-171Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 23 April 2019 c Author(s) 2019.CC BY 4.0 License.concentration of 209.7 pptv.Bannan et al. (2015) observed a peak value of 724 pptv in London where CMAQ predicts a concentration of 806.3 pptv.Simulations with the GEOS-Chem model (Sherwen et al., 2017) reported maximum values of 110 pptv and 140 pptv at Hessen and 230 London, respectively.CMAQ predicts higher values of ClNO2 and are closer to the observations compared to the GEOS-Chem model, probably due to the finer grid resolution and different uptake coefficient for heterogeneous hydrolysis of N2O5 in CMAQ which facilitates the model in capturing the local formation of ClNO2.BrO measurements have been reported at ground-based sites and during the ship cruises which 235 generally demonstrate a range of 0.5 to 2.0 pptv maximum values for land measurements and 3.0 to 3.6 pptv for ship measurements (Saiz-Lopez and von Glasow, 2012).BrO observations have been reported at several coastal sites in Europe.BrO level of up to 6.5 pptv(Saiz-Lopez et al., 2004) and 7.5 pptv(Mahajan et al., 2009) were reported in Mace Head and Brittany, respectively.CMAQ predicts 1.8 pptv and 0.2 pptv at those locations, which are lower than the 240 measurements.Sherwen et al. (2017)  also predicted similar values with a maximum of 0.8 pptv in Mace Head and 0.5 pptv in Brittany.An extremely high level of BrO, ~100 pptv, was

Figure 3 .
Figure 3. Monthly average OH and HO2 concentration in the BASE simulation, and changes due to chlorine (CHL)

Fig. 3
Fig. 3 shows the monthly average concentrations of the OH and HO2 radicals predicted by the BASE simulation and the impact of chlorine chemistry (CHL-BASE), and the full halogen chemistry (HAL-BASE), on the simulated OH and HO2 levels.In the BASE simulation, the

Figure 4 .
Figure 4. Monthly average NO3 and Cl radical concentrations in the BASE simulation, and changes induced by

Fig. 4
Fig. 4 presents the monthly average prediction of NO3 and Cl radicals in the BASE scenario and the influence of chlorine (CHL-BASE) and halogen chemistry (HAL-BASE) on the levels of NO3 and Cl.The BASE simulation predicted relatively high NO3 concentrations over the Mediterranean Sea along the busy shipping tracks.Although concentrations as high as 72.4 pptv

325Figure 5 .
Figure 5. Monthly average of daily-maximum concentrations of OH and HO2 in the BASE simulation, and changes due to chlorine (CHL) and full halogen chemistry (HAL).

Figure 6 .Figure 7 .
Figure 6.Monthly average of daily-maximum concentrations of NO3 and Cl radical in the BASE simulation, and changes induced by chlorine (CHL) and full halogen chemistry (HAL).

Fig. 7 .
Fig. 7.The BASE simulation produced many hot spots of NO2 over Europe, in the vicinity of the

Figure 8 .
Figure 8. Monthly average AOT40 index in the BASE and HAL simulations, and absolute and relative changes 425partitioning of OH/HO2 and NO/NO2.The budgets of HOx and NOx are key parameters to accurately simulate the formation of O3 and its response to the reductions of the precursors, namely NOx and VOCs (e.g.,Li et al., 2018).Air quality models are predominantly used to formulate air pollution control policy by examining the responses of O3 levels to various Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2019-171Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 23 April 2019 c Author(s) 2019.CC BY 4.0 License.reduction rates of NOx and/or VOCs.These models do not include the comprehensive halogen 430 chemistry, potentially leading to unrealistic simulation of O3 concentration responsiveness to the predicted NOx and/or VOCs emission changes in Europe.
450 the level of Cl radical and increases the levels of OH, HO2, NO3, O3, and NO2.The combined halogen chemistry marginally increases the level of OH and reduces HO2, NO3, and O3.The impact of halogen chemistry on ambient concentration of NO2 is smaller but non-negligible.Halogen chemistry significantly influences the atmospheric oxidation capacity throughout the day by imposing the highest effect on Cl in the early morning, maximum effects on OH and HO2 455 Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2019-171Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 23 April 2019 c Author(s) 2019.CC BY 4.0 License.indaytime, and largest effect on NO3 at night.Halogen chemistry can have a strong influence on atmospheric composition over oceanic and coastal regions but also some noticeable impacts over continental Europe.This study highlights the potential benefit of incorporating halogen chemistry into air quality models for policy development.

Table 1 .
Statistical summary of model performance Note: MB = Mean Bias, RMSE = Root Mean Square Error, r = correlation coefficient, IOA = Index of Agreement.1903.2SimulatedhalogenspeciesAveragesurface concentrations of the inorganic halogen species predicted in the HAL simulation over the ocean are summarized in Table2.HCl is the dominant chlorine species with 195 an average level of 237.5 pptv representing over 97% to the total inorganic chlorine (Cly) while 205 Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2019-171Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 23 April 2019 c Author(s) 2019.CC BY 4.0 License.

Table 2 .
Simulated average concentrations of inorganic halogen species over the ocean

Table 3 .
The comparison of observed and simulated halogen species current air quality management in Europe has two main objectives: (1) to protect human health and (2) to protect the environment.While many plans and measures have prioritized PM 400 or NO2, policies to reduce O3 concentrations are still needed (EEA, 2018a).The WHO Air Quality Guidelines value for O3 (maximum daily 8-hour mean of 100 µg•m -3 ) was exceeded in 96% of all the reporting stations in Europe, although this is especially true for the areas near the Mediterranean Sea.According to the EEA latest report, 12% of the EU-28 urban population is exposed to O3 concentrations above the EU target value threshold (maximum daily 8-hour mean 405of 120 µg•m -3 not to be exceeded on more than 25 days/year, as set out by the Directive 2008/50/EC) in 2016.Apart from significant potential health effects Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2019-171Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 23 April 2019 c Author(s) 2019.CC BY 4.0 License.