Composited analyses of the chemical and physical characteristics of co-polluted days by ozone and PM 2 . 5 over 2013–2020 in the Beijing–Tianjin–Hebei region

. The co-polluted days by ozone (O 3 ) and PM 2 . 5 (particulate matter with an aerodynamic equivalent diameter of 2.5 µm or less) (O 3 –PM 2 . 5 PDs) were frequently observed in the Beijing–Tianjin–Hebei (BTH) region in warm seasons (April–October) of 2013–2020. We applied the 3-D global chemical transport model (GEOS-Chem) to investigate the chemical and physical characteristics of O 3 –PM 2 . 5 PDs by composited analyses of such days that were captured by both the observations and the model. Model results showed that, when O 3 –PM 2 . 5 PDs occurred, the concentrations of hydroxyl radical and total oxidant, sulfur oxidation ratio, and nitrogen oxidation ratio were all high, and the concentrations of sulfate at the surface were the highest among all pollution types. We also found unique features in vertical distributions of aerosols during O 3 –PM 2 . 5 PDs; concentrations of PM 2 . 5 decreased with altitude near the surface but remained stable at 975–819 hPa. Process analyses showed that secondary aerosols (nitrate, ammonium, and sulfate) had strong chemical productions at 913–819 hPa, which were then transported downward, resulting in the quite uniform vertical proﬁles at 975–819 hPa on O 3 –PM 2 . 5 PDs. The weather patterns for O 3 –PM 2 . 5 PDs were characterized by anomalous high-pressure system at 500 hPa as well as strong southerlies and high RH at 850 hPa. The latter resulted in the strong chemical productions around 850 hPa on O 3 –PM 2 . 5 PDs. The physical and chemical characteristics of O 3 –PM 2 . 5 PDs are quite different from those of polluted days by either O 3 alone or PM 2 . 5 alone and have important implications for air quality management


Introduction
Surface ozone (O 3 ) and PM 2.5 (particulate matter with an aerodynamic equivalent diameter of 2.5 µm or less) are important air pollutants in the atmosphere that have harmful effects on public health (Gao and Ji, 2018;Jiang et al., 2019), ecosystems (Ren et al., 2011;Yue et al., 2017), and crops (Wang et al., 2005(Wang et al., , 2007)).Surface O 3 is a secondary pollutant produced by photochemical oxidation of volatile organic compounds (VOCs) and nitrogen oxides (NO x ≡ NO + NO 2 ) in the presence of intense ultraviolet light, and the major PM 2.5 components (nitrate, NO − 3 ; am-monium, NH + 4 ; sulfate, SO 2− 4 ; black carbon, BC; and organic carbon, OC) are mainly caused by anthropogenic emissions of aerosols and aerosol precursors.Although surface O 3 and PM 2.5 have different formation mechanisms, they are coupled through the common precursors (NO x and VOCs) and photochemical reactions (Chu et al., 2020).Since 2013, stringent clean air actions have been implemented to improve air quality in China (Chinese State Council, 2013, 2018).However, O 3 concentrations increased unexpectedly, while PM 2.5 concentrations decreased drastically in the past years (M.Li et al., 2019).The co-polluted days by O 3 and PM 2.5 (concentrations of both O 3 and PM 2.5 exceed the national air qual-ity standards on the same day, hereafter referred to as O 3 -PM 2.5 PDs) were also reported (Dai et al., 2021).Therefore, it is fundamental to examine the chemical and physical characteristics of O 3 -PM 2.5 PDs.
The Beijing-Tianjin-Hebei (BTH) region is the most populated region in northern China.In the past few years, concentrations of O 3 and PM 2.5 in BTH were among the highest in China.The observations from China National Environmental Monitoring Center (CNEMC) showed that the mean and maximum MDA8 (daily maximum 8 h average) O 3 in North China in summer of 2019 were 83 and 129 ppb, respectively, and the summer mean MDA8 O 3 increased with a trend of 3.3 ppb a −1 over 2013-2019 (Li et al., 2020).Gong et al. (2020) reported that O 3 -polluted days (i.e., MDA8 O 3 concentration exceeds 80 ppb) in May-July in BTH increased from 35 d in the year 2014 to 56 d in 2018.As for observed PM 2.5 , the concentration averaged over BTH had a decreasing trend of 10 µg m −3 yr −1 over 2013-2019, and the mean value was 79 ± 17 µg m −3 over these years (Li et al., 2020).BTH also had the highest frequency and intensity of severe-haze-pollution days (i.e., days with daily mean PM 2.5 concentration exceeding 150 µg m −3 ) in China from 2013 to 2017, with an observed mean frequency of 21.2 d yr −1 and an observed mean intensity of 231.6 µg m −3 (Dang and Liao, 2019).
The interactions between O 3 and PM 2.5 have been reported in previous studies.Zhu et al. (2019) examined the spatiotemporal characteristics of the correlations between observed O 3 and PM 2.5 at 1497 sites in China for 2016 and found that O 3 -PM 2.5 had the highest positive correlations (correlation coefficients > +0.7) in July in southern China and the largest negative correlations (r values < −0.5) during January in northern China.K. Li et al. (2019) used the GEOS-Chem model to analyze the O 3 -PM 2.5 relationship in northern China and found that O 3 production was suppressed under high-PM 2.5 conditions (PM 2.5 concentrations > 60 µg m −3 ) because of the reactive uptake of hydrogen oxide radicals (HO x ) by aerosol particles.Chu et al. (2020) analyzed the observed daily PM 2.5 and O 3 concentrations in 114 cities in China during the years of 2013-2018 and found that the correlations between O 3 and PM 2.5 tended to change from negative in 2013 to positive in 2018 in China as air quality improved.
Few previous studies have examined the co-occurrence of O 3 and PM 2.5 pollution (MDA8 O 3 > 80 ppb and PM 2.5 > 75 µg m −3 ).Zong et al. (2021) used the obliquely rotated principal component analysis in the T-mode (T-PCA) method to identify the synoptic weather pattern associated with O 3 -PM 2.5 PDs in eastern China during summer of 2015-2018 and found that O 3 -PM 2.5 PDs were associated with a stable western Pacific subtropical high ridge, which brought warm and moist airflow from the East China Sea to eastern China to promote hygroscopic growth of fine particulate matter in BTH and the northern Yangtze River Delta (YRD).Dai et al. (2021) analyzed O 3 -PM 2.5 PDs in the YRD for April-October of 2013-2019 by using observations and reported that the co-polluted days occurred mainly in April (29.6 % of co-polluted days occurred in April), May (23.0 %), June (19.5%), and October (10.8 %) under meteorological conditions of higher relative humidity, higher surface air temperature, and lower wind speed relative to the days with O 3 pollution alone.Qin et al. (2021) investigated O 3 -PM 2.5 PDs by using the hourly observed concentrations of water-soluble ions, OC, and elemental carbon (EC) in 2019 in cities of Nanjing and Changzhou.They found that inorganic aerosols mainly existed as NH 4 NO 3 , and the correlation coefficients between the secondary components NO − 3 , NH + 4 , and SO 2− 4 were relatively high during O 3 -PM 2.5 PDs in 2019, indicating a significant formation of secondary inorganic aerosols.Although these studies have discussed the meteorological conditions and some chemical characteristics of O 3 -PM 2.5 PDs, the understanding of O 3 -PM 2.5 PDs was quite limited because of the limited observations of chemical species involved.
In this work, we take advantage of the comprehensive chemical mechanism of the global chemical transport model to have better understanding of O 3 -PM 2.5 PDs.We apply the 3-D global chemical transport model (GEOS-Chem) to simulate O 3 -PM 2.5 PDs in BTH in the years of 2013-2020 and investigate the chemical and physical characteristics of O 3 -PM 2.5 PDs by composited analyses of such days that are captured by both the observations and the model.The objectives of this study are (1) to examine the underlying chemical mechanisms for O 3 -PM 2.5 PDs in BTH for warm seasons (April-October) of 2013-2020 by comparing O 3 -PM 2.5 PDs with polluted days by O 3 alone or by PM 2.5 alone and (2) to identify the weather patterns that are associated with O 3 -PM 2.5 PDs in BTH.The observations, the reanalyzed meteorological data, the GEOS-Chem model, and the process analysis are described in Sect. 2. The observed O 3 -PM 2.5 PDs are presented in Sect.3.1.The evaluation of simulated concentrations of O 3 and PM 2.5 as well as the simulated pollution days by O 3 and/or PM 2.5 are shown in Sect.3.2.The underlying mechanisms of O 3 -PM 2.5 PDs are shown in Sect. 3.3. In Sect. 3.4, the meteorological conditions for the co-occurrence of O 3 and PM 2.5 pollution are investigated.The conclusions are presented in Sect. 4.

Methods
2.1 Observed O 3 and PM 2.5 concentrations Hourly concentrations of PM 2.5 and O 3 in China over the years of 2013-2020 were taken from the public website of CNEMC (https://air.cnemc.cn:18007/,last access: 17 November 2022).To ensure data quality, the daily mean PM 2.5 concentration was calculated when there were valid data for more than 20 h during that day, and the MDA8 O 3 concentration was calculated when there were valid data for at least 6 h for each 8 h.For the calculation of monthly and annual mean concentrations, the number of days with valid concentrations had to be more than 15 in each month.The spatial distribution of the 79 valid sites within BTH (37-41 • N, 114-118 • E; the black rectangle) is shown in Fig. 1.For model evaluation, the observed concentrations were averaged over sites within each of the 0.5 • latitude × 0.625 According to the National Ambient Air Quality Standard of China (GB3095-2012), O 3 (PM 2.5 ) concentration exceeds the national air quality standard if the MDA8 O 3 (daily mean PM 2.5 ) concentration is higher than 160 µg m −3 (75 µg m −3 ).In this study, we define O 3 polluted days (hereafter called "O 3 PDs") for days with MDA8 O 3 concentration > 160 µg m −3 , PM 2.5 polluted days (hereafter called "PM 2.5 PDs") as days with daily mean PM 2.5 concentration > 75 µg m −3 , and the co-pollution days by O 3 and PM 2.5 (O 3 -PM 2.5 PDs) as days with daily MDA8 O 3 concentration > 160 µg m −3 as well as a daily mean PM 2.5 concentration > 75 µg m −3 .

Reanalyzed meteorological fields
Meteorological fields were obtained from MERRA-2, which were generated by the NASA Global Modeling and Assimilation Office (GMAO).The MERRA-2 data have a horizontal resolution of 0.5 • latitude × 0.625 • longitude and 72 vertical layers (Molod et al., 2015).To analyze the meteorological conditions for O 3 -PM 2.5 PDs, vertical pressure velocity (OMEGA), planetary boundary layer height (PBLH), temperature (T ), relative humidity (RH), and surface incoming shortwave flux (SWGDN) are used.Note that the temporal resolution for PBLH, T , and SWGDN is 1 h, and that for OMEGA and RH is 3 h.Daily mean geopotential heights at 850 and 500 hPa from the National Center for Environmental Prediction (NCEP) and National Center for Atmospheric Research (NCAR) global reanalysis with a resolution of 2.5 • latitude by 2.5 • longitude are also utilized in this study.

GEOS-Chem model
We simulated O 3 and PM 2.5 using the nested version of the 3-D global chemical transport model (GEOS-Chem, version 11-01) driven by the MERRA-2 meteorological data.The nested domain was set over Asia (11 • S-55 • N, 60-150 • E) with a horizontal resolution of 0.5 • latitude × 0.625 • longitude, and the chemical boundary conditions were provided by the global GEOS-Chem simulation with 2.5 • latitude × 2.5 • longitude horizontal resolution.
The GEOS-Chem model includes fully coupled O 3 -NO xhydrocarbon and an aerosol chemistry mechanism (Bey et al., 2001;Pye et al., 2009) to simulate aerosols including SO 2− 4 (Park et al., 2004), NO − 3 (Pye et al., 2009), NH + 4 , BC and OC (Park et al., 2003), mineral dust (Fairlie et al., 2007), and sea salt (Alexander et al., 2005) as well as the gas-phase pollutants such as NO x and O 3 .Over the Asian domain, the anthropogenic emissions of OC, BC, carbon monoxide (CO), sulfur dioxide (SO 2 ), NO x , ammonia (NH 3 ), and VOCs were obtained from the Multi-resolution Emission Inventory for China (MEIC), which includes emissions from the industry, power, residential, and transportation sectors for the years of 2014-2017 (Li et al., 2017;Zheng et al., 2018), 2019, and 2020(Zheng et al., 2021).Emissions in 2018 were obtained by the interpolation of those in 2017 and 2019 for each grid due to the lack of publicly accessible emission inventories for that year.The biogenic emissions in GEOS-Chem are simulated using MEGAN v2.1 (Guenther et al., 2012).
The hourly O 3 and PM 2.5 concentrations for the years of 2013-2020 were simulated by the GEOS-Chem model and were driven by MERRA-2 meteorological fields.The model was spun up for 6 months before the integration over the studied time period.

Process analysis
Process analysis (PA) was applied to identify the relative importance of atmospheric processes on O 3 -PM 2.5 PDs.PA has been widely used in previous studies to examine the key processes contributing to air pollution episodes (Gonçalves et al., 2009;Dang and Liao, 2019;Gong and Liao, 2019)  ( Mu and Liao, 2014;Lou et al., 2015).Five major processes that influence O 3 and PM 2.5 concentrations were diagnosed at every time step, including net chemical production, dry deposition, horizontal advection, vertical advection, and diffusion, for the regional pollution days (days with more than half of the sites in BTH experiencing pollution).We carried out PA for O 3 SPDs (excluding O 3 -PM 2.5 PDs from O 3 PDs), PM 2.5 SPDs (excluding O 3 -PM 2.5 PDs from PM 2.5 PDs), and O 3 -PM 2.5 PDs over BTH.

Results
3.1 Observed polluted days by O 3 and PM 2.5 Figure 1a shows the spatial distributions of observed numbers of O 3 PDs, PM 2.5 PDs, and O 3 -PM 2.5 PDs summed over the warm seasons (April-October) of 2013-2020.The spatial distributions of polluted days in each year are shown in Fig. S1 in the Supplement.The numbers of O 3 PDs, PM 2.5 PDs, and O 3 -PM 2.5 PDs were high in BTH, which were, respectively, 524.3 344.6, and 128.1 d from observations, as the values were averaged over all sites in BTH.The high numbers of O 3 PDs, PM 2.5 PDs, and O 3 -PM 2.5 PDs in BTH were associated with the highest anthropogenic emissions (NO x and non-methane volatile organic compounds, NMVOCs) in this region (Dang et al., 2021).
Figure 1b shows the numbers of days averaged over all sites in BTH for non-polluted days (NPDs; MDA8 O 3 < 80 ppb and PM 2.5 < 75 µg m −3 ), O 3 PDs, O 3 -PM 2.5 PDs, and PM 2.5 PDs in each month of warm seasons Figure 2a shows the linear trends of observed O 3 PDs, PM 2.5 PDs, and O 3 -PM 2.5 PDs in warm seasons of 2013-2020 averaged over BTH.O 3 PDs showed an upward trend of 7.9 d yr −1 from 2013 to 2020.However, the numbers of PM 2.5 PDs and O 3 -PM 2.5 PDs decreased over 2013-2020, with linear trends of −11.2 and −3.4 d yr −1 , respectively.Figure 2b shows the changes in percentage of O 3 -PM 2.5 PDs on PM 2.5 PDs from 2013 to 2020 for each month.It should be noted that, when PM 2.5 PDs occurred, the proportions of O 3 -PM 2.5 PDs had an upward trend from 2013 to 2020.In May, June, August, and September of 2020, the proportions of O 3 -PM 2.5 PDs on PM 2.5 PDs reached 100 %, indicating that PM 2.5 pollution was accompanied by O 3 pollution in recent years.) in eight warm seasons have a normalized mean bias (NMB) of 7.9 % (10.6 %).The model generally captures the daily variations (peaks and troughs) in the observed MDA8 O 3 and PM 2.5 concentrations, with R values of 0.80 and 0.72, respectively.It should be noted that mineral dust and sea-salt aerosols were not considered in this study because they are not the major aerosol components in China, and the concentrations are generally low based on previous measurements (Xuan et al., 2000;Ye et al., 2003;Duan et al., 2006;Zhao et al., 2013).However, excluding dust and sea salt may lead to low biases in simulated PM 2.5 concentrations.
Due to the lack of publicly accessible long-term observations of PM 2.5 components in China, we compared the simulated SO 2 and NO 2 (precursors for SO 2− 4 and NO − 3 ) with observations from CNEMC in Fig. S2.The simulated daily mean concentrations of NO 2 (SO 2 ) agree well with the observations from CNEMC, with R of 0.82 (0.78) and NMB of −14.9 % (9.3 %).PDs because of the model's deficiency in capturing the peak concentrations of air pollutants.Such deficiency was also reported in previous studies that used the GEOS-Chem model or the weather Research and Forecasting with Chemistry (WFR-chem) model (Zhang and Wang, 2016;Ni et al., 2018;Gong and Liao, 2019;Dang and Liao, 2019).Therefore, to identify O 3 PDs, PM 2.5 PDs, and O 3 -PM 2.5 PDs using model results, we utilized lower thresholds by considering the NMBs of simulated MDA8 O 3 and PM 2.5 concentrations in each of the 18 grids of BTH.Taking the grid of Beijing as an example, simulated MDA8 O 3 and PM 2.5 had NMBs of −22.0 % and −26.9 %, respectively, as the simulated concentrations were compared with observations for days with observed concentrations higher than the national air quality standards over the warm seasons of 2013-2020.We then adjusted the threshold of O 3 PDs in this grid to be 62.4 ppb (80 ppb × 78 %) and that of PM 2.5 PDs to be 54.8 µg m −3 (75 µg m −3 × 73.1 %).These adjusted thresholds were also used to identify O 3 -PM 2.5 PDs.Such an approach was also used in previous studies to better capture the pollution events based on the simulations (Dang and Liao, 2019;Gong and Liao, 2019).With the adjusted thresholds, 56 %-93 % of the observed O 3 PDs, PM 2.5 PDs, and O 3 -PM 2.5 PDs can be captured by the model (Fig. S3e).

Simulated
where O i and M i are the observed and simulated concentrations, respectively; i refers to the ith day, and N is the total number of days.
pose of obtaining the characteristics of different polluted days.Figure 4a and b 2017).The concentrations of OH were the highest on O 3 SPDs, with an averaged value of 2.8 × 10 6 molec.cm −3 , followed by that on O 3 -PM 2.5 PDs (2.0 × 10 6 molec.cm −3 ) and on PM 2.5 SPDs (1.0 × 10 6 molec.cm −3 ).Due to the lack of publicly accessible observations of OH in BTH, we compare the simulated OH concentrations with observations reported in the literature (Table S1).The simulated OH concentrations agree closely with the observed values.In Wangdu of BTH, while the observed daily maximum OH concentrations in summer of 2014 were in the range of 5-15 × 10 6 molec.cm −3 (Tan et al., 2017), the simulated OH concentrations in the same time period in this work were 3.7-9.5 × 10 6 molec.cm −3 .In Beijing in summer of 2017, the observed daily mean OH concentration was 5.8 × 10 6 molec.cm −3 (Woodward-Massey et al., 2020), and our simulated value was 2.4 × 10 6 molec.cm −3 .The mean values of O x were, respectively, 178.7, 118.1, and 184.1 µg m −3 on O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs, indicating that the atmospheric oxidation capacity was strong on O 3 -PM 2.5 PDs, which favored the production of secondary components of PM 2.5 .Figure 5 also shows sulfur oxidation ratio (SOR; n-SO 2− 4 / (n-SO 2− 4 + n-SO 2 ), where n-SO 2− 4 and n-SO 2 are the concentrations of SO 2− 4 and SO 2 , respectively) and nitrogen oxidation ratio (NOR; n-NO − 3 / (n-NO − 3 + n-NO 2 ), where n-NO − 3 and n-NO 2 are the concentrations of NO − 3 and NO 2 , respectively).SOR and NOR are measures of the conversion degrees of sulfur and nitrogen, respectively (Zhu et al., 2019).On O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs, the averaged values of SOR were 50.0 %, 36.7 %, and 49.7 %, and those of NOR were 55.4 %, 70.0 %, and 70.2 %, respectively.The high SOR and NOR on O 3 -PM 2.5 PDs indicated the strong formation of SO 2− 4 and NO − 3 that was promoted by high atmospheric oxidation capacity.The monthly variations in OH, O x , and SOR were similar (Fig. 5), with the highest values in summer, owing to the high temperature that promoted high concentrations of oxidants and SOR.It is interesting that SOR and O x values were higher on O 3 -PM 2.5 PDs than on O 3 SPDs or on PM 2.5 SPDs during May-August.Similarly, NOR values were higher on O 3 -PM 2.5 PDs than on O 3 SPDs or on PM 2.5 SPDs in May and July-September.Overall, the O 3 -PM 2.5 PDs occurred with high levels of atmospheric oxidants, SOR, and NOR, leading to combined increases in O 3 and PM 2.5 concentrations.

Surface-layer concentrations of PM 2.5
components on O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs The simulated concentrations of PM 2.5 components (NO − 3 , NH + 4 , SO 2− 4 , BC, and OC, averaged over 18 grids of BTH, are shown in Fig. 6 for O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs in the warm seasons of 2013-2020 that were observed and also captured by the model.While the mean concentrations of NO − 3 , NH + 4 , BC, and OC were all the highest on PM 2.5 SPDs, SO 2− 4 concentration was the highest on O 3 -PM 2.5 PDs.On O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs, the mean concentrations of SO 2− 4 were 6.2, 9.4, and 11.97 µg m −3 , respectively, and the percentages of SO 2− 4 in PM 2.5 were 14.9 %, 9.0 %, and 15.0 %, respectively.In July and August, the concentrations of SO 2− 4 and MDA8 O 3 on O 3 -PM 2.5 PDs were the highest compared with those on O 3 SPDs and PM 2.5 SPDs (Fig. S5).
Figure 7 S6).Compared to O 3 SPDs and O 3 -PM 2.5 PDs, the PBLH of PM 2.5 SPDs was lower and uplifted 1 h later, which was more favorable for the accumulation of aerosols.During the daytime, PBLH on O 3 -PM 2.5 PDs was between O 3 SPDs and PM 2.5 SPDs.
It is worth noting that the diurnal variations in SO    8a).In these two cases, concentrations of O 3 increased from the surface to about 975 hPa, remained high between 975 and 819 hPa and decreased with altitude between 819 and 663 hPa.Although the magnitudes of O 3 were close at the surface (61.9 ppb on O 3 -PM 2.5 PDs and 58.1 ppb on O 3 SPDs), the concentration of O 3 averaged over 975 and 819 hPa was 10.4 % higher on O 3 -PM 2.5 PDs than on O 3 SPDs, which was a very unique feature of O 3 -PM 2.5 PDs.For the case of PM 2.5 SPDs, the concentrations of O 3 were the lowest among the three cases and increased gently with altitude above 975 hPa.
Figure 8b shows the vertical distributions of PM 2.5 components.In all the cases, PM 2.5 concentrations were the highest at the surface, and decreased with altitude from the surface to 975 hPa.However, concentrations of PM 2.5 were quite stable between 975 and 819 hPa for O 3 SPDs (about 36.4 µg m −3 ) and O 3 -PM 2.5 PDs (about 58.1 µg m −3 ), corresponding to the stable O 3 levels at these altitudes in these two cases (Fig. 8a).For PM 2.5 SPDs, while PM 2.5 concentration at the surface was the highest among the three cases, it decreased rapidly between 975 and 819 hPa.The averaged PM 2.5 concentration between 975 and 819 hPa was 52.4 µg m −3 on PM 2.5 SPDs, which was lower than that on O 3 -PM 2.5 PDs.
To further investigate the differences in vertical profiles of NO − 3 , NH + 4 , SO 2− 4 , BC, OC, and PM 2.5 on O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs, the ratios of concentration at 975 hPa to that at the surface as well as the concentration at 819 hPa to that at 975 hPa are shown in Table 1.The concentration of PM 2.5 decreased largely, with the ratio of PM 2.5(975 hPa) / PM 2.5(1005 hPa) of 0.78 on O 3 -PM 2.5 PDs and of 0.74 on PM 2.5 SPDs.For each of the PM 2.5 components, the ratios near the surface (from 1005 to 975 hPa, shaded gray area in Fig. 8) were close in the three types of pollution.While the ratios of NO − 3 , NH + 4 , BC, and OC were in the range of 0.65-0.80, the ratios of SO 2− 4 were about 0.93-0.98,indicating that SO 2− 4 concentrations were quite uniform from the surface to 975 hPa in all three types of pollution.
In the upper layers (975-819 hPa, red rectangle in Fig. 8), the changes in concentrations of pollutants with altitude on PM 2.5 SPDs were quite different from those on O 3 -PM 2.5 PDs and O 3 SPDs.The decline in PM 2.5 from 975 to 819 hPa was slow on O 3 -PM 2.5 PDs (PM 2.5(819 hPa) / PM 2.5(975 hPa) = 0.89) and O 3 SPDs (0.86) and fast on PM 2.5 SPDs (0.67).Considering that the variation in BC with altitude was mainly driven by meteorology (Sun et al., 2020)  Table 1.The ratios at 975 and 1005 hPa (shaded gray area in Fig. 8) and at 819 and 975 hPa (red frame in Fig. 8) of NO − 3 , NH + 4 , SO 2− 4 , BC, OC, and PM 2.5 on O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs in BTH region.that differed significantly from BC indicated the influences of chemical processes.On PM 2.5 SPDs, NO − 3 , NH + 4 , and OC had about the same ratio as BC (0.64) (with large decreases with height), except for SO 2− 4 concentration, which had a ratio of 0.81.On O 3 -PM 2.5 PDs, the ratios of NO − 3 , NH + 4 , and SO 2− 4 were 0.94, 0.91, and 0.87, respectively, which were much higher than the value of BC (0.79), indicating NO − 3 , NH + 4 , and SO 2− 4 were quite uniform in the layers of 975-819 hPa with the influence of chemical processes, which is discussed further in Sect. 3.3.4 below. 3.3.4Process analyses for O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs The process analysis (PA) is applied to identify the relative importance of atmospheric processes in the three types of pollution.Figure 9 shows the net changes in O 3 , NO − 3 , NH + 4 , and SO 2− 4 by the processes of chemical production (Chem), horizontal advection (Horizon-tal_adv), vertical advection (Vertical_avd), and diffusion (Diff; vertical PBL mixing process) in the GEOS-Chem model, as well as the total of all these processes (i.e., Chem + Diff + Horizontal_avd + Vertical_avd) on O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs.
For O 3 , the net changes in O 3 by all processes were positive at altitudes of 975-819 hPa on O 3 -PM 2.5 PDs and O 3 SPDs, in which Chem had the largest positive contribu-tion (about 1.5 Gg d −1 ), indicating O 3 is chemically produced at these layers.For NO − 3 and NH + 4 , the nets of all processes increased mass concentrations at 913-819 hPa on O 3 -PM 2.5 PDs and O 3 SPDs, in which Chem and Vertical_avd were positive, and Chem had the largest positive contribution.The vertical profiles of Chem were similar for NO − 3 and NH + 4 , both of which had the largest positive values at 913-819 hPa (2.83 Gg d −1 for NO − 3 and 0.88 Gg d −1 for NH + 4 ), leading to higher concentrations of NO − 3 and NH + 4 on O 3 -PM 2.5 PDs than on O 3 SPDs and PM 2.5 SPDs.Chem and Diff of SO 2− 4 were different from those of NO − 3 and NH + 4 .For SO 2− 4 , Chem was positive from the surface to 510 hPa, with a peak around 819 hPa, and Diff was positive at 819-771 hPa but negative from 819 hPa to the surface, which resulted in the uniform SO 2− 4 profile as shown in Fig. 8. Chem for SO 2− vertical transport from 819 hPa to near the surface, resulting in the quite uniform vertical profiles at 975-819 hPa on O 3 -PM 2.5 PDs.In addition, the vertical profiles of net changes in PM 2.5 over BTH are shown in Fig. S8 for these three cases.Since NO − 3 , NH + 4 , and SO 2− 4 were the major components of PM 2.5 , the PA of PM 2.5 is similar to that of each component.S7).The vertical profiles of temperature were similar in the three types of pollution, with the lowest temperature on PM 2.5 SPDs.The vertical profiles of OMEGA were different in the three cases.On O 3 SPDs and O 3 -PM 2.5 PDs, OMEGA had positive values around 819 hPa, indicating a strong sinking airflow, leading to a downward transport of pollutants.Under O 3 -PM 2.5 PDs, the average values of PBLH and SWGDN were 946.1 m and 257.2 W m −2 , respec-tively, which were higher (lower) than those on PM 2.5 SPDs (O 3 SPDs) (Fig. S10).
Figure 12 shows the composited weather patterns for regional O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs (over 50 % of the cities in BTH experienced the pollution) that were captured by the model in April-October of 2013-2020.The weather patterns of O 3 -PM 2.5 PDs were similar to some extent to those of O 3 SPDs but were quite different from those of PM 2.5 SPDs.On O 3 -PM 2.5 PDs, the BTH region was controlled by westerlies and an anomalous high-pressure system at 500 hPa (Fig. S11).At 850 hPa, BTH was at the western boundary of an anomalous anticyclone, and the associated strong anomalous southerlies at 850 hPa brought moist air to BTH (Figs.S12 and S13), resulting in a high RH that was beneficial to the aqueous chemical production of SO 2− 4 on O 3 -PM 2.5 PDs.On O 3 SPDs, BTH was under the influence of the high-pressure ridge of the western Pacific subtropical high (WPSH) at 850 hPa.Additionally, the northeast cold vortex was located to the southwest of BTH at 850 hPa on O 3 SPDs, leading to dry and warm conditions, which was favorable for the formation of O 3 .On PM 2.5 SPDs, the BTH region was under the influence of both the continental high and the WPSH at 850 hPa.At the surface, BTH was under the influence of a uniform high pressure with very weak winds and hence stagnate atmosphere, which was conducive to the accumulation of PM 2.5 .

Conclusions
We used the observed hourly concentrations of O 3 and PM 2.5 from CNEMC and the model results from the nested-grid version of the GEOS-Chem model to examine the chemical and physical characteristics of the co-polluted days by  The chemical characteristics of O 3 -PM 2.5 PDs were found to be different from those of O 3 SPDs and PM 2.5 SPDs at the surface.O 3 -PM 2.5 PDs occurred with high levels of atmospheric oxidants (high OH and O x ), with higher SOR and NOR compared to those on O 3 SPDs and PM 2.5 SPDs, leading to high concentrations of both O 3 and PM 2.5 .At the surface, the composited concentrations of NO − 3 , NH + 4 , BC, and OC were the highest on PM 2.5 SPDs, while the composited concentration of SO 2− 4 was the highest on O 3 -PM 2.5 PDs.There was a strong formation of SO 2− 4 during the daytime on O 3 -PM 2.5 PDs in the oxidative atmosphere.
We also found unique features of the vertical distributions of O 3 and PM 2.5 on O 3 -PM 2.5 PDs.Concentrations of PM 2.5 were stable and high between 975 and 819 hPa on O 3 -PM 2.5 PDs, unlike those on PM 2.5 SPDs that decreased rapidly with the altitude.On O 3 -PM 2.5 PDs, the vertical profiles of NO − 3 , NH + 4 , and SO 2− 4 were quite uniform at 975-819 hPa, corresponding to the stable O 3 concentrations at these altitudes.The process analysis (PA) showed that NO − Figure 13 summarizes the chemical and physical characteristics on O 3 -PM 2.5 PDs, O 3 SPDs, and PM 2.5 SPDs in the BTH region.On O 3 -PM 2.5 PDs, the strong chemical productions of O 3 , NO − 3 , NH + 4 , and SO 2− 4 occurred at high altitudes of 913-819 hPa, where RH was high, and the accompanied downward airflow caused the stable concentrations at 944-819 hPa.The composited PBLH on O 3 -PM 2.5 PDs was about 946.1 m, and the strong mixed diffusion underneath the PBLH led to high concentrations of pollutants at the ground level.In contrast, O 3 SPDs occurred in a hot and dry atmosphere with composited PBLH of 1073.5 m.Strong O 3 chemical production occurred around 819 hPa, and O 3 was then transported to the surface by downward airflow.The atmosphere was stable and stagnate when PM 2.5 SPDs occurred, with the lowest PBLH of 681.8 m.High RH (high chemical formation of PM 2.5 ) and the accumulation of aerosols led to the highest surface-layer PM 2.5 on PM 2.5 SPDs.
To summarize, O 3 -PM 2.5 PDs were characterized by high O x , SOR, and NOR; uniform vertical profiles at 975-819 hPa, which were caused by an anomalous high-pressure system at 500 hPa; strong southerlies; and high RH at 850 hPa.Meteorological parameters around 850 hPa promoted strong chemical production of secondary aerosols and downward transport, resulting in the unique vertical profiles and high surface concentrations on O 3 -PM 2.5 PDs.
Figure 1.(a) Spatial distributions of observed numbers of O 3 PDs, PM 2.5 PDs, and O 3 -PM 2.5 PDs summed over April-October of 2013-2020.The solid black rectangle indicates the BTH region.(b) The observed numbers of NPDs (non-polluted days, green), O 3 PDs (blue and purple with slashes), O 3 -PM 2.5 PDs (purple with slashes), and PM 2.5 PDs (pink and purple with slashes) averaged over all sites in BTH from April to October in 2013 to 2020.
from 2013 to 2020.O 3 PDs and O 3 -PM 2.5 PDs mainly occurred in May, June, and July, while PM 2.5 PDs mainly appeared in April and October.The monthly numbers of O 3 -PM 2.5 PDs (PM 2.5 PDs) declined from 2013 to 2020, with the fastest drop in June, from 7.5 (17.1) d in June 2013 to 1.8 (1.8) d in June 2020.In contrast, the numbers of O 3 PDs kept increasing, especially in June, from 10.9 d in June 2013 to 23.6 d in June 2020.The reductions in O 3 -PM 2.5 PDs were associated with the large reductions in PM 2.5 since the implementation of the Clean Air Action in 2013.

Figure 2 .
Figure 2. (a) The trends of observed O 3 PDs, PM 2.5 PDs, and O 3 -PM 2.5 PDs in warm seasons from 2013 to 2020 averaged over all sites in BTH.The blue, yellow, and purple solid lines (dashed lines) represent the numbers (liner trend) of O 3 PDs, PM 2.5 PDs, and O 3 -PM 2.5 PDs, respectively.(b) The percentage of O 3 -PM 2.5 PDs on PM 2.5 PDs for April to October in 2013 to 2020.The polluted days were averaged over all sites in BTH.

3. 2
Figure3a and bshow, respectively, the spatial distributions of simulated and observed surface-layer concentrations of MDA8 O 3 and PM 2.5 in China, as the concentrations are averaged over the warm seasons (April-October) of 2013-2020.The concentrations of MDA8 O 3 and PM 2.5 were both high in BTH.Averaged over BTH and the studied time period, the observed concentrations of MDA8 O 3 and PM 2.5 were 58.1 ppb and 60.3 µg m −3 , respectively, while the simulated values were 68.0 ppb and 61.1 µg m −3 , respectively.Figure3cand d compare the time series of observed and simulated daily MDA8 O 3 and PM 2.5 concentrations averaged over BTH.The simulated daily concentrations of MDA8 O 3 (PM 2.5 ) in eight warm seasons have a normalized mean bias (NMB) of 7.9 % (10.6 %).The model generally captures the daily variations (peaks and troughs) in the observed MDA8 O 3 and PM 2.5 concentrations, with R values of 0.80 and 0.72, respectively.It should be noted that mineral dust and sea-salt aerosols were not considered in this study because they are not the major aerosol components in China, and the concentrations are generally low based on previous measurements(Xuan et al., 2000;Ye et al., 2003;Duan et al., 2006;Zhao et al., 2013).However, excluding dust and sea salt may lead to low biases in simulated PM 2.5 concentrations.Due to the lack of publicly accessible long-term observations of PM 2.5 components in China, we compared the simulated SO 2 and NO 2 (precursors for SO 2− 4 and NO − 3 ) with observations from CNEMC in Fig.S2.The simulated daily mean concentrations of NO 2 (SO 2 ) agree well with the observations from CNEMC, with R of 0.82 (0.78) and NMB of −14.9 % (9.3 %).
FigureS3shows the capability of the model in capturing the polluted days.Although the GEOS-Chem model well reproduces the spatial distributions of observed MDA8 O 3 and PM 2.5 concentrations, it underestimates the numbers of O 3 PDs, PM 2.5 PDs, and O 3 -PM 2.5 PDs because of the model's deficiency in capturing the peak concentrations of air pollutants.Such deficiency was also reported in previous studies that used the GEOS-Chem model or the weather Research and Forecasting with Chemistry (WFR-chem) model(Zhang and Wang, 2016;Ni et al., 2018;Gong and Liao, 2019;Dang and Liao, 2019).Therefore, to identify O 3 PDs, PM 2.5 PDs, and O 3 -PM 2.5 PDs using model results, we utilized lower thresholds by considering the NMBs of simulated MDA8 O 3 and PM 2.5 concentrations in each of the 18 grids of BTH.Taking the grid of Beijing as an example, simulated MDA8 O 3 and PM 2.5 had NMBs of −22.0 % and −26.9 %, respectively, as the simulated concentrations were compared with observations for days with observed concentrations higher than the national air quality standards over the warm seasons of 2013-2020.We then adjusted the threshold of O 3 PDs in this grid to be 62.4 ppb (80 ppb × 78 %) and that of PM 2.5 PDs to be 54.8 µg m −3 (75 µg m −3 × 73.1 %).These adjusted thresholds were also used to identify O 3 -PM 2.5 PDs.Such an approach was also used in previous studies to better capture the pollution events based on the simulations(Dang and Liao, 2019;Gong and Liao, 2019).With the adjusted thresholds, 56 %-93 % of the observed O 3 PDs, PM 2.5 PDs, and O 3 -PM 2.5 PDs can be captured by the model (Fig.S3e).

Figure 3 .
Figure 3. Spatial distributions of simulated (shades) and observed (CNEMC, dots) surface-layer concentrations of (a) MDA8 O 3 (ppb) and (b) PM 2.5 (µg m −3 ) averaged over the eight warm seasons (April to October, 2013-2020).The solid black rectangle in (a) and (b) indicates the BTH region.Simulated and observed daily concentrations of surface-layer (c) MDA8 O 3 and (d) PM 2.5 averaged over BTH.The correlation coefficient (R) and normalized mean bias (NMB) are also shown for (c) and (d).NMB =

3. 3 . 1
Figure5shows the boxplots of daily concentrations of hydroxyl radical (OH) and total oxidant (O x = O 3 + NO 2 ) from the model for days of O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs that were observed and also captured by the model (samples in Fig.4b) in the warm seasons of 2013-2020 in 18 grids of BTH.The levels of OH and O x characterize the atmospheric oxidation capacity, followingLiu et al. (2020) andNan et al. (2017).The concentrations of OH were the highest on O 3 SPDs, with an averaged value of 2.8 × 10 6 molec.cm −3 , followed by that on O 3 -PM 2.5 PDs (2.0 × 10 6 molec.cm −3 ) and on PM 2.5 SPDs (1.0 × 10 6 molec.cm −3 ).Due to the lack of publicly accessible observations of OH in BTH, we compare the simulated OH concentrations with observations reported in the literature (TableS1).The simulated OH concentrations agree closely with the observed values.In Wangdu of BTH, while the observed daily maximum OH concentrations in summer of 2014 were in the range of 5-15 × 10 6 molec.cm −3(Tan et al., 2017), the simulated OH concentrations in the same time period in this work were 3.7-9.5 × 10 6 molec.cm −3 .In Beijing in summer of 2017, the observed daily mean OH concentration was 5.8 × 10 6 molec.cm −3 (Woodward-Massey et al., 2020), and our simulated value was 2.4 × 10 6 molec.cm −3 .

Figure 4 .
Figure 4. Spatial distributions of (a) observed numbers of O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs; (b) numbers of polluted days that were observed and also captured by the GEOS-Chem model with adjusted thresholds; and (c) percentages of observed polluted days that were captured by the model with adjusted thresholds.The values were calculated for the warm months (April to October) of 2013-2020.

Figure 5 .
Figure 5.The boxplots of surface-layer hydroxyl radical (OH, molec.cm −3 ), total oxidant (O x , µg m −3 ), sulfur oxidation ratio (SOR, %), and nitrogen oxidation ratio (NOR, %) for model-captured O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs in 18 grids of BTH in the months of April to October from 2013 to 2020.The whiskers represent the standard deviation, and the black line represents the mean value of the samples.
presents the hourly concentrations of NO − 3 , NH + 4 , SO 2− 4 , BC, OC, and O 3 for model-captured O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs over all 18 grids of BTH in the warm seasons from 2013-2020.Concentrations of NO − 3 and NH + 4 had similarities in diurnal variations, all of which reached the highest values in the early morning (05:00 local time, LT, on O 3 SPDs and O 3 -PM 2.5 PDs, 07:00-08:00 LT

Figure 6 .
Figure 6.The concentrations of PM 2.5 components (µg m −3 ) and percentages of PM 2.5 components (%) at the surface for NO − 3 , NH + 4 , SO 2− 4 , BC, and OC.The values were averaged over the model-captured O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs in the months of April to October of 2013-2020 in BTH.

Figure 7 .
Figure 7.The hourly concentrations of NO − 3 , NH + 4 , SO 2− 4 , BC, OC, and O 3 averaged over the model-captured O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs in BTH in the months of April to October of 2013-2020.
3.3.3Vertical distributions of O 3 and PM 2.5 on O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs The simulated vertical distributions of O 3 and PM 2.5 averaged over the 18 grids of BTH and the O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs in warm seasons of 2013-2020 are shown in Fig. 8.The vertical distribution of O 3 on O 3 SPDs was similar to that on O 3 -PM 2.5 PDs (Fig.

Figure 9 .
Figure 9.The vertical profiles of net changes in O 3 , NO − 3 , NH + 4 , and SO 2− 4 (Gg d −1 ) over BTH by each process and the total processes.The values were averaged over the model-captured regional O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs in April-October of 2013-2020.

Figure 12 .Figure 13 .
Figure 12.Composites of wind field (m s −1 ) with SLP (sea level pressure) and with geopotential height at 850 and 500 hPa for regional O 3 SPDs, PM 2.5 SPDs, and O 3 -PM 2.5 PDs that were captured by the model in April-October of 2013-2020.The solid black rectangle indicates BTH region.
3 , NH + 4 , and SO 2− 4 all had larger chemical productions at altitudes of 913-819 hPa on O 3 -PM 2.5 PDs compared to those on O 3 SPDs and PM 2.5 SPDs.The chemical production of SO 2− 4 had large positive values from the surface to about 500 hPa.The Vertical_avd also had positive contributions to the net changes in O 3 , NO − 3 , NH + 4 , and SO 2− 4 at 944-819 hPa on O 3 -PM 2.5 PDs.Therefore, the strong chemical productions at 913-819 hPa accompanied by the downward transport resulted in the quite uniform vertical profiles at 975-819 hPa on O 3 -PM 2.5 PDs.