Introduction

In recent years, the emissions of major air pollutants, such as sulfur dioxide (SO2), nitrogen oxides (NOx), carbon monoxide (CO), and primary PM2.5, have significantly decreased in Beijing-Tianjin-Hebei (BTH) region due to the implementation of the Chinaʼs Clean Air Action Plans and thus resulted in PM2.5 improvement (Zhang et al., 2019a, b). However, PM2.5 concentration has not yet reached the standard that is recommended by the World Health Organization despite in the hot season with a low level of PM2.5. Additionally, the surface O3 pollution has been worsening in this region in recent years (Li et al., 2019), which has become an important pollutant affecting the air quality following PM2.5, thus bringing about the frequent occurrence of co-pollution of O3 and PM2.5. Both O3 and PM2.5 cause adverse effects on health (Zhang et al., 2017), air quality (Angelevska et al., 2021; Fenech et al., 2019), visibility (Xu et al., 2020), and climate change (Yang et al., 2020). Therefore, it is of critical importance to explore the causes of surface O3 and PM2.5 pollution for addressing the issue of air pollution.

A large body of studies were conducted on either summer O3 or winter PM2.5 to understand the pollution characteristics and source apportionment (Zha et al., 2021; Cheng et al., 2021; Shu et al., 2020). It is found that high level of gaseous precursors was the prerequisites leading to the occurrence of O3 and PM2.5 pollution (Chen et al., 2020; Hughes et al., 2021; Kuerban et al., 2020). And the O3 pollution in summer was always accompanied with high temperature and low relative humidity (Ma et al., 2021b; Yu et al., 2019), while the PM2.5 pollution frequently occurred under the stagnant weather conditions (i.e., high relative humidity and small winds) (Ma et al., 2021b; Wang et al., 2021). Besides, the occurrence of O3 pollution was the joint effects of local photochemistry and regional transport, and the latter was usually dominated in most pollution episodes (Gong et al., 2020), because the O3 pollution is a regional issue. By contrast, the PM2.5 pollution was more a sensitive response to the local emission. Existing studies were mostly focused on O3 or PM2.5 pollution. However, study on the co-pollution of these two pollutants and the interaction between them is limited.

PM2.5 and O3 can affect each other in the atmosphere. Theoretically, as airborne aerosols, PM2.5 can weaken the photolysis processes through absorbing or scattering solar radiation and reduce the precursors through being consumed by aerosol chemistry, which would be adverse to photochemical O3 production (Liu & Wang, 2020a, b; Li et al., 2019). Thus, surface O3 is expected to increase (decrease) with the noticeable decline (rise) of PM2.5 concentrations. Shao et al. (2021) reported that O3 formation increased by 37% in 2006–2016 due to PM2.5 dropping in Beijing. Ma et al. (2021a, b) documented that aerosol radiative forcing accounted for ~23% of the total change in surface O3 over North China Plain in summer of 2013–2019. Previous study has revealed that the atmospheric oxidation capacity is the essential driving force of atmospheric chemistry in forming complex air pollution, which determines the removal rate of trace gases and also the production rates of secondary pollutants (Lu et al., 2019). O3 as an important atmospheric oxidant, can enhance the atmospheric oxidation capacity and thereby promote the conversion of gaseous precursors (i.e., SO2, NO2, and NH3, etc.) into secondary particles (i.e., SO42−, NO3− and NH4+, etc.), and eventually induce PM2.5 pollution (Zhao et al., 2020). Numerous studies have demonstrated that more than half of the components in PM2.5 are secondary particles in China (Liu et al., 2021; Wang et al., 2020), indicating the atmospheric oxidation capacity plays a crucial role in the air pollution of China. Besides, an increase in O3 is usually accompanied with the rise in OH radicals (the most chemically active component in atmosphere), and thereby further enhance the atmospheric oxidation capacity. This is because that the O(1D) generated from O3 photolysis reacting with H2O is another important source of OH (Shao et al., 2004). Overall, O3 pollution will further enhance PM2.5 pollution, and PM2.5 pollution will enhance or inhibit O3 pollution to some extent.

Therefore, to explore the characteristics and causes of co-pollution based on surface O3 and PM2.5 in the hot season in BTH, we researched the co-pollution of O3 and PM2.5 according to the observations in 2015–2019. We aimed to (a) obtain the variation trends of co-pollution under the conditions of continuous declined PM2.5 and aggravated O3, (b) investigate what the meteorological and chemical conditions when co-pollution occurred were; (c) figure out how the atmospheric oxidation capacity changed in co-pollution by analyzing the variations of secondary inorganic aerosols and its oxidation ratio; and (d) explore the interactions of O3 and PM2.5 in co-pollution. Overall, our research results provide a better understanding of co-pollution from the perspective of observations.

Materials and methods

Conception of co-pollution

As mentioned previously, PM2.5 and O3 can affect each other in the atmosphere. The observational data showed that the summer daily mean PM2.5 concentration in China in recent years is lower than 75 μg/m3 but far higher than 35 μg/m3 that is recommended by the World Health Organization guidelines. In addition, considering the national ambient air quality standards (NAAQS) Grade II (35 μg/m3) of annual mean PM2.5 concentration, a certain day in which the MDA8 O3 concentration being above 160 μg/m3 or/and the daily mean PM2.5 concentration above 35 μg/m3 was recognized as a polluted day in this study. The study period in May–September of 2015–2019 was divided into four groups based on the O3 and PM2.5 concentrations above-mentioned: O3-single pollution (MDA8 O3 > 160 μg/m3 and PM2.5 < 35 μg/m3), PM2.5-single pollution (MDA8 O3 < 160 μg/m3 and PM2.5 > 35 μg/m3), co-pollution (MDA8 O3 > 160 μg/m3 and PM2.5 > 35 μg/m3), and non-pollution (MDA8 O3 < 160 μg/m3 and PM2.5 < 35 μg/m3).

Pollutants and meteorological data

The hourly mass concentrations of pollutants, including surface O3, PM2.5, NO2, SO2, and CO in May–September of 2015–2019, were taken from the China National Environmental Monitoring Centre website (CNEMC, http://www.cnemc.cn/). The surface meteorological parameters at 1-h intervals, including wind speed (WS), wind direction, temperature (T), and relative humidity (RH), were obtained from the China Meteorological Administration observation network (http://data.cma.cn/).

Measurement of main compositions in PM2.5

Beijing-Tianjin-Hebei region is located at the northern of the North China Plain, surrounded by the Yanshan Mountains to the north, the Taihang Mountains to the west, the Bohai Bay to the east, and plains to the south. It encompasses two municipalities (Beijing and Tianjin) and one province (Hebei), as shown in Fig. 1. This region is one of the most economically vibrant regions in China. A detailed description of this region can be found in a previous study (Guo et al., 2020). In this study, three cities with obviously different emission structures were chosen to explore the characteristics in co-pollution. Beijing (BJ) is an international and post-industrialization megacity that emits less air pollutant due to less industries compared with other two cities. Tangshan (TS) is an industrial city, resulting in emitting many more air pollutants. Shijiazhuang (SJZ) is the capital city of Hebei province, which is dominated by light industry; thus the emissions of air pollutant level are between BJ and TS.

Fig. 1
figure 1

Location of the BTH region and the sampling sites (PM2.5 sampling site located in Beijing, Shijiazhuang, and Tangshan; NR-PM1 sampling located site in Beijing)

Full size image

Integrated PM2.5 samples were collected on a day-by-day schedule in BJ, SJZ, and TS in summers (July) from 2015 to 2019. The three sampling sites were surrounded by offices, residential areas, and traffic, and all were placed on the rooftop of a certain office building (approximately 20 m from the ground level), which can be a representative of the urban environment. PM2.5 samples were collected on the 90 mm cellulose filters (Whatman Inc. Maidstone, UK) lasting for 24 h per day (from 10:00 am to 9:00 am in the sample collection day) at a volumetric flow rate of 100 L/min. Finally, filters were collected to examine PM2.5 and its chemical compositions, comprising of water-soluble ions and carbonaceous materials. The filters were pre- and post-weighed using an electronic microbalance with an accuracy of 0.01 mg (Sartorius TB-215D, Germany) in a super-clean room with the constant temperature (20 ± 5 °C) and humidity (40 ± 2%) for 48 h. After weighing, the samples were stored in a refrigerator at −18 °C.

For analyzing of water-soluble ions concentrations, a quarter of each sample was extracted ultrasonically using 10-mL distilled-deionized water and oscillated for 40 min in a supersonic cleaner. Sulfates (SO42−), nitrates (NO3−), and ammonium salts (NH4+) were measured by ion chromatography (Metrohm 861 Advanced Compact IC, Switzerland). Carbonaceous concentrations, including organic carbon (OC) and elemental carbon (EC), were measured using a thermal/optical carbon analyzer (DRI Model 2001A, Desert Research Institute of the United States). All the above-mentioned operations were conducted in accordance with quality control standards to avoid any possible contamination in the membrane samples. More detailed analytical procedures for water-soluble ions and carbonaceous aerosols as well as quality control can be found in a previous study (Wang et al., 2015).

NR-PM1 components measurement at 1-h resolution

Previous study reported that secondary inorganic aerosols (SIA, including SO42−, NO3−, NH4+) are mostly gathered in particles with an aerodynamic diameter of less than 1 μm (Zhang et al., 2018). Thus, based on the daily PM2.5 results, non-refractory submicron aerosol (NR-PM1) components at a higher time resolution (at 1–h) were applied to explore the characteristics of secondary particles in co-pollution. Concentrations of NR-PM1 components, including SO42−, NO3−, and organics (Org) in the hot season from 2017 to 2019, were measured by an Aerodyne Aerosol Chemical Speciation Monitor (ACSM) at the Beijing University of Technology (BJUT). The BJUT sampling site (39.88° N, 116.49° E) was located between the southeastern third and fourth ring roads in Beijing, which is surrounded by traffic, residential, and business districts and does not have a stationary pollution source in the vicinity. The ACSM data were analyzed for mass concentration and composition using the ACSM standard data analysis software written in Igor Pro. The output data was at a 15-min time resolution, which was further computed into 1-h resolution to match the precursor concentrations. More detailed descriptions of this instrument and data analysis have been provided in a previous study (Han et al., 2019).

Results and discussion

Occurrence frequency and intensity of co-pollution

According to the definition of co-pollution of O3 and PM2.5, its occurrence frequency in 13 cities of BTH was firstly calculated in the hot season over 2015–2019, as well as that of O3-single pollution, PM2.5-single pollution, and non-pollution. Figure 2 illustrates the proportions of these pollution types in various years and months at regional and local scales. Generally, co-pollution accounted for 24.3–35.6% during 2015–2019 on 13-city average and became the most predominant type since 2017 (Fig. 2a). This is because that although PM2.5 pollution has been alleviated over the last few years due to the implementation of the China’s Clean Air Action Plans, PM2.5 concentration has not yet reached the standard that is recommended by World Health Organization guidelines; in addition, the O3 pollution have been worsening in recent years. Therefore, the co-pollution has become more and more predominant in the hot season. Regarding the monthly variations (Fig. 2b), co-pollution prevailed in May–July, whose proportion reached up to 35.7–44.6% and greatly higher than that in August–September (18.7–21.0%). One most possible reason is that the temperature in May–July is relatively higher than that in August–September, thus leading to the more enhanced atmospheric oxidation capacity due to more occurrence of O3 pollution and further inducing the co-pollution. Additionally, better atmospheric diffusion conditions are usually found in August–September than those in May–July, such as higher mixing layer height, more precipitations, and more windy days, which are conducive to the diffusion of pollutants. The corresponding interannual and monthly changes also presented a good similarity in BJ (international metropolis), SJZ (province capital city), and TS (industrial city), which implied the regional characteristics of co-pollution of O3 and PM2.5 in hot season over BTH.

Fig. 2
figure 2

Occurrence frequency (a, b) and intensity (c, d), concentrations of MDA8 O3 and PM2.5 (e, f) in co-pollution during the study period in BTH

Full size image

Then, we quantified the pollution intensity (PI) of co-pollution according to Eq. (1), in which O3 and PM2.5 pollutants were equally considered.

$$PI= \frac{{C}_{{\mathrm{MDA}8\;\mathrm{ O}}_{3}}}{160} \times 0.5+ \frac{{C}_{{\mathrm{daily\; PM}}_{2.5}}}{35} \times 0.5$$
(1)

Figure 2c shows that PI values of co-pollution presented a discernible decreasing trend, from 1.69 in 2015 to 1.31 in 2019 on 13-city average, which was mostly attributed to the greatly evident decline in PM2.5. To PI values, the contribution of O3 largely increased from 35.3% in 2015 to 48.2% in 2019; conversely, the corresponding contribution of PM2.5 gradually reduced in this period. These results indicated that combined roles of O3 and PM2.5 in air quality over BTH became more and more significant, further suggesting that more attention should be paid to co-pollution in the future. However, the PI values and their contributors varied little in different months (Fig. 2d), which might be balanced by the decreased PM2.5 and increased O3 occurred in various months. During the whole study period in BTH, the mean MDA8 O3 concentration in co-pollution type was close to that in O3-single pollution type but higher by 49.5% than that in PM2.5-single pollution type (Fig. 2e); while the mean of daily PM2.5 concentration in co-pollution type was higher than that in O3- and PM2.5-single pollution types, respectively, by 87.7% and 5.2% (Fig. 2f). It means that the more serious harm of co-pollution type compared to two single pollution types because of the integrated effects of PM2.5 and O3.

Characteristics of meteorology and precursors in co-pollution

Then, we further investigated the meteorological and chemical conditions of co-pollution in BJ, SJZ, and SJZ three cities. The average diurnal patterns of surface temperature (T), relative humidity (RH), and wind speed (WS) as well as precursors of SO2, NO2, and CO in May–September of 2015–2019 were summarized in Fig. 3. It demonstrated that co-pollution type was always accompanied by high T (daily mean of 25.8–28.2 ℃), moderate RH (daily mean of 55.4–63.5%), extremely high SO2 (daily mean of 5.9–31.1 μg/m3), and higher NO2 (daily mean of 36.1–49.8 μg/m3). During the co-pollution, the high T can facilitate O3 formation through directly controlling temperature-dependent photochemical reactions or indirectly promoting precursors emitting from anthropogenic and biogenic sources based on temperature (Liu & Wang, 2020b; Porter & Heald, 2019); in addition, under the conditions of high T and high gas precursors, the moderate RH can be conductive to the formation of secondary particles (i.e., SO42−, NO3−, etc.), thus leading to the increase in PM2.5 concentration. Unexpectedly, small winds were not found in co-pollution, in which the wind speed was comparable among all pollution types. Compared with co-pollution, O3-single pollution had the lower RH (lower by ~11.7% except TS), SO2 (lower by ~27.5%), and CO (lower by ~30.0%) but presenting the similar levels of T and NO2, while PM2.5-single pollution had the lower T (lower by ~11.4%), higher RH (higher by 14.9%), and lower SO2 (lower by 27.1% except TS) but the similar levels of NO2 and CO. It can be seen high T and abundant NO2 was the essential condition of O3-single pollution and co-pollution; and then the participation of high level of SO2 facilitated the PM2.5 increase through aerosol chemistry, thus eventually inducing the co-pollution of O3 and PM2.5. Additionally, CO concentrations in co-pollution were comparable to those in PM2.5-single pollution but far higher than those in O3-single pollution and non-pollution. These results also implied the important role of physical accumulation in co-pollution type, as well as in PM2.5-single pollution type. This is as a result of a fact that CO is not active photochemical, whose chemical life can last for several months; it can be seen as an inert pollution tracer in a short period and is mainly influenced by meteorological factors (Zhang et al., 2015). Therefore, the high level of CO means that poor atmospheric diffusion conditions occurred.

Fig. 3
figure 3

Diurnal variations of meteorological parameters and precursors in co-pollution during the study period in BJ, SJZ, and TS

Full size image

In addition, we applied multiple linear regression (MLR) method based on the Statistical Product and Service Solutions (SPSS) to further quantify the relationship among co-pollution and meteorological parameters and precursors during June, in which the co-pollution most frequently occurred compared with other months of the hot season (Fig. 2b). In this study, the values of confidence (i.e., Sig.) were all far lower than 0.05, indicating that the linear regression equations were meaningful. Besides, the variance inflation factor (VIF) was in the range of 1.1–4.5 and were lower than 10, demonstrating no collinearity among all variables (Fox & Monette, 1992). Table 1 summarizes the linear regression equation of hourly O3 or PM2.5 concentrations with meteorological parameters ([T], [RH], and [WS]) and precursors ([SO2], [NO2], and [CO]) under different pollution types.

Table 1 Linear regression equation of O3 or PM2.5 under different pollution types
Full size table

As for O3, the regression coefficients of all factors (except for WS) showed a comparable level in co-pollution (Co_O3) and O3-single pollution (Single_O3), further indicating that the causes of O3 formation in the two pollution types were similar, which was consistent with the results observed in Fig. 3. The coefficient value of WS in Co_O3 was significantly higher than that in Single_O3, which could imply that the effect of transport on O3 in co-pollution could be bigger than that in O3-single pollution. Regarding PM2.5, the regression coefficients presented significant differences between co-pollution (Co_PM2.5) and PM2.5-single pollution (Single_PM2.5), particularly the coefficients of T and RH being higher in co-pollution, which suggested that PM2.5 in co-pollution was more dependent on high T and RH than that in PM2.5-single pollution. Besides, the regression coefficient of SO2 was negative and more significant in co-pollution, which highlighted that the chemical formation of PM2.5 played a dominant role in co-pollution, while the regression coefficient of WS was negative and more pronounced in PM2.5-single pollution, indicating that the small wind was obvious and the physical accumulation might be the driving force in PM2.5-single pollution. In general, quantitative relationships among meteorological parameters and precursors would be helpful for better understanding the occurrence of co-pollution.

Characteristics of high level of oxidation products in co-pollution

SIA of daily PM2.5 in July

Both ozone and secondary PM2.5 are deeply related to atmospheric oxidation reactions. The major secondary inorganic aerosols (SIA) include SO42−, NO3−, and NH4+, in which SO42− and NO3− are directly generated from the oxidations of SO2 and NOx and NH4+ is converted from NH3 to neutralize the acid aerosols. Thus, the bigger the content of SIA in particles is, the more aged and oxidized the atmosphere is. According to the PM2.5 measurements in BJ, SJZ, and TS, the average proportions of SIA and other components in July were obtained and summarized in Fig. 4. It can be seen the total proportion of three inorganic salts was the highest in co-pollution type and accounted for 44.3–48.7% of PM2.5; PM2.5-single pollution (with proportion of 42.1–46.5%) and O3-single pollution (with proportion of 41.2–44.3%) followed, while, the corresponding proportion in non-pollution was the lowest (with proportion of 35.7–40.6%). These results further illustrated more formation of secondary particles in co-pollution under the condition of a strong oxidative air condition with high level of O3 concentrations.

Fig. 4
figure 4

Variation of main PM2.5 compositions in co-pollution over July of 2015–2019 in BJ, SJZ, and TS

Full size image

The increase of SIA proportion in co-pollution type compared with PM2.5-single pollution type was mainly derived from SO42− (increasing by 12.0–18.0%) and NH4+ (increasing by 4.3–7.7%), but negatively affected by NO3− (decreasing by 0.2–3.9%), which was associated with the fact that NO3− generated by photochemical reactions could not overcome or offset the effects of NO3− volatilization induced by the high temperature, particularly after noon (Zhang et al., 2018). While the increase of SIA proportion in co-pollution type compared with O3- single pollution type resulted from the enhancement of SO42−, NO3−, and NH4+, with contribution of 15.3–17.1%, 13.1–16.0%, and 11.5–12.9%, respectively, in general, the contribution of SO42− to PM2.5 played a more predominant role than NO3− and NH4+, particularly in the co-pollution type. This could be explained by the extremely high level of SO2 in co-pollution facilitating the evident increase in PM2.5 through aerosol chemistry (see in Fig. 3). In addition, the ratios of OC/EC varied among different pollution types, following the order of O3-single pollution (4.03 ± 0.37) > co-pollution (3.47 ± 0.27) > PM2.5-single pollution (3.12 ± 0.22), demonstrating that the atmosphere oxidation was relatively more intense in co-pollution and thus strengthen the production of oxidation products (i.e., SIA and OC) and finally led to PM2.5 increase.

SIA of hourly NR-PM1 in May–September

In this study, ACSM analyzer was applied in BJ to measure the compositions of particle with an aerodynamic diameter of less than 1 μm (NR-PM1) at the time resolution of 1-h. The temporal characteristics of the secondary aerosols in co-pollution are further explored according the NR-PM1 and its components, as shown in Fig. 5. NR-PM1 mass concentration in co-pollution during May–September of 2017–2019 was measured to be 26.5 μg/m3 on average (Fig. 5c), which was 1.05, 0.14, and 2.06 times higher than that in O3- single pollution, PM2.5- single pollution, and non-pollution, in respective. These differences in NR-PM1 among various pollution types were significantly bigger than those in PM2.5, in which PM2.5 mass concentration in co-pollution was measured to be 60.9 μg/m3 on average (Fig. 5b), being 1.22, 0.01, and 1.94 times higher than that in O3- single pollution, PM2.5- single pollution, and non-pollution, respectively. As a result of approximately 50% secondary aerosols gathering in NR-PM1 (Zhang et al., 2019b), it can better reflect the characteristics of secondary conversion than that of PM2.5.

Fig. 5
figure 5

Diurnal cycles of pollutants in different pollution types during May–September of 2017–2019 in BJ

Full size image

Diurnal variations in pollutant concentrations can provide insights into the interplay between emissions and chemical and physical processes that operate on a diurnal cycle (Xu et al., 2021). Thus, to better understand the diurnal changes of various secondary aerosols, their hourly concentrations were normalized to the concentration at 0:00, as shown in Fig. 5d−f. The SO42− behaved as a significant contributor in NR-PM1 in co-pollution, which was consistent with the above-mentioned results of PM2.5 (Fig. 4). Its corresponding contribution to NR-PM1 mass concentration was relatively lower at nighttime (23.7–38.3%), but rapidly rose since the morning and reached the peak in the afternoon (Fig. 5g). This pattern was well consistent with that of O3 concentrations, illustrating that the augmentation of atmospheric oxidation capacity with high level of O3 could strengthen the production of secondary particles. The SO42− proportion at daytime (8:00–18:00, LT) reached up to 35.5% on average in co-pollution, which was far higher than those in PM2.5- single pollution (31.8%) and O3- single pollution (27.5%), further demonstrating that the important role of SO42− in co-pollution. In addition, its proportion at nighttime was comparable to that of co-pollution and O3- or PM2.5-single pollution. These results implied that SO42− formation was closely related to daytime photochemistry and gas-phase reactions related to O3 could be the key channel on the condition of high level of SO2 and proper RH (Fang et al., 2019; Shon et al., 2012).

However, the behavior of NO3− was opposite to that of SO42−. The mass proportion of NO3− was relatively higher at nighttime than that at daytime. This is because that and the NO3− generated by photochemical reactions at daytime could not overcome or counteracted the effects of NO3− volatilization induced by the high temperature. Besides, a high level of the NO3− at nighttime was associated with the high level of precursor NO2 on condition of high RH and abundant oxidant (Fig. 3d, m); and heterogeneous hydrolysis of N2O5 on humid aerosol surfaces is an important pathway for the formation of NO3− at nighttime (Wang et al., 2017). During nighttime, NO3− would contribute 21.1% and 25.8% to NR-PM1 in co-pollution and PM2.5-single pollution, respectively, but only 15.8% to NR-PM1 in O3-single pollution. This difference might be caused by the fact that the highest level of SO2 and CO in co-pollution played a key role during this period (Fig. 3j, p).

For Org, its corresponding proportion was relatively higher at daytime (10:00–16:00) in O3-single pollution but higher at nighttime (19:00–23:00) in PM2.5-single pollution. The former reflected the more important role of photochemistry in the formation of secondary organic aerosols, while the latter might indicate the greater impacts of nighttime chemistry, whereas the Org mass concentration in co-pollution were both at a high level at daytime and nighttime. However, the corresponding proportion was lower in co-pollution and PM2.5-single pollution than those in O3-single pollution and non-pollution, indicating that the proportion of carbon components declined with increasing particle concentration. This decrease in the proportions of carbon components from low-level to high-level of particles were similar to other studies in BTH region (Zhao et al., 2019). Overall, the aged or oxidation process of the inorganic was stronger than that of the organic in co-pollution.

Interaction of O3 and PM2.5 in co-pollution

To further explore the interactions between O3 and PM2.5, we analyzed 21 co-pollution episodes occurring in BJ during the study period (Fig. 6a). Among them, 13 episodes were initiated by high level of O3 mostly occurring about at 13:00, and then the PM2.5 concentrations rapidly increased within 2–4 h. This indicated that the augmentation of atmospheric oxidation capacity strengthens the production of secondary particles and thus increased the PM2.5 concentrations. The O3-initiated co-pollution always lasted for approximately 2–7 days. In addition, another 8 episodes were initiated by high level of PM2.5 mostly occurring at nighttime (~ 21:00), leading to the accumulation of precursors, and then the O3 concentrations slowly increased within 3–6 h. This is mostly because that high level of aerosols causes cooling of the surface, resulting in reduced buoyant turbulence, enhanced atmospheric stratification, and suppressed boundary layer growth (Slater et al., 2020). The PM2.5-initiated co-pollution always continued for about 1–3 days, which was evidently shorter than that induced by O3. This is mostly probably that the co-pollution driven by O3 was closely related to the chemical processes referred to the augmentation of atmospheric oxidation capacity (Jia et al., 2017), while the co-pollution driven by PM2.5 was ascribed to the physical processes involved to unfavorable meteorological conditions. It is worth noting that both the mean MDA8 O3 and PM2.5 concentrations in O3-initiated co-pollution were significantly higher than those in PM2.5-initiated co-pollution (Fig. 6b), respectively, higher by 20.1 μg/m3 (10.0%) and 4.2 μg/m3 (7.1%), highlighting that the enhanced atmospheric oxidation capacity due to high level of O3 could be the driving force to the formation secondary particles. Moreover, the mean pollution intensity in O3-initiated co-pollution was slightly higher than that in PM2.5-initiated co-pollution, higher by 10.0%.

Fig. 6
figure 6

Characteristics of varied types of co-pollution during May–July of 2017–2019 in BJ

Full size image

As expected, a higher T (daily of 27.4 ± 3.0 °C) and lower RH (daily of 53.5 ± 11.0%) as well as lower WS (daily of 2.07 ± 0.31 m/s) were found in O3-initiated co-pollution (Fig. 6c), in which the higher T would greatly facilitate the occurrence of chemical processes and the lower WS promotes the accumulation of precursors. It is noteworthy that the precursors SO2 and NO2 in O3-initiated co-pollution episodes were slightly higher than those in PM2.5-initiated co-pollution episodes, respectively higher by 0.6 μg/m3 (8.8%) and 1.0 μg/m3 (2.7%). Such meteorological and chemical conditions were conductive to the formation of secondary particles under the condition of high level of O3. Combining the concentrations of the inorganic salts in NR-PM1 and the corresponding gaseous precursors, the sulfur oxidation ratio (SOR) and nitrogen oxidation ratio (NOR) were estimated to evaluate the aging of atmosphere in co-pollution, which was calculated through Eqs. (2) and (3), respectively (Lang et al., 2017; Zhao et al., 2019). Higher values of SOR and NOR reflect greater oxidation of gaseous SO2 and NO2, and more secondary aerosols exist in the atmosphere (Huang et al., 2016).

$$\mathrm{SOR}=\frac{n[{SO}_{4}^{2-}]}{n[{SO}_{4}^{2-}]+n[{SO}_{2}]}$$
(2)
$$\mathrm{NOR}=\frac{n[{NO}_{3}^{-}]}{n\left[{NO}_{3}^{-}\right]+n[{NO}_{2}]}$$
(3)

where n represents the molar concentration of the corresponding air pollutants (mol/m3). The SIA concentrations and their proportions in O3-initated episodes were substantially higher than those in PM2.5-initiated episodes, higher by 5.4 μg/m3 and 6.9%, respectively (Fig. 6d). The mean of daily SOR would reach up to 0.47 ± 0.24 and 0.40 ± 0.26 (Fig. 6e) and that of NOR were 0.11 ± 0.09 and 0.09 ± 0.05 (Fig. 6f) in O3-initated and PM2.5-initiated episodes, respectively. These results indicate that the atmosphere in O3-initiated episodes was much more aged and oxidized compared with that in PM2.5-initiated episodes. This was further approved by the higher “total oxidant” (Ox = O3 + NO2) production in O3-initiated episodes, being higher by 4.6% than that in PM2.5-initiated episodes.

It can be concluded that (i) in O3-initiated episodes, the high level of O3 at daytime induced the raise of PM2.5 through facilitating the formation of secondary sulfates on the condition of high level of SO2, thereby co-pollution occurring, in which chemical processes would be the driving force; (ii) while in PM2.5-initiated episodes, high level of PM2.5 firstly occurred under the more stable meteorological conditions (high RH and small winds), and then accumulation of gaseous precursors further induced high O3, in which the physical processes would play a key role.

Conclusions

The monitoring data of O3 and PM2.5 covering all cities of BTH in the hot season during 2015–2019 was analyzed to investigate the variation trends of co-pollution and its causes. It is found that co-pollution of O3 and PM2.5 has become the most predominant type of air pollutions in May–July since 2017 in BTH. And the co-pollution was always accompanied by high T (25.8–28.2 ℃), moderate RH (55.4–63.5%), extremely high SO2 (5.9–31.1 μg/m3), and higher NO2 (36.1–49.8 μg/m3). MLR analysis showed that O3 in co-pollution and O3-single pollution had the comparable dependence on meteorology and precursors, while PM2.5 in co-pollution was more sensitive responses to T, RH, and precursors but that in PM2.5-single pollution was more sensitive responses to small winds. These results implied that co-pollution seems to be more affected by atmospheric chemistry. Further analysis based on the PM2.5 measurements showed that SIA proportion was the highest in co-pollution and accounted for 44.3–48.7% of PM2.5 and then PM2.5-single pollution (42.1–46.5%) and O3-single pollution (41.2–44.3%) followed, further highlighting the relatively stronger atmospheric chemistry processes in co-pollution. We found that the extremely high proportion of SIA in co-pollution was mainly attributed to SO42−, which was observed to rapidly boom in NR-PM1 with high level of O3 at daytime.

Furthermore, the interactions of O3 and PM2.5 in co-pollution were explored based on 21 episodes occurring in Beijing. Most (~ 61.9%) co-pollution episodes were initiated by high level of O3 accompanied by high SO2 at daytime, in which the chemical processes played a key role; as for the rest of the episodes, high PM2.5 firstly occurred under the more stable meteorological conditions, and then accumulation of precursors further induced high O3. The SIA concentration in O3-initiated co-pollution was higher than that in PM2.5-initiated co-pollution, indicating that the atmospheric oxidation in co-pollution caused by chemical processes was stronger than that by physical processes. And it was further approved by the higher values of SOR and NOR in O3-initiated co-pollution. The key findings revealed that controlling O3 and precursor SO2 is the key to abating co-pollution in the hot season. Quantitative study on the control strategies needs to be further carried out by numerical simulation in future.