Atmospheric Oxidation Capacity and Its Impact on the Secondary Inorganic Components of PM2.5 in Recent Years in Beijing: Enlightenment for PM2.5 Pollution Control in the Future

Abstract

In recent years, the concentrations of PM_2.5 in urban ambient air in China have been declining; however, the strong atmospheric oxidation capacity (AOC) represents challenges to the further reduction of PM_2.5 concentration and the continuous improvement of ambient air quality in China in the future, since the overall AOC is still at a high level. For this paper, based on ground observation data recorded in Beijing from 2016 to 2019, the variation in AOC was characterized according to the concentration of odd oxygen (O_X = O₃ + NO₂). The concentrations of the primary and secondary components of PM_2.5 were analyzed using empirical formulas, the correlation between AOC and the concentrations of secondary PM_2.5 and the secondary inorganic components (SO₄²⁻, NO₃⁻, NH₄⁺, and SNA) in Beijing were explored, the impact of atmospheric photochemical reaction activity on the generation of atmospheric secondary particles was evaluated, and the impact of atmospheric oxidation variations on PM_2.5 concentrations and SNA in Beijing was investigated. The results revealed that O_X concentrations reached their peak in 2016 and reached their lowest point in 2019. The O_X concentrations followed a descending seasonal trend of summer, spring, autumn, and winter, along with a spatial descending trend from urban observation stations to suburban stations and background stations. The degree of photochemical activity and the magnitude of the AOC have a large influence on the production of atmospheric secondary particles. When the photochemical reaction was more active and the AOC was stronger, the mass concentrations of the secondary generated PM_2.5 fraction were higher and accounted for a higher proportion of the total PM_2.5 mass concentrations. In the PM_2.5 fraction, SNA accounted for 50.7% to 94.4% of the total mass concentrations of water-soluble inorganic ions in the field observations. Higher concentrations of the atmospheric oxidant O_X in ambient air corresponded to a higher sulfur oxidation ratio (SOR) and nitrogen oxidation ratio (NOR), suggesting that the increase in AOC could promote the increase of PM_2.5 concentration. Based on a relationship analysis of SOR, NOR, and O_X, it was inferred that the relationship between O_X and SOR and the relationship between O_X and NOR were both nonlinear. Therefore, when establishing PM_2.5 control strategies in Beijing in the future, the impact of the AOC on PM_2.5 generation should be fully considered, and favorable measures should be taken to properly regulate the AOC, which would be more effective when carrying out further control measures regarding PM_2.5 pollution.

1. Introduction

In order to address its heavy air pollution, China has promulgated a series of national action plans since 2013, including the Air Pollution Prevention Scheme and the Control Action Plan, the Three-Year Action Plan for Blue Sky Protection Campaign, and the Action Plan for the Elimination of Heavily Polluted Weather, the Prevention and Control of Ozone Pollution, and the Control of Diesel Truck Pollution. Through a series of air pollution prevention and control measures, the concentration of PM_2.5 (with aerodynamic diameters of ≤2.5 μm) in urban ambient air has dropped significantly. From 2016 to 2019, the pollution control effect was very significant in China; the annual average mass concentration of PM_2.5 decreased from 47 μg/m³ to 36 μg/m³. As an international metropolis, Beijing has also achieved remarkable results regarding air pollution control. From 2016 to 2019, the annual average mass concentration of PM_2.5 in Beijing decreased from 75 μg/m³ to 40 μg/m³, but there is still a long way to go before levels will reach the WHO guideline value (5 μg/m³). Meanwhile, during this period, the annual evaluation value of O₃ (the 90th percentile of the maximum daily 8-h moving average) in Beijing increased by 5% [1,2]; the AOC remained at a high level and the proportion of secondary components in PM_2.5 continued to increase [3,4,5].

The AOC is defined as the total reaction rate of primary pollutants (such as volatile organic compounds (VOCs)) and atmospheric oxidants, which can also be characterized by the O_X value. Secondary components in PM_2.5 include secondary inorganic aerosols (SNA, such as sulfate, nitrate, and ammonium) and secondary organic aerosols (SOA), while secondary formation caused by atmospheric chemical reactions has been proven to be a significant source of PM_2.5, which would be promoted by a strong AOC [6,7,8]. Wang et al. found that the AOC level of ambient air did not decrease significantly in China, and even increased slightly in the North China Plain and the Pearl River Delta during the period from 2013 to 2020 [8]. It was found that the strong AOC led to an increasing proportion of secondary components in PM_2.5 in the ambient air in China [9]. Zhu et al. [7] found that the levels of SOR, NOR, and SOA in summer have a strongly positive correlation with the concentrations of O_X and O₃, indicating that the AOC caused by strong photochemical reaction activity in the atmosphere was enhanced, which, in turn, increased the generation of secondary particles. Kim et al. [10] found that although the concentration of particulate matter in the atmosphere was low and the numbers of particulate matter over the standard limit were small in summer, heavy O₃ pollution occurred with a rapid increase in PM_2.5 concentration. Tian et al. [11] found that increased levels of AOC accelerated the generation of secondary particulate matter, thereby weakening the effect of precursor emission reduction during the autumn haze. Xiaodong et al. found that the concentration of NO₃⁻ in PM_2.5 in Beijing gradually increased from 2013 to 2019, with an increase rate of 1.3% per year [12].

Beijing faces many difficulties regarding atmospheric composite pollution, with an increase in the concentrations of secondary pollutants (such as O₃, SIA (Secondary inorganic aerosol, SIA), and SOA) [12,13]. Strong AOC levels can promote the generation of PM_2.5 secondary components; there is a complex atmospheric chemical coupling mechanism between AOC and the generation of secondary pollutants. Therefore, it is of great significance to explore their relationship, to further realize the coordinated control of PM_2.5 and O₃ in China. Based on this background, this research selected data from 2016 to 2019 to study the spatiotemporal variation characteristics of AOC in Beijing during this period, along with its correlation with PM_2.5 mass concentration, evaluate the impact of different degrees of atmospheric photochemical activity on the generation of secondary components in PM_2.5, and explore the impact of AOC on the generation of secondary inorganic components in PM_2.5 in summer and autumn by using the ambient air quality data from 35 monitoring stations and enhanced observation data from Beijing. This study aims to clarify the effect of AOC on the generation of secondary components in PM_2.5 and provide a point of reference for the government when making decisions to effectively control PM_2.5 concentration.

2. Experiment

2.1. Observation Sites and Period

2.1.1. Data on Conventional Pollutants

This study obtained real-time conventional pollutant data (SO₂, NO₂, O₃, CO, PM₁₀, and PM_2.5) from 35 monitoring stations, which was released by the Beijing Municipal Bureau of Ecological Environment (http://zx.bjmemc.com.cn (accessed on 1 March 2020)) from 2016 to 2019. According to their different monitoring functions and purposes, these stations can be divided into 12 urban environmental assessment stations, 11 suburban environmental assessment stations, 5 traffic pollution monitoring stations, and 7 background stations. These 35 stations cover all the administrative areas of Beijing; their spatial distribution is shown in Figure 1.

2.1.2. Enhanced Field Observation

The enhanced field observation site (Figure 1) was located on the roof of the Atmospheric Photochemical Smog Chamber Simulation Laboratory of the Chinese Research Academy of Environmental Sciences, Chaoyang District, Beijing (40.04 N, 116.42 E), about 8 m above the ground. The observation site was adjacent to major roads and subway stations, with large residential areas, offices, schools, hospitals, and shopping centers in close proximity, making the area highly trafficked and densely populated. The observation periods were in summer (4 July to 6 August 2019) and autumn (19 September to 11 October 2019). The online observations include conventional pollutants. Two PM_2.5 pollution processes were selected for offline sampling during the field observation period, with one in the daytime (07:30–19:00) and one at night (19:30–07:00). The water-soluble inorganic ion components in PM_2.5 were also analyzed.

2.2. Observation Items and Analysis Methods

2.2.1. Data on Conventional Pollutants

The gaseous pollutants, SO₂, NO₂, CO, and O₃, were monitored using an SO₂ analyzer, model 43i, a NOx analyzer, model 42i, a CO analyzer, model 48i, and an O₃ analyzer model 49i (Thermo Scientific, Waltham, MA, USA), respectively. PM₁₀ and PM_2.5 were monitored using a 5030 monitor (Thermo Scientific, Waltham, MA, USA). The sampling port of the monitor was equipped with cutting heads, corresponding to the collection of atmospheric particles in different particle size ranges for PM₁₀ and PM_2.5 sampling.

2.2.2. PM_2.5 Sampling and Inorganic Ion Component Analysis

PM_2.5 inorganic ion components were collected in the daytime (07:30–19:00) and nighttime (19:30–07:00) using an off-line sampling method, then the water-soluble inorganic ion components in PM_2.5 were analyzed. Since we replaced the filter manually, we needed to replace the filter and clean the sampler before the next sampling was conducted every day, creating a 30-min gap. The PM_2.5 samples were collected using quartz filter membranes (Whatman, 203 mm × 254 mm) with a high flow sampler (Thermo Scientific, Waltham, MA, USA). The mass concentration of PM_2.5 was analyzed using the gravimetric method. Quartz filter membranes were weighed with an analytical balance (CPA225D, 0.01 mg). The collected PM_2.5 filter membrane samples were analyzed using ICS-1000 and ICS-1100 ion chromatographs for water-soluble cations (Ca²⁺, NH₄⁺, Mg²⁺, Na⁺, and K⁺) and anions (SO₄²⁻, NO₃⁻, Cl⁻, and F⁻), respectively.

2.3. Quality Control

2.3.1. Online Monitoring of Conventional Pollutants

The ambient air automatic continuous monitoring system included a model 49i (O₃), model 48i (CO), model 43i (SO₂), model 42i (NO₂), and model 5030 SHARP monitor (for PM₁₀ and PM_2.5) (Thermo Scientific, Waltham, MA, USA). Data quality assurance and quality control (QA/QC) followed the regulations for automated methods for ambient air quality monitoring (HJ/T 193–2005) [14]. The model 42i, model 48i, and model 43i instruments were calibrated regularly using NO (2 μmol/mol), CO (4.24 μmol/mol), and SO₂ (54 μmol/mol) standard gases. NO standard gas is produced by Air Products and Chemicals, Inc. Nitrogen was used as the balance gas. The CO and SO₂ standard gases were produced by the Institute for Environmental Reference Materials (IERM) of the China Ministry of Environmental Protection. The model 49i instrument was calibrated annually using the standard reference photometer (SRP) calibration from the IERM. The paper tapes of the PM₁₀ and PM_2.5 particulate matter monitor (model 5030 SHARP monitor) were checked and the filter membranes were changed once a week. The cutting heads were cleaned once a week, which meets the requirements of the “Specifications and test procedures for ambient air quality continuous automated monitoring system for PM₁₀ and PM_2.5 (HJ 653—2021) [15]”. Leakage detection and flow correction were carried out once a week on the monitor.

2.3.2. Offline Sampling of PM_2.5 and Ion Component Analysis

Before PM_2.5 sampling, the key components of the high-flow sampler were cleaned with anhydrous ethanol; then, a layer of petroleum jelly was evenly applied to the upper surface of the impact plate, and the flow rate of the sampling system was tested and calibrated. During sampling, tweezers cleaned with anhydrous ethanol were used to replace the filters, and nitrile gloves were worn. After the sampling, the filters were wrapped with clean aluminum foil and stored in the refrigerator. The ICS-1000 and ICS-1100 ion chromatographs were verified and calibrated to ensure that the analysis requirements were met during the ion component analysis of PM_2.5. The correlation coefficients of the correction curves were above 0.995. While analyzing the samples, the standard solutions of the midpoint concentration of calibration curves were randomly tested, and the relative errors were controlled within the range of ±10%.

2.4. Data Analysis and Processing

Based on the data from the national urban sites that were monitored (excluding background sites), the over-standard conditions of PM₁₀, PM_2.5, and O₃ in Beijing from 2016 to 2019 were determined, according to the relevant requirements for the Ambient Air Quality Standards (GB 3095–2012) and Technical Regulation for Ambient Air Quality Assessment (On Trial) (HJ 663–2013) [16,17]. According to the revision notice of Ambient Air Quality Standards (GB 3095–2012) issued by the Ministry of Ecology and Environment in 2018, the data on conventional pollutants used in this study from 2016 to 2019 were recorded under conditions for monitoring data using the reference state (25 °C, 101.325 kPa), while the data of PM_2.5 and PM₁₀ were recorded when monitoring in the actual state.

2.4.1. Correlation Analysis

The hourly average mass concentrations of PM₁₀, PM_2.5, and SNA were converted into the daily average, seasonal average, and annual average using Microsoft Excel 2020 software (Microsoft, Redmond, WA, USA). The hourly average mass concentration of O₃ was converted into the maximum daily 8-h moving average (MDA8 O₃) and the annual evaluation value (90th percentile of the maximum daily 8-h moving average). The Pearson correlation coefficient (r) was calculated using SPSS 22 software (International Business Machines Corporation, Armonk, NY, USA) to quantify the correlation among PM_2.5, SNA, O₃, and O_X.

2.4.2. Estimation of Secondary Particle Generation

In this study, the primary and secondary component concentrations in PM_2.5 were analyzed using an empirical formula method. The specific method takes carbon monoxide as a tracer of the primary components, and MDA8 O₃ represents the degree of photochemical reaction activity [6,18]; the formula expressions are as follows:

(PM_2.5)_p,G,h = CO_G,h × (PM_2.5/CO)_p,E,h(1)

(PM_2.5)_p,L,h = CO_L,h × (PM_2.5/CO)_p,E,h(2)

(PM_2.5)_p,M,h = CO_M,h × (PM_2.5/CO)_p,E,h(3)

where E, G, L, and M represent different O₃ concentration ranges (E: MDA8 O₃ ≤ 100 μg/m³; G: 100 μg/m³ < MDA8 O₃ ≤ 160 μg/m³; L: 160 μg/m³ < MDA8 O₃ ≤ 200 μg/m³; M: MDA8 O₃ > 200 μg/m³); p: primary pollutant; h: any hour of the day. PM_2.5 and CO denote the hourly concentrations of PM_2.5 and CO, respectively.

The concentrations of the secondary components of PM_2.5 are the observed concentration of PM_2.5 minus the concentrations of the primary components of PM_2.5 obtained from Formulas (1)–(3). Therefore, the formulas for calculating the concentrations of the secondary components of PM_2.5 are:

(PM_2.5)_sec,G,h = (PM_2.5)_obs,G,h-(PM_2.5)_p,G,h(4)

(PM_2.5)_sec,L,h = (PM_2.5)_obs,L,h-(PM_2.5)_p,L,h(5)

(PM_2.5)_sec,M,h = (PM_2.5)_obs,M,h-(PM_2.5)_p,M,h(6)

where (PM_2.5)_sec,G,h, (PM_2.5)_sec,L,h, and (PM_2.5)_sec,M,h represent the concentrations of secondary components in PM_2.5 at different O₃ pollution levels.

In order to understand the impact of AOC and photochemical reaction activity on secondary particle generation, O₃ concentration is usually used as an indicator to reflect the photochemical reaction activity. In order to distinguish the impact of different photochemical reaction activity levels on the generation of secondary particles, the photochemical reaction activity levels were divided into four categories according to O₃ concentrations, based on values given in the literature [6,18]:

• (1). R0: MDA8 O₃ ≤ 100 μg/m³;

• (2). R1: 100 < MDA8 O₃ ≤ 160 μg/m³;

• (3). R2: 160 < MDA8 O₃ ≤ 200 μg/m³;

• (4). R3: MDA8 O₃ > 200 μg/m³.

2.4.3. Calculation of Sulfur Oxidation Rate, Nitrogen Oxidation Rate, and Atmospheric Oxidant (O_X) Concentrations

SOR and NOR are often used to measure the secondary conversion efficiency of gaseous precursors such as SO₂ and NOx. The higher the values of SOR and NOR, the higher the secondary transformation degree of SO₂ and NO₂ in the atmosphere. The formulas are as follows:

SOR = N₁/(N₁ + N₂)(7)

NOR = S₁/(S₁ + S₂)(8)

where N₁ and N₂ represent the concentrations of NO₃⁻ and NO₂ in mol/m³, respectively; S₁ and S₂ represent the concentrations of SO₄²⁻ and SO₂ in mol/m³, respectively.

O_X is commonly used to indicate AOC. In this study, O_X concentrations from 2016 to 2019 at four types of monitoring stations were calculated to characterize the AOC at different periods. The calculation method is as follows:

ρ(O_X) = ρ(O₃) + ρ(NO₂)(9)

where ρ(O_X), ρ(O₃), and ρ(NO₂) represent the concentrations of O_X, O₃, and NO₂ in μg/m³, respectively.

3. Results and Discussion

3.1. Changes in O_X Concentration and Its Correlation with PM_2.5 Concentration

3.1.1. Interannual Change and Seasonal Change of O_X Concentration

As seen in the variations in the monthly average concentrations of O_X in Beijing from 2016 to 2019, the O_X concentrations reached their highest levels in June (142–154 μg/m³) and the lowest in October–December (67–80 μg/m³), in these four years. This showed that the AOC was usually the highest in summer and the lowest in winter (Figure 2).

High O_X concentrations have long been considered an important feature of air pollution in Beijing [11,19,20]. As shown in Figure 3, the annual mean O_X concentrations at urban stations, suburban stations, traffic pollution monitoring stations, and background stations in Beijing from 2016 to 2019 were 92–107 μg/m³, 90–105 μg/m³, 96–114 μg/m³, and 84–98 μg/m³, respectively. The annual mean O_X concentrations of the four types of stations fluctuated little from 2016 to 2018, but the annual mean O_X concentration in 2019 showed a significant decline. Compared to those in 2018, the annual mean values at urban stations, suburban stations, traffic pollution monitoring stations, and background stations decreased by 11, 11, 12, and 13 μg/m³, respectively. The annual O_X values of the four types of stations, listed in descending order, were traffic pollution monitoring stations, urban stations, suburban stations, and background stations. Although the O₃ concentration at the traffic pollution monitoring stations was significantly lower than that of the other three types of monitoring stations, the O_X concentration for the traffic pollution monitoring stations was higher than those of the other three types of monitoring stations, due to their high NO₂ concentration.

From 2016 to 2019, the AOC of the four seasons in descending order were summer, spring, autumn, and winter, and the O_X concentrations in spring, summer, autumn, and winter were 130, 119, 88, and 86 μg/m³ at urban stations, 129, 116, 84, and 81 μg/m³ at suburban stations, 137, 123, 98, and 88 μg/m³ at traffic pollution monitoring stations, and 114, 113, 75, and 79 μg/m³ at background stations, respectively. As shown in Table 1, the O₃/O_X ratio was 0.74 in summer, which was significantly higher than the annual mean value. The correlation coefficient between O₃ and O_X was up to 0.96, indicating that the AOC was mainly affected by O₃ concentration in summer, which could significantly increase the level of O_X concentration in summer. The O_X concentration was lowest in winter, which indicated that the AOC was low in winter. The NO₂/O_X ratio was higher (0.36) than that in winter, and the correlation coefficient of NO₂ and O_X was up to 0.81, indicating that the change in O_X concentration in winter was mainly caused by the change in NO₂ concentration; NO₂ had a more important effect on the AOC value in winter.

3.1.2. Correlation between O_X and PM_2.5 Concentrations

In order to explore the distribution relationship between O_X and PM_2.5 concentrations, different colors were used in Figure 4 to show the distribution characteristics of O_X and PM_2.5 mass concentrations for four seasons. As shown in Figure 4, PM_2.5 and O_X were positively correlated in all four seasons, and PM_2.5 concentration increased with the increase in O_X concentration, especially in winter, when the slope of PM_2.5 and O_X was 0.18, indicating that the increase in PM_2.5 concentration in winter was more sensitive to changes in O_X concentration. This was mainly due to the higher NO₂/O_X ratio in winter, while the increase in PM_2.5 was mainly affected by the change in NO₂ concentration. Conversely, the slope of PM_2.5 and O_X in summer was 0.8, indicating that the increase in PM_2.5 in summer was less sensitive to changes in O_X concentration. However, due to the large fluctuation of O_X concentration in summer, when the O_X concentration was significantly higher than that in other seasons, the stronger AOC in summer also made a great contribution to the increase in PM_2.5.

PM_2.5 showed a good correlation with O_X in winter with a correlation coefficient of 0.61, followed, in summer, by a correlation coefficient of 0.39. The correlation coefficient between PM_2.5 and O_X in spring was 0.36. The lowest correlation coefficient was 0.24 in autumn. PM_2.5 showed a positive correlation with O_X in the four seasons, especially in summer and winter, indicating that the enhancement of AOC could promote an increase in PM_2.5 concentration, to a certain extent.

3.2. Evaluation of the Effect of Atmospheric Photochemical Reaction Activity on Atmospheric Secondary Particle Generation

3.2.1. Classification of Photochemical Reaction Activity

Figure 5 shows the correlation between CO and the daily mean mass concentration of PM₁₀ and PM_2.5 at different photochemical reaction activity levels. When MDA8 O₃ ≤ 100 μg/m³, there was a good linear relationship between CO and PM_2.5 and PM₁₀; the correlation coefficients were 0.88 and 0.72, respectively. However, when MDA8 O₃ was 100–160 μg/m³, 160–200 μg/m³, and greater than 200 μg/m³, the correlation between CO and PM₁₀ and PM_2.5 decreased, especially the correlation between CO and PM₁₀, which decreased significantly to 0.45, 0.41, and 0.39, respectively, along with the increase in O₃ concentration. This indicated that when the MDA8 O₃ value was less than 100 μg/m³, the photochemical reaction activity was less marked, and the observed particle concentration was mainly contributed by the particles from the primary emission. This only contained a few particles from the secondary generation, resulting in a significant correlation between CO, PM₁₀, and PM_2.5. The increase in MDA8 O₃ concentration also indicated that the photochemical reaction activity increased, which led to the generation of more secondary particles and further promoted the increase in PM₁₀ and PM_2.5 concentrations, resulting in the deterioration of the correlation between CO and PM₁₀ and PM_2.5. In addition, when MDA8 O₃ ≤ 100 μg/m³, the intercept of the regression line was low, indicating that CO and PM_2.5 had a certain homology, and the levels drawn from other sources were relatively low. In summary, when the photochemical reaction activity was low, CO showed a good correlation with PM₁₀ and PM_2.5, respectively. In this study, the linear fitting of PM_2.5 and CO was better than that of PM₁₀ and CO, and the ratio of PM_2.5/PM₁₀ from 2016 to 2019 was 72.05%, 62.18%, 61.5%, and 59.36%, respectively. Therefore, when evaluating the influence of the photochemical reaction activity on the generation of secondary particles, the ratio of PM_2.5/CO was more suitable for estimating the generation of secondary particles than PM₁₀/CO. The PM_2.5/CO ratio was selected as the tracer and used to evaluate secondary particle generation.

3.2.2. Effect of Different Photochemical Reaction Activities on the Generation of Atmospheric Secondary Particles

Based on the classification of photochemical reaction activity, the diurnal concentration variation curves of primary PM_2.5 emissions under different O₃ concentration ranges were calculated. In Figure 6, the values R1: 100 < MDA8 O₃ ≤ 160 μg/m³, R2: 160 < MDA8 O₃ ≤ 200 μg/m³, and R3: MDA8 O₃ > 200 μg/m³ correspond to the PM_2.5 mass concentration from primary emissions, at (PM_2.5)_p,G,h, (PM_2.5)_p,L,h, and (PM_2.5)_p,M and _H,h, respectively. As shown in Figure 6, the curves of the primary PM_2.5 emission concentrations corresponding to the three photochemical reaction activities had a similar trend, which rose rapidly after 05:00 and continued to peak at around 09:00, mainly because Beijing had a “morning peak” period between 06:00 and 09:00. The large volume of traffic flow led to a rapid increase in PM_2.5 emission concentration. After 09:00, primary PM_2.5 emissions began to decline gradually and then began to rise again around 17:00, during the “evening peak”. The primary PM_2.5 emission concentration increased with photochemical reaction activity, indicating that the primary PM_2.5 emission was overestimated when the photochemical reaction activity was strong. Therefore, the difference between the diurnal concentration variation curves of the primary PM_2.5 emission, when the photochemical reaction activity was R1, R2, and R3, and the curve when the photochemical reaction activity was R0 are a result of the overestimation of primary PM_2.5 emission, when the photochemical reaction activity was R1, R2, and R3, respectively. These differences were added to the calculation of secondary PM_2.5 generation concentrations to reduce errors in the estimation of secondary PM_2.5 generation.

Figure 7 shows the diurnal variation of the estimated mass concentration of secondary PM_2.5 generation under different photochemical reaction activity levels in Beijing from 2016 to 2019. As shown in Figure 7, the diurnal variation trends of the mass concentration of secondary PM_2.5 generation, corresponding to the three types of photochemical reaction activities R1, R2, and R3, were quite different. When the photochemical reaction activity was R1, 100 < MDA8 O₃ ≤ 160 μg/m³, the secondary PM_2.5 generation peaked at 15.14 μg/m³ at 04:00 and reached its minimum of 11.1 μg/m³ at 07:00. Then, it gradually rose until 17:00, when it reached its maximum of 17.72 μg/m³. Subsequently, it decreased slightly. When the photochemical reaction activity was R2, 160 < MDA8 O₃ ≤ 200 μg/m³, the secondary PM_2.5 generation peaked at 23.33 μg/m³ at 05:00 and reached a trough value of 19.88 μg/m³ at 07:00. Then, it began to rise rapidly, reached its maximum of 28.66 μg/m³ at 11:00, and then began to decline gradually. It continued to decline until 21:00 and reached its minimum of 14.77 μg/m³. When the photochemical reaction activity was R3, MDA8 O₃ > 200 μg/m³, secondary PM_2.5 generation began to rise from 00:00 until 03:00–07:00 and then remained stable at about 27 μg/m³. After 08:00, it began to rise rapidly and reached the maximum of 36.33 μg/m³ at 11:00. Then, it gradually decreased and continued to reach the minimum of 16 μg/m³ at 21:00. In conclusion, the stronger the photochemical reaction activity, the greater the secondary PM_2.5 generation. Secondary PM_2.5 generation increased rapidly with the increase in photochemical reaction activity in the daytime and reached the maximum value at 17:00, 11:00, and 11:00, respectively, under different photochemical reaction activity levels of R1, R2, and R3. In addition, the secondary PM_2.5 generation at night was much lower than that in the daytime, indicating that the strong photochemical reaction activity and AOC in the daytime were more conducive to the generation of secondary particles.

Figure 8 shows the mass concentrations and proportions of PM_2.5 from primary emission and secondary generation at different photochemical reaction activity levels during the whole study period. Meanwhile, in order to understand the proportion of secondary particles under strong AOC, the estimated values in summer, when the atmospheric oxidation was strongest, are shown separately in Figure 8b. The results showed that from 2016 to 2019, the secondary PM_2.5 generation corresponding to the photochemical reaction activities of R1, R2, and R3 accounted for 26%, 35%, and 42% of the total PM_2.5 mass concentration, respectively. This indicates that the stronger the atmospheric photochemical reaction activity, the more numerous the secondary particles generated, and the higher the proportion. Strong photochemical reactions in the atmosphere could promote the oxidation process of precursor substances such as SO₂, NO₂, and VOCs, and further generate SIA and SOA, thus promoting the generation of secondary particles. Especially in summer, the increase in AOC had a more obvious promoting effect on secondary particle generation. Intense solar radiation and higher temperatures in summer increased the AOC; therefore, the increased AOC could promote the formation of secondary particles. In summer, the secondary PM_2.5 generation corresponding to the photochemical reaction activities of R1, R2, and R3 accounted for 26%, 39%, and 49%, respectively. Stronger AOC values in summer led to the formation of more secondary particles, which might be the main factor behind the positive correlation between O₃ and PM_2.5 in summer.

3.3. Effect of AOC on the Formation of Secondary Inorganic Components in Atmospheric Particulate Matter

3.3.1. Composition Characteristics of Water-Soluble Inorganic Ions in PM_2.5

From 2016 to 2019, the mass concentrations of water-soluble inorganic ions in PM_2.5 at the enhanced observation site were as follows, from high to low: NO₃⁻, SO₄²⁻, NH₄⁺, Na⁺, Ca²⁺, Cl⁻, K⁺, Mg²⁺, and F⁻. The most important anion component was NO₃⁻, accounting for 40.69% of the total ions, followed by SO₄²⁻ (21.95%), Cl⁻ (3.37%), and F⁻ (0.37%). NH₄⁺ was the most important cationic component, accounting for 14.46% of the total ions, followed by Na⁺ (9.53%), Ca²⁺ (6.66%), K⁺ (2.11%), and Mg²⁺ (0.86%). SNA were the most important secondary water-soluble inorganic ions in PM_2.5 and accounted for 50.7–94.4% of the total mass concentration of water-soluble inorganic ions, indicating that SIA represent important components of PM_2.5.

The composition characteristics of PM_2.5 inorganic ions were different between the daytime and nighttime. The composition characteristics of water-soluble inorganic ions of PM_2.5 in the daytime and nighttime are shown in Figure 9. The concentration of water-soluble inorganic ions in PM_2.5 in the daytime was significantly higher than that at night, and the sums of mass concentrations of water-soluble inorganic ions in PM_2.5 in the daytime and nighttime were 34.85 μg/m³ and 30.36 μg/m³, respectively. The concentrations of 9 kinds of water-soluble ions in PM_2.5 in the daytime were all higher than those at night, especially the mass concentrations of SNA. This indicates that, compared with the levels at night, the generation of SNA in PM_2.5 was higher in the daytime, with higher photochemical reaction activity and AOC levels.

3.3.2. Sulfur and Nitrogen Conversion Rates of SO₂ and NO₂ in the Atmosphere

The oxidation pathways of SO₂ and NO₂ in the atmosphere are mainly a gaseous oxidation reaction and heterogeneous oxidation reaction. These two pathways are affected by atmospheric oxidants, such as OH, O₃, H₂O₂, and NO₃. Therefore, an increase in AOC can promote the oxidation reaction between SO₂ and NO₂, resulting in an increase in SOR and NOR values. Figure 10 shows the SOR and NOR distribution characteristics in the daytime and at nighttime during the observation period. The values for SOR (0.61) and NOR (0.26) in the daytime were higher than for SOR (0.55) and NOR (0.16) at nighttime, indicating that the conversion rates of SO₂ and NO₂ to SO₄²⁻ and NO₃⁻ were higher than those at nighttime. This was because the AOC was enhanced by active photochemical reactivity in the daytime; thus, the oxidation reaction of SO₂ and NO₂ was promoted. Compared with other domestic cities, the SOR value of Beijing from July to September was higher than that of Shanghai, Xi’an, and Taiyuan; the NOR value was higher than that of Shanghai and Taiyuan, but lower than that of Jiaozuo, Puyang, and Xi’an (Table 2).

3.3.3. Correlation between O_X and Secondary Inorganic Salts, SOR, and NOR in PM_2.5

SO₄²⁻ and NO₃⁻ are mainly generated through the secondary reaction of the gaseous pollutants SO₂ and NO₂, while the oxidation efficiency of SO₂ and NO₂ is related to the AOC [26,27]. The correlation coefficients of PM_2.5, O_X, SOR, NOR, and SNA were calculated, as shown in Table 3. PM_2.5 was positively correlated with O_X, SOR, NOR, and SNA. O_X was positively correlated with SOR and NOR, with correlation coefficients of 0.25 and 0.36, respectively (Figure 11). O_X was positively correlated with SO₄²⁻, NO₃⁻, and NH₄⁺, and the correlation coefficients were 0.35, 0.41, and 0.32, respectively, indicating that strong AOC could improve the conversion rates of SO₂ and NO₂ and, thus, promote the generation of SNA.

The coefficient of determination (R²) between SOR and O_X reached 0.91, while that between NOR and O_X was only 0.15. This may be due to the fact that nitrate generation during the observation period may not only have been affected by AOC but may also have been greatly affected by environmental factors, such as temperature and humidity. In contrast, environmental factors have little influence on sulfate production [4,28]. Due to the simultaneous increase in atmospheric oxidizing capacity, changes in precursor concentration, air temperature, relative humidity, and other conditions may have a certain impact on the generation of sulfate and nitrate. Therefore, the relationship between O_X and SOR, and the relationship between O_X and NOR, were both nonlinear. This study exponentiated the sulfate conversion rate and nitrate conversion rate against the change in O_X concentration. The results showed that when the O_X concentration was less than around 140 μg/m³, SOR and NOR significantly increased with the increase in O_X concentration, indicating that the atmospheric oxidizing capacity had a significant promoting effect on SOR and NOR. However, when the O_X concentration was higher than around 140 μg/m³, the magnitude of the increase in SOR and NOR was smaller, indicating that atmospheric oxidizing capacity may no longer be the determining factor for sulfate and nitrate generation. At this time, the generation of sulfate and nitrate may be influenced by other factors, such as precursor concentration, temperature, and relative humidity [29].

4. Conclusions

From 2016 to 2019, the annual O_X values in Beijing, listed in descending order, were traffic pollution monitoring stations, urban stations, suburban stations, and background stations. In 2019, the annual mean of O_X at all stations in Beijing showed a significant decline. Compared to those in 2018, the AOC values at urban stations, suburban stations, traffic pollution monitoring stations, and background stations decreased by 11, 11, 12, and 13 μg/m³, respectively. The AOC of the four seasons, in descending order, were summer, spring, autumn, and winter. The concentration of O_X was mainly affected by the concentration of O₃ in summer, while the concentration of NO₂ had the most important effect on O_X in winter.

The results of the PM_2.5 secondary generation evaluation showed that from 2016 to 2019, the secondary PM_2.5 generation corresponding to the photochemical reaction activities of R1, R2, and R3 accounted for 26%, 35%, and 42% of the total PM_2.5 mass concentration, respectively. It is indicated that the stronger atmospheric photochemical reaction activity promoted the generation of secondary particles, resulting in their higher proportion in PM_2.5.

Secondary PM_2.5 generation increased rapidly with the increase in photochemical reaction activity in the daytime and reached the maximum values at 17:00, 11:00, and 11:00, respectively, under the different photochemical reaction activity levels of R1, R2, and R3. SNA, SOR, and NOR values in the daytime were higher than those at nighttime, indicating that the daytime, with its higher photochemical activity and greater AOC, was more conducive to SNA generation. O_X showed a positive correlation with SNA, SOR, and NOR, indicating that the generation of SNA in PM_2.5 was promoted when the concentration of O_X was high, thereby leading to an increase in PM_2.5 concentration. When the O_X concentration was less than around 140 μg/m³, SOR and NOR significantly increased with an increase in O_X concentration. However, when the O_X concentration was higher than around 140 μg/m³, the magnitude of the increase in SOR and NOR was smaller, indicating the relationship between O_X and SOR, and the relationship between O_X and NOR, were both nonlinear.

When establishing emission reduction measures for precursors, it is necessary to comprehensively consider the impact of increasing AOC on the generation of secondary components in PM_2.5, so as to further reduce the emission of highly reactive precursors and reduce AOC. The decline in AOC will help achieve PM_2.5 control targets in China in the future.

Author Contributions

Conceptualization, W.C., L.L. and H.L.; data curation, W.C., L.L., Y.Z., X.Y. and Y.J.; formal analysis, W.C. and L.L.; funding acquisition, H.L.; investigation, W.C.; methodology, W.C., L.L. and H.L.; project administration, H.L. and F.C.; resources, Y.Z., Y.C., G.Z., X.Y. and Y.J.; supervision, H.L.; validation, W.C., L.L. and H.L.; visualization, W.C. and L.L.; writing—original draft, W.C. and L.L.; writing—review and editing, H.L. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The observation and modeling data are available from the corresponding author upon reasonable request.

Acknowledgments

The authors acknowledge all funders.

Conflicts of Interest

The authors declare no conflict of interest.

Footnotes

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Figures and Tables

Figure 1. Distribution map of the 35 ambient air quality monitoring stations in Beijing.

Figure 2. Inter-month variations in atmospheric oxidant capacity (OX) in Beijing from 2016 to 2019.

Figure 3. Inter-annual and seasonal changes of atmospheric oxidant capacity (OX) in Beijing from 2016 to 2019.

Figure 4. Mass concentration distribution characteristics of PM2.5 and atmospheric oxidant capacity (OX) in Beijing from 2016 to 2019. The vertical dashed lines in the figure represent different pollution levels: green—excellent; yellow—good; orange—light contamination; red—moderate pollution; purple—heavy pollution.

Figure 5. Linear fitting between the daily mean of PM10, PM2.5, and CO for different photochemical activities in Beijing from 2016 to 2019.

Figure 6. Estimation of the daily variation of primary PM2.5 emissions under different photochemical reaction activity conditions in Beijing, from 2016 to 2019.

Figure 7. Estimation of the daily variations in secondary generation PM2.5 under different photochemical reaction activity conditions in Beijing from 2016 to 2019.

Figure 8. PM2.5 mass concentration and the percentage of primary emission and secondary generation for different photochemical reaction activities in Beijing: (a) 2016–2019, (b) summer 2016–2019. R1: 100 < MDA8 O3 ≤ 160 μg/m3; R2: 160 < MDA8 O3 ≤ 200 μg/m3; R3: MDA8 O3 > 200μg/m3.

Figure 9. Daytime and nighttime PM2.5 water-soluble inorganic ion composition characteristics.

Figure 10. Daytime and nighttime SOR and NOR distribution characteristics.

Figure 11. Exponential fitting of OX concentration to SOR and NOR during the observation period.

Table 1

Ratio and correlation coefficients of NO₂ and O₃ to O_X over four seasons.

Seasons	OX Concentration (μg/m3)	O3/OX	NO2/OX	r(NO2)	r(O3)
Spring	116	0.6	0.4	0.41	0.75
Summer	128	0.74	0.26	0.29	0.96
Autumn	88	0.57	0.43	0.27	0.73
Winter	82	0.64	0.36	0.81	0.35

Note: r(NO₂) and r(O₃) represent Pearson correlation coefficients between NO₂ concentration and O_X concentration, and between O₃ concentration and O_X concentration, respectively.

Table 2

Comparison of SOR and NOR values reported in the literature during the observation period.

Cities	Periods	SOR	NOR	Sources
Beijing	July–September 2019	0.58	0.20	This paper
Jiaozuo	January 2018	0.43	0.35	Wang Liumin et al. [21]
Puyang	October 2017–January 2018	0.22	0.22	Chen Chu et al. [22]
Shanghai	October–November 2010	0.24	0.15	Zhou Min et al. [23]
Xi’an	June–August 2006	0.44	0.32	Shen Zhenxing et al. [24]
Taiyuan	May–June 2013	0.13	0.08	Guo Wendi et al. [25]

Table 3

Correlation coefficients between O_X, SOR, NOR, and SNA.

	PM2.5	OX	SO42−	NO3−	NH4+	SOR	NOR
PM2.5	1	0.46	0.71	0.9	0.86	0.79	0.85
OX		1	0.35	0.41	0.32	0.25	0.36
SO42−			1	0.75	0.78	0.82	0.8
NO3−				1	0.94	0.8	0.91
Ca2+					0.45	0.43	0.56
NH4+					1	0.83	0.89
SOR						1	0.83
NOR							1

References

1. Beijing Municipal Ecology and Environment Bureau. Beijing Ecology and Environment Statement 2019; Beijing Municipal Ecology and Environment Bureau: Beijing, China, 2019.

2. Beijing Municipal Ecology and Environment Bureau. Beijing Ecology and Environment Statement 2016; Beijing Municipal Ecology and Environment Bureau: Beijing, China, 2016.

3. Zang, H.; Zhao, Y.; Huo, J.; Zhao, Q.; Fu, Q.; Duan, Y.; Shao, J.; Huang, C.; An, J.; Xue, L. et al. High atmospheric oxidation capacity drives wintertime nitrate pollution in the eastern Yangtze River Delta of China. Atmos. Chem. Phys.; 2022; 22, pp. 4355-4374. [DOI: https://dx.doi.org/10.5194/acp-22-4355-2022]

4. Tan, Z.; Wang, H.; Lu, K.; Dong, H.; Liu, Y.; Zeng, L.; Hu, M.; Zhang, Y. An Observational Based Modeling of the Surface Layer Particulate Nitrate in the North China Plain during Summertime. JGR Atomspheres; 2021; 126, e2021JD035623. [DOI: https://dx.doi.org/10.1029/2021JD035623]

5. Liu, P.; Ye, C.; Xue, C.; Zhang, C.; Mu, Y.; Sun, X. Formation mechanisms of atmospheric nitrate and sulfate during the winter haze pollution periods in Beijing: Gas-phase, heterogeneous and aqueous-phase chemistry. Atmos. Chem. Phys.; 2020; 20, pp. 4153-4165. [DOI: https://dx.doi.org/10.5194/acp-20-4153-2020]

6. Jia, M.; Zhao, T.; Cheng, X.; Gong, S.; Zhang, X.; Tang, L.; Liu, D.; Wu, X.; Wang, L.; Chen, Y. Inverse Relations of PM2.5 and O₃ in Air Compound Pollution between Cold and Hot Seasons over an Urban Area of East China. Atmosphere; 2017; 8, 59. [DOI: https://dx.doi.org/10.3390/atmos8030059]

7. Zhu, J.; Chen, L.; Liao, H.; Dang, R. Correlations between PM2.5 and Ozone over China and Associated Underlying Reasons. Atmosphere; 2019; 10, 352. [DOI: https://dx.doi.org/10.3390/atmos10070352]

8. Wang, P.; Zhu, S.; Zhang, M.; Shao, T. Atmospheric oxidation capacity and its contribution tosecondary pollutants formation. Chin. Sci. Bull.; 2022; 67, pp. 2069-2078. [DOI: https://dx.doi.org/10.1360/TB-2021-0761]

9. Wen, L.; Xue, L.; Wang, X.; Xu, C.; Chen, T.; Yang, L.; Wang, T.; Zhang, Q.; Wang, W. Summertime fine particulate nitrate pollution in the North China Plain: Increasing trends, formation mechanisms and implications for control policy. Atmos. Chem. Phys.; 2018; 18, pp. 11261-11275. [DOI: https://dx.doi.org/10.5194/acp-18-11261-2018]

10. Kim, J.Y.; Song, C.H.; Ghim, Y.S.; Won, J.G.; Yoon, S.C.; Carmichael, G.R.; Woo, J.H. An investigation on NH3 emissions and particulate NH⁴⁺–NO³⁻ formation in East Asia. Atmos. Environ.; 2006; 40, pp. 2139-2150. [DOI: https://dx.doi.org/10.1016/j.atmosenv.2005.11.048]

11. Feng, T.; Zhao, S.; Bei, N.; Liu, S.; Li, G. Increasing atmospheric oxidizing capacity weakens emission mitigation effort in Beijing during autumn haze events. Chemosphere; 2021; 281, 130855. [DOI: https://dx.doi.org/10.1016/j.chemosphere.2021.130855]

12. Xie, X.; Hu, J.; Qin, M.; Guo, S.; Hu, M.; Wang, H.; Lou, S.; Li, J.; Sun, J.; Li, X. et al. Modeling particulate nitrate in China: Current findings and future directions. Environ. Int.; 2022; 166, 107369. [DOI: https://dx.doi.org/10.1016/j.envint.2022.107369]

13. Peng, J.; Hu, M.; Shang, D.; Wu, Z.; Du, Z.; Tan, T.; Wang, Y.; Zhang, F.; Zhang, R. Explosive Secondary Aerosol Formation during Severe Haze in the North China Plain. Environ. Sci. Technol.; 2021; 55, pp. 2189-2207. [DOI: https://dx.doi.org/10.1021/acs.est.0c07204]

14. Ministry of Ecology and Environment of the People’s Republic of China. Automated Methods for Ambient Air Quality Monitoring; Ministry of Ecology and Environment of the People’s Republic of China: Beijing, China, 2005.

15. Ministry of Ecology and Environment of the People’s Republic of China. Specifications and Test Procedures for Ambient Air Quality Continuous Automated Monitoring System for PM10 and PM2.5 (HJ 653–2021); Ministry of Ecology and Environment of the People’s Republic of China: Beijing, China, 2021.

16. Ministry of Ecology and Environment of the People’s Republic of China. Ambient Air Quality Standards (GB 3095-2012); Ministry of Ecology and Environment of the People’s Republic of China: Beijing, China, 2012.

17. Ministry of Ecology and Environment of the People’s Republic of China. Technical Regulation for Ambient Air Quality Assessment (On Trial) (HJ 663-2013); Ministry of Ecology and Environment of the People’s Republic of China: Beijing, China, 2013.

18. Chang, S.-C.; Lee, C.-T. Secondary aerosol formation through photochemical reactions estimated by using air quality monitoring data in Taipei City from 1994 to 2003. Atmos. Environ.; 2007; 41, pp. 4002-4017. [DOI: https://dx.doi.org/10.1016/j.atmosenv.2007.01.040]

19. Wei, W.; Wang, Y.; Bai, H.; Wang, X.; Cheng, S.; Wang, L. Insights into atmospheric oxidation capacity and its impact on PM2.5 in megacity Beijing via volatile organic compounds measurements. Atmos. Res.; 2021; 258, 105632. [DOI: https://dx.doi.org/10.1016/j.atmosres.2021.105632]

20. Lu, K.; Fuchs, H.; Hofzumahaus, A.; Tan, Z.; Wang, H.; Zhang, L.; Schmitt, S.H.; Rohrer, F.; Bohn, B.; Broch, S. et al. Fast Photochemistry in Wintertime Haze: Consequences for Pollution Mitigation Strategies. Environ. Sci. Technol.; 2019; 53, pp. 10676-10684. [DOI: https://dx.doi.org/10.1021/acs.est.9b02422]

21. Wang, L.; Wang, X.; Wang, M.; Yu, G. Spatial and Temporal Distribution and Potential Source of Atmospheric Pollution in Jiaozuo City. Res. Environ. Sci.; 2020; 33, pp. 820-830. [DOI: https://dx.doi.org/10.13198/j.issn.1001-6929.2019.04.22]

22. Chen, C.; Wang, T.; Li, Y.; Ma, H.; Chen, P.; Wang, D.; Zhang, Y.; Qiao, Q.; Li, G.; Wang, W. Pollution Characteristics and Source Apportionment of Fine Particulate Matter in Autumn and Winter in Puyang, China. Environ. Sci.; 2019; 40, pp. 3421-3430. [DOI: https://dx.doi.org/10.13227/j.hjkx.201901119]

23. Zhou, M.; Chen, C.; Wang, H.; Huang, C.; Su, L.; Chen, Y.; Li, L.; Qiao, Y.; Chen, M.; Huang, H. et al. Chemical characteristics of particulate matters during air pollution episodes in autumn of Shanghai, China. Acta Sci. Circumstantiae; 2012; 32, pp. 81-92. [DOI: https://dx.doi.org/10.13671/j.hjkxxb.2012.01.019]

24. Shen, Z.; Huo, Z.; Han, Y.; Cao, J.; Zhao, J.; Zhang, T. Chemical Composition of Water-Soluble Ions in Aerosols over Xi’an in Heating and Non-Heating Seasons. Plateau Meteorol.; 2009; 28, pp. 151-158.

25. Guo, W.; Wang, K.; Guo, X.; Yan, Y. Characteristics of sulfur and nitrogen conversion in the aerosol, Taiyuan. Environ. Chem.; 2016; 35, pp. 11-17. [DOI: https://dx.doi.org/10.7524/j.issn.0254-6108.2016.01.2015060305]

26. Wang, Z.; Li, Y.; Chen, T.; Zhang, D.; Sun, F.; Wang, X.; Huan, N.; Pan, L. Analysis on diurnal variation characteristics of ozone and correlations with its precursors in urban atmosphere of Beijing. China Environ. Sci.; 2014; 34, pp. 3001-3008.

27. Huang, X.-F.; Cao, L.-M.; Tian, X.-D.; Zhu, Q.; Saikawa, E.; Lin, L.-L.; Cheng, Y.; He, L.-Y.; Hu, M.; Zhang, Y.-H. et al. Critical Role of Simultaneous Reduction of Atmospheric Odd Oxygen for Winter Haze Mitigation. Environ. Sci. Technol.; 2021; 55, pp. 11557-11567. [DOI: https://dx.doi.org/10.1021/acs.est.1c03421]

28. Xue, J.; Yuan, Z.; Griffith, S.M.; Yu, X.; Lau, A.K.H.; Yu, J.Z. Sulfate Formation Enhanced by a Cocktail of High NOx, SO₂, Particulate Matter, and Droplet pH during Haze-Fog Events in Megacities in China: An Observation-Based Modeling Investigation. Environ. Sci. Technol.; 2016; 50, pp. 7325-7334. [DOI: https://dx.doi.org/10.1021/acs.est.6b00768]

29. Li, K.; Jacob, D.J.; Liao, H.; Zhu, J.; Shah, V.; Shen, L.; Bates, K.H.; Zhang, Q.; Zhai, S. A two-pollutant strategy for improving ozone and particulate air quality in China. Nat. Geosci.; 2019; 12, pp. 906-910. [DOI: https://dx.doi.org/10.1038/s41561-019-0464-x]