Contribution of Atmospheric Diffusion Conditions to the Recent Improvement in Air Quality in China

This study analyzed hourly mass concentration observations of PM2.5 (particulate matters with diameter less than 2.5 μm) at 512 stations in China from December 2013 to May 2015. We found that the mean concentrations of PM2.5 during the winter and spring of 2015 Dec. 2014 to Feb. 2015 and Mar. 2015 to May 2015) decreased by 20% and 14% compared to the previous year, respectively. Hazardous air-quality days decreased by 11% in 2015 winter, with more frequent good to unhealthy days; and the good and moderate air-quality days in 2015 spring increased by 9% corresponding to the less occurrence of unhealthy conditions. We compared the atmospheric diffusion conditions during these two years and quantified its contribution to the improvement of air quality during the first half of 2015 over China. Our results show that during the 2015 winter and spring, 70% and 57% of the 512 stations experienced more favorable atmospheric diffusion conditions compared to those of previous year. Over central and northern China, approximately 40% of the total decrease in PM2.5 during the 2015 winter can be attributed to the favorable atmospheric diffusion conditions. The atmospheric diffusion conditions during the spring of 2015 were not as favorable as in winter; and the average contributions of the atmospheric conditions were slight.

shows the seasonal mean PM 2.5 mass concentrations during the winters and springs of 2015 and 2014 and their relative differences. A higher seasonal PM 2.5 concentration occurred at the intersection region of Hebei, Shandong, Shangxi and Henan Provinces, particularly during the 2014 winter, with better air quality in the east coastal area than in the inland regions, and better quality in the spring than in the winter. The PM 2.5 concentrations decreased considerably in the first half of 2015 compared with the same period of 2014, by 20% and 14% during the winter and spring, respectively (summarized in Table 1). A total of 88.5% stations reported decreased PM 2.5 concentrations during the winter of 2015, and 86.9% stations for the 2015 spring compared to that of 2014.
The occurrence of different air-quality levels is shown in Fig. 2. Six categories of air-quality levels (as shown in Table 2 and Table S1) are defined based on the air-quality index (AQI) of the U.S. Environmental Protection Agency 40 ; we combined them into three major levels with increasing air pollution conditions: good and moderate, unhealthy for sensitive and unhealthy, and very unhealthy and hazardous group. Compared with the winter of 2014, it showed a substantial decrease in hazardous air quality occurrences in 2015 winter but more frequent unhealthy and good air-quality occurrences, indicating that the decreased seasonal mean PM 2.5 concentration was primarily due to the infrequency of severely polluted days. Approximately 40% of the stations did not record hazardous air quality in both spring seasons. As summarized in Table 2, good and moderate air-quality occurrences were 11% more frequent in the spring of 2015, corresponding to a 9% decrease in unhealthy condition occurrences.  Table 1. This figure was produced by Matlab version 7.13 (http://www. mathworks.com/products). Contribution of Atmospheric Diffusion Conditions to the Improved Air Quality. Air pollutants emissions from a particular area do not typically change over a short period, but local meteorological patterns can strongly affect the accumulation of air pollutants via removal and transport. Pollution episodes are usually related to the presence of air stagnation conditions with poor atmospheric horizontal transport or vertical diffusion abilities 21,34,[41][42][43][44] . Another important factor for contaminant scavenging is wet deposition due to precipitation [45][46][47] .  Similarly, the occurrence of good and moderate air quality increased in the spring of 2015 due to the decrease in unhealthy conditions. Stations without specific air-quality levels in both years are not shown. The statistical results of this figure are summarized in Table 2. This figure was produced by Matlab version 7.13 (http://www. mathworks.com/products). Figure 3 shows the variation of PM 2.5 and the simultaneous atmospheric conditions at a Beijing station from Oct. 1 to Dec. 15, 2014, during which the PM 2.5 concentration could dramatically increased from below 10 μ gm −3 to approximate 400 μ gm −3 . Precipitation always tends toward the deposition of pollutants with a lower PM 2.5 concentration during rainy days. Sustained weak surface wind and a shallow atmospheric boundary layer can trigger an air pollution episode, which can be broken by heavy surface wind or a well-developed boundary layer. Air stagnation is typically used to describe the atmospheric capability for air pollutant diffusion, which involves horizontal transport, vertical diffusion and wet deposition; however, air stagnation criteria are sensitive to local meteorological conditions, and existing air stagnation definitions are empirical 43,46 . Therefore, we proposed a quantitative criterion to identify the occurrence of air stagnation condition based on the daily boundary layer height (BLH), surface wind speed and the present of precipitation (as shown in Fig. 4, more detailed information in Method). Daily surface wind speed and boundary layer height were used to indicate the horizontal and vertical diffusion ability of the atmosphere, and the present of precipitation is considered to be benefit to the wet deposition by default. Figure 4 shows that when there is no effective precipitation, near-surface PM 2.5 concentrations decreased with the increase of surface wind speed and BLH, i.e., shallow boundary layer and weak surface wind restricts the diffusion of near-surface pollutants. Thus, a day was considered as air stagnation day when the weak surface wind (2.7 ms −1 for winter and 3 ms −1 for spring) companied with shallow BLH (350 m for winter and 450 m for spring) and no-precipitation. As shown in Fig. 3, persistent air stagnation conditions can lead to air pollution episodes; a stagnation event was claimed to last at least 3 days in this study (the sensitivity of air stagnation event duration was discussed in the SI). Figure 5 shows the relative difference in PM 2.5 concentrations between stagnation and no-stagnation periods, which considered as the effect of the air stagnation event on air pollutant accumulation. The effect of air stagnation on PM 2.5 was greater in the high-pollution scenario, i.e., stronger in 2014 than in 2015, and greater in winter than in spring due to the more frequent severe pollution occurrences, as shown in Fig. 2. PM 2.5 concentrations during stagnation events were 49% higher than during no-stagnation events during the winter of 2014 and 40% higher for the winter of 2015 (Table 3). Air stagnation effects were relatively weaker during the springtime, with

Table 2. Comparison of the occurrence of different air-quality levels between 2014 and 2015.
A positive value indicates a more frequent occurrence of the specific air quality in 2015 than in 2014 (detailed thresholds of each air quality level were summarized in Table S1). a 26-32% higher PM 2.5 concentration during stagnation events than that in no-stagnation period. The difference of PM 2.5 concentrations between air stagnation and no-stagnation period was small and even reversed at some stations over northwestern China during the springtime, which may be due to the increased likelihood of sand and dust storms with heavy winds in the arid and semi-arid regions in no-stagnation events. PM 2.5 accumulates on air stagnation days and exacerbates local pollution. More frequent air stagnation leads to worsening air quality under similar emission conditions, i.e., lower air-stagnation frequency indicates better atmospheric diffusion ability. Figure 6 exhibits the occurrence frequency of air stagnation events during the study periods, i.e., the proportion of the total number of days during all the stagnation events in the available seasonal records. It shows more frequent air stagnation events during winter than spring. The higher frequency of air stagnation over inland China contrasts with the infrequent stagnations along the coastlines during winter, which may be due to a heavy land-sea breeze. The spatial difference in air stagnation occurrences was slight during the spring, with a relatively higher frequency over Xinjiang, Gansu and Sichuan provinces.
There were significantly fewer air stagnation events during the winter of 2015 than that of 2014, with 70% of the stations showing lower air stagnation frequency in the winter of 2015 (shown in Fig. 6 and Table 3). However, in spring, only 57% of the stations displayed a better atmospheric diffusion environment in 2015 compared with the previous year, and these stations were primarily located over northern and northeastern China. In addition, the magnitude of decreased air stagnation frequency was greater in winter than in spring (Fig. 5). The greater reduction of air stagnation occurrences over the majority of stations during the winter of 2015 may led to the relatively stronger improvement in air quality during winter than spring (20% decrease in PM 2.5 for winter and 14% decrease for spring, as shown in Fig. 1).
We used the following method to quantify the contribution of changed atmospheric diffusion conditions to the improvement of air quality in China:  air stagnation or no-stagnation events in 2014. Seasonal concentrations and frequencies during winter and spring were used separately. We compared the relative difference of C if2015 to the actual observations in 2015 to estimate the variation of PM 2.5 due to the different atmospheric diffusion condition in 2015:  Table 3. Results for air stagnation events, defined as the presence of an air stagnation day (at least one stagnation day), are shown in Fig. S1. This figure was produced by Matlab version 7.13 (http://www.mathworks.com/products).    (Fig. 6). Overall, the average contribution of atmospheric conditions was slight throughout the entire country.

Conclusions and Discussion
China's overall air quality significantly improved during the winter and spring of 2015 compared with that of 2014, with the near-surface PM 2.5 concentrations decreased by 20% and 14%, respectively. The occurrences of very unhealthy and hazardous air quality decreased by 11%, which corresponds to a 6% more frequent normal unhealthy conditions in the winter of 2015 compared with 2014. In the spring of 2015, good and moderate air-quality days occurred 11% more frequently than in the same period of the previous year, with a corresponding infrequency of unhealthy conditions.  Table 3. This figure was produced by Matlab version 7.13 (http://www.mathworks.com/products).
Scientific RepoRts | 6:36404 | DOI: 10.1038/srep36404 A seasonal dependence of PM 2.5 concentrations on surface winds, boundary layer height and the occurrence of precipitation was established and used to define the presence of air stagnation conditions. Air stagnation events were defined as comprising at least 3 continuous air stagnation days. It showed higher PM 2.5 concentrations during stagnation periods compared with no-stagnation events, particularly in the winter of 2014, with 49% higher PM 2.5 concentration during air stagnation periods. Lower frequency of air stagnation events occurred at 70% and 57% of stations during the 2015 winter and spring, respectively, compared with the same periods of the previous year. Compared with the spring of 2015, the more significant improvement in air quality during the winter because the reduced air stagnation event frequency was observed at more stations in winter of 2015. The favorable atmospheric diffusion conditions contributed to approximately 40% of the total decreased concentrations during the winter of 2015 compared with the same period of the previous year over central and northern China (Hebei, Henan, Shandong and Shanxi Provinces). The atmospheric diffusion ability during the spring of 2015 was not as favorable as in the winter, and the average contribution of atmospheric conditions was slight throughout the country.
Significant northerly wind anomalies occurred over northern China in 2015 compared with the previous year (as shown in Fig. 8), providing cleaner air and strengthening the horizontal transport of highly polluted regions. The stronger and northerly wind field anomalies could be attributed to the strong El Niño that developed in 2015 13,48 . This study only considered the physical removal of meteorological processes; however, the specific atmospheric conditions in a similar atmospheric diffusion condition could be different (e.g., temperature and humidity) which can effect certain chemical reactions during the formation of secondary aerosol 1,22,49,50 . In addition, high concentrations of surface air pollutants can locally reduce the amount of solar radiation reaching the earth's surface; thus suppressing the development of an atmospheric boundary layer. The consequent shallow boundary layer height increases surface PM 2.5 concentrations by compressing their diffusion volume 51,52 . Moreover, the increase in absorptive aerosols can enhance the atmospheric stability, thus impairing the downward transport of momentum and leading to worsened diffusion conditions with low-velocity surface winds 53,54 . The contribution of meteorological conditions to the chemical reaction of aerosol and the feedback between air pollutants and boundary layer processes are complicated and require further exploration.

Data and Method
Data. The real-time hourly PM 2.5 concentration dataset has been continuously recorded by the Chinese Ministry of Environmental Protection and made publicly accessible since January 2013. The mass concentrations of PM 2.5 are measured using either the micro-oscillating balance method or the β absorption method with commercial instruments. The instrumental operation, maintenance, data assurance and quality control are conducted according to HJ 655-201 Standards 5 . Some meteorological variables involved in this study (i.e., boundary layer height and surface wind) were obtained from a 0.25°*0.25° daily four-time ERA-Interim reanalysis dataset. Precipitation events were defined as the presence of rain, hail or snow rather than the daily precipitation amounts because of its high number of missing values over China, based on the US National Climate Data Center (NCDC) Global Surface Summary of the Day (GSOD) database.
Method to quantify the criterion of air stagnation condition. Figure 3 shows that the sustained weak surface wind and a shallow boundary layer can trigger an air pollution episode, which can be broken by heavy surface wind or a well-developed boundary layer. Daily surface wind speed and boundary layer height were used to indicate the horizontal and vertical diffusion ability of the atmosphere. Daily PM 2.5 concentrations were normalized by the monthly mean of the specific year to avoid the effects of spatial and temporal variation. All of the precipitation events were excluded in the dependence relationship to rule out the effect of wet deposition, i.e., the air stagnation condition will be terminated only if it rains. In general, near-surface PM 2.5 concentrations decreased with the increase of surface wind and BLH. A shallow boundary layer (350 m for winter and 450 m for spring) restricts the diffusion of near-surface pollutants. And weak surface wind (2.7 ms −1 for winter and 3 ms −1 for spring) is favorable for the accumulation of air pollutants; however, this stagnation condition can be disrupted by a well-developed boundary layer during the winter. Atmospheric conditions in which the average daily normalized PM 2.5 concentrations exceeded 100% were considered as air stagnation conditions, as shown in the lower-left corner of the dashed line in Fig. 4. All rainy days were defined as no-stagnation days because of the wet deposition affect.
A day can be designated as a stagnation day or a no-stagnation day based on its specific surface wind, BLH and the presence of precipitation. A stagnation event was claimed to last at least 3 days in this study, and the differences of PM 2.5 concentrations between air stagnation and no-stagnation events were compared to estimate the effect of air stagnation event on air pollutant diffusion. Sensitivity of air stagnation event duration was discussed in the SI, which indicated that the one-day air stagnation showed the same pattern with that of three-day air stagnation events, but with a small magnitude of impact on air pollutant dispersion.