PM10 Observed at a Meteorological Station in Beijing: Historical Trend and Implications

Particulate matter (PM) suspended in the atmosphere and having aerodynamic diameters of 10 μm or smaller are generally coded PM₁₀. Particulate matter contains microscopic solids or liquid droplets that are so small that can be inhaled and cause short-term and long-term health problems (Englert ^[1]; Chu et al. ^[2]; Zhang et al. ^[3]). Fine particles are also the main cause of reduced visibility and haze (Trijonis ^[4]; Chen et al. ^[5]; Song et al. ^[6]). These particles come in many sizes and shapes and can be made up of hundreds of different chemicals, which are emitted directly from sources (primary particles) or formed in the atmosphere as a result of complex reactions of gases (secondary particles) such as sulfur dioxide (SO₂), nitrogen oxides (NO_X), and certain organic compounds (Heintzenberg ^[7]; Bernd et al. ^[8]; Lu et al. ^[9]; Lu et al. ^[10]). Meteorological conditions like wind, temperature, ventilation, and height of the boundary layer are key factor affecting the accumulation, transport, dilution and sinking of particulate (Haagenson ^[11]; Kassomenos et al. ^[12]; Guo et al. ^[13]; Nan et al. ^[14]; Wu et al. ^[15]), and the change of meteorological conditions under different synoptic systems could influence the inter-annual, seasonal, and monthly atmospheric characteristics of atmospheric pollution (Fan et al. ^[16]; Wei et al. ^[17]; Zhang and Niu ^[18]; Ren et al. ^[19]; An et al. ^[20]).

As a developing megacity, Beijing has experienced a rapid development in the past two decades. The great economic prosperity and urban growth have also resulted in the deterioration of the city's air quality. Among which, severe atmospheric particles matter pollutions are derived from both natural and anthropogenic activities (Guo et al. ^[21]; Sun et al. ^[22]; Zheng et al. ^[23]). PM pollution in Beijing has become a serious concern of the government and the public (Zhang et al. ^[24]; Shi et al. ^[25]; Hou et al. ^[26]; Zhang et al. ^[27]; Wang et al. ^[28]; Lv et al. ^[29]; Song et al. ^[30]; Wang and Li ^[31]). To tackle severe air pollution, Beijing has launched comprehensive air pollution control programs in phases since 1998 (Xie ^[32]; Ji et al. ^[33]). Beijing has continuously promoted end-of-pipe control and energy structure adjustment over the past 20 years. Coal combustion in power plants, boilers, and residential use were controlled simultaneously with achieved remarkable progress (Zhao et al. ^[34]; Yang et al. ^[35]). For the prevention and control of vehicle pollution, Beijing has constantly implemented a series of local emission standards and comprehensive control measures, as well as strengthened traffic management and economic incentives. Beijing has implemented an intensive pollution control during 2013-2017 under the"Action Plan on Prevention and Control of Air Pollution"and the implementation has led to an unprecedented air quality improvement (Vu et al. ^[36]; Chen et al. ^[37]; Xue et al. ^[38]). With the constant efforts in air pollution control, emission intensity has decreased year by year and air quality has improved significantly. On-ground observational data shows that the annual average concentrations of SO₂, NO₂ and PM₁₀ decreased greatly (UN Environment Programme ^[39]; Huang and Wang ^[40]).

The dramatic decline of PM₁₀ mass concentration was detected in Beijing after the 2008 Olympics, with the average rate of about 4 μg m^-3 per year during 2002-2016 (Shi et al. ^[25]). PM₁₀ concentration was the highest in spring, followed by winter, and the lowest in summer. The annual mean PM₁₀ concentrations in Beijing were high in March, April, May, November, and December, while relatively low in July, August, and September during 2008-2012 (Hou et al. ^[26]). The variation in air quality was significantly influenced by meteorological conditions. There was a significant correlation between PM₁₀ mass concentration and wind speed, relative humidity, and surface air temperature. Besides, the correlation varies with seasons and year-to-year (You et al. ^[41]; Guo et al. ^[13]; Zhang et al. ^[24]). Transport pathways and potential sources of PM₁₀ in Beijing have been studied (Liang et al. ^[42]; Zhu et al. ^[43]). There was a clear seasonal and spatial variation of the potential source areas for Beijing (Li et al. ^[44]).

PM₁₀ in Beijing have been studied by many researchers using short-term experimental data or long-term monitoring data. Most of the long-term monitoring data were downloaded from the real-time air pollution monitoring information released online by the Ministry of Environmental Protection of China (MEPC). Although some data quality control had been carried out based on the obtained real-time data themselves, there were still some limitations because of the absence of the detail measurement information during the monitoring processes. In this study, PM₁₀ observed from 2006 to 2017 at a meteorological station located near the southern 5th-ring road in Beijing, is analyzed to understand the level, variation, and trend. The data are independent from the results from MEPC's sites, and will constitute a third-party assessment.

As the capital city of China, Beijing is located in the northern edge of the North China Plain, with an area of 16800 km², a resident population over 20 million, and more than 6 million vehicles circulating. It has a typical warm and temperate monsoon climate with distinct four seasons. The observation site used in this work is located in the yard of Beijing station. The site is located in the southeast of urban Beijing city and thus it is also named Nanjiao (NJ) observation station (Fig. 1). This is a long-term observation site with nearly 300 years history, where the observational data have been made available for academic exchange; thus, the PM₁₀ data here can be used to conduct an independent evaluation of PM₁₀ variations and the pollution control effect in Beijing.

Figure 1.  The location of Nanjiao (NJ) observation site in Beijing.

Ambient PM₁₀ mass concentrations have been continuously measured since 2006 using a TEOM1400A PM analyzer (Thermo Scientific, USA), which is based on the principle of Tapered Element Oscillating Microbalance method. The inlet height is 2.5 meter above the ground level. The lowest limit detection is 0.1 μg m^-3 in 1-hour average. The meteorological records of ambient air temperature (T), wind speed (WS), wind direction (WD), relative humidity (RH), and pressure (P) are from the routine operation measurement of the co-located meteorological station. Specifications of instruments are described in Table 1. QA/QC for the instruments follows the guideline of China Meteorological Administration.

Backward trajectory, which describes the paths air parcels take, was calculated using the HYSLPIT4 (version 4.9) model (Draxler and Hess ^[45]) and the meteorological data from the Global Data Assimilation System. The trajectory endpoint height was set to 100 m above ground level and the 5-day backward trajectories were calculated at 0, 6, 12, and 18 UTC every day during 2016-2017. The vertical motion method in the calculations used the default model selection. A total of 2924 backward trajectories were obtained and clustered with the clustering tool in the HYSLPIT4 software. The total spatial variance (TSV), the sum of all the cluster spatial variances, was calculated and a plot of percent change in TSV vs. number of clusters was created. The large increase in the change of TSV gave the final number of clusters. Each trajectory corresponded to one PM₁₀ concentration value averaged in three hours, which were used to calculate the average PM₁₀ concentration in each cluster.

From the year 2006 to the year 2017, a total of 85651 valid hourly mean data were obtained and a total of 3622 valid daily mean data were obtained, considering a valid daily mean data as a day with at least 11 valid hourly mean data. Moreover, valid monthly data were considered, as those months with at least 15 days' valid daily mean data. The records covered the whole period from 2006 to 2017 except from June 2013 to February 2014, when data was missing due to instrument malfunctions. Values were all reported at ±1 σ (standard deviation).

For the whole set of PM₁₀ concentrations data, the average of hourly mean was 138.7 ± 107.5 μg m^-3, the median value was 116.9 μg m^-3, the maximum was 3446.3 μg m^-3, the minimum was 0.1 μg m^-3 and the average ± 1σ of daily mean was 138.8 ± 96.1 μg m^-3, the median value was 118.6 μg m^-3, the maximum was 978.3 μg m^-3, and the minimum was 0.6 μg m^-3. PM₁₀ concentration exceeded in 1307 days the limited daily mean value (150 μg m^-3) of China National Ambient Air Quality Standard (GB3095-2012).

Lorentz curve fitting was applied to obtain the frequency distributions of daily mean PM₁₀ concentrations as plotted in Fig. 2. The open circles in the figure indicate the numbers of hourly concentrations in the corresponding concentration bin, and the solid line is the fitting Lorentz curve. As shown in Fig. 2, the frequency distribution pattern of the PM₁₀ concentration was of a single-peaked Lorentz curve. The peak concentration of PM₁₀ was of 96.6 ± 1.6 μg m^-3. The peak in frequency distribution obtained through statistical analysis identified the most probable concentrations, often used to represent the regional background concentration of PM₁₀ under a well-mixing assumption (Lin et al. ^[46]), which indicated a high regional background of PM₁₀ in Beijing. The peak concentration of the distribution curve was less than the mean (median) values and a long tail of the Lorentz curve with PM₁₀ concentrations was up to 1000 μg m^-3, which indicates that the site was often under heavy PM₁₀ pollutions in the observed period.

Figure 2.  PM₁₀ concentration frequency distribution and Lorentz curve fitting.

Figure 3 shows the time series variation in PM₁₀ daily mean concentrations from 2006 to 2017. PM₁₀ mass concentrations exhibited a fluctuating downward trend, which shows that achievements had been made in PM₁₀ pollution reduction in recent years. The linear fitting shows a significant downward trend with 7.0 μg/m³/yr of PM₁₀ mass concentration. The trend observed is higher than the one reported by Shi ^[25], in which the PM₁₀ concentration declined with the average rate of about 4 μg m^-3 per year.

Figure 3.  The time series variations of PM₁₀ daily mean concentration from 2006 to 2017 with linear fitting.

The valid PM₁₀ monthly data was obtained only when there were more than 15 days'valid daily mean data in the corresponding month. Average monthly variations in PM₁₀, ambient temperature (T), relative humidity (RH), rain, and wind speed (WS) are shown in Fig. 4. The maximum monthly mean PM₁₀ concentration was in December with a value of 170.9 ± 83.3 μg m^-3, and the second high value was in April with 167.0 ± 67.8 μg m^-3. The low values were in summer with the lowest one of 105.3 ± 35.0 μg m^-3 in August. The annual mean value of PM₁₀ concentration was 139.6 ± 21.5 μ m^-3, a higher value than the limit of 50 μg m^-3 prescribed by the China National Ambient Air Quality Standard (GB3095-2012).

Figure 4.  Average monthly variations in PM₁₀, T, RH, rain, and WS at NJ observation station in Beijing.

Relatively higher PM₁₀ concentrations were found in winter, spring, and autumn and lower values in summer. Similar seasonal changes were also reported by Shi ^[25]. Normally, summer is the cleanest season with more rains and abundant cover of vegetables, which help to clean the air and prevent the occurrence of fugitive dust. Relatively stronger wind speeds occur in spring when the air is rather dry. Thus, natural sources of dust and occasional sandstorms could contribute to the observed high PM₁₀ concentrations in spring. In winter, there are much more emissions due to heating and the lower boundary heights, which increases the accumulation of PM₁₀.

The meteorological parameters have directly influenced the atmospheric diffusion, transport, transmission, and photochemical reactions of pollutants. Here, we performed the analysis in the impact of surface wind and long-range air transport on surface concentrations of PM₁₀.

As seen in Fig. 5 (left), the prevailing surface wind directions at NJ station were from southwest (prevailed in summer) and northeast (prevailed in winter) sections. Stronger wind prevailed in the north and northwest sections. Fig. 5 (right) shows the contour map of PM₁₀ concentrations, wind direction and wind speed. Under low wind speed, PM₁₀ mass concentration distribution in all directions was relatively uniform. When the wind speed was greater than 4 m s^-1, the concentration gradient gradually increased, and the relatively high contributions were in the southeast and the north section. When the wind speed was over 6 m s^-1, higher PM₁₀ concentrations occasionally occurred in the southeast and northwest. In Beijing, windy weather might often lead to high PM₁₀ concentrations, especially the strong wind from northwest direction in spring and winter, which was closely related to dust activities (Zhang et al. ^[24]).

Figure 5.  The wind rose map (left) and the contour map of PM₁₀ vs wind direction and wind speed (right) at NJ station in Beijing.

Figure 6 shows PM₁₀ variations in different wind speed bins from 2006 to 2017. At low wind speed (< 4 m s^-1), the concentrations of PM₁₀ exhibited a downward trend with -19 μg m^-3 per unit of wind speed, but when wind speed rose (> 4 m s^-1), the concentrations of PM₁₀ exhibited an increasing trend with 49 μg m^-3 per unit of wind speed. It was observed that, in general, increasing wind speed caused the dilution of primary pollutants. The phenomenon in the increasing PM₁₀ concentrations with the increase of wind speed demonstrated how natural dust is a relevant source for PM₁₀ in Beijing.

Figure 6.  PM₁₀ variations in different wind speed bins.

The influence of regional airmass transports on surface PM₁₀ was conducted by using backward-trajectory-clustering analysis, with 7 clusters grouped. The mean clusters map is showed in Fig. 7 together with the percentage of each cluster to the total backward trajectories. The monthly percent distribution of trajectories in different clusters is shown in Table 2. The cluster analysis on backward trajectory showed that airmasses originated from the northwest (cluster 1-4, 6) had a dominant influence (accounting for 66% of all trajectories), especially in winter, spring, and autumn on surface air pollutants. In summer, air masses originated from the south (cluster 5) at shorter distances had a relevant influence. From May to September, air masses occasionally from northeast also had some influence.

Figure 7.  Mean trajectories of different clusters at 100 m a.g.l.

The average concentrations of PM₁₀ (PM₁₀-avg) for each cluster were calculated and are listed in Table 2, too. The average concentrations for each cluster were the overall mean of the PM₁₀ concentrations during the years of 2015-2016 for specific origins of air mass. Some trajectories were more frequent in certain season, with PM₁₀ concentrations showing associated seasonal variation. Therefore, the corresponding concentrations for certain clusters were often affected by the inherent seasonal change in PM₁₀ concentrations. Here, we use the PM₁₀-ref to filter seasonal variations, defined as

where a_n is the percent value of trajectories at the corresponding month in each cluster and C_n is the average concentration of PM₁₀ in the corresponding month.

The differences between PM₁₀-avg and PM₁₀-ref were then obtained and are listed in Table 2, in which the positive difference indicates a net contribution of long-range contribution and the negative one a net dilution effect. As a result, airmasses from northwest (cluster 3-4, 6), especially those originated from the Gobi desert and other desert areas, had net contributions; from north northwest (cluster 1-2) and northeast (cluster 7) the dilution effects occurred. Air massed from the south exhibited a more local signal. Long-range transport could enhance the surface PM₁₀ concentrations as high as a factor of 29% according to the difference in cluster 6.

Figure 8 shows the time series variations in PM₁₀ monthly mean concentration from 2006 to 2017. From 2006 to the winter of 2012, PM₁₀ concentration showed a linear downward trend, while at the end of 2012, the concentration rose sharply, followed by a steady downward trend after 2013. Overall, PM₁₀ concentration showed a downward trend from 2006 to 2017, and the linear fitting showed a significant downward trend with -8.0 μg/m³/yr. The trend result is close to that fitted by daily mean data of - 7.0 μg /m³/yr as plotted in Fig. 6. However, the annual mean PM₁₀ in 2017 was still about 94 μg m^-3, which was much higher than the limited annual value of 70 μg m^-3 prescribed by the China National Ambient Air Quality Standard (GB3095-2012).

Figure 8.  Time series variations in the monthly means of PM₁₀ concentrations from 2006 to 2017 with linear fitting.

Figure 9 shows the comparison of annual mean PM₁₀ data measured at NJ station and nearby regions (Fengtai, Daxing, and Chaoyang districts), as well as the whole mean values for Beijing. The different data showed a similar downward trend, but there was more fluctuation in the NJ data. The result further proved the effective controls in PM₁₀ in recent years in Beijing, but it also indicated that PM₁₀ behaviors in different area might be very different.

Figure 9.  Comparison of annual mean PM₁₀ data measured at NJ station and at nearby Fengtai, Daxing, and Chaoyang districts, as well as the whole mean values for Beijing.

Figure 10 shows the time series variations in the monthly mean PM₁₀ concentrations in different seasons from 2006 to 2017. The linear fittings show that the higher downward trends were found in autumn (-9.9 μg/m³/yr) and summer (-8.6 μg/m³/yr), and relatively less downward trends were found in winter (-5.2 μg/m³/yr) and spring (- 5.7 μg/m³/yr). The significant downward trend of PM₁₀ reflects the preliminary achievement in air pollution control over recent years. Since 2009, a series of strict actions has been taken to control air pollution, including national plans and policies to tackle air pollution, updating the national ambient air quality standard, and extending national air quality monitoring network (Li et al. ^[47]). In spring, PM₁₀ mass concentration fluctuated significantly with the lowest concentration in 2009 and 2011, which may be related to the atmospheric circulation (Tian et al. ^[48]). In summer, the concentration showed an obvious downward trend, and the lowest in 2011. In autumn, it decreased significantly from 2006 to 2010, and steadily declined after rising in 2011. Thereinto, the decline before 2011 was significantly greater than that after 2011. In the winter, PM₁₀ concentration dropped sharply between 2006 and 2010, rose to its peak in 2013, and then gradually declined with the most significant decline in 2017.

Figure 10.  Time series variations in the monthly mean PM₁₀ concentration in different seasons from 2006 to 2017 with linear fittings.

According to the changes in average diurnal variation of PM₁₀ concentrations in different years, the data were grouped into two datasets by different time periods. One period is from 2006 to 2011, and the other is from 2012 to 2017. The average diurnal variations in T, RH, WS and PM₁₀ in two different periods are reported in Fig. 11, which shows more humidity and warmer air, but weaker wind during 2012-2017 than 2006-2011. Higher temperatures (0.3 ℃), higher relative humidity (2%), and lower surface wind speeds (-0.2 m s^-1) indicated the influence of urbanization, which increased the surface roughness and the intensity of the urban heat island. Very significant difference of diurnal changes in PM₁₀ concentrations was found in these two periods. During 2006-2011, PM₁₀ concentrations showed a little diurnal variation with a bit higher level in the daytime. However, during 2012-2017, the diurnal variation of PM₁₀ concentrations showed a feature with a minimum in the daytime and a second peak during the morning rush hours. The lowest concentration in the afternoon minimum occurred around 2 pm, which is due to the favorite air conditions of dilution and dispersion. During 2006-2011, local sources such as fugitive dust might be a greater contribution to the daytime PM₁₀ under strong wind since the surrounding areas were not developed or under development. Similar conclusion has been drawn by Chen ^[49]. By contraries, the surrounding area were in a relatively mature development during 2012-2017 and greatly reduced natural fugitive dust by land solidification and vegetation cover.

Figure 11.  Average diurnal variations in T, RH, WS and PM₁₀ in two different periods.

Variations in the PM₁₀ concentrations from 2006 to 2017 at an observation station in Beijing were analyzed. The average ± 1 standard deviation of daily mean PM₁₀ concentrations was 138.8 ± 96.1 μg m^-3 and the median value was 118.6 μg m^-3. There were 1307 days (accounting for 34.7% of the total days) with the PM₁₀ concentration exceeding the limited daily mean value (150 μg m^-3) of the China National Ambient Air Quality Standard (GB3095-2012). At low wind speed (< 4 m s^-1), the concentrations of PM₁₀ exhibited a downward trend with -19 μg m^-3 per unit of wind speed, but when wind speed rose (> 4 m s^-1), the concentrations of PM₁₀ exhibited an increasing trend with 49 μg m^-3 per unit of wind speed. The phenomenon in the increasing PM₁₀ concentrations with the increase of wind speed demonstrated that how the natural dust is a relevant source for PM₁₀ in Beijing. Air massed from northwest, especially originated from the Gobi Desert and other desert areas had net contributions. Long-range transport could enhance the surface PM₁₀ concentrations as high as a factor of 29% during 2015-2016.

Overall, PM₁₀ mass concentration showed a significant declining trend with -8.0 μg/m³/yr from 2006 to 2017, and the higher downward trends were found in autumn (-9.9 μg/m³/yr) and summer (-8.6 μg/m³/yr), while relatively minor reduction was found in winter (- 5.2 μg/m³/yr) and spring (- 5.7 μg/m³/yr). Similar downward trends were found at nearby districts and the whole Beijing, but with wider data fluctuation. The result delivered an independent proof of the efficacy of the PM₁₀ reduction actions of recent years in Beijing. Further efforts should be done to reduce the presence and concentration of PM₁₀, especially by controlling fugitive dust.

The data that support this study can be shared on reasonable request to the corresponding author.