Figure 1 shows the spatial distribution of the average PM2.5 and PM10 concentrations in the JJJ area in December 2013–16. The concentrations of PM2.5 and PM10 decrease from the south to the north of the study region. The average concentration of PM2.5 (PM10) ranges from 36.7 (86.2) to 194.9 (299.9) μg m–3. The highest concentration of PM2.5 (PM10) appears in BD (HD), and the lowest in ZJK. As reported in previous studies (e.g., He et al., 2017b), several PM emission sources are located in southeastern JJJ, which explains the high concentrations of PM2.5 and PM10 in this area. A spatial difference in meteorology is another factor affecting the PM concentration distribution. Because of the winter monsoon, the wind speed in northern JJJ is often greater than in southern JJJ. In addition, a large amount of wind often occurs along the coast because of the land–sea distribution. A large amount of wind is beneficial for pollutant dispersion and results in relatively low concentrations of PM2.5 and PM10 in northern and eastern JJJ.
Figure 2 shows the daily mean PM concentrations in December 2013–16. The days featuring serious pollution, based on the daily mean PM2.5 concentration (larger than 150 μg m–3), reach 13 days in December 2013–15, and 9 days in December 2016. Only 4 days are seriously polluted in December 2014. A serious pollution period reach-ing 6 days is found in December 2013, with maximum daily mean PM2.5 and PM10 concentrations of 391.9 and 485.4 μg m–3, respectively, over the JJJ area.
Figure 2. Daily mean (a) PM2.5 and (b) PM10 concentrations in December 2013–16 over the JJJ area. The gray dotted lines represent the critical value for serious air pollution.
In recent years, central and local governments have made notable efforts to control air pollution. The interannual variation of pollutant concentrations is a pressing issue for the government and the public. Table 1 exhibits the monthly mean concentrations of PM2.5 and PM10 in December 2013–16 over the JJJ area. Compared to December 2013, a significant decrease in PM concentration was detected in December 2014, with a decrease of 33% and 51% for PM2.5 and PM10, respectively. However, increasing PM concentrations appeared in December 2015, and the concentrations of PM2.5 and PM10 in December 2016 were largely the same as those in December 2015. Overall, total emissions decreased year by year in the JJJ area (National Bureau of Statistics and Ministry of Environmental Protection, 2015). Therefore, the unusual increase in PM concentration was most likely related to unusual meteorological conditions.
2013 2014 2015 2016 PM2.5 148.6 100.1 140.5 141.7 PM10 236.4 166.4 204.5 203.1 PM2.5/PM10 0.63 0.60 0.69 0.70
Table 1. Monthly mean PM concentrations (μg m–3) and ratios of monthly mean concentrations between PM2.5 and PM10 in December over the JJJ area
Based on the Kolmogorov–Smirnov test, daily average PM concentrations and some meteorological parameters do not satisfy the normal distribution in the JJJ area. Therefore, Spearman’s rank correlation coefficient was used for correlation analysis. Before analyzing the relationship between meteorological parameters and PM concentrations, the correlation coefficients between daily mean meteorological parameters are discussed (Table 2). Winter temperature was significantly correlated with winter monsoon, with a correlation coefficient of –0.52. RH2 was significantly correlated with other meteorological parameters (except temperature). It should be noted that the enhancement of atmospheric diffusion conditions (large WS10 and BLH) reduced RH2 in winter over the JJJ area. The correlation coefficient between WS10 and BLH was 0.79, which implies that the development of the atmospheric boundary layer (or turbulence) is mainly caused by dynamic rather than thermodynamic mechanisms in winter over the JJJ area. BLH was also significantly correlated with the SHI, with a correlation coefficient of 0.48. Attaining deep insight into the relationships between meteorological parameters can help us in understanding the relationship between PM concentrations and meteorological parameters, as shown in the following analysis.
T2 RH2 WS10 BLH SHI SHPI T2 1.0 0.11* –0.04* –0.14* –0.52 –0.17* RH2 1.0 –0.61 –0.52 –0.22 0.47 WS10 1.0 0.79 0.28 –0.38 BLH 1.0 0.48 –0.27 SHI 1.0 0.05* SHPI 1.0 *Correlation coefficient is not significant at the 95% confidence level.
Table 2. Correlation coefficients between daily mean meteorological parameters in December 2013–16
WS10 and BLH reflect the turbulent mixing and dispersion capability of the atmosphere. High wind speeds and a high BLH are beneficial to the horizontal and vertical dispersion of pollutants. PM2.5 and PM10 concentrations were negatively correlated with WS10 and BLH (Fig. 3), and were positively correlated with T2 and RH2. Hygroscopic condensation growth of aerosol results in a positive correlation between PM concentrations and RH2. On the other hand, significant negative correlations between RH2 and WS10 and BLH (Table 2) may have resulted in the positive correlation between PM concentrations and RH2. A strong winter monsoon contributes to the rapid dispersion of pollutants in the JJJ area. The SHI (SHPI) was negatively (positively) correlated with PM concentration in the JJJ area. These findings regarding the correlation between PM concentrations and meteorological parameters are consistent with previous studies (Tai et al., 2010; Jia et al., 2015; He et al., 2017a; Liu et al., 2017). The correlation coefficients were significant at the 95% confidence level, except the SHI. Basically, the correlations between meteorological parameters and PM2.5 were higher than those with PM10, except for T2 and SHI.
Figure 3. Correlation between PM concentrations and meteorological parameters in December 2013–16 in the JJJ area. The dashed lines indicate statistical significance at the 95% confidence level, based on the t-test.
Table 3 shows the monthly meteorological parameters for December 2013–16 over the JJJ area. When the winter monsoon was strong (high SHI), prevailing northwesterly winds are favorable for the dispersion and transport of pollutants, resulting in good air quality in the JJJ area. The winter monsoon was strongest in December 2014, according to the SHI. A relatively high (low) SHI (SHPI) was responsible for the low PM2.5 and PM10 concentrations in December 2014. Based on the local meteorological conditions, T2 and RH2 were abnormally low and WS10 and BLH were abnormally high in December 2014. These abnormal meteorological conditions resulted in low concentrations of PM2.5 and PM10. With high T2 and RH2, and low WS10 and BLH, the meteorological conditions were more adverse for pollutant dispersion in December 2015 and 2016 than in December 2013 and 2014. However, the concentrations of PM2.5 and PM10 in December 2015 and 2016 were lower than in December 2013 (Table 1), which implies that control measures have resulted in an effective improvement in air quality.
2013 2014 2015 2016 T2 (K) 270.4 269.7 271.3 271.8 RH2 (%) 48.1 39.8 63.4 63.6 WS10 (m s–1) 2.8 3.3 2.7 2.4 BLH (m) 406.5 548.5 404.7 305.9 SHI (hPa) 1027.3 1031.0 1025.5 1025.0 SHPI 100.7 101.7 104.8 104.7
Table 3. Monthly mean meteorological parameters in December over the JJJ area
The ratio of fine PM to coarse PM (PM2.5/PM10) increased year by year, except in December 2014 (Table 1), when it was affected by meteorological conditions and emissions. The average wind speed was higher than 3.2 m s–1 in December 2014, which enabled the formation of local dust and sand storms, and the weather phenomena of floating dust was recorded (for December 2014 only). This resulted in a relatively low ratio of fine to coarse PM (Table 1). The increasing RH2 had an adverse influence on visibility under the same loading of PM, and it promoted the formation of secondary PM from gaseous species (Liu et al., 2017). High RH2 (> 60%) is an important factor for understanding the large ratio of fine to coarse particles recorded in December 2015 and 2016 (Table 1). Control measures, such as road watering and covering construction sites, decrease dust emissions (coarse particles). This is another reason for the increase in the ratio of fine to coarse particles in recent years.
Air quality in northern China is prominently correlated with pressure systems (Chen et al., 2008). Six circulation types were identified in this study using the T-mode PCA method combined with K-means clustering method (Fig. 4). The mean SLP for the six circulation types implied that a strong cold air process with a large pressure gradient controlled the JJJ area for CT1, followed by CT2 and CT3. Static weather with a negligible pressure gradient controlled the JJJ area for CT4 and CT6. The concentration anomalies of PM2.5 and PM10 (daily mean concentration minus monthly mean concentration) for the six circulations types are shown in Fig. 5. A strong cold air process is beneficial for the rapid dispersion and transport of pollutants, and the lowest concentrations of PM2.5 and PM10 were detected in CT1, followed by CT2 and CT3. In other words, the concentration anomaly is a positive value for static weather, which indicates that the weather adversely affects the dispersion and transport of pollutants. For CT5, a low-pressure system was located in Northeast China, resulting in southerly winds in the JJJ area that formed serious air pollution due to regional transport and convergence.
Figure 5. Box plots of concentration anomalies of (a) PM2.5 and (b) PM10 for the six circulation types.
The occurrence frequencies of the six circulation types are exhibited in Table 4. The ratios of static weather (CT4 and CT6) to CT5 were 41.9%, 32.3%, 48.4%, and 54.8% in December 2013–16, respectively. The average ratio of static weather to CT5 in December 2013–16 was 44.4%. The good air quality recorded in December 2014 (Table 1) was most likely caused by a high ratio of cold air. While poor air quality was recorded in December 2015 and 2016, this was most likely caused by a high ratio of static weather. These data indicate that the interannual variations of meteorological conditions determine the interannual variations of pollutant concentrations during winter.
CT1 CT2 CT3 CT4 CT5 CT6 2013 12.9 32.3 12.9 3.2 35.5 3.2 2014 41.9 12.9 12.9 9.7 12.9 9.7 2015 12.9 19.4 19.4 19.4 12.9 16.1 2016 12.9 16.1 16.1 35.5 9.7 9.7 2013–16 20.2 20.2 15.3 16.9 17.7 9.7
Table 4. Occurrence frequencies (%) of the six circulation types in December
Numerical simulation is an important method to identify emissions changes based on the difference between the observed and simulated PM concentration change (using the same emissions inventory) in different years (Yang et al., 2016; Liu et al., 2017). The Chinese Unified Atmospheric Chemistry Environment (CUACE) model coupled with the fifth-generation Penn State/NCAR mesoscale model (MM5) was used in this study. The model settings, initial and boundary conditions, meteorological forcing fields, and emissions inventory, were consistent with previous studies (He et al., 2016d, 2017b; Liu et al., 2017). The performance of MM5–CUACE has been verified in previous research (He et al., 2016d; Liu et al., 2017), and so an evaluation was not performed again here. Table 5 shows the PM2.5 concentration change ratio and emissions change ratio over the JJJ area. Because of the same emissions inventory (2013) for December from 2013 to 2016, the simulated concentration change only represents the impact of meteorological conditions. The observed concentration change represents the impact of both meteorological conditions and emissions. Therefore, the emissions change ratio can be inferred from the difference between the simulated and observed change ratio.
Observation Simulation Emissions change 2014 vs 2013 –32.6 –19.9 12.7 2015 vs 2014 40.3 48.9 8.6 2016 vs 2015 0.9 9.2 8.3
Table 5. Observed and simulated PM2.5 concentration change ratio and emissions change ratio (%) over the JJJ area
With the considerable efforts in terms of emissions control, the emissions of PM2.5 decrease by 12.7%, 8.6%, and 8.3% for December 2014, 2015, and 2016, as compared with each previous year, over the JJJ area. The PM2.5 concentration increases by about 50% in December 2015 compared with 2014 due to the change in meteorological conditions. Moreover, the impact of meteorological conditions on the interannual variation of the monthly mean PM2.5 concentration is larger than the impact of the emissions change. It should be noted that the model simulation contains uncertainties, and the inferred emissions changes should be regarded as approximations only.