Multi-Method Observation and Numerical Simulation of a PM 2 . 5 Pollution Episode in Beijing in October , 2014

Multi-method observation and numerical simulation were applied to analyze a PM2.5 pollution episode in Beijing in October, 2014. The results of vertical observation showed that surface-level backscatter signal and extinction coefficient increased during the episode, suggesting that air pollutants accumulated near the ground. The main meteorological factors during this episode could be described as calm wind, high relative humidity and low surface pressure. The evolution of PM2.5 concentrations in this episode was divided into four stages, including two-steps type concentration climbing stages (P1 and P2), high concentration maintenance stage (P3) and rapid cleanup stage (P4). Analysis on ground-based observation, satellite remote sensing and atmospheric general circulation showed that regional transport, including crop residue burning, was the main incentive of this pollution episode. Subsequently, local pollutants emission and regional transport maintained and aggravated the episode under unfavorable meteorological conditions. Temporal variation of OX was in close agreement with that of PM2.5 and the concentration peaks of OX occurred few hours before those of PM2.5, which indicated that strong atmospheric oxidation could promote the formation of secondary PM2.5. The results of numerical simulation showed that during 8–10 October, the average contribution of regional transport to PM2.5 in the five sites exceeded 50%.


INTRODUCTION
With the rapid development of the social economy and continuously increasing energy consumption, air pollution has become a serious problem in China in recent years (Chan and Yao, 2008;Lin et al., 2008;Yang et al., 2009;Rose et al., 2010).Beijing has a population of 16 million within an area of 16,800 km 2 , making it one of the largest and most densely populated cities in China.Beijing is located at the northwestern border of the North China Plain and is bound by mountains on the north, east, and west.Many heavily populated industrialized cities are located near Beijing to the southwest and southeast (Xu et al., 2011).Unfavorable geographical conditions and rapid growth in traffic emissions as well as regional pollutant transport have made Beijing one of the most seriously polluted cities in China.
Particulate matter (PM), especially PM 2.5 (fine particles with aerodynamic diameters less than 2.5 µm), plays important roles in atmospheric visibility reduction, acid deposition, and climate change (Yan et al., 2008;Garland et al., 2009;Ma et al., 2011).PM 2.5 also has adverse health effects.Exposure to high concentrations of PM 2.5 has been found to result in increased hospitalizations and higher mortality rates (Michaels and Kleinman, 2000;Dockery, 2001;Schwartz et al., 2002;Zhang et al., 2010).The studies on the characterization of PM 2.5 in Beijing have been carried out in the last two decades (Winchester and Bi, 1984;Dong and Yang, 1998;Yang et al., 2000;Cao et al., 2002;Zhang et al., 2012).Several studies have examined the general characteristics of PM 2.5 chemical components and discussed their seasonal variations (He et al., 2001;Sun et al., 2004;Song et al., 2006).Several studies have focused on correlations among PM 2.5 components and the formation of secondary particles (Yao et al., 2002;Dan et al., 2004;Huang et al., 2006;Pathak et al., 2009;Wang et al., 2009;Ianniello et al., 2011).Several studies have discussed the evolution process and formation mechanism of PM 2.5 heavy pollution (Sun et al., 2006;An et al., 2007).There have also been several studies aimed at revealing the health effects of PM 2.5 (Zhang et al., 2000;Guo et al., 2009;Kipen et al., 2010;Wu et al., 2010).In addition, lidar, satellite remote sensing and numerical simulation were also applied to investigate the vertical profile, the biomass burning emission and the regional transport of PM 2.5 , respectively (Aaron et al., 2006;Yu et al., 2008;Tang et al., 2015).
In this study, multi-observation method, including laser ceilometer, micro-pulse lidar, satellite remote sensing and online observation of meteorological factors and air pollutants as well as numerical simulation were applied in a PM 2.5 pollution episode in Beijing during 7-11 October, 2014.The purpose of this study is to (1) represent the evolution process and variation characteristics of PM 2.5 in this pollution episode; (2) discuss the relationship between PM 2.5 and meteorological factors as well atmospheric oxidation; and (3) investigate the formation mechanism of this pollution process and offer advice for air pollution control policy in Beijing.

Observation Sites
Five monitoring stations from the automatic air quality monitoring network in Beijing were selected for this study, including Yufa (YF), Guanyuan (GY), Gucheng (GC), Dongsi (DS) and Changping (CP).Concentrations of PM 2.5 , NO 2 and O 3 were monitored in the five sites.Laser ceilometer, micro-pulse lidar, visibility instrument and online PM 2.5 chemical composition analyzer were located at Beijing Municipal Environmental Monitoring Center (JCZX), which was less than 2 km away from GY site.Groundbased meteorological data was monitored in Guanxiangtai (GXT) (Fig. 1).

Observation Instruments
Monitoring instruments from Thermo Fisher Corporation, USA, were used to measure O 3 (49C), NO/NO 2 /NO x (42C), and PM 2.5 (48C), the detection limit of which was 1.0 ppb, 0.05 ppb and 3.0 µg m -3 , respectively.Daily zero/span checks were automatically performed using dynamic gas calibrators combined with zero air suppliers and standard gas mixtures for NO.Multipoint calibrations of PM 2.5 and NO x analyzers were performed every week using standard gases.An O 3 calibrator (49IPS) was used to calibrate the O 3 analyzers at the sites.The calibrator is traceable to the Standard Reference Photometer maintained by the WMO World Calibration Centre in Switzerland (EMPA).Sampling-heads of the instruments are about 3-5 m high from the ground.
The RT-4 OC/EC analyzer (Sunset Lab, USA) and URG9000S ion analyzer (Thermo Fisher, USA) were applied to measure chemical composition of PM 2.5 .Before each chemical component measurement, standard samples from the Institute for Environmental Reference Materials of the Ministry of Environmental Protection were applied to calibrate the instruments.A multi-point calibration was done every week.Parallel samples contributed at least 10% of the total number of samples.The WXT520 meteorological instrument (Vaisala, Netherlands) and the FD12 instrument (Vaisala, Netherlands) were used to measure meteorological factors and visibility.The CL31 laser ceilometer (Vaisala, Netherlands) and the EV-LIDAR micro-pulse lidar (Everise, China) were also applied in this study.All meteorological data and air pollutants data used in this study are hourly data.

Model Configurations and Evaluation
The MM5-CAMx model modeling system (Wang et al., 2015a) was applied in this study to simulate concentrations of PM 2.5 and chemical compositions of PM 2.5 during this pollution episode.The PAST module in CAMx was used to study the regional transport of PM 2.5 in the monitoring sites.The first modeling domain of CAMx covered East Asia region with a 36 × 36 km grid resolution, and the domain origin was 36°N, 104°E.The second modeling domain covered North China region with a 12 × 12 km grid resolution.MM5 model used the same domain origin and grid resolution as CAMx.The MM5 modeling domain was four grid cells broader on each side of the CAMx domain.First-guess fields and the initial conditions for MM5 were taken from the National Center for Environmental Prediction (NCEP) final analysis data sets with a spatial resolution of 1° × 1° and a temporal resolution of 6 h.The MEGAN model (Wang et al., 2011) was used to generate biogenic pollutant emission inventory for CAMx.Anthropogenic pollutant emission was from the MEIC inventory with a resolution of 0.25°× 0.25° (Zhang et al., 2009).A spin-up period of 5 days was used for all model simulations to reduce the influence of initial conditions on the model results.
As shown in Fig. 2, the simulated data was in close agreement with observed data.Overall, the simulated data was lower than the observed data, which might be caused by model grid resolution or the errors in emission inventory.In addition, several statistical parameters (○ 1 -○ 4 ) were calculated to provide a quantitative assessment of simulation (Table 1).These statistical parameters were similar to previous studies (Morris et al., 2005;Boylan and Russell, 2006).Thus, CAMx performed adequately for this study.

Analysis on Vertical Observation
Laser ceilometer measured the height of cloud by calculating the backscatter signal.In addition to the cloud, haze and particulate matter can also cause increase of laser backscatter signal.Thus, laser ceilometer is also applied to investigate the concentration level and concentration distribution of air pollutants.Previous study has shown that aerosol extinction coefficient has a positive correlation with aerosol mass concentration (Weinzierl et al., 2009).Thus, micro-pulse lidar is used to investigate air pollutants by measuring aerosol extinction coefficient.The surface-level backscatter signal increased gradually in the afternoon of 7 October and maintained a high level during 8-10 October (Fig. 3), which suggested that air pollutants accumulated near the ground and concentrations of PM 2.5 increased obviously.Simultaneously, surfacelevel extinction coefficient increased in the afternoon of 7 October, indicating that visibility decreased (Fig. 4).Vertical observation showed that air pollutants accumulated near the ground during 7-11 October.

Analysis on Ground-Based Observation
Fig. 5 represented the meteorological factors and hourly concentrations of PM 2.5 in the monitoring sites during 7-11 October, 2014.The prevailing wind was northeast wind in the morning and at night and turned to southeast wind in the afternoon, which maintained a small wind speed.Horizontal meteorological conditions were not favorable for the diffusion of air pollutants.Relative humidity was high (60%-99%) during this pollution episode, especially at night (> 90%), which was favorable for the formation of secondary PM 2.5 (Khare et al., 2011).In addition, Beijing was controlled by low pressure (1012-1016 hPa) and the variation of pressure was small.Visibility decreased in the early stage of the episode and maintained lower than 2 km during the episode.In the afternoon of 11 October, this pollution episode ended with the increase of wind speed, surface pressure, visibility and the decrease of relative humidity.
The evolution of PM 2.5 concentrations was divided into four stages according to its variation characteristics, including two-steps type concentration climbing stages (P1 and P2), high concentration maintenance stage (P3) and rapid cleanup stage (P4).In the afternoon of 7 October, along with the unfavorable meteorological conditions, concentrations of PM 2.5 in the five sites increased rapidly, especially in YF.Concentration of PM 2.5 in YF reached the first peak (281.1 µg m -3 ) at 20:00 in 7 October, subsequently, concentrations peaks of PM 2.5 in GY (272.8 µg m -3 ) and CP (222.3 µg m -3 ) occurred at 22:00 in 7 October and at 05:00 in 8 October, respectively.PM 2.5 peaks in the three sites in north-south direction occurred first with the highest concentration in the southern site, following by the central site and the northern site, indicating significant impact of regional transport.The first-step concentration climbing stage P1 referred to the time period that from the beginning of the pollution episode to the first concentration peaks in the monitoring sites.After P1, concentrations of PM 2.5 maintained stable.
From 10:00 in 8 October, concentrations of PM 2.5 increased rapidly once again.Concentration peaks of PM 2.5 in YF (425.5 µg m -3 ), GY (413.4 µg m -3 ) and CP (385.0 µg m -3 ) occurred at 15:00 in 8 October, 22:00 in 8 October and 04:00 9 October, respectively.The second-step concentration climbing stage P2 referred to the time period that from the end of P1 to the second concentration peaks in the monitoring sites.After the concentration accumulation in P1, precursors of PM 2.5 also reached high concentrations.Thus, Concentration peaks in P2 were caused by regional transport, local accumulation and formation through photochemical reaction.When the concentration peaks in P1 and P2  occurred, the meteorological conditions could be described as calm wind, high relative humidity and low atmospheric pressure.In stages P1 and P2, concentration variation trends of PM 2.5 in the three sites in east-west direction (GC, GY and DS) were more consistent and the concentration peaks occurred at almost the same time.This phenomenon verified that the contaminated air mass moved in the north-south direction, thus, regional transport from the southern region of Beijing played an important role in stage P1 and P2.
After P2, concentrations of PM 2.5 maintained high level (> 200 µg m -3 ) for more than 60 hours, and this time period was the high concentration maintenance stage P3.In P3, wind speed kept small and relative humidity exceeded 90% at night.From 18:00 in 11 October, concentrations of PM 2.5 decreased rapidly from north to the south along with a strong cold air.The prevailing wind turned to northwest wind with a strong wind speed, and relative humidity decreased, and the surface pressure enhanced.This time period was the rapid cleanup stage P4.
The Spearman correlation coefficients between PM 2.5 and meteorological factors in GY sites were calculated (Table 2).For the whole pollution episode, PM 2.5 showed significant positive correlations with temperature and relatively humidity and significant negative correlations with wind speed and surface pressure.In all the four stages, PM 2.5 showed significant negative correlations with surface pressure, which  -) indicates no significant correlation; * correlation was significant at the 0.05 level (two tailed); unless noted, correlation was significant at the 0.01 level (two-tailed).
indicated that surface pressure had significant impact on concentrations of PM 2.5 .In stage P1-P3, PM 2.5 showed no significant correlation with relatively humidity.The possible reason was that relatively humidity maintained high level (> 60%) during

Analysis on Atmospheric Oxidation and Chemical Compositions in PM 2.5
Atmospheric oxidants OX (NO 2 + O 3 ) can be applied to evaluate atmospheric oxidation capacity (Berchet et al., 2013;Gressent et al., 2014;Wang et al., 2015b).As shown in Fig. 6, temporal variation of OX was in close agreement with that of PM 2.5 and the concentration peaks of OX occurred few hours before those of PM 2.5 , which indicated that strong atmospheric oxidation could promote the formation of secondary PM 2.5 .Fig. 7 represented concentrations of chemical compositions of PM 2.5 and the ratio of K + /OC.In stage P1, concentration of Cl -increased first and maintained a relatively high level during the episode.Concentrations of K + increased in the afternoon of 8 October and maintained a high level until the afternoon of 10 October.Previous studies have shown that Cl -and K + could be applied as tracer elements of biomass burning (Cheng et al., 2013).Considering the special characteristics of pollution emissions in North China in autumn, it could be concluded that crop residue burning had a significant impact on this episode.Duan et al. (2004) found that the ratio of K + /OC increased due to biomass burning.The variation of K + /OC in this episode also indicated the impact of crop residue burning on the episode.
Concentrations of SO 4 2-, NO 3 -, NH 4 + and OC increased at the same time and at a close concentration level.In stages P2 and P3, concentrations of NO 3 -was higher than other chemical compositions, which indicated that local pollution emission, represented by motor vehicles, played an important role (Song et al., 2006;Yang et al., 2015).In high concentration maintenance stage P3, SO 4 2-, NO 3 -, NH 4 + and OC were the major chemical compositions in PM 2.5 , concentrations of which accounted for 14.5%, 20.3%, 15.5% and 13.0% of PM 2.5 concentration, respectively.The Spearman correlation coefficients between chemical compositions of PM 2.5 and PM 2.5 (GY site), OX, visibility were calculated, respectively (Table 3).For the whole episode, all monitored chemical compositions showed significant positive correlations with PM 2.5 , among which the correlations between PM 2.5 and secondary ions (SO 4 2-, NO 3 -, and NH 4 + ) were the strongest.Chemical compositions showed significant negative correlations with visibility, among which the correlation between NH 4 + and visibility was the strongest.Secondary ions and OC in PM 2.5 showed significant positive correlations with OX, which verified that strong atmospheric oxidation could lead to the increase of PM 2.5 concentration by promoting the formation of secondary ions and SOC (secondary organic carbon).In P3 and P4, most chemical compositions had not showed significant correlations with visibility and OX due to the relatively stable concentrations 10-07 00:00 10-07 12:00 10-08 00:00 10-08 12:00 10-09 00:00 10-09 12:00 10-10 00:00 10-10 12:00 10-11 00:00 10-11 12:00 10-12 00:00 and the few sample numbers, respectively.In the early stage of the episode (P1), the correlations between chemical compositions and PM 2.5 , visibility, OX were the strongest in the four stages, respectively.

Analysis on Satellite Remote Sensing, Back-Trajectories and Numerical Simulation
Fig. 8 represented the open biomass-burning retrieved from the MODIS data (http://rapidfire.sci.gsfc.nasa.gov)following the method of an enhanced contextual fire detection algorithm (Giglio et al., 2003) during this episode.The results showed that crop residue burning was serious in the southern regions of Beijing during this episode.The 72-hour back-trajectories calculated by HYSPLIT model (Heintzenberg et al., 2013;Draxler and Rolph, 2015) also verified that air mass went through the crop residue burning regions (Fig. 8).Thus, satellite remote sensing, atmospheric general circulation and ground-based observation of PM 2.5 supported each other, suggesting that this episode was affected by crop residue burning.
The results of numerical simulation verified the analysis of ground-based observation (Figs.9-10).The contributions of emission sources outside Beijing to PM2.5 concentration in Beijing were applied as the contribution of regional transport.In the early stage of this episode (P1), the contribution of regional transport to PM 2.5 in YF increased first (> 70%) and maintained a high level during the episode (> 60%).During 8-10 October, the average contribution of regional transport to PM 2.5 in the five sites exceeded 50%.In the rapid cleanup stage P4, contribution of regional transport decreased.Simultaneously, Fig. 10 represented that regional transport had a significant impact on this PM 2.5 pollution episode.The simulated PM 2.5 was near the ground and the back-trajectories air mass was at 500 m height.In addition, the simulated PM 2.5 was daily average concentration, whereas, the air mass was 72-hour back-trajectories at 22:00 on 8 October, 2014.Thus, the sources of PM 2.5 in the two method might be different.

SUMMARY AND CONCLUSIONS
In this study, a PM 2.5 pollution episode in Beijing during 7-11 October, 2014 was selected to analyze through multiobservation method and numerical simulation.The major results and conclusions are as follows.1.In the early stage of this pollution episode, surface-level backscatter signal and extinction coefficient increased, which suggested that air pollutants accumulated near the ground.The main meteorological factors during this episode could be described as calm wind, high relative humidity and low surface pressure.2. The evolution of PM 2.5 concentrations was divided into four stages according to its variation characteristics, including two-steps type concentration climbing stages (P1 and P2), high concentration maintenance stage (P3) and rapid cleanup stage (P4).Regional transport had a significant influence on concentrations of PM 2.5 in P1 and P2. 3. Analysis on temporal variation of PM 2.5 and OX and The ratio of K + /OC 10-07 00:00 10-07 12:0010-08 00:0010-08 12:0010-09 00:00 10-09 12:00 10-10 00:0010-10 12:00 10-11 00:0010-11 12:0010-12 00:00 0.328 --( -) indicates no significant correlation; 1) correlation was significant at the 0.05 level (two tailed); unless noted, correlation was significant at the 0.01 level (two-tailed).correlation between chemical compositions of PM 2.5 and OX indicated that strong atmospheric oxidation could promote the formation of secondary PM 2.5 and led to the increase of PM 2.5 concentrations.Concentrations of Cl -and K + and the ratio of K + /OC increased during this episode, which suggested that the episode was affected by crop residue burning in the region.The results of satellite remote sensing and atmospheric general circulation verified this conclusion.4.During 8-10 October, the average contribution of regional transport to PM 2.5 in the five sites exceeded 50%.
According to the multi-method observation and numerical simulation, it could be included that regional transport, including crop residue burning, was the main incentive of this pollution episode.Subsequently, local pollutants emission and regional transport maintained and aggravated the episode under unfavorable meteorological conditions.On the basis of the monitoring results and analyses, several recommendations have been made for understanding and improving the air quality in Beijing.
• Regional transport played a critical role in this PM 2.5 pollution episode.Thus, the future direction of air pollution control in Beijing is regional joint prevention and control.• Strong atmospheric oxidation could promote the formation of secondary PM 2.5 .Thus, reducing the concentration of O 3 was a synergistic control of PM 2.5 , which should be put enough focus on.• In the autumn harvest season, crop residue burning was frequent and large-scale in North China, which had a significant impact on air quality.Some measures should be formulated to recycle the crop residues and use them for cleaner production.• NO 3 -was the most abundant composition in P2 and P3, suggesting the important contribution of motor vehicles.A total number restriction of motor vehicles should be implemented in Beijing and elimination of yellow-tag vehicles (Highly polluting vehicles with exhaust exceeding the European Union criteria I) should be speeded up.

Fig. 1 .
Fig. 1.Locations of the monitoring sites.YF (39.5°N, 116.3°E) is a regional transport monitoring site, which is located in Daxing District near the southern boundary of Beijing.GY (40.0°N, 116.3°E),GC (39.9°N, 116.2°E) and DS (39.9°N, 116.4°E) are three urban environment monitoring sites, which is located in the central region (Xicheng District), western region (Shijingshan District) and eastern region (Dongcheng District), respectively.CP (40.2°N, 116.2°E) is a suburban environment monitoring sites, which is located in Changping District in the northern suburbs of Beijing.

Fig. 8 .Fig. 9 .
Fig. 8. Open biomass-burning retrieved from the MODIS data in the central and eastern regions of China during 7-11 October, 2014 and 72-hour back-trajectories (the blue line) at 500 m height in Beijing at 22:00 on 8 October, 2014

Fig. 10 .
Fig. 10.The daily average contribution of regional transport to PM 2.5 concentration in Beijing.

Table 1 .
The statistical results of observed and simulated data.

Table 2 .
The Spearman correlation coefficients between PM 2.5 and meteorological factors in GY sites during the episode.

Table 3 .
The Spearman correlation coefficients between chemical compositions of PM 2.5 and PM 2.5 (GY site), OX, visibility during the episode, respectively.