Investigating the impact of regional transport on PM 2 . 5 formation using vertical observation during APEC 2014 Summit in Beijing

During the APEC (Asia-Pacific Economic Cooperation) Economic Leaders’ 2014 Summit in Beijing, strict regional air emission controls were implemented, providing a unique opportunity to investigate the transport and formation mechanism of fine particulate matter (PM2.5). This study explores the use of vertical observation methods to investigate the influence of regional transport on PM2.5 pollution in Beijing before and during the APEC Summit. Vertical profiles of extinction coefficient, wind, temperature and relative humidity were monitored at a rural site on the border of Beijing and Hebei Province. Three PM2.5 pollution episodes were analyzed. In episode 1 (27 October to 1 November), regional transport accompanied by the accumulation of pollutants under unfavorable meteorological conditions led to the pollution. In episode 2 (2–5 November), pollutants left from episode 1 were retained in the boundary layer of the region for 2 days and then settled down to the surface, leading to an explosive increase of PM2.5. The regional transport of aged aerosols played a crucial role in the heavy PM2.5 pollution. In episode 3 (6–11 November), emissions from large point sources had been controlled for several days while primary emissions from diesel vehicles might have led to the pollution. It is found that ground-level observation of meteorological conditions and air quality could not fully explain the pollution process, while vertical parameters (aerosol optical properties, winds, relative humidity and temperature) improved the understanding of regional transport influence on heavy pollution processes. Future studies may consider including vertical observations to aid investigation of pollutant transport, especially during episodic events of rapidly increasing concentrations.


Introduction
With rapid economic development and increases in energy consumption, a large quantity of emissions has caused serious particulate matter pollution in China.Monitoring data shows that the Beijing-Tianjin-Hebei (BTH) region is one of the most polluted regions in China (B.Zhao et al., 2013;Wang et al., 2014).The region was home to 8 of the 10 most polluted Chinese cities in 2014 according to the reports published by Ministry of Environment Protection (MEP, 2015).In 2014, the annual average PM 2.5 (particulate matter with aerodynamic diameter less than 2.5 µm) concentration reached 95 µg m −3 in the BTH region.With 21.5 million residents and 5.3 million vehicles in 2014, Beijing has frequently been burdened with severe pollution episodes in recent years (Beijing Municipal Bureau of Statistics, 2015).The capital is surrounded by mountains in three directions (north, west and east).The top three most polluted cities in China (Baoding, Xingtai and Shijiazhuang) are located south of Beijing.Polluted air masses from the south contribute to PM 2.5 pollution in Beijing (Wang et al., 2015).Source apportionment by the Beijing Environmental Protection Bureau indicates that regional transport contributed 28-36 % to PM 2.5 in Beijing in 2012-2013.During some severe pollution periods, regional contribution was more than 50 % (http://www.bjepb.gov.cn/bjepb/413526/331443/331937/333896/396191/index.html).Quite a few studies have researched the causes of heavy pollution episodes in the BTH region, and they show that regional transport plays an important role in pollution formation.The sharp PM 2.5 buildup events in Beijing were unique, while pollution accumulated more slowly in other cities in the region.This indicated that PM 2.5 was probably transported to Beijing from other cities (Zheng et al., 2015;Ji et al., 2014;Tao et al., 2014; X.J. Zhao et al., 2013).Meanwhile, most severe pollution occurs under stable synoptic meteorological conditions in Beijing (Sun et al., 2015;Zheng et al., 2015; X.J. Zhao et al., 2013).The low wind speed and stable synoptic meteorological conditions at ground level cannot explain the reason why regional transport contributes significantly to severe pollution.A previous study showed that the secondary aerosol in Beijing probably mainly forms over regional transport, according to a vertical observation from the ground to 260 m height (Sun et al., 2015).Therefore, vertical profiles of meteorology and air quality might help us to understand the impacts of regional transport on heavy pollution during stagnant conditions.
As in other megacities with local sources and regional transport, air quality in Beijing is affected by many factors, including emissions inside the city, formation of secondary pollutants, atmospheric mixing and regional transport.It is well known that the strength of each factor varies according to emissions and/or weather conditions.Therefore, it is challenging to pinpoint the major contributors to clean or polluted episodes in any given time period.This is especially difficult in the BTH region, considering the complicated emission sources and transport processes.
Emission control measures implemented during some events provide a unique opportunity to investigate the impact of various factors that influence air quality.One of them was the APEC (Asia-Pacific Economic Cooperation) Economic Leaders' 2014 Summit held in Beijing from 5 to 11 November 2014.A strict emission pollution control plan was carried out in the BTH region to improve air quality in Beijing from 3 to 11 November for APEC.According to a conservative estimate by MEP, production of 9289 plants was paused and 3900 plants were running at reduced capacity in six provinces (Beijing, Tianjin, Hebei, Shanxi, Shandong and Inner Mongolia), and more than 40 thousand construction sites were shut down temporarily (http://www.zhb.gov.cn/gkml/hbb/qt/201411/t20141115_291482.htm).Other measures included traffic control (50 % of private passenger vehicles and 70 % of buses were off-road) and frequent road sweeping and cleaning in Beijing.More detailed emission control measures are supplied in the Supplement.Studies found that regional emission control effectively reduced air pollutant concentrations during the summit (Wen et al., 2015;Tang et al., 2015;Han et al., 2015;C. Chen et al., 2015;Sun et al., 2016a).The significantly reduced local emissions led to reduced complexity of particulate matter pollution processes, thus providing a unique opportunity to investigate the influence of transport events on PM 2.5 levels in Beijing.
The objective of the study is to investigate the impact of regional transport on PM 2.5 in Beijing using both groundlevel and vertical observations.Field observation was conducted at a rural site (Liulihe) southwest of Beijing before and during the control period of the APEC 2014 Summit.Vertical profiles of temperature, RH (relative humidity), wind speed and direction, and extinction coefficient were observed as well as pollutant concentration and meteorological parameters on the ground.The characteristics of three PM 2.5 pollution episodes were analyzed.Findings of this study will help explore vertical observation methods for in-depth analysis of meteorological and transport influence.Furthermore, these findings can aid the development of future air quality management strategies in BTH and other regions around the globe, including emission control and air surveillance.
2 Field observation and analysis methods

Field observation site and sampling methods
Beijing is surrounded by mountains in the west, north and east, which blocks pollutants from spreading.The open air corridor in the south exposes the capital to air masses that are passing Hebei Province (Fig. S1 in the Supplement), a heavily polluted area in China.To investigate the impact of regional transport on Beijing, a rural site (Liulihe site, 116 • 2 E, 39 • 36 N) southwest of Beijing was chosen.It was located on the border of Beijing and Hebei Province (Fig. S1).
The field campaign was conducted from 27 October to 12 November 2014, including both ground-level and vertical observations.All the times discussed in this article are Local Time.Detailed information of instruments at the Liulihe site is provided in Table S1.Ground-level observations included meteorological parameters, mass concentration of PM 2.5 /PM 10 , SO 2 , NO x and O 3 as well as physical and chemical properties of PM.PM 2.5 -PM 10 mass concentration was determined by the tapered element oscillating microbalance (TEOM) method.Particle size distribution from 3 nm to 10 µm was measured by a spectrometer assembled in-house, including one nano scanning mobility particle sizer (NSMPS), one scanning mobility particle sizer (SMPS), and one aerodynamic particle sizer (APS) (Liu et al., 2014).
ACSM (Aerosol Chemical Speciation Monitor), a lowmaintenance aerosol mass spectrometer, was used to measure non-refractory (NR) particulate matter with aerodynamic diameters smaller than 1 µm (PM 1 ) (Ng et al., 2011).The ACSM data were calibrated with a collection efficiency (CE) value to compensate for the particle loss.The CE value of 0.45 recommended by Middlebrook et al. (2012) based on the monitoring site condition (see Supplement) was used in this study.The NR-PM 1 concentration measured by ACSM tracks well with PM 2.5 measured by the TEOM (R 2 = 0.91), and the regression slope is 0.43 (Fig. S2).Positive matrix factorization (PMF) with the PMF2.exealgorithm was used to distinguish different components of OA measured by ACSM (Paatero and Tapper, 1994).The PMF was performed and evaluated following the PMF analysis guide (http://cires1.colorado.edu/jimenez-group/wiki/index.php/PMF-AMS_Analysis_Guide).Three factors were distinguished (Fig. S3), i.e., HOA (hydrocarbon-like organic aerosol), SVOOA (semi-volatile oxygenated organic aerosol) and LVOOA (low-volatile oxygenated organic aerosol).
Beyond ground-level concentrations of routinely monitored air pollutants and meteorological parameters, the assessment was aided by vertical observations, including vertical extinction coefficient profiles as well as vertical wind, RH and temperature profiles.The vertical extinction coefficient profiles depict the distribution of PM, which could be used to infer the mixing process of particles transported from high evaluations and those near the ground.Vertical wind profiles indicate the transport direction.Vertical RH profiles reflect the strength of heterogeneous reaction at different layers.Vertical temperature profiles provide information on the stability of and mixing in the boundary layer.Lidar was used to observe the vertical optical properties of atmospheric aerosols at the Liulihe site.The lidar consisted of three parts, including emitting system, receiving system and signal analogue system (Z.Chen et al., 2015).The laser source emitted a pulse at 355/532 nm.The pulse energy was 30 MJ at 355 nm and 20 MJ at 532 nm.The pulse repetition was 20 Hz.The telescope for the receiving system was based on a Cassegrain design.Diameter of the telescope was 200 mm with a vertical resolution of 7.5 m.The particle backscatter coefficient and extinction coefficient was retrieved using the Fernald method (Fernald, 1984).CFL-03 phased array wind profile radar was used to monitor the vertical wind speed and direction with resolutions of 50 m (0-1 km) and 100 m (1-5.5 km).Parameters of these instruments can be found in another paper (Wang et al., 2013).There was a 300 m blind area for CFL-03.Vertical profiles of atmospheric temperature and humidity were derived by profiling radiometers.The channel center frequencies were 22-32 GHz (K band) and 51-59 GHz (V-Band).The vertical resolutions were 60 m (0-4 km) and 120 m (4-10 km).

Back-trajectory analysis
TrajStat, a GIS-based software into which the HYSPLIT (Hybrid Single Particle Lagrangian Integrated Trajectory) model was loaded (Wang et al., 2009), was used to calculate the back-trajectory.The model was run every 6 h in a 24 h mode back-trajectory mode at 1000 m above sea level from the Liulihe site to identify the origins and pathways of air mass.The meteorology data used in the mode were obtained from the Global Data Assimilation System (GDAS) model (http://www.ready.noaa.gov/READYamet.php).

Quantification of regional transport contribution
A novel technique was used to quantify the contribution of regional transport (Jia et al., 2008).The diurnal trend of PM 2.5 in Beijing often exhibits "saw-tooth cycles" with a smoothly increasing or decreasing baseline upon which daily cycles are superimposed.Ancillary measurements around Beijing show that the baselines represent regional aerosols, while the daily cycles represent local aerosols.Following Jia et al. (2008), the total contribution is defined as the area under the concentration line (A t ), while its regional component is defined as the area under the baseline curve (A r ).Both areas are approximated using trapezoid numerical integration as Eq. ( 1): where N is the total number of hourly PM 2.5 concentrations in a specific time period, C i is the total concentration (for A t ) or baseline concentration (for A r ) value at time t i (i = 1, N − 1).The baseline concentration curve is the line connecting daily afternoon minimal values.The percentage regional contribution (R) is expressed as the following Eq.( 2): The uncertainty evaluation mainly includes systematic errors, random errors and sensitivities.The major systematic errors depend on the calibration of instruments for PM 2.5 concentration measurement.Minor systematic errors might come from judging the location and height of the daily minima, and the sensitivity analysis suggests that these errors are less than 10 %.Random errors include data measurement and quantification steps, such as identifying the daily minima properly, dealing with days without less obvious afternoon minima and using linear interpolation between the daily minima.All these errors were evaluated by Jia et al. (2008).As a whole, this technique has an uncertainty of 40-50 % for the results of daily regional transport.
3 Results and discussion

General characteristics of atmospheric pollution before and during the APEC Summit
To investigate the changes in air quality during the APEC Summit, average pollutant concentrations and the rates of change were calculated.Concentrations of PM 2.5 , SO 2 and NO 2 decreased significantly during the emission control (3-12 November) compared to the period before control (27 October to 2 November), shown in Fig. S4a.Large rates of reduction were observed for NO 2 (37 %) and SO 2 (36 %), while the reduction in PM 2.5 was smaller (21 %) but still significant (Fig. S4b).Three pollution episodes were selected to discuss the pollution characteristics during the observation (Fig. S5).Episode 1 (27 October to 1 November) represents the period before the emission control.Episode 2 (2-5 November) was the first pollution episode during the emission control.Episode 3 (6-11 November) was the second pollution episode during the emission control.PM 2.5 concentration at the Miyun site (located in northern Beijing, shown in Fig. S1, data source: Beijing EPB) is shown in Fig. S5 alongside Liulihe to demonstrate the synchronism of PM 2.5 levels at different sides in Beijing.At Liulihe, PM 2.5 concentration was the highest in episode 1 (140 ± 70 µg m −3 ) before implementation of emission control, whereas the mean values were close during the last two episodes (91 ± 75 and 89 ± 61 µg m −3 ).
The average concentration of online non-refractory PM 1 chemical components was shown in Fig. 1.Average concentrations of OM (organic matter), NH + 4 , SO 2− 4 and NO − 3 were the highest in episode 1 before emission control.During episode 2, those compounds decreased by 32-60 %.In episode 3, the average concentrations remained similar except for NH + 4 , which decreased by 12 %.HOA (related to primary emission), LVOOA and SVOOA were distinguished.Compared with episode 1, HOA, LVOOA and SVOOA decreased by 22, 58 and 28 %, respectively, in episode 2. After that, LVOOA kept decreasing by 10 % in episode 3, while HOA and SVOOA increased by 39 and 5 %.
Overall, most meteorological parameters changed little during the three episodes except for RH (Fig. S6).The average ground-level RH (69 %) in episode 1 was higher compared with those in episode 2 (50 %) and in episode 3 (58 %).Wind speed remained low during the entire observation.The average wind speed was 0.5, 0.8 and 0.7 m s −1 in episode 1, episode 2 and episode 3, respectively.The dominant wind direction was southwest during the 17 days of observation.The frequency of southwestern wind was above 60 % during each of the three episodes, with the highest occurrence of 81 % observed during episode 2.
The significant reduction in pollutant concentrations during APEC implies that the emission control was effective.However, the general characteristics derived from ground-  level observation are insufficient to identify the leading cause of particulate matter pollution, local emissions, regional transport or both.Furthermore, the significant differences in particle chemical component changes from episode 2 to episode 3 under similar ground-level meteorological conditions and local emission intensity suggest different transport or formation mechanisms during those two episodes.Therefore, vertical observations will be used to aid further investigation in each of the three episodes in the following section.

Characteristics of heavy PM 2.5 pollution episodes
and contribution of regional transport

Pollution process in episode 1
Episode 1 (27 October to 1 November) was before emission control.The average concentration of PM 2.5 reached 140 ± 70 µg m −3 .This high level of PM 2.5 is typical in Beijing during autumn.There were two unique features in this episode.One is the continued increase of PM 2.5 mass and PM 1 component concentrations during the first 4 days, with OM showing a more distinct diurnal cycle (Figs.2, 3 and S5).
Another is the rapid increase of OM on 29 October (Fig. 3).Both suggest that other mechanisms, except for secondary formation, might impact the OM growth and need further investigation.Various parameters collected during episode 1 are shown in Fig. 4. Combining the ground-level observation and vertical observation, it is evidenced that the pollution was caused by regional transport and later pollutant accumulation.Vertical extinction coefficient data observed at the Yongledian site ( 116• 47 E, 39 • 43 N) near the Liulihe site were used (Fig. 4a) because the optical lidar at Liulihe did not work in October.High levels of PM appeared at approximately 2 km above ground (Fig. 4a) and remained there for 1 day.The air mass came from the southwest, where emissions were high (see horizontal wind direction profile, Fig. 4c).Backtrajectories also show air masses from the southwest arriving in Liulihe as well as Yongledian (Fig. S7).Then pollutants settled down (see downward vertical wind direction in Fig. 4b) and mixed with aerosols on the ground (Fig. 4a).The online particle size distribution also implied transport processes.During the same period (from 13:00 to 20:00 LT on 28 October), a new group of particles appeared and mixed with existing particles, indicating the arrival of aged aerosols (Fig. 4e).As mentioned above, except for secondary formation, other mechanisms might impact OM increase.The increase of OM might come from freshly emitted organic particles that are transported to the site instead of aged particles.One piece of evidence is that both HOA and OOA increased significantly.Another is that the OM peak appeared after the transport occurrence, much earlier than SNA.Notice that even though wind direction on the ground changed to north early in the morning on 29 October, it still remained from the southwest above 500 m, indicating significant influence of regional transport.
In the next 2 days (30-31 October), vertical wind direction was downward, which was unfavorable for pollutant diffusion (Fig. 5a).Weather Research & Forecasting Model (WRF) modeling results also showed that the whole region was under the control of weak downward wind from late on the night of 30 October (Fig. S8, modeling parameters are provided in the Supplement).Furthermore, both the atmosphere press and wind speed decreased at the same time (Fig. S6).This indicates that the site was probably in the rear of a cold anticyclone.The steady weather conditions promoted pollutant accumulation.Meanwhile, high RH on the surface (Fig. S6) enhanced the formation of SA (secondary aerosol), as pointed out by Pathak et al. (2009).Under this condition, NH + 4 , SO 2− 4 and NO − 3 concentrations increased at rates of 0.26, 0.21 and 0.58 µg m −3 h −1 , respectively.The peak of NH + 4 , SO 2− 4 and NO − 3 concentrations was 2 days later than OM.This also proved that the organic particles were transported to Beijing and reached the peak on 29 October and that secondary formation became severe later, both of which promoted the pollution occurrence.
To quantify the impact of regional transport, the transport component is calculated with the method introduced in Sect.2.2.The baseline needs to be defined first, especially for pollution end timing.The vertical observation and ground observation were combined to discuss when the pollution ended (see Supplement).The regional component is calculated based on the determination of the baseline.For episode 1, the regional component accounted for 75 % of PM 2.5 mass concentration observed at the Liulihe site, indicating the important influence of regional transport on the pollution.It can be seen that episode 1 was a pollution episode influenced by transport processes in Beijing.RH was high, wind speed was continuously low and wind direction was mainly from the southwest on the surface.Vertical observation showed that the pollutants transported from the southwest settled down.OM concentration increased significantly when the transport PM was observed.After that the low wind speed and high RH easily promoted pollutant accumulation, and downward vertical wind was unfavorable for pollutant diffusion.

Pollution process in episode 2
Episode 2 (2-5 November) saw a lower mean PM 2.5 concentration (91 ± 75 µg m −3 ) due to the implementation of emission control since 2 November.Unlike the gradual accumulation of PM observed in episode 1, PM 2.5 , OM and SNA had a sharp increase from 4-5 November.The concentrations of NH + 4 , SO 2− 4 and NO − 3 increased at rates from the lowest to the highest of 0.88, 0.43 and 1.64 µg m −3 h −1 , respectively, much faster than the rates in episode 1. OOA also increased much more significantly during this episode.The explosive increases of PM components, mainly SA, in such a short period of time is contrary to lower RH values in this episode, leading to a less heterogeneous reaction.Thus, such rapid increases in PM levels could be due to the transport of aged aerosol from other regions, as hypothesized by previous studies where the transport process was not directly observed (Yue, et al., 2009;Massling et al., 2009;Sun et al., 2014Sun et al., , 2016b)).
With the aid of vertical observation, an in-depth investigation revealed atmospheric processes leading to the peak concentrations during 4-5 November.Firstly, after episode 1 ended on 1 November, relatively high PM levels still resided at 1000 m (from 2 to 3 November), as shown in the vertical extinction coefficient (Fig. 7).Furthermore, a band of  high PM centered around 750 m was observed (Fig. S9) on 3 November at another site (Baoding site, 115 • 31 E, 38 • 52 N; shown in Fig. S1) in the BTH region, suggesting a widespread PM aloft in the region.During the next 2 days, the pollutants were transported in the region and the slow winds (average speed of 4.8 m s −1 at 1000 m) allowed aerosols ample time to age in their journey.Backtrajectories showed transport of air mass from the southwest on the night of 3 November (Fig. S10), consistent with the vertical wind profile observed at Liulihe (Figs. 5 and 9).On 3 and 4 November, the downward motion of air mass around 1000 m above ground intensified, bringing the aged aerosols down and mixing them with the aerosols on the ground.The well-mixed boundary layer with regard to aerosol is evidenced in Fig. 8 with a fairly uniform distribution from the ground to 900 m.Consequently, secondary chemical component concentrations of PM 1 (Figs. 2 and 3) started ascending at remarkably fast rates.
Dry and clean air mass from the north arrived early in the morning on 5 November.RH started to increase significantly at 10:00 LT and wind speed became higher at 12:00 LT .At the same time, PM 2.5 concentration started to decrease.Based on the analysis, the pollution episode ended at 12:00 LT .The calculation shows that regional transport contributed 62 %, relatively lower than that during episode 1 (Fig. 6).
Rather than chemically reacting, aged aerosols settled down and contributed significantly to the high PM 2.5 concentration in episode 2. Vertical observations found that the aged aerosol settled down and caused the explosive increase of SNA in such a short time, which cannot be explained by the ground-level observations.It was also noticed that the high PM 2.5 level appeared when the emission control just started, which means this episode was partly caused by regional transport before control.Even when local emission control was conducted effectively, the uncontrolled regional  emission still led to severe particulate matter pollution in Beijing.

Pollution process in episode 3
During episode 3 (6-11 November), the Liulihe site recorded a relatively high average PM 2.5 concentration of 89 ± 61 µg m −3 .Furthermore, this episode is characterized by much more and faster increases in OM concentrations than SNA (Figs. 2 and 3).Specifically, concentrations of aerosol related to fuel combustion (HOA) increased significantly, while SNA increased slowly (NH + 4 and NO − 3 ) or changed little (SO 2− 4 ).All of these indicate that primary emission rather than the formation of SA was the dominant cause.
Vertical extinction coefficient shows that pollutants appeared at 2000-2500 m on 7 November.The air mass came from the northwest and the vertical convection bringing them down on 7 and 8 November (Figs. 7,5b and 9).Air mass trajectories at 1000 m also show that the air mass arrived in Beijing from the south on 7 November but changed to the northwest on 8 November (Fig. S11).Because of less pollution in the northwest and the effective emission control in the BTH region during the APEC, the regional transport of PM was weakened.This is supported by an estimated regional contribution of 53 % to PM 2.5 in Beijing, much lower than in episode 1 (75 %) and episode 2 (63 %).Concentrations of BC, a marker of vehicular emission in urban settings, had two peaks every day.One was early in the morning and another was after morning rush hour around 10:00-11:00 LT .The first peak might have resulted from diesel vehicle emissions (Westerdahl, et al., 2009).This is because transportation of goods to Beijing via heavy-duty diesel vehicles has been permitted at night only, and the number of trucks was large.The second peak might have resulted from vehicles from outside Beijing coming into the city.Vehicles not registered in Beijing are banned from coming into Beijing during the rush hour (07:00 to 09:00 LT ), which reduces the morning peaks and smoothes the traffic flow.The vehicles coming into Beijing reach a peak after morning rush hour (http://max.book118.com/html/2016/1022/60419427.shtm).As a result, a second peak appeared late in the morning at the Liulihe site, which is close to the entrance from Hebei Province into Beijing.When the regional emission control was conducted effectively and air mass came from relatively clean areas, traffic emissions in and around the city became the dominant source.

Conclusion
This study indicates that sometimes the meteorological conditions on the ground could not explain the air pollution process, especially the air pollution episodes significantly impacted by regional transport of air pollutants.Vertical observation can provide the vertical meteorological and optical profile, which can help identify the regional transport episodes.Combining the ground-level observation with information from radar, we can determine the regional transport influence on air quality.
Three episodes of different types under similar ground meteorological conditions were discussed in this study.In episode 1, particle concentration accumulated under the unfavorable meteorological condition after transport occurred.The transport pollutants later brought organic aerosol and SNA increased under high RH.In episode 2, pollutants left  from episode 1 were retained in the boundary layer in the region.When vertical wind direction changed to downward, the pollutants settled down.As a result, OM and SNA increased explosively.In episode 3, when the control had been conducted for several days, SNA and OA concentration increased much less, while HOA and increased significantly.
The pollution might be caused by the primary emission from diesel vehicles.
Our research suggests regional transport of air pollutants has significant contribution (up to 70 %) to severe secondary particle pollution, even when local emission was controlled effectively (53 %, such as during the APEC summit).Although lots of attention was paid to air quality management in Beijing, equal efforts need to be paid to regional emission to ensure the clean air.Furthermore, diesel vehicle emission at night in Beijing might be an important pollution source and needs further investigation.

Data availability
The data of this paper are available upon request to Shuxiao Wang (shxwang@tsinghua.edu.cn) or Yang Hua (huay09jane@126.com).
The Supplement related to this article is available online at doi:10.5194/acp-16-15451-2016-supplement.

Figure 1 .
Figure 1.Non-refractory PM 1 chemical components at the Liulihe site during the three episodes.(a) Average non-refractory PM 1 chemical components, (b) differences of chemical components among episodes.

Figure 2 .
Figure 2. Temporal changes of non-refractory PM 1 chemical components at the Liulihe site.

Figure 3 .
Figure 3.The temporal changes of organic components in PM 1 at the Liulihe site.

Figure 4 .
Figure 4. Characteristics of particulate matters and meteorological parameters during episode 1.(a) Vertical profile of extinction coefficient (Yongledian site), (b) vertical profile of wind vertical direction and speed, (c) horizontal wind direction profile, (d) wind direction on the ground, (e) particle size distribution, (f) NR-PM 1 chemical components.

Figure 5 .
Figure 5. Vertical profile of wind at the Liulihe site.(a) Wind vertical speed, (b) wind horizontal direction profile.

Figure 6 .
Figure 6.Regional and local components of the three episodes at the Liulihe site.

Figure 7 .
Figure 7. Vertical profile of extinction coefficient at the Liulihe site.

Figure 10
Figure 10 depicts black carbon (BC) concentrations measured by an aethalometer and OM concentrations measured by ACSM.They tracked each other well during this episode.Concentrations of BC, a marker of vehicular emission in urban settings, had two peaks every day.One was early in the morning and another was after morning rush hour around 10:00-11:00 LT .The first peak might have resulted from diesel vehicle emissions(Westerdahl, et al., 2009).This is because transportation of goods to Beijing via heavy-duty diesel vehicles has been permitted at night only, and the number of trucks was large.The second peak might have resulted from vehicles from outside Beijing coming into the city.Vehicles not registered in Beijing are banned from coming into Beijing during the rush hour (07:00 to 09:00 LT ), which reduces the morning peaks and smoothes the traffic flow.The vehicles coming into Beijing reach a peak after morning rush hour (http://max.book118.com/html/2016/1022/60419427.shtm).As a result, a second peak appeared late in the morning at the Liulihe site, which is close to the entrance from Hebei Province into Beijing.When the regional emission control was conducted effectively and air mass came from relatively clean areas, traffic emissions in and around the city became the dominant source.

Figure 8 .
Figure 8. Characteristics of particulate matters and meteorological parameters during episode 2. (a) Vertical profile of extinction coefficient, (b) vertical profile of wind vertical direction and speed, (c) horizontal wind direction profile, (d) wind direction on the ground, (e) NR-PM 1 chemical components.

Figure 9 .
Figure 9. Characteristics of particulate matters and meteorological parameters at the Liulihe site during episode 3. (a) Vertical profile of extinction coefficient, (b) vertical profile of wind vertical direction and speed, (c) horizontal wind direction profile, (d) wind direction on the ground, (e) NR-PM 1 chemical components.

Figure 10 .
Figure 10.BC and OM concentrations of PM 1 at the Liulihe site during episode 3.