A Severe Air Pollution Event from Field Burning of Agricultural Residues in Beijing , China

Air pollutant emissions from agricultural burning are observed every year after harvest in China. While agriculture is not the main contributor to air pollution in China, agricultural activities can cause severe pollution events. Recognizing the key mechanisms involved in this process offers an opportunity to minimize pollution events caused by agricultural burning. In this paper, we review the meteorological conditions present during a selected air pollution episode and discuss those conditions using standard meteorological observations. The spatio-temporal variations of PM2.5 concentrations following agricultural burning in Beijing were measured from October 4 to October 7, 2013. This time period coincides with a Chinese public holiday and was selected because the influence of other anthropogenic emissions on air quality was strongly reduced during those days. As a result, we were able to identify the key sources and progress of a severe air pollution event. On October 4, average PM2.5 concentration in Beijing continuously increased from 49.7 μg m at 1:00 to 302.5 μg m at 23:00. Heavily polluted air (> 300 μg m) initially appeared in southeastern Beijing on the afternoon of October 4. On October 5 and in the early morning of October 6, heavily polluted air masses moved into central Beijing, the inner suburbs, and the suburbs. From 0:00 on October 6 to 15:00 on October 7, the average PM2.5 concentration in Beijing decreased from 291.6 μg m to 19.2 μg m. Active fire information derived from the MODIS sensors and back trajectory analysis show that field burning of agricultural residues after a harvest triggered and massively contributed to this severe air pollution event. The results improve our understanding of PM2.5 air pollution development processes, and they provide scientific support for the Chinese government to accelerate emission reductions from the field burning of agricultural residues.


INTRODUCTION
Small particulate matter air pollution has significantly increased in China during the last 10 years due to rapid industrialization and urbanization (Huang et al., 2014;Fu et al., 2014).Changes in the industrial structure, especially in the energy sector, have accelerated the deterioration of China's air quality; and satellite-based aerosol optical depth research indicates that eastern China has the highest concentrations of airborne fine particulate matter (PM 2.5 ) in the world (van Donkelaar et al., 2010).Furthermore, severe air pollution has resulted in significant negative impacts on public health, particularly in megacities (Schwartz et al., 1996;Donaldson et al., 2005;Parrish and Zhu, 2009;Burgan et al., 2010;Cao et al., 2014), and it has direct and indirect effects on climate change (IPCC, 2001;Lohmann and Feichter, 2005;Li et al., 2011).
Several studies have correlated aerosol dynamics with primary meteorological and micrometeorological parameters (Sun et al., 2006;Vardoulakis and Kassamenos, 2008;Zhao et al., 2009;Donateo et al., 2012;Chang et al., 2013).Zhang et al. (2009) studied the correlation between air pollution and meteorological elements by using the PLAM (Parameter Linking Air Quality and Meteorological Elements) index, and they found that "bad" air quality days with elevated PM 10 (> 150 µg m -3 ) in the summer were always associated with high temperature, high humidity, low wind speed, and high stability.Air stagnation conditions (e.g., temperature inversions, light winds, a lack of precipitation, nocturnal mixing height) trap air pollutants close to the ground and allow them to build up and spread horizontally.As a result, air pollution events are directly related to regional and local meteorology.
Integrated near-ground air pollutant monitoring and satellite remote sensing allow to effectively analyze spatial and temporal variations in PM 2.5 concentrations during severe air pollution events, as has been reported by numerous researchers (Liu et al., 2009;Blanchard et al., 2011;Chudnovsky et al., 2013;Nordio et al., 2013;Li et al., 2014;Pope and Wu, 2014).The spatial distributions of different emission sources may generate different spatial and temporal variation patterns for pollutants.Pronounced monthly, weekly, and hourly PM 2.5 concentration variations have been observed from ground stations in Beijing (Song et al., 2006).PM 2.5 concentrations near the ground layer in Beijing have been shown to peak near midnight in autumn (Yang et al., 2005), while PM 2.5 and PM 10 concentrations at street locations in urban areas were found to be higher than those at other urban and rural locations (Holst et al., 2008;Liu et al., 2011;Contini et al., 2012;Keuken et al., 2013).PM 2.5 concentrations were also typically found to be lower at elevated sites surrounding the San Joaquin Valley than at monitoring sites located in the valley (McCarthy et al., 2005).Huang et al. (2014) analyzed the chemical composition of and sources responsible for PM 2.5 collected at urban sites in Beijing during high pollution events from January 5 to 25, 2013, and they found that secondary organic-rich and inorganic-rich aerosols were the primary pollutant contributors.During the post-harvest season, field burning of agricultural residues is the primary producer of secondary aerosols.Field burning of agricultural residues contributed 37% of the PM 2.5 , 70% of the organic carbon and 61% of the elemental carbon found in a heavy haze episode during a post-harvest season in the Yangtze River Delta, China (Cheng et al., 2014).Particles from field burning of agricultural residues contributed more than 35% of aerosol optical depth over regions of the Jiaodong Peninsular, the North Plain, East China and other areas, and exceeded 60% in some areas of Shandong, Henan and Jiangsu provinces (Zha et al., 2013).Based on recent source distribution results, Zhang and Cao (2015) suggested that strengthening the regulations on emissions from biomass burning in both urban and rural areas is needed in order to meet China's rigorous emissions reduction target.
In this paper, we analyze the meteorological conditions and the hourly spatio-temporal variation of PM 2.5 concentrations in Beijing during a heavy air pollution episode that occurred from October 4 to 7, 2013.Active fire information derived from the MODIS sensors revealed the spatial distribution of post-harvest fires.We also analyze the movement of pollutants using back trajectory analysis.This study improves our understanding of the heavy air pollution event development processes and aims to accelerate emission reductions from agricultural field burns.

Study Area
Beijing is located on the northern tip of the North China Plain, near the junction of the Xishan and Yanshan Mountain ranges, and has a total area of 16410.54km 2 .The metropolis had a population of 21.1 million in 2013 (BMBS, 2014).Beijing currently comprises 14 urban and suburban districts and two rural counties.Twelve expressways and eleven national highways run across Beijing, and urban transportation is dependent upon six "ring roads" that are concentrically distributed in southern Beijing.In 2012, there were 5.2 million registered automobiles in Beijing (BMBS and SOB-NBS, 2013).Mountains surround 62% of the region, from southwest to northeast.

Data Analysis
Average wind speed and relative humidity were calculated using daily data obtained from the China Meteorological Administration (http://cdc.nmic.cn/home.do).The monitoring height for the meteorological data is 10-12 m.The planet boundary layer height (PBLH) data were available directly from the National Center for Environmental Prediction (NCEP)'s website at http://rda.ucar.edu/cgi-bin/datasets/dataaccess?dsnum=083.2.Upper air temperature, relative humidity, wind speed and wind direction profiles were derived from the sounding data provided by the University of Wyoming (UTC 0:00 and UTC 12:00) (http://www.weather.uwyo.edu/upperair/sounding.html).
Thirty-five air quality monitoring stations are located in Beijing (Fig. 1) and provide hourly PM 2.5 concentration data.Five stations (Qianmen East Street, Yongdingmennei Street, Xizhimen North Street, Southern 3nd Ring Road, Eastern 4th Ring Road) monitor traffic pollution.The data from these stations are released by the Beijing Municipal Environmental Monitoring Center (http://zx.bjmemc.com.cn).PM 2.5 concentration data have a sampling height of 3-15 m above the ground (MEP, 2012).Traffic pollution monitoring stations have a sampling height of 2-5 m.If the average building height of buildings within a radius of 300-500 m around a monitoring station exceeds 20 m, the sampling height is 15-25 m.The average PM 2.5 concentration within Beijing is calculated using hourly PM 2.5 concentration data from the 35 air quality monitoring stations.
We have chosen six air quality monitoring stations to represent different PM 2.5 variation patterns among the 35 sites during the studied air pollution episode.Two stations (Fengtai Garden and Qianmen East Street) are near ring roads within the urban center (inside five ring roads).Two stations (Yongledian and Donggaocun) are located in southeastern and southern Beijing (rural areas), where heavily polluted air masses appeared on the afternoon of October 4. PM 2.5 concentration data interpolation was performed using 300 × 300 grid files in Surfer 10 (Golden Software, Inc.) and a linear variogram model based on an ordinary (point) Kriging method.Then, the grid files were converted to DEM files, which were used to draw isoline maps of PM 2.5 concentration in ArcGIS 10.1.Daily variations in PM 2.5 concentrations are usually higher around midnight than at any other time, but they may vary with emission source activities, local atmospheric oxidation capacity (i.e., OH radical level) and the meteorology at the boundary layer (i.e., PBLH, inversion).We generated isoline maps of PM 2.5 concentrations at 2:00, 10:00, and 18:00 local time every day during the severe air pollution event, representing air pollution conditions in Beijing during the morning, afternoon and night, respectively.

Spatial Distribution of Fires
We used the Fire Information for Resource Management System (FIRMS) to identify the spatial distribution of PM 2.5 emissions from agricultural residue field burns.FIRMS provides active fire information derived from the MODIS sensors onboard NASA's Aqua and Terra satellites (https://firms.modaps.eosdis.nasa.gov/download/request.php).Each active fire location represents the center of a 1 km pixel that is flagged by the algorithm as containing one or more fires within that pixel.

Back Trajectory Analysis
We used the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model developed by the Air Resources Laboratory of the National Oceanic and Atmospheric Administration (NOAA) to perform a back trajectory analysis (http://ready.arl.noaa.gov/hypub-bin/trajtype.pl?runtype=archive).Heavily polluted air initially appeared in southeastern Beijing at approximately 18:00 in the afternoon on October 4. Hence, the starting time for the back trajectory analysis was set to be at 10:00 UTC (Coordinated Universal Time) on October 4, 2013.The back trajectory was calculated using GDAS (Global Data Assimilation System) meteorological data for the preceding 48 h in southeastern Beijing (39.7°N, 116.8°E) at 10:00 UTC (18:00 local time) on October 4-7 at 500, 1000, and 2000 m altitudes, respectively.

Meteorological Conditions
The local circulation of air masses influences the spatiotemporal variation of PM 2.5 concentrations.On October 4-6, a low pressure system was located over Beijing and the area was subject to southwestern and southeastern winds.The stable synoptic conditions led to high PM 2.5 concentrations and weak dispersion throughout the region.On October 7, a shift in the low pressure system and steadily building high pressure quickly dispersed the high concentration PM 2.5 pollutants in Beijing.The surface and 850 mb weather charts on October 5 and October 7 were described in detail in the supplementary material file.
Slight winds and high humidity may suppress the diffusion of air pollutants and facilitate the accumulation of PM 2.5 particles.The average wind speed near the ground layer (10-12 m) was 1.3 m s -1 in Beijing, 1.8 m s -1 in Hebei, 2.3 m s -1 in Tianjin, 2.4 m s -1 in Shandong, 1.5 m s -1 in Henan, 3.2 m s -1 in Inner Mongolia, and 1.6 m s -1 in Shanxi from October 4-7.The average wind speed near the ground layer in Beijing was 1.7 m s -1 on October 3 and 1.5 m s -1 on October 8.During the pollution episode, the average relative humidity near the ground in Beijing was 79%.Heavy pollution rapidly emerged when the near-ground wind speed in Beijing was low (below 2 m s -1 ) and the relative humidity was high (greater than 60%).
PBLH plays an important role in the enhancement of air pollution events.The average PBLH during the studied episode was 654 m.At night, the average PBLH was 594 m, which was lower than the average measured during the day (713.2m).Fig. 2 shows that PBLH dramatically decreased from 1278 m at 0:00 on October 4 to 197 m at 18:00 on October 7.The change in PBLH was associated with the development of the air pollution event.The low altitude of the PBLH favored the accumulation of air pollutants.
Profiles of temperature, relative humidity, wind speed, and wind direction from 250 hPa to 1000 hPa are presented in Fig. 3 during the study period.The temperature profile indicates that on October 4, a clear temperature inversion from 971 hPa (371 m) to 882 hPa (1183 m) existed at 8:00 am local time (UCT:0:00).The daytime temperature inversion layer slightly dropped on October 5, rose above 1000 m on October 6, and dramatically dropped on October 7. Temperature inversions during the study period were low at night and located at approximately 1000 hPa, and 850 hPa on October 4; 1000 hPa, 820 hPa, and 650 hPa on October 5; 1000 hPa, 780 hPa, and 320 hPa on October 6; and 1000 hPa, and 700 hPa on October 7.
The relative humidity profile shows that during the episode, high humidity (> 50%) levels were recorded inside the PBL above Beijing, except at night on October 7.During the night of October 7, the relative humidity below 2076 m was less than 50%.During the day, the vertical thickness of the high humidity layer increased from 1000 m on October 4 to 1683 m on October 6, and decreased to 863 m on October 7.At night, the vertical thickness of the high humidity layer increased from 1414 m on October 4 to 1667 m on October 5, decreased to 845 m on October 6, and increased again to between 2315 m and 4519 m on October 7.

Spatio-Temporal Variation of PM 2.5 Concentration
The average PM 2.5 concentration in Beijing was 89.5 µg m -3 in 2013 (BMEPB, 2014).The average PM 2.5 concentration in Beijing from October 4 to October 7, 2013 was 182.7 µg m -3 , over two times higher than the annual average.A clear spatial pattern was identified during the episode.Heavily polluted air initially appeared in southeastern Beijing on the afternoon of October 4 (Fig. 4).On October 5  th 02:00 4 th 10:00 4 th 18:00 5 th 02:00 5 th 10:00 5 th 18:00 6 th 02:00 6 th 10:00 6 th 18:00 7 th 02:00 7 th 10:00 7 th 18:00 and 6, heavily polluted air masses moved into central Beijing, the inner suburbs, and the suburbs.On the morning of October 5, air pollution in central Beijing intensified and lasted until the morning of October 6.After the morning of October 6, air quality in the urban and rural areas began to improve, with monitored hourly PM 2.5 concentrations rapidly returning to the national air quality standard of less than 50 µg m -3 .Highly polluted areas were spatially identified.At 9:00 on October 4, PM 2.5 concentration exceeded 150 µg m -3 to the southwest of the 4 th Ring Road, at the Fengtai Garden station.Due to regional transportation and continuous local emissions (traffic and some industrial factories), this area remained continuously polluted (PM 2.5 concentration >150 µg m -3 ) for 72 hours (from 9:00 on October 4 to 9:00 on October 7).The Qianmen East Street traffic monitoring station was the first station where PM 2.5 concentrations exceeded 250 µg m -3 , which occurred at 13:00 on October 4. With the exception of 6 minor decreases (208-248 µg m -3 ) on October 6, PM 2.5 concentrations at the Qianmen East Street station remained over 250 µg m -3 until 3:00 on October 7 (57 hours).PM 2.5 concentrations remained over 250 µg m -3 for 55 hours at the Xizhimen North Street, 50 hours at the Fengtai Garden station, 18 hours at the Yongledian station, and 13 hours at the Donggaocun station.This indicates that the worst pollution occurred close to the streets in the core urban areas (Qianmen East Street and Xizhimen North Street) and suggests that traffic emissions contributed to the identified severe air pollution event.Mountain stations (Dingling, Changping, and Yanqing) measured comparably lower PM 2.5 concentrations.
The temporal variation in total average PM 2.5 concentrations in Beijing revealed an obvious trend during the event.Total average PM 2.5 concentrations continuously increased on October 4 and then decreased from October 6-7.From 1:00 to 23:00 on October 4, average PM 2.5 concentrations in Beijing continuously increased from 49.7 µg m -3 to 302.5 µg m -3 .From 0:00 to 23:00 on October 6, average PM 2.5 concentrations in Beijing decreased from 291.6 µg m -3 to 175.2 µg m -3 .From 0:00 to 15:00 on October 7, average PM 2.5 concentrations in Beijing decreased from 171.4 µg m -3 to 19.2 µg m -3 .
More specifically, three variation patterns for different stations were identified during the episode, which is shown in Fig. 5. On October 5-6, PM 2.5 concentrations for stations near roads (Fengtai Garden and Qianmen East Street) increased during the morning and evening commute hours.Significant peaks (500 µg m -3 ) occurred around midnight, following the mixing of accumulated and source emissions (Type 1).This variation pattern shows that traffic emissions locally contributed to the air pollution event.PM 2.5 concentrations at Yongledian, which is located in southeastern Beijing, and Donggaocun, which is located in eastern Beijing, reached their maximum values on the afternoon and evening of October 4, respectively.Afterwards, these values continually decreased, exhibiting small fluctuations (Type 2).This pattern suggests that emission sources had a considerable impact on air quality at these stations on October 4.However, the sources were not persistent because, after October 4, PM 2.5 concentrations continually decreased at these stations.Hourly PM 2.5 concentration patterns at Yanqing (suburb) and Wanliu (core urban area) varied  distinctly from the two patterns identified above.Here, peak PM 2.5 concentrations appeared on the evening of October 5 (Type 3), indicating that PM 2.5 accumulated slowly over a period of hours during the severe air pollution event.The peaks lagged because these two stations were comparably farther from the agricultural residue field burning emission sources, and the dispersal of pollutants was blocked by nearby mountains.

Fires from Agricultural Residue Field Burns
According to China's National Air Pollution Prevention and Control Law (revised in 2000), field burning of agricultural residues, fallen leaves, etc. is only prohibited in densely inhabited areas, and near airports, traffic trunk lines, and certain other areas designated by local governments.However, poor supervision, management, technical, and related policy support systems have resulted in an inefficient and irregular implementation of this law.In China, farmers only have a few days in autumn after the corn, cotton, and peanut, etc. harvest to plough the land and sow wheat or other crops.Burning agricultural residues in the field is the most economical and effective choice for farmers who need to quickly make way for their next crop.
The spatial distributions of fires in China from October 2-7 2013 were identified using FIRMS (Fig. 6).Fires were identified in southern Beijing just prior to the air pollution event (October 2-3).Most fires were located in the Shandong, Henan, Shanxi, Hebei, Hubei, and Guangdong provinces.When the episode began in Beijing and reached its peak on October 4-5, the number of fires had increased in Shandong, Hebei, Anhui, Jiangsu, and Inner Mongolia.On October 6-7, the number of fires had dramatically decreased across China.Accordingly, and supported by changing meteorological conditions, the average PM 2.5 concentration in Beijing decreased.

Back Trajectory Analysis of PM 2.5
A back trajectory analysis of PM 2.5 shows that from  October 4-6 PM 2.5 particles were transported into Beijing at a height of 500 m from Anhui, Shandong, Henan, and Hebei; at a height of 1000 m from Shaanxi, Shanxi, Shandong, Henan, and Hebei; and at a height of 2000 m from Inner Mongolia, Shanxi, Shandong, and Hebei, where fires were intensively distributed (Fig. 7).A back trajectory of PM 2.5 concentrations on October 7 shows that then air was transported into Beijing from Inner Mongolia, where the air was clean and very few fires were identified.Because October 4-7 was a holiday, most factories, companies, and government departments were closed.Moreover, the heating season does not begin until November 15 in Beijing.No cooking, dust storms, or major construction events were reported around southeastern Beijing during the holiday, eliminating increasing coal burning, cooking, and dust-related emissions as potential major contributors to this event.Traffic was a potential major cause of smog because people are known to travel to visit relatives or for leisure on the holiday.Expressways and national highways in Beijing are distributed in every direction, with major roads in south and southeastern Beijing.The traffic jam index during Chinese National Day was 0.9-2.2(unimpeded) (BMCT, 2013), even though some roads near Beijing's entrance and exit gates were impeded on Oct. 1, 2, 6, and 7.This suggests that traffic was also not a large contributor to the studied severe air pollution event of October 4. We can conclude that the field burning of agricultural residues upwind of Beijing triggered and massively contributed to an air pollution event when local anthropogenic emissions were suppressed during a public holiday.However, traffic did contribute to local pollution concentration increases during the event.

CONCLUSION AND DISCUSSION
We reviewed the meteorological conditions present during an air pollution episode using meteorological observations, analyzed the spatio-temporal variation of PM 2.5 concentrations, and used satellite images and back trajectory analyses to identify the sources and progression of a severe air pollution event.
The stable synoptic conditions, temperature inversion, low PBLH, high humidity within the PBL, weak wind speed, and southern winds led to high PM 2.5 concentrations and weak dispersal over Beijing on October 4-6.Following the occurrence of a high pressure system, a decrease in relative humidity, an increase in wind speed, a change in wind direction, and a decrease in fires on October 7, the high PM 2.5 concentrations quickly dissipated.
The air pollution began in southeastern Beijing on the afternoon of October 4. Total average PM 2.5 concentrations continuously increased on October 4. Heavily polluted air then gradually moved from southeastern to central Beijing.On October 5 and 6, heavily polluted air fluctuated between the central urban area and the suburbs.On October 6-7, the total average PM 2.5 concentration in Beijing decreased, with monitored hourly PM 2.5 concentrations finally returning to levels within the national air quality standard of less than 50 µg m -3 on the afternoon of October 7. Three spatio-temporal variation patterns were identified and each pattern revealed different emission source activities and concentration processes during the air pollution event.The main source that triggered and then amplified this severe pollution event was the field burning of agricultural residues upwind of Beijing.This can be shown by comparing the FIRMS and back trajectory results with the spatiotemporal variation of PM 2.5 concentrations, while taking into account that other local anthropogenic emissions were reduced during a public holiday and that coal burning for heating had not yet stated.
The relative contribution of agricultural field burning in this event be assessed as follows.According to the PM 2.5 concentration isoline map, PM 2.5 concentrations increased from the interval 50-100 to the interval 400-450 µg m -3 on October 4 in southeastern Beijing.That this increase was due to agricultural field burning follows from the fact that due to the fall holiday conditions other air pollution was nearly exclusively from local transportation, industry, and cooking, none of which increased dramatically in those days.This implies that the contribution of agricultural field burning to the subsequent pollution levels in southeastern Beijing was around 80%, because the increase of PM 2.5 concentration (350 µg m -3 ) accounted for 77.8%-87.5% of the total PM 2.5 concentration (400-450 µg m -3 ).If we do the same calculation for northern Beijing, which is comparably farther away from the fires and located in the mountains, the contribution of field burning to increased pollution level was around 60%, because the increase of PM 2.5 concentration (100 µg m -3 ) accounted for 50%-66.7% of the total PM 2.5 concentration (150-200 µg m -3 ).Thus, the contribution of agricultural field burning to air pollution in this episode was around 60-80%.
Unlike the persistent emissions characteristic of industry, transportation and cooking, or the uncertainty of emissions from dust storms and forest fires, field burning of agricultural residues only occurs immediately after a harvest season.As a result, these emissions are comparably easier to control.Though a National Air Pollution Prevention and Control Law and other government documents ban the field burning of agricultural residues, many fires were identified across China during the post-harvest season.In the case of air pollution events caused by agricultural residues field burning, governments can and should intervene quickly to reduce air pollution.Such actions can create positive pollution control experiences that are important also for broader policy measures.Available technologies that turn agricultural residues into diesel and biochar should be made more effective and profitable to enable farmers to participate in sustainable agriculture.Under unfavorable meteorological conditions, the ban on agricultural field burning upwind of Beijing (and similar cities) should be rigorously enforced.In addition to policy and legal guidance, public participation, interest litigation, and positive market-based incentives should be encouraged.Our results provide clear scientific support for the Chinese government to accelerate emission reductions from the field burning of agricultural residues during post-harvest seasons.

Fig. 1 .
Fig. 1.Topographic map of the locations of air quality monitoring stations (AQMS), and of road distribution in Beijing (1 Xizhimen North Street; 2 Guanyuan; 3 Dongsi; 4 National Agriculture Exhibition Center; 5 Qianmen East Street; 6 Wanshouxigong; 7 Tiantan; 8 Yongdingmennei Street).Green denotes elevation.The blue marked AQMS sites are chosen to analyze temporal variation of PM 2.5 concentration during the air pollution episode in Fig. 5.

Fig. 4 .
Fig. 4. Isoline map of PM 2.5 concentrations in Beijing during the air pollution event.The black lines are isolines of PM 2.5 concentration.The white lines are ring roads.

Fig. 5 .
Fig. 5. Temporal variation of PM 2.5 concentrations in representative AQMS sites in Beijing during the air pollution event.

Fig. 7 .
Fig. 7. Back trajectories of PM 2.5 from October 4-7, 2013 at 500 m, 1000 m, and 2000 m altitudes, respectively.The blue lines are 48 hours back trajectories on October 4. The green lines are 48 hours back trajectories on October 5.The red lines are 48 hours back trajectories on October 6.The purple lines are 48 hours back trajectories on October 7.