|Home | About | Journals | Submit | Contact Us | Français|
Size fractionated particulate matter (PM) samples (including PM2.5 and PM10) were collected at Peking University in Northwestern Beijing, China for a 2 week period prior to the Olympics, during the 2 week period of the Olympics, and for a 4 week period following the 2008 Olympics, during both source control and non-source control period. PM10 concentrations in this study were high correlated with, but a factor of 1.3 times higher than, the Beijing Environmental Protection Bureau's PM10 concentrations at near-by sites because of differences in the measurement methods used. The mean PM2.5 and PM10 concentrations were statistically different, and lower by 31 and 35%, during the Olympic period compared to the non-Olympic period. However, the PM concentrations were not statistically different between the source control and non-source control periods. While meteorological parameters (air masses from the south and precipitation) accounted for 40% of the total variation in PM10 concentration, source control accounted for 16%, suggesting that meteorology accounted for more of the variation in PM concentration than source control measures. The PM10 concentrations in Beijing during the Olympic period were 2.9, 3.5, and 1.9 times higher than those in Atlanta, Sydney and Athens. In addition, the PM2.5 and PM10 concentrations during the Olympic period exceeded the WHO 24-hour guideline 100% and 81% of the time, respectively. Finally, the PM10 concentrations in October, November, and December 2008 were reduced by 9% to 27% compared to the same months in 2007, suggesting that the Olympic source control efforts (and possibly a down turn in the economy) have resulted in lower PM10 concentrations in Beijing.
Atmospheric particulate matter (PM) is composed of a mixture of various sized particles that have potentially toxic chemicals sorbed to them. These particles are mainly distributed in two fractions: fine particles (aerodynamic diameters <2.5μm) and coarse particles (aerodynamic diameters >2.5μm) (1,2). Fine particles are a concern for human lungs because they can be deposited more deeply in the lungs than coarse particles (2,3). Several studies have also shown that fine particles are linked to increased respiratory morbidity and mortality (2,3). Recently, Kan et al. (4) found that, in Shanghai, fine particles (PM2.5) were associated with increased death rates from all causes, including cardiorespiratory diseases. In addition, PM-polluted air is believed to be responsible for about one million premature deaths per year in China (5).
Beijing (39°56′N and 116°20′E) is the capital of China with a population of over 17 million people and an area of 16,800 square kilometers. Beijing is surrounded by the Yanshan Mountains in the west, north and northeast which effectively trap air pollutants. In addition, the rapid urbanization and motorization of Beijing (1.39 million motor vehicles in 1999 compared to 3.3 million motor vehicles in 2007) has contributed to the high PM pollution in Beijing (2, 6-11). Dust storm events in the spring, and re-suspended dust from traffic and construction activities, further exacerbate the PM pollution in Beijing (2).
Beijing hosted the 2008 Summer Games of the 29th Olympiad from August 8-24, 2008. In view of the significant air pollution in Beijing, the local governments cooperated to improve air quality during the 2008 Olympic Games. During the Olympic and the Paralympic Games, from July 20 to September 20, 2008, source control measures in Beijing included removing approximately one-half of the cars (~ 1.5 million cars) off the road on alternate days under an even-odd license plate system, closing pollution emitting factories, and slowing down construction activities. In the neighboring Tianjin municipality, Hebei, Shanxi and Shandong provinces, and Inner Mongolia Autonomous Region, polluting factories were closed and high-emission cars were removed from roads (12). In addition, in Beijing, stricter vehicle emission standards are being implemented, businesses are switching from coal power to natural gas, and heavy-duty truck traffic has gradually been reduced (12).
Given the significant international interest in improving the air quality of Beijing during the 2008 Olympic Games, some studies have attempted to predict the air quality during this period based on modeling (13,14). Using the US Environmental Protection Agency's Models-3/Community Multi-scale Air Quality (CMAQ) model over the Beijing region, Streets et al. (14) predicted that about 34% of PM2.5 (on average) at the Olympic Stadium site can be attributed to sources outside of Beijing. The UNEP recently released an environmental assessment report of the Beijing 2008 Olympics games, including an assessment of air quality, which suggests that PM10 concentration was reduced by 20% during the source control periods (15). Although not at ground level, the interpretation of satellite-retrieved aerosol optical thickness for the Beijing region during the Olympic period suggests that the magnitude of aerosol load reduction during this period was low (10 to 15%) compared to meteorological variability (16).
The objectives of this study were to: (1) determine Beijing's atmospheric PM pollution during the Olympic and non-Olympic periods, as well as source control and non-source control periods; (2) assess the influence of meteorology on PM concentrations; and (3) determine the source regions that influenced Beijing's PM concentrations and the effectiveness of source control strategies in reducing PM pollution.
The sampling site was located on the roof of the 7-story Geology Building on the Peking University (PKU) campus, about 25 meters above ground. PKU is located in northwestern Beijing and is primarily a residential and commercial area without major industrial sources (Figure SI.1). Local PM emission sources within 1 km of PKU include vehicular traffic, fuel combustion for domestic cooking and heating, and construction. The 2008 Olympic table tennis, marathon, and bicycle races were held on or near the PKU campus.
PM was collected and size fractionated using a three stage High Volume Cascade Impactor (Series 230, Tisch Environmental, Cleves, OH) that operated in accordance with procedures established by USEPA (CFR40, Part 50.11, Appendix B, July 1, 1975, pages 12-16) and ASTM Specification D2009. The Impactor stages consisted of two aluminum plates, with outside dimensions of 6 inches × 7 inches that held 5.625 inch × 5.375 inch filters with ten perforated slots, and a base aluminum plate that held an 8 × 10 inch filter. The High Volume Cascade Impactor was calibrated according to the manufacturer's instructions before use and once per week during the sampling periods. Airborne particles were size fractionated based on their aerodynamic particle diameter and are abbreviated as follows: <2.5μm (PM2.5) (bottom 8 inch × 10 inch filter), 2.5-10μm (PM2.5-10) (middle 5.625 inch × 5.375 inch filter), and >10μm (<PM10) (top 5.625 inch × 5.375 inch filter). We calculated the PM10 concentration by summing the PM2.5 and PM2.5-10 concentrations.
Sixty-three sets of size fractionated PM samples were collected continuously over 24 h periods (~1500 m3) from July 28 to September 3, 2008 and from September 13 to October 7, 2008. Six field blanks were also collected during these periods. Samples were not collected from September 4 to 12, 2008 because of sampler motor failure. Quartz fiber filters (Tisch Environmental, Cleves, OH), that were baked in a furnace at 350°C prior to use, were used for sample collection and were weighed before and after sample collection.
A subset of the filters (~33%) was weighed before and after sample collection, and before and after dessication for 24 hours, in accordance with USEPA Method 5 of 40 CFR Part 60 (http://www.epa.gov/ttn/emc/methods/method5.html). The percent change in filter mass due to dessication (0 to 20%) was correlated with the ambient relative humidity (40 to 95%) during the sample collection period and this relationship was used to correct the change in filter mass (see Supporting Information). Only a subset of samples was tested by dessication and the filters were not dried at 105 °C after sample collection in order to minimize analyte loss and/or sample contamination that would affect subsequent semi-volatile organic compound analysis of the filters. Detailed information on the filter mass humidity correction is given in the Supporting Information (Figure SI.2).
Meteorological data, including wind speed (WS), wind direction (WD), precipitation (Precip), atmospheric pressure (P), temperature (T), and relative humidity (RH), were simultaneously measured every minute by HOBO Auto Meteorological Station (Onset Company, Pocasset, MA) at the sampling site during all sample collection periods.
The Beijing Environmental Protection Bureau's (Beijing EPB's) Air Pollution Index (API) PM10 data was used for comparison. The Beijing EPB's PM10 measurements were made continuously using Tapered Element Oscillating Microbalances (TEOM® RP1400a) as described in USEPA Method IO-1.3 of EPA/625/R-96/010a (http://www.epa.gov/ttn/amtic/inorg.html). We compared our PKU data to the Beijing EPB's data at the following sites: the WanLiu (WL) PM10 Auto-monitoring Station (located 3 km from our PKU air monitoring site); the Olympic Center (OC) PM10 Auto-monitoring Station (located 3 km from our PKU site), the Beijing Average API data (the average from all Beijing EPB sites shown in Figure SI.1), and the average API data for all of the urban Beijing EPB sites within the 5-Ring Road of Beijing (Figure SI.1). The Beijing EPB's API PM10 values were converted to PM10 concentrations (ConWL, ConOC, ConBeijing and Con5-Ring) in μg/m3 as described in reference 17. PM2.5 concentrations were not reported by the Beijing EPB during this period so no comparison could be made to our PKU PM2.5 concentrations.
Four-day air mass back trajectories were calculated using National Oceanic and Atmospheric Administration (NOAA)'s ARL HYSPLIT 4.0 model (with meteorological data from the Global Data Assimilation System, GDAS) (18). For each 24 h sample, seven trajectories were calculated (one every 4 h), including the sample start and stop times. The trajectories were calculated at 50, 150, 250 and 350 m asl and S-PLUS version 8.0 (insightful, Seattle, WA) was used for statistical analysis.
We chose to compare our PKU PM10 concentrations to the Beijing EPB PM10 concentrations measured at the WanLiu PM10 Auto-monitoring station (ConWL) because of its close proximity to our PKU site and the Olympics Center PM10 Auto-monitoring station (ConOC) because of its proximity to many Olympic events. In addition, we compared our PKU PM10 concentrations to the Beijing average PM10 concentrations (ConBeijing) reported by the Beijing EPB and the more urban Beijing EPB sites within the 5-Ring Road of Beijing (Con5-Ring).
There was significant correlation (p<0.001) between the PKU PM2.5 and PM10 concentrations (Table SI.1). In addition, there was significant correlation (p<0.001) between our PKU PM10 concentrations and the Beijing EPB PM10 concentrations at the WanLiu and Olympic Center sites, as well as the average PM10 concentrations for Beijing and the PM10 concentrations within the 5-Ring Road (Table SI.1 and Figure SI.3). However, our PKU PM10 concentrations were a factor of 1.3 times higher than the Beijing EPB PM10 concentrations for all comparison sites (Figure SI. 3). A factor of 1.3 between gravimetric PM10 methods and TEOM methods has been previously reported and is likely due to the loss of volatile and semi-volatile components of the PM during the TEOM measurement (19-21).
The mean PKU PM2.5 and PM10 concentrations and Beijing EPB PM10 concentrations during the Olympic and non-Olympic periods are given in Table 1. In addition, Figure 1 shows the temporal variation in the PKU PM2.5 and PM10 concentrations. Our 2008 PKU PM2.5 and PM10 concentrations are comparable to, or slightly lower than, previously reported concentrations in Beijing (including previous measurements at PKU) (6-11)
The mean PKU PM2.5 and PM10 concentrations during the Olympic period were significantly different and 31% to 35% lower than the PM concentrations measured during the non-Olympic period (Table 1). In addition, the mean PKU >PM10 and PM2.5-10 concentrations were reduced by 42%, while the PM2.5 concentrations were reduced by only 31%.
Based on the PKU PM10 concentration for individual days during the Olympic period, 12% of days exceeded the U.S. National Ambient Air Quality Standard (NAAQS) (150 μg/m3) for a 24-hour period and China's Ambient Air Quality Standard (CAAQS) for cities (grade II 150 μg/m3) and 81% of days exceeded the WHO 24-hour PM10 air quality guideline (AQG) (50 μg/m3) and CAAQS Grade I (Table SI.2). Based on the Beijing average EPB PM10 concentration for individual days during the Olympic period, no days exceeded the NAAQS and CAAQS grade II and 44% of days exceeded the WHO AQG and CAAQS Grade I. However, based on the PKU PM2.5 concentration for individual days during the Olympic period, 88% of days exceed the U.S. NAAQS (35 μg/m3) and 100% of days exceeded the WHO AQG (25 ug/m3) (Table SI.2)
The mean PM2.5/PM10 ratio was not significantly different between the Olympic (0.77) and non-Olympic (0.75) periods (Table 1). Our results indicate that fine particles (PM2.5) accounted for 62 – 86% of the inhalable particles and were the major component of the PM10 concentrations. Previous studies measured a PM2.5/PM10 ratio of 0.25 to 0.88 in Beijing (2). High PM2.5/PM10 ratios (larger than 0.6) are generally attributed to NO3−, SO42− and NH4+ and secondary organic aerosols, while low ratios are attributed to soil or construction dust (2). Although Beijing's PM concentrations were significantly reduced during the Olympic period, the similarity of the PM2.5/PM10 ratio during the Olympic and non-Olympic periods suggests that combustion sources were a significant contributor to PM concentrations during both periods.
Although it is interesting to compare the PM concentrations during the Olympic and non-Olympic periods from an athlete and spectator health standpoint, a comparison of the source control and non-source control periods is more relevant from an atmospheric chemistry standpoint. The mean PKU PM2.5 and PM10 concentrations and Beijing EPB PM10 concentrations (ConWL, ConOC, and ConBeijing) during the source control and non-source control periods are given in Table 2. Figure 1 shows that some of the lowest PM concentrations were measured during the non-source control period, while some of the highest PM concentrations were measured during the source control period. This suggests that factors other than source control, such as meteorological conditions, were important in determining the day-to-day variation in Beijing's PM concentrations.
Unlike the Olympic and non-Olympic periods, only the mean PKU PM2.5-10 concentration was significantly different between the source control and non-source control periods, with a 31.2% reduction in concentration (Table 2). Similar to the Olympic and non-Olympic periods, the source control measures were less effective at reducing PM2.5 concentrations than >PM10 and PM2.5-10 concentrations.
The mean PKU PM2.5 and PM10 concentrations during the Olympic period were significantly different, and 16.9% to 26.3% lower, than during the source control periods before and after the Olympic period (Tables (Tables11 and and2).2). This suggests that additional source control measures may have been taken during the Olympic period. Overall, the mean PM concentrations were Olympic period < source control period < non-Olympic period < non-source control period.
Based on the PKU PM10 concentration for individual days during the source control period, 22% of days exceed the NAAQS and CAAQS Grade II and 88% of days exceeded the WHO AQG and CAAQS Grade I. Based on the Beijing average EPB PM10 concentration during the source control period, 7% of days exceeded the NAAQS and CAAQS grade II and 68% of days exceeded the WHO AQG and CAAQS Grade I. Based on the PKU PM2.5 data for individual days, 93% of days exceed the NAAQS and 100% of days exceeded the WHO AQG during the source control period.
Finally, similar to the Olympic and non-Olympic periods, the PM2.5/PM10 ratio was not significantly different between the source control (0.76) and non-source control (0.75) periods (Table 2) and was similar to the Olympic and non-Olympic periods (Table 1). These results suggest that the source control measures had little effect on the proportion of PM that was highly respirable (PM2.5).
Although PM sources in Beijing clearly contribute to PM concentrations in the city, it has been shown that surrounding areas also contribute to Beijing's PM concentrations (14, 15). Four day air mass back trajectories were calculated for each air mass sampled at PKU and were used to determine the impact of different source regions on Beijing PM concentration. These source region impact factors (SRIFs) represents the percentage of time an air mass spent in a given source region (Northeast, Northwest, South and East of Beijing) in the last 4 days prior to arrival in Beijing. The higher the SRIF percentage, the more time the sampled air mass spent in a given source region. Details of the SRIF calculation are reported elsewhere (22).
The four potential source regions to Beijing, South (SRIFSouth), Northeast (SRIFNortheast), Northwest (SRIFNorthwest) and East (SRIFEast) of Beijing, and representative 4-day air mass back trajectories are shown in Figure 2. The correlations between the PKU PM2.5 and PM10 concentrations and the Beijing EPB PM10 concentrations (ConWL, ConOC, and ConBeijing) and air mass SRIFs were calculated to determine if changes in Beijing PM concentrations were significantly correlated with air mass time in different source regions (Table SI 3). If a source region had a significant influence on Beijing's PM concentration, then a significant correlation would exist. For the entire dataset (both source control and non-source control periods), there was a significant positive correlation between PM concentrations and SRIFSouth and a significant negative correlation between PM concentrations and SRIFNorthwest (Figure SI.4). This suggests that air masses from regions south of Beijing, including Hebei, Shandong, Shanxi and Tianjin, increased PM concentrations in Beijing, while air masses from regions northwest of Beijing, including Russia and Mongolia, decreased PM concentrations in Beijing. Other researchers have also shown that the area south of Beijing was a source for PM pollution to Beijing (14-16).
The correlation between PM concentration and meteorology (including wind speed, wind direction, temperature, precipitation, and relative humidity) was investigated (Table SI.4). The precipitation amount (on the day prior to the air sampling day), was significantly negatively correlated with PM concentration due to wet deposition (p < 0.01) (Table SI.4). The relative humidity (excluding the days with precipitation) was significantly positively correlated with PM concentration (p < 0.05) (Table SI.4) and SRIFsouth (p < 0.05) (Table SI.4). High relative humidity can promote the formation of secondary organic aerosols (7,23). In addition, all PM concentrations (excluding >PM10) were significantly positively correlated (p-value< 0.01) with wind from the south, and significantly negatively correlated (p-value<0.05) with winds from the north. This is consistent with our interpretation of SRIFs discussed above.
To assess the impact of meteorological conditions, in addition to the source control measures, on PM concentrations, a multivariable linear regression (MLR) model was developed. MLR models have been previously used to assess the effect of meteorological conditions on PM and other pollutant concentrations (22, 24, 25).
Autocorrelation was used to reduce the number of meteorological parameters for the MLR model. Because there was significant autocorrelation between SRIFSouth, relative humidity, and wind direction (Table SI.4), only SRIFSouth was used in the MLR model. In addition, to describe the varying levels of source control during the sampling period, indicator variables were used to describe the Olympic, Source Control (non-Olympic) and Non-Source Control periods:
Where CPM is the PM concentration for the entire dataset, SRIFSouth is SRIF from the south of Beijing, Precip (mm/day) is the precipitation amount for the day prior to the sampling day, Oly is indicator variable for the Olympic period (coded “1” for an “Olympic” day and “0” for all other days), SCnonOly is the indicator variable for the Source Control period outside of the Olympic period (coded “1” for a “Source Control” day outside of the Olympic period and “0” for all other days), and NSC is the indicator variable for the Non-source Control period (coded “1” for a “Non-source Control” day and “0” for all other days).
The MLR model results are shown in Table SI.5. Meteorological parameters (SRIFSouth and precipitation) accounted for 40% of the total variation in PM10 concentration, while source control accounted for only 16%. The results were similar for PM2.5 (Table SI.5). These data suggest that, at ground level, meteorology accounted for a larger portion of the variation in PM concentration than source control measures. Using an analysis of aerosol load from the Moderate Resolution Imaging Spectroradiometer (MODIS) aerosol optical thickness product for the Beijing region during the Olympic period, Cermak and Knutti (16) recently reported a statistically significant reduction in aerosol load that decreased in magnitude and significance with increasing region size. They also reported that the magnitude in aerosol load reduction was low (10 to 15%) compared to meteorological variability (16). Their findings are consistent with our findings at ground level and reflect the importance of PM emissions and transport on a regional scale in controlling Beijing's PM pollution.
Table SI. 6 summarizes the PM10 concentrations in Beijing, China, Atlanta, Georgia, Sydney, Australia, and Athens, Greece one month before, during, and after their respective Olympic Games (26-28). During the Olympic periods, the mean PM10 concentrations were 28.1 μg/m3, 23.7 μg/m3, 44.3 μg/m3, 82.4 μg/m3 and 53.7 μg/m3 in Atlanta, Sydney, Athens, PKU PM10, and Beijing EPB average, respectively. The mean PM10 concentrations in Beijing were 1.9 to 3.5 times higher than the other three Olympic cities. Of these Olympic cities, only Beijing exceeded the WHO standard for PM10. However, Beijing had the most significant decrease in PM10 concentrations (40%), compared to Atlanta (29% reduction), Sydney (72% increase), and Athens (9% decrease), during the Olympic period.
Beijing PM concentrations are dependent on seasonal weather conditions and seasonal heating (8, 11). As a result, PM concentrations are generally higher in the winter months (2). In order to evaluate the current status (post-Olympics) of PM concentrations in Beijing, and if the source control measures implemented during the Olympics were still effective, we compared the monthly mean Beijing EPB ConBeijing values for October, November and December 2007 and 2008 (Table SI.7). There was a 8.7%, 27.0% and 26.9% reduction in the mean ConBeijing concentration in October, November and December 2008, respectively, compared with the same 2007 periods (p value <0.1).
During the post-Olympic period, traffic was reduced by 20% and this, along with a down turn in the economy and energy production, appears to have resulted in a 9-27% decrease in PM10 concentrations compared to the end of 2007. However, the monthly mean ConWL, ConOC, and ConBeijing concentrations during October, November and December, 2008 were 47.9% to 65.5% higher than the PM10 concentrations during Olympic period and 30.5 to 52.7% higher than during the source control period.
The authors thank the Air Protection Branch of Georgia EPA for supplying the 1996 Atlanta, Georgia PM10 data and Susan Alber and Cliff Pereira of Oregon State University for statistical guidance. Funding for this research was provided by the China Scholarship Council (to Wentao Wang), the U.S. National Science Foundation (ATM-0239823), and National Scientific Foundation of China (40710019001). This publication was made possible in part by grant number P30ES00210 from the National Institute of Environmental Health Sciences (NIEHS), NIH and NIEHS Grant P42 ES016465. Its contents are solely the responsibility of the authors and do not necessarily represent the official view of the NIEHS, NIH.
Supporting Information Available
Details of the relative humidity correction of the quartz fiber filters, correlation analysis, a map of the Beijing sampling sites and calculated SRIFs are provided in the Supporting Information. This material is available free of charge via the Internet at http://pubs.acs.org.