Decreasing concentrations of carbonaceous aerosols in China from 2003 to 2013

Carbonaceous aerosols were characterized in 19 Chinese cities during winter and summer of 2013. Measurements of organic carbon (OC) and elemental carbon (EC) levels were compared with those from 14 corresponding cities sampled in 2003 to evaluate effects of emission changes over a decade. Average winter and summer OC and EC decreased by 32% and 17%, respectively, from 2003 to 2013, corresponding to nationwide emission control policies implemented since 2006. The extent of carbon reduction varied by season and by location. Larger reductions were found for secondary organic carbon (SOC, 49%) than primary organic carbon (POC, 25%). PM2.5 mass and total carbon concentrations were three to four times higher during winter than summer especially in the northern cities that use coal combustion for heating.


Method
Sampling sites and descriptions. Nineteen cities (Supplemental Table S1) including five megacities (i.e., Beijing [BJ], Tianjin [TJ], Shanghai [SH], Chengdu [CD], and Chongqing [CQ]) were selected to represent economically developed and developing urban regions, as shown in Fig. 1. The sampling network is documented in Table 1. In addition to those from the 2003 study 24 , three northern cities were added that represent devel- www.nature.com/scientificreports/ oping industries (Taiyuan, TY), high altitudes (Xinning, XN), and arid western (Urumqi, UR) environments, along with two southern cities that represent developing urban environments (Chengdu, CD and Nanjing, NJ). Sampling sites represent urban-scale exposures, with most located at university campuses or research centers (> 100 m from local sources such as major roadways).
Sample collection. Integrated daily, 24 h PM 2.5 sampling (0900 to 0900 local standard time) was conducted during winter (5-26 January) and summer (1-31 July) of 2013 (Table S2). PM 2.5 samples were collected on 836 prefired (900 °C for 3 h) 47 mm Whatman QMA quartz-fiber filters, using mini-vol air samplers (Airmetrics, Eugene, OR, USA) at a flow rate of 5 L min −1 . These samplers were located on rooftops at varying heights (~ 6-20 m) above the ground level (Table 1). Quartz-fiber filters were analyzed gravimetrically for mass concentrations 25 using a Sartorius MC5 electronic microbalance with a ± 1 μg sensitivity (Sartorius, Gottingen, Germany). To minimize particle volatilization and aerosol liquid water biases, these filters were weighed after 24-h equilibration at a constant (within ± 2 °C) temperature between 20 and 23 °C and within ± 5% relative humidity (RH) between 30 and 40% following U.S. EPA guidelines 26 . Nominal values of 20 °C and 30% RH best conserve particle deposits during sample weighing 26 . Each filter was weighed at least twice before and after sampling to www.nature.com/scientificreports/ ensure reproducibility. The differences between the replicated weights were < 10 μg per filter for laboratory blanks and < 20 μg for exposed samples.  33 propose the use of the minimum R squared (MRS) method, as the MRS method presents a quantitative criterion to determine the primary OC and EC ratio ((OC/EC) pri ), which minimizes the uncertainties in estimating SOC. In comparing five linear regression techniques, Wu and Yu 34 further recommended the use of Deming regression, orthogonal distance regression, and York regression to estimate SOC.
As formation of SOC, often referred to secondary organic aerosol (SOA), involves complicated atmospheric processes of photochemical oxidation, gas/particle partitioning and nucleation/condensation, it presents a large challenge to accurately estimate SOC. There are pros and cons of applying different methods to estimate SOC with uncertainties associated with each. The underlying principle of the tracer approach 31 is based on the assumption of a fixed relationship between primary OC and EC, while in reality, the OC and EC ratios and background concentrations vary over time. To ensure adequate validity of the (OC/EC) pri , data from extreme weather conditions (e.g., rain or storms) and from combustion dominated environment (with high OC/EC ratios) are excluded, and OC/EC measurements with the lowest potential in photochemical production (i.e., lowest 20% of OC/EC ratios) are retained to estimate (OC/EC) pri 24,35,36 . The EC tracer approach used in this study intends to maintain consistency in SOC comparison with the 2003 study. Table S3 summarizes linear regressions between OC and EC, with similar findings for both years. During winter, good OC/EC correlations (R 2 : 0.66-0.95) were found for all but two northern industrial cities (i.e., Changchun and Xi'an), consistent with contributions from a mixture of sources (e.g., residential and commercial coal combustion, and motor vehicle exhaust). Correlations between OC and EC (R 2 : 0.16-0.92) were variable in summer, implying different source mixtures.
The OC/EC slopes were 2.34 and 2.08 in winter for the northern and southern cities, respectively, similar to those of 2.81 and 2.10 in 2003. As shown in Table S4, the summer OC/EC ratios of 0.97 and 0.51 during 2013 in the northern and southern cities are lower than the corresponding 2003 values of 1.99 and 1.29. Table S5 summarizes
In southern China, average wintertime POC concentrations were higher (19.6 ± 8.3 µg m −3 , ~ 63% of TC) than those in summer (4.6 ± 1.5 µg m −3 , ~ 40% of TC), with ~ 86% of OC as POC and ~ 14% as SOC. Minimal differences were found for average SOC between winter (3.3 ± 6.4 µg m −3 , ~ 11% of TC) and summer (3.1 ± 3.2 µg m −3 , ~ 27% of TC).   (Table 3). POC reductions were more pronounced in summer (~ 61% in the north and ~ 26% in the south) than the corresponding ~ 28% and ~ 22% in winter. Summer SOC decreased by ~ 44% and ~ 58% for northern and southern China, respectively. Wintertime SOC experienced a ~ 73% reduction in southern China, but a ~ 10% increase was observed for northern China from 2003 to 2013, possibly due to less stringent VOC and NO x controls for small boilers 14,19 . Past studies found that VOCs and NO x emissions influence winter air quality in China 32,38 . Average 14-city carbonaceous aerosols constituted ~ 38% and 42% of PM 2.5 in 2003 and 2013, respectively. From 2013 to 2003, SOC and POC decreased by ~ 49% and 25%, respectively, whereas EC decreased by ~ 17%. Figure 3 illustrates nationwide pollution reduction, with most of the average 2013/2003 ratios less than unity for PM 2.5 mass, OC, and EC. Concentration reductions are most apparent in the three northern (i.e., Jinchang, Xi'an, Yulin) and five southern (i.e., Chongqing, Guangzhou, Hangzhou, Shanghai, and Wuhan) cities. However, increased wintertime OC concentrations were found in Beijing and Changchun, whereas increased EC levels were found during winter in Tianjin and during summer in Qingdao and Xiamen. Hong Kong showed increases in OC and EC from 2003 to 2013 for both seasons. Figure 1 shows that wintertime carbon decreased for most cities since 2003 24 . The most apparent wintertime OC reductions (~ 49-62%) were found for Shanghai (~ 62%), Guangzhou (~ 53%), and Chongqing (~ 49%); EC also decreased (34-58%) with ~ 34% for Chongqing, ~ 40% for Shanghai, and ~ 58% for Guangzhou. Reduction of wintertime carbonaceous aerosols for the remaining cities ranged from 13 to 37% for OC and 2-49% for EC. Shanghai, Guangzhou, and Chongqing represent the three most developed regions in southern China: the Yangtze River Delta (YRD) region, the Pearl River Delta (PRD) region, and the Chong-Yu (CY) region, respectively. Control measures such as industrial plant closures and traffic controls during the 2010 (May to October) World Exposition in Shanghai and the 2010 (November) Asian Games in Guangzhou demonstrated the effectiveness of regional air quality management strategies that were implemented on larger scales.

Changes of OC and EC concentrations between 2013 and 2003.
Increases in wintertime OC (~ 46%) and EC (~ 18%) in Beijing and the OC increase (~ 26%) in Changchun may be associated with severe haze episodes under sluggish weather systems that occurred over central and eastern China in 2013. During the two January episodes (9th-15th and 25th-31st), maximum hourly PM 2.5 mass    Table 1 Figure 4 shows variations in OC/EC ratios as a function of EC concentrations, ranging 1.2-11.6 during winter and 0.01-6.9 during summer of 2013. Similar variability was found for 2003 (red circles in Fig. 4) with lower OC/EC ratios in winter for concentrations < 20 µg m −3 . Changes in PM carbon properties may influence the relative amounts of particle light scattering and absorption 24 related to visibility impairment and climate change.
Higher winter OC/EC ratios (6.0 to 11.6) during 2013 (barely seen in 2003) were found in the northwestern cities of Urumqi, Yulin, Xining, and Xi'an. This may be due to changes in the source mixtures. Abundances of the thermal carbon fractions also changed. In wintertime Xi'an, high temperature carbon fractions of OC3 and OC4 (480 °C and 580 °C) were most abundant in 2013, whereas lower temperature OC2 (280 °C) and OC3 fractions dominated in 2003. These changes may reflect changes in the composition of organic compounds in the combustion emissions.
Ten northern cities (with the exception of Taiyuan) reported higher wintertime OC/EC ratios than southern cities. Taiyuan reported the lowest winter OC/EC ratio of 2.4, close to the ratio of 2.0 for primary fossil fuel combustion from the national emission inventory 43 . This is reasonable considering that Taiyuan is China's coal capital, with coal energy, metallurgy, chemicals, and machinery industries 44 .
While the distribution pattern of OC/EC ratios during summer in 2013 was similar to that of 2003, the average ratio is 23% lower. This is attributed to the decrease in OC from 13.8 μg . m −3 in 2003 to 7.1 μg . m −3 in 2013 as the average EC concentrations (3.4-3.6 μg . m −3 ) remained similar. The summer OC/EC ratios were the highest in Beijing (3.0-4.8), followed by Guangzhou (1.9-6.9), with the lowest in Hong Kong (0.6-0.8). There are different vehicle fuel formulations among the cities. The higher proportion (~ 38%) of diesel vehicle emissions in Hong Kong resulted in lower OC/EC ratios. This is consistent with the average OC/EC ratio of 2.7 for gasoline and LPG vehicles and 0.5 for diesel vehicles 45 .

Conclusions
Decreasing OC and EC concentrations were apparent at a national scale from 2003 to 2013; but the extent of reduction varied by season and location. Most inland cities showed decreases in carbon concentrations with the exception of Beijing and Changchun (capital of Jilin in northern China). Shanghai showed an apparent decrease, with a less distinct reduction in Qingdao and Xiamen for OC and slight increases for EC. The exception was found in Hong Kong, with a 2003 to 2013 increase of ~ 29% for OC and ~ 122% for EC.
Carbonaceous aerosols constituted ~ 37.5% and ~ 41.5% of PM 2.5 mass for 2013 and 2003, respectively. Majority of the TC is in OC, with average OC/EC ratios of 3.1 and 4.0 for 2013 and 2003, respectively. Summertime PM 2.5 mass and OC concentrations were ~ 25-35% of those in winter with less variability in EC. Wintertime carbon concentrations were higher in northern than southern China, elevated in inland cities and low in coastal and non-urban desert cities. Winter in northern China is the most polluted season with ~ 64% primary organic carbon (POC) and ~ 36% of secondary organic carbon (SOC) in OC.
Wintertime OC/EC ratios were higher in most northern than southern cities, indicating the impact of haze episodes and fossil fuel combustion from heating. Elevated 2013 OC/EC ratios (6.0 to 11.6) in northwestern cities such as Urumqi, Yulin, Xining, and Xi'an, were barely seen in 2003. Changes in abundances among eight