Seasonal Variation of Mass Absorption Efficiency of Elemental Carbon in the Four Major Emission Areas in China

As an important site-specific optical parameter widely used in climate models, the mass absorption efficiency (MAE) of elemental carbon (EC), varies dramatically with the source types and governs the direct radiative forcing (DRF) estimation. In this study, the MAE of EC for ambient samples collected from four major emission areas in China, i.e., Beijing-Tianjin-Hebei area (BTH), Yangtze River Delta area (YRD), Sichuan Basin area (SB), and Pearl River Delta area (PRD), as well as emissions from burning of residential honeycomb briquette, firewood and rice straw were investigated by using a filter-based method. The annual mean MAEEC over the four major emission areas is 7.51 m/g. MAEEC in BTH and YRD during summer appears significantly higher than MAEEC in other seasons, while seasonal variations of MAEEC in SB and PRD suggest MAEEC in summer and autumn is higher than that in winter and spring. MAEEC for samples from fossil fuels burning and biomass open-burning is 2.10 times higher than that from residential biofuel burning, which could be one of the reasons for the higher MAEEC values during the seasons heavily affected by fossil fuels burning and biomass open-burning (i.e., summer and autumn) than winter for the four locations. Difference between the measured and AeroCom median value of MAEEC may cause underestimation of DRFEC over the studied area by a factor of 0.13. Spatial and temporal variations of MAEEC would also result in underestimations of DRFEC to different degrees varying with seasons and areas.


INTRODUCTION
Light-absorbing carbon (LAC) is getting much more attentions from researchers and policy-makers for its considerable effect on global and regional climate change (Chung and Seinfeld, 2005;Ramanathan and Carmichael, 2008), air quality and human health (Jansen et al., 2005;Suglia et al., 2008), and its reduction has been targeted for near-term global warming mitigation (Shindell et al., 2012).Black carbon (BC) defined optically is a useful qualitative description when referring to light absorbing carbon in atmospheric aerosols (Petzold et al., 2013).Elemental carbon (EC) derived from thermal-optical methods is also recommended to be used in air quality and source apportionment studies by atmospheric chemists (Bond et al., 2004;Bond and Bergstrom, 2006).BC and EC will be used interchangeably in the paper.
Mass absorption efficiency (MAE) is a widely used optical parameter of aerosol particles, which is usually calculated from the light absorption coefficient divided by the mass concentration.MAE is a key factor to derive radiative forcing of aerosol particles in climate models, and its values range from 2.3 to 18 m 2 /g (Koch et al., 2009;Jacobson, 2012).Quinn and Bates (2005) conducted a series of intensive experiments with identical sampling and analysis protocols and gave values of MAE of elemental carbon (MAE EC ) ranging from 6 to 20 m 2 /g.China and India are the two largest anthropogenic aerosol emitting countries in the world (Lu et al., 2011).Recently, a few investigations on MAE EC have been conducted in these two areas.Reported work on the MAE EC in rural and urban sites in India suggests higher MAE EC values in urban areas (Ram and Sarin, 2009).Lan et al. (2013) reported a MAE EC with 6.5 ± 0.5 m 2 /g at 532 nm in South China.According to a study conducted in Xiamen, MAE EC in suburban areas was only one third of that in urban areas (Wang et al., 2013).In Beijing, the MAE EC measured at the wavelength of 632 nm was 8.45 ± 1.71 m 2 /g and 9.41 ± 1.92 m 2 /g during winter and summer respectively (Cheng et al., 2011).Bond et al. (2013) reviewed the modeled global mean MAE EC used in radiative forcing estimates, ranging from 4.3 to 15 m 2 /g at 550 nm, with the AeroCom (Aerosol Comparisons between Observations and Models) median around about 6.5 m 2 /g.The AeroCom project documents differences of aerosol component modules of sixteen global models and assembles data sets for aerosol model evaluation (Koffi et al., 2012).Based on the joint study of AeroCom models, modelers would explore the uncertainties in the direct radiative forcing estimates.Taking MAE EC changes due to internal mixing into consideration, forcing would increase by about 50% (Bond et al., 2013).
It was estimated that 1957 Gg of BC were emitted in China in 2007 with high emission densities in the North China Plain, Northeast Plain, Shanxi Highland, Henan Mountains, Sichuan Basin-Guizhou Plateau and almost all large cities, where population, vehicles, and industries are concentrated, as shown in Fig. 1 (Wang et al., 2012).Beijing-Tianjin-Hebei area (BTH), Yangtze River Delta area (YRD), Sichuan Basin area (SB), and Pearl River Delta area (PRD) are the four major BC emission areas in China since the population, vehicles, and industries are highly concentrated, and a large amount of carbonaceous fuels are consumed in these areas, which have significant impact on the regional climate and air quality (Table S1).As the top four developed areas, BTH, YRD, SB, and PRD account for 10.44%, 10.06%, 7.76%, and 3.19% BC emissions, respectively (Cao et al., 2006).However, there are fewer studies on the regionally dependent scale factor, MAE EC , in these four areas.
Therefore, the main objective of this study is to investigate the seasonal variation of MAE of ambient EC, and evaluate the potential impact of MAE EC seasonal variations on direct radiative forcing estimation.Given that elemental carbon is formed in incomplete combustion of carbonaceous matter, residential coal burning, crop residue and firewood are the top three residential source of ambient black carbon, accounting for 27.5%, 12.7% and 9.5% of total BC emission in China respectively (Wang et al., 2012).MAE EC from burning of honeycomb briquette, firewood and rice straw are also quantified to evaluate the effects of source emission on MAE seasonal variation of ambient elemental carbon.

Ambient Samples
A new approach of sampling sites selection was applied in this work.Four satellite cities including Wuqing (WQ), Haining (HN), Deyang (DY) and Zhongshan (ZS) were selected to represent the BTH, YRD, SB and PRD, respectively (Fig. 1).Satellite city sites, other than downtown sites, were chosen to investigate the MAE EC in the urban agglomerations because: a) the chosen sites locate on the diffusion path of EC from the corresponding megacities (cities marked with black triangular symbols in Fig. 1: Beijing, Tianjin, Shanghai, Guangzhou, Shenzhen, Chengdu, Chongqing et al.); b) the sampling sites are not too close to the central cities (approximate 100 kilometers) so as to ensure atmospheric aging process of EC.Therefore, PM 2.5 collected in the chosen sites could be considered as containing aged EC from corresponding megacities while avoiding the fresh EC from the concentrated EC sources like coal burning power plants and vehicles.Ambient PM 2.5 samples were collected on the roofs of the buildings with the height of about 10-30 meters in November 2012, January, April and July 2013 as shown in Table 1.All of the ambient PM 2.5 samples were collected on quartz fiber filters (Ф90 mm, Millipore, Billerica, Ireland) by employing a mediumvolume air sampler (TH-150C III, Tianhong, Wuhan, China) and a PM 2.5 impactor (TH-PM 2.5 -100, Tianhong, Wuhan, China).Each sample was continuously collected for no less than 22 hours at the operating flow rate of 100 L/min.

Fuel Source Samples
In order to investigate the MAE EC from the carbonaceous fuel sources, three types of solid fuels that are commonly used for residential heating or cooking including honeycomb briquette, firewood and rice straw were combusted in a residential cooking and heating stove in the laboratory.A self-designed sampling set-up equipped with a PM 2.5 impactor (MiniVol TM TAS, AIR METRICS, OR, USA) and a portable low-volume PM 2.5 sampler (MiniVol TM TAS, AIR METRICS, OR, USA) was used to collect the fine particles emitted from combustion of carbonaceous fuels.As shown in Fig. 2, hot flue gas emitted from burning of carbonaceous fuels diffused into the mixing chamber through a few meters of aluminum foil pipes.Warm air masses were then cooled down to fulfill the operating requirement of PM 2.5 impactor and sampler.The operating flow rate of MiniVol TM TAS was 5 L/min and the PM 2.5 samples were collected on quartz filters (Ф47 mm, Whatman, General Electric Company, USA).The PM 2.5 sampling periods for burning of honeycomb, firewood and rice straw briquette last 2-10 hours, 5-35 min and 1-7 min, respectively.Honeycomb briquette used in this study are made of coal powder and clay, and formed by pressing the mixture into a mold.This kind of coal is widely used for cooking and/or heating in rural areas.All the briquette samples were dried in the air before burning.PM 2.5 samples were collected during flaming combustion of the honeycomb briquette while the air inlet on bottom of the stove was completely open.Firewood was collected from forest in southwest China and Xiamen suburbs, consisting of various trees and bushes.Firewood was cut into small pieces (about 10-20 cm in length and 3-8 cm in diameter) before burning.Both honeycomb briquette and firewood were ignited with a band of hay or wood chips and inserted into the combustion chamber.Newly produced rice straw was collected from the farmland and naturally dried.During every combustion period, rice straw was inserted into the burning chamber in batches.All the solid fuels were combusted without any forced air-blast.
Before each PM 2.5 sampling event, the carbonaceous fuels were combusted for 5-10 minutes, so that the aluminum foil pipes and mixing chambers were filled with the flue  gas emitted from burning.Then, the fuel inlet was closed to avoid intake of background air.So the collected PM 2.5 could be treated ideally as being freshly emitted from burning of carbonaceous fuels.

Determination of Mass Absorption Efficiency of Elemental Carbon Calculation of Mass Absorption Efficiency of Elemental Carbon
When measuring MAE EC , we assume that: (a) EC is the only light-absorbing component in PM 2.5 ; (b) all EC particles have diameter less than 2.5 µm.MAE EC for every sample is calculated from the measured optical attenuation (ATN, unitless) and EC loading (L EC , µg/cm 2 ) based on the following Eq.( 1) (Liousse et al., 1993): where C and R(ATN) are two empirical factors that are often used to correct the artifacts in filter based measurement due to the multiple scattering and shadowing effects.2.14 was used for C (Weingartner et al., 2003) in this work and R(ATN) was determined using Eq. ( 2): where R(ATN) depends on parameter f.In this work, f = 1.103 for ATN > 10% and the samples were collected during wintertime (December-March) and from carbonaceous fuels exhaust.For the other three seasons, f was set to =1.114 to calculate R(ATN) (Sandradewi et al., 2008).Statistics of MAE EC were then obtained using MAE EC for each data point.Details of the determination of MAE EC can be found in our previous work (Wang et al., 2013).

Measurements of ATN and L EC
The collected samples were analyzed for L EC with a Semi-Continuous OC/EC Carbon Aerosol Analyzer (Sunset Laboratory Inc.,) following the NIOSH TOT protocol and measured for ATN at 880nm with a LED transmissometer.The LED transmissometer consists of a LED module (Model SL2420, 880nm, Advanced Illumination, Rochester, UK) and the photon detector (Model DET36A, Thorlabs, New Jersey, USA).A blank filter was placed between the LED source and the photon detector, with voltage output being recorded by a laptop connected with the photo detector.Subsequently, the same operation was repeated with a particle-loaded filter.ATN of the particle layer was then calculated according to Eq. ( 3): ATN = -100 × ln(I/I 0 ) (3) where I 0 , I were the voltage outputs through the blank filter and the deposited filter respectively.
To avoid the effect of relative humidity on ATN measurement, all the sampling filters were balanced in a desiccator for over 24 hours.
Factors contributing to the uncertainty of MAE EC measurement were investigated in our previous work (Wang et al., 2013).The results showed that measurement uncertainty of MAE EC using an 880 nm LED transmissometer and a Semi-Continuous OC/EC Carbon Aerosol Analyzer (Sunset Laboratory Inc.) was 6.9%.

Seasonal Variations of MAE EC
The results of MAE EC for ambient samples over the four major emission areas revealed a remarkable spatial and temporal variability.The lowest value of 5.18 ± 0.38 m 2 /g was found in Sichuan Basin during spring, and the highest value of 9.67 ± 0.70 m 2 /g was found in Yangtze River Delta area in summer.Annual mean values of MAE EC in BTH, YRD, SB, and PRD are 6.80 ± 1.13, 8.11 ± 1.87, 7.37 ± 1.93, and 7.77 ± 1.51 m 2 /g, respectively.Seasonal patterns of MAE EC over the four areas can be classified into two types: MAE EC in summer is significantly higher than MAE EC in other times (BTH and YRD, Fig. 3), and MAE EC in summer and autumn higher than MAE EC in winter and spring (SB and PRD, Fig. 4).

Beijing-Tianjin-Hebei area and Yangtze River Delta
Fig. 3 shows the seasonal variations of MAE EC in BTH and YRD.The highest MAE EC in BTH was found in summer with the value of 8.48 ± 1.25 m 2 /g, significantly higher (p < 0.05) than MAE EC in autumn (6.36 ± 1.82 m 2 /g), winter (5.99 ± 1.08 m 2 /g) and spring (6.37 ± 1.39 m 2 /g).
The lowest and highest MAE EC values in YRD were found in autumn and summer respectively, with MAE EC of 5.82 ± 0.47 m 2 /g and 9.67 ± 0.70 m 2 /g.MAE EC in winter was another high value of the annual cycle, 9.60 m 2 /g, but the high value of standard deviation (± 3.60 m 2 /g) indicated that MAE EC in YRD during winter is highly variable.

Sichuan Basin and Pearl River Delta
Seasonal mean MAE EC values in SB and PRD fluctuate  following a similar trend, with the highest MAE EC found during autumn, 9.46 ± 1.61 m 2 /g and 9.22 ± 1.27 m 2 /g in SB and PRD, and significant lower MAE EC during winter and spring.

Source-Specific MAE EC and Seasonal Variation of MAE of Ambient EC
It has been reported that MAE EC varies with location and time.Sharma et al. (2002) revealed a wide range of median MAE EC from 6.4 to 20.1 m 2 /g in different regions of Canada and suggested that the variability would be connected to the sources distribution and processes contributing to the carbonaceous aerosols at the sites.The large variability discovered in India was attributed to the predominance of aerosol species derived from biomass burning emissions (Ram and Sarin, 2009).In our study, the seasonal mean values of MAE EC over the four major BC emission areas are 6.48 (spring), 8.85 (summer), 7.72 (autumn) and 7.00 (winter) m 2 /g.It is also interesting to look into reasons of the seasonal variations in China with respect to various BC sources.

MAE EC for Source Samples from Various Carbonaceous Fuels Burning
Three types of carbonaceous fuel were also investigated in our work to look into the source-specific light absorption property of black carbon aerosols.MAE EC for source samples from burning of honeycomb briquette and rice straw burning were 11.33 ± 2.57 and 3.95 ± 0.19 m 2 /g, respectively.No linear relationship was found between ATN and L EC for samples collected from firewood burning emissions (Fig. S2), indicating that MAE EC measured using Eq. ( 1) is not valid.The reason why ATN was not positively linear correlated with L EC was discussed briefly in the supplement information.
We reviewed the reported works on MAE EC for variable carbonaceous fuels burning and calculated the modified MAE EC in 880 nm with Eq. ( 4): where λ is the wavelength used in ATN determination, Å is the Ångstrom exponent.The theoretical value of Å is 1.0 for particles that are small relative to the wavelength and that have a constant refractive index (van de Hust, 1957).Kirchstetter et al. (2004) revealed the spectral dependence of light absorption by aerosols, and recommended Ångstrom exponents of 2 and 1 for biomass smoke aerosols and motor vehicle aerosols based on their measurements.While light absorption of particles emitted from hard coal briquettes and lignite is low and has a strong spectral dependence, absorption efficiency of fresh aerosols emitted from bituminous coal combustion is high and the Ångstrom exponent approaches the theoretical value of 1.0 (Bond et al., 2002).Ångstrom exponents of 2.0 and 1.0 were used to modify the reported MAE EC derived from biomass burning (wood, crop residues, biomass, forest fires, crop straw, pellet and firewood) and fossil fuel burning (diesel exhaust, coal and honeycomb briquette), respectively, to MAE EC at 880 nm.

Impact of Source-Specific Elemental Carbon on Seasonal Variation of MAE EC
Modified MAE EC in 880 nm for source samples varies from 1.31 to 11.33 m 2 /g, as shown in Table 2. MAE EC values for fossil fuel burning (diesel, coal, honeycomb briquette) and biomass open burning (forest fire and open burning of agricultural waste) show 2.10 times higher compared with the values for residential bio-fuel burning, with 6.55 m 2 /g for fossil fuels burning and biomass open-burning and 3.12 m 2 /g for residential biofuel burning.In a given site or area, MAE of ambient EC supposed to be determined by EC from various carbonaceous fuels burning with high or low MAE EC .According to a monthly EC emission inventory in China from 1996 to 2010, the average ratios of EC from fossil fuel burning and biomass open-burning to EC from biofuel burning were 2.42, 2.99, 2.58 and 1.23 in spring, summer, autumn and winter, respectively (Lu et al., personal communication), suggesting that ambient EC aerosols from fossil fuel burning and biomass open-burning account for a bigger proportion of total EC than EC from biofuel burning in summer and autumn.This provides an evidence to support the effect of source-specific EC on MAE EC seasonal variations.
The emission inventory cited here was based on data throughout China.To further discuss the source-emission effects on EC optical property, source appointment investigations of EC in specific areas during different seasons are badly needed.It should be noted that other factors such as air pollutants and local meteorological conditions would also have impact on the ambient aerosol particles' MAE EC values and its temporal variations.The discussion above just provides an aspect of factors influencing the measured MAE EC .Comprehensive investigation of the mechanisms of MAE EC temporal and spatial variation is badly needed but out of the scope of this paper.

Implications on Direct radiative Forcing Estimation due to Variation of MAE EC
Direct radiative forcing of EC (DRF EC , W/m 2 ) can be expressed as the product of three factors (Schulz et al., 2006;Bond et al., 2013): where [EC] is the atmospheric burden of EC (product of global mean emission rate and lifetime of EC, g) governed by EC emissions and removal, and AFE is the absorption forcing efficiency (forcing per aerosol absorption optical depth, W/m 2 /AAOD).Values of both [EC] and AFE can be estimated based on appropriate models.Based on our measurements, annual mean MAE EC over four areas is 7.51 ± 0.56 m 2 /g, higher than MAE EC value of 6.5 m 2 /g, the AeroCom median value reviewed by Bond et al. (2013).Therefore, DRF EC in China may be underestimated by a factor of 0.13 because of the difference between measured and modeled MAE EC .Taking the seasonal variations of MAE EC into consideration, DRF EC in summer would be underestimated by a factor as high as 0.27.Similarly, DRF EC of the four areas would be underestimated to various degrees when modeling DRF EC using the AeroCom median MAE value of 6.5 m 2 /g.
As a governing factor in DRF EC estimation, variations of MAE EC in space and time can indeed cause spatial and temporal variations of DRF EC .Our work is limited to the four major EC emission areas in China, so more observation sites in east Asia should be considered in the future work to fulfill the requirements in regional DRF EC estimation.

CONCLUSION
Field campaigns for measuring mass absorption efficiency of elemental carbon in BTH, YRD, SB area and PRD were conducted from Oct. 2012 to Jul. 2013.The mean MAE EC measured over the four major emission areas during the four seasons is 7.51 m 2 /g.Seasonal patterns of MAE EC in BTH and YRD shows that MAE EC in summer is significantly higher than MAE EC in other times, while seasonal variations of MAE EC in SB and PRD suggests MAE EC in summer and autumn is higher than that in winter and spring.
The seasonal variations of MAE EC could be affected by the different EC emission patterns since light absorption of elemental carbon from various sources varies.Modified MAE EC for samples from fossil fuels burning and biomass open-burning show 2.10 times higher than that from residential bio-fuel burning, which might result in higher in MAE EC during the seasons heavily affected by fossil fuels burning and biomass open-burning (summer and autumn).
MAE EC governs the estimation of direct radiative forcing of EC.The AeroCom median value, 6.5 m 2 /g, used to estimate DRF EC could cause over-or under-estimations without considering the seasonal and spatial variations of MAE EC .Taking the region-and season-specific values of MAE EC into consideration may be a good way to obtain DRF EC with better accuracy by climate models.

A2. Determination of elemental carbon loading
When measuring L EC with a Semi-Continuous OC/EC Carbon Aerosol Analyzer (Sunset Laboratory Inc.), a piece of filter sample of 2.0 cm 2 is punched off to analyze for EC loading as mass per unit area following the temperature protocol listed in

Fig. 1 .
Fig. 1.Description of four sampling sites.BC emission map of China cited from Wang et al. (2012) was used to describe the sampling sites' location.

Fig. 2 .
Fig. 2. Schematic diagram of the self-designed sampling set-up (keep the fuel inlet close during every sampling period).

Fig. 3 .
Fig. 3. Seasonal variations of MAE EC in BTH and YRD.Non-linear relationship between ATN and L EC in YRD during winter and summer (Fig. S2) may cause great uncertainty in determination of MAE EC.

Fig. 4 .
Fig. 4. Seasonal variations of MAE EC in SB and PRD.

Table 1 .
Details of the sampling sites.BTH: Beijing-Tianjin-Hebei area; YRD: Yangtze River Delta area; SB: Sichuan Basin; PRD: Pearl River Delta area.b Only 2 valid samples were collected in spring in Deyang because of the interruption by the earthquake on 20th, April 2013. a

Table 2 .
Comparison of mass absorption efficiency of elemental carbon for various source samples.
Shen et al. (2013)n the filters were determined using DRI thermal/optical carbon analyzer following IMPROVE temperature protocol which is different from the Sunset EC/OC analyzer used inShen et al. (2013)'s work and this study.

Table S2 .
The protocol listed here is the default one recommended by the manufacture.