Emissions of black carbon in East Asia ... - Wiley Online Library

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 116, D16201, doi:10.1029/2011JD015637, 2011

Emissions of black carbon in East Asia estimated from observations at a remote site in the East China Sea Y. Kondo,1 N. Oshima,2 M. Kajino,2 R. Mikami,3 N. Moteki,1 N. Takegawa,3 R. L. Verma,1 Y. Kajii,4 S. Kato,4 and A. Takami5 Received 13 January 2011; revised 24 April 2011; accepted 13 May 2011; published 16 August 2011.

[1] East Asia, including China, is the largest source of anthropogenic black carbon (BC). In estimating the BC emissions from this region, it is advantageous to use BC mass concentrations measured at remote locations on the ocean appropriately distant from the large sources because of spatially uniform distributions through mixing during transport. We made continuous measurements of the BC mass concentration with an accuracy of about 10% at Cape Hedo on Okinawa Island, Japan, in the East China Sea, from February 2008 to May 2009, simultaneously with carbon monoxide (CO). The seasonal median BC concentrations at Hedo were highest (0.23–0.31 mg m−3 at standard temperature and pressure) in winter and spring when plumes from China, predominantly northern China north of 33°N, were often transported to the site. A three‐dimensional chemical transport model is used to calculate the mass concentration of BC using the annual mean emission inventory of Zhang et al. (2009) for the base year 2006. The model results and the observed BC‐CO correlation are used to exclude the BC data substantially influenced by wet deposition. The calculated BC mass concentrations agree with those observed to within about 30% in air strongly affected by emissions in China for winter and spring on average. We estimate the annually averaged BC emission flux over the whole of China to be 1.92 Tg yr−1 with an uncertainty of about 40%. This value is very close to the value of 1.81 Tg yr−1 estimated by Zhang et al. (2009). The overall uncertainty of 40% of the present estimate is a substantial improvement in the uncertainty (208%) of the bottom‐up inventory. Citation: Kondo, Y., N. Oshima, M. Kajino, R. Mikami, N. Moteki, N. Takegawa, R. L. Verma, Y. Kajii, S. Kato, and A. Takami (2011), Emissions of black carbon in East Asia estimated from observations at a remote site in the East China Sea, J. Geophys. Res., 116, D16201, doi:10.1029/2011JD015637.

1. Introduction 1.1. Objectives [2] Black carbon (BC) particles are emitted by incomplete combustion of fossil fuels and biomass [e.g., Streets et al., 2003; Bond et al., 2004], alter the global radiation budget by absorbing solar visible radiation [Jacobson, 2001; Ramanathan et al., 2001, 2007; Intergovernmental Panel on Climate Change (IPCC), 2007], and, when deposited in the polar regions, change snow albedo [Hansen and Nazarenko, 2004]. Asia was the largest BC source region in 2000, according to current emission inventories [Streets et al., 2003; Ohara et al., 2007; Zhang et al., 2009]. Therefore, it is 1 Department of Earth and Planetary Science, Graduate School of Science, University of Tokyo, Tokyo, Japan. 2 Meteorological Research Institute, Tsukuba, Japan. 3 Research Center for Advanced Science and Technology, University of Tokyo, Tokyo, Japan. 4 Division of Applied Chemistry, Faculty of Urban Environmental Sciences, Tokyo Metropolitan University, Tokyo, Japan. 5 National Institute for Environmental Studies, Tsukuba, Japan.

Copyright 2011 by the American Geophysical Union. 0148‐0227/11/2011JD015637

critically important to estimate or assess the emission rates from major source regions in Asia to improve our understanding of the effect of BC on climate. [3] The uncertainty of the most recent emission estimates for the base year of 2006 is still as large as 208% for China [Zhang et al., 2009], although the uncertainty was substantially improved in comparison with that of 484% for a previous inventory [Streets et al., 2003]. Here we make independent estimates of BC emissions by comparing BC mass concentrations calculated by 3‐D chemical transport models (CTMs) with those observed in the atmosphere. BC concentrations are influenced not only by BC emissions but also by transport processes and removal during transport. The estimate of BC emissions using ambient BC data needs to include the uncertainties of predicting transport and removal of BC in the models. [4] There have been attempts to evaluate BC emissions in Asia. Carmichael et al. [2003] compared the observed and model‐calculated BC concentrations during the NASA Transport and Chemical Evolution over the Pacific (TRACE‐P) aircraft campaign to find that the Sulfur Transport Eulerian Model 2001 (STEM‐2K1) model underestimated the mean BC concentration by 20% for flight

D16201

1 of 14

D16201

KONDO ET AL.: BLACK CARBON EMISSIONS IN EAST ASIA

D16201

in 2008–2009 (R. L. Verma et al., Transport of black carbon and carbon monoxide from the Asian continent to the western Pacific in the boundary layer, submitted to Journal of Geophysical Research, 2011). Second, we excluded data substantially influenced by wet deposition by using carbon monoxide (CO) data as a water‐insoluble tracer, combined with 3‐D CTM calculations. Third, we made intercomparisons of the BC mass concentrations calculated by two CTMs with different formulations for atmospheric transport and aerosol processes to evaluate the uncertainties associated with the variability in transport and removal processes of BC. The vertical distributions of BC calculated by the models were also supported by the comparison of the BC vertical profiles obtained by aircraft observations over the East China Sea in spring 2009. We show that improved BC measurements and models have enabled quantitative estimates of BC emissions in China. Figure 1. Location of the Cape Hedo observatory and model grids (solid circles) used in the Community Multiscale Air Quality (CMAQ) model simulations. The gray scale model grids represent the annual anthropogenic emissions of black carbon (BC) for the year 2006 with a spatial resolution of 0.5° × 0.5° by Zhang et al. [2009]. The regions of north China and south China defined in this study are bounded by bold lines. altitudes below 2 km. However, the period of the comparison was too short for assessing BC emissions from Asia with sufficient statistical significance. Uno et al. [2003] compared model BC values predicted by the Regional Atmospheric Modeling System (RAMS)/Chemical Weather Forecast System (CFORS) model system with those measured at the surface on five remote islands in Japan during the Asian Aerosol Characterization Experiment (ACE Asia) campaign. These models predicted the patterns of the temporal variations of the observed BC well, leading to improved understanding of the modes of transport of BC from the Asian continent. Hakami et al. [2005] utilized four‐ dimensional variational data assimilation (4D‐Var) for the estimation of BC emissions from the observed BC values on five islands and on board a ship during ACE‐Asia. However, the accuracies of the BC measured during ACE‐Asia have not been fully quantified [e.g., Huebert and Charlson, 2000]. [5] In addition to the uncertainties of the BC measurements, the effects of wet removal of BC from the atmosphere were not fully taken into account in these studies. In order to assess the bottom‐up emission inventories accurately by comparison of the BC mass concentrations calculated by 3‐D CTMs with those observed, factors potentially leading to substantial uncertainties have to be evaluated and improved. They are (1) the accuracies of the observed BC values and (2) the uncertainties in model calculations, such as horizontal and vertical diffusion and the microphysical properties of BC (size distribution and hygroscopicity), which affect the rate of wet removal of BC. [6] We have greatly improved the uncertainties in the observed and model BC values to make estimates of BC emissions in a more reliable way compared with the previous studies. First, we measured BC at Cape Hedo on Okinawa Island in the East China Sea with an accuracy of 10%

1.2. BC Emissions in East Asia [7] In previous studies, estimates of BC emissions were made for the East Asian region for the year 2000 with a spatial resolution of 1° × 1° [Streets et al., 2003]. The inventory was updated by Zhang et al. [2009] for the year 2006 with a spatial resolution of 0.5° × 0.5°. Figure 1 shows the spatial distribution of BC (grams/second/grid) for the East Asian region [Zhang et al., 2009]. In an estimate by Bond et al. [2004], the amount of BC emitted from East Asia is about 30% of the total global anthropogenic BC emissions. According to Zhang et al. [2009], the total BC emissions from the East Asian region in 2006 are about 1.99 Tg. About 91% (1.81 Tg) of BC in East Asia is emitted from China [Zhang et al., 2009]. Because of the dominance of China in the BC emissions, we focus on the estimate of the BC emissions from China. In China, a large fraction of BC is emitted from the residential (55%) and industrial (32%) sectors, while a smaller fraction is from transport (11%) and power (2%) [Zhang et al., 2009]. These inventories are subject to an uncertainty of 208% for BC. [8] Ohara et al. [2007] estimated that BC emissions in Asia increased by 5% between 2000 and 2010. We estimate BC emissions based on the BC measurements made in 2008–2009. Our estimate should not depend on the reference BC inventory (in this case Zhang et al. [2009]) as long as the pattern of spatial distribution and seasonal variation remains unchanged. In any case, the change in the BC emissions between 2006 [Zhang et al., 2009] and 2008– 2009 (this study) is estimated to be less than 2% by linear interpolation.

2. Observations 2.1. BC and CO at Hedo [9] We made surface measurements of BC and CO at Cape Hedo observatory (26.9°N, 128.3°E, 60 m above sea level) on Okinawa Island, Japan, in the East China Sea, from 20 February 2008 to 31 May 2009. The location of the site is shown in Figure 1, together with the map of the BC emission rate estimated by Zhang et al. [2009]. The Hedo site is 600 km distant from the main island of Japan, 1000 km from Korea, and 800 km from the coast of southern China. The Hedo site is on a remote island in Asian outflows, surrounded by ocean, typically within a few days of transport

2 of 14

D16201


from the source region, thus allowing unambiguous identification of Asian BC sources. BC transported from China was clearly detected according to a detailed analysis of the temporal variations of BC and CO combined with trajectory calculations (Verma et al., submitted manuscript, 2011). More than 45% of air masses measured in winter and spring are estimated to have been influenced by BC emissions from China. [10] BC mass concentrations in the fine mode (PM2.5, i.e., particles with aerodynamic diameters smaller than 2.5 mm) were measured using a filter‐based absorption photometer, the Continuous Soot Monitoring System (COSMOS), with an inlet heated to 400°C [Miyazaki et al., 2008; Kondo et al., 2009, 2011]. The instrument monitors changes in transmittance across an automatically advancing quartz fiber filter tape at 565 nm wavelength (l). The changes in the transmittance are converted to BC mass concentrations using the mass absorption cross section (MAC) determined by comparison with a single‐particle soot photometer (SP2) based on the laser‐induced incandescence (LII) technique and the thermal‐optical transmittance (TOT) technique [Kondo et al., 2009, 2011]. The stability of MAC is achieved by removing volatile aerosol components before BC particles are collected on filters with the use of an inlet heated at 400°C. The BC mass concentrations are given in the units of mass and volume per unit volume of air at standard temperature and pressure (STP; 273.15 K and 1013 hPa), unless otherwise stated. The accuracy of the BC mass concentrations measured by COSMOS has been estimated to be about 10% by comparison with those measured by the SP2 in Tokyo and those with TOT instruments at six sites in Asia, including Jeju Island, which is located downstream of the Asian continent [Kondo et al., 2009, 2011]. About 9%–13% of the total BC mass concentration measured by the SP2 was in the size range below the detection limit of 70 nm, and the unmeasured BC was corrected for. [11] BC was measured with an integration time of 1 min. The estimated lower limit of detection (LOD) was about 0.047 mgC m−3. We used 1 h average BC data for the present analysis. [12] CO was measured continuously using a commercial instrument, TECO Model 48C (Thermo Environmental Instruments, Inc.), based on a nondispersive infrared absorption (NDIR) technique [Suthawaree et al., 2008]. The instrument was calibrated at least once per calendar year using a 1.80 parts per million by volume (ppmv) CO standard gas (Nippon Sanso, Inc.). Changes in the calibration factor between each calibration did not exceed about 5%. Zero air generated by a TECO Model 96 was sampled every hour for 20 min to correct for the drift of the background signal of the instrument. The 10 min data obtained between zero and ambient air were discarded. The measurement uncertainty was 8 parts per billion by volume (ppbv), or 6%, whichever was greater. 2.2. Vertical Distributions of BC [13] Aircraft measurements of BC mass concentrations were made during the Aerosol Radiative Forcing in East Asia (A‐FORCE) campaign conducted over the East China Sea in spring 2009 (N. Oshima et al., Wet removal of black carbon in Asian outflow: Aerosol Radiative Forcing in East Asia (A‐FORCE) aircraft campaign, submitted to Journal of

D16201

Geophysical Research, 2011). Number concentrations of BC were measured by SP2. The measurement principle and schematic diagram of the SP2 have been described previously [Schwarz et al., 2006; Gao et al., 2007; Moteki and Kondo, 2007, 2008, 2010], and detailed descriptions of the calibration of the SP2 used for A‐FORCE are given elsewhere [Moteki and Kondo, 2010; Kondo et al., 2011]. In brief, aerosol particles are introduced to the Nd:YAG laser (l = 1064 nm) beam. BC particles strongly absorb laser light and are heated to their vaporization temperature and incandesce, producing the LII signal. The peak intensity of the LII signal is a function of BC mass. The number concentrations of BC particles were measured in the diameter range of 75–850 nm. The total mass concentrations of BC were derived by integrating the number size distributions with respect to diameter. The BC mass concentrations measured by COSMOS and SP2 agreed to within 10% in Tokyo [Kondo et al., 2011]. [14] On board the King‐Air aircraft, sample air was provided to the SP2 by the use of a forward‐facing, roof‐ mounted inlet probe. Real‐time measurements of air speed, static pressure, and temperature were used to maintain isokinetic flow through the inlet tip to minimize inertial enhancement.

3. Three‐Dimensional Models 3.1. Community Multiscale Air Quality Model [15] We used the Community Multiscale Air Quality (CMAQ) model [Byun and Ching, 1999; Binkowski and Roselle, 2003], driven by the Weather Research and Forecasting (WRF) model [Skamarock et al., 2005], with a horizontal resolution of 81 km on a Lambert conformal map projection, that consisted of 117 × 69 grid cells with the center at 30°N and 110°E, covering the whole Asian region. There were 21 vertical levels from the ground surface up to 100 hPa on terrain‐following coordinates. The model simulation was performed so as to cover the entire period of BC measurement at Cape Hedo (1 February 2008 through 31 May 2009). The Asian anthropogenic emission inventory by Zhang et al. [2009] with a grid resolution of 0.5° × 0.5° in latitude and longitude developed for the year 2006 was used in the simulation. We also used the biomass burning (BB) emission inventory developed for the year 2000 by Streets et al. [2003] with a grid resolution of 1° × 1° in latitude and longitude in the simulation. Seasonal variation of BC emissions from BB was not available and was therefore not included in this study. However, the ratio of estimated annual total BC emissions from BB to the total (anthropogenic + BB) BC emission estimate is 0.13 in East Asia. Therefore, neglecting the seasonal variation of BB should not have substantially influenced the conclusions of the present study. The initial and boundary conditions of BC mass concentrations for the model domain were set to zero. A 15 day spin‐up period was used to minimize the influences of the initial conditions in the CMAQ simulation. [16] Seasonal variation of BC emissions from fossil fuel and biofuel uses was discussed by Jeong et al. [2011]. We have not included this effect directly in the model calculations because the spatial variability of the amplitude of the seasonal variation was not available. However, we took into

3 of 14

D16201


Table 1. Summary of All CMAQ Model Simulations Conducted in This Studya Simulation Name

Wet Deposition

Emission Region

CMAQ‐Modified (baseline) CMAQ‐Original CMAQ‐NoWetDep CMAQ‐China CMAQ‐NorthChina

modified scheme original scheme none modified scheme modified scheme

East Asia East Asia East Asia China north China

ing transport by comparing with the baseline simulation (CMAQ‐Modified). For the estimate of the effects of wet deposition of BC, the transport efficiency of BC, TEBC, was defined as the ratio of the BC mass concentration in the CMAQ baseline simulation to that in the CMAQ‐NoWetDep simulation. TEBC ¼

a

Emissions of black carbon (BC) from anthropogenic and biomass burning are included in all simulations. CMAQ, Community Multiscale Air Quality.

account this effect assuming linearity between emissions and ambient BC concentrations (see section 6). [17] We modified the CMAQ model to improve the calculation of wet deposition of accumulation‐mode aerosols. The Regional Acid Deposition Model (RADM) scheme in the original CMAQ model calculates wet deposition of aerosols based on precipitation water content (sum of rainwater, snow, and graupel content) and precipitation amount, when the sum of cloud water, rainwater, and graupel content exceeds a prescribed threshold value (WCTV) [Byun and Ching, 1999]. It does not distinguish rainout (in‐cloud scavenging) and washout (below‐cloud scavenging) of aerosols and does not properly take into account these removal processes. More concretely, the original CMAQ RADM scheme erroneously calculates removal of aerosols located below precipitating clouds, assuming this process as rainout, which should be treated as washout. This causes excessive removal of accumulation‐mode aerosols, because these particles are much less efficiently removed by collision of precipitating raindrops below clouds [Seinfeld and Pandis, 2006]. [18] The modified version of the CMAQ model used in this work ignores wet deposition of accumulation‐mode aerosols when the cloud water content is smaller than a prescribed threshold value. The threshold is assumed to be the same as the WCTV value used in the original scheme. Therefore, it can separately treat rainout and washout processes and ignore washout process of accumulation‐mode aerosols. The effect of this improvement is discussed in section 4.2. We note here that all aerosol species in each mode in CMAQ are assumed to be internally mixed and the accumulation‐mode aerosols are assumed to be hydrophilic, although CMAQ treats the aging processes of BC, such as condensation and coagulation [Byun and Ching, 1999; Binkowski and Roselle, 2003]. [19] A total of five CMAQ simulations, namely, one baseline and four additional sensitivity simulations, were conducted in this study. Table 1 shows a summary of all simulations conducted in this study. Emissions of BC from BB were included in all the simulations. The baseline simulation (CMAQ‐Modified) included BC emissions from the entire region of Asian countries and adopted the modified wet deposition scheme. We also performed a simulation using the CMAQ original wet deposition scheme (CMAQ‐ Original) to evaluate the effect of the modification of our wet deposition scheme. [20] Another simulation without including the effect of wet deposition (CMAQ‐NoWetDep) was also conducted to evaluate the effects of removal of BC by precipitation dur-

D16201

½ BC ðCMAQ Modified Þ : ½ BC ðCMAQ NoWetDepÞ

ð1Þ

The TEBC value is used to exclude BC data substantially influenced by wet deposition, as described in section 5.3. [21] In this study, the emission region of China was divided into north China (north of 33°N) and south China (south of 33°N) (Figure 1), identical to the divisions used for the trajectory analysis of Verma et al. (submitted manuscript, 2011). In order to evaluate the relative contributions of the BC emissions from the whole of China and the north China region, the simulations with BC emissions including only the whole of China (CMAQ‐China) and the north China region (CMAQ‐NorthChina), respectively, were performed, using the modified wet deposition scheme (Table 1). The relative contributions of BC emissions from these regions to the BC concentrations at Hedo were derived directly from these sensitivity simulations. We define fChina and fN‐China as the relative contributions of BC emitted from the whole of China and the north China region, respectively, to the CMAQ baseline simulation, as follows: fChina ¼

fN-China ¼

½ BC ðCMAQ ChinaÞ ½ BC ðCMAQ Modified Þ

½ BC ðCMAQ NorthChinaÞ : ½BC ðCMAQ Modified Þ

ð2Þ

ð3Þ

3.2. Eulerian, Multiscale Tropospheric Aerosol Chemistry and Dynamics Simulator [22] We also conducted another 3‐D model simulation using the Eulerian, Multiscale Tropospheric Aerosol Chemistry and dynamics Simulator (EMTACS) [Kajino and Kondo, 2011] to assess the uncertainty of the CMAQ results. EMTACS is also driven by the WRF model but uses different horizontal and vertical grid scales compared with WRF‐CMAQ, as described below. The EMTACS model is a size, chemical, mixing‐state, and soot‐shape resolved CTM, which simulates Brownian, turbulent, and sedimentation coagulation between various aerosol categories and thus represents a variety of atmospheric aerosol compositions and processes using the Modal Aerosol Dynamics model for multiple Modes and fractal Shapes of mass fractal agglomerates (MADMS) [Kajino, 2011]. The CMAQ and EMTACS models differ in the representations of atmospheric dynamical and microphysical processes. First, they differ in calculating dispersion processes, namely, the formulation in horizontal and vertical turbulent diffusion, numerical diffusion due to horizontal and vertical grid resolution, and transport algorithms. Second, they are different in representing the processes of evolution in size and chemistry, namely, differences in condensation and coagulation. Third, scavenging processes are represented differ-

4 of 14

D16201

D16201


Figure 2. Temporal variation of the hourly BC mass concentration observed at Hedo for data from the whole observation period (black) and data after the selection with fChina > 0.8 and TEBC > 0.8 (red). See sections 3.1, 5.2, and 5.3 for details. ently, by the use of different treatments of dry and wet deposition. The evolution of the size distribution of aerosols plays important roles in their cloud condensation nuclei activity and therefore scavenging of aerosols [Oshima et al., 2009a, 2009b]. [23] The simulation using EMTACS was conducted to evaluate the model variability from the comparisons between the two model results. The horizontal grid resolution was 60 km on a Lambert conformal map projection that consisted of 100 × 70 grid cells with a center at 35°N and 115°E, covering the East Asian region. There were 27 vertical levels from the ground surface to 10 km on terrain‐following coordinates. The EMTACS simulation was conducted only for spring 2009 (between 10 March and 30 April). The same Asian anthropogenic emission inventory was used in the EMTACS simulation [Zhang et al., 2009]. The initial and boundary conditions of BC mass concentrations for the model domain were set to zero. The model formulation, settings, and evaluation of performance are described in detail by Kajino and Kondo [2011].

4. Observed and Calculated BC 4.1. Temporal Variations [24] Figure 2 shows the temporal variation of the hourly BC mass concentrations observed at Cape Hedo throughout the entire observation period. The average and median values of BC mass concentrations for each season are summarized in Table 2. The BC concentrations were highest in winter and spring, with the largest variability (Figure 2 and Table 2). The periods of the variations were of the order of a few to several days. The lowest average BC concentrations were observed in summer. In summer, the low‐level circulation was characterized by the prevailing southeasterly flow in association with the Pacific high, transporting clean marine air to the East China Sea. Verma et al. (submitted manuscript, 2011) have given a more detailed analysis of the seasonal variation of BC and its variability. [25] Figure 3 shows the spatial distributions of the seasonally averaged mass concentrations of BC calculated by

the CMAQ baseline simulation at the surface for summer (June–August 2008), winter (December 2008 to February 2009), and spring (March–May 2009). In winter, BC was predicted to be highest over the entire Asian continent (China and Korea) because of the reduced upper boundary of the planetary boundary layer (PBL) over this area. The BC emission due to domestic heating in winter is anticipated to be higher in the northern part of the continent than the flux in summer according to Zhang et al. [2009]. However, no seasonal variations in the BC emissions were included in the CMAQ simulations. In spring, BC was still higher over the East China Sea and the southern part of Japan because synoptic‐scale migratory cyclones, which frequently transported pollutants from the Asian continent, were active in this season. In summer, in contrast, inflow of clean marine air reduced BC over the East China Sea, consistent with the trajectory analysis (Verma et al., submitted manuscript, 2011). [26] Table 3 summarizes the average and median BC mass concentrations calculated by the CMAQ baseline simulation for each season. The model reproduced well the observed seasonal variation of BC (Tables 2 and 3). In spring and winter, the fChina values were greater than 0.5, indicating that BC concentrations at Hedo were largely influenced by Chinese BC emissions. The median values of the model simulation agreed with the observations to within 15% for these seasons. 4.2. Effect of Wet Deposition [27] Figure 4 shows temporal variations of the hourly BC mass concentration observed at Hedo and those calculated by the three CMAQ simulations, namely, the CMAQ‐ Modified, CMAQ‐NoWetDep, and CMAQ‐Original simulations (see Table 1), in spring from March to May 2008, when Asian outflows were frequent. The mean BC concentrations and standard deviations (in parentheses) observed in spring 2008 and those predicted by the CMAQ‐Modified, CMAQ‐NoWetDep, and CMAQ‐Original simulations were 414 (377), 431 (269), 642 (324), and 327 (266) ng m−3, respectively. The CMAQ‐Modified simulation generally reproduced well the observed temporal variations of BC (4% overestimation of the mean observed BC concentration), suggesting that regional‐scale transport of pollutants was represented well in the simulation. [28] In comparison with the CMAQ‐Modified simulation, the BC values by the CMAQ‐Original simulation were occaTable 2. Average and Median Values of BC Mass Concentrations Observed at Hedo During Each Season Period

a

Spring (MAM) 2008 Summer (JJA) 2008 Fall (SON) 2008 Winter (DJF) 2008–2009 Spring (MAM) 2009 Entire period

Number of Average BCc Data Pointsb (mg m−3) 2,169 2,208 2,124 2,154 2,208 11,020

0.41 0.12 0.18 0.38 0.36 0.30

(0.38) (0.11) (0.21) (0.50) (0.31) (0.37)

Median BCd (mg m−3) 0.31 (0.091–0.77) 0.081 (0.034–0.21) 0.12 (0.041–0.31) 0.23 (0.091–0.56) 0.27 (0.10–0.65) 0.17 (0.058–0.52)

a MAM, Mar‐Apr‐May; JJA, Jun‐Jul‐Aug; SON, Sep‐Oct‐Nov; DJF, Dec‐Jan‐Feb. b Number of hourly observed BC data points used for the statistics. c Values in parentheses are the standard deviations. d Values in parentheses are the central 67% ranges.

5 of 14

D16201


D16201

sionally zero when air parcels had likely been influenced by precipitation during transport, substantially underestimating the observed BC concentrations. The CMAQ‐Original 7simulation underestimated the mean observed BC concentrations by 21%. [29] The BC values derived from the CMAQ‐NoWetDep simulation were systematically greater than those from the CMAQ‐Modified simulation. The differences in these two simulations indicate the effects of wet deposition of BC during transport. The simulation excluding wet deposition calculations overestimated the mean observed BC concentrations by 55%, indicating the importance of wet deposition during transport on BC mass concentrations downwind of Asian outflows. The CMAQ‐Modified simulation generally reproduced the observed BC values. However, a more detailed comparison suggests that the CMAQ‐Modified simulation overestimates the observed BC to some extent, as discussed in section 5.3.

Figure 3. Seasonally averaged mass concentrations of BC at the surface calculated by the CMAQ baseline (CMAQ‐ Modified) simulation for summer (June–August 2008), winter (December 2008 to February 2009), and spring (March–May 2009). The white dot denotes the location of Cape Hedo.

4.3. Overall Uncertainties in the Model Calculations 4.3.1. Comparison With Surface BC Data [30] Figure 5 shows comparisons of the hourly BC mass concentrations predicted by the CMAQ‐Modified and EMTACS simulations with those observed at Hedo in April 2009. The two model calculations used the same emissions inventory [Zhang et al., 2009]. Both model simulations generally reproduced the observed temporal variations of BC well. Mean BC concentrations and standard deviations (in parentheses) observed in April 2009 and those predicted by CMAQ‐Modified and EMTACS were 421 (316), 469 (296), and 316 (239) ng m−3, respectively. [31] Figure 6 shows the monthly mean surface BC mass concentrations over East Asia predicted by the CMAQ‐ Modified and EMTACS simulations and the correlation coefficients (r2) of time‐dependent surface BC concentrations between the two model results for April 2009. Despite the differences in the model formulations between CMAQ and EMTACS, the horizontal distributions of BC calculated by the two models were similar in their patterns and absolute values. A correlation coefficient (r2) of 0.69 was obtained for the location of Hedo (Figure 6). These results give a measure of the validity of the CMAQ simulations at this location. [32] Over the north China region where the anthropogenic BC emissions were large (Figure 1), BC mass concentrations were mostly controlled by near‐field emissions. However, it is practically difficult to evaluate the BC emissions with a spatial resolution of 0.5° × 0.5° accurately from the observations near the large sources over the continent, because spatial and temporal variations of BC concentrations over the source regions were much larger than those over the East China Sea. A number of observational sites over a large area are required to obtain BC values representing average values over the source regions. On the other hand, over the downwind area, such as the East China Sea, where there are no large BC emissions, the BC mass concentrations are mainly controlled by transport from the Chinese continent and wet deposition during transport. Air parcels from the continent were well‐mixed vertically and horizontally during transport, resulting in relatively uniform spatial and temporal variations of surface BC concentrations around Cape Hedo. We note that the relatively higher values of the

6 of 14

D16201

D16201

KONDO ET AL.: BLACK CARBON EMISSIONS IN EAST ASIA Table 3. Average and Median Values of Model‐Calculated BC for Hedo During Each Season, Ratios of the Median Values of the Modeled BC to Those of Observed BC, and the Contribution of BC Emissions From China Period

Number of Data Pointsa


2,169 2,208 2,124 2,154 2,208 11,020

Average BCb (mg m−3) 0.43 0.23 0.32 0.40 0.39 0.36

(0.27) (0.11) (0.30) (0.44) (0.28) (0.32)

Median BCc (mg m−3) 0.36 0.21 0.22 0.23 0.29 0.25

(0.20–0.68) (0.14–0.32) (0.10–0.52) (0.12–0.70) (0.13–0.69) (0.13–0.59)

Model(med)/ Obs(med)d

fChinae

1.15 2.54 1.90 0.99 1.07 1.48

0.53 0.068 0.25 0.53 0.52 0.37

a

Number of hourly modeled BC data points used for these statistics. Values in parentheses are the standard deviations. c Values in parentheses are the central 67% ranges. d Ratios of the median values of modeled BC to those of observed BC. e Ratios of the median BC values calculated including BC emissions only from China to those including emissions from all of East Asia, defined by equation (2). b

correlation coefficients (r2 = 0.6–0.8), shown in Figure 6, were obtained because of the mixing processes during transport, although some BC particles were removed from the atmosphere because of deposition. These results indicate that Cape Hedo is one of the most appropriate locations to measure the emission flux and transport from the Chinese continent if we can exclude the effects of wet deposition during transport. [33] Figure 7 shows the correlation between the observed BC concentrations with those calculated by CMAQ‐Modified for the entire period. They are moderately correlated (r2 = 0.39) with a slope of 0.8. The model BC is suggested to be lower based on the slope of model‐observed BC correlation. However, the ratio of median values of modeled BC to that of observed BC, denoted as the model(med)/obs(med) ratio, for the entire period (1.48 in Table 3) indicates the reverse. These results suggest that the uncertainty in statistically estimating of the model(med)/obs(med) ratios was about ±50%. 4.3.2. Comparison With Aircraft BC Data [34] Figure 8 shows vertical profiles of the median values of the BC mass concentrations observed at latitudes between 26°N and 33°N over the East China Sea in March–April 2009 during the A‐FORCE aircraft campaign. In Figure 8 the BC concentrations are not in units of mg STP m−3 but rather with respect to ambient temperature and pressure because of the use of column BC amounts, discussed below. The median BC values predicted by the CMAQ‐Modified and EMTACS simulations are also shown for comparison. Within 0–2 km in altitude, the model‐calculated BC median values agreed with those observed to within 21% and 17% for CMAQ and EMTACS, respectively. Above 2 km, the CMAQ‐Modified simulation systematically overestimated BC values, although the BC profiles were generally well reproduced. [35] It is possible that vertical transport of BC from the PBL to the free troposphere (FT) was overestimated in CMAQ, because CMAQ tended to give higher values of the vertical eddy diffusion coefficient, as was suggested by comparison between CMAQ and WRF‐chem model simulations over Beijing, China [Matsui et al., 2009]. Another possibility is that the CMAQ‐Modified simulation underestimated wet deposition of BC to some extent, as discussed in section 5.3. The effect of wet deposition on BC concentrations is much larger in the FT than in the PBL. The uncertainties in the model estimate of wet deposition

have little influence on the analysis of the surface BC data, because the BC data substantially influenced by wet deposition were excluded in this study, as discussed in section 5.3. [36] By integrating the median values of observed BC mass concentrations, the contribution of the BC amounts

Figure 4. Comparisons of the hourly BC mass concentrations observed at Hedo (black) and those calculated by the CMAQ‐Modified (red), CMAQ‐NoWetDep (green), and CMAQ‐Original (blue) simulations for spring 2008.

7 of 14

D16201


D16201

Figure 5. Comparison of the hourly BC mass concentrations observed (black) and those calculated by CMAQ‐ Modified (red) and Eulerian, Multiscale Tropospheric Aerosol Chemistry and dynamics Simulator (EMTACS) (blue) at Hedo in April 2009. below 2 km to the total column amounts was 78% (Figure 8). CMAQ and EMTACS predicted the contribution of BC below 2 km to be 59% and 63%, respectively. This indicates the limited influence of the BC in the FT on the budget of BC near the surface.

5. Selection of BC Data 5.1. Methods of Selection [37] We selected the BC data observed at Hedo strongly affected by emissions from China. For this purpose we used the fChina value calculated by the CMAQ simulations (see equation (2)). In addition to this selection, the BC data least influenced by wet deposition during transport were selected to minimize the uncertainty in the prediction of the effect of wet deposition. This selection was made by using the transport efficiency, TEBC, calculated by the CMAQ simulations (see equation (1)). This data selection enabled the comparison of BC transported from China, maintaining statistical reliability and ensuring minimal model uncertainty in estimating the effect of wet deposition. 5.2. Selection by Emissions [38] Figure 9 (top) shows the observed BC‐CO correlation for the entire observational period, color coded by fChina. The slope of all data samples is 3.9 ng m−3 ppbv−1, with r2 = 0.62. It was found that high BC concentrations (>1.0 mg m−3) at Hedo were associated with the transport of air masses originating from China. [39] In order to select the observed BC data strongly affected by Chinese emissions, we used an fChina value of 0.8 as the threshold. The statistics of the data remaining after the selection of fChina > 0.8 are summarized in Table 4. After the selection, the number of remaining data points was reduced to 19%. As discussed in section 4.1, the low‐level circulation over Cape Hedo was characterized by the prevailing southeasterly flow in summer and early fall. This led to median fChina values in summer and in fall as small as

Figure 6. Comparison of the monthly mean surface BC mass concentrations over East Asia calculated by (top) the CMAQ‐Modified and (middle) the EMTACS simulations for April 2009. (bottom) The correlation coefficients (r2) of time‐dependent surface BC mass concentrations between the two model results. The white dot denotes the location of Cape Hedo.

8 of 14

D16201


D16201

0.068 and 0.25, respectively (Table 3). No data remained after the selection for summer and only 13% of the data remained for fall. The model(med)/obs(med) ratio for all the data was 1.25, indicating good agreement of the model simulations with the observations. The median fN‐China/ fChina ratios were larger than 0.6 throughout the period, with a maximum of 0.83 in fall. 5.3. Selection by Wet Deposition [40] Figure 9 (bottom) shows the observed BC‐CO correlation for the data selection with fChina > 0.8, color coded by TEBC. The slope of the all data samples is 5.3 ng m−3 ppbv−1, with r2 = 0.61. The slope of the data with TEBC > 0.8 is systematically higher than that for the whole data set, supporting the validity of the selection method. [41] Figure 10a shows the slope of the BC‐CO correlations (DBC/DCO ratio) and r2 versus the threshold of TEBC. For example, when we adopted a threshold value of 0

Figure 7. Correlation plots of the observed hourly BC values at Hedo with those calculated by CMAQ‐Modified for the entire period of the observations. The slope of the regression line, the correlation coefficient, and the number of data points are also shown.

Figure 8. Comparisons of the altitude distributions of the median BC mass concentrations observed at 26°N–33°N in spring 2009 by aircraft during A‐FORCE (solid circles) and those calculated by the CMAQ‐Modified (open circles) and EMTACS (open triangles) simulations. The values are averages for each 1 km in altitude. Horizontal lines denote the central 67% ranges of the BC values. The median and mean values of hourly BC values at Hedo during the same period are also shown for comparison. The BC values shown in this figure are in units of mg m−3 at ambient temperature and pressure.

Figure 9. (top) Scatterplot of hourly CO and BC concentrations observed at Hedo for all data, color coded by fChina. (bottom) Scatterplot of hourly CO and BC concentrations observed at Hedo after data selection with fChina > 0.8, color coded by TEBC.

9 of 14

D16201

D16201

KONDO ET AL.: BLACK CARBON EMISSIONS IN EAST ASIA Table 4. Median Values of Observed and Model‐Calculated BC at Hedo During Each Season, Ratios of Median Values of Modeled BC to Those Observed, and the Contributions of BC Emissions From North China for Data Selected by fChina > 0.8 Period


Observed BCb (mg m−3)

Modeled BCb (mg m−3)

Model(med)/ Obs(med)

fN‐Chinac/fChina


624 0 271 714 377 2091

0.56 (0.32–1.17)

0.69 (0.45–1.00)

1.22

0.79

0.28 0.45 0.64 0.51

0.55 0.57 0.61 0.64

1.96 1.26 0.96 1.25

0.83 0.61 0.75 0.75

(0.13–0.75) (0.19–1.02) (0.40–0.95) (0.24–1.06)

(0.35–1.19) (0.19–1.32) (0.29–0.99) (0.28–1.07)

a

Number of hourly observed BC data points used for these statistics. Values in parentheses are the central 67% ranges. c Ratios of the median BC values calculated including emissions only from north China to those including emissions from all of East Asia, defined by equation (3). b

for the TEBC value (i.e., data selection for TEBC > 0), the slope of the BC‐CO correlation was 5.3 ng m−3 ppbv−1 with r2 = 0.61, as shown in Figure 9 (bottom). The slope of the BC‐CO correlation and r2 increased with the increase in the threshold of TEBC, quantitatively supporting the validity of the data selection using TEBC. [42] Figure 10a also shows the number of the remaining data points versus TEBC. The number of the data points decreased strongly with the increase in the threshold of TEBC, giving a measure of statistical significance depending on TEBC. The decrease in the DBC/DCO ratio at TEBC > 0.85 is interpreted as due to the decrease in the statistical significance, as indicated by the corresponding decrease in r2. We have selected data with a threshold TEBC of 0.8 as an optimal value, based on this interpretation. For this data set, the DBC/DCO ratio was about 7.5 ng m−3 ppbv−1, with r2 = 0.82. [43] The TEBC values underwent pronounced seasonal variations, with a minimum in late summer and maximum in winter and spring. The pattern of the seasonal variations of the TEBC values was similar to the DBC/DCO ratios (Verma et al., submitted manuscript, 2011). It should also be noted that the DBC/DCO ratios for TEBC > 0.8 showed some seasonal and year‐to‐year variations (Verma et al., submitted manuscript, 2011), and DBC/DCO = 7.5 ng m−3 ppbv−1 is considered to be the average value of the entire observational period. [44] Table 5 summarizes the statistics of the remaining data points after the selection of fChina > 0.8 and TEBC > 0.8. After the selection, the total number of remaining data points was reduced to 10%. More than 1000 samples were still available for the analysis. The model(med)/obs(med) ratio was 1.09 for the entire subsample (fChina > 0.8 and TEBC > 0.8). The fN‐China/fChina ratios were larger than 0.75 throughout the period, with a maximum of 0.95 in the spring of 2009. This result indicates that BC emissions from north China had significant influences on BC concentrations at Hedo. [45] The fN‐China/fChina ratios were generally higher for TEBC > 0.8 than those without selection by TEBC, as can be seen from Tables 4 and 5. The average fN‐China/fChina ratios increased from 0.75 to 0.80 by the selection for TEBC > 0.8. This indicates that wet deposition of BC was more efficient for south China air than for north China air. This is generally consistent with more precipitation at lower latitudes over China on average (Verma et al., submitted manuscript, 2011).

Figure 10. (a) Changes in the slopes of the observed BC‐ CO correlations (solid circles), their correlation coefficients (open circles), and the number of remaining data points (solid triangles) with the changes in the threshold value of TEBC. (b) Changes in the ratios of median values of modeled BC to those observed after the data selection with fChina > 0.8 (open squares) and the number of remaining data points (solid triangles) with changes in the threshold value of TEBC.

10 of 14

D16201

D16201

KONDO ET AL.: BLACK CARBON EMISSIONS IN EAST ASIA Table 5. Median Values of Observed BC at Hedo During Each Season with fChina > 0.8 and TEBC > 0.8, Those of Model‐Calculated BC, Ratios of the Median Values of Modeled BC to Those Observed, and the Contributions of BC Emissions From North China Period


Observed BCb (mg m−3)

Modeled BCb (mg m−3)


fN‐China/ fChina


428 0 176 268 214 1135

0.68 (0.32–1.20)

0.71 (0.42–1.08)

1.04

0.77

0.35 0.56 0.65 0.60

0.65 0.52 0.53 0.66

1.86 0.92 0.81 1.09 RMS error = 0.42

0.84 0.75 0.95 0.80

(0.18–0.93) (0.25–1.31) (0.40–0.95) (0.27–1.15)

(0.33–1.43) (0.21–1.42) (0.27–0.95) (0.28–1.17)

a

Number of hourly observed BC data points used for these statistics. Values in parentheses are the central 67% ranges.

b

[46] It should be noted here that the model(med)/obs(med) ratios are lower if the data with TEBC > 0.8 are selected for spring and winter (Tables 4 and 5). A similar trend is seen for all seasons. For a more detailed investigation of this relationship, Figure 10b shows the model(med)/obs(med) ratio and the number of the remaining data points versus the threshold of TEBC using the data selected with fChina > 0.8. The model(med)/obs(med) ratio was nearly constant (1.25) at a threshold TEBC < 0.7. The ratio started to decrease rapidly with the increase in TEBC at TEBC ∼ 0.7. These results suggest an underestimation of the effect of wet deposition of BC by the CMAQ‐Modified simulation for data with TEBC < 0.7. [47] It should be noted that the BC‐CO slope also started to increase rapidly with the increase in TEBC at TEBC ∼ 0.7 (Figure 10a). At 0.8 < TEBC < 0.9, the BC‐CO slope changed little, while the model(med)/obs(med) ratio decreased by about 6.3% (= 1.02/1.09). This gives a measure of the uncertainty in the model(med)/obs(med) ratio

Figure 11. Correlation plots of the hourly BC values observed at Hedo with those calculated by CMAQ‐Modified after data selection with fChina > 0.8 and TEBC > 0.8 for the entire period of the observations. See sections 3.1, 5.2, and 5.3 for details.

associated with the choice of the threshold TEBC (in this case, 0.8). [48] Despite the practical usefulness of the model(med)/ obs(med)‐TEBC relationship, the interpretation of the relation and its comparison with the BC‐CO slope are not necessarily straightforward. First, the threshold TEBC was derived from model calculations. Second, the TEBC values shown in Figures 10a and 10b are used as thresholds. Therefore, the three parameters (BC‐CO slope, TEBC, and the model(med)/obs(med) ratio) may not be totally independent but probably interrelated. [49] As shown in section 4.2, the CMAQ simulation using the original wet deposition scheme has been shown to be an overestimate of wet deposition effects. Therefore, the actual wet deposition rates might be between the values calculated by the two schemes, when testing the emission inventory.

Figure 12. Histogram of the hourly BC mass concentrations observed at Hedo (solid circles) and those calculated by the CMAQ‐Modified simulation (open circles) after data selection with fChina > 0.8 and TEBC > 0.8 for the entire period of the observations. The number of data points is given at 0.2 mg m−3 intervals. See sections 3.1, 5.2, and 5.3 for details.

11 of 14

D16201

Figure 13. Comparison of the median and the central 67% ranges of hourly BC mass concentrations observed at Hedo in winter, fall, spring, and the entire period (February 2008 to May 2009) (solid circles) and those calculated by the CMAQ‐Modified simulation with (open squares) and without (open circles) taking into account the seasonal variation of BC emissions after data selection with fChina > 0.8 and TEBC > 0.8. See sections 3.1, 5.2, 5.3, and 6 for details.

This confirms the importance of the selection of the BC data least affected by wet deposition to minimize uncertainties caused by errors in estimating the effect of wet deposition by the model.

6. Model Comparison With Selected Data [50] The remaining BC data after the selection (fChina > 0.8 and TEBC > 0.8) are shown in Figure 2 with the entire observed BC data set. Figure 2 shows that some of the peaks in the BC concentrations corresponded to the selected data. The results of the CMAQ baseline simulation after the same data selection were compared with these data in three different ways. [51] First, Figure 11 shows the correlation between the observed BC concentrations with those calculated by CMAQ‐Modified for the entire period. The correlation coefficient (r2 = 0.36) did not improve, and the slope of 0.77

Table 6. Anthropogenic Emissions of BC in China for Different Seasons in 2006 and Their Ratios to the Annually Averaged Emissions Period

BC Emissiona (Gg)

Cseasonb

Spring (MAM) Summer (JJA) Fall (SON) Winter (DJF) Total (year)

405 348 408 650 1811

0.895 0.769 0.901 1.44 1.00

was similar even after the data selection. This suggests that the major cause of the variability of the correlation was not due to the uncertainty in the model prediction of wet removal of BC. It was partly due to errors in precisely predicting the timing of the transport of plumes from China with time resolutions of a few hours. [52] Second, in order to avoid the uncertainty of the precise prediction of transport by the CMAQ model, we compared the frequency distributions of observed and model‐calculated BC mass concentrations at Hedo, as shown in Figure 12. The shapes of the distribution of the observed and modeled BC values were very similar, with the most frequent values of about 0.2–0.6 mg m−3. The very low frequency of the observed BC exceeding 1.6 mg m−3 was also predicted by the model. In addition, we also compared the lowest part of the BC values. The average observed value of the selected BC within the lowest fifth percentile was 113 ng m−3. The corresponding model mean value was 127 ng m−3, indicating that the model predicted the background BC values of the plumes transported from China to within 12%. Comparison of the frequency distributions is advantageous in that it is not influenced by the uncertainties in predicting the timing of the transport of small‐scale plumes from China to Cape Hedo. [53] The agreement of the frequency distributions provides support in using statistical parameters such as median and mean values for quantitative comparison between the model and observations. Therefore, third, we compared the median values of the observed and modeled BC values after the data selection for winter and fall (2008) and spring (2008 and 2009), as shown in Figure 13 and Table 5. The median values of the modeled BC agreed with those observed to within 20% for winter and spring. In addition, the central 67% ranges of the observed and model calculated BC values agreed well, as seen from the bars shown in Figure 13 and Table 5. The model(med)/obs(med) ratio was 1.09 for the whole period, as mentioned. The deviation of the ratio from this value was largest for fall. The cause of the large discrepancy for fall is not understood. [54] The contribution of the north China region (>33°N) was estimated to be 80%, indicating that outflows from north China had dominant influences on the BC concentrations at Hedo. The frequency of the transport of BC emitted from south China was much smaller than that for north China, as shown by trajectory analysis (Verma et al., Table 7. Median Values of the Season‐Corrected Model‐Calculated BC at Hedo During Each Season With fChina > 0.8 and TEBC > 0.8, and the Ratios of Median Values of Season‐Corrected Modeled BC to Those Observeda Period

Number of Data Pointsb

Modeled BCc (mg m−3)



428 0 176 268 214 1135

0.65 (0.38–0.98)

0.95

a

Values taken from Zhang et al. [2009]. The season‐corrected coefficient (Cseason) is defined as the ratio of the BC emission for each season to that allocated equally for each season (= 1811/4 = 453 Gg per 3‐month period). For example, the Cseason value of 0.895 for spring (MAM) is given as 405/453. b

D16201


0.59 0.73 0.48 0.62

(0.31–1.29) (0.29–2.00) (0.25–0.86) (0.29–1.18)

1.68 1.29 0.74 1.03 RMS error = 0.38

a The season‐corrected modeled BC mass concentration for each season ([BC]season) is defined by equation (4). b Number of hourly observed BC data points used for these statistics. c Values in parentheses are the central 67% ranges.

12 of 14

D16201


Table 8. Contributions of the Individual Uncertainties to the Overall Uncertainty in Estimating BC Emissions From China Individual and Overall Uncertainties

Values

BC measurements by COSMOS Choice of the threshold TEBC Seasonal variations of BC emissions Variability in the model(med)/obs (med) ratio Overall

10% 6.3% 2.8% 38% 40% (RMS error)

submitted manuscript, 2011). A substantial portion of pollutants emitted from south China is transported to higher altitudes due to more frequent convective activity in spring [Oshima et al., 2004]. [55] Thus far, we used the annual mean emission inventory of Zhang et al. [2009]. However, strong seasonal variations of BC emissions were also shown by Zhang et al. [2009]. Table 6 shows the anthropogenic BC emissions in China estimated for different seasons in 2006. The BC emissions were largest in winter and smallest in summer. The season‐corrected coefficients (Cseason), defined as the ratio of the BC emission for each season to that allocated equally for each season (= 1811/4 Gg per 3‐month period; see Table 6), were 1.44 and 0.895 in winter and spring, respectively. We calculated the season‐corrected modeled BC mass concentration for each season ([BC]season) using Cseason and the results from the two CMAQ simulations (see Table 1). ½BCseason ¼ ½BCðCMAQ-ModifiedÞ ½BCðCMAQ-ChinaÞ þ ½BCðCMAQ-ChinaÞ Cseason ¼ ½BCðCMAQ-ModifiedÞ þ ½BCðCMAQ-ChinaÞ ðCseason 1Þ:

ð4Þ

The median values and central 67% ranges of the [BC]season values for each season are shown in Figure 13 and Table 7. It should be noted that the estimate of the seasonal variation of the BC emissions was given for the whole of China, and we could not include possible differences in the seasonal variation between north China and south China. The model (med)/obs(med) ratio for the entire period was 1.03.

7. Estimate of BC Emissions From China [56] The model(med)/obs(med) ratios for the entire period were 1.03 and 1.09 with and without, respectively, seasonal variation of BC emissions, as discussed in section 6. Considering the possible uncertainty of the seasonal variation, especially the difference between north China and south China, we consider the average of the two ratios as a representative central value, namely, 1.06 ± 0.03. [57] In addition to the 2.8% (= 0.03/1.06) uncertainty estimated from the difference of the central values of the model(med)/obs(med) ratios, we estimated the overall uncertainty of the comparison between the model and observations using the deviations of the model(med)/obs(med) ratios for different seasons from the central values. The uncertainties estimated from the RMS error were calculated to be 37% (= 0.38/1.03; see Table 7) and 38% (= 0.42/1.09; see Table 5) with and without, respectively, seasonal vari-

D16201

ation of BC emissions. The overall uncertainty is estimated to be 40% as the RMS error of the individual uncertainties discussed above. As summarized in Table 8, they are 38% variability in the model(med)/obs(med) ratios, 10% uncertainty of the BC observations (section 2.1), 6.3% uncertainty in the model(med)/obs(med) ratios associated with the choice of the threshold TEBC (section 5.3), and 2.8% uncertainty of the central model(med)/obs(med) ratios. [58] This means that we estimate the total BC emission flux over China to be 1.06 of the estimate of Zhang et al. [2009], with an uncertainty of 40%. Although the absolute values of the two estimates turned out to be very similar, the uncertainty of our estimate is much smaller than the 208% uncertainty given by Zhang et al. [2009].

8. Conclusion [59] We have estimated the emission rate of BC from China, which is the largest source of BC in East Asia. This has been conducted by comparing BC values observed in the Asian outflows and those predicted by 3‐D CTMs that used the emissions inventory of Zhang et al. [2009]. For this purpose, we used BC mass concentrations measured by COSMOS at Cape Hedo from February 2008 to May 2009. Hedo is located sufficiently distant from China to represent a relatively uniform field of BC over the East China Sea while still capturing signals of BC emissions from China. The seasonal median BC at Hedo was highest in winter (0.23 mg m−3) and spring (0.31 mg m−3), when high‐BC plumes from China were often transported to the site. The CMAQ‐Modified simulation was used to calculate the mass concentrations of BC using the annual mean emission inventories of Zhang et al. [2009] for the base year of 2006. The model reproduced well the temporal variations of surface BC mass concentrations observed at Hedo during the whole observation period and the median BC vertical profile obtained by aircraft over the East China Sea in spring 2009. [60] In order to assess the model variability in transport and removal processes of BC, we intercompared CMAQ‐ Modified and EMTACS simulations, in which formulations of atmospheric dispersion and aerosol processes are different. The difference in the predicted BC mass concentrations between the models was small around Cape Hedo, with r2 = 0.6–0.8, indicating the smaller uncertainty in the model predictions of BC at the location of Hedo. [61] We selected BC data strongly affected by emissions from China (fChina > 0.8) and least affected by wet deposition during transport (TEBC > 0.8). After these selections, r2 for the BC‐CO correlation increased to 0.82, with a substantial amount of data (N > 1000) remaining for statistical analysis. The model(med)/obs(med) ratios were derived by taking into account the seasonal variations of BC emissions. The model(med)/obs(med) ratio for the entire period was 1.06 ± 0.03. The overall uncertainty of the ratio was estimated to be 40%. [62] This means we estimate the annually averaged BC emission flux over all of China to be 1.92 Tg yr−1, with an uncertainty of about 40%. The value is very close to the value of 1.81 Tg yr−1 estimated by Zhang et al. [2009]. Although the absolute values of the two estimates turned out to be very similar, partly by coincidence, the uncertainty of

13 of 14

D16201


our estimate is much smaller than the 208% uncertainty given by Zhang et al. [2009]. [63] Acknowledgments. This work was supported by the Ministry of Education, Culture, Sports, Science, and Technology (MEXT); the strategic international cooperative program of the Japan Science and Technology Agency (JST); and the global environment research fund of the Japanese Ministry of the Environment (A‐0803 and A‐1101). M.K. and N.O. were supported by Research Fellowships of the Japan Society for the Promotion of Science (JSPS) for Young Scientists. CO measurement was supported by the Okinawa Prefecture Institute for Health and Environment. We thank K. Kawana for her assistance with the field measurements.

References Binkowski, F. S., and S. J. Roselle (2003), Models‐3 Community Multiscale Air Quality (CMAQ) model aerosol component: 1. Model description, J. Geophys. Res., 108(D6), 4183, doi:10.1029/2001JD001409. Bond, T. C., D. G. Streets, K. F. Yarber, S. M. Nelson, J.‐H. Woo, and Z. Klimont (2004), A technology‐based global inventory of black and organic carbon emissions from combustion, J. Geophys. Res., 109, D14203, doi:10.1029/2003JD003697. Byun, D. W., and J. K. S. Ching (1999), Science algorithms of the EPA Models‐3 Community Multiscale Air Quality (CMAQ) modeling system, Rep. EPA/600/R‐99/030, U. S. Environ. Prot. Agency, Washington, D. C. Carmichael, G. R., et al. (2003), Evaluating regional emissions estimates using the TRACE‐P observations, J. Geophys. Res., 108(D21), 8810, doi:10.1029/2002JD003116. Gao, R. S., J. P. Schwarz, K. K. Kelly, D. W. Fahey, L. A. Watts, and T. L. Thompson (2007), A novel method for estimating light‐scattering properties of soot aerosols using a modified single‐particle soot photometer, Aerosol Sci. Technol., 41, 125–135, doi:10.1080/02786820601118398. Hakami, A., D. K. Henze, J. H. Seinfeld, T. Chai, Y. Tang, G. R. Carmichael, and A. Sandu (2005), Adjoint inverse modeling of black carbon during the Asian Pacific Regional Aerosol Characterization Experiment, J. Geophys. Res., 110, D14301, doi:10.1029/2004JD005671. Hansen, J., and L. Nazarenko (2004), Soot climate forcing via snow and ice albedos, Proc. Natl. Acad. Sci. U. S. A., 101, 423–428, doi:10.1073/ pnas.2237157100. Huebert, B., and R. J. Charlson (2000), Uncertainties in data on organic aerosols, Tellus, 52B, 1249–1255. Intergovernmental Panel on Climate Change (IPCC) (2007), Climate Change 2007: The Physical Science Basis—Contribution of Working Group I to the Fourth Assessment Report of the IPCC, Cambridge Univ. Press, Cambridge, U. K. Jacobson, M. Z. (2001), Strong radiative heating due to the mixing state of black carbon in atmospheric aerosols, Nature, 409, 695–697, doi:10.1038/35055518. Jeong, J. I., R. J. Park, J.‐H. Woo, Y.‐J. Han, and S.‐M. Yi (2011), Source contributions to carbonaceous aerosol concentrations in Korea, Atmos. Environ., 45, 1116–1125, doi:10.1016/j.atmosenv.2010.11.031. Kajino, M. (2011), MADMS: The Modal Aerosol Dynamics model for multiple Modes and fractal Shapes in the free‐molecular and the near‐continuum regimes, J. Aerosol Sci., 42, 224–248, doi:10.1016/j. jaerosci.2011.01.005. Kajino, M., and Y. Kondo (2011), EMTACS: Development and regional‐ scale simulation of a size, chemical, mixing type, and soot shape resolved atmospheric particle model, J. Geophys. Res., 116, D02303, doi:10.1029/ 2010JD015030. Kondo, Y., et al. (2009), Stabilization of the mass absorption cross section of black carbon for filter‐based absorption photometry by the use of a heated inlet, Aerosol Sci. Technol., 43, 741–756, doi:10.1080/ 02786820902889879. Kondo, Y., L. Sahu, N. Moteki, F. Khan, N. Takegawa, X. Liu, M. Koike, and T. Miyakawa (2011), Consistency and traceability of black carbon measurements made by laser‐induced incandescence, thermal‐optical transmittance, and filter‐based photo‐absorption techniques, Aerosol Sci. Technol., 45, 295–312, doi:10.1080/02786826.2010.533215. Matsui, H., et al. (2009), Spatial and temporal variations of aerosols around Beijing in summer 2006: Model evaluation and source apportionment, J. Geophys. Res., 114, D00G13, doi:10.1029/2008JD010906. Miyazaki, Y., Y. Kondo, L. K. Sahu, J. Imaru, N. Fukushima, and M. Kano (2008), Performance of a newly designed continuous soot monitoring system (COSMOS), J. Environ. Monit., 10, 1195–1201, doi:10.1039/ b806957c.

D16201

Moteki, N., and Y. Kondo (2007), Effects of mixing state on black carbon measurement by laser‐induced incandescence, Aerosol Sci. Technol., 41, 398–417, doi:10.1080/02786820701199728. Moteki, N., and Y. Kondo (2008), Method to measure time‐dependent scattering cross sections of particles evaporating in a laser beam, J. Aerosol Sci., 39, 348–364. Moteki, N., and Y. Kondo (2010), Dependence of laser‐induced incandescence on physical properties of black carbon aerosols: Measurements and theoretical interpretation, Aerosol Sci. Technol., 44, 663–675, doi:10.1080/02786826.2010.484450. Ohara, T., H. Akimoto, J. Kurokawa, N. Horii, K. Yamaji, X. Yan, and T. Hayasaka (2007), An Asian emission inventory of anthropogenic emission sources for the period 1980– 2020, Atmos. Chem. Phys., 7, 4419–4444, doi:10.5194/acp-7-4419-2007. Oshima, N., et al. (2004), Asian chemical outflow to the Pacific in late spring observed during the PEACE‐B aircraft mission, J. Geophys. Res., 109, D23S05, doi:10.1029/2004JD004976. Oshima, N., M. Koike, Y. Zhang, Y. Kondo, N. Moteki, N. Takegawa, and Y. Miyazaki (2009a), Aging of black carbon in outflow from anthropogenic sources using a mixing state resolved model: Model development and evaluation, J. Geophys. Res., 114, D06210, doi:10.1029/ 2008JD010680. Oshima, N., M. Koike, Y. Zhang, and Y. Kondo (2009b), Aging of black carbon in outflow from anthropogenic sources using a mixing state resolved model: 2. Aerosol optical properties and cloud condensation nuclei activities, J. Geophys. Res., 114, D18202, doi:10.1029/ 2008JD011681. Ramanathan, V., P. J. Crutzen, J. T. Kiehl, and D. Rosenfeld (2001), Aerosols, climate, and the hydrological cycle, Science, 294, 2119–2124, doi:10.1126/science.1064034. Ramanathan, V., M. V. Ramana, G. Roberts, D. Kim, C. Corrigan, C. Chung, and D. Winker (2007), Warming trends in Asia amplified by brown cloud solar absorption, Nature, 448, 575–578, doi:10.1038/nature06019. Schwarz, J. P., et al. (2006), Single‐particle measurement of midlatitude black carbon and light‐scattering aerosols from the boundary layer to the lower stratosphere, J. Geophys. Res., 111, D16207, doi:10.1029/ 2006JD007076. Seinfeld, J. H., and S. N. Pandis (2006), Atmospheric Chemistry and Physics: From Air Pollution to Climate Change, Wiley‐Interscience, New York. Skamarock, W. C., J. B. Klemp, J. Dudhia, D. O. Gill, D. M. Barker, W. Wang, and J. G. Powers (2005), A description of the advanced research WRF version 2, Tech. Note NCAR/TN‐468+STR, Natl. Cent. for Atmos. Res., Boulder, Colo. [Available at http://wrf‐model.org/ wrfadmin/publications.php.] Streets, D. G., et al. (2003), An inventory of gaseous and primary aerosol emissions in Asia in the year 2000, J. Geophys. Res., 108(D21), 8809, doi:10.1029/2002JD003093. Suthawaree, J., S. Kato, A. Takami, H. Kadena, M. Toguchi, K. Yogi, S. Hatakeyama, and Y. Kajii (2008), Observation of ozone and carbon monoxide at Cape Hedo, Japan: Seasonal variation and influence of long‐range transport, Atmos. Environ., 42, 2971–2981, doi:10.1016/j. atmosenv.2007.12.053. Uno, I., G. R. Carmichael, D. G. Streets, S. Satake, T. Takemura, J.‐H. Woo, M. Uematsu, and S. Ohata (2003), Analysis of surface black carbon distributions during ACE‐Asia using a regional‐scale aerosol model, J. Geophys. Res., 108(D23), 8636, doi:10.1029/2002JD003252. Zhang, Q., et al. (2009), Asian emissions in 2006 for the NASA INTEX‐B mission, Atmos. Chem. Phys., 9, 5131–5153, doi:10.5194/acp-9-51312009. Y. Kajii and S. Kato, Division of Applied Chemistry, Faculty of Urban Environmental Sciences, Tokyo Metropolitan University, 1‐1 Minami‐ Ohsawa, Tokyo 192‐0397, Japan. ([email protected]; [email protected]) M. Kajino and N. Oshima, Meteorological Research Institute, 1‐1 Nagamine, Tsukuba 305‐0052, Japan. (kajino@mri‐jma.go.jp; oshima@ mri‐jma.go.jp) Y. Kondo, N. Moteki, and R. L. Verma, Department of Earth and Planetary Science, Graduate School of Science, University of Tokyo, Hongo 7‐3‐1, Tokyo 113‐0033, Japan. ([email protected]‐tokyo.ac.jp; [email protected]‐tokyo.ac.jp; [email protected]‐tokyo.ac.jp) R. Mikami and N. Takegawa, Research Center for Advanced Science and Technology, University of Tokyo, 4‐6‐1 Komaba, Meguro‐ku, Tokyo 153‐8904, Japan. ([email protected]‐tokyo.ac.jp; takegawa@atmos. rcast.u‐tokyo.ac.jp) A. Takami, National Institute for Environmental Studies, 16‐2 Onogawa, Tsukuba City, Ibaraki 305‐8506, Japan. ([email protected])

14 of 14