Temporal variations of elemental carbon in Tokyo



[1] Mass concentrations of elemental carbon (EC) in fine mode and mixing ratios of carbon monoxide (CO) were measured at the University of Tokyo campus in Tokyo in different seasons in 2003–2005. Measurements of EC were made using a semicontinuous thermal-optical analyzer. The mass concentrations of nonvolatile aerosol measured by the calibrated scanning mobility particle sizer combined with a heated inlet agreed with the independent EC measurements with a systematic difference of about 4%, demonstrating that the mass concentrations of nonvolatile aerosol well represent those for EC. A majority of the nonvolatile aerosol and therefore EC mass concentration was in volume equivalent diameters between 50 and 200 nm, peaking at around 130 nm. The correlation of EC and CO was generally compact throughout the measurement period because of the similarity in sources. The slope of the EC-CO correlation (ΔEC/ΔCO) is therefore a useful parameter in validating EC emission inventories. The EC concentration and ΔEC/ΔCO showed distinct diurnal variation. On weekdays, EC and ΔEC/ΔCO reached maximum values of about 3 μg m−3 and 9 ng m−3/parts per billion by volume, respectively, in the early morning (0400–0800 local time), when the traffic density of heavy-duty trucks with diesel engines was highest. In addition, these values were lower by a factor of 2 on Sundays. The heavy truck traffic showed similar diurnal and weekday/weekend variations, indicating that exhaust from diesel engines is an important source of EC. Monthly mean ΔEC/ΔCO showed a seasonal variation, reaching broad maximum values in spring-autumn and reaching minimum values in midwinter, following the seasonal variation in temperature, as observed in Maryland, United States (Chen et al., 2001). This temperature dependence is likely due to the temperature dependence of EC emissions from diesel engines on intake air temperature. More stringent regulation of emissions of particles from diesel cars started in the Tokyo Metropolitan Area in October 2003. The ΔEC/ΔCO values did not change, however, exceeding the natural variability (10%) after 1 year from the start of the new regulations, when the temperature dependence is taken into account. This indicates that the regulation of particle emissions in the Tokyo Metropolitan Area was not effective in reducing the EC concentrations after 1 year.

1. Introduction

[2] Elemental carbon (EC), sometimes referred to as black carbon (BC), is produced by incomplete combustion of carbon-based fuels, principally fossil fuels used in transportation, heating, power generation, and industrial processes; wood for residential heating; and agricultural biomass. Natural wildfires at temperate and boreal latitudes are also significant EC sources [e.g., Streets et al., 2003; Bond et al., 2004]. Soot particles have been shown to be agglomerates of spherical EC particles with diameters of 15–50 nm coated with organic carbon (OC) [Seinfeld and Pandis, 1998; Park et al., 2004a, 2004b; Slowik et al., 2004]. A significant fraction of EC particles has OC coatings.

[3] EC also comprises a significant portion of nanoparticles, which are harmful to human health at high concentrations in megacities. EC particles act as carriers of the organic compounds, especially polycyclic aromatic hydrocarbons (PAHs), that can be allergens or carcinogens [Lighty et al., 2000]. EC has also been identified as making important contributions to the radiative heating of the atmosphere [Haywood et al., 1997; Myhre et al., 1998; Jacobson, 2001, 2002]. It can also contribute to climate forcing by changing snow and ice albedos [Hansen and Nazarenko, 2004]. Therefore reductions of EC are beneficial to air quality and global climate [Hansen and Sato, 2001].

[4] Emissions of EC inside megacities are important for local air quality and that in the surrounding regions, considering the magnitude of emissions. In addition, EC transported from megacities can influence regional and global climate. Emissions of EC in megacities are relatively well known in that the major sources are general automotive traffic and industrial fuel combustion, as compared with EC emissions from less well characterized sources, such as combustion of biofuels and biomass, which are much more variable [e.g., Streets et al., 2003]. The behavior of EC has not been fully characterized, however, even in urban areas, including those in Japan, mainly because of a paucity of sufficiently time-resolved EC data for extended time periods. Uncertainties in the estimates of EC emissions in these areas, including those from road traffic, are still significant [e.g., Schaap et al., 2004], partly because of the lack of detailed EC information. Obtaining systematic and accurate EC data is an important step in characterizing EC emissions in urban regions. Measurements of EC were made in Tokyo in 2003–2004, through the series of Integrated Measurement Program for Aerosol and Oxidant Chemistry in Tokyo (IMPACT) campaigns. The IMPACT campaigns were conducted within the framework of the International Global Atmospheric Chemistry Project (IGAC), Mega-Cities: Asia. Here we show that long-term, high-accuracy measurements of EC and its tracers are useful in understanding emissions and transport of EC in urban areas.

2. Measurements

[5] EC concentrations were measured along with other tracers, including CO and CO2 near the urban center of Tokyo during periods from May 2003 to February 2005. Air samples were taken about 20 m above ground level from a building at the Research Center of Advanced Science and Technology (RCAST) campus of the University of Tokyo (35.66°N, 139.66°E) in Japan. The configurations of the instruments used for the ambient measurements and calibration are shown in Figure 1. RCAST is located about 10 km west of the Tokyo Bay coastline and is near the southeastern edge of the Kanto Plain, composed of prefectures labeled 1, 2, 3, 6, and, 7 and southern parts of prefectures 4 and 5 in Figure 2. The population of Tokyo (labeled 2 in Figure 2) is 12 million, and the population in the whole Tokyo Metropolitan Area (TMA labeled 1, 2, 3, and 7 in Figure 2) is 41 million. The 10-km-resolution CO emission rates, estimated by the Japanese Ministry of Environment (JMOE), are also shown in Figure 2. RCAST is located in the highest-CO-emission regions in the TMA.

Figure 1.

(a) Configurations of the instruments used for the ambient air measurements. A gas-phase organics denuder is used to remove gas-phase organic carbon species. Heaters are used to control the air temperature up to 400°C. (b) Test and calibration systems for nonvolatile aerosol.

Figure 2.

Locations of the sampling points at RCAST (diamond), Iogi tunnel (triangle), Ring Road 8 (thick line), Tsukuba city (square), and Kisai city (circle). Prefectures numbered 1–7 belong to the Tokyo metropolitan area (Kanto area) by administrative definition. Shaded areas represent the CO emission rates. Tokyo (2) and Kanagawa (1) prefectures constitute major CO emission regions.

[6] EC mass concentrations were measured every hour with a semicontinuous EC and organic carbon (OC) analyzer manufactured by Sunset Laboratory Inc. (Beaverton, Oregon, United States), using a thermal-optical method. The inlet for air sampling was equipped with a PM1 (1-μm-diameter cutoff size) cyclone (Model URG-2000-30EHB, URG Inc., United States) and a gas-phase organics denuder (Sunset Laboratory Inc.) to remove gas-phase organic carbon species. Ambient aerosol, including EC and OC particles, was collected on a quartz fiber filter with an effective collection area of 0.69 cm2 for 45 min at a flow rate of 133 cm3 s−1. Collected aerosols were analyzed on the basis of the thermal-optical transmittance method in 15 min. The total number of days of these observations was 277.

[7] For this analysis, we used the temperature protocol proposed by the National Institute for Occupational Safety and Health (NIOSH) [Birch and Cary, 1996]. Part of the OC is pyrolytically converted to EC during the heating of this filter in an oxygen-free helium atmosphere in four temperature steps up to 870°C. After completion of the oxygen-free heating stages, the filter is heated up to 900°C in the presence of 2% oxygen. During this phase, both the original EC and that produced by pyrolysis of OC are converted to CO2, which is measured by a nondispersive infrared (NDIR) CO2 detector. The correction for the pyrolytic conversion of OC to EC was performed by monitoring the transmittance of a pulsed diode laser beam at 670 nm through a quartz fiber filter during the sample analysis. More detailed descriptions of this EC-OC analyzer are given elsewhere [Birch and Cary, 1996; Bae et al. 2004]. The sensitivity of the EC-OC analyzer was measured by a routine addition of a known amount of CH4 every hour. Overall calibration was made by adding a known amount of sucrose (86 μg) onto the filters every week. The uncertainty of this calibration was estimated to be 3% from the ratio of the sensitivities determined from sucrose calibration and routine CH4 calibration. The zero level was determined by removing aerosol in the sample air using a filter once per week. The detection limit was determined to be 0.4 μg m−3 from the variability (2σ) of the zero level.

[8] The uncertainties associated with the aerosol sampling by the semicontinuous instrument were assessed by comparison with filter samples taken at RCAST during the IMPACT campaign conducted in July–August 2004, followed by EC-OC analysis in the laboratory. The systematic uncertainty and the detection limit of the semicontinuous EC measurement were estimated to be 4% and 0.48 μg m−3 (2σ), respectively. A similar analysis was made by comparing semicontinuous and filter-based measurements in East St. Louis, Illinois, United States [Bae et al., 2004]. The systematic difference of the two EC measurements was estimated to be 5% for an average EC concentration of 0.70 μg m−3.

[9] In addition, a comparison with the temperature protocol proposed by the Interagency Monitoring of Protected Visual Environments (IMPROVE) of the Desert Research Institute was performed. The mass concentrations of EC based on the NIOSH protocol were found to be lower by approximately 21% than those based on the IMPROVE protocol, under typical ambient conditions in Tokyo. This is considered to be a measure of the uncertainty associated with the separation of EC and OC. The overall accuracy of the present EC measurement is estimated to be 22% from uncertainties associated with the sensitivity calibration, aerosol sampling, and temperature protocol.

[10] Nonvolatile aerosol has been found to contain a significant portion of EC in outflows from the Asian continent [e.g., Clarke et al., 2004], although some contribution of dust was indicated, depending on the meteorological conditions. EC is expected to constitute a much larger part of nonvolatile aerosol in Tokyo, where concentrations of dust should be much smaller than those of EC. We have measured size distributions of PM1 nonvolatile aerosol using a heated inlet system, which is described below, and a scanning mobility particle sizer (SMPS; TSI Inc., Model 3034, St. Paul, Minnesota, United States), which is composed of a 1-μm diameter cut off size impactor, differential mobility analyzer (DMA), and condensation nuclei counter (CNC) (Figure 1a). About six sets of size distributions at aerosol mobility diameters (Dm) between 10 and 487 nm were obtained every hour. Dm measured by the SMPS was calibrated using standard polystyrene latex (PSL) particles (Duke Scientific Corp., Palo Alto, California, United States) with diameters of 102, 199, and 404 nm. The PSL particles were nebulized in purified water. The uncertainty of the PSL particle diameter, traceable to the National Institute of Standards and Technology (NIST), was certified to be 1–3% by the manufacturer. The uncertainty of the number size distribution (dN/dlogDm) measured by this SMPS was estimated to be about 10%, from a comparison with two other SMPS systems (TSI Inc., Model 3936). A 21-cm-long section of the inlet, made of 7-mm inner diameter stainless steel tubing, was heated to 400°C, using a method similar to that used by Clarke et al. [2004]. The residence time of sampled air in the heated section was 0.48 s. In order to confirm evaporation of volatile components, the chemical composition of ambient aerosol was measured using an Aerodyne aerosol mass spectrometer (AMS) [e.g., Takegawa et al., 2005] by varing inlet temperatures between 20° and 400°C in 20°C increments, when the ambient PM1 aerosol concentrations were stable at about 39 ± 3 μg m−3. At 400°C, about 97 ± 5% of inorganic (SO42−, NO3, Cl, and NH4+) and organic components were estimated to have evaporated (Y. Komazaki, unpublished data, 2004) from the measured decreases in their mass concentrations by heating. This result confirms high evaporation efficiency at this temperature, although some of the organic compounds might have pyrolized. The possible error in the estimate of the EC mass concentration by the pyrolized organic compounds is estimated to be less than about 10%, as discussed in section 4.1. Loss of EC particles through the heated inlet was measured by comparison with measurements at ambient temperature using glassy carbon particles. The loss was measured to be 0, 2, and 5% with an uncertainty of about ±5% for Dm = 100, 60, and 30 nm, respectively. No detectable loss was observed at larger diameters.

[11] A relationship between mass and Dm was calibrated for nonvolatile particles in ambient air at the National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba (Figure 2), Ibaraki, Japan, using an aerosol particle mass analyzer (APM) [Ehara et al., 1996]. The uncertainty of the aerosol mass measurement by the DMA-APM system is estimated to be about 5% [McMurry et al., 2002]. The mass of ambient aerosol after being heated to 400°C was measured by the APM mounted downstream of the DMA, as shown in Figure 1b. This aerosol mass calibration system is similar to that used for diesel exhaust particles by Park et al. [2003]. The difference is that we used ambient air instead of direct diesel exhaust. The average mass (M) of the nonvolatile aerosol measured by this system is shown as a function of Dm in Figure 3. We define the volume equivalent diameter Dve as

equation image
Figure 3.

Average mass of nonvolatile aerosol versus the mobility diameter (Dm) measured by the heater-DMA-APM system shown in Figure 1b.

[12] Here, ρ is the density of EC. Park et al. [2004b] showed that the density of diesel soot particles was 1.77 g cm−3, after they were heated to 300°C to remove volatile organic compounds. Using this reported density and equation (1), the relationship between Dm and Dve can be derived from the Dm-M calibration data. Once dN/dlogDm is measured, it can be converted to mass size distribution dM/dlogDm using the Dm-Dve relationship.

[13] CO concentration was measured by an NDIR gas analyzer (Thermo Electron Inc., Model 48C, United States) with an integration time of 1 min. In order to reduce the interference from water vapor, the sample air for the CO instrument was dried (dew point < 0°C) using a Nafion dryer (Perma-Pure, Inc., United States). The background (zero) signals were routinely measured every 1 or 2 hours by supplying purified (dry) air into the sample line. The zero signals measured using the purified air agreed well with those measured using a Hopcalite scrubber (CO removal catalyst) after passing the sample air through the Nafion dryer. The uncertainty in determining the zero levels was estimated to be about 30 ppbv (2σ). Calibrations were performed once every 3 weeks by supplying a CO standard (5 parts per million by volume (ppmv) CO in air) manufactured by the Nissan-Tanaka Corp., Japan. The stability of the instrument sensitivity (i.e., stability of the calibration signals) was about 1.5% (1σ). The overall precision and accuracy of the 1-min CO data were estimated to be 4 ppbv and 20 ppbv, respectively, at a CO mixing ratio of 400 ppbv.

[14] CO2 was measured using an NDIR-based instrument with an integration time of 10 s (Model LI7000, Li-Cor, Inc., United States). The sample line for the CO2 instrument was the same as that for the CO instrument, so as to reduce the interference from H2O. A detailed description of the calibration of the CO2 instrument is given by Takegawa et al. [2006]. The precision and accuracy of the 10-s CO2 data were estimated to be 0.3 ppmv and 2 ppmv, respectively, at the CO2 mixing ratio of 400 ppmv.

3. Meteorological Fields

[15] Meteorological conditions for the measurements are described briefly. From late spring to midsummer, a sea-land breeze circulation was often driven by the heating and cooling of the Kanto Plain during daytime and nighttime, respectively. On clear, calm days, southerlies typically dominated during the daytime, and air over the ocean was brought to the observational site and further transported northward. From midnight to early morning, weak northerlies dominated and air was transported from northern Kanto Plain. The typical altitude of the top of the boundary layer was about 1200–1800 m during midday and 300–500 m during midnight in July 2004, according to low-powered laser radar (Ceilometer TXK-6, Meisei Electric. Co. Ltd.) observations at the National Institute of Information and Communications Technology (NICT) in Koganei city, 20 km west of RCAST (M. Yasui, unpublished data, 2004). In winter, the sea-land breeze circulation did not develop. Instead, the northwesterlies associated with the strong Siberian high-pressure system generally dominated over the Japanese islands. As a result, air was transported to the observational site primarily from the northwest, after passing over the northern part of the Kanto Plain.

4. Results and Discussion

4.1. Size Distribution

[16] Size distributions of nonvolatile aerosols measured by the SMPS with a heated inlet were measured between 24 July and 15 August 2004. The average mass size distributions for 0700, 1400, and 2200 local time (LT) are shown in Figure 4 as a function of Dm and Dve. Nonvolatile aerosol had a unimodal size distribution, and a majority of the mass concentration was in the size range of Dve = 50–200 nm (Dm = 50–480 nm), peaking at around Dve = 130 nm. Changes in the peak Dve were small, ranging over 130 ± 20 nm during the whole of the observational period. These mass size distributions were integrated to derive total mass concentrations of nonvolatile aerosol (SMPS mass), assuming smooth extrapolation beyond 480 nm. For this estimation, lognormal size distributions were derived by fitting to the observed values. The error associated with this extrapolation is estimated to be about 1%. Loss of EC through the heated inlet for Dm < 100 nm introduces errors less than 1%. The nonvolatile SMPS mass concentrations are compared with the PM1 EC concentrations measured simultaneously with the EC-OC analyzer in Figures 5a (time series) and 5b (scatterplot). The SMPS mass concentrations agree very well with the EC concentrations with a slope of 0.96 and r2 = 0.88. The systematic difference of about 4% between the two measurements is within the combined accuracy of the EC (22%) and SMPS (about 10%) measurements discussed in section 2, demonstrating that the mass concentrations of nonvolatile aerosol are representative of those for EC. From this result, we can conclude that size distributions of nonvolatile aerosols (Figure 4) represent those for EC particles. Particles with 50 < Dve < 200 nm constituted a majority of the total EC mass in Tokyo. Mass size distributions of EC particles in urban air were measured previously by using impactors in the Los Angeles basin, United States [Venkataraman and Friedlander, 1994], Pittsburgh, United States [Cabada et al., 2004], and Helsinki, Finland [Viidanoja et al., 2002]. In most cases, a majority of EC mass was in the submicron range, basically in agreement with the present measurements. However, it should be noted that the diameters of particles collected by impaction in the United States and Finland are those for internally mixed aerosols, which contain volatile species in addition to the EC component. By contrast, the present data provide mass size distributions of the EC component in internally mixed aerosols. In addition, the present size distributions were obtained at higher size and time resolutions than the previous studies. Therefore the mass size distribution is not smeared out by integration, and a more accurate distribution was obtained in this study.

Figure 4.

Average mass (M) size distribution of nonvolatile aerosol measured as a function of Dm and volume equivalent diameter (Dve) at 0700, 1400, and 2200 LT in July–August 2004. Bars indicate the 1σ values.

Figure 5.

(a) Time series plot of PM1 EC mass concentration measured by the Sunset EC-OC analyzer and the mass concentration of nonvolatile aerosol (SMPS mass concentration) for the data obtained in July–August 2004. (b) Correlation between the EC and SMPS mass concentrations shown in Figure 5a.

4.2. EC-CO Correlation

[17] The EC concentrations observed during July–August 2004 are plotted versus CO concentration in Figure 6. EC was generally correlated well with CO during this and other periods (r2 = 0.62). Correlation coefficients improve to r2 = 0.70 when the data sets are selected by local time because of a diurnal variation of the slope, as discussed in section 4.3. EC is known to be correlated with CO in other urban areas as well [Chen et al., 2001; Baumgardner et al., 2002; Park et al., 2005], because both species are released by incomplete combustion of carbon-based fuels, especially fossil fuel in the case of the present study. According to the emission inventory by JMOE shown in Figure 2, 80% of CO is emitted from vehicles (cars and trucks) and the rest from industry and commercial facilities in the TMA, suggesting that EC is also largely emitted from vehicles. It should be noted that emission ratios of EC/CO are very different depending on the types of emission sources, even in the same region [Bond et al., 2004]. For example, EC/CO emission ratios are known to be much higher for diesel engines as compared with those for gasoline engines. Considering the high traffic densities of both diesel and gasoline vehicles, the slope of the EC-CO correlation (ΔEC/ΔCO) in this study is considered to have resulted from the mixing of air impacted by emissions with different EC/CO emission ratios in the TMA. The uniformity of air impacted by these sources is evaluated by hourly mean CO concentrations observed at monitoring stations in the TMA in May 2003, as shown in Figure 7. The recordings of the CO data at the monitoring sites were made with a resolution of 100 ppbv. CO concentrations measured at 7 locations 3–22 km distant from RCAST were rather uniform and showed highly correlated variations, with r2 = 0.44–0.88 and an average r2 = 0.77. Although the EC measurement was made only at RCAST, CO and EC values measured here represent those for mixtures of air impacted by different sources, rather than those strongly influenced by a few individual sources in highly localized areas. ΔEC/ΔCO is better suited for detecting average changes in EC emissions in the TMA than the absolute value of EC alone, because changes in ΔEC/ΔCO through mixing with background air were much less than those in EC, as discussed in section 4.3.1. In order to derive ΔEC/ΔCO, we determined background values of EC and CO (EC0, CO0), defined as the median of the values below the 3-σ range, namely the 1.25th percentile of the EC and CO values for each month. ΔEC/ΔCO was calculated in three different ways: (1) the slope of the least squares fitting of the EC-CO scatterplots; (2) the slope of the least squares fitting of the (EC − EC0) − (CO − CO0) scatterplots, assuming a zero offset; and (3) median values of the slope of each data point, i.e., (EC − EC0)/(CO − CO0). For this calculation, data with CO − CO0 < 100 ppbv were excluded. The three ΔEC/ΔCO values generally agree well, and the median values within these three estimates were used for the present study. In most cases ΔEC/ΔCO derived by the method 2 became the median values. EC0 and CO0 are the values for air masses flowing into the TMA and thus can depend on various factors, including season and day of week. They are generally lower in the summer because of the southerly flow of relatively clean air from the Pacific Ocean and higher in the other seasons, especially winter, influenced by the outflows from the Asian continent. The seasonal variations of EC0 and CO0 are largely taken into account by using their monthly mean values. Variability with timescales shorter than a month can lead to uncertainty in ΔEC/ΔCO.

Figure 6.

EC-CO correlation using all the data obtained in July–August 2004. The correlations obtained at 0400–0800 LT and 2200–0200 LT are also shown. The least squares fit is shown for reference.

Figure 7.

CO temporal variations measured at RCAST and two monitoring stations at Setagaya (3 km from RCAST) and Arakawa (14 km from RCAST) in Tokyo in May 2003. The recordings of the CO data at the monitoring sites were made with a resolution of 100 ppbv.

[18] Emission ratios of trace species, including EC and CO, versus CO2 are often used to relate them to fuels burned [e.g., Andreae and Merlet, 2001; Streets et al., 2003]. At RCAST, it has been found that CO2 was tightly correlated with CO in autumn and winter (r2 = 0.91–0.92) [Takegawa et al., 2006] and CO2 is a useful parameter for comparison with the emission inventories discussed in section 4.5.1. The slopes of the CO2-CO correlation (ΔCO/ΔCO2 values) in autumn and winter were 11.6 and 10.7 ppbv/ppmv, respectively.

4.3. Diurnal Variation of EC and ΔEC/ΔCO

4.3.1. Dynamical Effects

[19] Figure 8 shows the diurnal variations of the median values of EC, CO, ΔEC/ΔCO, and wind speed for four seasons. For winter, the values for two periods (February 2004; December 2004 to February 2005) are shown separately. The EC–wind speed correlation is also shown. The diurnal variations of the median values of EC, CO, and ΔEC/ΔCO for the entire observational period are shown in Figure 9, for weekdays (Monday–Friday), Saturday, and Sunday, for the consideration of possible differences in EC emissions, depending on the day of the week. The EC concentration did not show a significant dependence on the day of the week Monday–Saturday, with an average value of 1.9 μg m−3, and it was about 40% lower than this value on Sunday, similar to that observed in Atlanta, Georgia, United States [Lim and Turpin, 2002]. These variations are interpreted in terms of diurnal patterns of EC emissions and transport below.

Figure 8.

Diurnal variations of the median values of EC, CO, ΔEC/ΔCO, and wind speed for each season. The blue open and solid circles represent the values for February 2004 and for December 2004 to February 2005, respectively. Bars for EC, CO, and wind speed, representing the 1σ values, are shown only for spring for the sake of clarity of the figures. The ranges of the bars are similar for the other seasons. Bars for ΔEC/ΔCO indicate the range of the values derived by three methods (see text). The correlations of EC and CO with wind speed are also shown.

Figure 9.

Diurnal variations of median values of ΔEC/ΔCO for weekdays, Saturdays, and Sundays. Bars indicate the range of the values derived by the three methods.

[20] As seen in Figure 8, EC started to increase rapidly around 0400 LT, reaching maximum values of about 2–3 μg m−3 around 0700 LT, because of increases in traffic, especially diesel trucks, as discussed in more detail below. It then decreased throughout the day, in summer and to a lesser degree in spring. The decrease is less pronounced in autumn and winter. The average nighttime EC values were about 1.5 μg m−3, half of the early morning peak values. The decrease during the daytime was also seen for CO in summer and spring. During these seasons, southerly wind speed increased from morning to afternoon, associated with the development of the sea breeze. The period of EC decrease corresponds to that of the increase in wind speed, as shown in Figure 8. Increases in the horizontal wind speed allow for a shorter time for air to accumulate emitted EC before reaching RCAST. In addition, the boundary layer height increased in the early morning in summer, as described earlier, indicating that vertical mixing became more effective during daytime. Active venting of air during daytime also dilutes higher EC and CO concentrations near the surface by mixing with overlying air with lower concentrations.

[21] In autumn and winter, the sea-land breeze circulation did not develop and the wind speed did not show a systematic diurnal variation similar to that for summer and spring, although the wind speeds were generally higher during daytime than nighttime. As a result, the diurnal variation of EC between 0600 and 1800 LT was less pronounced, especially in winter. Despite the seasonal difference in the diurnal pattern of the local winds, EC decreased with wind speeds similarly to spring and summer (Figure 8). It should be noted that, in summertime, the timing of the increases in the wind speed in the late morning coincides with the decrease in EC emissions, as discussed in section 4.3.2. Therefore the EC–wind speed relationship in summer might naturally have been more pronounced than that for other seasons. Similar dynamical effects (wind speed and boundary layer height) on EC concentrations were observed also in the close vicinity of a high-traffic road around Paris, leading to a continued decrease in EC throughout the daytime in summer, after reaching maximum values in the early morning [Ruellan and Cachier, 2001].

[22] As seen in Figure 9, ΔEC/ΔCO reached maximum values of about 8–9 ng m−3/ppbv on weekdays around 0700 LT similarly to EC. However, in summer and spring, ΔEC/ΔCO values did not show a subsequent decrease during the daytime, in contrast to EC, as seen from Figure 8. The decrease in EC in the afternoon is associated with the decrease in CO to a large degree, resulting in the reduced decrease in ΔEC/ΔCO. After 1800 LT, ΔEC/ΔCO started to decrease, reaching minimum values of about 5 ng m−3/ppbv around midnight, irrespective of the season. Although the diurnal wind pattern was different for autumn, the diurnal patterns of ΔEC/ΔCO became similar. In winter, ΔEC/ΔCO was generally lower than the other seasons, especially in the early to late morning. The average temperature in the early morning (0400–0800 LT) in midwinter was about 3°C, which was more than 10°C lower than the other seasons. EC/CO emission ratios for vehicles are lower under the cold conditions in winter, as discussed in more detail in section 4.4.

4.3.2. EC Emissions

[23] According to the estimates of EC emissions by Streets et al. [2003], transportation and power generation constitute 42 and 27%, respectively, of EC emissions in Japan as a whole, with uncertainties of about 83%. Uncertainties in EC emission inventories with finer spatial resolutions are even larger (D. Streets, private communication, 2005). Attempts to estimate contributions of traffic EC emissions to background EC levels in Tokyo were made by the Tokyo Metropolitan Research Institute for Environmental Protection (TMRIEP) and an outline of these studies are summarized below. PM2.5 (particles smaller than 2.5 μm in diameter) and PM10 aerosols were collected from filter samples at 10 monitoring sites (5 residential and 5 roadside sites) in Tokyo between April 2001 and February 2002 and were also collected intensively at 4 monitoring sites (2 residential and 2 roadside sites) in Tokyo in March 2002 to measure concentrations of total aerosol mass, metallic elements, water-soluble ions, EC, and OC. In addition, chemical source profiles consisting of the same particulate components were obtained for steel engineering, waste incineration, heavy oil burning, vehicles, sea salt, as well as for paved road dust. The derived data on relative emissions of 7 elements (Na, K, Ca, V, Al, Mn, and EC) from the above sources were combined with the observed concentrations of these elements in a chemical mass balance (CMB) receptor model [e.g., Zheng et al., 2002] for source apportionment of EC. The results of the CMB modeling showed that the major sources of EC at general/residential sites were emissions from diesel vehicles (59–97% by mass, 86% on average). Larger contributions from diesel vehicles to the observed EC (87–95%, 95% on average) were derived for roadside areas. It should be noted, however, if EC emissions from sources other than those considered here are significant, the contributions of diesel vehicles are overestimated.

[24] We now investigate how temporal variations in traffic are reflected in the variations in ΔEC/ΔCO, considering the significant contributions of traffic. Figures 10 and 11 show one-way traffic density of different types of vehicles (passenger cars, light trucks, and heavy-duty diesel trucks) monitored in the Iogi tunnel on Ring Road 8 (marked as triangle in Figure 2) in Tokyo on Saturday, Sunday, Monday, and Tuesday during the periods 10–13 March 2001 and 8–11 November 2003 by TMRIEP (K. Ishii, unpublished data, 2001 and 2003). Ring Road 8 is one of the busiest roads in Tokyo, and the tunnel is located 8 km north-northwest of RCAST. Although the monitoring was made only for 8 days in total, the fluctuation in the traffic density was very small, indicating that these data are representative for ordinary conditions in spring and autumn. The diurnal pattern of the measured ratios of diesel trucks to total vehicles will probably be similar to those averaged for the TMA, considering the intensity of the traffic. However, because of the lack of traffic data for other roads in the TMA, our interpretation of the diurnal variation of EC in terms of the diurnal variation of diesel trucks is qualitative.

Figure 10.

Average diurnal variations of the traffic density of total vehicles measured in March 2001 and November 2003 for weekdays, Saturdays, and Sundays. Bars indicate the range of values.

Figure 11.

Average diurnal variations of the fractions of all vehicles, comprising cars, light-duty trucks, heavy-duty trucks, and all trucks, for March 2001 and November 2003 for weekdays, Saturday, and Sunday.

[25] We have roughly estimated the CO and EC emissions from gasoline cars and diesel trucks from existing data. Emission factors of EC (fg(EC) and fd(EC)) and CO (fg(CO) and fd(CO)) for gasoline and diesel vehicles respectively, in g km−1, estimated by Streets et al. [2003], are summarized in Table 1, which show that fd(EC) is much greater than fg(EC) (fd(EC)/fg(EC) = 10–160), although fg(CO) and fd(CO) are similar. From the CO measurements in the Iogi tunnel (K. Ishii, unpublished data, 2004), fg(CO)/fd(CO) was derived to be close to unity. K. Miyamoto (unpublished results, 2001) reported fg(CO)/fd(CO) = 0.5–0.9 for Tokyo. The similarity in fg(CO) and fd(CO) was also observed from measurements in the a tunnel in Osaka, Japan, where fg(CO)/fd(CO) = 0.86 [Funasaka et al., 1998] and in the tunnels in Maryland and Pennsylvania, United States, where fg(CO)/fd(CO) = 0.65–0.81 [Pierson et al., 1996]. A significantly lower ratio of fg(CO)/fd(CO) = 0.32 was reported by tests in Hong Kong [Tong et al., 2000]. Although values of fg(CO) and fd(CO) vary greatly, depending on various factors (types of cars, periods used, etc.), it is very likely that fg(CO) is comparable to fd(CO) on average in Japan. fd(EC) was observed to be much higher than fg(EC) in the tunnel in Japan [Funasaka et al., 1998] and in California, United States [Miguel et al., 1998], consistent with Streets et al. [2003].

Table 1. Emission Factors of EC and CO for Gasoline Cars and Heavy-Duty Diesel Trucksa

[26] Considering that fd(EC) ≫ fg(EC), the EC/CO emission ratio for vehicles is expressed as

equation image

where Ng and Nd are the traffic density of gasoline and diesel vehicles, respectively.

[27] Assuming that fg(CO) = fd(CO), equation (2) is approximated as

equation image

where Nt = Ng + Nd is the total traffic. If ΔEC/ΔCO reflects EC/CO emission ratios, it should vary depending on Nd/Nt. The total traffic densities of cars, light-duty trucks, and heavy-duty trucks, are shown in Figure 10. The traffic of different types of vehicles relative to total traffic is shown in Figure 11. Light-duty trucks are considered to include both diesel and gasoline vehicles. The total traffic showed rapid increases in the early morning and a decrease in the early evening. For weekdays, the relative car traffic during nighttime was comparable or higher than that during daytime and does not agree with the diurnal variation of ΔEC/ΔCO. The fraction of heavy-duty trucks peaked at around 0400 LT, 2 hours prior to the morning peak of the total traffic. A significant portion of these trucks arrive in the TMA by this time to complete freight shipments from outside the TMA undisturbed by the morning high traffic density. Only the diurnal pattern of traffic densities of heavy-duty trucks or total trucks agrees with that of ΔEC/ΔCO. These traffic densities increased by 60–100% from midnight to the early morning, in agreement with the corresponding change in ΔEC/ΔCO.

[28] In addition to the diurnal pattern on weekdays, the changes in relative traffic densities on Saturday and Sunday from those on weekdays can be compared with the corresponding changes in ΔEC/ΔCO. Again, only the traffic densities of heavy-duty trucks or total trucks were lowest on Sundays. To be more precise, the truck fraction for Saturday and Sunday were normalized to those for weekdays and are compared with the corresponding ratios of ΔEC/ΔCO in Figure 12. Diurnally averaged Saturday/weekday and Sunday/weekday ratios for all tucks were 0.80 and 0.63, which are in reasonable agreement with the ratios of 0.92 and 0.69 for ΔEC/ΔCO, respectively. If the traffic of heavy-duty trucks is used for this comparison, the Saturday/weekday and Sunday/weekday ratios are significantly lower, 0.56 and 0.31, respectively. Therefore, for Sundays at least, considering some contribution of light-duty trucks to EC emissions in addition to the heavy-duty trucks improves agreement. However, precise data of emission factors and traffic density of diesel light-duty trucks are necessary to assess their contributions more quantitatively.

Figure 12.

(left) Truck fraction ratios for Saturday and Sunday, normalized to those for weekdays. (right) Ratios of ΔEC/ΔCO for Saturday and Sunday normalized to those for weekdays.

[29] A shift of 1–2 hours between the peaks of the traffic of heavy-duty trucks and ΔEC/ΔCO reflect delocalized sources of EC. EC emitted at different locations are accumulated and mixed during transport prior to sampling at RCAST. A delay in the ΔEC/ΔCO peak is reasonable considering the time required to accumulate emitted EC. In addition, the diurnal pattern of the traffic of heavy-duty trucks shown in Figure 11 may be somewhat different at different locations in Tokyo, although similar in general. These factors can lead to a shift in the peak times of ΔEC/ΔCO and the traffic on Ring Road 8. Despite this shift, ΔEC/ΔCO generally followed the diurnal pattern of the traffic both for weekdays and Sunday. Off-road vehicles equipped with diesel engines are unlikely to be important sources of EC in the early morning, because they are generally operational from 0800 LT or later hours throughout the daytime (working hours) in Tokyo. In addition, an increase in EC should occur at even later hours, considering the time required for accumulation. From these analyses, it is concluded that emissions from heavy-duty diesel trucks are likely the major sources of EC in Tokyo. Some contributions from light-duty trucks are also likely.

4.3.3. Scavenging of EC by Rain (Rainout)

[30] While EC is hydrophobic immediately after emission, it gradually becomes hydrophilic after being coated by inorganic and organic compounds. The time constant for aging by sulfate under daytime conditions during summer is estimated to be about 8 hours in regions close to sources, for example, by mesoscale model calculations, although the model predicts a significant variability [Riemer et al., 2004]. Aged EC is removed by uptake by clouds, followed by precipitation. However, cloud formation was not frequent in the boundary layer. Therefore removal of EC by cloud processing should not be important for EC near the surface in Tokyo. EC is also scavenged by collisions with and adsorption onto raindrops (rainout). However, the time constant for removal of submicron aerosol by falling raindrops is estimated to be longer than 4 days [Seinfeld and Pandis, 1998], suggesting that precipitation scavenging of EC is not important. Therefore it is likely that ΔEC/ΔCO observed at RCAST is considered to be mainly controlled by emissions, mixing, and transport out of Tokyo. In order to confirm this, the diurnal variation of ΔEC/ΔCO on rainy days was compared with the average values. Rainy days were defined here as days when the amount of precipitation exceeded 8 mm. In total, 14 rainy days were identified between May and October in 2003 and 2004. The diurnal variation of ΔEC/ΔCO on rainy days agreed well with the average values, to within the range of the variability (bars in Figure 9). The median value during 24 hours on rainy days differed by only 0.1 ng m−3/ppbv (2%) from the value for the entire data set, confirming that scavenging of EC by falling raindrops had little effect on the average EC diurnal variation.

4.4. Seasonal Variation of EC and ΔEC/ΔCO

[31] We now investigate possible seasonal variations of EC and ΔEC/ΔCO. Figure 13 shows monthly median values of the EC and CO 24-hour average concentrations in 2003–2005. It is clearly seen that changes in EC corresponded well to those in CO, indicating that transport processes are important factors in controlling median EC concentrations, as discussed in section 4.3.1. Again, ΔEC/ΔCO is more appropriate in detecting possible seasonal variations in EC emissions than EC itself. The median EC and CO values during the entire 2003–2005 period were 1.8 ± 1.8 μg m−3 and 368 ± 274 ppbv, respectively.

Figure 13.

Monthly median values of EC (solid circles) and CO (open circles) for all the data obtained.

[32] The ΔEC/ΔCO values for the data obtained during 0400–0800 LT are shown in Figure 14. In Figure 14, the median temperatures during this local time are also plotted for comparison. These values represent the highest values during the course of the day, which are most strongly influenced by EC emissions from traffic in Tokyo, as discussed in section 4.3.2. The lowest ΔEC/ΔCO values at 2200–0200 LT and the diurnal average for the whole period are also summarized in Table 2 for comparison. The ΔEC/ΔCO ratio in the early morning ranged over 4.6–9.3 ng m−3/ppbv, with an average of 7.2 ng m−3/ppbv during this period. It reached broad maximum values between spring and autumn and minimum values in midwinter. The median temperature showed a similar seasonal pattern. To be more quantitative, the median ΔEC/ΔCO values in the early morning observed in Tokyo are plotted versus median temperature (T) in Figure 15. The August 2004 data (open circle) deviated significantly from the other data points. This data was persistently influenced by southerly maritime air associated with a high-pressure system located east of Tokyo during this period. The ΔEC/ΔCO value showed a systematic increase with temperature between 2° and 25°C, and the linear regression line, excluding the August 2004 data, is expressed as

equation image

The ΔCO/ΔCO2 values in autumn and winter showed little temperature dependence (11.6 and 10.7 ppbv/ppmv, respectively), as discussed in section 4.2., indicating that the temperature dependence of the CO emission factor is small. This in turn indicates that changes in EC emission factor caused the changes in ΔEC/ΔCO.

Figure 14.

Monthly median values of ΔEC/ΔCO for the early morning (0400–0800 LT). Bars indicate the range of values derived from three different methods (see text). The r2 values are also shown for reference.

Figure 15.

Monthly median values of ΔEC/ΔCO for the early morning plotted versus temperature (T). The August 2004 data are shown as an open circle. The linear regression line was derived excluding the August 2004 data, which was persistently influenced by southerly maritime air associated with a high-pressure system located east of Tokyo during this period.

Table 2. ΔEC/ΔCO for Tokyo at Different Local Times for the Whole Period
Local TimeΔEC/ΔCOaTbnb
  • a

    Units are ng m−3/ppbv.

  • b

    T and n represent the median temperature and number of data points, respectively.

0400–08007.2 (±2.6)16.7 (±9.6)800
2200–02004.3 (±1.7)17.6 (±9.0)840
0000–23005.7 (±1.4)18.6 (±9.0)3860

[33] The temperature dependence of ΔEC/ΔCO was also observed at a suburban site in Maryland, United States from observations between July 1999 and July 2000 by Chen et al. [2001]. This temperature dependence has been ascribed to corresponding changes in intake air density for diesel vehicles. In Maryland, ΔEC/ΔCO showed a strong increase above 15°C. By contrast, the increase in ΔEC/ΔCO was more uniform in Tokyo over the range of 2°–25°C. The dependence of ΔEC/ΔCO on temperature was lower in Tokyo: 0.17 ng m−3/ppbv/°C for 2°–25°C in Tokyo versus 0.4 ng m−3/ppbv/°C for 15°–25°C in Maryland. Despite these differences, the temperature dependence of the EC emissions from heavy-duty diesel engines is very likely a common feature. The differences in the ΔEC/ΔCO–temperature relation can depend on various factors, including differences in the design and performance of diesel engines, the quality of fuels used, and the driving conditions in the two countries.

[34] New regulations that restrict the use of diesel vehicles (buses, trucks, and special category vehicles) with high aerosol emissions were adopted in the TMA (prefectures labeled as 1, 2, 3, and 7) on 1 October 2003. Limitations on the maximum allowable particulate emissions became more stringent, depending on the weight and registration dates of vehicles. Vehicles that did not meet the limitations were required to mount diesel particulate filters (DPFs) to remove particles. We have used the ΔEC/ΔCO ratio to detect the possible effect of this regulation on ambient EC. The ΔEC/ΔCO for 0400–0800 LT and the 24-hour average during May and June 2003 are 8.8 ± 0.8 and 6.5 ± 0.7 ng m−3/ppbv, respectively. For comparison with these data, we have chosen data obtained at similar temperatures to minimize the effect of the temperature dependence of ΔEC/ΔCO. The corresponding ΔEC/ΔCO values in May–June 2004 are 8.0 ± 1.2 and 6.6 ± 1.2 ng m−3/ppbv, respectively. If the data during May–October 2004 (excluding August) are used, these values are 7.9 ± 0.9 and 6.4 ± 1.1 ng m−3/ppbv. These values are only 0–10% lower than the preregulation values. This means that the possible decrease in ΔEC/ΔCO caused by the additional regulation is within the 10% variability of the data. After these regulations were adopted, about 10% of the diesel vehicles in the TMA were equipped with passive regeneration-type DPFs, which are composed of a catalyst to remove particles by oxidation and a particulate filter. This type of DPF removes particles emitted from diesel vehicles with high efficiencies [Yokota et al., 2003]. However, about 90% of diesel vehicles in the TMA were equipped with DPFs that use only catalysts. It is likely that this type of DPF does not remove EC efficiently. In addition, the EC removal efficiency of any type of DPF might depend on the condition of vehicles using the roadways. These possibilities need to be validated by further tests or measurements of emissions directly from vehicles on the roads. The present study demonstrates the importance of accurately measuring ambient EC and its tracer CO on a long-term basis in detecting the effects of EC emission changes on ambient EC concentrations. There are no other systematic data sets of EC and CO comparable to those obtained by the present study in the TMA. In this regard, the continuation of the EC and CO measurements will be useful in detecting the effects regulations of EC emissions.

4.5. ΔEC/ΔCO at Other Locations

4.5.1. Nagoya City

[35] Here we make comparisons of the ΔEC/ΔCO values in Tokyo with those obtained over Nagoya city and its vicinity by aircraft in March 2003. The population of Nagoya city (including surrounding major cities) is 2.20 (3.37) million and the locations of aircraft sampling in this region are shown in Figure 16. In situ measurements of CO, CO2, and EC were made onboard the Gulfstream-II aircraft during the Pacific Exploration of Asian Continental Emission (PEACE)–C campaign, conducted between 22 and 27 March 2004, within the framework of the atmospheric chemistry project of the Earth Observation Research Center (EORC) of Japan Aerospace Exploration Agency (JAXA). CO and CO2 were measured using a vacuum ultraviolet (VUV) resonance fluorescence instrument [Takegawa et al., 2001] and an NDIR instrument [Machida et al., 2002], respectively, with a time resolution of 1 s.

Figure 16.

Flight tracks for PEACE-C aircraft measurements made near Nagoya city, shown as thick and dashed lines. The data obtained in the region north of 34.5°N (thick lines) were used for the present analysis.

[36] EC was measured using a particle soot absorption photometer (PSAP) manufactured by Radiance Research Inc. (Seattle, Washington, United States) and used for previous aircraft missions [Liley et al., 2002]. The PSAP measures optical extinction of light at 565 nm by absorbing aerosols accumulated on a filter. For PEACE-C, sampled air was heated to 400°C to remove volatile aerosol components, in a way similar to that described in section 2. The specific absorption coefficient σae (m2 g−1) is defined as the ratio of the optical absorption coefficient (m−1) to the EC mass concentration (g m−3). Previous studies have shown significant variability in σae values [e.g., Liousse et al., 1993; Martins et al., 1998; Sharma et al., 2002], due partly to differences in the mixing state of EC particles. However, by heating the sampled air, the specific absorption coefficient σae has been found to become much more stable under different conditions and not influenced by changes in the mixing state of EC. The average σae, after correction following the procedures described by Bond et al. [1999], has been determined to be 8.9 m2 g−1 by comparison with the simultaneous EC measurements in Tokyo in July–August 2004. In addition, during the same period, σae was derived to be 8.7 m2 g−1, by similar measurements in Kisai city (Figure 2), 50 km north of Tokyo, where the degree of EC mixing with volatile aerosols was greater (Y. Kondo, unpublished data, 2004). From these comparisons in Tokyo and Kisai, the uncertainty in σae is estimated to be smaller than 10%. The precision of the EC measurements by PSAP is estimated to be 0.3 μg m−3 for an integration time of 10 s.

[37] The sample air for the PSAP was aspirated via a forward facing inlet system. The flow velocity in this system was slowed to 4–5 m s−1 from an air speed (aircraft speed relative to wind speed) ranging over 100–150 m s−1 below 1 km. The flow velocity of 4–5 m s−1 was comparable to the sampling velocity of the PSAP. This enabled isokinetic air sampling by the PSAP for submicron aerosols. In this analysis we used the data obtained below 1 km in the region north of 34.5°N (Figure 16), within distances of about 50 km from Nagoya. In total, 8 profiles obtained on 4 different flight days were used.

[38] The 1-min-averaged EC concentrations below 1 km are plotted versus CO values in Figure 17. The ΔEC/ΔCO derived from the slope of the least squares fitting is 6.3 ± 0.5 ng m−3/ppbv. To check the variability in the CO source types between the two regions, the slope of the observed CO-CO2 correlation was compared. Over Nagoya, ΔCO/ΔCO2 = 12.6 ± 0.7 ppbv/ppmv, which is similar to the value of 16 ± 2 ppbv/ppmv obtained over Nagoya in January 2002 during PEACE-A [Takegawa et al., 2004]. These values, together with ΔEC/ΔCO2, are compared with those obtained in Tokyo in Table 3. In this table, EC/CO, CO/CO2, and EC/CO2 emission ratios averaged over all of Japan estimated by Streets et al. [2003] are also shown for comparison. The observed ratios in Nagoya and Tokyo are similar, indicating uniformity of the emission ratios averaged over these areas. The average EC/CO and EC/CO2 ratios given by Streets et al. [2003] agree with the observed ΔEC/ΔCO and ΔEC/ΔCO2 values to within about 40 and 30%, respectively, providing a measure of the overall uncertainty of the emission inventories averaged over Japan. Further improvements in the estimate of the CO/CO2 emission ratio could lead to a better understanding of the difference between the observed ΔEC/ΔCO (ΔEC/ΔCO2) ratios and the EC/CO (EC/CO2) emission inventory ratios.

Figure 17.

EC-CO correlation obtained in the boundary layer over the region shown in Figure 16 (north of 34.5°N). The number of the data points (n) is 101. The least squares fitting is shown for reference.

Table 3. Comparison of the Average ΔEC/ΔCO, ΔCO/ΔCO2, and ΔEC/ΔCO2 Measured in Tokyo and Nagoyaa
  • a

    ΔEC/ΔCO is given in ng m−3/ppbv, ΔCO/ΔCO2 is given in ppbv/ppmv, and ΔEC/ΔCO2 is given in ng m−3/ppmv. Emission ratios derived from emission inventories by Streets et al. [2003] are also shown for comparison.

Tokyo5.7 (±0.9)11.2 (±0.4)64 (±13)
Nagoya6.3 (±0.5)12.6 (±0.7)79 (±11)
Streets et al. [2003]9.3 (±4.2)8.8 (±1.5)82 (±35)

4.5.2. Other Urban Sites

[39] We have compared the median EC and ΔEC/ΔCO values (24-hour averaged) obtained in Tokyo with those measured in urban and suburban sites on the North American continent in Table 4. Generally, the ΔEC/ΔCO value in the United States was lower than the values from the present work. In Atlanta, Georgia, Fort Meade, Maryland, and Baltimore, Maryland, United States, the ΔEC/ΔCO value ranged between 2.9 and 4.1 ng m−3/ppbv [Lim and Turpin, 2002; Chen et al., 2001; Park et al., 2005]. These values are 1.5–2.1 times lower than those measured in Tokyo and Nagoya. The ΔEC/ΔCO in Mexico City [Baumgardner et al., 2002] is about 7 times lower than that in Tokyo. The low value in Mexico City has been confirmed by observations during the Mexico City Metropolitan Area field campaign in April 2004 [Jiang et al., 2005], although the values were 80% higher than those in 2000. In Mexico City, a majority of vehicles are equipped with gasoline engines, leading to lower EC emissions than in Tokyo. CO concentrations in Mexico City were much higher than those in Tokyo [Baumgardner et al., 2002], partly contributing to the lower ΔEC/ΔCO. In summary, ΔEC/ΔCO in Tokyo was 2–3 times higher than or comparable to those measured in other urban or suburban areas in the United States and Mexico.

Table 4. Comparison of ΔEC/ΔCO in Different Industrial/Urban Regions
Tokyo2003–20055.7 ± 0.9thermal-opticalthis work
NagoyaMarch 20036.3 ± 0.5light absorptionthis work
Atlanta, Georgia, United StatesAug.–Sept. 19993.2thermal-optical1
Fort Meade, Maryland, United StatesJuly 19994.1 ± 1.6thermal-optical2
Baltimore, Maryland, United StatesMarch–Nov. 20022.5 ± 1.6thermal-optical3
Mexico CityJan.–Feb. 20000.88light absorption4
Mexico CityApril 20041.6light absorption5

5. Summary and Conclusions

[40] PM1 EC concentrations and ΔEC/ΔCO were measured on an hourly basis in Tokyo between May 2003 and February 2005. The mass concentrations of nonvolatile aerosol measured by the calibrated SMPS combined with a heated inlet agreed with the independent EC measurements, with a systematic difference of about 4%, demonstrating that the mass concentrations of nonvolatile aerosol are representative of those for EC. A majority of the nonvolatile aerosol and therefore EC mass concentration was in the size range of Dve = 50–200 nm (Dm = 50–480 nm), peaking at around Dve = 130 nm.

[41] EC and CO were well correlated throughout the measurement period because of a similarity in sources. CO and CO2 were also well correlated in autumn and winter, indicating that both CO and CO2 are good tracers of EC. EC generally decreased with increasing wind speed, indicating the importance of dilution by vertical mixing and horizontal transport in controlling their near surface concentrations. Because the ΔEC/ΔCO ratio is not affected by dilution, it is a more suitable parameter for detecting changes in the strength of EC emissions than EC measurements alone. EC and ΔEC/ΔCO values showed diurnal variation, peaking in the early morning (0400–0800 LT) and reaching minimum values around midnight. The peak EC and ΔEC/ΔCO values were 2 times greater than the minimum values at midnight. This diurnal pattern is similar to that of the traffic density of diesel trucks. ΔEC/ΔCO for Sundays was lower by about 50% than weekday values, in reasonable agreement with changes in truck traffic density. These results indicate that diesel trucks, especially heavy-duty trucks, are the dominant sources of EC in Tokyo.

[42] The median EC and CO values of the whole data set obtained in 2003–2005 were 1.8 ± 1.8 μg m−3 and 368 ± 274 ppbv, respectively. Because the variations of monthly median EC and CO values were correlated, the average monthly ΔEC/ΔCO ratio was stable at 7.2 ± 2.6 ng m−3/ppbv for 0400–0800 LT. ΔEC/ΔCO showed a seasonal variation, reaching broad maximum values in spring-autumn and reaching minimum values in midwinter, following the seasonal variation in temperature. The overall dependence of ΔEC/ΔCO on temperature was 0.17 ng m−3/ppbv/°C for 2°–25°C. The dependence of ΔEC/ΔCO on temperature was also observed in Maryland, United States [Chen et al., 2001], and is likely due to the temperature dependence of EC emissions from diesel engines on intake air temperature.

[43] More stringent regulation of emissions of particles from diesel vehicles started in the Tokyo metropolitan area in October 2003. Changes in the ΔEC/ΔCO values did not exceed the natural variability (10%) after 1 year from the start of the new regulations, when this temperature dependence is taken into account.

[44] ΔEC/ΔCO and ΔEC/ΔCO2 measured in the boundary layer over Nagoya in March 2004 were close to those observed in Tokyo. EC/CO and EC/CO2 emission inventory ratios averaged over Japan [Streets et al., 2003] agree with the measured 24-hour average values of ΔEC/ΔCO and ΔEC/ΔCO2 to within 60 and 20%, respectively, providing a measure of the uncertainties in the emission inventories.


[45] This research was funded by the Japanese Ministry of Education, Culture, Sports, Science and Technology (MEXT) and the Japanese Science and Technology Agency (JST). We acknowledge the ceilometer data provided by M. Yasui and the traffic and CO data in the Iogi tunnel provided by K. Ishii. We are indebted to all of the PEACE-C participants for their cooperation and support. Special thanks are due to the flight and ground crews of the G-II aircraft of Mitsubishi Diamond Air Service Co. CO2 data obtained during PEACE-C were provided by T. Machida.