Semicontinuous measurements of submicron water-soluble organic carbon (WSOC) aerosol were made simultaneously with organic carbon (OC) and elemental carbon (EC) in the Tokyo urban area in winter, summer, and fall 2004. The measurements of WSOC and OC/EC were made every 6 min and 1 hour, respectively, using a particle-into-liquid sampler (PILS) with a total organic carbon (TOC) analyzer and with an EC-OC analyzer using a thermal-optical technique. The PILS and 12-hour integrated filter measurements of WSOC agreed to within 12%. The WSOC mass concentrations and WSOC/OC ratio showed diurnal variations with peaks at 1200–1400 LT in summer and later in the afternoon in winter. On average, the WSOC/OC ratio was 0.20 and 0.35 μg C/μg C for winter and summer/late fall, respectively. The difference in the winter and summer frequency distributions of the WSOC/OC ratio suggests that the sampled air masses in summer and fall were more photochemically processed than those in winter. Secondary organic carbon (SOC) concentrations were estimated using the EC-tracer method. The measured WSOC was highly correlated with the derived SOC (r2 = 0.61–0.79), with WSOC/SOC slopes of 0.67 to 0.75 μg C/μg C for each season. These results suggest that the WSOC and SOC were similar in their chemical characteristics in this study. Water-insoluble organic carbon (WIOC) ( = OC–WSOC) correlated well with EC and CO (r2 = 0.59–0.73). The diurnally averaged WIOC/EC ratios were nearly constant (1.1 ± 0.1 μg C/μg C) throughout the study periods, suggesting that motor vehicle emissions were an important source of WIOC. A dominant portion (about 90% or more) of the POC was water-insoluble, consistent with previous studies of POC.
 Organic compounds constitute a major fraction (10–70%) of total fine particle mass in urban air [e.g., Turpin et al., 2000]. Previous studies have shown that water-soluble organic carbon (WSOC) can significantly alter the hygroscopicity of aerosols and may be important in determining the ability of particles to act as cloud condensation nuclei (CCN) [Novakov and Penner, 1993; Saxena et al., 1995; Facchini et al., 1999]. Therefore, WSOC can play an important role both in direct and indirect effects on radiative forcing in the troposphere. WSOC is also considered to be associated with a major fraction of secondary organic aerosol (SOA), which is formed by oxidation of volatile organic compounds (VOCs) followed by condensation on existing particles and/or nucleation. Warneck  has recently suggested that oxalic acid, which is generally the most abundant dicarboxylic acid in the atmosphere [Kawamura and Sakaguchi, 1999], may be formed in cloud droplets through heterogeneous reactions in marine environments. Knowledge of the relative abundance of WSOC provides insight into processes leading to the formation of SOA as well as CCN, which have important implications for regional air quality and global climate.
 Measurement of the water-soluble fraction of total organic carbon provides valuable clues regarding the chemistry of organic aerosol and its interaction with water vapor in ambient air [Saxena and Hildemann, 1996; Jacobson et al., 2000]. Zappoli et al.  found that WSOC in fine particles with diameters (Dp) less than 1.5 μm, accounted for 77 and 65% of the fine particle organic carbon (OC) at European background and polluted sites, respectively. Seasonal variation in fine particles in the Po Valley, Italy, showed that WSOC accounted for between 38% in winter and 50% in summer of OC [Decesari et al., 2001]. These measurements were made by integrated filter-based sampling (typically 12–24 hours) followed by offline chemical analysis. Sullivan et al.  first made online measurements of WSOC and found that the average WSOC in a residential area of St. Louis, MO, contributed to 64, 61, and 31% of OC in June, August, and October, respectively.
Decesari et al.  have proposed that WSOC can be separated into three main classes according to the following acid/base characteristics: neutral compounds, mono-/dicarboxylic acids, and polyacids. Previous studies have also shown that a considerable portion of WSOC is composed of polyacidic compounds or humic-like substances (HULIS) [e.g., Havers et al., 1998; Kiss et al., 2002; Mayol-Bracero et al., 2002]. Observational studies of WSOC have focused mainly on dicarboxylic acids, keto acids, and dicarbonyls. For urban aerosols, e.g., these include C2–C9α, ω-dicarboxylic acids; C2–C9α-oxocarboxylic acids; pyruvic acid; and C2–C3 dicarboxylic acids [e.g., Sempere and Kawamura, 1994].
 Experimental data on WSOC in ambient air, however, are still lacking, especially in urban regions, and the behavior of WSOC is not well characterized because emission, production, and transport processes of aerosols have dynamic timescales ranging from minutes to hours. This situation emphasizes the need for time-resolved aerosol measurements, which can provide data that lead to a much better understanding of the chemical properties of WSOC. Moreover, it is generally believed that under certain conditions (e.g., lack of biomass burning influence), total SOA is nearly equivalent to WSOC because SOA produces oxygenated chemical functional groups resulting in OC that is water-soluble, although this has not been experimentally demonstrated for ambient air.
 This paper presents the results of high time-resolution ground-based measurements of WSOC, OC, and elemental carbon (EC) in the Tokyo Metropolitan Area during the winter, summer, and fall of 2004. The results are interpreted in terms of the properties of WSOC and its related parameters in urban Tokyo, including its diurnal and seasonal profiles. We also present the relationship between WSOC and estimated secondary OC (SOC) based on the EC-tracer method. In this study we use different terminologies for organic compounds in aerosol particles. These include WSOC and SOC, depending on the measurement techniques used and aerosol chemical characteristics. A summary of the definitions is given in the Notation section.
 Measurements were made on the top floor of a building (∼20 m above ground level and 55 m above sea level) on the campus of the Research Center for Advanced Science and Technology (RCAST), University of Tokyo (35.4°N, 139.4°E). This site is located ∼8 km west of the center of Tokyo and ∼20 km northwest of the coast (Tokyo Bay). Further details on the sampling location and meteorological conditions are described elsewhere [Kondo et al., 2006a; Takegawa et al., 2006]. The simultaneous measurements of WSOC, OC, and EC were made during three periods: 25 January to 6 February 2004 (winter), 1–17 August 2004 (summer), and 3–18 November 2004 (fall).
2.1. Particle-Into-Liquid Sampler (PILS)–WSOC
 Semicontinuous measurements of WSOC were made using a particle-into-liquid sampler (PILS) [Weber et al., 2001; Orsini et al., 2003] followed by online quantification of total organic carbon (TOC) every 6 min using a TOC analyzer (Model 810; Sievers, Boulder, Colorado). Because Sullivan et al.  have presented the principles and details of PILS-WSOC operation, only a brief description is given here. Ambient aerosol was sampled at a flow rate of 16.7 l min−1 by the PILS, which used a steam saturator to grow the aerosol to sizes that can be collected by inertial impaction. The liquid sample (1.3 cm3 min−1) was filtered by a 0.5-μm stainless steel meshed liquid filter. The carbonaceous compounds in the liquid sample were then quantified online with the TOC analyzer. In the TOC analyzer, organic compounds dissolved in deionized water were oxidized to form carbon dioxide (CO2) using an ultraviolet (UV) lamp and ammonium persulfate as a chemical oxidizing agent. The CO2 formed was then measured using a membrane-based conductivity detection technique. The term WSOC in the present study is technically defined as particles sampled by the PILS and detected by the TOC analyzer after penetrating the liquid filter. In the PILS-WSOC instrument, the organic mass is diluted to typical concentrations of the order of 104 g per 1-l water.
 Measurements of carbonaceous aerosol are difficult because of possible interferences from the following two types of interactions with vapor-phase compounds: adsorption of VOCs that contribute to the aerosol measurements (positive artifact) and volatilization of semi-VOCs during the sampling (negative artifact). In order to reduce the positive artifact, a parallel plate diffusion denuder [Eatough et al., 1993] was placed upstream of the PILS. For the negative artifact, evaporative losses of semi-VOCs during the PILS sampling have not been evaluated in this study.
 Internal calibration was periodically performed using oxalic acid as a standard. The concentrations of dissolved TOC in water ranged between 30 and 230 parts per billion by mass (ppbm) in the present study. The zero levels (blanks) of WSOC were measured automatically for 30 min every 4 hours by passing the sample air through a Teflon® filter upstream of the PILS. The uncertainty in the WSOC measurements in this study was estimated to be between ±5 and 9%. A PM1 (1.0-μm-diameter cutoff size) cyclone (URG-2000-30EHB; URG Corp., Chapel Hill, NC) was used during the measurement period.
2.2. Comparisons to Integrated Filter Measurements
 The traditional technique for measuring WSOC is manual extraction of collected mass from filters using purified water followed by TOC analysis. In addition to the WSOC measurements by the PILS, 12-hour integrated filter samples were taken to determine WSOC concentrations and provide a comparison of the two techniques during the period from 9–15 August 2004. A PM1 cyclone (URG-2000-30EHB; URG Corp.) and Teflon® filter pack (URG-2000-30FG; URG Corp.) were used for the filter sampling. The sampling was made at a flow rate of 16.7 l min−1. We did not use a carbon denuder during the sampling. The quartz filters were extracted with organic-free ultrapure water. The water extracts were then filtered with a syringe disk filter (Miller-GV, 0.22 μm; Millipore, Billerica, MA), which had been rinsed under ultrasonication in pure water three times and then flushed with pure water five times to remove potential contaminants. The WSOC in the water extracts was measured using a Shimadzu (Kyoto) TOC-5000A carbon analyzer [Wang et al., 2005]. The analytical error of this measurement was 15% with a detection limit of 0.1 μg C m−3.
Figure 1a shows the comparison of WSOC measured by the PILS together with WSOC data obtained by the filter sampling (Filter-WSOC). The PILS-WSOC captured the short-term temporal variations of WSOC that were averaged in the Filter-WSOC. For quantitative comparison, individual PILS–WSOC data were merged into the 12-hour filter sampling periods and the merged data are plotted against the Filter-WSOC (Figure 1b). The PILS–WSOC regression slope to the integrated Filter-WSOC (forced through zero) was 0.88 ± 0.10 (r2 = 0.89; 95% confidence interval). Concentrations of PILS-WSOC agreed with those of Filter-WSOC to within 12% on average, indicating that the PILS–WSOC measurements in the current study were almost identical to the filter-based measurements of WSOC. Slightly lower PILS–WSOC measurements may have resulted from several factors, including sampling artifacts associated with each measurement (e.g., absorption of volatile organic compounds that contributed to the particle sampling), differences in the solubility of various organic compounds, and differences in the TOC analytical methods. These possible factors, however, were not evaluated in the current study. Sullivan and Weber [2006b] also compared the PILS-WSOC measurements to 24-hour integrated high-volume filter measurements in Atlanta, Georgia. They reported that the High-Volume samplers were generally larger than the PILS-WSOC measurements by 10%, which is similar to the results obtained in this study.
2.3. OC and EC Measurements
 Mass concentrations of OC and EC were measured using a semicontinuous EC/OC analyzer (Sunset Laboratory Inc., Tigard, OR) [Birch and Cary, 1996; Bae et al., 2004]. The inlet for air sampling was equipped with a PM1 cyclone (Model URG-2000-30EHB; URG Inc.) and denuder that is identical to that used for measuring WSOC. In this instrument, ambient aerosol particles are collected on a quartz fiber filter for 45 min at a flow rate of 8 l min−1, followed by analysis based on the thermal-optical transmittance method for 15 min. Temperature of a filter sample was increased at four steps up to 870°C (National Institute for Occupational Safety and Health (NIOSH) protocol [Birch and Cary, 1996]), first in an inert (helium, He) environment. After completion of the oxygen-free heating stages, the filter sample was heated to 870°C in an oxidizing (2% oxygen in He, He/Ox) atmosphere. Ideally, OC volatilizes in the He mode and converted to CO2, which is measured by a nondispersive infrared (NDIR) CO2 detector. However, a fraction of the OC does not volatilize during this mode but instead pyrolyzes. In the He/Ox mode, both the original EC and the pyrolyzed OC are converted to CO2 and measured by the NDIR detector. Pyrolyzed OC was estimated by using the transmittance of a pulsed He-Ne diode laser beam at 670 nm through the quartz fiber filter during the sample analysis. Background levels of OC were periodically checked by placing a particle filter upstream of the denuder.
 Prior to this study, we made measurements of OC and EC in both PM1 and PM2.5 (2.5-μm-diameter cutoff size) for 4 days in fall 2003 at the same sampling site. The measurements showed that on average the mass of OC and EC in PM1 accounted for 75% and 95% of those in PM2.5, respectively. This result indicates that the OC and EC mass concentrations were dominantly in the PM1 range at the sampling site.
 The overall accuracies of the OC and EC measurements were 16 and 22%, respectively, for the study period [Kondo et al., 2006a; Takegawa et al., 2005]. The detection limits of these measurements were also assessed as 1 μg m−3 for OC and 0.2 μg m−3 for EC. Water-insoluble organic carbon (WIOC) is defined as WIOC = OC–WSOC. The uncertainty of WIOC was estimated to be 26%, by combining errors of the OC and WSOC measurements. This paper reports 1-hour average OC, EC, and WIOC concentrations.
2.4. NOx, NOy, CO, and O3
 Nitrogen oxides (NOx = NO + NO2), total reactive nitrogen (NOy), and carbon monoxide (CO) were simultaneously measured in this study to determine primary combustion-generated carbon emissions. Measurements of NOx and NOy were conducted at 1-s time resolution using a NO-O3 chemiluminescence detector combined with a photolytic converter and a gold tube converter [Kondo et al., 1997]. The accuracy of the measurements of NO, NO2, and NOy were 10, 16, and 12%, respectively (Fukuda et al., manuscript in preparation).
 An NDIR gas analyzer (Model 48; Thermo Environmental Instruments (TEI), Franklin, MA) was used to measure CO at an integration time of 1 min. The detection limit for CO was 10 parts per billion by volume (ppbv). The overall precision and accuracy of the 1-min CO measurements were estimated to be 4 ppbv and 20 ppbv, respectively, at a CO mixing ratio of 400 ppbv [Takegawa et al., 2006]. Ozone was measured using the UV absorption technique with an integration time of 1 min (Model 1101; Dylec, Tokyo). All of these gas-phase data as well as WSOC were merged into the time interval of OC and EC sampling (∼45 min).
2.5. Estimation of Primary and Secondary OC
 To compare the measured WSOC with SOC, we estimated primary OC (POC) and SOC by using the measured EC and OC following the EC-tracer method [e.g., Turpin and Huntzicker, 1995]. EC is often used as a good tracer of primary combustion-generated carbon emissions [e.g., Turpin and Huntzicker, 1991]. The EC-tracer method assumes a representative ratio of primary OC/EC for a given area because EC and primary OC typically have the same sources. By this method, the POC concentration ([POC]) can be defined as
where [OC/EC]prim is the estimated primary OC/EC ratio and c is to account for noncombustion sources contributing to the POC and sampling artifacts [Turpin and Huntzicker, 1995; Strader et al., 1999]. The SOC concentration [SOC] can be estimated as
where [OC] is the measured OC concentration.
 As a first step in determining [OC/EC]prim, we identified the data representing ambient concentrations dominated by primary emissions. We selected data in which the EC concentrations were larger than 0.5 μg C m−3. In the present study, NOx/NOy ratios and CO were used as tracers of combustion-related primary emissions. The inverse of the NOx/NOy ratio is also a good indicator of the photochemical processing of an air mass. Figure 2 shows the relationship between OC/EC and NOx/NOy ratios for the winter and summer periods. The OC/EC ratios decreased with the increase in the NOx/NOy ratios, indicating an increased influence of primary emissions over the study area. CO concentrations were also used as an indicator of primary emissions. We selected data with NOx/NOy and CO levels larger than threshold values determined for each season to extract OC data strongly influenced by primary emissions. As shown in Table 1, the threshold values for NOx/NOy ratios were chosen to be 0.9 (0.8) for winter (summer) and those for CO were 700 ppbv (400 ppbv) for winter (summer). The threshold values for NOx/NOy are close to the median values, and those for CO levels correspond to the highest 80% CO values for each season. The larger threshold value for the NOx/NOy ratios in winter than that for summer is due to the higher average NOx/NOy ratio (Figure 2). The larger average NOx/NOy ratio in winter is caused by the slower NOx oxidation rate at the lower OH concentrations. The diurnally averaged NOx lifetimes are estimated to be ∼1.3 days in winter and ∼0.6 days in summer for the same sampling site [Takegawa et al., 2006]. The larger threshold for CO value in winter is due to suppressed vertical mixing of air within the boundary layer. The lower boundary layer height in winter leads to less effective dilution of air and elevated surface CO concentrations [Kondo et al., 2006a].
Table 1. Estimated Parameters for the Linear Fit of POC and EC Concentrations
Thresholds for OC and EC concentrations influenced by anthropogenic primary emissions.
Winter (Jan–Feb 2004)
Summer (Aug 2004)
Fall (Nov 2004)
 The data with the OC/EC ratios lower than the median values of the data set already filtered by the NOx/NOy ratios and CO values were further selected for each season to represent the data set dominated by primary emissions. Linear regression of OC–EC was then fitted to the datasets of the primary-dominated concentrations. This slope represents the [OC/EC]prim in equation (1), which was then used to estimate SOC by equation (2). Table 1 summarizes the estimated [OC/EC]prim ratio, and other parameters for each season. Figure 3 shows the OC-EC correlations for the data dominated by primary emissions (open circles) and the rest of the data (shaded circles) for each study period. The estimated ratios of [OC/EC]prim were 1.46 ± 0.12, 1.36 ± 0.12, and 1.33 ± 0.13 μg C m−3 (μg C m−3)−1 for winter, summer, and fall, respectively, with 95% confidence intervals. Streets et al.  estimated emission ratios of OC/EC from anthropogenic sources to be 1.40 g C/g C over Japan for the year 2000. The estimated [OC/EC]prim in the current study was similar to the OC/EC emission ratio from Streets et al., although the uncertainty in their emission ratio was large (83% for EC and 181% for OC at 95% confidence intervals). Cabada et al.  applied the EC-tracer method to estimate POC and SOC during summer near downtown Pittsburgh, PA, by introducing NO, NOx, and CO as combustion-related tracers. They used measurements (with time resolutions of 2–6 hours) of OC and EC to derive a [OC/EC]prim ratio of 1.70 ± 0.20, a result that is similar to but slightly higher than the current estimate.
 We now examine possible uncertainties in the estimation of POC and SOC mass concentrations. Uncertainties were mainly introduced by the assumptions that primary noncombustion OC was negligible and that [OC/EC]prim was constant for each season. First, if primary noncombustion (e.g., biogenic) OC had substantially contributed to SOC concentrations, the SOC concentrations derived here would be overestimated. From the intercept of the linear regression fit in Figure 3, the non-EC associated POC was estimated to be 0.3, 0.1, and 0.1 μg C m−3 for the winter, summer, and fall periods, respectively. These values are as low as 7% (winter) and <1% (summer and fall) of the mean SOC values estimated in this study. Second, variation of the source strength and local meteorology can change [OC/EC]prim even within a few hours to 1 day [Harley et al., 2005]. The uncertainties of [OC/EC]prim (95% confidence intervals) correspond to ±10–12% of the SOC concentrations for the entire periods estimated above. Possible uncertainties of the POC were estimated to be ±24% by combining errors of the EC measurements and the uncertainties of [OC/EC]prim. Overall, these considerations suggest possible uncertainties of ±28% for the SOC estimated in this study.
 It is noted that EC particles immediately after emissions are hydrophobic and then become hydrophilic by condensation of water-soluble components onto them over time scales of about 0.3–1.2 days [e.g., Cooke et al., 1999; Riemer et al., 2004]. The hydrophilic EC particles are incorporated into clouds or fogs and removed from the atmosphere by precipitation scavenging (rainout). However, cloud or fog formation near the surface in Tokyo was not frequent during the whole observational periods and the effects of rainout on the surface EC concentrations could not be detected [Kondo et al., 2006a]. Therefore, the effects of EC removal on the determination of [OC/EC]prim were neglected in this study. Quantitative comparisons of the derived SOC and the measured WSOC concentrations are made in section 3.3.
3. Results and Discussion
3.1. Temporal Variations
Figure 4 shows time series plots of the measured mass concentrations of WSOC, OC, and EC for the winter, summer, and fall periods. Wind speeds and directions are also shown. The time series show the range of mass concentrations of these species during the study periods and the amount of data available for each measurement period. Mean values of WSOC, WIOC, OC, EC, and the ratio of WSOC to OC for the three study periods are summarized in Table 2. Throughout the study periods, patterns of the temporal WSOC variations were similar to those of OC. The WSOC concentrations ranged from levels below the detection limit to maximum values of approximately 5, 8, and 10 μg C m−3 for the winter, summer, and fall periods, respectively. The OC concentrations ranged between 2 and 15 μg C m−3 (winter), 0.2 and 13 μg C m−3 (summer), and 1 and 23 μg C m−3 (fall). Diurnal variations of WSOC, OC, WSOC/OC, WIOC, EC, and CO averaged for the winter and summer periods are also shown in Figure 5. Diurnal variations for the fall are not shown because the sampling site was affected by synoptic-scale meteorology (on a time scale of 3 to 5 days) leading to no distinct diurnal variations.
Table 2. Mean Values and ±Standard Deviations of PM1 WSOC, WIOC, OC, EC, and the WSOC/OC Ratios for the Three Study Periods in Tokyo
WSOC, μg C m−3
WIOC, μg C m−3
OC, μg C m−3
EC, μg C m−3
Winter (Jan–Feb 2004)
1.05 ± 0.91
4.05 ± 1.99
5.00 ± 2.78
2.46 ± 1.33
0.19 ± 0.08
Summer (Aug 2004)
1.28 ± 1.18
2.70 ± 1.45
3.98 ± 2.68
1.69 ± 1.17
0.35 ± 0.15
Fall (Nov 2004)
2.38 ± 1.84
3.57 ± 2.30
5.95 ± 2.37
3.52 ± 2.32
0.37 ± 0.14
 In winter, local surface winds were dominated by northerlies, which transported air from inland of the Kanto plain with low wind speed (<2 m s−1) in general. The WSOC concentrations were generally anticorrelated with wind speeds. The highest WSOC and OC concentrations reached 5 and 16 μg C m−3, respectively, on 2 February (Figure 4a). On this day, a surface low was developing just south of the Japanese main island of Honshu. The sampling site was located behind a surface cold front associated with the low-pressure system, causing the transport of pollutants inland from the Kanto Plain.
 For the winter period, WSOC began to increase at ∼0600 LT, reaching maximum concentrations of ∼1 μg C m−3 in the late afternoon (1400–1800 LT) on average (Figure 5). In addition, a second peak in WSOC was observed at night (2200–2400 LT). The WSOC/OC ratios reached maximum values of 0.20 μg C/μg C at ∼1600 to 1800 LT and minimum values of 0.13 μg C/μg C at ∼0600 LT, mainly driven by the temporal variations of WSOC concentrations.
 In summer, WSOC concentrations were generally less than 2 μg C m−3 during the period of 1–10 August mainly because of strong southerly winds with speeds of ∼3 m s−1 on average. These southerly winds were driven by the development of the high-pressure system centered over the Pacific, bringing relatively clean marine air from the south to the sampling site. By contrast, during the period of 11–16 August, a sea-land breeze circulation prevailed under low general winds. As a result, the weak (∼1 m s−1) northerly winds (land breeze circulation) was observed typically from midnight until morning, while in the afternoon the southerly wind (sea breeze circulation) brought air over the ocean to the sampling site, as seen from Figure 4b.
 WSOC concentrations increased up to 8 μg C m−3 at ∼1200 LT from 12–14 August (Figure 4b), corresponding to the weak winds. In the afternoon and evening (1300–2200 LT), WSOC decreased to ∼2 μg C m−3, mainly because of the dominance of the southerly winds with increased wind speeds (>2 m s−1). The OC/EC ratios were relatively high at 2 to 7 μg C/μg C, with maximum O3 concentrations exceeding 100 ppbv, indicating high photochemical activity during this period. The median WSOC concentrations started to increase in the early morning (∼0400 LT), showing broad peaks of ∼1.1 μg C m−3 at ∼0800 to 1400 LT (Figure 5). The daytime peak values were ∼2–3 times larger than values during the night. The OC concentrations also showed broad peaks at 0800 to 1400 LT with maximum values of 3.0 μg C m−3. The WSOC/OC ratios had a maximum value of 0.45 μg C/μg C at 1200 to 1400 LT and a minimum value of 0.30 μg C/μg C at 0000 to 0200 LT.
 The observed WSOC levels could be interpreted as resulting from a combination of locally produced WSOC and that transported from an upwind region. To examine the relative degree of photochemical processing, WSOC/EC ratios were used to correct for the effect of dilution to some degree. During summer, the WSOC/EC ratio correlated well with O3 (r2 = 0.62). In addition, the r2 value increased considerably (r2 = 0.72) if daytime data (0600–1800 LT) were selected. Therefore, a major fraction of WSOC in summer was likely produced by photochemical processes, which are also linked to the O3 production. Kawamura and Yasui  measured specific water-soluble organics (these included diacids, ketoacids, and dicarbonyls) in the Tokyo Metropolitan Area for 6 days in summer and winter of 1989. These organic compounds accounted for <2.5% of the total carbon. Despite the limited number of samples, they also showed diurnal variations of these species with a maximum during the day, which were interpreted as the photo-oxidation of aromatic hydrocarbons and cyclic olefins.
 In winter, the WSOC and OC concentrations increased from the late afternoon until midnight. The increase in the late afternoon was also seen for the CO concentrations, reaching ∼670 ppbv. In addition, the WSOC/EC ratio was poorly correlated with O3 (r2 < 0.01). These results suggest that the elevated WSOC levels in winter were more strongly controlled by the transport of WSOC produced in the region north of the sampling site and its accumulation under lower boundary layer height conditions than in situ production at the sampling site.
3.2. Seasonal Variations in WSOC/OC Ratios
Figure 6 shows frequency distributions of the WSOC/OC ratios during the winter, summer, and fall periods. Overall, the observed WSOC/OC ratios ranged from 0.02 to 0.80 μg C/μg C. A seasonal variation in the modal ratios is apparent. The WSOC/OC ratios were 0.02–0.45, 0.02–0.86, and 0.05–0.85 μg C/μg C for the winter, summer, and fall periods, respectively. Median WSOC/OC ratios were 0.19 μg C/μg C in winter and 0.35 μg C/μg C in summer and fall. Considering that the average WSOC levels in summer and fall were also higher than those in winter, the results suggest that the sampled air masses in summer and fall were more photochemically processed than those in winter. Moreover, the WSOC/OC ratios for summer and fall had broader distributions than those for winter, partly as a result of the stronger effects of the photochemical formation of WSOC in these seasons. Air masses with different chemical aging or processing will broaden the WSOC/OC distributions.
 Generally, the WSOC/OC ratios obtained in this work were lower than those found at other urban and rural sites in previous studies. Decesari et al.  showed seasonal variations in fine particles (Dp < 1.5 μm) in the Po Valley, Italy, where WSOC accounted for 38 and 50% of OC in winter and summer, respectively. Kiss et al.  measured WSOC (Dp < 1.5 μm) by filter-based samplings to determine that WSOC accounted for 71% of OC on average at a rural site in Hungary from January to September 2000. Sullivan and Weber [2006a] reported that mean WSOC/OC ratios in PM2.5 were ∼0.50 and 0.60 μg C/μg C in winter and summer, respectively, for urban measurements in St. Louis, MO and Atlanta, Georgia. At an urban site in Nanjing, China, WSOC accounted for 30% of OC for PM2.5 in winter [Yang et al., 2005]. More comprehensive summaries of WSOC/OC ratios are given by Mader et al.  and Jaffrezo et al. .
 The lower WSOC/OC ratios in the present study may reflect a larger contribution of local WIOC sources to the OC concentrations, compared to those reported in previous studies. For example, the mean WSOC value of 1.57 μg C m−3 for the entire period is similar to that (1.98 μg C m−3) in urban Atlanta [Sullivan and Weber, 2006a], whereas the average WIOC value of 3.44 μg C m−3 in urban Tokyo is much larger than the value of 1.24 μg C m−3 for Atlanta. This may have been due to the location of the sampling site in the vicinity of large urban sources of WIOC in Tokyo [Kondo et al., 2006a]. In fact, Ruellan and Cachier  also observed low mean WSOC/OC values of 0.13 μg C/μg C near a high-traffic road around Paris during summer and fall. This value is slightly lower than the winter WSOC/OC ratios reported in this work. A possible major source of WIOC in Tokyo is discussed in section 3.4. Another possibility is that production of WSOC from biogenic VOCs was low in urban Tokyo. For example in urban Atlanta in summer, radiocarbon measurements (14C) of filter-extracted WSOC indicate that approximately 70 to 80% of WSOC is from biogenic sources and 20 to 30% is from fossil fuels (R. Weber, unpublished data). However, the contribution of biogenic WSOC was not evaluated in this study, because WSOC compounds were neither speciated into functional groups nor characterized at the molecular level.
3.3. Comparison to Derived SOC
 We now examine the relationship between the measured WSOC and derived SOC in the previous section for quantitative comparison of these two parameters. On average, the derived SOC accounted for 34 ± 17, 47 ± 21, and 37 ± 20% of OC in winter, summer, and fall, respectively. SOC/POC ratios were 0.52 ± 0.27, 0.89 ± 0.41, and 0.59 ± 0.33 μg C/μg C in winter, summer, and fall, respectively. Figure 7 shows the temporal variations of the measured WSOC and the derived SOC concentrations for each season. Scatterplots of WSOC and SOC are also shown in Figure 7. WSOC and SOC were highly correlated, with r2 = 0.61–0.79 for all seasons. The slopes with 95% confidence intervals of the correlation in winter, summer, and fall were similar: 0.67 ± 0.29, 0.74 ± 0.21, and 0.73 ± 0.22 μg C/μg C, respectively. In contrast, WSOC was poorly correlated with POC (r2 = 0.30–0.36) for each season. These results indicate that SOC and WSOC are very similar in their chemical characteristics. It is interesting to note that the average WSOC/SOC ratios were smaller than unity for all the seasons, even if the uncertainties of ±28% in the SOC estimates (section 2.5) are taken into account. This may suggest that not all the SOC compounds were necessarily water-soluble; some SOC compounds are considered to have large carbon-hydrogen functional groups, leading to insolubility [e.g., Saxena and Hildemann, 1996]. Some studies, however, suggest that the EC-tracer method tends to overestimate SOC [Yuan et al., 2006]. This may also account for, at least to some degree, the discrepancy between derived SOC and measured WSOC.
Kondo et al. [2006b] compared the measured WSOC with oxygenated organic aerosols (OOA), which may be a good approximation of secondary organic aerosol at the same sampling site. Organic aerosol was measured with an Aerodyne aerosol mass spectrometer (AMS; Aerodyne Research, Billerica, MA) at the same sampling site in winter and summer [Kondo et al., 2006b; Takegawa et al., 2005]. OOA and hydrocarbon-like organic aerosols (HOA) were quantified separately from the AMS mass spectral time series based on custom principal component analysis by using a newly developed algorithm by Zhang et al . Kondo et al. [2006b] showed that the measured WSOC was highly correlated (r2 = 0.87–0.93) with OOA and quantified total carbon concentrations in OOA (defined as OOC) derived from the OOA mass spectra, showing that ∼88% of OOC is WSOC (i.e., WSOC/OOC ∼ 0.88 μg C/μg C). The WSOC/OOC ratio is similar to the average WSOC/SOC ratio of about 0.71 μg C/μg C. These results indicate that WSOC, OOC, and SOC are very similar and likely represent a very similar set of chemical species in the aerosols.
3.4. Water-Insoluble Organic Carbon (WIOC)
 We now examine possible sources of WIOC, since WIOC accounts for a significant part of OC at the sampling site in urban Tokyo. The average WIOC concentrations ranged between 2.7 and 4.1 μg C m−3 during the entire period. The WIOC/OC ratios were 0.81, 0.65, and 0.63 μg C/μg C in winter, summer, and fall, respectively. Scatterplots of WIOC and EC are shown in Figure 8 for each season. Generally, WIOC correlated well with EC during the entire period, with r2 ranging between 0.62 and 0.76. The average slope of the correlation between WIOC and EC was stable to within ±10% for all the seasons (1.0–1.2 μg C/μg C).
 WIOC also correlated with CO (r2 = 0.54–0.63; not shown), which is often used as a tracer of motor vehicle emissions in urban areas [e.g., Harrison et al., 1997; Lim and Turpin, 2002]. Most WIOC in urban areas is likely composed of incomplete combustion products such as aliphatic hydrocarbons, long-chain ketones, alkanols, and polycyclic aromatic hydrocarbons (PAHs) [e.g., Simoneit et al., 2004]. In fact, WIOC observed at the same sampling site was positively correlated with a signal at m/z 57 of the AMS mass spectra [Kondo et al., 2006b], which is a fragment (C4H9+) typical of saturated hydrocarbon compounds mainly from combustion sources [Allan et al., 2003; Alfarra et al., 2004]. These results indicate that motor vehicle emissions significantly contributed to the observed WIOC.
 The WIOC/EC ratios showed diurnal variations, as seen in Figure 8. They increased from 0.90–1.09 μg C/μg C between 0000 and 1200 LT (r2 = 0.65–0.78) to 1.20–1.31 μg C/μg C between 1200 and 2400 LT (r2 = 0.59–0.73) for all the seasons. The ratios ranged from 0.88–1.03 μg C/μg C at 0600–1200 LT to 1.28–1.41 μg C/μg C at 1800–2400 LT. This diurnal variation is driven by the diurnal variation of EC to some extent; the average EC concentrations in the morning (0400–1200 LT) were higher by a factor of ∼2 than those in the afternoon (1200–2400 LT) in summer, as seen in Figure 5. A morning peak in EC concentrations has been typically observed in urban areas [e.g., Turpin and Huntzicker, 1991; Allen et al., 1999; Kondo et al., 2006a]. Kondo et al. [2006a] showed that in Tokyo, exhaust from heavy-duty diesel trucks was a major source of EC, reaching its maximum in the early morning (0400–0800 LT). The amplitude of the diurnal variations of the WIOC/EC ratios was as low as ±19% or smaller, which was much smaller than that for the EC/CO ratios (by a factor of ∼2) [Kondo et al., 2006a]. The high correlations (r2 = 0.59–0.78) between WIOC and EC and the stable WIOC/EC ratios suggest that motor vehicle emissions were an important source of the WIOC.
 POC was defined to be nearly proportional to EC by equation (1) and therefore was highly correlated (r2 = 0.65–0.75) with WIOC. The WIOC/POC ratio was determined by the slope of the least squares fitting of the WIOC-POC scatterplots. The uncertainty of the WIOC/POC ratios was estimated to be ±32%, by combining the errors of the WIOC and POC concentrations, discussed in section 2. The WIOC/POC ratios with 95% confidence intervals of the correlation were 1.01 ± 0.49 μg C/μg C (winter), 0.84 ± 0.41 μg C/μg C (summer), and 0.85 ± 0.32 μg C/μg C (fall). These results indicate that a predominant portion of POC was water-insoluble.
 On the other hand, some portion (about 30%) of SOC was estimated to be water-insoluble, as discussed in section 3.3. The WIOC/POC ratios should therefore be larger than 1, because the concentrations of SOC and POC were estimated to be comparable on average. The upper limit of the WIOC/POC ratio is estimated to be about 1.28 from the average ratio of ∼0.90 ± 0.38 μg C/μg C, with its uncertainty of 32%.
 Organic aerosols directly emitted from motor vehicles were analyzed by Rogge et al.  and Tobias et al. . Most of them consist of unburned fuel and/or lubricating oil [Tobias et al., 2001]. These organic aerosols in motor vehicle exhaust were found to be predominantly water-insoluble, as anticipated from their chemical compositions. Water insolubility of the POC derived in this study is understood by considering that most of the POC represent organic aerosols emitted from motor vehicles in Tokyo.
 Semicontinuous measurements of WSOC in the fine mode were made in the Tokyo urban area in January, August, and November 2004 using a PILS followed by online quantification of WSOC every 6 min using a TOC analyzer. OC and EC were simultaneously measured using a semicontinuous thermal-optical carbon analyzer. The PILS measurements were compared to 12-hour integrated filter-based measurements. These two measurements agreed to within 12%. The WSOC concentrations and WSOC/OC ratios showed diurnal variations with peaks at 1200–1400 LT in summer and at 1400–1800 LT in winter. The average WSOC/OC ratios were 0.20 and 0.35 μg C/μg C for winter and summer/fall. The maximum values of WSOC/OC reached 0.37 and 0.86 μg C/μg C in winter and summer, respectively. The seasonal variations of the frequency distributions of the WSOC/OC ratio suggest that the air masses sampled in summer and fall were more photochemically processed than those in winter.
 The concentrations of secondary organic carbon (SOC) were estimated using the EC-tracer method. The NOx/NOy ratios and CO mixing ratios were used to determine the OC/EC emission ratio from combustion sources. The mass concentrations of WSOC and SOC were highly correlated, with r2 = 0.70–0.79. The WSOC/SOC ratios (linear regression slopes) ranged from 0.67 to 0.75 μg C/μg C. In contrast, the measured WSOC showed poor correlations with primary organic carbon (POC) with r2 = 0.30–0.36 throughout all of the periods. These results indicate that WSOC and SOC likely represent a similar set of chemical species in the aerosol.
 WIOC correlated well with EC and CO, with r2 = 0.59–0.73, depending on local time. The diurnally averaged WIOC/EC ratios (linear regression slopes) showed little seasonal variation (1.03–1.17 μg C/μg C). These results suggest that motor vehicle emissions were an important source of the WIOC. The average WIOC/POC ratio was estimated to be about 0.90 ± 0.38 μg C/μg C, with an uncertainty of 32%.
 In summary, the present study shows that about 71% of SOC was water-soluble, with an uncertainty of ±28%. On the other hand, a dominant portion (about 90% or more) of the POC was water-insoluble, consistent with previous chemical analysis of POC.
Definitions of Organic Aerosol
elemental carbon (μg C m−3).
hydrocarbon-like organic aerosol (μg m−3).
organic carbon (μg C m−3).
oxygenated organic aerosol (μg m−3).
primary organic aerosol (μg C m−3).
secondary organic carbon (μg C m−3).
water-insoluble organic carbon, WIOC = OC–WSOC (μg C m−3).
water-soluble organic carbon (μg C m−3).
 The authors thank A. Sullivan for providing the information on the PILS measurements. We also thank T. Watanabe and T. Nakajima for their help in the filter-based measurements of WSOC. This research was supported by global environment research fund of the Japanese Ministry of the Environment (C-051), the Japanese Ministry of Education, Culture, Sports, Science and Technology (MEXT), and the Japanese Science and Technology Agency (JST).