Removal of sulfur dioxide and formation of sulfate aerosol in Tokyo

Transcription

1 Click Here for Full Article JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 112,, doi: /2006jd007896, 2007 Removal of sulfur dioxide and formation of sulfate aerosol in Tokyo T. Miyakawa, 1 N. Takegawa, 1 and Y. Kondo 1 Received 9 August 2006; revised 4 December 2006; accepted 16 February 2007; published 12 July [1] Ground-based in situ measurements of sulfur dioxide (SO 2 ) and submicron sulfate aerosol (SO 4 ) together with carbon monoxide (CO) were conducted at an urban site in Tokyo, Japan from spring 2003 to winter The observed concentrations of SO 2 were affected dominantly by anthropogenic emissions (for example, manufacturing industries) in source areas, while small fraction of the data (<30%) was affected by large point sources of SO 2 (power plant and volcano). Although emission sources of CO in Tokyo are different from those of SO 2, the major emission sources of CO and SO 2 are colocated, indicating that CO can be used as a tracer of anthropogenic SO 2 emissions in Tokyo. The ratio of SO 4 to total sulfur compounds (SO x =SO 2 +SO 4 ) and the remaining fraction of SO x, which is derived as the ratio of the linear regression slope of the SO x -CO correlation, is used as measures for the formation of SO 4 and removal of SO x, respectively. Using these parameters, the average formation efficiency of SO 4 (i.e., amount of SO 4 produced per SO 2 emitted from emission sources) are estimated to be 0.18 and 0.03 in the summer and winter periods, respectively. A simple box model was developed to estimate the lifetime of SO x. The lifetime of SO x for the summer period (26 h) is estimated to be about two times longer than that for the winter period (14 h). The seasonal variations of the remaining fraction of SO x, estimated formation efficiency of SO 4, and lifetime of SO x are likely due to those of the boundary layer height and photochemical activity (i.e., hydroxyl radical). These results provide useful insights into the formation and removal processes of sulfur compounds exported from an urban area. Citation: Miyakawa, T., N. Takegawa, and Y. Kondo (2007), Removal of sulfur dioxide and formation of sulfate aerosol in Tokyo, J. Geophys. Res., 112,, doi: /2006jd Introduction [2] It is well known that atmospheric sulfate aerosol has important impacts on ecosystems and climate change. Deposition of sulfate has adverse effects on ecosystems through acidification of soils, lakes, and marshes [e.g., Schindler, 1988; Gerhardsson et al., 1994]. The direct and indirect radiative effects caused by sulfate have strong impacts on global and regional climate [IPCC, 2001]. Global and annual mean direct radiative forcing by sulfate aerosol has been estimated to be between 0.26 and 0.4 W m 2 [Haywood et al., 1997; Myhre et al., 1998; IPCC, 2001]. Lohmann et al. [2000] estimated that the global mean annual average indirect radiative forcing of sulfate is from 0 to 0.4 W m 2. [3] Fossil fuel combustion is the major source of sulfur dioxide (SO 2 ) and sulfate aerosol in urban areas. Megacities in Asia are large emission sources of SO 2. Guttikunda et al. [2003] estimated that megacities in Asia cover less than 2% 1 Research Center for Advanced Science and Technology, University of Tokyo, Tokyo, Japan. Copyright 2007 by the American Geophysical Union /07/2006JD007896$09.00 of the land area but emitted about 16% of the total anthropogenic sulfur in Asia for the years of 1975 to They also estimated that urban sulfur emissions contributed over 30% to the regional pollution levels of the sulfur concentrations in large parts of Asia during the same period. [4] Anthropogenic SO 4 is produced mainly by oxidation of SO 2.SO 2 is oxidized to gas-phase sulfuric acid (H 2 SO 4 ) by gas phase reactions. SO 2 þ OH þ M! HOSO 2 þ M HOSO 2 þ O 2! HO 2 þ SO 3 SO 3 þ H 2 O þ M! H 2 SO 4 þ M [5] M is a third body. In the formation of H 2 SO 4, equation (1) is the rate-limiting step. The resultant H 2 SO 4 condenses almost irreversibly onto the preexisting aerosols. It also nucleates with water vapor (H 2 O) and/or H 2 O and gaseous ammonia (NH 3 ) to form sulfate aerosol [Kulmala et al., 2004; and references therein]. In addition, sulfate aerosol is also produced efficiently by aqueous phase reac- ð1þ ð2þ ð3þ 1of13

2 tions, where ozone (O 3 ) and hydrogen peroxide (H 2 O 2 ) act as important oxidants [Seinfeld and Pandis, 1998]. SðIVÞþO 3! SðVIÞþO 2 HSO 3 þ H 2O 2 $ SO 2 OOH þ H 2 O SO 2 OOH þ H þ! H 2 SO 4 These reactions oxidize SO 2 in the cloud and fog droplets and on particles. H 2 SO 4 in the particle phase can take the form of either SO 4, HSO 4,orH 2 SO 4 depending on the availability of NH 3 and liquid water content of the particle. SO 2 is also removed efficiently by dry deposition. Sulfate aerosol is removed predominantly by wet deposition [Seinfeld and Pandis, 1998]. [6] The formation of sulfate aerosol from SO 2 and loss of the sulfur compounds have been studied extensively. The spatial distributions and temporal variations of SO 2 and sulfate at various locations have been investigated mainly by filter-based observations in previous studies [e.g., Hidy et al., 1978; Husar and Patterson, 1980]. These studies showed that the sulfate mass concentration and the molar ratio of sulfate to total sulfur compounds (SO x =SO 2 + sulfate) showed maximum values in summer and minimum values in winter. Chemical processes of sulfur compounds were also studied by aircraft observations of power plant plumes [Hewitt, 2001; and references therein]. The major findings from these studies are that (1) under cloud-free conditions, chemical loss of SO 2 in power plant plumes was dominated by reaction with hydroxyl radical (OH) during daytime, (2) under cloudy conditions, SO 2 was removed rapidly by aqueous phase reactions, and (3) the conversion rate of SO 2 to sulfate varied from 1 to3% h 1. Wojcik and Chang [1997] summarized a number of studies on the atmospheric residence time of SO x t SOx and suggested that (t SOx ) estimated in the previous studies were about 1 5 days. Koike et al. [2003] estimated the large-scale transport efficiency of SO x from the East Asian source regions using the data from the NASA Transport and Chemical Evolution over the Pacific (TRACE-P) experiment. The transport efficiency of SO x (i.e., remaining fraction of SO x exported from emission sources) and formation efficiency of sulfate (i.e., amount of sulfate produced per SO 2 from emission sources) in the boundary layer were estimated to and 10 25%, respectively, for the TRACE-P period. [7] Although a number of studies of sulfur chemistry have been made, including those discussed above, detailed studies of sulfur chemistry near the source areas are still limited. The (t SOx ) and formation efficiency of sulfate near the source areas determine the spatial distribution of sulfate in surrounding regions. Therefore an improved understanding of these processes is essential for evaluating the impacts on air quality and climate on regional and global scales. In the present study, we estimate these parameters for Tokyo, which is one of the largest megacities in East Asia, using data obtained from April 2003 to February For this purpose, time- and size-resolved measurements of sulfate aerosol were performed using an Aerodyne aerosol mass spectrometer (AMS). The formation efficiency of sulfate ð4þ ð5þ ð6þ was estimated using data sets of SO 2 and sulfate obtained during the intensive measurements. A simple box model was developed to quantitatively estimate (t SOx ) in Tokyo. The seasonal variations of these parameters are also discussed. 2. Measurements [8] Intensive measurements of aerosols and gases were conducted at an urban site in Tokyo during the spring (April May 2003), summer (July August 2003), fall (September October 2003), and winter (January February 2004) periods. The observations were made on the campus of the Research Center for Science and Technology (RCAST), University of Tokyo ( N, E), 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, Mega-Cities: Asia. Detailed descriptions of the observational site and measurement campaigns are given in the works of Kondo et al. [2006] and Takegawa et al. [2006]. [9] The mass loadings and size distributions of nonrefractory submicron (NR-PM 1 ) aerosols (vaporized at 600 C in the vacuum) were measured using an Aerodyne AMS with a time resolution of 10 min. The AMS can provide highly time-resolved (5 10 min) measurements of sizeresolved mass loadings of NR-PM 1 aerosols [Jayne et al., 2000; Jimenez et al., 2003]. The performance of our AMS was characterized on the basis of intercomparison with a Particle Into-Liquid Sampler combined with Ion Chromatography (PILS-IC) [Takegawa et al., 2005]. The mass concentrations of inorganic aerosols measured by the AMS agreed well with those measured by PILS-IC to within 26%. SO 2 was measured using a pulsed-uv fluorescence technique with an integration time of 1 min (model 43C, Thermo electron corporation (TECO), USA). The zero levels were measured every 2 or 3 days by supplying SO 2 -free air generated by a charcoal filter. The limits of detection (2s of zero signals) were calculated to be 0.03 ppbv (parts per billion by volume) for each observation period using 8 sets of 5-min data. Calibrations were performed every week by supplying an SO 2 gas standard (1 part per million by volume (ppmv), Nissan-Tanaka Corp., Japan) diluted with purified dry air to parts per billion by volume (ppbv). The sensitivity of this instrument was stable (within 5%) from April 2003 to February Carbon monoxide (CO) was measured using a nondispersive infrared absorption technique with an integration time of 1 min (model48, TECO, USA). The performance of the CO instrument is given by Takegawa et al. [2006]. Meteorological parameters, including wind speed and wind direction, were measured by an automatic weather station (model HMP45DX; Vaisala Inc., Helsinki, Finland) with an integration time of 10 min. For the analysis in the following sections, SO 2 and CO concentrations are averaged according to the AMS integrating time (10 min). 3. Mass Loadings of NR-PM 1 Aerosols in Tokyo [10] Figure 1 shows the average chemical composition of NR-PM 1 aerosols measured by the AMS in Tokyo. There was no significant seasonal variation in the average mass 2of13

3 Figure 1. Pie chart representation of the average chemical composition of NR-PM 1 aerosols measured using an AMS during the summer and winter periods. concentration of NR-PM 1 aerosols during the measurement periods, as described by Takegawa et al. [2006], and were 14 ± 8 and 17 ± 10 mg m 3 in the summer and winter periods, respectively. Takegawa et al. [2006] investigated the molar ratios of sulfate, nitrate, and chloride to ammonium (NH ) observed by the AMS and showed that NH 4 was almost fully neutralized by the observed sulfate, nitrate, and chloride aerosols. This result suggested that in this study sulfate aerosol took the form of (NH 4 ) 2 SO 4 or SO 4 in solid or liquid particles, respectively. Hereafter, sulfate aerosol is referred to as SO 4 for convenience. Organic aerosol accounted for 40 50% of NR-PM 1 aerosols in both periods. SO 4 and nitrate were the dominant inorganic anions in the summer and winter periods, respectively. The behaviors of nitrate aerosols are discussed in detail by Morino et al. [2006]. 4. Emission Sources, Chemical Conversion, and Loss of SO x in Tokyo [11] Brief descriptions of the emission sources of SO 2 in Tokyo are shown in section 4.1. It is indicated that the observed concentrations of SO 2 at the observation site are affected by the natural and anthropogenic emissions of SO 2. [12] In order to interpret the time variations of SO 2 concentration, CO is used as a tracer compound of anthropogenic emissions of SO 2 for two reasons. First, it is well known that the chemical lifetime of CO is long enough (2 months) to be a good tracer of transport. Second, emission sources of CO and SO 2 in Tokyo are colocated, as discussed in section 4.1. Carbon dioxide (CO 2 ) can also be a good tracer in addition to CO. However, the mixing ratio of CO 2 was not measured for the summer period. CO 2 was well correlated (r 2 = 0.9) with CO for the fall and winter periods as discussed by Takegawa et al. [2006], indicating that the time variations of CO 2 are nearly the same as those of CO. Therefore CO was used as a tracer of anthropogenic SO 2 in the present study. [13] Removal of SO x is investigated using the emission ratio of SO 2 to CO in Tokyo and the linear regression slope of the correlation of SO x versus CO (SO x /CO slope). Formation of SO 4 is investigated using the molar ratio of SO 4 to SO x (SO 4 /SO x ). In sections 4.3, 4.4, 4.5, and 4.6, we discuss the relationship between the SO 4 /SO x and SO x /CO slopes to investigate the removal of SO x and formation of SO 4 in Tokyo Sources of Sulfur Compounds in Tokyo [14] Figure 2 shows the SO 2 emission inventory (10 10-km grid) for Japan developed by Kannari et al. [2004], used for the present analysis. The region inside the rectangle (50 70 km) in Figure 2 ( E, N) is considered as the major source region (hereafter referred to as the source region) for the air sampled at RCAST. Observed surface winds for the observation periods generally blew from the south or north of RCAST. Therefore we use the above-defined rectangular region for the source area. Large emission sources of SO 2 are located in Figure 2. Map of Tokyo metropolitan area and surrounding regions with SO 2 annual emission (kg grid 1 year 1 ). The color code depicts the annual emissions of SO 2 from each grid box. The cross symbol depicts the location of RCAST ( N, E). The black rectangular region ( N, E) is the source region (see text). 3of13

4 Figure 3. (a) The time series of CO (black, left axis) and SO 2 (shaded, right axis). (b) Time series of horizontal wind speed (black line, left axis) and wind direction (shaded circles, right axis). the coastal area, which are mainly due to the power plant emissions. The commercial institutions, manufacturing industries, waste incineration, and motor vehicles also make significant contributions to the total anthropogenic emissions of SO 2 in the Tokyo metropolitan Area. In addition to these anthropogenic emissions, volcanic emissions of SO 2 from Miyakejima Island, which is located about 180 km south of RCAST, can significantly affect the concentrations of SO 2 and SO 4 over Tokyo [Kajino et al., 2004]. Mt. Oyama on Miyakejima Island (Miyakejima Volcano) erupted in July 2000 and has continued emitting SO 2 since then. We estimated the potential effects of the Miyakejima Volcano and concluded that the observed sulfur concentrations were not strongly impacted by the volcano (<10%). Detailed descriptions of the volcanic source of SO 2 are discussed in Appendix A1. [15] On the other hand, 80% of CO is emitted from motor vehicles and the rest from manufacturing industries, and commercial institutions in the Tokyo metropolitan area [Kondo et al., 2006]. Therefore the major sources of SO 2 observed at RCAST are different from those of CO in some cases. However, the emission inventory and its relationship with the observations indicated that the emission sources of CO and SO 2 in Tokyo were colocated and mainly comprised of commercial institutions, manufacturing industries, and motor vehicles (Appendix A2, Figure A1). Therefore we conclude that CO can be used as a tracer of anthropogenic SO 2 emissions. The emission ratio of SO 2 to CO (ER SO2 /CO) value was calculated to be ( ) g g 1 = ( ) mol mol 1, where the values in the parentheses are the uncertainties of ER SO2 /CO used to evaluate the uncertainties caused by the estimated ER SO2 /CO in section 4.5. Detailed descriptions of the anthropogenic sources of SO 2 are shown in Appendix A Data Selection [16] We show time variations of CO, SO 2, and winds (speed and direction) during a part of the summer period in Figure 3 in order to illustrate characteristics of emission sources of SO 2 in Tokyo and their relationship with observed surface wind patterns. Figure 3a shows the time series plots of CO (left axis) and SO 2 (right axis) for 1 13 August Wind speed (left axis) and wind direction (right axis) for the same period are also shown in Figure 3b. The mixing ratios of SO 2 varied from 0.1 ppbv to 10 ppbv, in correlation with CO, during this period. However, there were some high-so 2 episodes that were uncorrelated with CO (for example, 10:00 12:00 LT on 9 August), suggesting SO 2 emissions from different types of sources. Figure 4 depicts the relationship between SO 2 mixing ratio and wind speed for the summer period. These high values of SO 2 (shown inside the rectangles in Figure 4) were often observed under southerly wind conditions with relatively high wind speeds (>3 ms 1 ). Moreover, the time duration of these large SO 2 plumes were 1 3 h, indicating that, according to the characteristics of some high-so 2 periods, we can classify the data affected by the large point sources of SO 2 located south of RCAST (power plants and/or volcanic emissions). These data constituted 20, 24, and 29% of the entire data set in the summer, fall, and winter periods, respectively. They were excluded from 4of13

5 Figure 4. Relationship between SO 2 and horizontal wind speed. Black and open circles denote the data sets with southerly and northerly wind patterns, respectively (southerly wind, ; northerly wind, 0 90 and ). the present analysis. It is suggested that the observed air masses were mostly affected by the anthropogenic emissions (for example, manufacturing industries) near RCAST Average Formation Efficiency of SO 4 [17] Previous studies used SO 4 /SO x ratios as a parameter to represent SO 4 formation processes [e.g., Kaneyasu et al., 1995]. The SO 4 /SO x ratio increases with air mass aging because of the chemical conversion of SO 2 to SO 4 and a faster SO 2 loss rate than that of SO 4 by dry deposition. Figure 5 depicts histograms of the SO 4 /SO x ratios observed in different seasons. The modal SO 4 /SO x ratios showed a significant seasonal variation as follows: (summer), (fall), and (winter). Similar seasonal variation of the SO 4 /SO x ratio was observed in previous studies [e.g., Bari et al., 2003]. The removal rate of SO 2 via equation (1), which is also the formation rate of H 2 SO 4, is given as k[oh][so 2 ], where k is the reaction rate coefficient of equation (1) calculated using average temperature and pressure for each period [DeMore et al., 1997]. If SO 4 is predominantly produced via reaction (R1), k[oh][so 2 ] can be regarded as the formation rate of SO 4. The average values of k are calculated to be 8.89, 9.24, and cm 3 s 1 for summer, fall, and winter, respectively. Therefore k[oh] should exhibit significant diurnal and seasonal variation depending mainly on OH concentration. Using an empirical method to estimate OH concentration [Ehhalt and Rohrer, 2000], the estimated diurnally averaged lifetime (e-folding) of SO 2 by equation (1) was calculated to be 7 days (15 days) for the summer (winter) periods. Therefore localized in situ gas-phase formation was not important in controlling the SO 4 levels observed at RCAST. However, the seasonal variations of the SO 4 /SO x ratios can be, at least partly, due to those of the photochemical activity (i.e., OH), considering that equation (1) should also influence regional scale SO 4 /SO x ratios. [18] The SO x /CO slope should decrease with air mass aging mainly due to dry deposition of SO 2, this and can be used as a diagnostic for the removal of SO x. The SO x /CO slope should be also affected by the air masses exported from outside of the source region that have significantly different emission ratios from the source area. The relative contribution of spatially different SO 2 emission sources at the observation site is proportional to the emission rate divided by the square of the distance between the observation site and emission sources, indicating that the emission sources apart from the observation site do not strongly affect the SO x /CO slope observed at the observation site. Therefore we consider the observed changes of the SO x /CO slope as being controlled by the loss of SO x during transport from the source area. Figure 6 shows the correlation of SO x versus CO for the summer period. The data sets in Figure 6 were classified according to the SO 4 /SO x as follows: and Background concentrations of SO x (BG- SO x ) and CO (BG-CO) are the reference points of the enhancement of each compound in order to calculate the SO x /CO slope. The BG-CO values were determined as the average of the data lower than the fifth percentile of observed CO concentrations [Kondo et al., 2006; Takegawa et al., 2006]. BG-SO x were determined as the averaged value of the same data set used to determine BG-CO. The derived BG-SO x and BG-CO for the observation periods are summarized in Table 1. The SO x /CO slope in each air mass is determined by least squares fitting forced through the background data point (the open square in Figure 6). The SO x /CO slope in fresh emissions was larger than that in processed air, indicating that SO x was removed with air mass aging. [19] In order to quantify the loss of SO x, we estimated the remaining fraction of SO x, denoted as RSO x, using the SO x / CO slope and ER SO2 /CO. The RSO x means the ratio of SO x to total SO 2 emitted and is expressed as RSO x ¼ðSO x =CO slopeþ=er SO2=CO Figure 5. Histograms of the observed SO 4 /SO x in the summer (black), fall (dashed), and winter (shaded) periods. ð7þ 5of13

6 Figure 6. Correlations of SO x versus CO distinguished using the values of the observed SO 4 /SO x ratios during the summer period. The square and thin black line depict the background values of SO x and CO and the estimated emission ratio of SO 2 to CO (see text), respectively. Fresh emissions and processed air correspond to the black and shaded circles, respectively. The bold black and shaded lines show the linear regression slopes of the fresh emissions and processed air data, respectively. S in this figure denotes a value of the linear regression slope of each data. If loss of SO x does not occur, the SO x /CO slope is equal to ER SO2 /CO, and RSO x is estimated to be 1. On the other hand, a loss of SO x results in the decrease of SO x /CO slopes, and RSO x is estimated to be less than 1. The average values of RSO x were estimated using the average SO x /CO slopes and are summarized in Table 2. It is suggested that t SOx for the summer was longer than that for the winter and about 60 % (80%) of SO x was removed in the summer (fall and winter) period in the boundary layer during transport from source regions to RCAST in all observation periods. The average RSO x defined here is equivalent to the transport efficiency of SO x given by Koike et al. [2003]. The transport efficiency of SO x in the continental outflow from the East Asia in spring was estimated to be 25 45% in the boundary layer, using the data measured during TRACE-P. The transport efficiency of SO 4 given by Koike et al. was comparable to the RSO x values for the winter and summer periods ( ). [20] The average formation efficiency of SO 4 (e av )is defined as the average molar ratio of the observed SO 4 to the total SO 2 emitted. It was estimated from the averaged values of the RSO x (RSO x ) and SO 4 /SO x ratios ( SO 4 SO x ) for each period, in the similar way as given by Koike et al. [2003], namely: e av ¼ RSO x SO 4 SO x ð8þ The product of the average values of the remaining fraction of SO x which means the ratio of SO x to total SO 2 emitted and the ratio of SO 4 to SO x represents the average value of the ratio of SO 4 to the total SO 2 emitted. [21] The derived e av values for the different seasons are given in Table 2, together with the SO 4 /SO x, SO x /CO slope, and RSO x values. The derived e av values were 0.18 and 0.03 for the summer and winter periods, respectively. It is suggested that 18% (3%) of total SO 2 emitted from the emission sources in Tokyo in the summer (winter) period was converted to SO 4 during the transportation. The key parameters estimated here, namely RSO x and e av, are discussed in detail in section 4.7. It should be noted that all of the physical and chemical processes involved in converting SO 2 to SO 2 4 are included in the observed SO 4 /SO x. Therefore e av represents the efficiency of SO 4 formation through gas-phase and aqueous-phase oxidation of SO 2. [22] We next describe the two sources of error for the RSO x and e av estimated above. First, we discuss the assumed constant background levels of CO, which can affect the calculation of the SO x /CO slope, because the SO x /CO slope is determined by least squares fitting forced through the background data point. Regional background levels of SO x and CO can be variable. If the variability of BG-CO is within 50%, that of the SO x /CO slope is calculated to be within 21% for all observation periods. The uncertainties derived from this assumption were also discussed by Takegawa et al. [2006]. The wintertime export of polluted air masses from Asian continental regions to Tokyo should affect BG-CO, because there is a difference in the DCO/DCO 2 ratios between China (28 48 ppbv ppmv 1 ) and Tokyo (10.7 ppbv ppmv 1 ), as given by Takegawa et al. [2006]. Tight correlation between CO and CO 2 in Tokyo (r 2 = 0.9) indicates that observed CO concentrations and BG-CO are not strongly affected by polluted air masses from China. Second, we discuss the uncertainty of the estimated ER SO2 /CO. This uncertainty was estimated to be between 23% and 15% in section 4.1. The total uncertainty of RSO x and e av due to the uncertainties in the constant BG-CO (<21%) and the estimated ER SO2 /CO (<23%) is estimated to be <31% Removal of SO x and Formation of SO 4 in Tokyo Formulation of the Chemical and Physical Processes [23] We developed a simplified box model in order to derive the removal and formation rates of SO 2 and SO 4 from the observational data. The chemical and physical processes considered in the model are the gas-phase oxidation of SO 2 (equations 1, 2, and 3), aqueous phase oxidation of SO 2 (equations 4, 5, and 6), and the dry deposition of SO 2. The dry and wet deposition of SO 4 are neglected Table 1. Background Concentrations of CO and SO x During the Observation Periods BG-CO (ppbv) BG-SO x (ppbv) Period Average 1s Average 1s Summer Fall Winter of13

7 Table 2. Average SO 4 /SO x, RSO x, and Formation Efficiency During the Observation Periods Period SO 2 4/SO x (ppbv ppbv 1 ) b SO x /CO slope (10 3 ppbv ppbv 1 ) c RSO x (ppbv ppbv 1 ) Formation Efficiency a,d Summer 0.44 (±0.20) 9.62 (±0.40) 0.40 (±0.02) 0.18 (±0.08) Fall 0.38 (±0.19) 5.21 (±0.19) 0.22 (±0.01) 0.08 (±0.04) Winter 0.17 (±0.09) 5.19 (±0.19) 0.21 (±0.01) 0.03 (±0.02) a Formation efficiencies (e av ) are estimated by multiplying the average SO 4 /SO x. b The values inside the parentheses are the standard deviation of the observed SO 4 /SO x. c The values inside the parentheses are derived from the 95% confidence interval of the linear regression slope of SO x. d The values inside the parentheses are derived from the standard deviation of the observed SO 4 /SO x. because they are considered to be slower than the dry deposition of SO 2, as discussed in detail later in this section. The errors associated with this assumption are discussed in section 4.5. Then, the temporal variations of SO 2 and SO 4 are expressed as d½so 2Š dt ¼ k½so 2 Š½OHŠþC½SO 2 ŠþL½SO 2 Š d½so 4 Š ¼ k½so 2 Š½OHŠþC½SO 2 Š; dt ð9þ ð10þ where C is the rate constant of the aqueous-phase reaction of SO 2, and L is the first-order loss rate for the dry deposition of SO 2.SO 4 is also produced efficiently by aqueous-phase oxidation of SO 2 in the cloud or fog droplets and heterogeneous reactions on the aerosol particles, as discussed in section 1. The reaction rate constant for aqueous-phase oxidation of SO 2 depends on various parameters, including the concentrations of oxidizing agents, liquid water content, and the ph of the aqueous phase. It is difficult to quantify all of the parameters involved with the aqueous phase-reaction from groundbased observations only. Therefore we expressed the rate constant of the aqueous-phase reaction as C symbolically. Seinfeld and Pandis [1998] suggested that aqueous-phase oxidation in the cloud or fog droplets and wet deposition are important processes for the production and removal of SO 4, respectively. [24] On the other hand, gas phase oxidation of SO 2 is believed to be the primary process for SO 4 formation in industrial plumes in summertime, under cloud-free conditions [McMurry et al., 1981]. Brock et al. [2002] observed particle growth in fair weather with air mass aging in the SO 2 -rich plumes from power plants. They found that the particle growth can be explained by the condensation of precursor gases from growth law analysis [Friedlander, 2000], without including the aqueous phase reactions. Pandis et al. [1995] calculated the timescale of aerosol processing (for example, transport, SO 4 production, and deposition) in the various atmospheric types and suggested that aerosol phase SO 4 production was of secondary importance in urban environments compared to gas- and aqueous-phase reaction of SO 2. The residence time of SO 4 aerosol is generally governed by wet deposition [Koch, 2001] and is estimated to be 4 6 days in modeling and field studies [e.g., Koch et al., 1999; Millet et al., 2004]. The timescale taken into account in this study (less than about a day) is shorter than the lifetime of SO 4. [25] Considering these past studies, we assumed that gasphase oxidation of SO 2 (equations 1, 2, and 3) was the major formation process of SO 4 (i.e., C was set to 0), and the loss of SO 4 could be neglected. In accordance with this assumption, we excluded data with precipitation or heavy cloud cover, which constitute 5, 5, and 1% of all data points in the summer, fall, and winter periods, respectively Solution to the Box Model Equations [26] The analytical solution to equations (11) and (12) are given as follows. ½SO 2 Š t ¼½SO 2 Š t¼0 expð k Total tþ ð11þ ½SO 4 Š t ¼ e½so 2Š t¼0 f1 expð k Total tþg ð12þ k Total and e are defined as follows. k Total ¼ k½ohšþl e ¼ k½ohš L þ k½ohš ð13þ ð14þ k Total is the total loss rate of SO 2. e represents the fraction of SO 2 molecules reacting with OH and producing SO 4 to total SO 2 molecules emitted in the simple box model. Using equation (13) and (14), SO 4 /SO x and RSO x are expressed as a function of time in the following equations (15) and (16). SO 4 ¼ SO x ðrso x t ½SO 4 Š t ½SO 2 Š t þ½so 4 Š t 1 expð k Total tþ ¼ 1 þ 1 e expð k Total tþ e ð15þ Þ t ¼ ½SO 2Š t þ½so 4 Š t ¼ e þð1 eþexpð k Total tþ ð16þ ½SO 2 Š t¼0 Using equations (15) and (16), e av defined in section 4.3 is expressed approximately as a function of a time in equation (17). eðtþ ¼ ½SO 4 Š t ½SO 2 Š t¼0 ¼ e f1 expð k Total tþg ð17þ Thus for moderately or fully processed air, e(t) tends close to e. Therefore we denote e(t) also as the SO 4 formation efficiency. 7of13

8 Figure 7. Relationship between the SO 4 /SO x ratio and CO in the summer (black) and winter (shaded) periods. Each marker corresponds to the average value of SO 4 / SO x and CO mixing ratio in each 0.2-SO 4 /SO x bin (i.e., 0 0.2, , , and ). The error bars depict the standard deviations in each bin. The dashed lines on the graph depict the average values of the observed CO concentrations in the summer (black) and winter (shaded) periods. Solid (open) markers in this figure are (not) used in the analysis. [27] As discussed above, we now assume that C equals zero in equations (16) and (17). Thus the unknown variables for equations (18) and (A1) become L and t. However, it is impossible to express these variables (L and t) as functions of known quantities in equations (18) and (A1), namely, SO 4 /SO x and RSO x. Instead, we calculate SO 4 /SO x and RSO x using L and t as input parameters. Then, the calculated SO 4 /SO x and RSO x values are compared with the observed values to estimate the L and t by least squares fitting. [28] As a first step, we need to estimate k[oh], in order to calculate k Total and e as a function of L. For this purpose, we used the diurnally averaged OH concentration calculated by the empirical formula given by Ehhalt and Rohrer [2000]. The diurnally averaged concentrations of OH were , , and molecules cm 3 for the summer, fall, and winter periods, respectively. These OH values correspond to the SO 2 oxidation rates of 0.40, 0.20, and 0.19% h 1 in the summer, fall, and winter periods, respectively. An SO 2 oxidation rate of 1% h 1 in the lower troposphere was suggested according to the reviews of previous studies, which derived the SO 2 to H 2 SO 4 conversion rate by field observations (conducted almost in summer season; one case in winter season) [Warneck, 1999]. The SO 2 oxidation rates estimated in the present study are somewhat smaller than this value, because we use diurnally averaged OH concentrations, while Warneck [1999] used daytime OH concentrations. [29] For comparison with the calculations, we selected SO 4 /SO x and RSO x values that were not significantly influenced by mixing, because the box model did not include this effect. Mixing of air masses with different ages can generally change the SO 4 /SO x and RSO x values [McKeen and Liu, 1993; McKeen et al., 1996]. The SO 4 / SO x ratio and RSO x varies because of mixing of relatively fresh emissions with background air with high SO 4 /SO x Figure 8. Fitting results. The markers and lines in this figure are the observed and calculated values of SO 4 /SO x and RSO x, respectively. Circles, triangles, and square depict the observed SO 4 /SO x and RSO x in the summer, fall, and winter, respectively. Black, dashed, and shaded lines depict the calculations in the summer, fall, and winter, respectively. ratios and low CO-mixing ratio. Therefore the relationship between SO 4 /SO x and CO can be convenient for diagnosing the influence of background air. Figure 7 depicts the relationships between SO 4 /SO x and the CO-mixing ratio for the summer and winter period. The CO data were averaged in each 0.2 SO 4 /SO x bin. The fall data were similar to those for the summer data. The data with large SO 4 /SO x ratios (for example, during the summer period) corresponded to low CO concentrations (for example, 250 ppbv during the summer period), indicating that the observed air masses with large SO 4 /SO x were strongly influenced by mixing with background air. In order to minimize possible effects of the air mass mixing on SO 4 / SO x and RSO x, we utilized only the data with CO-mixing ratios larger than the average for each period (closed markers in Figure 7) to compare calculations with observations. [30] Figure 8 shows the comparison of the observed relationships between SO 4 /SO x and RSO x (circles, squares, and triangles) with the model calculations (lines) for the different seasons. The calculated curves are the best fits to the observations obtained by varying L and t. The calculated line for the winter period was forced through the observed data point, because there is only one data available for the analysis of the winter period. The e-folding lifetime of SO x (t SOx ) is defined as t when RSO x is reduced to 1/e. The values of L, t SOx, and e derived by this comparison are shown in Table 3. The estimated L showed significant seasonal variations as follows: and h 1 in the Table 3. Summaries of the Estimated Values of L, t SOx, and e a Period L (hour 1 ) t SOx (hours) e Summer ( ) 26 (18 33) 0.16 ( ) Fall ( ) 19 (15 23) 0.07 ( ) Winter ( ) 14 (9 19) 0.05 ( ) a The values in the parentheses are the ranges of those derived from the variability of the observed SO 4 /SO x. 8of13

9 Table 4. Comparison Between Present Study and Previous Studies for k Total and e Study k Total (hr 1 ) e Region Method Present Study Tokyo Measurements and simple box model Summer ( ) Fall ( ) Winter ( ) Previous Studies Eliassen and Saltbones [1975] United Kingdom Measurements and air trajectory model Annual 0.08 ± Prahm et al. [1976] United Kingdom Measurements and air trajectory model Winter 0.07 ± Henmi and Reiter [1978] E. United States Calculation using observed boundary Summer layer heights and specified decay Winter rates from literatures summer and winter periods, respectively. The loss rate of dry deposition of SO 2 is expressed as L = v d /H, where v d is the dry deposition velocity of SO 2 and H is the boundary layer height. The seasonal variations of L should be due to seasonal changes in the boundary layer height, assuming no significant seasonal change in v d. Typical L values are estimated be h 1, independently using the typical values of H (200 m for nighttime and 1500 m for daytime) and v d (0.5 cm s 1 )[Zhang et al., 2003]. These L values are comparable to those estimated by the present study. Correspondingly, the values of t SOx were derived to be about 26 and 14 h in the summer and winter periods, respectively. It is indicated that significant fraction of SO x was removed in the boundary layer over Tokyo within about a day for all observation periods. The values of e showed maximum values of in the summer period and minimum values of in the winter period. The values of e are close to those of e av because k Total ô SOx was about , depending on the season. Interpretations of these results are discussed in section Evaluations of the Simplified Box Model [31] Here we estimate uncertainties of the derived L and e, specifically associated with the uncertainties of the key parameters, namely, ER SO2 /CO and OH concentrations. We also show the limitations of the analysis technique given in the present study. [32] ER SO2 /CO is used in deriving RSO x in equation (10), and OH concentrations determine k Total, together with L in equation (16). The uncertainty of ER SO2 /CO is estimated to be between 23 and +15% from the range of the lower and upper cases, respectively, as discussed in section 4.1. The estimated L and e changed by about 40 and 50%, respectively, if the assumed ER SO2 /CO was decreased by 23%. [33] Next, we evaluate the uncertainty associated with the assumed OH concentrations. The in situ measurements of OH were made during the winter period using a laserinduced fluorescence technique [Kanaya et al., 2001]. The OH concentrations derived by the empirical method were found to agree with those observed to within 40% in the winter period. The sensitivity to the assumed OH concentrations was evaluated by varying the OH concentration by 40%. There was no significant change in the estimated e. The values of e are close to those of e av for the moderately processed air. e av does not depend on pathways or rates of SO 4 formation from, as discussed in section 4.3, leading to insensitivity of e to OH. By contrast, the increases in OH lead to proportional increases in L. This is explained in equation (14), which is rewritten as L ¼ k½ohšð 1 1Þ e ð18þ With the increases in OH and L, both oxidation and depositional loss of SO 2 proceed faster maintaining the same correlation of SO 4 /SO x versus RSO x in Figures 8. [34] For the present analysis, we neglected possible contributions of aqueous-phase oxidation processes of SO 2 to the production of SO 4 by setting C = 0 in equations (14) and (15), as discussed above. Inclusion of this effect is equivalent to the increase in the OH concentrations. The dependence of L on OH indicates that L estimated for C =0 may be underestimated to some extent, although data for rainy or heavily overcast conditions were excluded, indicating that the derived L values are considered to be lower limits. On the other hand, the derived e values do not depend on the assumption of ignoring aqueous-phase oxidation processes, as discussed above. The uncertainty of the formation efficiency of SO 4 is therefore mainly determined by the uncertainty of ER SO2 /CO. [35] The data analysis techniques used in this study could be applicable to the data obtained in a location where the spatial distribution of the emission sources of SO 2 and its tracer is well characterized (that is, a reliable emission inventory of SO 2 and its tracer is available). However, if there are a lot of sources around the observation site that have a variety of emission ratios of SO 2 to its tracer compound, it would be quite difficult to interpret the data obtained using the techniques given by this study Comparisons of k Total and e With Previous Studies [36] We show the comparisons of the two parameters derived here, namely, k Total and e, with those using different methods given in the previous studies. The values of k Total and e and methods of the estimation are summarized in Table 4. [37] First, the values of k Total are compared with those obtained by previous studies. Eliassen and Saltbones [1975] compared trajectory calculations with observations at six stations in Western Europe based on an area emission inventory. Eliassen and Saltbones estimated k Total to be 9of13

10 0.08 ± 0.05 h 1, somewhat larger than the present study. Prahm et al. [1976] estimated k Total to be 0.07 ± 0.04 h 1 over the Atlantic Ocean in winter 1975 by the method using an air trajectory model and aerosol observations at the Faeroe Islands. Henmi and Reiter [1978] calculated the residence time of SO 2 over the eastern United States using annual climatological data and decay rates due to dry deposition and chemical conversion to be h ( h 1 ) and h ( h 1 ) in summer and winter seasons, respectively. Seasonal changes of k Total given by the present study are qualitatively consistent with the results given by Henmi and Reiter. The present k Total values are somewhat higher than those given by Henmi and Reiter, although the difference is within the uncertainties of the present estimate of about ±50%. [38] The values of e are also compared with those obtained by previous studies, although the previous studies did not directly derive e. e in Table 4 were calculated using the decay rates due to the removal and chemical conversion of SO 2 given by the previous studies. The e values from Henmi and Reiter [1978] showed similar seasonal variations, maximum in summer and minimum in winter. These comparisons suggest that e varied from 0.05 to 0.2 depending on the location and season. Aircraft observations also provided decay rates due to the removal and chemical conversion of SO 2 [e.g., Fisher and Callander, 1984; Gotaas, 1982]. However, because of shorter observation periods (several hours) over a variety of locations and meteorological conditions, previous studies based on aircraft observations showed a wide range of the residence times of SO 2, from about 10 to 100 h, and e from about 0.1 to 0.6. Therefore these values were not used for comparison with the present study Synthetic Interpretation of the Key Parameters (RSO x, e av, and t SOx ) [39] Here we discuss the interpretation of RSO x (remaining fraction of SO x ), e av (formation efficiency of SO 4 ), and t SOx (removal lifetime of SO x ) and their importance in atmospheric chemistry. Together with Tables 2 and 3, the seasonal variations of RSO x and e av are illustrated in Figure 9 for an overall understanding of the relevant processes. For the summer period, 18% of the emitted SO 2 was converted to SO 4, and about 60% was lost by deposition in Tokyo. On the other hand, for the winter period, only 3% of the emitted SO 2 was converted to SO 4, and about 80% was lost by deposition. The loss of SO x was estimated to occur with timescales of 1 day and 0.5 days for the summer and winter periods, respectively. [40] Possible factors controlling these seasonal variations are investigated. For the summer period, the SO 2 oxidation rate by equations (1), (2), and (3) was estimated to be about two times higher than for the winter period because of the higher photochemical activity. This contributed to the higher e av value for the summer period. Since the removal lifetime of SO 2 is much longer than that of SO 4, the smaller fraction of SO 2 led to the longer removal lifetime of SO x, resulting in the higher RSO x for the summer period. In addition, to the SO 2 oxidation rate, the boundary layer height could be another important factor affecting the seasonal variations of these parameters. The higher boundary layer height for summer reduced the loss rate of SO 2 by Figure 9. Seasonal variations of the average RSO x (RSO x ) and e av for the (a) summer and (b) winter periods. RSO x =1 and SO 4 /SO x = 0 upon SO 2 emission by definition. By the times of the observations, the RSO x (e av ) values decreased (increased) from these initial values. The fractions of SO 2 (black) and SO 4 (shaded) were derived using the average values of SO 4 /SO x for the observation periods (see text for details). dry deposition, leading to the longer t SOx and higher RSO x. Therefore SO 2 can be oxidized to SO 4 more efficiently before loss onto the surface, resulting in the higher e av for the summer period. Consequently, the seasonal variations of the values of RSO x, e av, and t SOx were likely controlled by the combined effects of photochemistry and vertical mixing within the boundary layer. [41] The key parameters (RSO x, e av, and t SOx ) discussed in this study are useful in validating the formation and loss processes of sulfur compounds calculated using chemical transport models. RSO x can be used to assess the amount of SO x deposited on the ground in source areas, and e av can be used to assess the amount of SO 4 exported from sources. The validation of the formation and loss processes of SO x is important for estimating the direct/indirect forcing by SO 4 over the source regions and their surroundings. 5. Summary and Conclusions [42] The formation of PM 1 SO 4 and removal of SO x near urban sources were investigated by the ground-based measurements of SO 2,SO 4, and CO at RCAST in Tokyo between April 2003 and February Comparisons between the emission inventory and the observations of CO, CO 2, and SO 2 in Tokyo revealed the following two points. First, the observed concentrations of SO 2 were affected predominantly by anthropogenic emissions (for 10 of 13

11 example, manufacturing industries) in the source region, while small fraction of the data (<30%) was affected by large point sources of SO 2, namely, power plants and a volcano, located south of the observation site. Second, although emission sources of CO in Tokyo are different from those of SO 2, the major emission sources of CO and SO 2 are colocated in the source region, indicating that CO can be used as a tracer of anthropogenic SO 2 emissions in Tokyo. We excluded data affected by some point sources of SO 2 and used CO as a tracer of anthropogenic SO 2 emissions in Tokyo. [43] The SO 4 /SO x ratio, which is a good measure to represent the formation of SO 4, showed a significant seasonal variation as follows: modal values were for the summer period and for the winter period. In order to quantify removal of SO x, we estimated the remaining fraction of SO x combining the linear regression slope of SO x -CO correlation and the emission ratio of SO 2 to CO in Tokyo. The average value of the remaining fraction of SO x (RSO x ) was estimated to be 0.40 and 0.21 for the summer and winter periods, respectively. Using the averaged value of SO 4 /SO x and RSO x, we estimated the values of the average formation efficiency of SO 4 (e av ), which is the average molar ratio of the observed SO 4 to the total SO 2 emitted. The e av value for the summer period (0.18) was much larger than that for the winter period (0.03). [44] A simple box model was developed in order to estimate the values of removal lifetime of SO x (t SOx ). The dominant loss process of SO x is dry deposition of SO 2,if wet deposition of SO 4 is neglected. The derived t SOx value for the summer period (26 h) was about two times longer than that for the winter period (14 h). The values of RSO x ; e av, and t SOx suggest that more than about 60% of SO x was removed in the boundary layer within about a day, and less than about 20% of the SO 2 formed SO 4 over Tokyo during the observation periods. The seasonal variation of these parameters is likely due to the seasonal variations of the boundary layer height and SO 2 oxidation rate. The results derived from the present study provide useful insights into the formation and removal processes of sulfur compounds exported from an urban area. Appendix A: Tokyo Emission Sources of SO 2 and CO in A1. Volcanic Source of SO 2 in Tokyo [45] In Appendix A1, we discuss volcanic source of SO 2 and its impacts on the observed concentration of sulfur compounds in Tokyo. SO 2 emission flux from Miyakejima Volcano has been observed by the Seismological and Volcanological Department of the Japan Meteorological Agency using a correlation spectrometer [Kazahaya, 2001]. The volcanic SO 2 emissions have continued to decrease since December 2000 and the average SO 2 emission flux during the observation periods was about one-forth of that for This corresponds to about 10% of the sum of the anthropogenic SO 2 emissions in the source region shown in Figure 2. The volcanic SO 2 and SO 4 were transported to the Japan Archipelago by the southerly winds associated with the subtropical high-pressure system dominating over the Pacific Ocean in the summer season. Transport of air masses strongly impacted by point sources is also possible, as discussed below. On the other hand, in winter, northwesterly wind prevailed and volcanic SO 2 and SO 4 were transported southward and rarely reached RCAST. The average concentrations of the volcanic SO 4 over Japan in February 2000 and August 2001 were estimated to be close to 0 and 3 mg m 3 (30 40% of anthropogenic and volcanic SO 4 ), respectively [Kajino et al., 2004]. Assuming that the concentration of volcanic SO 4 is proportional to the emission flux of volcanic SO 2, the relative contribution of volcanic SO 4 to the total SO 4 during the observation periods is estimated to be less than 10%. These estimations suggest that the observed sulfur concentrations for all observation periods were not strongly affected by Miyakejima Volcano. A2. Calculations of the Emission Ratio of SO 2 to CO in Tokyo [46] Appendix A2 focuses on identifying the emission sources of SO 2 and calculating the emission ratio of SO 2 and CO in the source region. Figures A1a and A1b show the relationships of annual emissions for SO 2 -CO and CO-CO 2 in the source region. The data points in the correlations correspond to the values estimated for each grid. The average SO 2 /CO and CO/CO 2 emission ratios from the emission inventory given by Streets et al. [2003] (hereafter referred to as Streets ER SO2 /CO and Streets ER SO2 /CO, respectively) are g g 1 = ppbv ppbv 1 and Gg Tg 1 = ppbv ppmv 1. The Streets ER SO2 /CO agrees well with the DCO/DCO 2 ratios observed by Takegawa et al. [2006], within the ranges of uncertainty of the inventory given by Streets et al. [2003]; 34% for CO and 7% for CO 2. However, the CO/CO 2 emission ratios in some areas reported by Kannari et al. [2004] are considerably lower than the observed DCO/DCO 2 ratios, as shown by the data points in the shaded circles in Figure A1b. These low CO/ CO 2 emission ratios correspond to the high SO 2 emission rates representing emissions from power plants in the coastal area (data in the shaded circles in Figure A1a). The nearly constant DCO/DCO 2 ratios observed at RCAST indicate that the observed CO and SO 2 values were not significantly influenced by localized power plant emissions. The data points in Figure A1a can be classified into two groups; CO/CO 2 emission ratios (mainly SO 2 emission > kg grid 1 year 1 ) lower than the observed DCO/DCO 2 ratios and data with CO/CO 2 emission ratios (mainly SO 2 emissions < kg grid 1 year 1 ) close to the observed DCO/DCO 2 ratios. The CO/CO 2 emission ratios close to the observed DCO/DCO 2 ratios represent emissions from the commercial institutions, manufacturing industries, waste incineration, and motor vehicles. The SO 2 /CO emission ratios are nearly constant for these sources in Figure A1a, indicating that although emission sources of CO and SO 2 are not necessarily the same, these sources are colocated. [47] The observed concentrations of SO 2 and CO should result from multiple injections from different sources as air mass passed over the source regions. Therefore we calculated ER SO2 /CO in the source region taking this into account. For these calculations, we considered only emission data 11 of 13