Net photochemical production rates of oxidant (Ox = O3 + NO2), F-D(Ox), were determined in Tokyo during the winter and summer of 2004 using observed and calculated HO2 radical concentrations. In both cases, calculated RO2 (organic peroxy) radical concentrations were used. The rates calculated using the two HO2 data sets are similar. In summer, morning F-D(Ox) values on smog days (those with midday O3 concentrations exceeding 100 ppbv) were higher than those on smog-free days (with typical midday O3 concentrations of 30 ppbv); however, the amount of ozone produced in a single day, as estimated by integrating F-D(Ox) over the daytime, was not significantly different for the two periods. This analysis suggests that the occurrence of smog events in the city center cannot readily be explained by day-to-day variations in the strength of in situ photochemistry. On smog days, the coupling of photochemistry and meteorology appears to be important, as air masses in which oxidants accumulated over successive days arrive at the city center at approximately midday, transported by land-sea breeze circulation. The average maximum daytime F-D(Ox) values in summer, 11 and 13 ppbv h−1 using observed and calculated HO2 levels, respectively, were only 1.5 and 2.2 times higher than those in winter (8 and 6 ppbv h−1). In winter, an underestimation of HO2 levels at high NO concentrations resulted in an underestimation of F-D(Ox) when calculated using modeled HO2. While the model predicted a volatile organic compounds (VOC)-limited regime for Ox production in winter, F-D(Ox) based on observed HO2 did not show features of the VOC-limited regime and only steadily increased with increasing NO mixing ratio, even when it exceeded 20 ppbv. In summer, the dependence of F-D(Ox) on nonmethane hydrocarbons (NMHCs) and NOx concentrations was similar in the two cases, in which observed and calculated HO2 levels were used. A VOC-limited regime, predicted on smog-free days, changed to a NOx-limited regime on smog days. The F-D(Ox) values determined for Tokyo are also compared with values for other cities.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Photochemical smog, which is characterized by elevated concentrations of ozone and other oxygenated species and low visibility due to elevated concentrations of aerosol particles, has been one of the most critical issues in large cities for more than three decades. Smog events usually arise from intense chemical production of ozone in the atmosphere coupled with favorable meteorological conditions. It is known that the photochemistry of NOx and hydrocarbons is essential for the chemical production of ozone. Previous simulations of photochemical reactions in chambers and model-based studies have yielded a detailed chemical mechanism in which OH, HO2, and RO2 radicals catalyze the production of ozone. These studies have described the nonlinear dependence of ozone production on concentrations of NOx and NMHCs; plots that show this relation are termed ozone isopleth diagrams [e.g., Milford et al., 1994].
 In general, the current tropospheric chemistry mechanism dictates that NMHCs/NOx ratios in the range of 10–20 ppbvC/ppbv, where ppbv is parts per billion by volume and ppbC is parts per billion by volume on carbon basis, are the most favorable in terms of the summertime production of O3 in the urban/suburban areas in the midlatitudes and that the rate of production decreases when the ratio is lower or higher than this range [e.g., Sillman, 1999]. For lower ratios, where the concentration of NMHCs is relatively low, it is expected that the production rate of ozone is limited by the NMHCs concentration (VOC-limited regime) and that the rate decreases with increasing NOx concentration. For higher ratios, where the concentration of NOx is relatively low, the production rate of ozone is mainly limited by NOx (NOx-limited regime), while still being positively dependent on the concentration of NMHCs. On the basis of this understanding, strategies aimed at reducing ozone have been investigated over recent decades.
 It should be noted, however, that it has been implicitly assumed in previous analyses that the radicals behave as described by the currently used tropospheric chemistry mechanism. The validity of this assumption is not necessarily apparent in the real atmosphere; if the radicals do not behave as described by the mechanisms, the production rate could differ from that predicted. In addition, the NMHC/NOx ratios that result in the maximum production of ozone could be different from that expected, and our understanding of the two regimes might need to be revised. Accordingly, we urgently require tests, based on direct measurements of radical concentrations, of the chemical mechanisms that operate in the urban atmosphere.
 Since 1970, when incidents related to photochemical smog were first reported in Tokyo, various observational and modeling studies have focused on the metropolitan area. Most previous field studies have highlighted the role of meteorology, including the role of the land-sea breeze, vertical stability of the atmosphere, and the inland transportation of air masses [e.g., Uno et al., 1984; Chang et al., 1989, 1990; Ueda et al., 1987; Wakamatsu et al., 1983]. In comparison, few studies have addressed the problem of smog from the perspective of atmospheric chemistry.
Wakamatsu et al. [1996, 1999] studied trends in NOx and NMHCs concentrations between 1978 and 1990 and suggested that a decrease in the NMHCs/NOx ratio in this period resulted in an increase in ozone formation potential. This in turn might have been responsible for the movement of the location of the daily maximum oxidant concentration away from areas in which the precursors were emitted. Over this period, the NMHCs/NOx ratio in high-emission area within Tokyo (0600–0900 local standard time (LST: UT + 9 h), April–September) decreased from 13 to 8 ppbC/ppb, associated with an increase in NOx and a decrease in NMHCs concentrations. From 1990, the NMHCs/NOx ratio continued to decrease, falling to 6 ppbC/ppb by 1996, accompanied by a reduction in both NOx and NMHC concentrations. The ratio increased slightly to 7 ppbC/ppb by 2002, reflecting a decrease in NOx concentrations that exceeded the magnitude of the decrease in NMHCs concentrations.
 The results of a statistical analysis contained in the final report of the Committee on Photochemical Oxidant Control [Tokyo Metropolitan Government, 2005] revealed that the most recent increase in the NMHCs/NOx ratio was one reason for a rapid increase in the number of events of high oxidant concentrations after 2000 and that a greater reduction in NMHCs emissions was required. These past studies of photochemistry in Tokyo neglected to investigate instantaneous ozone production rates. Moreover, the production rate has yet to be calculated on the basis of radical concentrations, in particular directly observed radical concentrations.
 We measured a sufficiently full suite of chemical species to enable a study of ozone and aerosol chemistry in Tokyo during the Integrated Measurement Program for Aerosol and oxidant Chemistry in Tokyo (IMPACT) campaign. Measurements of OH and HO2 concentrations were performed in the fourth phase of this campaign (IMPACT-IV, January–February 2004) and in a Lagrangian campaign (IMPACT-L, July–August 2004). An accompanying paper compares the observed concentrations of OH and HO2 radicals with model predictions [Kanaya et al., 2007]. In the present paper, we extend the analysis to highlight the influence of radical chemistry on ozone production.
 In section 2, we present a method to estimate the net production rate of oxidant based on either observed or calculated HO2 radicals. Section 3.1 considers temporal variations in oxidant concentrations and the estimated production rate. Production rates were compared between smog and smog-free conditions and between summer and winter. Section 3.2 deals with the behavior of the rate of HO2 + NO reaction as a function of NO concentration for both winter and summer. This analysis is extended in section 3.3, which considers the dependence of the measured concentrations of OH and HO2, the reaction rate of HO2 + NO, and the production rate of oxidant on concentrations of NO and NMHCs. We examine whether the observed radical concentrations and reaction rates determined on the basis of observed radical concentrations depend on NO and NMHCs concentrations in the same manner as modeled concentrations and rates. Section 3.4 deals with the dependence of the oxidant production rate on concentrations of NOx and NMHCs; this is done to make the analysis compatible with conventional ozone isopleth diagrams. In section 3.5, the rates determined for Tokyo are compared with those for other cities throughout the world where the rates have been determined in a similar way.
2. Calculations of F-D(Ox)
 NO2 is present in the urban atmosphere at concentrations comparable to O3, with each species being rapidly converted to the other. Thus, in the present study oxidant (Ox) is defined as a group of O3 and NO2 (the production of Ox is discussed in the following sections):
 It has been established that the formation of Ox occurs in the daytime when NO is oxidized into NO2 by the peroxy radicals HO2 and RO2. The produced NO2 molecules undergo photolysis, producing O(3P) and thus O3 via a reaction between O(3P) and O2:
where k is the reaction rate coefficient of the pertinent reaction indicated by the suffix and ϕ is the yield of NO2 from the RO2 + NO reaction. The destruction of Ox occurs either by the reaction of O(1D), which is formed by the photolysis of O3, with H2O, by the reactions of OH, HO2, or olefins with O3, or by the reaction of OH with NO2:
 Concentrations of OH and HO2 are key parameters in both the formation and destruction of Ox and thereby in the net production rate, F(Ox)-D(Ox) (herein stated simply as F-D(Ox)). In this paper, F-D(Ox) is calculated in two ways, using observed or calculated HO2 concentrations for the IMPACT campaigns in winter and summer.
 Observations of OH and HO2 concentrations by laser-induced fluorescence at the Komaba site (35°39′N, 139°41′E) in Tokyo are described in the accompanying paper [Kanaya et al., 2007]. Only the daytime data are used in the present analysis. A photochemical box model used to predict OH and HO2 levels is also described in the accompanying paper. Briefly, the model is based on the Regional Atmospheric Chemistry Mechanism (RACM) [Stockwell et al., 1997] with updated kinetic parameters. The heterogeneous loss of HO2 is not taken into account in the base runs. For the winter and summer campaigns, we used the following observed concentrations and parameters as constraints for the model: O3, CO, H2O, SO2, NO, NO2, CH4, NMHCs, PANs, temperature, ambient pressure, and J values. Measured wintertime concentrations of HONO, HCHO, and CH3CHO were also used to constrain the model. Summertime concentrations of HONO were calculated in the model. Estimated concentrations of HCHO and CH3CHO were used for the summer period. For the winter campaign, radical concentrations estimated by model run 2, in which the concentrations of internal olefin (OLI in RACM) and reactive alkanes (HC8) were increased by factors of 3 and 5, respectively, were used to calculate F-D(Ox) values, as they showed a better agreement with observed OH and HO2 levels during the daytime than the results obtained from the base run (run 1).
 Concentrations of RO2 (organic peroxy) radicals were not measured during the campaign. Thus, the model-derived RO2 concentrations were always used even in the calculations of F-D(Ox) based on observed HO2. The model runs predicted that the [HO2]/[HO2 + RO2] ratio at midday was in the range of 0.3–0.5 and that the contribution of RO2 to ozone production was less than that of HO2, on the basis of the fact that the rate coefficients of the reactions of the dominant RO2 with NO are lower than for HO2 in the mechanism (see Figure 4). The daytime (0900–1500 LST) median values of HO2 and RO2 mixing ratios were 0.54 and 0.91 pptv in Run 2 in winter, respectively, and 5.3 and 6.9 pptv in the model for summer.
 Another set of model runs, hereafter termed “sensitivity model runs,” was performed to construct model-derived contour diagrams that illustrate how the concentrations of OH and HO2, the rate of the HO2 + NO reaction, and F-D(Ox) depend on NO and NMHCs concentrations given the currently used tropospheric chemistry mechanism. In these runs, concentrations of species other than NOx and NMHCs observed at 1100 LST on 28 January and 1000 LST on 9 August were used as constraints. The concentrations of O3, partial pressure of H2O, J(O1D) and temperature values at the two timings (31.2 ppbv, 2.7 hPa, 7.3 × 10−6 s−1, and 8.3°C for 1100 LST on 28 January and 26.6 ppbv, 23.8 hPa, 2.3 × 10−5 s−1, and 31.7°C for 1000 LST on 9 August) were close to the median values of the midday (1000–1400 LST) data in winter (25.0 ppbv, 3.2 hPa, 6.1 × 10−6 s−1, and 8.2°C) and those of 6-h (0900–1500 LST) data under clear-sky conditions (J(NO2)/J(NO2)clearsky of >0.7) in summer (32.4 ppbv, 24.5 hPa, 2.5 × 10−5 s−1, and 31.5°C), respectively. NOx and NMHC concentrations were artificially shifted from the observed values at these times to cover wide ranges of imaginary NO and NMHCs concentrations. We used the shifted NO and NMHCs concentrations to calculate the radical concentrations, rate of HO2 + NO, and F-D(Ox). By interpolating the results, we obtained smooth contour diagrams showing the degree of sensitivity of these values to NOx and NMHCs concentrations. Concentrations of NMHCs categorized into HC3, HC5, HC8, ETE, OLT, OLI, DIEN, ISO, TOL, and XYL (see the accompanying paper [Kanaya et al., 2007, Table 2] for details of hydrocarbon categories) were increased or decreased by the simultaneous use of a single factor. At the same time, CO, CH4, and ethane concentrations were scaled by the same factor with consideration of their background concentrations. A different independent factor was used to scale NOx concentrations. Concentrations of HONO in winter, constrained by observations in the model, were multiplied by the same scaling factor as that used for NOx. As concentrations of HONO were not measured during summer, they were predicted in the model using the assumed amount of NOx. The sensitivity runs undertaken on the basis of these assumptions do not show a better performance in simulating OH andHO2 levels at each time of day than point-specific model runs described in the accompanying paper [Kanaya et al., 2007], which were fully constrained by ancillary observations undertaken at the relevant times of day. However, the results of the sensitivity runs are useful in studying the general behavior of the quantities that depend on the concentrations of NMHCs and NOx.
3. Results and Discussion
3.1. [Ox] and F-D(Ox) During the Campaigns
 Temporal variations in F-D(Ox) values calculated using observed and modeled HO2 concentrations during the winter campaign are shown in Figure 1a. For reference, concentrations of O3 and Ox (O3 + NO2) observed at the Komaba site are shown in Figure 1b. The values of F-D(Ox) derived from observed and calculated HO2 concentrations agreed well with each other. The typical midday peak values of F-D(Ox) were in the range of 5–10 ppbv h−1 for both data sets, although higher values (>10 ppbv h−1) were obtained on two days, 21 and 27 January. Agreement between the two sets of values was usually better in the afternoon than in the morning. The observed Ox concentration showed regular diurnal variation, with an average daytime increase of 10 ppbv. This increase can be explained by photochemical formation, although vertical and horizontal mixing and dry deposition might also have significantly influenced the Ox concentration.
 The corresponding figures for the summer campaign are shown in Figures 1c and 1d. The two sets of estimated F-D(Ox) values showed stronger agreement than for the winter case. The midday F-D(Ox) peak was typically in the range of 10–15 ppbv h−1, with both estimates exceeding 20 ppbv h−1 on 6 and 9 August. The observed daily maximum concentrations of ozone were 12–93 ppbv (31 ppbv as median value) during the period of 26 July to 11 August, but increased to 120, 103, and 141 ppbv on the last three days of the campaign (12, 13, and 14 August). On the three days, the environmental standard level in Japan of 60 ppbv as an hourly value was significantly exceeded. In this study, the former and latter periods are denoted “smog-free” and “smog” periods, respectively.
3.1.1. Comparison Between Smog and Smog-Free Days in Summer
 One hypothesis concerning summertime was that the F-D(Ox) values increased during the smog period by a factor of >2 as inferred from the ratio of Ox concentrations during the smog period to the smog-free period; however, the results of our analysis do not support this hypothesis. In Figure 2, the average diurnal variations in F(Ox) and F-D(Ox) calculated using observed HO2 concentrations are shown separately for the smog and smog-free days. F(Ox) and F-D(Ox) values on the smog days in the morning (0500–0800 LST) were at least twice as high as those for smog-free days; the integrated amount of F-D(Ox) over the period 0500–0900 LST on smog days was higher by 15 ppbv. In contrast, after 1000 LST the F(Ox) and F-D(Ox) values on smog days were similar to or even lower than those on smog-free days. The integrated amounts of F(Ox) and F-D(Ox) over the day (using observed HO2 concentrations) were 117 ± 11 (1σ range) and 85.1 ± 9.1 ppbv on smog days and 107 ± 11 and 90.1 ± 9.8 ppbv on smog-free days, respectively, with no significant difference between the two periods. D(Ox) showed daytime maximum values of ∼5 ppbv h−1 on smog days, slightly higher than the ∼2 ppbv h−1 recorded on smog-free days, compensating for the higher values of F(Ox) on smog days. However, the small difference in D(Ox) does not fully explain the fact that F-D(Ox) on smog days was similar to that on smog-free days. These results did not differ significantly when comparing F(Ox) and F-D(Ox) values calculated using modeled HO2 concentrations.
 On 12, 13, and 14 August, the Ox concentration recorded at Komaba increased from 0500 to 1300 LST by 104, 81, and 116 ppbv, respectively. The integrated values of F-D(Ox) over the 8 h were just 89, 65, and 58 ppbv, respectively, on these three days, as calculated using modeled HO2 levels. The corresponding values obtained using observed HO2 levels were 78 and 51 ppbv for 12 and 13 August, being smaller than the increase in Ox concentrations (the value for 14 August was not estimated because HO2 was not measured in the early morning). Thus, the local photochemistry cannot quantitatively explain the increase in oxidant concentrations on these days. On the basis of the above analyses, we conclude that the local photochemistry alone cannot readily explain the high ozone or Ox concentrations recorded during the smog period.
 The occurrence of smog in central Tokyo appears to be strongly controlled by meteorological factors in combination with photochemistry. Land-sea breezes occurred under a high-pressure system during the smog period. It is likely that pollutants were trapped in the air mass that remained over the Tokyo area, with only slight movement of the air mass associated with weak land-sea breezes; ozone was produced within this largely stationary air mass over a couple of days. The air mass that contained elevated oxidant concentrations was advected to the Pacific Ocean by the nighttime land breeze but reentered central Tokyo with the sea breeze on the following day [e.g., Uno et al., 1984]. Low NOx/NOy ratios were observed at the Komaba site on the smog days, suggesting that the air mass was aged and had been influenced by photochemistry for an extended period (Y. Kondo et al., Formation and transport of oxidized reactive nitrogen, ozone, and secondary organic aerosol in Tokyo, submitted to Journal of Geophysical Research, 2008). It is also likely that ozone remained at relatively high concentrations in the residual layer (above the nocturnal boundary layer) throughout night and that vertical mixing in the following morning period resulted in increases in the ozone level observed at surface level.
 In contrast, smog-free days were marked by persistent southerly winds, bringing a continuous supply of clean air from the Pacific Ocean to the Tokyo Metropolitan Area. On these days, a buildup of ozone concentration was commonly observed in Saitama Prefecture, located to the north of Tokyo. The O3 concentration observed at the Kisai site in Saitama, approximately 50 km north of the Komaba site, increased to ∼100 ppbv on 9 August during the afternoon [Takegawa et al., 2006]. In comparison, the peak ozone concentration at Komaba on this day was only 38 ppbv. Considering the observed wind speed of ∼3 m s−1, the transport time of the air mass between the two sites is estimated to be 4–5 h [Takegawa et al., 2006]. The 60 ppbv difference in O3 concentrations is in rough agreement with the ozone production rate of 15 ppbv h−1 without considering mixing and dry deposition.
3.1.2. Comparison Between Summer and Winter
Figures 3a and 3b show composite diurnal variations in F-D(Ox) and the rate of the HO2 + NO reaction in winter and summer, respectively. For each quantity, values are shown for observed and calculated HO2 concentrations. Before noon in winter and before 1000 LST in summer, the rate of the HO2 + NO reaction and F-D(Ox) are underestimated when based on the modeled HO2 concentrations. This underestimation was caused by an underestimation of HO2 concentrations at high NO concentrations, as investigated in the accompanying paper [Kanaya et al., 2007]. This will be discussed in more detail in sections 3.2 and 3.3 in examining the ozone production regimes. During 1000–1600 LST in summer, an overestimation of HO2 levels by the model resulted in an overestimation of the rate of HO2 + NO reaction and F-D(Ox). The accompanying paper demonstrates that the degree of overestimation was similar to the degree that could be explained by the potential maximum rates of heterogeneous loss of HO2 on aerosol surfaces. In winter, the average diurnal variations in F-D(Ox) and the rate of the HO2 + NO reaction based on calculated HO2 levels showed maxima at 11H (the hour between 1100 and 1200 LST) or 12H (the hour between 1200 and 1300 LST). In contrast, the values based on observed HO2 levels showed earlier maxima at 9H (the hour between 0900 and 1000 LST) in winter. In summer, their maxima occurred in midday in both cases, although the rates based on the observed HO2 levels had shoulders in the morning. In winter, the daytime average maximum values of F-D(Ox) and the rate of the HO2 + NO reaction were 6.0 and 4.3 ppbv h−1, respectively, as calculated using modeled HO2 levels; the values calculated using observed HO2 levels were 7.7 and 6.2 ppbv h−1, respectively (Table 1). In summer, the average values were 13.1 and 8.7 ppbv h−1 using calculated HO2 and 11.2 and 6.8 ppbv h−1 using observed HO2. The integrated amount of Ox produced over a day in winter was estimated to be 37.5 ppbv using calculated HO2 and 49.6 ppbv using observed HO2, of which 29.3 and 39.2 ppbv were derived from the HO2 + NO reaction. The corresponding amounts of Ox in summer were 91.2 and 92.3 ppbv, respectively, of which 62.0 and 61.7 ppbv were derived from the HO2 + NO reaction. It is interesting to note that F-D(Ox) and the rate of the HO2 + NO reaction in summer are only 1.1–2.4 times larger than those in winter. It should be noted that the determined rates are calculated for the air mass at ground level and do not necessarily represent ozone production within the entire boundary layer.
Table 1. F-D(Ox), HO2 + NO Rate, and F(Ox) in Tokyo During Summer and Winter Compared to Values for Other Citiesa
HO2 + NO Rate
Daytime Peak (ppbv h−1)
6-h Median (ppbv h−1)
Daily Production (ppbv)
Daytime Peak (ppbv h−1)
6-h Median (ppbv h−1)
Daily Production (ppbv)
Daytime Peak (ppbv h−1)
6-h Median (ppbv h−1)
Daily Production (ppbv)
Values based on modeled HO2 levels are shown in parentheses.
Figures 4a and 4b show a breakdown of the net production of Ox on 29 January and 1 August. At all times during the two days the HO2 + NO reaction made the largest contribution to F(Ox), in excess of 50%. The OH + NO2 reaction was the largest term in D(Ox) in both winter and summer. The olefin + O3 reaction and O3 photolysis were of secondary importance in D(Ox) in winter and summer, respectively.
 Here we estimate the contribution of isoprene with biogenic origin [Shirai et al., 2007] to ozone production. The contribution of the ISOP (peroxy radical formed from isoprene) + NO reaction to F(Ox) was a maximum of 13% at midday in summer; however, isoprene chemistry should produce HO2 and thereby contribute further to F(Ox) via the HO2 + NO reaction. The maximum contribution of the rate of the OH + isoprene reaction to the total rate of OH reactions that propagate radical chain reactions was 28%, which corresponded to the maximum contribution of biogenic isoprene to ozone production at midday.
3.2. Dependence of HO2 + NO Reaction Rate on NO
 As mentioned in the accompanying paper and in the preceding subsection, our model underestimated daytime HO2 levels at high NO concentrations in winter. The consequence of this underestimation in terms of the ozone production rate and regimes is discussed in sections 3.2 and 3.3. First, we consider the dependence of the rate of the HO2 + NO reaction on the NO concentration in winter (Figure 5a) and in summer (Figure 5b). In winter, the rate of the HO2 + NO reaction obtained using modeled concentrations of HO2 shows a maximum when the NO mixing ratio is around 10 ppbv, decreasing with further increases in the NO mixing ratio. This trend is observed in both model runs (run 1 and 2), where different concentrations of NMHCs are assumed.
 In contrast, the reaction rate obtained using observed concentrations of HO2 increases steadily with increasing NO mixing ratio, even when it exceeds 20 ppbv. This contrast in behavior stems from an underestimation of HO2 levels at high NO concentrations in winter, as shown in the accompanying paper [Kanaya et al., 2007, Figure 9b]. During daytime in summer, this difference is not clearly recognized, partially because high NO concentrations were not observed. The steady increase in the reaction rate with increasing NO concentrations is not unique for Tokyo in winter; similar results have been reported for Nashville and New York City, USA, in summer [Martinez et al., 2003; Ren et al., 2003] and in New York City in winter [Ren et al., 2006].
 In Figures 5a and 5b, however, which only show the dependence on NO concentrations, we were unable to investigate the dependence on NMHCs concentration, another important precursor of O3 that potentially covaries with NO concentration or even varies independently of NO concentration. Therefore, in the following subsection we undertake an extended analysis of the dependence of the rate on the concentration of NMHCs simultaneously with an analysis of the dependence on NO concentration. This analysis is also useful in linking the analysis of NO dependence to an analysis of ozone production regimes using an O3 isopleth diagram.
3.3. Dependence of OH, HO2, HO2 + NO Rate, and F-D(Ox) on the Concentrations of NMHCs and NO
 The dependence of concentrations of OH and HO2, rate of the HO2 + NO reaction, and F-D(Ox) on the concentrations of NMHCs and NO is shown for winter (Figure 6) and summer (Figure 7). The observed data points (10-min data for 1000–1400 LST for the winter period and 0900–1500 LST for the summer period) are indicated by circles whose colors are dependent on the magnitude of the plotted quantity. In summer, only data points under clear-sky conditions (J(NO2)/J(NO2)clearsky of >0.7) are used in the analysis. In winter, most of the data were associated with clear-sky conditions and no further data selection was undertaken. The color contours in Figures 6 and 7 are produced from the sensitivity model runs described in section 2. The color codes used for the contours are the same as those used for the circles; agreement between the colors of the circles and the areas in which the circles are located indicates agreement between the observed and modeled levels of the plotted quantity. However, it should be noted that quantitative comparison between the observed and modeled concentrations of OH and HO2, and the rate of HO2 + NO reaction and F-D(Ox) is not intended here; rather, the dependence of these observed and modeled values on the concentrations of NO and NMHCs is compared.
3.3.1. Wintertime Case
 For OH concentrations in winter (Figure 6a), the colors of the circles roughly match those of the areas in which the circles are located, indicating a general agreement between observations and the model predictions. The observed OH concentration has a weak decreasing trend with increasing NO concentration in the range of 1–10 ppbv, as with the model predictions. The modeled OH concentration increases slightly with increasing NMHCs concentration because the enhanced production of OH via the OLI (internal olefins) + O3 reaction affects OH concentrations more strongly than the amplified OH loss. This feature is not clearly seen with the observed data.
 For HO2 concentrations in winter (Figure 6b), the observed data points associated with low NO concentrations (1–5 ppbv) and low NMHCs concentrations (100–200 ppbvC) plot in areas of the same color; however, the observed HO2 levels associated with higher NO concentrations (>10 ppbv) and higher NMHCs concentrations (>200 ppbvC) are higher than the modeled levels, indicating that the model tends to underestimate HO2 levels in these concentration ranges of NO and NMHCs.
 This discrepancy is more pronounced in the rate of the HO2 + NO reaction (Figure 6c). Using the dashed black line in Figure 6c as a guide, which represents covariations in the observed NO and NMHCs concentrations, the model predicts that the rate of the HO2 + NO reaction does not change significantly; the colors of the area along the line are green or yellow-green and do not change significantly. In contrast, the reaction rate calculated using observed HO2 levels increases with increasing NO and NMHCs concentrations. In Figure 5a, we examined the same difference between the model- and observation-based rate of the HO2 + NO reaction but without considering the concentration of NMHCs. Next, we observe the dependence of the reaction rate on NO concentration along a horizontal line with a constant NMHCs level of 200 ppbvC (Figure 6c). The rate increases if based on observed HO2 concentrations but decreases if based on model-predicted concentrations. The dependence of the rate on the concentration of NMHCs along a vertical line with constant NO levels is not clear if based on observed HO2 concentrations, whereas the model suggests an increase in the rate with increasing concentration of NMHCs.
 The difference between observation- and model-based results evident in Figure 6c is also apparent in the F-D(Ox) diagram (Figure 6d). The contours in Figure 6d reveal the NMHCs/NO ratio at which F-D(Ox) is maximized (“ridge”); this is accompanied by adjacent NOx-limited and VOC-limited regimes. The position of the ridge for F-D(Ox), with consideration of NO and NMHCs concentrations, is similar to that found in the OH plot (Figure 6a). For the observed concentration ranges of NO and NMHCs, the model suggests that F-D(Ox) is VOC-limited. In this regime, the model states that F-D(Ox) increases with increasing NMHCs concentrations (∂(F-D(Ox))/∂[NMHCs] > 0) but decreases with increasing NO concentration (∂(F-D(Ox))/∂[NO] < 0); however, F-D(Ox) based on the observed HO2 levels does not show the predicted behavior. For example, when the data points along a constant NMHCs level (e.g., 200 ppbvC) are considered, the color of the data points changes from orange to red with an increase in NO; namely, ∂(F-D(Ox))/∂[NO] > 0. This is NOx-limited-like behavior, being strongly inconsistent with the model prediction. This discrepancy stems from the model's underestimation of HO2 levels associated with high NO and NMHCs concentrations. Potential causes of this discrepancy are investigated in the accompanying paper [Kanaya et al., 2007]. Briefly, one possibility is that HOx-NOx chemistry, including HNO4 chemistry, is not correctly understood, while another possibility is the presence of an HOx source unaccounted for in the model whose rate is proportional to the NO concentration. It should be noted that the chemistry of HONO has already been taken into account in the model according to its observed concentrations in winter; however, any unexpected behavior of HONO, if present, might be able to explain the difference in the model-based and observation-based F-D(Ox) values. The dependence of F-D(Ox) on the concentration of NMHCs along vertical lines with constant NO levels is not clearly visible from the observed data points.
3.3.2. Summertime Case
 In the plots for summer (Figure 7), observed NO and NMHCs concentrations do not show a strong correlation compared with that in winter; thus, the observational data points show a largely random scatter. For OH concentrations in summer (Figure 7a), observed data for NO concentrations higher than 1 ppbv show a predictable behavior with NO, in that OH concentrations decrease with increasing NO concentrations. In contrast, the colors of the data points for NO concentrations less than 1 ppbv are more strongly red than the colors of the surrounding areas, indicating underestimations by the model. These data points are for smog days (12–14 August).
 As shown in the accompanying paper [Kanaya et al., 2007], the model did not reproduce the observed OH levels on these smog days, even when all of the ancillary observations were used as constraints in the model. An additional reason for the underestimation may be the difference in the production rate of radicals. In the sensitivity model runs performed to construct the contours, O3 concentrations were assumed to be 27 ppbv independently of NO and NMHCs concentrations, as observed at 1000 LST on 9 August, whereas the observed O3 concentrations on smog days were much higher (>100 ppbv). The high concentrations of ozone and possibly carbonyl species on these days resulted in higher radical production rates [Kanaya et al., 2007, Figure 11], affecting the OH concentrations. Therefore, a meaningful comparison was not possible for these days, especially for OH. At any constant NO concentration of >1 ppbv, the model predicts that OH concentrations are largely independent of NMHCs concentrations; this is consistent with the behavior of observed OH.
 The model predicts that HO2 concentrations in summer decrease with increasing NO concentrations at any constant NMHC levels (Figure 7b). This behavior is also found in the observed HO2. An overestimation of HO2 concentrations is found when NO concentrations are lower than 1 ppbv. This can be explained by the heterogeneous loss of HO2 on aerosol surfaces, as reported in the accompanying paper [Kanaya et al., 2007]. A slight underestimation of HO2 concentrations is found when NO concentrations are higher than 10 ppbv. This might represent the same phenomenon as that found in winter; however, the number of data points with high NO concentrations was small in summer and thus it was difficult to study the chemistry with high NO. In a vertical cross section with a constant NO level of 2 ppbv, both the observations and the model record a slight increase in HO2 concentrations with increasing NMHCs concentrations.
 The plot that shows the rate of the HO2 + NO reaction (Figure 7c) does not show the severe mismatch in color coding recorded for winter, while the rates based on observations are lower in the yellow area and higher in the light blue area (with high NO concentrations). In a horizontal cross section with a constant NMHCs concentration of 200 ppbvC, the rate based on observed HO2 levels shows a maximum with a yellow level (14.7–21.5 ppbv h−1) when NO concentrations were 2–3 ppbv, and a decrease with further increases in NO concentration. This is similar to the model prediction, although the modeled maximum occurred at slightly lower NO concentrations of 1–2 ppbv and the rate decreased with NO more sharply in the model than in measurements as found in Figure 5b. The concentrations of VOCs might have been higher than assumed in the base run for summer, as we assumed higher HC8 and OLI concentrations in winter. Increased VOCs in summer would shift the model results in the direction of NOx-sensitive chemistry. This may also improve model-measurement agreement in Figures 5b and 7c.
 Finally, in the F-D(Ox) plot (Figure 7d) the model predicts a “ridge” with adjacent regimes (NOx-limited and VOC-limited), as seen in Figure 6d. The position of the ridge for F-D(Ox) in terms of NO and NMHCs concentrations is similar to that found in the OH plot (Figure 7a). The general behavior of F-D(Ox) using observed HO2 levels is similar to that obtained using calculated HO2, except that several green circles fall into an orange area and orange circles are located in a yellow area. The locations of the green circles may reflect the heterogeneous loss of HO2, as stated above. In areas that contain data points, the model predicts that the Ox production regime is generally VOC-limited, changing to a NOx-limited regime under smog conditions (data for smog days are shown in Figure 7d by symbols with black outlines). Similar regimes are deduced from F-D(Ox) using observed HO2 concentrations. In the high NO concentration range (>3 ppbv), we find the following behavior: ∂(F-D(Ox))/∂[NO] < 0 and ∂(F-D(Ox))/∂[NMHCs] > 0, as predicted by the model.
3.4. Dependence of F-D(Ox) on NOx and NMHCs
 In this subsection, the behavior of F-D(Ox) is again investigated, using NOx concentrations as the x axis rather than NO concentrations. This makes the analysis more compatible with conventional ozone isopleth diagrams.
Figure 8a shows the dependence of F-D(Ox) on NOx and NMHCs concentrations in winter. The model predicts that the ridge is associated with a NMHCs/NOx ratio of ∼50 ppbC/ppb. Almost all of the features apparent in Figure 6d are retained in Figure 8a. In the area with observed data points, the model suggests that ozone production is limited by NMHCs concentrations; however, F-D(Ox) based on the observed HO2 levels increases with increasing NOx concentration, showing NOx-limited-like behavior. This is clearly inconsistent with the model predictions.
Figure 8b shows the dependence of F-D(Ox) on NOx and NMHCs concentrations in summer. The model predicts that the ridge is associated with a NMHCs/NOx ratio of ∼20 ppbC/ppb. This is lower than that in winter and is roughly consistent with the range suggested by Milford et al.  for summer (10–20 ppbC/ppb). The critical ratio of ∼20 ppbC/ppb proposed in the present study is slightly higher than the ratio of 10 ppbC/ppb proposed by Wakamatsu et al.  as representing the maximum ozone concentration in Tokyo during 1981. These discrepancies between the different studies may reflect the fact that the studied quantity differs (ozone production rate in this paper and the maximum ozone concentration given by Wakamatsu et al. ), that the composition of NMHCs in 1981 differs to that observed in the present study, and that the two studies employed a different chemical mechanism (Wakamatsu et al.  employed a Carbon–Bond Mechanism). Kannari  estimated NMHCs/NOx ratios of 7–16 in summertime Tokyo for which the ozone production regime changed from NOx-limited to VOC-limited; this was achieved by analyzing the relationship between observed O3 concentrations and NMHCs/NOx ratios on weekdays and weekends.
 From the area that contains observational data points under smog conditions (color circles with black outlines in Figure 8b), it is deduced that the smog conditions are associated by higher NMHCs concentrations (200–500 ppbvC) than those recorded in the smog-free period. The range of NOx concentrations (4–20 ppbv) was similar to that for the smog-free period; thus, the NMHCs/NOx ratios were slightly higher on smog days than on smog-free days. For these ranges of NOx and NMHCs concentrations, our sensitivity model suggests that ozone production takes place under a “ridge” or NOx-limited conditions and that its rate is higher (>20 ppbv h−1) than that on smog-free days; however, F-D(Ox) based on observed HO2 levels are not as high, being less than 15 ppbv h−1. This result reflects the fact that NO/NOx ratios observed on smog days were lower than those in the sensitivity model; consequently, the NO levels associated with a given level of NOx were lower than those in the sensitivity model, leading to slower rates for the HO2 + NO and RO2 + NO reactions.
 As the Komaba site is separated from the source region of NO, it is likely that the NO concentration in the air mass arriving at the site has already been affected by photochemical processing. The model results suggest that the ozone production rate close to the NOx source region might have been higher during the smog period than the smog-free period. As well as meteorological factors, the slight increase in the NMHCs/NOx ratio and potentially active photochemistry associated with the higher ratio value could be important factors to high O3 levels. This topic is analyzed in more detail in a separate paper (Kondo et al., submitted manuscript, 2008), which reports that the ozone production efficiency, the number of ozone molecules produced by a NOx molecule, was higher on the smog days.
 Although the smog days (12–14 August) occurred on a Thursday, Friday, and Saturday, they took place during the summer vacation period in Japan, when industrial activity within the city was significantly below average. It is likely that the level of emissions from central Tokyo at this time changed from that expected on normal weekdays to that usually found on weekends, leading to an increase in the NMHCs/NOx ratio [Kannari, 2006]. Kannari  reported that the ozone production regime in Tokyo changes from VOC-limited on weekdays to NOx-limited on weekends; this is consistent with the results of the present study.
 The concentrations of NMHCs might also have increased via accumulation in the air mass that persisted over Tokyo for a couple of days in association with land-sea breezes, as discussed above. The averaged reactivity toward OH per the amount of NMHCs (s−1 ppbvC−1) slightly decreased during the smog period (by 14–27%), because reactive NMHCs were selectively lost during the aging of the air mass. This lower reactivity could be another reason for the fact that the F-D(Ox) based on observed HO2 levels were lower than those based on calculated HO2. However, this slight change did not alter the above discussion significantly.
 On the basis of the above analyses, we conclude that the production of Ox on smog-free days in summer is limited by NMHCs; consequently, a reduction in the concentration of NMHCs is advisable to reduce the production rates of Ox in central Tokyo. The modeled contours in Figure 8b indicate that a reduction in NMHCs concentrations is helpful even on smog days under the “ridge” or NOx-limited conditions. This finding is largely consistent with the recommendation made by the Tokyo Metropolitan Government in 2005 that a greater reduction in NMHCs concentrations is needed in addition to a reduction in NOx. In addition, we point out that a reduction in NOx, if unaccompanied by a reduction in NMHCs, can result in an unwanted increase in the ozone production rate on smog-free days.
3.5. Intercity Comparisons
 In this section, F-D(Ox) and related values determined in Tokyo are compared to the values obtained in a similar way for other cities, using observed and calculated radical concentrations (Table 1). In Nashville and New York City, USA, the rate of the HO2 + NO reaction showed early morning maximum at around 14 ppbv h−1 [Martinez et al., 2003] and 20 ppbv h−1 [Ren et al., 2003], respectively, when based on observed HO2 levels. The shift in the timing of the maximum from noon to an earlier period is similar to that found in Tokyo during winter. The rates obtained in these earlier studies are higher than those in Tokyo in summer (6.8 ppbv h−1). The average daily ozone production from HO2 was about 130 ppbv in Nashville and 140 ppbv in New York City, again higher than the 62 ppbv recorded in Tokyo during summer. The HO2 + NO rate calculated using the modeled HO2 levels in New York City was slightly lower than the rate obtained using observed HO2 levels; however, the midday maximum rate was ∼12 ppbv h−1, still slightly higher than that in Tokyo (8.7 ppbv h−1).
 Daytime NOx (around 20 ppbv) and NMHCs levels (100 ppbvC) in summer in New York City [Ren et al., 2003, 2006] were higher and lower than those in Tokyo during summer (12 ppbv and ∼190 ppbvC), respectively. The ozone production rate for New York City, when estimated from the contours in Figure 8b on the basis of the NMHCs and NOx levels, is lower than that in Tokyo; this is inconsistent with the above comparison. This discrepancy might stem from the different compositions of NMHCs in the two cities and the higher mixing ratio of ozone in New York City (∼50 ppbv), resulting in higher rates of radical production and in turn higher rates of ozone production. In both Nashville and New York City in summer, observed HO2 levels were underestimated at high NO concentrations by model predictions, as was found in Tokyo during winter; this resulted in higher F-D(Ox) values where based on the observed HO2 concentrations.
 In Mexico City, Mexico, the median rate of the HO2 + NO reaction calculated from measured HO2 levels peaked at 48 ppbv h−1, while the rate calculated from modeled HO2 levels peaked at 86 ppbv h−1. Both peaks were broad and occurred close to 1000 CST (Central Standard Time). Peak values on a number of days were greater than 100 ppbv h−1. The cumulative daily ozone production from HO2 was 319 ppbv when calculated using measured HO2 [Shirley et al., 2006], almost 5 times larger than that in Tokyo in summer. One of the reasons for these differences is that the production rate of the radicals, to which the photolysis of high concentrations of HCHO made a contribution, was markedly high in Mexico City.
Emmerson et al.  reported that the midday (1100–1500 LST) average F(Ox) during the UK TORCH campaign, performed at a site located ∼50 km northeast of London in the summer of 2003, was 7.2 ppb h−1, associated with average NOx concentration of 10.8 ppb and NMHCs concentrations of 32.7 ppbC. The F(Ox) values are smaller than the 6-h median value of F(Ox) in summertime Tokyo (14.3 ppbv h−1).
 In rural Pennsylvania during May and June, P(O3) values (largely equivalent to F-D(Ox) in our notation) calculated on the basis of observed HO2 increased rapidly in the morning and reached a maximum at about 1000 eastern daylight time (EDT) [Ren et al., 2005]. The values remained at the average maximum level of ∼5 ppbv h−1 for the next 6 h before decreasing in the late afternoon. The daily O3 production was about 48 ± 47 ppbv d−1, of which O3 production from the HO2 + NO reaction accounted for about 75% [Ren et al., 2005]. These values are lower than those observed in Tokyo during summer, but are in part comparable to the values determined in Tokyo during winter.
 The BERLIOZ campaign, performed at a site located ∼50 km from Berlin during August, found midday maximum P(O3) values (largely compatible to F(Ox) in our notation) of 2–3 ppb h−1 on the basis of measured HO2 + RO2 concentrations derived from two instruments based on matrix isolation/electron spin resonance and chemical amplification; the two methods were in good agreement [Volz-Thomas et al., 2003; Mihelcic et al., 2003]. This production rate is lower than that in Tokyo during summer (12.8 ppbv h−1) and winter (8.7 ppbv h−1). The low rate is likely explained by reduced NOx concentrations under rural conditions.
Kleinman et al.  summarized the P(O3) (F(Ox) values in the present study) values in five metropolitan areas in the United States based on calculated radical concentrations that were in turn based on measurements of precursors in the boundary layer obtained using an aircraft. The median values were 6.2 ppb h−1 for Nashville (June and July 1995), 4.3 ppb h−1 for New York City (July 1996), 3.5 ppb h−1 for Phoenix (May and June 1998), 11.3 ppb h−1 for Philadelphia (July and August 1999), and 11.3 ppb h−1 for Houston (August and September 2000). The median F(Ox) for 6-h midday periods in Tokyo during summer was 14.3 ppb h−1 using modeled HO2 and 12.8 ppb h−1 using observed HO2 levels, higher than the above-ground values measured over the five cities listed above.
 The comparisons based on observed radical concentrations can be affected by systematic biases in the instruments used to measuring the radicals, if present. Standardization of the instruments, either by intercomparisons or use of a common standard calibrator, is required to enable for precise comparisons.
 Net production rates of Ox (F-D(Ox), [Ox] = [O3] + [NO2]) in Tokyo during the winter and summer of 2004 were determined using observed and calculated HO2 concentrations. The two values roughly agree with each other for most of the analysis period except for an underestimation in the morning during winter and an overestimation at midday in summer when the rates were estimated using modeled HO2 levels. In summer, the F-D(Ox) values on smog days were higher in the morning than those on the other days, while the midday rates on smog days were even lower than those on smog-free days. The occurrence of smog events in central Tokyo is not explained by day-to-day variations in the strength of in situ photochemistry alone: the effects of meteorology (land-sea breeze) coupled with photochemistry are also important. The daytime maxima of F-D(Ox) and the rate of HO2 + NO reaction in summer were only 1.1–2.4 times higher than those in winter. The dependence of the observed daytime concentrations of OH and HO2, rate of HO2 + NO reaction, and F-D(Ox) on NOx and NMHCs concentrations was studied in detail and compared with the dependence of modeled values. An underestimation of HO2 levels at high NOx concentrations in winter resulted in an underestimation of the rate of the HO2 + NO reaction and thus F-D(Ox). In the observed NO and NMHCs concentration ranges, the model predicted that ozone production takes place in the VOC-limited regime in winter. In contrast, F-D(Ox) based on observed HO2 levels increased with increasing NO concentration, even when it exceeded 20 ppbv, indicating a NOx-limited condition in winter.
 This analysis strongly indicates that cases exist where the behavior of the radicals can depart from that expected from our currently used tropospheric chemistry mechanism. In these cases, regimes of ozone production may be incorrectly determined even when they are deduced from observation-based models (OBM) in which observed concentrations of NMHCs and NOx are used as constraints and thereby the radical concentrations are calculated with the tropospheric chemistry mechanism. Clearly, the behavior of the radicals needs to be studied further in additional urban-based studies. In summer, the behavior of F-D(Ox) was similar for both cases in which observed and modeled HO2 levels are used. The production of ozone occurred in the VOC-limited regime under smog-free conditions in summer, whereas it changed into a “ridge” or NOx-limited regime under smog conditions. It is assumed throughout this paper that the modeled RO2 values are legitimate. A further test is required to investigate if discrepancies between modeled and actual RO2 values exist and contribute to the discrepancies in F-D(Ox). The F-D(Ox) values for Tokyo in summer were lower than the ground-based values obtained for Nashville, New York City, and Mexico City, but were higher than the ground-based values obtained for rural Pennsylvania and rural sites located near Berlin and London.
 The authors are grateful to all of the participants in the IMPACT campaigns. This study was supported financially by the Ministry of Education, Science, Sports, and Culture of Japan (RR2002 of the Kyosei Project).