Analyses of the Ozone Weekend Effect in Tokyo, Japan: Regime of Oxidant (O3 + NO2) Production

The ozone weekend effect (OWE) in Tokyo, the capital of Japan, was analyzed from April 2005 to March 2008 with respect to the photochemical ozone production, focusing on the ozone production regime. For most periods, the OWE was confirmed in the wintertime as well as in the summertime. The comparison of Ox ([Ox] = [O3] + [NO2] – 0.1[NOx]) concentrations between weekdays and weekends suggests that there are some periods when the photochemical ozone production on the weekends could be larger than on the weekdays (defined as the POP group). We compared NOx (the sum of NO and NO2) and Ox concentrations on the weekends with those on the weekdays for several distinct ranges of the non-methane hydrocarbons (NMHCs) concentrations, and NOx and Ox concentrations on the weekends are significantly lower and higher than those on the weekdays in the POP group. Therefore, the ozone production regime would be NMHC-limited in the POP group. The ozone production regime in the summertime in Tokyo could be in the boundary area between NMHCand NOx-limited in terms of the NMHCs/NOx ratios. The NMHCs/NOx ratios in the NMHC-limited regime are smaller than those in the NOx-limited regime in general. However, the NMHCs/NOx ratios in the POP group were not smaller than those in the non-POP group.


INTRODUCTION
Tropospheric ozone, a main component of photochemical smog, is a crucial air pollutant in the atmosphere.Ozone causes health problems such as respiratory diseases (Ho et al., 2007;Karakatsani et al., 2010;Neidell and Kinney, 2010).In addition, ozone has harmful effects on vegetation (Heath et al., 2009;Watanabe et al., 2010).Ozone in the troposphere is generated by the reactions involving nitrogen oxides (NO x = NO + NO 2 ) and volatile organic compounds (VOCs) as represented by non-methane hydrocarbons (NMHCs) in the presence of solar ultraviolet.In particular, O 3 is formed via the photolysis of NO 2 : NO 2 + hν (λ < 420 nm) → NO + O( 3 P), (1) where M represents a third-body molecule such as N 2 and O 2 in the troposphere.The inverse of these two reactions regenerates NO 2 and O 2 , as described by the following reaction: NO + O 3 → NO 2 + O 2 . (3) Reactions ( 1)-(3) form the photostationary state of NO x in the daytime (Hauglustaine et al., 1996;Carpenter et al., 1998;Matsumoto et al., 2006).The rate of the photochemical ozone production is determined substantially by the reaction of NO with peroxy radicals (RO 2 and HO 2 ): where R, R' and R'' indicate organic groups.Peroxy radicals are generated by the reaction of OH with NMHCs and CO: OH + CO → H + CO 2 , ( H + O 2 + M → HO 2 + M. ( 10) Reactions ( 4)-( 10) form a chain reaction centered on RO x (= OH + HO 2 + RO 2 ) radicals to generate tropospheric ozone.The main primary source of RO x radicals is the reaction of water vapor with O( 1 D) atoms which are produced by photodissociation of O 3 .Two OH radicals per an ozone molecule are generated via the following reactions: In summary, NO x and NMHCs are precursors of O 3 and the appropriate reductions of both are important to control ozone concentrations in the troposphere.However, ozone concentrations have been recently increasing, despite a decrease in the concentrations of NO x and NMHCs in some regions (e.g.Chou et al., 2006).In Japan, for example, the recent ozone increase is confirmed by continuous measurements at the numerous air monitoring stations in Tokyo (Tokyo Metropolitan Government, 2005).
The ozone "weekend effect" (e.g.Bronnimann and Neu, 1997;Atkinson-Palombo et al., 2006;Sadanaga et al., 2008;Stephens et al., 2008;Tonse et al., 2008) appears to be analogous to the recent increase in ozone concentrations.Ozone concentrations during the weekends are higher than those on the weekdays despite the presence of lower concentrations of NO x and NMHCs.CARB (California Air Resource Board, 2001) reports several hypotheses for the cause of the ozone weekend effect (OWE).Two of these hypotheses, "ozone quenching hypothesis" and "NO x reduction hypothesis", appear to be important in verifying the relationship between ozone and its precursors (NMHCs and NO x ) for the OWE.The "ozone quenching hypothesis" is based on the titration of ozone by NO (reaction (3)), which is emitted more abundantly in the urban area on weekdays than on weekends.The "NO x reduction hypothesis" is related to the chain reaction that produces ozone as described by reactions (4)-( 10).NMHCs play a role of the propagator in the chain reaction.The role of NO x in the ozone production is more complex.NO x is the direct ozone precursor as described in reactions (1) and (2).On the other hand, NO x is also an important terminator of the ozone-producing chain reaction because of the following reaction: In the case of low NO x mixing ratios, ozone production rates rise when NO x concentrations increase, that is socalled "NO x -limited" regime.In the NO x -limited regime, reactions (4) and ( 6) are important, and an increase in NO increases the ozone production rate.In the case of high NO x mixing ratios, ozone production rates decrease when NO x concentrations increase, that is so-called "NMHClimited" regime.Reaction (13) becomes significant in the OH radical reactions due to competition between reactions (7) and ( 13) in the NMHC-limited regime.Therefore, the increase in NO x concentrations decreases the chain length of the ozone-producing chain reaction.An important factor in determining the ozone production regime (i.e., NO xlimited or NMHC-limited) is the ratio of NMHCs to NO x .This ratio indicates that the ozone production rate is determined with respect to the balance between reactions (7) and ( 13).In the case of the low NMHCs/NO x , reaction (13) is favorable and an increase of NO x suppresses the ozone production rate, that is, the ozone production regime is NMHC-limited. CARB (2001) reports that the boundary area between NMHC-and NO x -limited regimes is the NMHCs/NO x ratio of 8-10 ppbvC/ppbv, where "ppbv" is parts per billion by volume and "ppbvC" is parts per billion by volume on carbon basis.
In Japan, the OWE analyses were conducted in Tokyo and Osaka, the largest cities in Eastern and Western Japan, respectively (Sadanaga et al., 2008).Sadanaga et al. (2008) confirmed the existence of the OWE in both Tokyo and Osaka.The OWE in Osaka was found to be due to the titration of O 3 by NO according to the quenching hypothesis.On the other hand, increased photochemical production of ozone due to the NO x reduction hypothesis could not be neglected as a cause of the OWE in Tokyo.This paper reports analyses of the OWE in Tokyo in terms of photochemical ozone production, focusing on the ozone production regime.In addition, the OWE in wintertime is also analyzed in this article although the report by Sadanaga et al. (2008) stated the OWE in only summertime.

SITE DESCRIPTION AND DATA ANALYSES
Fig. 1 shows the location of selected air quality stations in the Tokyo Metropolitan area.Tokyo is the capital of Japan, and is the largest metropolitan area in Japan, with a population in 2010 of approximately 13,000,000.Central Tokyo, the "special ward", is the most crowded area in Japan (the east area in Tokyo, see Fig. 1).As of 2010, central Tokyo's population was about 8,800,000, with an average population density of about 14,100 people/km 2 .
The OWE was analyzed using data measured in the monitoring stations for air pollution administrated by the government.We selected 23 air monitoring stations in Tokyo (Fig. 1).The stations in the "special ward" in Tokyo were numbered 1-13 (see Fig. 1).We classified stations 1-13 (in special ward) into "urban" and the remaining stations into "suburbs".We used hourly ozone, NO, NO 2 , NO x and NMHCs data from April 2005 to March 2008.VOC concentrations would be more appropriate to analyze the OWE, however, only the NMHCs data were available.Measurement principles of O 3 , NO x , and NMHCs are based on UV absorption, NO-O 3 chemiluminescence, and flame ionization detector with GC-backflush system to separate CH 4 off, respectively.NMHCs in this article mean the total NMHCs.The data were divided into "weekday" and "weekend," where "weekend" was defined as Saturday, Sunday and national holidays.Averaged diurnal patterns, as depicted in Fig. 2, were calculated for periods A (from January to March), B (from April to June), C (from July to September) and D (from October to December).For example, in this article "2006B", is defined as the period from April

RESULTS AND DISCUSSION
Existence or Nonexistence of the OWE Fig. 2 shows an example of the diurnal patterns of ozone, NO x and NMHCs on weekdays and weekends.For ozone precursors (i.e., NO x and NMHCs), their concentrations at the time of the peak value of NO were used, which is the primary product of vehicle exhaust and usually peaks between 06:00 and 09:00 in the morning.On the other hand, the ozone concentrations have a peak value between 13:00 and 16:00.These are typical diurnal variations in the urban atmosphere in Japan.
Table 1 shows weekday-weekend differences of ozone, NO x and NMHCs in summertime (see the supplementary data for each site).In this paper, "summertime" means "from April to September".The statistical significance was tested using the non-parametric Wilcoxon-Mann-Whitney test.In this article, "significantly" means "with statistical significance".NO x concentrations on the weekends were significantly lower than those on the weekdays for all periods.On the other hand, there are some periods when NMHCs concentrations on the weekends were significantly higher than those on the weekdays.Concentrations of O 3 on the weekends were found to be significantly larger than those on weekdays, except in the total and suburbs in 2006B, in the suburbs in 2007C.This result clearly shows that the OWE was present in most cases.
Table 1 also lists weekday-weekend differences of the "oxidant" (O x ), which indicates the sum of O 3 and NO 2 concentrations (e.g., White, 1997;Itano et al., 2007;Sadanaga et al., 2008).It should be noted that NO 2 in O x should represent the concentration of NO 2 generated secondarily by the oxidation of NO in the atmosphere.NO 2 emitted primarily from combustion sources such as vehicle exhaust should therefore be excluded in order to evaluate O x correctly.We defined 10% of NO x as the primarily emitted NO 2 , which is based on our previous report (Sadanaga et al., 2008).In summary, we defined O x by the following equation: As described in our previous report (Sadanaga et al., 2008), O x is an indicator to evaluate the "ozone quenching hypothesis".The O x concentration is conserved in the titration of ozone by NO (reaction (3)).Therefore, O x on weekend is  4.12 e a WD, WE and Delta indicate weekday, weekend and weekend minus weekday, respectively.b Urban, suburbs and total mean the averaged values of the sites 1-13, 14-23 and 1-23, respectively.B and C mean the periods from April to June and from July to September, respectively.c Statistically significant with P < 0.1.d Statistically significant with P < 0.05.e Statistically significant with P < 0.01.not necessarily higher than that on weekday if the "ozone quenching hypothesis" is a main cause of the OWE.
As listed in Table 1, O x concentrations on the weekends were found to be significantly greater than those on weekdays in 2005B, 2005C, 2006C and 2007B. In this article, the periods 2005B, 2005C, 2006C and 2007B are brought together into a group defined as the "photochemical ozone production" group (POP group).In the daytime, the O x concentration increases via the reaction of NO with peroxy radicals: Ozone is produced via the photolysis of NO 2 .In the daytime, reaction (4) is the rate-determining step for ozone production, so that this reaction governs the rate of photochemical ozone production (Sadanaga et al., 2005;Kanaya et al., 2008).
The decrease in O x concentration is mainly due to the reaction of OH with NO 2 : This reaction functions as an inhibitor against the photochemical ozone production.Hence, O x increase and decrease indicate promotion and inhibition of photochemical ozone production, respectively.In other words, O x concentrations are an indicator of photochemical ozone production.In the POP group, O x concentrations on weekend are greater than that on weekday, so that the photochemical ozone production on weekend would be larger than that on weekday.
Table 2 represents weekday-weekend differences of ozone, NO x , NMHCs and O x in wintertime (see the supplementary data for each site).In this paper, "wintertime" means "from October to March".Both NO x and NMHCs concentrations on the weekends were lower than those on the weekdays for all periods.On the other hand, ozone concentrations on the weekends were greater than those on the weekdays for all periods.It is noteworthy that the OWE was confirmed in the wintertime as well as in the summertime.O x concentrations on the weekends were not significantly higher than those on weekdays.Therefore, OWE in wintertime can be explained by the "ozone quenching hypothesis".

Diagnosis of Photochemical Ozone Production: Relationship between O x and its Precursors
The relationship between photochemical ozone production and the concentrations of ozone precursors (i.e., NO x and NMHCs) is non-linear and complex.As described in Introduction, in the NO x -limited regime, the production rate of ozone rises when NO x concentrations increase.Conversely, the production rate of ozone decreases when NO x concentrations increase, in the NMHC-limited regime.Taking this point into account, we discuss the ozone production regimes as follows.We compared NO x and O x concentrations on the weekend with those on weekdays for several distinct ranges of the NMHCs concentrations.If NO x concentrations are significantly lower and O x concentrations are significantly higher on weekends compared to weekdays, then the ozone production regime could be NMHC-limited.It should be noted that O x instead of O 3 was used in this analysis because O x concentrations are more appropriate indicator of photochemical ozone production as described above.weekday, weekend and weekend minus weekday, respectively.b Urban, suburbs and total mean the averaged values of the sites 1-13, 14-23 and 1-23, respectively.A and D mean the periods from January to March and from October to December, respectively.c Statistically significant with P < 0.1.d Statistically significant with P < 0.05.e Statistically significant with P < 0.01.
Table 3 shows the weekday-weekend differences for NO x and O x for several distinct ranges of NMHCs concentrations in summertime.NO x and O x concentrations on the weekends are significantly lower and higher than those on the weekdays, respectively, for the periods of 2005B, 2005C, 2006C and 2007B, which are exactly the POP group.This result would be valid considering the following three points.First, higher O x concentrations on the weekend than those on the weekdays (i.e. the POP group) mean that the OWE cannot be explained only by the titration of O 3 by NO.Second, the O x increase is due to reactions (4) and ( 6) and reflects the photochemical ozone production.Third, NO x reduction causes the rise of the photochemical ozone production in the NMHC-limited regime.In summary, the POP group must be in NMHCslimited regime, so that the "NO x reduction hypothesis" is one of the factors of the OWE for the POP group.The non-POP group would be in NO x -limited because the both NO x and O x concentrations on the weekends are significantly lower than those on the weekdays for several distinct ranges of NMHCs concentrations.It should be noted that the regime for O 3 might be different from those inferred for O x here.For example for the non-POP group, O x concentrations are smaller on weekend than those on weekday but that is not the case for O 3 concentrations.Nonetheless, we focus on the regime for O x here that should describe the photochemical tendency more correctly.CARB (2001) reports that the boundary area between NMHC-and NO x -limited regimes is the NMHCs/NO x ratio of 8-10 ppbvC/ppbv, which is roughly in agreement with that suggested by Milford et al. (1994) (10-20 ppbvC/ppbv).Hence such ozone production regimes are discussed in terms of NMHCs/NO x ratios.Table 3 also lists the weekday-weekend differences of NMHCs/NO x ratios for several distinct ranges of the NMHCs concentrations.The NMHCs/NO x ratios on the weekend were higher than those on the weekdays in all the cases aside from statistical significance.The NMHCs/NO x ratios in summertime ranged from 5.10 to 18.81 ppbvC/ppbv, which are roughly in agreement with the boundary area between NMHC-and NO x -limited regimes reported by Milford et al. (1994) andCARB (2001).Therefore, the ozone production regime in the summertime could be in the boundary area between NMHC-and NO x -limited with respect to the NMHCs/NO x ratios.In general, the NMHCs/NO x ratios in the NMHClimited regime are smaller than those in the NO x -limited regime.In the summertime, the NMHCs/NO x ratios ranged from 5.10 to 11.22 ppbvC/ppbv in the POP group.On the other hand, the NMHCs/NO x ratios in the non-POP group ranged from 6.12 to 18.81 ppbvC/ppbv.The NMHCs/NO x ratios in the POP group were not significantly smaller than those in the other periods.In order to discuss ozone production rates more specifically, the use of the VOC reactivity, that is, the reactivity of VOCs with OH radicals, is more appropriate than the VOC concentration (Sillman, 1999).In addition, the critical NMHCs/NO x ratio separating the two regimes can vary with the primary RO x production rate (P ROx ) (Thornton et al., 2002).The variation of the boundary area with P ROx , and the composition of NMHCs might cause the present result in summertime.
Table 4 represents the weekday-weekend differences for NO x and O x for several distinct ranges of NMHCs concentrations in wintertime.Both NO x and O x concentrations on the weekends are lower than those on the weekdays, that is, the ozone production regime in wintertime is in NO x -limited for most of the periods.Table 4 also shows the weekday-weekend differences of NMHCs/NO x ratios for several distinct ranges of the NMHCs concentrations.The NMHCs/NO x ratios on the weekend were higher than those on the weekdays in all the cases aside from statistical significance.In wintertime, the NMHCs/NO x ratios ranged from 3.54 to 7.26 ppbvC/ppbv, and the ozone production regime is categorized into NMHC-limited if only the NMHCs/NO x ratio is considered.In addition, it should be noted that the NMHCs/NO x ratio in the boundary area between NMHC-and NO x -limited regimes in wintertime becomes higher than that in summertime.Kanaya et al. (2008) predicts the boundary associated with the NMHCs/ NO x of ~20 and ~50 ppbvC/ppbv in summertime and wintertime, respectively.However, the ozone production regimes in wintertime were not NMHC-limited in many cases.Kanaya et al. (2008) reported that net O x production rates in winter in Tokyo based on observations showed NO xlimited-like behavior, whereas the box model calculation (Kanaya et al., 2008) suggested that the ozone production regime was in NMHC-limited.Our result is consistent with the result of Kanaya et al. (2008).

CONCLUSIONS
We analyzed the OWE in Tokyo, in terms of photochemical ozone production from April 2005 to March 2008.For most periods, the OWE was confirmed with statistical significance in the wintertime as well as in the summertime.In summertime, O x concentrations on the weekend were also significantly higher than those on the weekdays in periods 2005B, 2005C, 2006C and 2007B (defined as the POP group).This result suggests the photochemical ozone production on the weekend could be larger than on the weekdays in these periods.In order to diagnose the ozone production regime as a cause of the OWE, we compared NO x and O x concentrations on the weekend with those on the weekdays for several distinct ranges of the NMHCs concentrations and we concluded that the ozone production regime would be NMHC-limited and NO x -limited in the POP group and otherwise, respectively.We also compared the NMHCs/NO x ratios on the weekend with those on the weekdays for several distinct ranges of the NMHCs concentrations.The NMHCs/NO x ratios in summertime ranged from 5.10 to 18.81 ppbvC/ppbv, which are roughly in agreement with the boundary area between NMHC-and NO x -limited regimes reported by Milford et al. (1994) andCARB (2001).Therefore, the ozone production regime in the summertime could be in the boundary area between NMHC-and NO x -limited with respect to the NMHCs/NO x ratios.On the other hand, the NMHCs/NO x ratios in the POP group were not significantly smaller than those in the other periods although the NMHCs/NO x ratios in the NMHC-limited regime are smaller than those in the NO x -limited regime in general.This might be because the boundary area can vary with P ROx and the composition of NMHCs.In wintertime, O x concentrations on the weekend were not greater than those on the weekdays in most periods, so that the OWE in wintertime can be explained by the titration of ozone by NO.The ozone production regime during the wintertime could be NMHC-limited in terms of the NMHCs/NO x ratio but is actually NO x -limited in many periods.It might be insufficient that the ozone production regime in the wintertime is discussed using only the NMHCs/NO x ratio.

Fig. 1 .
Fig. 1.Locations of selected air monitoring stations in Tokyo.Dashed lines indicate the boundary lines of municipalities.Numbers represent the station code numbers used in this paper.The "special ward" in Tokyo is the east of the bold line.

Table 3 .
Weekday-weekend differences of NO x , O x and ratios of NMHCs to NO x for the same intervals of NMHCs concentrations in summertime.WD, WE and Delta are weekday, weekend and weekend minus weekday, respectively.b A and D mean the periods from January to March and from October to December, respectively.c Statistically significant with P < 0.1.d Statistically significant with P < 0.05.e Statistically significant with P < 0.01.f Number of samples.g The "POP group" periods are underlined in this column.

Table 4 .
Weekday-weekend differences of NO x , O x and ratios of NMHCs to NO x for the same intervals of NMHCs concentrations in wintertime.WD, WE and Delta are weekday, weekend and weekend minus weekday, respectively.b A and D mean the periods from January to March and from October to December, respectively.c Statistically significant with P < 0.1.d Statistically significant with P < 0.05.e Statistically significant with P < 0.01.f Number of samples.