Health Risk of Increased O 3 Concentration Based on Regional Emission Characteristics under the Unusual State of the COVID-19 Pandemic

: Photochemical oxidant concentration increases with the decrease in nitrogen oxide (NO x ) concentration in volatile organic compound (VOC)-sensitive areas with several automobiles and factories. We aimed to quantify the changes in health risks from ozone (O 3 ) and nitrogen dioxide (NO 2 ) using disability-adjusted life years (DALY) in Osaka City, which is one of the major cities in Japan. ADMER-PRO version 1.0, an atmospheric model for secondary products, was used to estimate the concentration distribution of NO 2 , VOC, and O 3 using the year-on-year change of trafﬁc during the declaration of the state of emergency in response to the coronavirus disease 2019 (7 April to 21 May 2020). NO 2 concentration decreased by an average of 0.962 ppb in 88.9% of the grids in Osaka City, whereas O 3 concentration increased by an average of 1.00 ppb in all the grids with a 26–28% reduction of trafﬁc volume due to the pandemic. We also found three intensities for the VOC-sensitive condition depending on the different regional emission characteristics, with the DALYs of health risks from the decrease in NO 2 exceeding those from the increase in O 3 , reaching 811.4 and 55.90 total DALYs in the city, respectively.


Introduction
The coronavirus disease 2019 (COVID-19) pandemic [1] has caused a significant reduction in traffic due to the lockdown to prevent its spread in many countries [2]. Particularly in Japan, the traffic volume decreased by up to 80% compared with the previous year during the period of declaration of the state of emergency [3]. As a result of the reduction in energy consumption and oil demand worldwide, the air quality showed improvement, with a reduction in NO 2 concentration of about 20% [4]. On the other hand, the lockdown significantly reduced the concentrations of carbon monoxide (CO) and NO 2 , but increased ozone (O 3 ) concentration in several areas, including Rio de Janeiro and Baghdad [5][6][7]. The increased O 3 concentration triggered by the COVID-19 lockdown is a major problem; Bell et al. [8] reported an increase in the daily mortality due to continuous O 3 exposure.
The mechanism by which the decrease in NO 2 concentration often leads to an increase in O 3 concentration, mainly in urban areas [5], is described below.
The nitric oxide (NO) emitted from automobiles is oxidized in the atmosphere to NO 2 , which is photodegraded into NO and mono-oxygen (O) by ultraviolet energy, whereas O 3 is formed by the combination of O and dioxygen (O 2 ): where M is the third body, stabilizing the produced molecules by sapping the chemical energy generated by the reaction. N 2 and O 2 play this role in the atmosphere. The formed O 3 reacts with NO to produce NO 2 again. Thus, if only nitrogen oxides (NO x ) are present in the environment, the O 3 concentration will be a constant value due to the balance of reactions that produce and consume O 3 : However, when volatile organic compounds (VOC) are present in the atmosphere, they react with hydroxyl radical (OH) to produce alkylperoxy radical (RO 2 ). In addition, the alkoxy radical (RO), which is produced by the reaction of RO 2 with NO, reacts with O 2 to produce a hydroperoxy radical (HO 2 ). The reaction of these peroxy radicals with NO in (R1) produces NO 2 without destroying O 3 : The NO 2 produced in (R4,5) undergoes the reactions in (R1,2) again. The chain reaction cycle of O 3 production is triggered, and the O 3 concentration increases.
When the NO x concentration is lower than the VOC concentration, the peroxy radicals do not react with NO, but with themselves, producing peroxides and other oxygenated compounds. Thus, NO 2 is not produced in (R4, 5), and the reaction of (R1,2) is delayed. Therefore, despite the further reduction in the VOC emissions, the O 3 concentration will remain unchanged given that the NO 2 supplied to (R1) will not change. Meanwhile, the further reduction of NO x emissions delays the reactions of (R1,2), resulting in a decreased O 3 concentration. Thus, the generation rate of O 3 is sensitive to the supply of NO x and is referred to as being "NO x -sensitive" [9]: P(O 3 ) = 2k 1 P(HO x ) 2k 2 1 2 [NO], (1) where P(O 3 ) and P(HO x ) are the generation rates of O 3 and hydrogen oxide radicals (HO x ), respectively, k 1 and k 2 are the rate constants, and [NO] is the concentration of NO. In general, the NO x -sensitive intensity is high in suburban areas due to the large amount of VOC emissions from plants [10]. On the other hand, when NO x concentration is higher than the VOC concentration, the reaction of NO 2 with OH to produce nitric acid (HNO 3 ) becomes dominant, and this condition interferes with the chain reaction cycle from (R4,5) to (R1), (R2), which starts from the reaction of OH with VOC, resulting in a decreased O 3 production efficiency: When NO x emissions are reduced, the reaction of (R6) is stagnated, and the chain reaction cycle starting from the reaction of OH and VOC is accelerated. If the decrease in NO 2 supplied to (R1) due to the reduction of NO x is balanced with the increase in NO 2 produced in (R4,5) due to the acceleration of the chain reaction cycle, the O 3 concentration will not change. However, if the latter amount exceeds the former, the O 3 concentration will increase. Meanwhile, the reduction in VOC emissions further reduces the amount of peroxy radicals in the atmosphere, resulting in a low production of O 3 , which requires the peroxy radical in the chain reaction cycle from (R4,5) to (R1), (R2). Thus, the generation of O 3 is sensitive to the supply of VOC and is referred to as being "VOC-sensitive" [9]:  2 , and M, respectively. In general, the intensity of VOC-sensitivity is high in urban areas due to the large amount of NO x emissions from automobiles and factories [10]. As described above, the increase in O 3 concentration with the decrease in NO 2 concentration depends on sensitive conditions in the area. However, the changes in O 3 concentration due to the decrease in traffic in COVID-19 and health risk as a result have not been quantitatively clarified for Japanese cities, which are considered to be in VOC-sensitive condition. Therefore, in this study, we first measured the sensitive conditions in a major city in Japan with a largely exposed population and then calculated the changes in O 3 and NO 2 concentrations caused by the reduction in traffic volume as a countermeasure against COVID-19. Finally, we quantitatively clarified the effects of these different types of pollutants on human health risks using the same criteria.

Selection of the Case Study Area
In this study, given that NO 2 and VOC cause the formation of O 3 , Osaka City ( Figure 1) was selected as the case study area, where many emission sources of such precursors exist [11][12][13][14][15]. The Hanshin Industrial Zone, one of the three major industrial zones (Keihin, Chukyo, and Hanshin Industrial Zone), extends along the coast of Osaka Prefecture. The Hanshin Industrial Zone is the second largest in terms of the shipment value of manufactured goods, after the Chukyo Industrial Zone, which is centered on the automobile industry [11]. The NO x emissions from smoke in Hyogo Prefecture, which is included in the Hanshin Industrial Zone, was the third largest after Yamaguchi and Hokkaido Prefectures in fiscal year (FY) 2018 [12], which indicates the presence of large stationary sources. The industrial zone is also a source of VOC, and VOC emissions in Osaka Prefecture in FY2018 was 31,789 t/year, the largest in western Japan [13]. In addition, Osaka Prefecture had the greatest transport ton-kilometers (12,705,503 thousand ton-kilometers) in FY2019 among the six major prefectures (Tokyo, Kanagawa, Aichi, Osaka, Hyogo, and Fukuoka Prefectures) [14] and the second greatest transport person-kilometers (5,160,352 thousand person-kilometers) in FY2019, following Tokyo Prefecture, the capital of Japan [15]; thus, mobile sources are also sufficient in the case study area. Thus, although the existence of NO x emissions is sufficient for there to be a VOCsensitive condition, the results of sensitivity regime diagnostics with satellite data show that the area around Osaka City has a mixed sensitivity, which is intermediate between NO xand VOC-sensitive conditions [16], or there are two sensitive conditions (NO x -and VOCsensitive ones) [17]. However, it has not been revealed which state is currently predominant in the city. Therefore, this study selected Osaka City, which has a large, exposed population and where the risk of increased O 3 concentration caused by decreased NO 2 concentration is unknown.

Estimation of Atmospheric Chemical Concentration and Measurement of Sensitive Conditions
In this study, ADMER PRO Version 1.0 (ADMER-PRO) [18], which was developed by National Institute of Advanced Industrial Science and Technology (AIST) in Tsukuba, Japan, was used in estimating the concentrations of VOC, O 3 , and NO x , including NO and NO 2 . This is an atmospheric model for evaluating regional concentration distributions and can estimate the atmospheric concentration distributions of chemical substances, including secondary products, by using the chemical transport model [18]. ADMER-PRO includes emissions and meteorological data in 2005 ( Figure 2). The meteorological data were processed from observed data by the National Centers for Environmental Prediction [19]. For stationary VOC sources, stationary NO x sources, and mobile NO x sources, we can simulate recent situations by comparing the actual emission data in 2005, which were included in the model, with the latest available emission data and by inputting the rate of change by prefecture. Given that NO x emissions from automobiles, which are mobile NO x sources, are not reported, the rate of the 24 h average traffic volume was substituted as the rate of NO x emissions. However, biogenic VOC (BVOC) emissions under standard meteorological conditions [20,21] were used and substantially revised significantly upward according to the results of field surveys and numerical experiments conducted by the AIST [22]. For stationary VOC sources, the VOC emission inventory data [13] were used in calculating the rate of change in VOC emissions from FY2005 to FY2018 by prefecture and industry ( Figure 2). These data are published two years after Ministry of the Environment (MOE) estimates VOC emissions by industry for each fiscal year for the purpose of monitoring the progress of VOC emission control [13]. Here, we used the latest published VOC emissions in FY2018 [13]. For stationary NO x sources, a comprehensive survey on air pollutant emissions [12,23] was used in calculating the rate of change in NO x emissions from factories and power plants from FY2005 to FY2017 by prefecture ( Figure 2). The survey is a statistical survey conducted every three years by the MOE for the monitoring of the emission status of air pollutants emitted from apparatus exhausting flue gas [12,23]. Here, we used the latest published NO 2 emission data in FY2017 [12]. For mobile NO x sources, the Road Traffic Census [24,25] was used in calculating the rate of change in 24 h average traffic volume on weekdays from FY2005 to FY2015 by prefecture ( Figure 2). This survey is a statistical survey conducted by Ministry of Land, Infrastructure, Transport and Tourism every five years [24,25]. Here, we used the latest published 24 h average traffic volume in FY2015 [25]. However, given that the traffic volume decreased by only 1.7% from FY2010 to FY2015 [26], we supposed that the data in FY2015 were sufficient to reproduce more recent situations. Using the emission data corrected by the above calculations, the average values from April to September (see Section 3.2 for the explanation on the selection of this period) of the concentrations of VOC and NO x and the sum of NO and NO 2 were outputted using CB-IV 99, which is a built-in model that aggregates the formation of secondary pollutants, such as O 3 and aldehydes, by grouping VOC into 37 chemical species and 93 chemical formulas based on their carbon bonding states [20]. In particular, given that the O 3 concentration varies greatly with time, the daily maximum hourly value, which is related to health risk, was outputted using CB-IV 99. [5,27,28] was used as an indicator to measure the sensitive conditions. When the value of this indicator is relatively small, the intensity of VOC-sensitivity is high, whereas when the value of this indicator is relatively large, the NO x -sensitive intensity is high [27]. Kannari [28] estimated that in the Kansai region, including Osaka Prefecture, the boundary value for the sensitive conditions is 6 to 12 for non-methane hydrocarbons (NMHC)/NO x concentration ratio. Since methane is not included in the VOC output by ADMER-PRO, we refer to its value as a boundary in this study.

Setting of Meteorological Conditions and Grids
The photochemical oxidant (OX) concentration increases during warmer seasons [29]. The total time when the hourly OX concentration in Osaka City exceeded the environmental quality standard of 60 ppb was highest from April to September [30]. Therefore, we set these months as the calculation period in this study. OX is a general term for oxidizing substances, excluding NO 2 , that can be measured using the neutral buffered potassium iodide method and include not only O 3 but also peroxyacetylnitrate. In this study, OX and O 3 concentrations were assumed to be equal, because O 3 accounts for most of OX. ADMER-PRO is equipped with a "long-term analysis mode" that uses the classification of weather patterns to estimate the concentration over the target period in a relatively short time [22]. In the long-term analysis mode, the weather patterns of the target areas, which are grouped by weather pattern, are characterized on each day, and the appearance frequency of each weather pattern is calculated until the appearance frequency reaches a specified value [31]. In this study, the top five weather patterns with the highest appearance frequencies from April to September were used to shorten the calculation time, which includes three days of weather by each weather pattern. The AIST indicates that the coefficient of determination is 0.77 when the top five weather patterns are considered [22].
In addition, two grids were set up: one for the area to be analyzed and the other for the larger area covered with a rough resolution. For Osaka City, the target area, a 1 km × 1 km grid was set. For the larger area, a 5 km × 5 km grid was set for Kinki region (Mie, Shiga, Kyoto, Osaka, Hyogo, Nara, and Wakayama Prefectures), assuming that O 3 and its precursors are transported from outside the target area [32].

Reduction of Traffic Volume during the COVID-19 Pandemic
In this study, we compared and analyzed two scenarios: a scenario reflecting the normal situation in recent years (usual scenario) and a scenario for reduced traffic due to the declaration of the state of emergency for COVID-19 (COVID-19 scenario), as shown in Figure 2 with the dataset and analysis procedure. The NO x emission sources include power plants and refuse incineration plants, which support the lives of citizens [33,34]; they were assumed to have been operating normally during the declaration of the state of emergency. To simulate the conditions based on the reduction of traffic volume due to COVID-19, at first, we extracted the rate of year-on-year change in traffic volume of major sections of expressways (West Nippon Expressway, Honshu-Shikoku Bridge Expressway, and Hanshin Expressway) in the Hanshin area only on weekdays and averaged over the period of declaration of the state of emergency [3]. Next, the rate of year-on-year change in traffic volume of general roads was calculated by extracting the rate of year-on-year change in traffic volume of National Route 1 on weekdays and averaged over the period of the declaration of the state of emergency [3]. The rate of year-on-year change of expressways and general roads was set uniformly in the Kinki region, including Osaka City. Finally, we multiplied the 24 h average traffic volume on expressways and general roads in FY2015 by the rate of year-on-year change in traffic volume for each road type and calculated the rate of change in traffic volume during the period of the declaration of the state of emergency, with weighting of the traffic volume for each road type by prefecture: where R i is the rate of change of traffic volume in the COVID-19 scenario, and a and b are the rates of year-on-year change in traffic volume for expressways and general roads, respectively. TE i and TG i are the usual traffic volumes for expressways and general roads, respectively. The rate of year-on-year change in traffic volume in the Kinki region was 72.5% for expressways and 93.7% for general roads during the period of the declaration of the state of emergency [3]. Based on these values, we set the rate of change in traffic volume during the period of the declaration of the state of emergency as a = 0.7 and b = 0.9 and estimated that the rates were 0.72 in Nara Prefecture, 0.73 in Mie and Shiga Prefectures, and 0.74 in other prefectures (Kyoto, Osaka, Hyogo, and Wakayama Prefectures). These values were multiplied by the rate of change in mobile NO x sources by prefecture in the usual scenario and inputted into ADMER-PRO.
In this study, the annual change in health risk was estimated with the assumption that the reduction in traffic volume during the period of the declaration of state of emergency (7 April to 21 May 2020) would continue for one year. We used the reduction in traffic volume during the period of the declaration of the state of emergency to estimate the annual change in health risk because the impact of COVID-19 is still continuing, and the continuous reduction in traffic volume should be considered. In addition, the health risk is comparable with those of other diseases, which are generally estimated using the annual loss of disability-adjusted life years (DALY) as an indicator.

Validation of the Model
In this study, the reproducibility of the model was evaluated using the normalized mean bias (NMB), normalized mean error (NME), and the probability of factor 2 (Pfac2), which are widely used in the evaluation of atmospheric model performances [35,36]. The smaller the absolute value of NMB (Equation (4)) and the smaller the value of NME (Equation (5)), the higher the reproducibility [35].
where Cal k is the estimated value, Obs k is the observed value, and N is the number of data. In this study, as substances for estimating health risks, the average NO 2 concentrations estimated from April to September and the average O 3 concentrations estimated at 14:00 of the specific weather patterns from April to September were compared with the average concentrations observed in FY2018 in the grid where each concentration was estimated. In Osaka City, NO 2 and O 3 concentrations were observed at 25 and 14 stations, respectively, in FY2018 [37].

Estimation of Health Risk Derived from O 3
In this study, a 0.32% increase in mortality with a 10 ppb increase in the 24 h value of O 3 concentration [8] was converted to a 0.17% increase in mortality, with the adoption of a 10 ppb increase in the daily maximum hourly value of O 3 [38]. These values were based on the results of Bell et al. [8], who observed that the percentage increase in all-cause mortality associated with a 10 ppb increase in O 3 concentration was 0.32% for a 1-day lag across the 98 U.S. communities. Nakanishi et al. [38] also adopted the value reported by Bell et al. [8], which was considered to have less uncertainty. On the other hand, a positive correlation exists between the daily maximum hourly value and the number of deaths due to respiratory disease [39]. Therefore, we calculated the average concentration at each hour from the observed values, excluding missing, unobserved, and error values, at 14 stations in Osaka City and used the estimated values for health risk assessment at 14:00 when the concentration was the highest [40] (Figure 3).  [40]. The concentration at 14:00 was the highest, at 44.9 ppb. DALY, as an index of health risk, refers to the years lost due to death or disability, and it can comprehensively represent the burden of disease [41]. By evaluating health risk in terms of DALY, the health risk can potentially be compared with other diseases. DALY is the sum of years of life lost (YLL) and years lost due to disability [41]; however, given that the endpoint of this study increased mortality, only YLL was considered. YLL is calculated as the product of the number of deaths and life expectancy per age group [41]. The vital statistics [42] were used for the number of deaths and the life tables [43] for the life expectancy; the YLLs were calculated for each 5-year age group and finally summed to calculate the change in health risk: where ∆DALY is the change in health risk, Case j is the number of deaths from causes other than injury disease at age j, RR is the relative risk of a 10 ppb increase in O 3 concentration, and L j is the life expectancy at age j. Based on the values of Bell et al. [8] and Nakanishi et al. [38], we calculated the health risk for usual and COVID-19 scenarios by using RR-1 of 0.0017. ∆C is the difference between the average O 3 concentration estimated at 14:00 of the specific weather patterns from April to September averaged over Osaka City and the background concentration. We adopted the minimum concentration estimated for the usual scenario by ADMER-PRO as the background concentration [38].

Estimation of Health Risk Derived from NO 2
Chen et al. [44] showed a 1.63% increase in daily mortality with a 10 µg/m 3 increase in the 24 h value of NO 2 concentration from the exposure-response relationship between NO 2 concentration and daily deaths in 17 Chinese cities. In this study, the mass concentration of NO 2 (a 1.63% increase with a 10 µg/m 3 increase in the 24 h value) was converted to a volume concentration (a 3.08% increase with a 10 ppb increase in the 24 h value) at 23.8 • C, the observed temperatures in Osaka City from April to September in FY2018 was averaged [40]. However, the 0-4-year-old group was excluded from this study [44]. Based on the values, the health risks for usual and COVID-19 scenarios were evaluated in terms of DALY in Equation (6) and the estimation of health risk due to O 3 . Meanwhile, RR-1 was set at 0.0308; ∆C is the difference between the average NO 2 concentrations estimated from April to September averaged over Osaka City and the minimum concentration estimated for usual scenarios by ADMER-PRO as the background concentration [38].

Validation of the Model
The average NO 2 concentrations estimated from April to September by ADMER-PRO were compared with the average concentrations observed for the same period in FY2018 in the grid where the concentrations were estimated. The results were 0.315 for NMB and 0.339 for NME. These findings can be attributed to the difference in the distribution of emission sources from FY2005. Thus, certain emission sources existed in FY2005 that were absent in FY2018 due to environmental pollution control or population decrease, and the estimated values were overestimated compared with the observed values. However, Pfac2 was 100%, and all estimated values were contained within two times the observed value.
Similarly, for O 3 concentrations, NMB and NME were both 0.301. The estimated values were higher than the observed ones at all stations, whereas Pfac2 was 100%. These differences between the estimated and measured values may have been caused by the difference in the distribution of emission sources and weather patterns of the observed values that were not included in the top five most frequently appearing weather patterns from April to September in this study. As a result, the mean of the observed values decreased. Based on these results, we considered this model to be sufficiently applicable for risk assessment, although it slightly overestimates the estimated values.  (Figure 4b). The low O 3 concentrations, especially in the coastal areas, is attributed to the reaction of NO 2 , which also showed high concentrations in these areas, with OH interfering with the chain reaction cycle from (R4,5) to (R1), (R2), which starts from the reaction of OH with VOC, resulting in a decrease in O 3 production efficiency. Table 1 lists the VOC/NO x concentration ratios (in ascending order) obtained from the average VOC and NO x concentrations estimated from April to September for each ward, and Figure 5 shows the VOC/NO x concentration ratio with the location of each ward. Since the VOC/NO x concentration ratios of all the wards were below 6 to 12, as indicated by Kannari [28], overall, Osaka City had a high intensity of VOC-sensitivity. Therefore, we divided the ward into three groups of eight wards, starting from the ward with the smallest concentration ratio and the highest intensity of VOC-sensitivity, and classified them based on high (Group High), middle (Group Middle), and low (Group Low) intensities of VOC-sensitivity ( Figure 5). The average VOC concentrations in each group were 71.4, 85.7, and 90.7 ppb, and the average NO x concentrations were 33.8, 26.8, and 23.7 ppb. Group High mainly included coastal areas, and Group Middle to Group Low comprised inland wards. The top ten wards with the highest VOC concentrations were distributed among the three groups, whereas seven of them with NO x concentrations were included in Group High. This finding indicates that NO x concentration contributed significantly to the group to which a ward belonged. Two types of NO x emission sources exist: stationary sources, such as factories and power plants, and mobile sources, such as automobiles [33,34]. With regard to stationary NO x emission sources, Figure 6 shows the breakdown of NO x emissions in Osaka City in FY2018 by industry [12]. The production, transmission, and distribution of electricity accounted for the largest share at 31%, followed by the manufacture of iron and steel at 23%, waste disposal businesses at 21%, building heating and other offices at 8%, and the manufacture of pulp, paper, and paper products at 4% [12]. Konohana Ward (No. 1 in Figure 5), which has the smallest VOC/NO x concentration ratio in Osaka City and was included in Group High, has a refuse incineration plant (Maishima Plant) and a sewage sludge treatment plant (Maishima Sludge Center). In addition, Suminoe Ward (No. 2 in Figure 5), which had the next smallest VOC/NO x concentration ratio, has the only natural gas-fired power plant (Nanko Power Plant) in Osaka City, from which 165 t of NO x , corresponding to 6% of annual NO x emissions in Osaka City, were emitted in FY2018 [45]. These large-scale emissions of NO x are thought to have resulted in high NO x concentrations in the wards with the smallest VOC/NO x concentration ratios. In addition, Taisho Ward (No. 4 in Figure 5), which was included in Group High, had the largest, at 190,610 million yen, value of manufactured goods shipments of iron and steel in Osaka City in 2019 [46]; the other wards in Group High are also engaged in the industry (Table 1). By contrast, most of the wards in Group Middle and Group Low have no engagement in the industry (Table 1).

Estimation of Atmospheric Chemical Concentration and Measurement of Sensitive Conditions
With regard to mobile NO x emission sources, Table 1 shows the total value of the product of section extension and 24 h traffic volume for each ward in FY2015 [25]. Six of the top ten wards were included in Group High. These results indicate that wards with high NO x concentrations had high emissions from stationary and mobile sources of NO x , which contributed to the VOC-sensitive condition.

Comparison of NO 2 and O 3 Concentrations Between Scenarios
In the COVID-19 scenario, NO 2 concentrations decreased by an average of 0.962 ppb in 88.9% of the grids in Osaka City compared with the usual scenario due to the reduction in traffic caused by COVID-19 (Figure 7a). Regarding the characteristics of each group, seven out of eight wards belonging to Group Low were among the top ten wards with the largest decrease in NO 2 concentration (Table 2). On the other hand, Group High did not include any of these wards (Table 2) given the absence of significant stationary emission sources in Group Low (Table 1). Thus, the proportion of emissions from mobile sources was large, and the decrease in traffic volume in the COVID-19 scenario may have caused a large decrease in NO 2 concentration in the wards belonging to Group Low. By contrast, Group High included significant stationary emissions, such as the power plant and refuse incineration plant. However, given that these operations continued in the COVID-19 scenario, the overall decrease in NO 2 concentration was considered to be relatively small, although only the emissions from mobile sources decreased with the decrease in traffic. The NO 2 concentration in Konohana Ward was 33.1 ppb in the usual scenario and 33.4 ppb in the COVID-19 scenario, resulting in an increase of 0.261 ppb. However, considering the possibility of overestimation using the NME in Section 4.1, the NO 2 concentration was at its lowest at 22.1 ppb in the COVID-19 scenario. With the increase in 0.261 ppb being within the margin of error, it is unclear as to whether only NO 2 the concentration in Konohana Ward followed an increasing trend.
On the other hand, O 3 concentrations in the COVID-19 scenario increased by an average of 1.00 ppb in all grids compared with the usual scenario (Figure 7b). The top 12 wards with the highest intensity of VOC-sensitivity included the top 10 wards with the largest increase in O 3 concentration. In particular, Nishi (No. 8 in Figure 5), Naniwa (No. 7 in Figure 5), and Minato Wards (No. 3 in Figure 5), which belonged to Group High, had large increases of 1.39, 1.35, and 1.28 ppb, respectively. By contrast, one ward in Group Low, with the top 10 wards, showed the largest increase in O 3 concentration. The wards in the former group were considered to be in phase A in Figure 8, where O 3 concentration increased in the VOC-sensitive condition. If the reduction of NO 2 emissions continues, the O 3 concentration is expected to remain almost constant as the amount of NO 2 supplied to (R1) decreases, while the amount of NO 2 produced in (R4) will increase due to the acceleration of the chain reaction cycle. However, given the continuity of VOC-sensitive conditions, the O 3 concentration may increase further if VOC emissions increase in this condition. By contrast, the wards in the latter group were considered to be in phase B in Figure 8, given the small increase in O 3 concentration. If the decrease in NO 2 emissions continues further, the condition will shift to the NO x -sensitive condition in phase C in     Table 3 shows the health risks from NO 2 and O 3 for the usual and COVID-19 scenarios, and the increase/decrease in health risk from the usual scenario to the COVID-19 scenario. For the usual scenario, the health risks of NO 2 and O 3 for the whole of Osaka City were 9454 and 1310 DALYs, respectively. In the COVID-19 scenario, the health risk from NO 2 was 811.4 DALYs (8.6%) lower, at 8643 DALYs, whereas that from O 3 was 55.90 DALYs (4.3%) higher, at 1365 DALYs. Given that Chen et al. [44] only showed mortality rates for the combination of NO 2 , particulate matter (PM 10 ), and sulfur dioxide, the endpoint in this study was the increase in mortality when only the relevant substances increased (single-pollutant model), in order to reduce uncertainty. Therefore, considering these health risks simultaneously is difficult. However, decreased NO x emissions will increase the O 3 concentration and the health risk from O 3 , but will substantially decrease the health risk from NO 2 . This condition is due to the greater loss in health risk per concentration of NO 2 (RR-1 = 0.0308) compared with that of O 3 (RR-1 = 0.0017), although the change in O 3 concentration was larger than that in the NO 2 concentration in Osaka City. In this study, we estimated the health risk for the whole of Osaka City. The daytime population of Nishi, Naniwa, Minato, and Nishinari Wards (No. 6 in Figure 5) in Group High and Chuo Ward (No. 13 in Figure 5) in Group Middle, where the increase in O 3 concentration, was large, and exceeded the night-time population; the daytime population of Chuo Ward is about five times larger than the night-time population [48]. Therefore, the health risk in these wards may be high, because the population coming in from outside the ward during daytime may be exposed to increased O 3 .

Health Risk Derived from NO 2 and O 3
When we used the NME to calculate the health risks from O 3 and NO 2 by correcting for the overestimated values of the two scenarios, the decrease in health risk from NO 2 was 14.5 times greater than the increase in health risk from O 3 before the correction and 13.7 times greater after the correction. This result indicates that errors in the model caused no effects on the results obtained in this study. On the other hand, given that we estimated DALYs per year using O 3 concentration estimated for a specific weather pattern from April to September in this study, the DALYs were possibly overestimated, considering that O 3 concentrations decrease in winter. However we concluded that this overestimation is acceptable.

Conclusions
In this study, we quantitatively clarified that Osaka City was in a VOC-sensitive condition, where the sensitive condition has not been clearly measured. The study also demonstrated that different sensitive conditions can occur because of differences in the distributions of NO x emission sources and emissions, even in narrow areas of the same city. Moreover, the results of the study enabled us to predict O 3 concentration variations due to reductions in NO 2 emissions and to quantitatively evaluate associated health risks with DALY that were previously unknown. The risk assessment method and the results obtained through ADMER-PRO and DALY produced valuable findings for determining future policies on air pollution control.
The VOC/NO x concentration ratio was calculated for each ward, and the results showed that the intensity of VOC-sensitivity was higher in coastal areas and lower toward inland areas. In Group High, with a high intensity of VOC-sensitivity, large-scale emissions of NO x , such as those from a power plant, a refuse incineration plant, a sewage sludge treatment plant, and a manufacturer of iron and steel, were located nearby, contributing to the VOC-sensitive condition. The top 12 wards with the highest intensities of VOC-sensitivity exhibited large increases in O 3 concentrations, with a 26-28% reduction in NO x emissions from mobile sources in the Kinki region. However, it is unlikely that O 3 concentrations will increase rapidly in the future, despite the further reduction in NO 2 emissions. Still, if NO 2 decreases in these areas, the VOC-sensitive condition is expected to continue for a time, and O 3 concentrations may increase further as VOC concentrations increase, which may increase the health risk from O 3 . Furthermore, given that the daytime population exceeds the night-time population in coastal and central areas where the increase in O 3 concentration was large, the health risk in those wards may be greater. On the other hand, Group Low showed a large reduction in NO 2 concentrations in the COVID-19 scenario because of the relatively large amount of emissions from mobile sources. In addition, the increase in O 3 concentration was relatively small in the wards belonging to this group, suggesting that the continuous decrease in NO 2 emissions in the future may shift to the NO x -sensitive condition and trigger a decrease in O 3 concentration.
As a result of traffic reduction, the health risk from NO 2 decreased by 811.4 DALYs, and that from O 3 increased by 55.90 DALYs, for the entirety of Osaka City.
This study showed that in Osaka City, which is a VOC-sensitive area, the decrease in health risk due to the reduction in NO x emissions from mobile sources exceeds the increase in health risk due to the increased O 3 concentration, regardless of the sensitive conditions, indicating that the reduction in NO 2 emissions reduces the net health risk from air pollutants. Therefore, the health risk is expected to decrease even under (i) a lockdown due to further spread of COVID-19, (ii) a decrease in traffic volume due to road damage caused by large-scale natural disasters, or (iii) a decrease in NO x emissions due to alternative means of transportation, such as the widespread use of next-generation vehicles. In addition, given the regional variation of vegetation, which are the major source of VOC emissions, the health risk from NO 2 and O 3 in other regions can be estimated, provided that the appropriate amount of BVOC emissions is set.

Data Availability Statement:
The data presented in this study are available upon request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.