Introduction

Photochemical oxidants (Ox), which mainly consist of tropospheric ozone (O3), are related to chemical reactions involving nitrogen oxides (NOx) and volatile organic compounds (VOCs)1. Ox causes urban smog, poses major risks to human health and the natural environment, and is an important greenhouse gas2. The Japanese environmental quality standard (EQS) for Ox was established in 1973, and hourly Ox values should not exceed 0.06 ppm (118 μg/m3)3. In addition, the Air Pollution Control Law (Paragraph 1, Article 23) stipulates that when hourly Ox concentration exceeds 0.12 ppm and the status is expected to continue due to weather conditions, a warning should be issued to prevent damage to human health or the living environment. The EQS has not been satisfied at most monitoring stations in Japan, and this is a continuing issue that should be solved4,5. In general, high Ox concentrations and subsequent Ox warnings have generally occurred in spring over western Japan and in summer toward eastern Japan (e.g., Osaka, Nagoya, and Tokyo). This seasonal variation of Ox in Japan is related to the transboundary Ox transport from the Asian continent during spring and local photochemical production in Japan during summer6,7,8,9,10,11. A recent report on Japanese local photochemical production found a decreasing trend in the 3-year average of the annual 99th percentile of the daily maximum 8-h concentration (a new index for evaluating Ox), especially over the Kanto region (e.g., Tokyo)12, and this result suggested that transboundary Ox pollution has an important effect over Japan. The worsening of Ox pollution in East Asia has been a concern recently13,14,15; thus, the effect of transboundary transport on Ox behaviour is the focus of the present work.

Over Kyushu Island, western Japan, high Ox concentrations and related warnings are generally issued during spring. The summary of the number of Ox warning days from 2000 to 2021 over Kyushu Island is presented in Fig. 1. The Ox warning days have mostly been observed in May, and sometimes in April and June. In Nagasaki Prefecture, located on western Kyushu Island, an Ox warning was issued for the first time in May 2006. However, there was one warning day in September 2020 in Nagasaki Prefecture (Fig. 1). This warning was issued on Goto Island (westernmost Nagasaki Prefecture, Fig. 1) on 8 September, 2020 at 15:00 local time when an Ox concentration of 123 ppb was observed, the maximum concentration was recorded as 126 ppb at 16:00 local time, and the warning was canceled at 19:00 local time. In the present study, this severe Ox concentration event was analysed because this is the first case of a warning issued in autumn in Nagasaki Prefecture.

Figure 1
figure 1

Record of the Ox warning days for severe concentrations (see, text for this definition) from 2000 to 2021 over seven prefectures on Kyushu Island, Japan. The colours indicate the month. At the start of this study, the confirmed values of AEROS were reported up to 202112, and no days were reported in 2000–2005 and 2021 over all prefectures in Kyushu Island. The maps were generated with GMT (https://docs.generic-mapping-tools.org/dev/index.html).

Full size image

Results

First, the meteorological conditions in September 2020 were analysed. Typhoons Maysak (2020) and Haishen (2020) occurred in this period16. Typhoon Maysak (2020) formed on 27 August, 2020 over the eastern Philippines and passed over western Kyushu Island on 2–3 September, and Typhoon Haishen (2020) formed on 31 August, 2020 over the Northwest Pacific and passed over western Kyushu Island on 6–7 September (Supplementary Fig. S1). The Ox warning on Goto Island in September 2020 was issued the day after Typhoon Haishen (2020) passed. Typhoons are usually associated with hazards, such as strong winds, storm surges, and rainfall, that can cause considerable damage and disruption. Although such a strong wind field could simply result in good ventilation and reduce air pollutant concentrations, several studies have reported that typhoons degraded air quality in terms of PM2.517,18, Ox19,20, and deposition21,22. Because the Ox warning issued in September 2020 occurred in a remote area of Goto Island, westernmost Nagasaki Prefecture, the horizontal distribution over East Asia should also be examined. Thus, a numerical modelling simulation covering the whole of East Asia (Supplementary Fig. S2) was applied to analyse this severe Ox event further (see “Methods” section for details of the modelling).

The surface pressure anomaly, which is a unique characteristic of typhoons, was investigated and evaluated with the typhoon best-track data16. The typhoon tracks (Supplementary Fig. S1) and the simulated surface pressure anomaly are shown in Fig. 2. In this comparison, the anomaly was calculated by the difference between the surface pressure during each typhoon and the 1-month average (from 15 August, 2020 to 14 September, 2020; see “Methods” section). The typhoon best-track data and the simulated lower anomaly associated with Typhoons Maysak (2020) and Haishen (2020) agreed well. The comparison shown in Fig. 2 indicated that the general features of typhoon movement were captured well in the present modelling simulations.

Figure 2
figure 2

Best-track data of (top) Typhoon Maysak (2020) and (bottom) Typhoon Haishen (2020) with surface pressure anomalies simulated by the WRF meteorological model. The maps were generated with GMT (https://docs.generic-mapping-tools.org/dev/index.html).

Full size image

The meteorological parameters were measured by the Automated Meteorological Data Acquisition System (AMeDAS) at the Fukue observatory, which is on Goto Island (Fig. 1), and the corresponding simulated results are shown in Fig. 3. The pressure (Fig. 3a) was simulated well, and the decrease in pressure was larger in the period when Typhoon Haishen (2020) passed compared with that when Typhoon Maysak (2020) passed. The precipitation (Fig. 3b) was generally underestimated but the timing was well captured. The wind speed (Fig. 3c) and direction (Fig. 3d) were also well captured, although the wind speed tended to be overestimated. This overestimation was mainly due to the insufficient wind reduction over the land. In addition to the case at Fukue in Fig. 3, another validation at Nomozaki, southern Nagasaki Prefecture is presented in Supplementary Fig. S3. The Nomozaki site observed a record-breaking maximum instantaneous wind velocity (59.4 m/s, the corresponding wind velocity in 1 h was 43.7 m/s) during Typhoon Haishen (2020), and such wind speeds were generally captured by the modelling system. These evaluations of the meteorological parameters showed that the present modelling system reproduced the meteorological field well, especially the features when the typhoons passed.

Figure 3
figure 3

Comparison of meteorological fields of (a) pressure, (b) precipitation, (c) wind speed, and (d) wind direction from observations (grey) and the model (black) at Fukue, AMeDAS station in Nagasaki Prefecture. The purple shading indicates the days affected by Typhoon Maysak (2020) and Typhoon Haishen (2020). The maps were generated with GMT (https://docs.generic-mapping-tools.org/dev/index.html).

Full size image

The modelling performance for O3 concentration in Nagasaki Prefecture is shown in Fig. 4. The O3 concentration was lower during two typhoon periods (purple shading, see also Fig. 3). After Typhoon Maysak (2020) passed on 3 September, 2020, the O3 concentration was slightly increased to 60–80 ppbv at all sites in Nagasaki Prefecture. Subsequently, due to Typhoon Haishen (2020) passing, O3 concentration was decreased to 20 ppbv on 7 September, 2020. There was a sharp increase in O3 concentration after Typhoon Haishen (2020) passed, and the O3 concentration reached its peak from 8 to 9 September, 2020. This maximum concentration was close to 100 ppbv in cities in the main parts of Nagasaki Prefecture, whereas on the remote islands of Nagasaki Prefecture (Tsushima, Iki, and Goto) it was close to 120 ppbv, which is the alert level for severe O3 pollution. Of these three remote sites, only Goto Island issued an Ox alert, and the present modelling system captured this high O3 concentration exceeding 120 ppbv well. The statistical metrics scores (see the “Methods” section for their definitions) are also shown in Fig. 4. The correlation coefficient (R) was from 0.61 to 0.91, the normalized mean bias (NMB) was from + 5.8% to + 31.2%, and the normalized mean error (NME) was from + 15.9% to 31.2%. Except for the grid corresponding to Nagasaki city (#3 in Fig. 4), these metrics met the model performance criteria. These evaluations of O3 concentration also demonstrated that the present modelling system reproduced O3 pollution in Nagasaki Prefecture. In addition to these validations in Nagasaki prefecture, the vertical O3 profiles are compared in Supplementary Fig. S4 and show that O3 concentration from the surface to the middle stratosphere (approximately 20 km) was also captured well by the modelling system. The NO2 column density observed by the satellite is compared in Supplementary Fig. S5 and a high NO2 concentration in the eastern part of China was generally captured. Based on these results, we confirmed the modelling system in this study captured the meteorological field and O3 concentration in early September 2020. The reasons for the Ox alert event over Goto Island despite the typhoon passing are discussed.

Figure 4
figure 4

Comparison of O3 concentrations from observations (red) and the model (black) at all 17 sites in Nagasaki Prefecture. The purple shading indicates the days affected by typhoons (Fig. 3). The dark-red shading indicates the period of Ox alerts. When the measurement sites were allocated in the same model grid, the averaged values were used to evaluate the model, and the minimum and maximum concentrations are indicated by red shading. The statistical scores are shown in the box. The maps were generated with GMT (https://docs.generic-mapping-tools.org/dev/index.html).

Full size image

Discussion

To identify the reasons for the high O3 concentration that triggered the alert, the horizontal distribution pattern from 7 to 9 September, 2020 is shown in Fig. 5. At midnight on 7 September (Fig. 5a), high O3 concentration (70–90 ppbv; yellow to orange, which exceeds the Japanese EQS) was seen over the East Asian ocean adjacent to the Asian continent. In addition, the typhoon was over Goto Island, and westerly or north-westerly wind fields prevailed over the East China Sea. At this time, the O3 concentrations over Goto Island and other sites in Nagasaki Prefecture were low (20 ppbv; light green, see also Fig. 4). The high O3 concentrations stretched into the East China Sea (Fig. 5b), and O3 was produced during the daytime and the concentration increased on 7 September (Fig. 5c). During the night of 7 September (Fig. 5d), the high O3 concentration over mainland China was decreased to a low concentration by NO titration and deposition, whereas the high O3 concentration (120 ppbv; dark red, alert level) remained over the East China Sea. During this time, Typhoon Haishen (2020) was over the northern part of the Korean peninsula, and southerly and south-westerly wind fields transported this high O3 concentration to Kyushu Island (Fig. 5e and f). During the daytime on 8 September (Fig. 5g), this air mass with O3 higher than 120 ppbv reached Goto Island on the western edge of Kyushu Island; hence, the Ox alert was issued. Then, the air mass was transported to the Tsushima Strait (between Kyushu Island and the Republic of Korea) due to the south-westerly wind affected by Typhoon Haishen (2020) in Liaoning Province, northeastern China (Fig. 5h and i). This analysis of the simulated O3 spatial distribution suggested transboundary O3 transport from the Asian continent to the western part of Japan, and the wind system caused by Typhoon Haishen (2020) was important in determining the air mass transport. Based on the analysis at the top of the boundary layer (approximately 750 hPa) shown in Supplementary Fig. S5, this high transboundary O3 transport was limited near the surface level.

Figure 5
figure 5

Simulated spatial distributions of O3 concentration at the surface level over East Asia before, during, and after the Ox alert on Goto Island, Nagasaki Prefecture. The maps were generated with gtool3 (http://www.gfd-dennou.org/library/gtool/index.htm.en).

Full size image

The transboundary O3 transport after the typhoon passing was clarified, and mitigation of this severe pollution is a concern. Finally, the effects of emission regulations are discussed based on the sensitivity simulations. The changes in O3 concentration when 20% reductions in Chinese VOC and NOx emissions (\(\Delta{C}_{V}\) in Eq. (4) and \(\Delta{C}_{N}\) in Eq. (5), see “Methods” section) were applied are shown in Figs. 6 and 7, respectively, and those for simultaneous VOC and NOx emission reductions are shown in Supplementary Fig. S6. The reduction in anthropogenic Chinese VOC emissions always helped to reduce O3 concentration during the analysis period (Fig. 6). In contrast, the reduction in anthropogenic Chinese NOx emissions had complex effects because it caused both O3 decreases and increases (Fig. 7). The increase in O3 was observed only over mainland China from night to early morning and was associated with the weakening NO titration effect due to the NOx emission reduction. During the daytime, Chinese NOx emission reduction led to an O3 decrease, and moreover, it always led to an O3 decrease over the downwind region of Japan and this impact was greater than Chinese VOC emission regulation. Therefore, NOx emission regulations in China should consider these negative aspects. Recently, the aggravating effect of NOx emission regulations on O3 pollution has been reported in China23,24, and the present results are also consistent with these findings.

Figure 6
figure 6

Simulated spatial distributions of change in O3 concentration with anthropogenic VOC emissions in China decreased by 20% (\(\Delta{C}_{V}\) in Eq. (4)). Red indicates increased O3 concentration in the sensitivity simulation (i.e., emission reduction disbenefit), whereas blue indicates decreased O3 concentration (i.e., emission reduction benefit). The maps were generated with gtool3 (http://www.gfd-dennou.org/library/gtool/index.htm.en).

Full size image
Figure 7
figure 7

Simulated spatial distributions of change in O3 concentration with anthropogenic NOx emissions in China decreased by 20% (\(\Delta{C}_{N}\) in Eq. (5)). Red color indicates increased O3 concentration in the sensitivity simulation (i.e., emission reduction disbenefit), whereas blue color indicates decreased O3 concentration (i.e., emission reduction benefit). The maps were generated with gtool3 (http://www.gfd-dennou.org/library/gtool/index.htm.en).

Full size image

Due to global warming, the frequency of typhoons will decrease, whereas their intensity will increase25,26. Similar future changes are also reported for the western North Pacific, and will lead to destructive threats by stronger wind fields and intense precipitation27,28. As shown in the present study, high levels of Ox over East Asia could be related to the meteorological fields associated with typhoons. Although precursor emission reduction could mitigate both global warming and air quality29, the behaviour of air pollutants related to the meteorological field should be continuously paid attention in future studies.

Methods

Observations

Ground-based observation of air pollutants that are regulated by EQS (carbon monoxide (CO), sulfur dioxide (SO2), nitrogen dioxide (NO2), Ox, suspended particulate matter (SPM), and particulate matter with an aerodynamic diameter less than 2.5 μm (PM2.5)) are routinely measured by the Atmospheric Environmental Regional Observation System (AEROS) (https://soramame.env.go.jp). AEROS is divided into monitoring of ambient air quality by ambient air pollution monitoring stations (APMSs) and of air pollution, particularly from automobiles, by roadside air pollution monitoring stations (RAPMSs). In this study, APMSs data was used in Nagasaki Prefecture. The Ox measurements were obtained by the neutral-buffered potassium iodide method or ultraviolet absorption spectrometry. Although ultraviolet absorption spectrometry only detects O3, it was approved in 1996 as the standard method for measuring Ox concentrations. For the consistency of the measurement dataset, Ox concentrations based on the ultraviolet absorption spectrometry at APMSs were used in this study. Nagasaki Prefecture, which is the focus of this study, is on the western coast of Japan (Fig. 1). Previous studies have clarified transboundary air pollution using measurement data here because of its location, especially at remote island sites (e.g., Goto Island) where domestic emissions are small30,31,32. There are 17 sites in total in Nagasaki Prefecture, and the model grids numbered 1 to 4 (Fig. 4) include 1, 6, 5, and 2 APMSs measurements sites in the same model grid, respectively, and the averaged Ox concentration in the same model grid was compared with the simulation result.

In addition to the ground-based observation in Nagasaki Prefecture, vertical O3 profiles measured by ozonesonde distributed from the World Ozone and Ultraviolet Radiation Data Centre (WOUDC) (https://woudc.org/data/explore.php?dataset=ozonesonde) were used. During the first part of September 2020, the available ozonesonde data in East Asia were from 2 and 9 September, 2020 at King’s Park, Hong Kong (22.31° N, 114.17° E) and 3 and 8 September, 2020 at Pohang, the Republic of Korea (36.03° N, 129.38° E). A total of four vertical profile data were compared with the model result. In terms of the precursor of O3, satellite-derived tropospheric NO2 column densities can be a proxy of NOx emissions33. The level 2 of swath measurement data taken by Ozone Monitoring Instrument (OMI) onboard the Aura was used34, and the information of the averaging kernel was applied for model output for fair comparison.

Modelling simulation

To understand the horizontal distribution pattern of Ox, the modelling simulation was conducted using the regional chemical transport model of the Community Multiscale Air Quality (CMAQ) version 5.3.335. The simulation domain covering East Asia was configured by 215 × 120 grid points with a horizontal resolution of 36 km (Supplementary Fig. S2), and 44 non-uniformly spaced layers from the surface to 50 hPa with a vertical resolution of approximately 20 m for the surface layer36. The meteorological fields for driving CMAQ were prepared by the Weather Research and Forecasting (WRF) model version 4.437. Modelling simulations were started from 8 August, 2020 and analysed from 15 August, 2020 to 14 September, 2020, discarding the first week as a spin-up period. SAPRC07tic38 and aero739 were used for gas-phase and aerosol chemistry, respectively, and the NO2 aqueous-phase reaction in the SO42− oxidation process40 and better consideration of trace metal ion-catalysed O2 oxidation41 were included based on previous studies. The initial and lateral boundary conditions were prepared from autumn-averaged values of the extended CMAQ over the northern hemisphere (H-CMAQ)42.

The emissions inventories were prepared as follows. Anthropogenic emissions except those in China were based on Hemispheric Transport of Air Pollution (HTAP) version 343. The latest available year is 2018 and this dataset with monthly variation is used. For China, the emission projections were based on the Multi-resolution Emission Inventory in China (MEIC), which includes the variation of Chinese emissions during the COVID-19 pandemic43. From this estimation, the anthropogenic emissions had almost recovered in August–September to the levels before the COVID-19 period, consistent with the author’s previous investigation into the transboundary PM2.5 transport31. Because the ship SO2 emissions were regulated according to the International Convention for the Prevention of Pollution from Ships (MARPOL) from 1 January, 2020, a reduction in SO2 emissions of 77% was applied based on a report by the International Maritime Organization (IMO)45. Biogenic emissions were prepared from the Model of Emissions of Gases and Aerosols from Nature (MEGAN)46 using the WRF-simulated meteorological field, and biomass-burning emissions were taken from the Global Fire Emissions Database (GFED) version 4.147. SO2 emissions from 17 volcanoes in Japan were prepared from a measurement report by the Japan Meteorological Agency48.

The model performance was evaluated statistically using metrics R, NMB, and NME, defined as

$$R = \frac{{\mathop \sum \nolimits_{1}^{N} \left( {O_{i} - \overline{O}} \right)\left( {M_{i} - \overline{M}} \right)}}{{\sqrt {\mathop \sum \nolimits_{1}^{N} \left( {O_{i} - \overline{O}} \right)^{2} } \sqrt {\mathop \sum \nolimits_{1}^{N} \left( {M_{i} - \overline{M}} \right)^{2} } }} ,$$
(1)
$$NMB = \frac{{\mathop \sum \nolimits_{1}^{N} \left( {M_{i} - O_{i} } \right)}}{{\mathop \sum \nolimits_{1}^{N} O_{i} }} ,$$
(2)
$$NME = \frac{{\mathop \sum \nolimits_{1}^{N} \left| {M_{i} - O_{i} } \right|}}{{\mathop \sum \nolimits_{1}^{N} O_{i} }} ,$$
(3)

where N is the total number of paired observations (O) and models (M), and the corresponding averages are \(\overline{O }\) and \(\overline{M }\), respectively. The recommended metrics for 1-h O3 based on a literature review for the United States49 reported as the model performance goals are R > 0.75, NMB <  ± 5%, and NME <  + 15%, and the model performance criteria are R > 0.50, NMB <  ± 15%, and NME <  + 25% with no cutoff for R and a 40 ppbv cutoff for NMB and NME.

To estimate the change in O3 production from the VOC and NOx precursors, sensitivity simulations that perturbed anthropogenic VOC emissions, anthropogenic NOx emissions, and simultaneous anthropogenic VOC and NOx emissions in China were performed in addition to the base-case simulation, which includes all emission sources. The perturbation range was a 20% emission reduction from the base-case, which is typical for potential emission controls8. The change in O3 concentration from the base-case simulation was calculated based on the following three definitions. In this study, the changes in concentration were directly shown and not scaled to be expressed as sensitivities.

$${\Delta }C_{V} { = } C_{{{\text{Chinese VOC emission }}20{\text{\% reduction case}}}} - C_{{{\text{Base}} - {\text{Case}}}} ,$$
(4)
$${\Delta }C_{N} { = }C_{{{\text{Chinese NOx emission }}20{\text{\% reduction case}}}} - C_{{{\text{Base}} - {\text{Case}}}} ,$$
(5)
$${\Delta }C_{NV} { = } C_{{{\text{Chinese VOC and NOx emission }}20{\text{\% reduction case}}}} - C_{{{\text{Base}} - {\text{Case}}}} ,$$
(6)