Dramatic improvement of aerosol pollution status over the East Asian ocean: from the establishment of Japanese environmental quality standard for PM2.5 in 2009 to its achievement in 2021

The severe aerosol pollution in East Asia has been a focus of much research. In Japan, the environmental quality standard (EQS) for PM2.5 was established in 2009 (daily average, 35 μg m−3; annual average, 15 μg m−3), and its achievement rate was below 50% during the early 2010s. Then, the PM2.5 concentration gradually decreased, the achievement rate improved, and the EQS for PM2.5 was finally achieved (100%) in fiscal year (FY) 2021. Because transboundary aerosol pollution is an important factor in Japanese air quality, here we analysed the long-term dataset of the satellite-measured fine-mode aerosol optical depth (AODf) over the East Asian ocean to reveal the changes in the transboundary aerosol over East Asia. Overall, a decrease in AODf was seen over the entire East Asian ocean during the period analysed. A gradual declining trend in AODf was measured (−4% to −5%/year over the adjacent ocean around Japan) and corresponded well to the trend in PM2.5 concentration observed in Japan (−5.3%/year) during FY2010–FY2021. Due to the domestic contribution in Japan, the negative trend was slightly greater for Japanese PM2.5 concentration than for AODf over the adjacent ocean around Japan, and we concluded that the main reason for the dramatic air quality improvement in PM2.5 in Japan was driven by the improvement of transboundary aerosol pollution over East Asia. In addition, the 12 year analysis period (FY2010 to FY2021) was divided into three parts: stagnation (FY2010 to FY2014), in which PM2.5 and AODf remained the same as they were in FY2010; improvement (FY2015 to FY2018), in which PM2.5 and AODf declined dramatically; and achievement (FY2019 to FY2021), in which PM2.5 and AODf declined further.


Introduction
Aerosol pollution in East Asia has attracted attention because of its severity (Carmichael et al 2002, 2008, Chen et al 2019, 2020).One of the severest pollution events occurred in January 2013 and was a haze over North China with maximum hourly concentrations of particulate matter with an aerodynamic diameter of 2.5 µm or less (PM 2.5 ) of 200-1000 µg m −3 (Zheng et al 2015).Although total anthropogenic emissions grew rapidly in the 2000s, they declined in the 2010s (Li et al 2017, van der A et al 2017, Zheng et al 2018, Kurokawa and Ohara 2020).Among these variations in anthropogenic emissions in China, a decline in PM 2.5 concentration has been reported in China since 2013 (e.g.Zheng et al 2017, Zhai et al 2019, Zhang et al 2019a, Conibear et al 2022).Because transboundary aerosol pollution in East Asia is a concern in the downwind region of Japan and the PM 2.5 concentration shows a longitudinal gradient from high concentrations in the west to low concentrations in the east (Itahashi et al 2017, 2018, Uno et al 2020, Chatani et al 2020 ), it is important to understand conditions including the upwind region.
In Japan, the environmental quality standard (EQS) for ambient PM 2.5 was established in 2009 as a daily average concentration of 35 µg m −3 and an annual average concentration of 15 µg m −3 (MOE n.d.).The air pollutants regulated by the EQS (carbon monoxide, sulphur dioxide, nitrogen dioxide, photochemical oxidants, and suspended particulate matter) have been observed routinely by using the Atmospheric Environmental Regional Observation System (AEROS) at ground-based measurement sites.After the EQS for PM 2.5 was established, PM 2.5 concentration has also been measured intensively in Japan by using AEROS.The AEROS measurement sites are categorized as ambient air pollution monitoring stations (APMSs) for ambient air quality and roadside air pollution monitoring stations (RAPMSs) for air pollution, focusing on automobiles.Only 46 sites (34 APMSs and 12 RAPMSs) measured PM 2.5 in FY2010 (FY is from April to March of the next year as Japanese fiscal year), but the number of measurement stations has been increasing, surpassing 500 sites in FY2014 and 1000 sites in FY2016, and a total of 1098 sites (858 APMSs and 240 RAPMSs) measured PM 2.5 in FY2021.The EQS of PM 2.5 was judged to have been achieved when both daily and annual concentration were lower than the target values at each measurement site, and then the country-level achievement rate was calculated.The annual average concentration of PM 2.5 and the achievement rate of Japanese EQS at APMSs and RAPMSs from FY2010 to FY2021 is summarized in figure 1 based on the report from the Ministry of the Environment, Japan (MOE 2023).The finalized data of FY2021 was published in March 2023 as the latest monitoring data.From FY2010 to FY2014, the PM 2.5 concentration was around 15 µg m −3 for APMSs and above 15 µg m −3 for RAPMSs and the achievement rate of the EQS was 30%-40% for APMSs and 10%-30% for RAPMSs.Thus, it was difficult to maintain the EQS in Japan during the early 2010s.However, in FY2015, PM 2.5 concentration declined to 13.1 µg m −3 at APMSs and 13.9 µg m −3 at RAPMSs, and the achievement rate dramatically improved to 74.5% at APMSs and 58.4% at RAPMSs.From FY2016 to FY2018, PM 2.5 concentration showed a gradual decline to 11-12 µg m −3 and the achievement rate was around 90% at both APMSs and RAPMSs.In FY2019-FY2020, PM 2.5 concentration reached below 10 µg m −3 at APMSs and around 10 µg m −3 at RAPMSs, and the achievement rate was 98%-99%.Finally, the EQS was achieved (100%) in FY2021 at both APMSs and RAPMSs, with a PM 2.5 concentration of 8.3 µg m −3 at APMSs and 8.8 µg m −3 at RAPMSs.It took 12 years from the establishment to the achievement of the EQS for PM 2.5 in Japan.
Japanese anthropogenic emissions showed a gradual reduction during this period (Kurokawa and Ohara 2020), and such regulations will contribute to improving the PM 2.5 air quality in Japan (Wakamatsu et al 2013).However, transboundary aerosol pollution is also an important factor in causing these changes in Japan.Therefore, in this study, to focus on the changes in the transboundary aerosol pollution over East Asia, we analysed the long-term trend in the satellite-measured fine-mode aerosol optical depth (AOD f ) over the East Asian ocean because the behaviour of AOD is a proxy for aerosol (van Donkelaar et al 2010, Itahashi et al 2012, 2016, 2021).Section 2 describes the methodology for the analysis of AOD f .Section 3 presents the changes in AOD f over the East Asian ocean.Section 4 discusses the detailed analysis of the changes in AOD f .Section 5 ends this paper with conclusions and future perspectives.

Methodology
The AOD represents the attenuation of sunlight by aerosols and is an important measure of the aerosol column concentration (Kaufman et al 2002).One of the longer periods of coverage of spacebased AOD has been achieved by the retrieval from the Moderate Resolution Imaging Spectroradiometer (MODIS) onboard NASA's Terra and Aqua satellites.Terra was launched on 18 December 1999, and the data are available from 24 February 2000 onward.Aqua was launched on 4 May 2002, and the data are available from 5 July 2002 onward.Terra crosses the equator in a descending mode (southbound) at about 10:30 a.m.local time and Aqua crosses the equator in an ascending mode (northbound) at about 1:30 p.m. local time.
In this study, we analysed both MODIS datasets from the Terra and Aqua satellites.Level 2 of the MOD04_3K product for Terra (NASA n.d.a) and the MYD04_3K product for Aqua (NASA n.d. b) in the latest Collection 6.1 (Levy et al 2013) were used.This 3 km product applied the same algorithm to the 10 km product for the cloud identification (Remer et al 2013).In the MOD04_3K and MYD04_3K products, the AOD (parameter name: 'Optical_Depth_Land_And_Ocean') was taken and analysed.The algorithm for determining aerosol characteristics at 550 nm has been validated with co-located surface observations from direct sun/sky radiometers from the Aerosol Robotic Network (AERONET), and the expected errors are ±0.05 ± 0.20 AOD over land and ±0.03 ± 0.05 AOD over the ocean (Remer et al 2013).The errors over land are larger compared with those over the ocean, and AOD itself includes coarse particles, such as mineral dust, which is another factor causing aerosol pollution.Therefore, the data of the small particle ratio at 550 nm provided over only the ocean (parameter name: 'Optical_Depth_Ratio_Small_Ocean_055micron') were multiplied by AOD, and we calculated the AOD for fine-mode aerosol (AOD f ).Additionally, calculated AOD f data at locations adjoining the land grid were not used to reduce the effect of errors in retrieval over land.The usefulness of AOD f for detecting anthropogenic aerosol has been reported previously (Anderson et al 2005, Kaufman et al 2005).Based on this approach, we focus on the aerosols attributed mainly to anthropogenic sources.
Level 2 of the swath data along the satellite orbit was converted into 0.05 • × 0.05 • gridded data for analysis.The AOD f was calculated each day for Terra and Aqua.If the value of AOD f was available both for Terra and Aqua, the averaged value was taken, and if the value of AOD f was available for only Terra or Aqua, the single value was used.Then, the AOD f data for each day were used to calculate the annual mean by averaging 365 (or 366) data through the fiscal year (hereafter called the annual mean).In addition, because the achievement of EQS for PM 2.5 is judged by the 98th percentile of daily average values in the year, the 98th percentile value was also calculated for AOD f (hereafter called the 98%ile daily mean).This approach using the 98%ile value would be also useful to reduce the uncertain high AOD f value.

Results
The long-term trend of the spatial distributions of the annual mean AOD f is shown in figure 2. The gradient of the annual mean AOD f was high in the west (>0.5 close to the Asian continent; dark red in figure 2) and low in the east (<0.1; blue in figure 2).From the East China Sea and the Sea of Japan to mainland Japan, the value of annual mean AOD f was 0.2-0.3(green in figure 2) in FY2010.However, as the highest value of annual mean AOD f along the coastline of the Asian continent declined, AOD f around the ocean around Japan also declined.The area of relatively high annual mean AOD f (green to red in figure 2; from Asian continent to mainland Japan) shrank.For example, the area-averaged AOD f value over the Yellow Sea and East China Sea (here simply defined as the box area of 120-125 • E and 25-45 • N) declined from 0.38 in FY2010 to 0.26 in FY2021 (corresponding to 0.67 when normalized to the value in FY2010).Comparing the spatial distributions of annual mean AOD f for  FY2010 and FY2021 showed that aerosol pollution over the entire East Asian ocean has diminished.
The improvement of aerosol pollution suggested by the annual mean AOD f (figure 2) was also seen in 98%ile daily mean AOD f (figure 3).Due to the timing of the available satellite measurements, although 98%ile daily mean AOD f exhibited a mosaic spatial distribution (e.g. a higher/lower AOD f value in near grid) compared with the annual mean value, 98%ile daily mean AOD f also showed a declining trend from FY2010 to FY2021.A high value (>0.5; dark red in figure 3) stretched over mainland Japan in FY2010, whereas it only reached western Japan in FY2021.The analysis of the year-to-year variation of 98%ile daily mean AOD f also showed that aerosol pollution over the East Asian ocean has gradually improved.

Discussion
The trend in annual and 98%ile daily mean AOD f was analysed based on linear regression analysis (figure 4).Similar to the year-to-year variation of AOD f (figures 2 and 3), the annual and 98%ile daily mean AOD f also declined.The negative trend was observed for annual mean AOD f over the entire domain over the East Asian ocean (figure 4(i)).The negative trend was also observed for 98%ile daily mean AOD f except for over the southern part of the domain, especially over eastern Taiwan, where the trend was complex and ambiguous (figure 4(ii)).The trend over the adjacent ocean around Japan was −4% to −5%/year both for annual and 98%ile daily mean AOD f , with the value slightly higher in 98%ile daily mean AOD f .The ground-based PM 2.5 at APMSs in Japan (figure 1) showed a −5.3%/year trend from FY2010 to FY2021 against the entire period averaged PM 2.5 concentration.The greater negative trend for PM 2.5 concentration than for AOD f was caused by the difference between the surface PM 2.5 and vertically accumulated AOD f and by the domestic contribution in Japan.Including the Japanese domestic contribution would increase the reduction of AOD f over eastern Japan; however, the declining trends of the annual and 98%ile daily mean AOD f were −4% to −5%/year over the whole analysed domain.Therefore, this analysis suggested that the improvement of PM 2.5 in Japan was dominated mainly by the change in East Asian aerosol pollution.
AOD f had a large longitudinal dependency, with a western high and eastern low (figures 2 and 3).To investigate the year-to-year variation of AOD f focusing on the longitudinal change, AOD f was averaged over 25-45 • N (fully covering Japan) (figure 5).Both annual and 98%ile daily mean AOD f showed a gradual reduction from the early 2010s (warm colours in figure 5) to FY2021 (cold colours in figure 5).This year-to-year reduction was also supported by the analysis of AOD f normalized to the value for FY2010.The normalized AOD f showed no change or a slight increase during the early 2010s (warm colours in figure 5), almost no change or a slight decrease during the middle 2010s (green in figure 5), and then a clear decrease during the late 2010s and to FY2020 and FY2021 (cold colours in figure 5).This gradual reduction was observed over all longitudes.The rates of decrease in FY2010 and FY2021 around Japan corresponded well with those over the Yellow Sea and East China Sea, as discussed in section 3. Including the Japanese domestic change would cause a further reduction eastward at 130 • E, but the decreasing ratios in the normalized AOD f showed almost no variation from 125 to 145 • E. Thus, this result also Finally, the regional averaged AOD f over the ocean around Japan (25-45 • N and 125-145 • N; grey region in figure 5) were analysed.Figure 6 shows the domain-averaged values of annual and 98%ile daily mean AOD f and their correspondence to PM 2.5 concentration at APMSs sites in Japan (figure 1).This scatter plot showed that AOD f and PM 2.5 concentration corresponded well, with a correlation coefficient (r) of 0.95 and 0.94 for annual and 98%ile daily mean AOD f , respectively.The F-test was conducted, and then Welch's t-test was used to calculate the significance level.The results indicated that both cases were statistically significant (p < 0.001).Based on the source-receptor analysis of the numerical simulation, it was reported that the contribution of China to PM 2.5 in Fukuoka (western Japan) exceeded 60% and the contribution within Japan was approximately 20% (Uno et al 2020).It seems reasonable that the decline in Chinese emissions has a linear relationship with AOD f in a downwind region such as Japan.AOD f averaged over the ocean around Japan and normalized to the value for FY2010 is shown in figure 7. The gradual year-to-year decline in aerosol pollution was observed over the ocean around Japan.Based on the comparison with the groundbased PM 2.5 observation data at APMSs shown in figure 1, the long-term 12 year trend from FY2010 to FY2021 was divided into three periods (dotted lines in figures 6 and 7).The first period from FY2010 to FY2014 (5 years) was stagnation, during which both PM 2.5 concentration (figure 1) and AOD f (figure 6) remained similar to those in FY2010.The second period from FY2015 to FY2018 (4 years) was improvement, in which both PM 2.5 concentration (figure 1) and AOD f (figure 6) declined dramatically.The third period from FY2019 to FY2021 (3 years) was achievement, in which both PM 2.5 concentration (figure 1) and AOD f (figure 6) showed further decline and the EQS for PM 2.5 was achieved in FY2021.This result supported the close correspondence between PM 2.5 in Japan and AOD f over the ocean around Japan.The aggressive anthropogenic emission controls introduced by the Chinese government after  broad impact over East Asia.The year-to-year variation could be also affected by the change in meteorology; however, previous studies found that the emission variation was the main reason for the decline in PM 2.5 over China (Chen et al 2020, Zhang et al 2019b) and its downwind area (Cheng and Hsu 2019, Kang et al 2020, Cha et al 2023).Based on these previous studies and our study showing the dramatic decline of AOD f over entire East Asian ocean, we concluded that the main driver for the improvement of aerosol pollution in Japan could be the improvement of transboundary aerosol pollution in East Asia.

Conclusions
We analysed AOD f over the East Asian ocean from FY2010 (the establishment of the EQS of PM 2.5 in Japan) to FY2021 (the achievement of the EQS).The analysis of AOD f showed a clear improvement in aerosol pollution over the entire East Asian ocean.The declining trends in PM 2.5 and AOD f agreed well, and the 12 year period from FY2010 to FY2021 was divided into three periods: stagnation (FY2010 to FY2014), improvement (FY2015 to FY2018), and achievement (FY2019 to FY2021).Based on the analysis throughout this study, we clarified that the aerosol pollution in Japan was probably driven mainly by the improvement in the long-range transport over the East Asian region.
In this study, we provided an overview of the long-term trend in PM 2.5 concentration from AOD f ; however, the impact of the long-range transport over East Asia and the domestic contribution from Japan should be estimated quantitively.Numerical modelling could be used for estimations to distinguish these two factors.In addition, we did not fully discuss the substantial change caused by two important factors after 2020, namely, the ship SO 2 emission regulation (IMO n.d., Sakurai et al 2021, Hayami et al 2022) and the impact of COVID-19 (Liu 2020, Venter et al 2020, Gkatzelis et al 2021, Wang et al 2021).These two factors could affect the aerosol pollution around Japan over different time periods (Itahashi et al 2021).The impact of Chinese lockdown during COVID-19 on the transboundary aerosol pollution was probably limited in FY2019 (Itahashi et al 2022), and the economic recovery and related anthropogenic emissions should be examined carefully.Therefore, long-term, detailed emission inventories considering such factors must be developed as important input datasets for numerical modelling (Zheng et al 2021, Chatani et al 2023).

Figure 1 .
Figure 1.PM2.5 concentrations (open circles) and the achievement rate of EQS for APMSs (grey bars) and RAPMSs (dark red bars) in Japan from FY2010 to FY2021.

Figure 2 .
Figure 2. Annual mean AOD f measured by MODIS over the East Asian ocean from to FY2021.

Figure 3 .
Figure 3. 98%ile daily mean AOD f measured by MODIS over the East Asian ocean from FY2010 to FY2021.

Figure 4 .
Figure 4. Trends in (i) annual and (ii) 98%ile daily mean AOD f over the East Asian ocean from FY2010 to FY2021 based on linear regression analysis.The trend is calculated against the entire period averaged (i) annual and (ii) 98%ile daily mean AOD f at each panel.

Figure 5 .
Figure 5. Variation of (i) annual and (ii) 98%ile daily mean AOD f for each year averaged over 25-45 • N and shown as the longitudinal dependency.In each panel, the top graph is absolute AOD f values and the bottom graph is the AOD f values normalized to the value for FY2010.

Figure 6 .
Figure6.Correspondence of (i) annual and (ii) 98%ile daily mean AOD f averaged over ocean around Japan (grey area in figure5) and PM2.5 concentration at APMSs sites in Japan (figure1).

Figure 7 .
Figure 7. Change in (i) annual and (ii) 98%ile daily mean AOD f normalized to the value for FY2010 averaged over ocean around Japan (grey area in figure 5).