Returning impact of long-range PM2.5 transport over East Asia in 2021 after Chinese economic recovery from the COVID-19 pandemic

Changes in the aerosol composition of sulfate (SO 42− ) and nitrate (NO 3− ) from 2012 to 2019 have been captured as a paradigm shift in the region downwind of China. Specically, SO 42− dramatically decreased and NO 3− dramatically increased over downwind locations such as western Japan due to the faster reduction of SO 2 emissions than NO x emissions and the almost constant trend of NH 3 emissions from China. Emissions from China sharply decreased during COVID-19 lockdowns in February–March 2020, after which China’s economic situation seemed to recover going into 2021. Given this substantial change in Chinese emissions, it is necessary to clarify the impact of long-range PM 2.5 transport over East Asia. In this study, ground-based aerosol compositions observed at three sites in western Japan were analysed. The concentrations of PM 2.5 , SO 42− and NO 3− decreased in 2020 (during COVID-19) compared with 2018–2019 (before COVID-19). In 2021 (after COVID-19), PM 2.5 and NO 3− increased and SO 42− was unchanged. This suggests the returning impact of long-range PM 2.5 transport in 2021. From numerical simulations, the status of Chinese emissions during COVID-19 did not explain this returning impact in 2021. This study shows that the status of Chinese emissions in 2021 was close to that before COVID-19.


Introduction
Before the outbreak of COVID-19, strong emission regulations in China had led to a rapid decrease in which had previously dominated aerosol pollution over East Asia 5,6 . This feature is recognised as a paradigm shift in PM 2.5 composition. Such variations with increasing importance of NO 3 − were also found from long-term analyses of the precipitation chemistry over East Asia 7,8 . These changes were due to the different reduction rates of NO x and SO 2 emissions with the faster reduction in SO 2 emission and the steadily high rate of NH 3 emission in China. This trend was expected to continue.
However, unexpected changes in air quality occurred because of the COVID-19 outbreak in December 2019 in the city of Wuhan in Hubei, China 9 . In response to the pandemic, lockdowns were imposed on many cities in China during February and March 2020, and these measures limited human activities such as travel and economic activity 10,11 . Accordingly, strong reductions in anthropogenic emissions were observed 12,13 , and subsequent improvements in air pollution were reported over China 14,15 . Because the long-range transport (i.e. trans-boundary transport) of air pollutants in East Asia is an important environmental issue 12,16,17 , the reduction in anthropogenic emissions in China during COVID-19 will further impact the changes in air pollution over not only China but also downwind regions such as the Korean Peninsula and Japan. This leads to another question: How is the situation after COVID-19? Based on China's estimated GDP, its economy has recovered since the 2020 lockdowns and is projected to keep growing 18 . Unlike the OECD countries, China achieved positive GDP growth in 2020, and this economic recovery likely led to a rebound in anthropogenic emissions. Against this background, it is worth investigating and clarifying, from the perspective of the downwind region, the changes in long-range PM 2.5 transport that resulted from the substantial emission changes that happened before, during and after COVID-19 in China. In the present study, we analysed the changes in aerosol components over western Japan based on a combination of ground-based and satellite observations and sensitivity experiments in numerical modeling simulations.

Results
We  Figure 1 shows the NO 2 column over East Asia as measured by the Ozone Monitoring Instrument (OMI) (see Methods) before, during, and after COVID-19. A high NO 2 column was found over north-eastern China before COVID-19 (Fig. 1a). Compared with before COVID-19 status, the NO 2 column decreased dramatically during COVID-19 over the whole of eastern China (Fig. 1b). Then, although some parts showed decreases, almost all parts of China showed an increased NO 2 column after COVID-19 compared to during it (Fig. 1c). These satellite results suggest substantial deceases in emissions over China in 2020 and then increased emissions along with the economic recovery in 2021.   (Fig. 1c), the returning impact of long-range transport can be assumed. However, because PM 2.5 concentration and composition can also be affected by the meteorological eld and emission sources other than anthropogenic sources, further discussion is required, and we conducted modeling simulations to clarify this point.

Discussion
To identify a potential reason for the variation in air pollutants found in 2021, we conducted numerical simulations for 2018-2021. The experiments are described in performance based on the correlation coe cient (R), the normalised mean bias (NMB) and the normalised mean error (NME), and the modeling performance was judged based on metrics from a review of modeling applications (see Methods for de nitions). The results of the statistical analyses are given in Table 2. For PM 2.5 , the model showed moderate correlation with observations but tended to underestimate the observed concentration; in some cases, the criteria for acceptable model performance were judged as not having been met. This underestimation stemmed mainly from the underestimation of organic aerosol in comparison with a model intercomparison study 19 . For SO 4 2−    and NME < +35%, and the model performance criteria (marked by * ) are R > 0.4, NMB < ±30% and NME < +50%; for NO 3 − , the model performance goals (marked by ** ) are NMB < ±15% and NME < +65%, and the model performance criteria (marked by * ) are NMB < ±65% and NME < +115% 56 . Where M21E19 is superior to M21E20 in a statistical score, the value is given in bold font. and NME < +35%, and the model performance criteria (marked by * ) are R > 0.4, NMB < ±30% and NME < +50%; for NO 3 − , the model performance goals (marked by ** ) are NMB < ±15% and NME < +65%, and the model performance criteria (marked by * ) are NMB < ±65% and NME < +115% 56 . Where M21E19 is superior to M21E20 in a statistical score, the value is given in bold font.
For 2021, two experiments were conducted because of the lack of emission information for that year (Table 1). In these experiments, anthropogenic emissions from China have been perturbed. One was M21E20, using the meteorological eld for 2021 but the Chinese emissions for 2020 (as during  with the natural emissions for 2021; the other was M21E19, also using the meteorological eld for 2021 but now the Chinese emissions for 2019 (as before COVID-19) with the natural emissions for 2021.
Based on these two experiments, the impact of supposing emission changes in 2021 can be estimated. From comparing the simulation results and the ACSA observations for Tsushima and Oki in 2021 (Fig. 2 showed better agreement with the observations. These analyses show that M21E19 outperformed M21E20 in capturing the changes observed after COVID-19. In terms of statistical scores (Table 2), M21E19 also outperformed M21E20. These results clarify that the emission status in 2021 was not the same as during COVID-19 and was close to the status before COVID-19. Chinese economic activity and anthropogenic emissions recovered in 2021 and impacted the air quality in downwind regions through long-range PM 2.5 transport.
Note that experiment M21E20 gave negative changes after COVID-19. Based on the differences between M20E20 and M21E20 (i.e. the M21E20 results minus the M20E20 ones), the impact due to the meteorology and changes in natural emission in 2021 can be investigated. the Korean Peninsula and western Japan are shown clearly (Fig. 4a-c). Based on the analyses of changes in natural emissions, volcanoes distributed mainly over the western island of Kyushu 23,24 showed SO 2 emissions of 605 Gg/period in 2020 and 307 Gg/period in 2021. This decrease could partly explain the negative change in 2021 as found in SO 4 2− , but a negative change in SO 4 2− will lead to a positive change in NO 3 − as examined previously as a paradigm shift in aerosol composition. Therefore, the negative changes seen for PM 2.5 , SO 4 2− and NO 3 − can be attributed to changes in the meteorological eld. To clarify this point, several meteorological elds are shown in Fig. S1 as the difference between 2021 and 2020 (i.e. the M21 meteorological elds minus the M20 ones) overlaid with the observational results. The positive relative humidity over the East China Sea may enhance the aqueous-phase reaction of SO 4 2and did not relate to negative changes in PM 2.5 and SO 4 2− (Fig. S1a). The change in precipitation was inhomogeneous over the East China Sea to western Japan (Fig. S1b). Lower planetary boundary layer (PBL) height found over oceans will lead to suppressed pollutants and did not explain the negative concentrations (Fig. S1c). The change in the 2-m temperature partly explains the NO 3 − change as understood by thermodynamic NO 3 − production. The positive and negative temperatures over western and eastern Japan (Fig. S1d) are generally consistent with the negative and positive changes in NO 3 − concentration over these areas ( Fig. 4c and Fig. S1d). The positive change in the 10-m wind speed could contribute to the negative changes in concentrations because of the faster transport of polluted air mass (Fig. S1e). The most plausible explanation comes from the change in the 10-m wind direction (Fig. S1f), which was positive over eastern China to the East China Sea and western Japan. Further detailed illustrations of wind eld are shown in Fig. 4d and e. During winter, the seasonal wind pattern is generally in the northwest direction 5 , and a northwest wind direction was simulated over the East China Sea to western Japan in 2020 (Fig. 4d); however, a north wind direction dominated in 2021 over western Japan (Fig. 4e). Therefore, a positive difference in wind direction with positive wind speed was calculated over western Japan (Fig. 4f and Figs. S1e and f). The results observed at Tsushima and Oki are also displayed (Fig. S2), and the observations at Tsushima show clearly the change in wind direction from NNW in 2020 to N in 2021. These ndings suggest that the meteorological conditions in 2021-especially the wind direction-prevented long-range PM 2.5 transport. Nevertheless, the observations showed increases in Asia is an issue of concern both currently and for the future 25,26 . The atmospheric input of nitrogen via dry and wet deposition processes over East Asia has been estimated in previous studies, and its signi cance has been clari ed [27][28][29][30] . Targeting the proper management of nitrogen, the nitrogen burden during and after COVID-19 should be clari ed in future work.

Observations
To obtain the spatial distribution of the NO 2 column, which can be regarded as a proxy for NO x emissions 31 , space-borne OMI measurements using the level-3 daily global nitrogen dioxide product (OMNO2d version 3.0) gridded into 0.25°×0.25° 32 was used. We analysed the tropospheric column with clouds screened under the condition of a cloud fraction of less than 30%.
For ground-based observation of PM 2.5 and its components, ACSA observation datasets for 10 sites in Japan were available 33 . The Japanese Ministry of the Environment (MoE) created this observation network in April 2017, and two remote observation sites at Goto and Oki, located in western Japan, were analysed in the present study. The ACSA measured PM 2.5 and the secondary inorganic components of SO 4 2− and NO 3 − with an hourly temporal resolution; this high temporal resolution minimises the effect of the volatilisation of NO 3 − 34 . Observations at the Goto site were unavailable from the autumn of 2020 onwards because of a typhoon; to compensate for these missing observations at Goto, observations at the Tsushima site were used for 2020 and 2021. The locations of the three sites at Goto, Tsushima and Oki are shown in Fig. 1c.
The lockdown in Wuhan began on 23 January 2020 and lasted until 8 April 2020, and those in other Chinese cities occurred within this period 35 . In Japan, a state of emergency was proclaimed for seven prefectures including Tokyo on 7 April 2020, and this was extended nationwide on 16 April 2020; it ended on 14 May 2020 for 39 prefectures except for mega-cities such Osaka and Tokyo, and on 25 May 2020 for all prefectures in Japan 36 . As also seen from the Stringency Index released from One World Data 37 , the countermeasures against COVID-19 were the strictest in China from late January to the end of March, whereas other countries in East Asia had relatively loose measures in place during this period. Given these differences between China and Japan, the period analysed in the present study was chosen as February to March to focus on the variation of long-range PM 2.5 transport according to Chinese anthropogenic emissions.

Numerical modeling simulations
The numerical modeling simulations were performed using the regional chemical transport model of the in Japan were prepared from a measurement report by the Japan Meteorological Agency (JMA) 52 .
Sulphur emissions from ships could in uence air quality, especially in areas close to the sea 53 . Because of the limited ship emission inventory, HTAP version 2.2 was used for 2018 and 2019, and a reduction in SO 2 emissions of 77% for 2020 and 2021 was applied based on a report by the International Maritime Organization 54 . This reduction rate was consistent with a high-resolution emission inventory developed for Japan 55 . To our knowledge, the emission situation in 2021-regarded as after the COVID-19 pandemic -is not yet available, so to consider the emission status in the region downwind of China from the observational facts, two experiments were conducted: (i) M21E20, which assumed the emission levels during COVID-19 for those in 2021, and (ii) M21E19, which assumed the emission levels before COVID-19 for those in 2021. By comparing these two experiments, the emission status in 2021 was investigated.
The model performance was evaluated statistically using the metrics of R, NMB and NME de ned as