The impacts of COVID-19, meteorology, and emission control policies on PM_2.5 drops in Northeast Asia

Since the first reported case of COVID-19 in Wuhan, China in December 2019, the disease spread rapidly and national emergencies were declared in China and South Korea in mid-January 2020. Economic activities and traffic were limited until restrictions eased (China in April 2020; South Korea in May 2020), leading to reductions in anthropogenic emissions related to outdoor activities. In Northeast Asia, high PM2.5 and haze events frequently occur in winter and early spring, especially in China and South Korea where emissions are concentrated. From December–March in the three years prior to COVID-19 (non-COVID-19 period), surface PM2.5 averaged 62.2 µg/m3 in China and 31.6 µg/m3 in South Korea; during December 2019–March 2020 (COVID-19 period), means were 50.0 µg/m3 and 24.5 µg/m3, reductions of 19.6% and 22.6%. Trends from 2016–2020 showed decreasing PM2.5 in China and a slight increase in South Korea before COVID-19, with sharp decreases during COVID-19. However, attributing changes solely to COVID-19-related activity reductions is difficult because PM2.5 arises from complex interactions between emissions and meteorology. Governments have implemented emission reduction policies since the early 2010s, yielding steady declines, and climate variability has altered synoptic conditions affecting pollutant levels. Most studies to date have documented reductions via observations and satellites or assumed emission changes in models. This study aims to quantitatively separate and attribute observed PM2.5 reductions during the COVID-19 period to meteorological variability, ongoing emission control policies, and COVID-19-related reductions in anthropogenic activities by combining four winters of observations with meteorology–air quality modeling.

Prior work documents frequent severe winter/early-spring PM2.5 and haze events in East Asia, with meteorology (e.g., wind speed, boundary layer, humidity) strongly modulating pollutant levels and climate change increasing the frequency of conditions conducive to haze (e.g., weakened East Asian winter monsoon). Regional emission control actions, such as China’s Air Pollution Prevention and Control Action Plan (2013–2017), have driven downward trends in primary pollutants and their precursors since 2010. Studies have updated and constrained emissions using surface and satellite data and examined ozone and SO2 responses. During the early COVID-19 outbreak, satellite NOx observations and modeling analyses reported significant changes in pollution levels, but attribution often relied on assumed emission reductions or observational contrasts without fully separating meteorology and policy effects. This study fills that gap by explicitly decomposing observed PM2.5 changes into meteorology-driven, policy-driven, and COVID-19-driven components.

Temporal design: The analysis period was split into a first half (December 2019–January 2020) and a second half (February–March 2020), corresponding to pre- and peak social distancing. For comparison, three non-COVID-19 winters were used: December 2016–March 2017, December 2017–March 2018, and December 2018–March 2019. Observations: Surface PM2.5 observations were obtained from the China National Environmental Monitoring Center (CNEMC) and Air Korea. Meteorological observations were taken from MADIS stations in China and South Korea. Modeling system: The Community Multiscale Air Quality model (CMAQ v4.7.1) simulated PM2.5 over Northeast Asia on a 174×128 grid at 27-km resolution. Six subregions were analyzed: China (and its North, Central, South subregions), South Korea, and the Seoul Metropolitan Area (SMA). Meteorology was generated by WRF v3.9.1 using NCEP GFS 1.0° data (6-hourly) for initial and boundary conditions. Anthropogenic emissions used the CREATE 2015 inventory processed via SMOKE v3.1; biogenic emissions were from MEGAN v2.1. Land cover and vegetation fields were provided to MEGAN. Attribution framework: CMAQ simulations were performed with fixed anthropogenic emissions (no interannual changes) for all winters (2016–2020) to isolate meteorological effects. The simulated change between COVID-19 and non-COVID-19 periods (ΔC_sim) thus represents the meteorology-driven PM2.5 change (ΔC_met). The observed PM2.5 change (ΔC_obs) between the same periods was computed from period means. The emissions-driven contribution (ΔC_emis) was calculated as ΔC_obs − ΔC_met, encompassing both policy-driven emission trends (ΔC_policy) and COVID-19-related activity reductions (ΔC_COVID). To separate ΔC_policy, the non-COVID-19-period linear trend in observed PM2.5 (Slope_obs) was compared to the simulated trend (Slope_sim). Slope_obs was corrected by the Obs/Model concentration ratio (C_obs/C_mod) to account for bias, and interpreted as the meteorology-driven slope; the difference between Slope_obs and Slope_sim over non-COVID-19 winters yielded ΔC_policy. Finally, for February–March 2020 (second half), ΔC_COVID was computed as ΔC_emis − ΔC_policy. Model performance was evaluated: temporal change trends in simulated and observed PM2.5 were similar; correlation coefficients exceeded 0.90 for 2-m air temperature and 10-m wind speed, and R>0.89 for PM2.5 during non-COVID-19. Region-specific metrics (e.g., Northern China second half RMSE 26.3 µg/m3, R 0.78) were noted.

- Observed seasonal means: Non-COVID-19 (Dec–Mar 2016–2019) PM2.5 averaged 62.2 µg/m3 (China) and 31.6 µg/m3 (South Korea). During COVID-19 (Dec 2019–Mar 2020), means were 50.0 µg/m3 and 24.5 µg/m3, reductions of 19.6% and 22.6%. - Monthly trend slopes: Pre-COVID-19 (Dec–Mar) China: −2.35 µg/m3/month; South Korea: +0.39 µg/m3/month. During COVID-19: China: −9.82 µg/m3/month; South Korea: −1.57 µg/m3/month. Early winter (Dec–Jan): pre-COVID China −4.59; South Korea +1.17; during early COVID China +3.46; South Korea −0.52. Late winter (Feb–Mar): during COVID China −7.60; South Korea −5.20; the regression slope decreased by 5.0 µg/m3/month (China) and 6.6 µg/m3/month (South Korea) relative to non-COVID late winters. - Observation–simulation anomaly differences: Non-COVID average difference: China 0.0%, South Korea 3.4%. COVID period average difference: China −17.2%, South Korea 32.3%; February differences were especially large (China 31.8%, South Korea 57.1%), consistent with unmodeled emission reductions during COVID-19. - Meteorology-driven impacts: First half (Dec–Jan) increased PM2.5 by +2% (China) and +5% (South Korea). Second half (Feb–Mar) decreased PM2.5 by −7% (China) and −5% (South Korea). In Northern China, PM2.5 increased by ~8.0 µg/m3 in the first half, with meteorology contributing +28% (17.1 µg/m3); wind speed averaged 2.3 m/s and relative humidity 69.6%, with wind 10.6% lower and RH 6.7% higher than the non-COVID period. - Emission control policy-driven impacts: China: −15% (−10.5 µg/m3) in the first half and −8% (−4.2 µg/m3) in the second half. South Korea: −9% (−2.9 µg/m3) in the first half and −4% (−1.3 µg/m3) in the second half. Compared to the first half, the policy-driven reduction in the second half decreased by 40% (China) and 44% (South Korea). In Northern China second half, policy effect appeared as +2% due to larger simulated than observed changes and higher model error. - COVID-19-driven impacts (Feb–Mar 2020): China overall −16% (−9.0 µg/m3), with −29% (North), −16% (Central), −12% (South). South Korea overall −21% (−7.0 µg/m3), including −22% in the Seoul Metropolitan Area. These COVID-related reductions exceeded meteorology and policy contributions during the peak period. - Overall, reductions in anthropogenic emissions during COVID-19 had greater influence on PM2.5 decreases than meteorological variability or ongoing policy-driven emission trends, with notable temporal and regional variability.

By combining observations with fixed-emissions meteorology–chemistry modeling, the study isolates the distinct roles of meteorology, policy-driven emission changes, and COVID-19-related activity reductions in shaping PM2.5 during winter 2019–2020. The pronounced declines in February–March 2020 are primarily attributable to COVID-19-driven reductions, exceeding the effects of meteorological conditions and ongoing policy measures. Meteorology contributed to higher PM2.5 early in the winter (reduced wind speeds and higher humidity limited dispersion) and to decreases later, underscoring the strong temporal modulation. Spatial heterogeneity (e.g., stronger COVID signals in Northern China and SMA) reflects differing emission sources and meteorological regimes. The divergence between observed and simulated anomalies during COVID-19, despite good baseline model performance, is consistent with the model’s fixed anthropogenic emissions not capturing real-time reductions. The attribution framework thus provides a quantitative basis to evaluate regulation efficacy and indicates that substantial additional emission reductions are necessary to achieve agreed air quality goals across Northeast Asia.

The study presents a quantitative decomposition of wintertime PM2.5 changes in Northeast Asia into meteorology-, policy-, and COVID-19-driven components using four winters of observations and CMAQ-WRF modeling with fixed anthropogenic emissions. During February–March 2020, COVID-19-related reductions dominated PM2.5 decreases (China −16%; South Korea −21%), surpassing meteorological (−7% China; −5% South Korea) and policy impacts (−8% China; −4% South Korea). The results emphasize temporal and regional variability and demonstrate an effective method to assess regulatory impacts. Achieving mutually acceptable air quality levels will require significant further reductions in PM2.5 and its precursors and enhanced cross-border collaboration within the shared airshed. Future work should relax the independence assumption between meteorology and emissions, incorporate dynamic emission inventories, and extend attribution across seasons and longer time spans.

- The attribution assumes independence between meteorological and emission effects, which may not fully hold in reality. - CMAQ simulations used fixed anthropogenic emissions across years; ΔC_sim captures only meteorological variability and does not include interannual policy-driven emission changes or COVID-19 reductions, necessitating indirect inference for emissions effects. - Model uncertainties arise from input data and parameterizations; despite good overall performance, regional/period-specific errors exist (e.g., Northern China second half RMSE 26.3 µg/m3, R 0.78), affecting local attribution (e.g., apparent +2% policy effect). - The COVID-19 impact was assessed primarily for February–March 2020; results may not generalize to other seasons or later pandemic phases. - Observational representativeness and network coverage can influence regional averages and trends.