Drivers for the poor air quality conditions in North China Plain during the COVID-19 outbreak

Abstract

China's lockdown to control COVID-19 brought significant declines in air pollutant emissions, but haze was still a serious problem in North China Plain (NCP) during late-January to mid-February of 2020. We seek the potential causes for the poor air quality in NCP combining satellite data, ground measurements and model analyses. Efforts to constrain COVID-19 result in a drop-off of primary gaseous pollutants, e.g., −42.4% for surface nitrogen dioxide (NO₂) and −38.9% for tropospheric NO₂ column, but fine particulate matter (PM₂₅) still remains high and ozone (O₃) even increases sharply (+84.1%). Stagnant weather during COVID-19 outbreak, e.g., persistent low wind speed, frequent temperature inversion and wind convergence, is one of the major drivers for the poor air quality in NCP. The surface PM_2.5 levels vary between −12.9~+15.1% in NCP driven by the varying climate conditions between the years 2000 and 2020. Besides, the persistent PM_2.5 pollution might be maintained by the still intensive industrial and residential emissions (primary PM_2.5), and increased atmospheric oxidants (+26.1% for ozone and +29.4% for hydroxyl radical) in response to the NO₂ decline (secondary PM_2.5). Further understanding the nonlinear response between atmospheric secondary aerosols and NO_x emissions is meaningful to cope with the emerging air pollution problems in China.

Graphical abstract

1. Introduction

In late December of 2019, a novel coronavirus pneumonia named “COVID-19” emerged in Wuhan, a capital of 11 million people in Hubei province, China. Within the next month, the novel coronavirus spread quickly throughout the country, helped by the Chinese New Year migration, with Wuhan being a transport hub and major rail interchange in China. As of 21 January, the official data reported 440 confirmed cases and 9 deaths of COVID-19 in mainland China (http://www.nhc.gov.cn/).

The Chinese government launched an aggressive campaign to contain the spread of COVID-19. On 22 January, Hubei Province initiatively announced the public health emergency response Level II, and a quarantine was issued to stop the travel in and out of Wuhan. As of 25 January, all the provinces in China had consecutively announced the public health emergency response Level I. The most serious epidemic control measures were implemented to contain the spread of COVID-19, including travel restrictions, factory closures, extended holidays, contact precaution, and closing of schools, stadiums, theatres and shopping malls. This pandemic was effectively controlled in early March, with the number of infected cases appearing each day has since plummeted in China.

China's efforts to control COVID-19 seem to have curbed energy consumption and air pollutant emissions. Ongoing efforts to contain COVID-19 suppressed China's industrial output by 15–30% across key industrial sectors (NBS, 2020). The traffic flow in major cities of China dropped by 60–90% during COVID-19 epidemic (http://jiaotong.baidu.com/top/). It was estimated that the carbon emission in China declined by 9.8% during COVID-19 compared to the same period of 2019, of which carbon emission from transportation sector declined mostly by 43.4% (Yue et al., 2020). Using tropospheric nitrogen dioxide (NO₂) column data, it was calculated that nitrogen oxides (NO_x) emission decreased by 50% in East China during 23 January to 9 February compared to that during 1–22 January 2020, and did not yet recover to their usual levels until late February in most provinces (Zhang et al., 2020).

The national lockdown brings significant air quality improvement (Chen et al., 2020; Sicard et al., 2020; Sun et al., 2020; Yue et al., 2020). The surface NO₂ concentrations and tropospheric NO₂ columns declined by 20–30% in China's main urban clusters during February to March of 2020 than the same period of 2019, and surface fine particulate matter (PM_2.5) and carbon monoxide (CO) declined by 24.9% and 17%, respectively, in China (Yue et al., 2020). Sicard et al. (2020) also found a substantial reduction in surface NO_x (56%) and PM_2.5 (42% in Wuhan and 8% in Europe) due to the global COVID-19 lockdown, but ozone (O₃) increased in all cities (36% in Wuhan and 17% in Europe). Despite that, haze was still a serious air pollution problem during COVID-19 outbreak in North China Plain (NCP). The entire NCP region was shrouded in thick smog in late-January to mid-February of 2020, with the hourly concentration of surface PM_2.5 remaining beyond 200 μg m⁻³ in the major cities of NCP (http://www.cnemc.cn/en/). The stifling haze was unexpected and aroused a hot debate in the public—why China's coronavirus control efforts did not eliminate the poor air quality in NCP?

The persistent particulate pollution in China is typically associated with emissions (Geng et al., 2017; Wang et al., 2013), enhanced secondary production (Fu et al., 2020; Huang et al., 2020) and unfavorable meteorology (Ji et al., 2014; Jiang et al., 2015; Li et al., 2019; Yang et al., 2015; Zhang et al., 2015). Regional emission controls are effective in reducing the emissions and hence improving the air quality in several specific events, e.g., Beijing Olympic Games (Wang et al., 2010) and Asia-Pacific Economic Cooperation Conference (Wang et al., 2016). But some studies pointed out that the oxidation production of secondary aerosols was sensitive to the atmospheric oxidants concentrations, implying complex impacts of NO_x and volatile organic compounds (VOC_s) emission controls on secondary aerosol formation (Leibensperger et al., 2011; Pun and Seigneur, 2001; Tsimpidi et al., 2008). They determined that the NO_x emission reduction might cause an increase of aerosol in the VOC_s-limited zone due to the enhanced atmospheric oxidation capacity. Recent evidence from field observations (Fu et al., 2020) also supported that the persistent heavy aerosol pollution in China over recent years was driven by the increased photochemical oxidants. Besides, the other causes for rapid haze increase include the stable boundary layer, weak ventilation and high humidity in winter that favored the accumulation and secondary formation of aerosol species (Wang et al., 2015; Yang et al., 2015; Zhang et al., 2015).

Because of the complex effects of emissions and meteorology, large unknowns still remain for the drivers of poor air quality in NCP during the COVID-19 outbreak. To address this issue, we try to seek the potential causes for the poor air quality in NCP and its link with the coronavirus control actions using satellite data, ground measurements and model analyses. Methods and data are presented in Sec. 2. Sec. 3 discusses the air quality trends in NCP and their potential drivers. Finally, a summary is presented in Sec. 4.

2. Methods and data

2.1. Observational data

Routine and comprehensive measurements of air pollutants are used to analyze the air quality trends in China during COVID-19 outbreak and to evaluate the model performance. The air pollutants data include: 1) Hourly mass concentrations of surface PM_2.5, sulfur dioxide (SO₂), NO₂, CO and O₃ spanning from 1 January to 22 February of 2020 at the air quality monitoring network in China (Fig. 1 ; http://www.cnemc.cn/en/); 2) Comprehensive measurements at Dianshanhu supersite (31.1°N, 121.0°E) in Shanghai, which provided hourly mass concentrations of PM_2.5, water-soluble ions (e.g., sulfate, nitrate, ammonium, chloride, sodium), carbonaceous aerosols (organic matter (OM) and elemental carbon (EC)) and gaseous pollutants (Shu et al., 2019); 3) The global tropospheric NO₂ column data provided by Ozone Monitoring Instrument (OMI) onboard the Aura satellite, with a spatial resolution of 0.125 ° × 0.125 ° (Gaudel et al., 2018). The error reported in OMI NO₂ data is 2.2 × 10¹⁵ molecules/cm² (Marchenko et al., 2015). The OMI NO₂ data are available at National Aeronautics and Space Administration (NASA) Goddard Space Flight Center (https://acd-ext.gsfc.nasa.gov).

Fig. 1.

Two-nested WRF-Chem modelling domain (a) and monitoring stations (b). The epidemic control time is marked with shaded color in panel (b). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Surface and sounding meteorological data in NCP are used to analyze the weather conditions during COVID-19 outbreak and to evaluate the model performance. The surface and sounding meteorological data (Fig. 1) in NCP include: 1) The land-based hourly observations of temperature, humidity and wind at 5 automatic stations in Beijing (40.1°N, 116.6°E), Tianjin (39.1°N, 117.2°E), Shijiazhuang (38.1°N, 114.4°E), Baoding (38.7°N, 115.5°E) and Tangshan (39.6°N, 118.1°E). These data are archived at the U. S. National Climatic Data Center (NCDC) (Smith et al., 2011); 2) The radiosonde observations in Beijing (39.9°N, 116.3°E) that is managed by World Meteorological Organization (WMO) (Durre et al., 2006). It provides air temperature and wind speed two times per day (08:00 and 20:00 Local Time) on the WMO mandatory pressure levels.

2.2. WRF-chem model setup

The Weather Research and Forecasting-Chemistry (WRF-Chem) model (version 4.1) developed by Grell et al. (2005) is used to predict the atmospheric dynamic and chemical processes. A two-nested modelling domain is setup, with the inner domain covering northern and mid-eastern China (Fig. 1). The model horizontal resolution is 36 and 12 km, and has 105 × 100 and 136 × 163 grids.

The WRF-Chem model is run with an 84-h model cycle, with the first 12 h discarded to allow for the influence of initial conditions (Berge et al., 2001). The model is regularly initialized from National Centers for Environmental Prediction final analysis fields (NCEP/FNL, available at http://rda.ucar.edu/datasets/ds083.2/), which has a temporal resolution of 6 h and a spatial resolution of 1 ° × 1 ° and 52 vertical layers. The NCEP/FNL data have been proven to provide realistic initial and boundary meteorological input, and lead to best estimates of wind regimes (Carvalho et al., 2014; Ghafarian et al., 2019). Model outputs of each model cycle are used to provide chemical initial conditions for the subsequent overlapping 84-h simulation. The complete simulation consists of thirteen 84-h model cycles that start from 12:00 UTC on 9 January to 00:00 UTC on 18 February in 2020, with the first 6 days as spin-up for chemistry and the remaining model outputs for analysis.

The gas-phase chemistry is simulated using the Carbon Bond Mechanism version Z (CBMZ) (Zaveri and Peters, 1999). The aerosol chemistry is based on the Model for Simulating Aerosol Interactions and Chemistry (MOSAIC) (Zaveri et al., 2008). The MOSAIC approach explicitly treats nine aerosol species, including sulfate, nitrate and ammonium (denoted as SNA), EC, OM, sodium chloride, calcium carbonate, water and unspecified inorganic aerosols. The aqueous-phase chemistry in MOSAIC is based on the Carnegie Mellon University (CMU) scheme including 50 species and more than 100 reactions (Fahey and Pandis, 2001). The aqueous SO₂ oxidation catalyzed by mineral ions and heterogeneous uptakes of SO₂, NO₂, nitrogen trioxide (NO₃), nitrogen pentoxide (N₂O₅) and nitric acid (HNO₃) on mineral aerosols were added in MOSAIC to better represent the secondary aerosol formation (Li et al., 2019). The physical options contain RRTMG radiation scheme (Iacono et al., 2008), Noah land surface scheme (Ek et al., 2003), Lin microphysics scheme (Lin et al., 1983) and YSU boundary layer scheme (Noh et al., 2003).

2.3. Numerical experimental designs and emission reduction scenarios

Ten WRF-Chem simulations are designed in Table 1 . In the BASE simulation, the anthropogenic emissions in China remain unchanged as the usual levels (E _BASE), without regard to the COVID-19 controls. The CTL20 simulation is the same as BASE, but designed with modulated emissions (E _CTL) considering the impacts of epidemic control measures on power, industry, transportation and residential sectors. The differences between CTL20 and BASE simulations are calculated to show the impacts of emission controls during the COVID-19 outbreak on particulate pollution. Four additional control simulations (CTL00–CTL15) are conducted with the same emissions (E _CTL) as CTL20, but driven by varied climatic data every five years in 2000–2015. The CTL00–CTL20 simulations are designed to calculate the impacts of climate conditions on particulate pollution in NCP. Besides, another four sensitivity simulations (CTL20-EM25, CTL20-EM50, CTL20-NO25 and CTL20-NO50) are designed to illustrate the responses of primary and secondary PM_2.5 components to different emission reduction scenarios. In the CTL20-EM25 and CTL20-EM50 simulations, the emission reductions for all trace gases and particulate matter due to the COVID-19 lockdown are increased by 25% and 50%, respectively. In the CTL20-NO25 and CTL20-NO50 simulations, the emissions reductions of NO_x due to the COVID-19 lockdown are increased by 25% and 50%, respectively.

Table 1.

WRF-Chem numerical experimental designs.

Experiment	Meteorological input	Anthropogenic emissions
BASE	FNL in 2020	Usual levels in 2020 (EBASE) without regard to COVID-19 controls
CTL20	FNL in 2020	In CTL00–CTL20, emissions from power plant, industry, transportation and residential sectors are modulated considering the epidemic control measures (ECTL) based on activity data
CTL15	FNL in 2015
CTL10	FNL in 2010
CTL05	FNL in 2005
CTL00	FNL in 2000
CTL20-EM25	FNL in 2020	Same as CTL20, but the total emission reductions due to COVID-19 lockdown are further increased by 25%, i.e., ECTL–25% × (EBASE–ECTL)
CTL20-EM50	FNL in 2020	Same as CTL20, but the total emission reductions due to COVID-19 lockdown are further increased by 50%, i.e., ECTL–50% × (EBASE–ECTL)
CTL20-NO25	FNL in 2020	Same as CTL20, but the emission reduction of NOx due to COVID-19 lockdown is further increased by 25%
CTL20-NO50	FNL in 2020	Same as CTL20, but the emission reduction of NOx due to COVID-19 lockdown is further increased by 50%

Anthropogenic emissions in China are adopted from Multi-resolution Emission Inventory for China (MEIC) that is developed by Tsinghua University (available at http://meicmodel.org), and updated to the usual levels of 2020 based on the annual statistical data of provincial emissions since 2016 (Chen et al., 2019; MEP, 2017; NBS, 2020). The timeline for the COVID-19 epidemic control in China is defined in Fig. 1. The epidemic control and consequent emission decline initially occurred near Wuhan and quickly spread to the whole country as more regions adopted quarantine measures (Fig. 1). The air pollutant emissions slowly rebounded to pre-pandemic levels in March of 2020 (Zhang et al., 2020). To estimate the emission reductions due to COVID-19, we scale China's emissions data from power, industry, transportation and residential sectors based on the social economic activity data between 2019 and 2020, in a similar way as some recent researches (Huang et al., 2020; Wang and Su, 2020; Yue et al., 2020). Traffic emissions are reduced using a province-level emission reduction ratio, according to the dynamic traffic flow monitored at major cities of China (Table 2 ; http://jiaotong.baidu.com/top/). The traffic flow dropped by 52–99% during the COVID-19 epidemic, and Hubei and Henan saw the most rigorous traffic restrictions. For power sector, the electricity demand remained far below its usual level, and the national thermal power generation in January–February of 2020 dropped by 8.9% than that in 2019 (NBS, 2020). Emission from power plants is then assumd to decline by 8.9%. For key industrial sectors, the outputs of cement, chemicals and rubber fell by 15–30%, while the output of steel remained nearly unchanged (NBS, 2020). The emission from industrial sector is thus assumed to decline by −15% during COVID-19 lockdown. For residential sector, emissions from commercial boilers and stoves were eliminated since the closing of entertainment venues, while emissions from residential heating increased by 10% because millions of people were on lockdown, according to the recent investigation by Liqiang (2020).

Table 2.

Monitored traffic flux changes after China's Spring Festival of 2020 in major regions (data available from http://jiaotong.baidu.com).

Region	Change	Region	Change
Hebei	−70.2%	Zhejiang	−58.5%
Beijing	−69.3%	Hubei	−98.5%
Henan	−93.4%	Hunan	−83.4%
Shanxi	−52.8%	Jiangxi	−75.0%
Shaanxi	−64.9%	Guangdong	−74.4%
Shandong	−71.7%	Guangxi	−68.3%
Jiangsu	−62.1%	Ningxia	−85.1%
Anhui	−78.9%	Gansu	−68.2%
Heilongjiang	−22.5%	Jilin	−51.5%
Liaoning	−27.7%	Inner Mongolia	−30.5%

3. Results and discussions

3.1. Air quality trends in NCP during COVID-19 outbreak from ground and satellite observations

We first look into the perturbations to atmospheric composition in response to COVID-19 (Fig. 2, Fig. 3, Fig. 4 ). The COVID-19 lockdown across the NCP region initiated on 24 January of 2020 (Fig. 1b), thereafter defined as a critical date for COVID-19 outbreak. The surface concentrations for main gaseous and particulate air pollutants had since then declined substantially in China, with a nationwide average of −18.5 μg m⁻³ for NO₂, −2.7 μg m⁻³ for SO₂, −0.3 mg m⁻³ for CO and −17.4 μg m⁻³ for PM_2.5 between 1 and 23 January and 24 January–22 February of 2020 (Fig. 2). The decrease of surface NO_x mixing ratios ranges from −50.8 to −16.4 μg m⁻³, particularly in southern NCP and mid-eastern China. Conversely the surface O₃ level increases nearly in all cities, with a nationwide average of 17.3 μg m⁻³, in a similar pattern as NO_x due to the nonlinear relationship to its precursor. Declining trends of surface PM_2.5 and CO concentrations are also visible at more than 75% sites beyond the cities adjacent to Beijing, with even faster reductions around Hebei, Henan and Shandong. The changes for surface SO₂ show large inhomogeneity and even opposite trends for adjacent areas, which could be possibly because of the still high industrial and power plant emissions. The results confirm the recently reported COVID-19 impacts on air quality across China (Sicard et al., 2020; Yue et al., 2020).

Fig. 2.

Spatial distributions for the changes of observed aerosol and gaseous pollutants concentrations in China before and after COVID-19 outbreak.

Fig. 3.

Observed aerosol composition and gaseous pollutants concentrations in five typical cities of NCP and Shanghai supersite in January–February of 2020.

Fig. 4.

Tropospheric NO₂ column (10¹⁵ molecules cm⁻²) from OMI in January–February of 2020 (a, b) and the same period in 2017–2019 (c, d).

Several cities that are most severely affected by this persistent haze in NCP are selected as targets, namely Beijing, Tianjin, Shijiazhuang, Baoding and Tangshan. Two-month monitoring data in NCP suggest that the efforts to control coronavirus did result in a drop of primary gaseous pollutants between 1 and 23 January and 24 January–22 February, e.g., −24.0 μg m⁻³ (−42.4%) for NO₂, −1.4 μg m⁻³ (−9.6%) for SO₂ and −0.3 mg m⁻³ (−16.5%) for CO averaged over the five typical cities (Fig. 3). Despite that, the surface PM₂₅ concentration still remains high (−6.7%), and surface O₃ even increases sharply (+84.1%) in NCP after the COVID-19 outbreak. The entire NCP region was shrouded in thick smog (154.7 μg m⁻³ for average PM_2.5 concentration) in late-January to mid-February of 2020, with two peaks around the Spring Festival (24–28 January) and Lantern Festival (8–13 February; Fig. 3). The high proportion of nitrate in PM_2.5, i.e., 22.2% for 24–28 January and 26.8% for 8–13 February, is in accordance with recent observations in northern China (Shao et al., 2018; Zhang et al., 2018) and synchronous measurements in Shanghai (19.5%; Fig. 3f).

Comparisons of satellite images from OMI in 2020 and the same period in 2017–2019 also show a dramatic effect of the “economic slowdown” during COVID-19 outbreak on tropospheric NO₂ in China, a gas primarily coming from traffic, industry and power plant (Fig. 4). OMI satellite data give a clear drop-off of tropospheric NO₂ column in NCP (−51.8%) from January to February in 2020 along with the full-scale promotion of epidemic control over all China. Typically, there is always lower NO₂ around this time of year, e.g., −39.2% for NCP from January to February in 2017–2019, as many household heating and industrial activities in China declines (Li et al., 2017). But it is noted that the NO₂ level in NCP declines by 38.9% in February 2020 compared to the same period in 2017–2019, attributed to the quarantine measures to stop the spread of coronavirus.

3.2. Drivers for the poor PM_2.5 air quality conditions in NCP

3.2.1. Model validation

The model results in CTL20 generally reproduced the spatial-temporal patterns of meteorology and air quality in NCP. Fig. S1 compares the modeled surface meteorology with real observations in five typical cities of NCP. The temporal patterns and magnitudes of temperature and relative humidity are reasonably well simulated, with a correlation coefficient of 0.79 and 0.64, and a mean bias (MB) of 0.52 °C and −16.5%, respectively. Underestimation of specific humidity is a common phenomenon in the WRF simulation, which might be attributed to the influence of land surface conditions (Li et al., 2017; Yan et al., 2020) and boundary layer parameterizations (Bhati and Mohan, 2018; Gomez-Navarro et al., 2015) on weather forecast. Slight overestimation of wind speed (0.18 m s⁻¹) in the simulation might be because of the unresolved topography in WRF (Jimenez et al., 2013; Li et al., 2014).

The predicted surface PM_2.5 mass concentrations in NCP also compare well with measurements (Fig. S2), with a correlation coefficient of 0.65, a total MB of −9.22 μg m⁻³ and a normalized mean bias (NMB) of −0.12. The model is particularly successful in predicting the locations of maximum PM_2.5 concentrations along the Beijing, Tianjin, southern Hebei and Henan city belt (Fig. 5 ). This region saw a sharp increase of daily-average PM_2.5 from tens of μg m⁻³ to more than 150 μg m⁻³ during COVID-19 outbreak, exceeding the 2nd limit of National Ambient Air Quality Standard (75 μg m⁻³) by more than twofold. But it tends to seriously underestimate the PM_2.5 observations in the vicinity of southern NCP on Lunar New Year's Eve (24 January), like Shanxi and Henan, which might be due to the missing of firework emissions in the emission inventory. Further evaluations of major gaseous pollutants (SO₂, NO₂, O₃ and CO) (Fig. S2) also indicate that the model is able to simulate the atmospheric pollution in NCP, with a MB of −16.3, 16.1 and 20.5 μg m⁻³ for O₃, NO₂ and SO₂, and −0.4 mg m⁻³ for CO, respectively. Comparisons between the CTL20 and BASE simulations confirm that the assumed emission reduction scenario results in better model performance for main gaseous pollutants in NCP. The model biases for NO₂, SO₂ and O₃ are reduced by 29.9%, 15.3% and 23.0% in CTL20 during the post-COVID-19 period compared to that in BASE (data not shown here).

Fig. 5.

Pattern comparisons of the simulated daily-average surface PM_2.5 concentrations (shaded areas) and observed data (dots).

3.2.2. Unfavorable weather conditions during COVID-19 outbreak

Local meteorology has deterministic impacts on pollution processes. Combining meteorological data and model results, we find that the weather conditions during COVID-19 outbreak are unconducive to the dispersion of air pollutants in NCP, e.g., persistently low wind speed (less than 1 m s⁻¹) lasting from late-January to early-February (Fig. S1), and sometimes strong surface temperature inversion (Fig. 6 ). From the temperature and wind speed profiles in Beijing (Fig. 6), the near-surface temperature inversion intensity reached up to 5–15 °C on these seriously polluted days (e.g., 26 January and 12 February), leading to a reduction of boundary layer height by 1/2–2/3 than the normal height on clean days. Meanwhile, the wind speeds on hazy days remained below 7 m s⁻¹ throughout the atmospheric boundary layer, nearly half that on clean days (Fig. 6b). The stagnant weather conditions allow the local accumulation of high PM_2.5 concentrations in NCP.

Fig. 6.

Observed vertical profiles of temperature and wind speed in Beijing on clean days (dashed lines) and polluted days (solid lines) during COVID-19 outbreak.

Analysis of the simulated surface PM_2.5 concentrations and wind fields (Fig. 7 ) also shows favorable circulation characteristics for persistent particulate pollution in NCP. In the “24–28 January” episode, the stagnant air and surface wind convergence in NCP allowed the rapid accumulation of air pollutants and secondary aerosol production. A severe regional haze layer formed stretching from the NCP region towards mid-eastern China through transport by the dominant north wind. In the “8–13 February” episode, affected by the southerly wind and topography effect of Yan Mountain, surface cyclonic circulations formed and trapped the polluted air masses near Beijing, Tianjin and southern Hebei, leading to a rapid increase of PM_2.5 in NCP. After then, the prolonged haze pollution was cleared by the sudden wind shift to dry, cold north wind and the fast increase of wind speed to nearly 10 m s⁻¹ (Fig. S1).

Fig. 7.

Simulated surface average PM_2.5 concentrations (shaded contour) and wind fields (vectors) for the two typical hazy processes during COVID-19 outbreak: (a) 24–28 January and (b) 8–13 February.

Model results from CTL00–CTL20 further illustrate the role of climate conditions in this persistent particulate pollution covering NCP. Fig. 8, Fig. 9 compare the simulated surface PM_2.5 concentrations, pressure fields and essential weather elements, i.e., temperature, humidity, planetary boundary layer height (PBLH) and wind in the CTL00–CTL20 experiments. Generally, climate anomaly determines the spatial patterns and concentration levels of surface PM_2.5 in China (Fig. 8). The PM_2.5 levels varied between −12.9~+15.1% in NCP driven by the varying climate conditions between 2000 and 2020. The simulated surface average PM_2.5 concentrations in NCP were 67.9, 57.4, 51.4, 58.1 and 59.9 μg m⁻³, respectively, in CTL00–CTL20. The particulate pollution in NCP was most serious for the years 2000 (67.9 μg m⁻³) and 2020 (59.9 μg m⁻³), with broader influencing sphere covering northern to mid-eastern China and higher PM_2.5 concentrations exceeding 80 μg m⁻³.

Fig. 8.

Simulated surface average PM_2.5 concentrations (shaded contour) and wind fields (vectors) during 24 January to 17 February in CTL00–CTL20.

Fig. 9.

(a) Simulated surface average PM_2.5 concentrations and weather elements in NCP in CTL00–CTL20. (b–c) Frequency distributions of surface wind speed and wind direction in CTL00–CTL20.

Analysis of the weather fields in CTL00–CTL20 (Fig. 9a) points out that the concentration level of PM_2.5 is negatively correlated to wind speed (R = −0.933) and boundary layer height (R = −0.927), and vice versa positively correlated to atmospheric humidity (R = 0.656). The calculated frequency distribution of wind in NCP for the CTL00–CTL20 climate scenarios shows that the dispersion conditions in CTL20 are favorable for the haze formation. The average wind speed in CTL20 remains low (3.9 m s⁻¹), and the frequency of light and calm wind (below 2 m s⁻¹) reaches up to 40.3%, compared to the mean levels of 27.7–41.2% in other climate scenarios (Fig. 9b). The NCP region is normally controlled by surface high pressure center in winter and the winds originate from the polluted industrial regions of Beijing-Tianjin-Hebei (Fig. 9c). Besides, in the climatologically dry winter season, the atmospheric humidity in CTL20 is larger than the usual levels, with a mean relative humidity of 52.2% near the surface and a maximum increase by 23.1%, facilitating multiphase reactions for aerosol formation and growth (Fig. 9a). Consistent with the increase in atmospheric humidity and the decrease in wind speed in CTL20, the atmospheric boundary layer height in NCP declined (403.7 m) during the COVID-19 lockdown, inducing a stable boundary layer and consequently producing high PM_2.5 levels. The climate conditions in CTL20 are featured by unusually stable atmosphere (404 m), stagnant air (3.9 m s⁻¹) and high humidity (52.2%; Fig. 9a), favoring the rapid accumulation and oxidation production of atmospheric aerosols.

3.2.3. Response of air quality to the coronavirus emission controls

The COVID-19 outbreak forces large changes in China's socio-economic activities, of which transportation sector seems to be most affected. The estimated emission reductions for main trace gases and primary PM_2.5 calculated from activity data are presented in Fig. S3. We observe a very strong decrease of NO_x, SO₂, CO and PM_2.5 emissions across northern and mid-eastern China attributed to the COVID-19 measures. The inferred emission reduction ratios range from −13.1% to −49.8% for NO_x, −0.4% to −15.6% for SO₂, −0.6% to −31.6% for CO, and −0.9% to −15.7% for primary PM_2.5 at the provincial level (Table 3 ). NO_x emitted from the combustion of coal fuels and the exhaust of urban vehicles is the major air pollutant affected by coronavirus controls, particularly in Hubei (−46.1%) and Henan (−49.8%) Provinces where the COVID-19 outbreak is first recorded or very strict lockdown regulations are adopted. In the NCP region, it is estimated that the unprecedented quarantine from 24 January to 17 February resulted in a dramatic reduction of air pollutants emissions, i.e., −6.9 × 10⁴ ton (−33.1%) for NO_x, −9.8 × 10³ ton (−9.3%) for SO₂, −2.6 × 10⁵ ton (−12.7%) for CO and −1.3 × 10³ ton (−7.4%) for PM_2.5. The estimated NO_x emission reduction in China using activity data (−13.1% to −49.8%) is comparable to the change in tropospheric NO₂ column from OMI during the lockdown period (−38.9% in NCP; Fig. 4) and some recent reports using a top-down emission inverse algorithm (−20% to −50% from early January to mid-February in most provinces) (Ding et al., 2020; Miyazaki et al., 2020).

Table 3.

Estimated emission reductions of air pollutants (%) in major regions of China from 24 January to 17 February in 2020 due to the COVID-19 control.

Region	NOx	SO2	CO	PM2.5	Region	NOx	SO2	CO	PM2.5
Beijing	−29.6	−11.8	−13.3	−5.0	Liaoning	−15.7	−10.9	−5.5	−4.4
Tianjin	−281	−13.1	−16.8	−6.2	Fujian	−36.5	−15.3	−21.7	−8.2
Hebei	−31.5	−9.0	−10.5	−5.7	Jiangxi	−42.5	−11.0	−14.2	−8.0
Shanxi	−21.3	−4.9	−6.5	−6.6	Guangdong	−41.0	−15.6	−31.6	−12.2
Inner Mongolia	−13.6	−4.5	−1.8	−3.0	Guangxi	−39.2	−14.4	−15.4	−8.1
Shandong	−33.3	−9.6	−13.4	−8.9	Hainan	−29.7	−14.6	−15.8	−4.3
Henan	−49.8	−11.0	−15.6	−9.8	Chongqing	−36.8	−11.8	−8.7	−3.6
Shanghai	−26.5	−12.9	−24.4	−15.7	Sichuan	−33.4	−11.8	−9.5	−1.9
Jiangsu	−29.1	−14.9	−15.8	−7.7	Guizhou	−28.8	−0.4	−0.6	−0.9
Zhejiang	−31.8	−14.6	−29.8	−12.4	Yunan	−40.9	−8.9	−15.4	−6.3
Anhui	−40.4	−11.6	−5.9	−3.0	Shaanxi	−28.1	−4.6	−8.8	−4.2
Hubei	−46.1	−5.7	−9.2	−4.7	Gansu	−30.0	−7.4	−5.5	−1.7
Hunan	−39.4	−5.2	−6.8	−3.0	Qinghai	−30.1	−8.7	−9.9	−5.2
Heilongjiang	−13.1	−2.9	−1.6	−1.4	Ningxia	−25.6	−10.4	−16.8	−8.3
Jilin	−20.2	−8.2	−3.8	−1.7	Xinjiang	−22.7	−11.1	−17.0	−10.1

Model results also find that the emission controls during COVID-19 clearly improve the air quality of major primary air pollutants in NCP (Fig. 10 ), such as the surface NO₂, SO₂ and CO concentrations by −31.3, −8.3 and −8.2%, respectively. Nevertheless, the surface O₃ mixing ratio in NCP increases sharply by 26.1% after the COVID-19 controls, resulting from the nonlinear relationship to its precursor NO_x (Seinfeld and Pandis, 2006). The response of particulate air quality to the coronavirus emission controls seems more complicated (Fig. 10), with surface PM_2.5 decreasing by 7.1% in NCP. The lockdown saw obviously lower emissions on road, but nearly unchanged emissions from emission-intensive industries (NBS, 2020) and household heating. The primary PM_2.5 components, which mainly come from industrial and residential sectors (e.g., 49.7% and 38.8% in MEIC), thus still remain high (−4.1%) during the COVID-19 outbreak (Fig. 11 ). The oxidation production of secondary aerosol components depends not only on the abundance of their precursors (e.g., NO_x and SO₂), but also the atmospheric oxidizing capacity (e.g., O₃ and hydroxyl radical (OH)). The emission controls of gaseous precursors NO_x (−33.1%) and SO₂ (−9.3%) seem to be less effective in reducing winter fine particles, which only lowers surface nitrate concentration by 11.7%, and even increases sulfate concentration (+2.4%) in NCP during COVID-19 outbreak.

Fig. 10.

Impacts of COVID-19 emission controls on surface air quality during 24 January to 17 February of 2020.

Fig. 11.

Impacts of six emission scenarios on the concentrations of PM_2.5 components, O₃, NO₂ and the main intermediate products of nitrate (N₂O₅ and HNO₃) during 24 January to 17 February of 2020.

The calculated nitrogen oxidation ratio (NOR) and sulfur oxidation ratio (SOR) on lockdown days are 10.6% and 16.9% higher in CTL20 than that in BASE. The increased NOR and SOR indicate that the decreased SNA formation in response to the COVID-19 emission controls is partly offset by the enhanced atmospheric oxidation capacity. To be specific, we focus on the response of nitrate to NO_x emission during COVID-19 outbreak. The formation of nitrate involves multiphase chemical reactions, including the NO₂+OH→HNO₃ gas-phase oxidation and the heterogeneous hydrolysis of nitrogen pentoxide (N₂O₅) (He et al., 2018; Pathak et al., 2011). NO₂ is key in atmospheric chemistry and serves as an important precursor for both ozone and secondary aerosol. The nonlinear response between nitrate and NO_x emission in NCP might be associated with the increased atmospheric oxidants in CTL20, e.g., O₃ (+26.1%) and OH (+29.4%), which subsequently facilitated the conversion of NO_x to nitrate. It is noted that despite the nitric acid (HNO₃) concentration in CTL20 drops by −25.6% in response to the COVID-19 NO_x emission control, the concentrations of N₂O₅ (+14.4%) and nitrogen trioxide (NO₃; +41.8%) increase markedly compared to that in BASE simulation (Fig. 11), which is produced from the $N{O}_2\stackrel{O_3}{\to }N{O}_3\stackrel{N{O}_2}{\to }{N}_2{O}_5$ chemical reaction and is a crucial intermediate product for nitrate.

The sensitivity simulations of CTL20-NO25 and CTL20-NO50 also highlight that the emission reduction of NO_x alone is less effective in altering the PM_2.5 concentration and components (Fig. 11). The concentrations of nitrate, sulfate, ammonium and PM_2.5 change by −6.2%, +15.4%, −3.8% and −1.8% after 25% cut of NO_x emission in CTL20-NO25, and by −24.9%, +34.6%, −18.3% and −8.5% after 50% cut of NO_x emission in CTL20-NO50 compared to CTL20. The insensitive response of secondary aerosols to NO_x emission may be explained by the increased atmospheric oxidizing capacity (e.g., +22.0% for O₃ and +36.4% for OH in CTL20-NO25) and the still high intermediate products for nitrate (e.g., +4.7% for N₂O₅ and −8.5% for HNO₃ in CTL20-NO25) compared to CTL20. Comparison of the emission reduction scenarios in CTL20-EM25 and CTL20-EM50 further confirms the complex chemical responses of primary and secondary PM_2.5 to emission control strategies (Fig. 11). More stringent emission reductions for trace gases and PM_2.5 by 25% and 50% on the basis of E _CTL result in concurrent declines of both primary and secondary PM_2.5 components (−19.0% in CTL20-EM25 and −39.4% in CTL20-EM50 for PM_2.5), yet with different responses for primary and secondary species. Secondary aerosols show much smaller changes (−14.1% in CTL20-EM25 and −28.9% in CTL20-EM50) with the emission reductions, compared with primary aerosols (−23.1% in CTL20-EM25 and −45.7% in CTL20-EM50). Recent evidence from intensive observation and numerical simulation also highlighted smaller response of secondary aerosols to the COVID-19 lockdown (Huang et al., 2020; Le et al., 2020; Sun et al., 2020).

4. Conclusion

China's efforts to contain COVID-19 suppressed the socio-economic activities and emissions, but the NCP region was still shrouded in thick smog in late-January to mid-February of 2020. We discuss the potential links between the poor air quality conditions in NCP and coronavirus control actions combining satellite data, ground measurements and model analyses.

The national lockdown brought dramatic declines in air pollutant emissions, e.g., −33.1%, −9.3%, −12.7% and −7.4% for NO_x, SO₂, CO and PM_2.5, respectively, from 24 January to 17 February in NCP. The COVID-19 effects on improved NO₂ air quality in NCP are confirmed by satellite data (−38.9% for tropospheric NO₂ column) and ground measurements (−42.4% for surface NO₂ concentration). Stagnant weather in NCP during the COVID-19 outbreak, e.g., persistent low wind speed, frequent temperature inversion and wind convergence, are the major drivers for the poor air quality. The surface PM_2.5 levels in NCP vary between −12.9~+15.1% driven by the varying climatic conditions from 2000 to 2020. Besides, the persistent PM_2.5 pollution might be maintained by the still intensive industrial and residential emissions (primary PM_2.5), and the enhanced atmospheric oxidation capacity (e.g., +26.1% for O₃ and +29.4% for OH) as a nonlinear response to NO_x declines (secondary PM_2.5). This work highlights that more scientific and stringent emission controls and climate adaptation strategies are required to attain the synergetic control of air pollution and to cope with the emerging air pollution problems in China.

Some uncertainties still exit in our assessment. The emission reductions due to COVID-19 are estimated based on activity data from key sectors. This method might obscure the detailed spatiotemporal emission information due to the difficulty in gathering fine-resolution statistical data. Instead, a top-down approach using an inverse algorithm and multiple satellite observations has been demonstrated to provide accurate and quick emission estimates during COVID-19 lockdown (Ding et al., 2020; Miyazaki et al., 2020). They found that the Chinese NO_x emission was reduced 20–50% for cities, 40% for power plants, and 15–40% for maritime transport.

Atmospheric ammonia (NH₃) that is emitted from fertilizer application, livestock, biomass burning, chemical industry, waste disposal and transportation acts as a critical neutralizing species for SNA production and efficient haze mitigation (Liu et al., 2019). China is known as the world's top emitter of NH₃, but the emission estimate for NH₃ has a large uncertainty of up to ±153% (Kurokawa et al., 2013; Liu et al., 2019). Compared to the process-based NH₃ estimates in China, e.g., PKU-NH₃ (Huang et al., 2012), the MEIC NH₃ inventory agrees well in temperate zones but is significantly higher in tropical zones, due to the discrepancies in derived emission factors (Li et al., 2017). Improved and updated emission estimates for NH₃ are needed to accurately simulate the secondary aerosol production and its emission responses.

Besides, secondary organic aerosol (SOA) formed from the oxidation of volatile organic compounds (Seinfeld and Pandis, 2006) is also an important component of PM_2.5 in China (Huang et al., 2014; Zhao et al., 2016). The concentration, composition and properties of SOA showed complicated responses to the emissions of NO_x and VOC_s (Feng et al., 2019; Tsimpidi et al., 2008; Xu et al., 2019). However, the simulation of SOA is usually seriously underestimated by more than tenfold in China because of the poor understanding of its precursors, chemical mechanisms and physicochemical characterization (Jiang et al., 2012; Li et al., 2011; Robinson et al., 2007). Further understanding and improvement of SOA simulation are needed to enhance the understanding of aerosol and photochemistry feedback in China.

CRediT authorship contribution statement

Mengmeng Li: Methodology, Writing - original draft, Software. Shu Li: Software, Supervision. Bingliang Zhuang: Resources, Data curation. Qingyan Fu: Data curation. Ming Zhao: Validation. Hao Wu: Validation. Jane Liu: Writing - review & editing. Eri Saikawa: Formal analysis. Kuo Liao: Visualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability statement

The WRF-Chem model version 4.1 is available at http://www2.mmm.ucar.edu/wrf/users/downloads.html. The NCEP FNL data are accessible at the National Center for Atmospheric Research (NCAR) Research Data Archive (http://rda.ucar.edu/datasets/ds083.2/). The surface meteorological data are downloaded from Integrated Surface Database (https://www.ncdc.noaa.gov/isd/data-access), and the sounding meteorological data are downloaded from Integrated Global Radiosonde Archive (https://www1.ncdc.noaa.gov/pub/data/igra/). The air pollutants and aerosol species data are provided by China National Environmental Monitoring Center and Shanghai Environmental Monitoring Center, which are archived at https://doi.org/10.6084/m9.figshare.12449663.v1. The OMI products are provided by the NASA Goddard Space Flight Center (https://acd-ext.gsfc.nasa.gov/Data_services). Model outputs from WRF-Chem could be accessible upon request via hard drives from the author (Dr. Mengmeng Li, Email address: mengmengli2015@nju.edu.cn).

Appendix A. Supplementary data

The following are the supplementary data to this article: