Radiative effects and feedbacks of anthropogenic aerosols on boundary layer meteorology and fine particulate matter during the COVID-19 lockdown over China

Abstract

The COVID-19 epidemic has exerted significant impacts on human health, social and economic activities, air quality and atmospheric chemistry, and potentially on climate change. In this study, an online coupled regional climate–chemistry–aerosol model (RIEMS-Chem) was applied to explore the direct, indirect, and feedback effects of anthropogenic aerosols on radiation, boundary layer meteorology, and fine particulate matter during the COVID-19 lockdown period from 23 January to 8 April 2020 over China. Model performance was validated against a variety of observations for meteorological variables, PM_2.5 and its chemical components, aerosol optical properties, as well as shortwave radiation flux, which demonstrated that RIEMS-Chem was able to reproduce the spatial distribution and temporal variation of the above variables reasonably well. During the study period, direct radiative effect (DRE) of anthropogenic aerosols was stronger than indirect radiative effect (IRE) in most regions north of the Yangtze River, whereas IRE dominated over DRE in the Yangtze River regions and South China. In North China, DRE induced larger changes in meteorology and PM_2.5 than those induced by IRE, whereas in South China, the changes by IRE were remarkably larger than those by DRE. Emission reduction alone during the COVID-19 lockdown reduced PM_2.5 concentration by approximately 32 % on average over East China. As a result, DRE at the surface was weakened by 15 %, whereas IRE changed little over East China, leading to a decrease in total radiative effect (TRE) by approximately 7 % in terms of domain average. The DRE-induced changes in meteorology and PM_2.5 were weakened due to emission reduction, whereas the IRE-induced changes were almost the same between the cases with and without emission reductions. By aerosol radiative and feedback effects, the COVID-19 emission reductions resulted in 0.06 °C and 0.04 °C surface warming, 1.6 and 4.0 μg m⁻³ PM_2.5 decrease, 0.4 and 1.3 mm precipitation increase during the lockdown period in 2020 in terms of domain average over North China and South China, respectively, whereas the lockdown caused negligible changes on average over East Asia.

Graphical abstract

The model simulated changes in (a) total radiative effect of anthropogenic aerosols at the surface (W m⁻²), and changes in the radiative feedback-induced (b) surface air temperature decrease (°C), (c) PBLH decrease (m), (d) surface PM_2.5 concentration increase (μg m⁻³) due to anthropogenic emission reductions during the COVID-19 lockdown from 23 January to 8 April 2020. The changes are derived from the BASE case minus EXP case. The numbers in the upper right corner of each panel denote averages over East China during the study period.

1. Introduction

Aerosols play a significant role in air quality, atmospheric chemistry, and climate change. Aerosols can affect radiation transfer directly by scattering/absorbing solar/infrared radiation, indirectly by acting as cloud condensation nuclei (CCN) and modifying cloud properties, and by heating atmosphere to modify atmosphere stability and cloud formation, which are regarded as the direct radiative effect, indirect effect, and semi-direct effect (Twomey, 1974; Albrecht, 1989; Ramanathan et al., 2001), respectively. China has been experiencing continuous economic growth in recent decades, along with large anthropogenic emissions and frequent occurrence of haze events. To tackle with air pollution problem, Chinese government launched strict emission control strategy since 2013, and evidently improved air quality, reducing PM_2.5 concentration by 25 %, 20 %, and 15 % in 2017 compared to the level in 2013 in the Beijing-Tianjin-Hebei, the Yangtze River Delta, and the Pearl River Delta, respectively. (Zhang et al., 2019).

In January 2020, COVID-19 epidemic outbroke and caused serious human health hazard in China. In order to mitigate the spread of COVID-19, Chinese government implemented lockdown measures from Wuhan city to the whole nation. During the lockdown period, economic, industrial, and social activities were restricted, leading to a significant reduction in anthropogenic emissions, which could significantly affect air quality, radiation, as well as meteorology. Zheng et al. (2021) estimated that anthropogenic emissions of SO₂, NO_x, CO, non-methane volatile organic compounds (NMVOC), and primary PM_2.5 decreased by 27 %, 36 %, 28 %, 31 %, and 24 %, respectively, in February 2020 compared to the same month in 2019 in China by applying a bottom-up approach. Zhang et al. (2021) reported that the nation-wide NO_x emission decreased by 53.4 % during the COVID-19 lockdown period, compared to 2019, by using a top-down approach. To limit the rapid spread of the COVID-19 globally, many countries also imposed lockdown measures, remarkably reducing human activities.

A lot of studies were carried out, focusing on the effects of the COVID-19 lockdown on air quality and atmospheric chemistry. Adam et al. (2021) reviewed observations in some countries during the COVID-19 lockdown in 2020, and found apparent reductions in the levels of airborne air pollutants, including gaseous species and particulate matter. Kanniah et al. (2020) used Himawari-8, Aura-OMI satellites, and ground-based observations to quantify changes in air pollutants associated with the shutdown due to COVID-19 in Southeast Asia, and found that tropospheric NO₂ decreased by ~27–34 % over urban agglomerations. Ren et al. (2022) investigated with an earth system model the effects of emissions reductions during the COVID-19 pandemic on fire weather in 2020 over the western United States and found that reductions in aerosols dominated the increases in wildfire risks. By using satellite data and a network of >10,000 air quality stations, Venter et al. (2020) found that lockdown events have reduced the population-weighted concentration of nitrogen dioxide and particulate matter levels by about 60 % and 31 % in 34 countries, with mixed effects on ozone, after accounting for meteorological variations. Silver et al. (2020) reported that PM_2.5 and PM₁₀ decreased by 10.5 % and 21.4 % during the lockdown period, respectively, while NO₂ decreased most (27.0 %) across China during the lockdown period, with the largest NO₂ reduction (51 %) in Hubei province based on China's air quality monitoring network. Xing et al. (2020) developed a response model based on air quality observations and emission-concentration response functions derived from chemical transport modeling and estimated that the anthropogenic emissions of NO₂, SO₂, VOC and primary PM_2.5 on the North China Plain were reduced by 51 %, 28 %, 67 % and 63 %, respectively, due to the COVID-19 shutdown. By applying a machine learning technique, Wang et al. (2020) estimated the COVID-19 lockdown reduced ambient NO₂ by 36–53 % in six megacities of China during the most restrictive periods. Shi and Brasseur (2020) found the mean levels of PM_2.5 and NO₂ in northern China have decreased by approximately 35 % and 60 %, respectively, between the period 1 and 22 January 2020 and the period 23 January and 29 February 2020 by using surface measurements at >800 monitoring stations operated by the China National Environmental Monitoring Center.

Huang et al. (2021) indicated that the large decrease in NO_x increased ozone and nighttime NO₃ radical formation, and in turn facilitated the formation of secondary particulate matter during the COVID-19 lockdown in China, based on analysis of ground measurements and WRF-Chem simulation. Le et al. (2020) indicated that because of nonlinear chemistry and titration of ozone in winter, reduced nitrogen oxides due to the lockdown resulted in ozone enhancement in urban areas, further increasing atmospheric oxidizing capacity and facilitating secondary aerosol formation in China from 23 January to 13 February 2020 through observation analyses and model simulations. Sun et al. (2020) analyzed aerosol composition measurements in Beijing and found the formation of secondary inorganic aerosols was significantly enhanced during the COVID-19 lockdown period. Li et al. (2021) used a random forest model to identify the relative contributions of meteorology and anthropogenic emissions to air quality changes during the COVID-19 lockdown, and indicated that the primary components in PM_2.5 significantly decreased, whereas the production of secondary aerosols was remarkably enhanced due to higher relative humidity and NH₃ level and lower temperature.

Only a few studies explored the effect of COVID-19 lockdown on radiation and meteorology over China (Yang et al., 2020; Wang et al., 2021; Zhu et al., 2021; Yang et al., 2022; Yu et al., 2022). Yang et al. (2020) used a global model CAM5 to simulate fast climate responses to aerosol emission reductions during the COVID-19 pandemic over the world and model results exhibited an anomalous warming of 0.05–0.15 K in eastern China in January–March, and surface temperature increases by 0.04–0.07 K in Europe, eastern United States, and South Asia in March–May 2020 relative to 2019. Zhu et al. (2021) investigated with WRF-Chem the effect of aerosol-radiation feedback on PM_2.5 and O₃ during January–February 2020 over the North China Plain, and found emission control during the COVID-19 epidemic enhanced PM_2.5 decrease and O₃ increase through the feedback effect, but direct and indirect effects of aerosols were not distinguished in their study. Wang et al. (2021) simulated with WRF-Chem that emission reduction during COVID-19 lockdown weakened aerosol-PBL interaction and thus led to a reduction of 25 μg m⁻³ (∼50 %) in PM_2.5 enhancement, without consideration of indirect aerosol effect.

As seen above, majority of previous studies concerned about the effects of COVID-19 on anthropogenic emission, air quality, and atmospheric chemical activity, whereas the impacts of COVID-19 lockdown on radiation, boundary layer meteorology, cloud and precipitation, as well as aerosol feedback effect were very limited. To our knowledge, this is the first attempt to explore synthetically the direct, indirect and feedback effects of anthropogenic aerosols during the COVID-19 lockdown period in winter-spring 2020 over China. By applying an online coupled regional climate–chemistry–aerosol model (namely RIEMS-Chem), firstly, this study estimates the direct and indirect radiative effects of anthropogenic aerosols during the lockdown period in 2020, then explores the aerosol radiative effects on planetary boundary layer meteorology, cloud and precipitation, as well as the radiative feedback on PM_2.5 and its components, and finally reveals the changes in aerosol radiative effects and feedbacks in response to emission reductions during the lockdown in 2020 relative to 2019. The results from this study would provide new insights into the environmental impact of COVID-19 epidemic on both regional and global scales.

2. Model description and configuration

2.1. RIEMS-Chem

RIEMS-Chem is an online coupled regional climate-chemistry-aerosol model developed based on the Regional Integrated Environmental Modeling System (RIEMS) (Fu et al., 2005). Land surface process is represented by a modified Biosphere-Atmosphere Transfer Scheme (BATS; Dickinson et al., 1993). Planetary boundary layer process is treated by the Medium-Range Forecasts scheme (MRF) (Hong and Pan, 1996). Cumulus convective process is parameterized by the Grell (1993) scheme, and a modified NCAR Community Climate Model, version CCM3 (Kiehl et al., 1996) is used to represent radiation transfer processes including aerosol attenuation effect. RIEMS has participated in the Regional Climate Model Intercomparison Project (RMIP) for Asia, which shows a good ability among models in predicting air temperature and precipitation over East Asia (Fu et al., 2005; Wang et al., 2015).

In recent years, a series of processes of chemical species including emission, transport, diffusion, multi-phase chemistry, dry and wet deposition, etc., have been incorporated into RIEMS and interact with existing dynamics and physical processes, forming an online coupled regional climate-chemistry-aerosol model, namely RIEMS-Chem (Han, 2010; Li and Han, 2016). The schemes for advection and diffusion of chemical species are identical to the scheme for moisture. An updated Carbon-bond mechanism (CB-IV, Gery et al., 1989) is used to represent gas chemistry. Photolysis rate is calculated by using the Tropospheric Ultraviolet-Visible (TUV) radiation model (Lee-Taylor and Madronich, 2007), which takes aerosol attenuation effect into account. Thermodynamic processes are represented by ISORROPIA II (Fountoukis and Nenes, 2007). Dry deposition velocity of aerosols is calculated as the inverse of the sum of resistances plus a gravitational settling term. Below-cloud wet scavenging of aerosols is parameterized as a function of precipitation rate and collision efficiency of particle by hydrometeor (Han et al., 2004). Heterogeneous reactions of gas species on mineral dust and sea salt surfaces are parameterized by the schemes from Li and Han (2010) and Li et al. (2018a). The conversion of SO₂ to sulfate on pre-existing hydrated aerosols is represented by the scheme of Li et al., 2018, Li et al., 2018a. A two-product model (Odum et al., 1997) is used to treat SOA formation processes. RIEMS-Chem calculates mass extinction coefficient, single scattering albedo and asymmetry factor of aerosols with a Mie-theory based parameterization developed by Ghan and Zaveri (2007) and basic physical parameters from the OPAC database (Optical Properties of Aerosols and Clouds) (Hess et al., 1998). Kappa (κ) parameterization (Petters and Kreidenweis, 2007) is applied to calculate aerosol hygroscopic growth. The bulk κ for internal mixture of aerosols is derived by volume-weighted average of κ of each aerosol component. Based on previous works, κ values for inorganic aerosols, BC, POA, SOA, dust and sea salt are chosen as 0.65, 0, 0.1, 0.2, 0.01 and 0.98, respectively. Nine types of aerosols are considered in RIEMS-Chem, including sulfate, nitrate, ammonium, black carbon (BC), primary organic aerosol (POA), secondary organic aerosol (SOA), anthropogenic primary PMs (PM_2.5 and PM₁₀), dust and sea salt. Size distribution of anthropogenic aerosols is prescribed mainly based on OPAC database. Dust deflation model is from Han et al. (2004), with dust size distribution represented by 5 size bins (0.1–1.0, 1.0–2.0, 2.0–4.0, 4.0–8.0, 8.0–20.0 μm). Sea salt generation is parameterized by the scheme of Gong et al. (1997), using the same size bins as dust particle. Anthropogenic aerosols are assumed to be internally-mixed while they are externally-mixed with dust and sea salt. A physically based scheme (namely A-G scheme) based on classical Köhler theory developed by Abdul-Razzak et al. (1998) is incorporated into RIEMS-Chem to represent activation of aerosols to form cloud droplets. Cloud droplet number concentration is estimated by several variables including aerosol number concentration, aerosol size distribution and composition, updraft velocity and ambient supersaturation etc. The maximum ambient supersaturation is calculated by solving supersaturation balance equation (Abdul-Razzak et al., 1998). The updraft velocity is represented by the sum of grid mean updraft velocity and sub-grid updraft velocity, which is diagnosed from vertical eddy diffusivity according to Ghan et al. (1997). Autoconversion rate from cloud water to rainwater perturbed by aerosols (second indirect effect) is parameterized by the scheme of Beheng (1994), in which the rate is expressed as a function of cloud droplet number concentration and cloud liquid water content. The effect of aerosols on ice nuclei and convective clouds is not treated yet in this model due to limitation in current knowledge. The calculated aerosol optical parameters and cloud droplet number concentration are transferred into radiation and cloud modules to perturb radiation transfer and cloud microphysics and thus meteorological variables. Chemical/aerosol modules are called every 2.5 min and the calculated aerosol induced changes in shortwave/longwave radiation feed back into dynamic/physical modules in RIEMS-Chem. Prediction for hourly CCN concentration using the A-G scheme coupled with RIEMS-Chem has been validated by cruise measurements from the marginal seas of China to the western Pacific Ocean south of Japan, which demonstrates a good ability, yielding a correlation coefficient of 0.87 with a normalized mean bias within 20 % (Han et al., 2019). RIEMS-Chem has participated in the Model Inter Comparison Study for Asia (MICS-Asia) phase III, which has demonstrated a good ability for aerosol concentrations and optical properties over East Asia (Gao et al., 2018). RIEMS-Chem has been widely applied in previous studies on distribution, evolution, radiative and feedback effects of anthropogenic and natural aerosols over East Asia (Han et al., 2012, Han et al., 2013, Han et al., 2019; Li et al., 2014; Li and Han, 2016; Li et al., 2019; Gao et al., 2020; Li et al., 2020, Li et al., 2022). More details on this model refer to Han (2010), Han et al. (2012) and Li et al. (2020).

2.2. Model configuration and numerical experiments

RIEMS-Chem is configured on a lambert projection with a sigma terrain-following coordinate. It has a horizontal resolution of 60 km and 16 layers in the vertical, with seven layers in the planetary boundary layer. The study domain covers China, the Korean Peninsula, Japan and northern parts of Southeast Asia (Fig. 1 ). The center of the model domain is at 110°E and 35°N, with 89 grid points in west-east direction and 75 grid points in south-north direction. The study period is from 1 January to 8 April 2020, covering the entire time period of the COVID-19 outbreak and lockdown. The first 10 days are taken as model spin-up and model results from 23 January to 8 April, which is the COVID-19 lockdown period, are analyzed and interpreted. Boundary conditions of chemical species are derived from MOZART-4 (Model for Ozone and Related chemical Tracers, http://www.acd.ucar.edu/wrf-chem/mozart.html) simulations and updated every 6 h. Fig. 1 shows the study domain, locations of megacities for model validation in this study.

Fig. 1.

The study domain, and locations of megacities over China. Full name of each city refers to Table 1. North China, South China, and East China are denoted by the yellow box, green box, and blue box, respectively, with detailed descriptions in Table 4.

To identify the radiative and feedback effects of anthropogenic aerosols and to explore the changes in these effects due to emission reduction during the COVID-19 lockdown, a series of numerical experiments are carried out. First, two cases are designed, one is the base case (BASE), which uses real-time meteorological fields and anthropogenic emissions in 2020, the second one is a sensitivity case (EXP), which keeps meteorological fields in 2020 but uses anthropogenic emissions in 2019 instead. The difference between the BASE and EXP cases can reflect the impact of emission reductions due to the COVID-19 lockdown on aerosol levels, optical properties, radiative and feedback effects, as well as meteorological fields. Besides, to distinguish the direct, indirect, and total radiative effects (DRE, IRE, and TRE) of anthropogenic aerosols and their impacts on meteorology and feedbacks on PM_2.5, four additional sensitivity simulations are conducted for both BASE and EXP.

The direct radiative effect of anthropogenic aerosols (DRE) is calculated by comparing two radiation calls with and without anthropogenic aerosols (in one simulation) with aerosol perturbation to cloud microphysics turned off. The indirect radiative effect of aerosols (IRE) is estimated in a similar way as in DRE by comparing two radiation calls with and without aerosols with aerosol direct radiative effect turned off. The total radiative effect of anthropogenic aerosols (TRE) is estimated by comparing two radiation calls with and without aerosols while both aerosol direct radiative effect and perturbation to cloud microphysics are active. The impact of individual aerosol radiative and feedback effects (by DRE, IRE, and TRE) on meteorological variables and PM_2.5 concentration is estimated by comparing each of the above three sensitivity simulations and a reference case (CASE0) without considering aerosol radiative feedback effects (on dynamics, cloud and chemistry). In the above sensitivity simulations, natural emissions (i.e., mineral dust, sea salt, biogenic VOC, biomass burning emissions) and boundary conditions are identical.

2.3. Emissions and observations

In this study, monthly mean anthropogenic emissions are derived from the Multi-resolution Emission Inventory model for China (MEIC, http://meicmodel.org/, last accessed: 23 May 2022) developed by Tsinghua University, which includes NO_x, SO₂, NMVOC, NH₃, CO, PM_2.5, PM_coarse, OC, and BC from 5 sectors. The inventory has a horizontal resolution of 0.5° and is interpolated to Lambert projection in RIEMS-Chem by bilinear interpolation. Note anthropogenic emissions for 2020 (during the COVID-19 period) used in this study are derived from the modification of anthropogenic emissions in 2017 (open version in the MEIC website) according to a recent work by Zheng et al. (2021) (a key member in the MEIC group, personal communication). Anthropogenic emissions outside China are derived from the MIX inventory (Li et al., 2017) for the base year 2010. Biomass burning emissions are derived from the Global Fire Emission Database version4 (GFEDv4), and biogenic VOC emissions are derived from the Global Emissions Inventory Activity (GEIA, http://www.geiacenter.org/).

Surface meteorological observations are derived from the National Meteorological Information Center (http://data.cma.cn/, last accessed: 24 May 2022), including air temperature at 2 m, relative humidity at 2 m, and wind speed at 10 m. Daily meteorological observations in 22 megacities of China are used for model validation. Hourly observations for gas species and particulate matter in 24 megacities of China are from the China National Environmental Monitoring Centre (http://www.cnemc.cn/, last accessed: 24 May 2022), and observation for each city is the average of all sites in that city. Note meteorological observations are not available in Shijiazhuang and Xi'an during the study period. Hourly PM_2.5 and its chemical components (sulfate, nitrate, ammonium, BC, and OC) are measured by using an Aerodyne time-of-flight aerosol chemical speciation monitor (ToF-ACSM) on a meteorological tower of the Institute of Atmospheric Physics, which is located between the third and fourth ring road in urban Beijing (39°58′N, 116°22′E). Ground observations for aerosol optical depth (AOD) are obtained from the three monitoring sites of AErosol RObotic NETwork (AERONET) (https://aeronet.gsfc.nasa.gov/, last accessed: 24 May 2022). Beijing-CAMS and Beijing-RADI sites are in urban area, whereas Xianghe site is located in Hebei province, about 60 km southeast of Beijing downtown. AOD retrievals from MODIS at 550 nm at 13:30 LST are obtained from the website (https://ladsweb.modaps.eosdis.nasa.gov/, last accessed: 24 May 2022). Monthly mean shortwave radiation flux at the top of the atmosphere is derived from the Clouds and Earth's Radiant Energy System (CERES) (https://ceres.larc.nasa.gov/, last access: 24 May 2022). Monthly observations of precipitation with a spatial resolution of 0.1° are provided by the Level 3 final monthly precipitation product (GPM_3IMERG) produced at the NASA Goddard Earth Sciences (GES) Data and Information Services Center (DISC) (https://earthdata.nasa.gov/eosdis/daacs/gesdisc, last access: 24 May 2022).

3. Model validation

3.1. Meteorological variables

The model performance statistics for air temperature at 2 m (T2), wind speed at 10 m (WS10), and relative humidity at 2 m (RH2) in the 22 megacities of China during the study period are presented in Table S1. The model reproduces meteorological variables generally well, with the correlation coefficients (R) of 0.97, 0.83, and 0.76, and the normalized mean biases (NMB) of 0.8 %, 2.3 %, 36.8 %, for T2, RH2, and WS10, respectively, for all sites. In some cities, including Beijing, Tianjin, Hefei, Guangzhou, and Nanning, the model performed quite well, although wind speed is somewhat overestimated. R and NMB in Beijing are 0.97 and 6.5 % for T2, 0.78 and 3.4 % for RH2, and 0.84 and 25.4 % for WS10, respectively. Fig. S1 presents the observed and model simulated daily mean near-surface air temperature, relative humidity, and wind speed in Beijing from 1 January to 8 April 2020. It shows that the model reasonably reproduces the day-to-day variations of these meteorological variables in Beijing.

3.2. Mass concentrations of PM_2.5 and aerosol components

Table 1 shows model performance statistics for hourly PM_2.5 concentration in the 24 megacities in China during the study period. The overall predictions are generally satisfactory, with the R of 0.53 and NMB of −8.1 % for all sites. The model reproduces PM_2.5 quite well in the cities over the North China Plain (e.g., Beijing, Tianjin, Shijiazhuang, Jinan, Zhengzhou), with R of 0.64–0.79 and NMB within ±17 %. The performance in Shanghai is also good, with R of 0.66 and NMB of −4.0 %, although the model tends to overpredict PM_2.5 by 30–50 % in other cities over the Yangtze River Delta (i.e., Nanjing, Hangzhou, Hefei). The model simulates PM_2.5 fairly well in the cities over the middle reaches of the Yangtze River (i.e., Wuhan, Changsha, Nanchang), with NMBs <20 %, and R of 0.3–0.5. Model underprediction is found in the cities of northeast China, with NMBs around −40 %, and R of 0.5–0.6. The above model biases could be mainly associated with uncertainties in anthropogenic emissions, because the degree of lockdown measure could be different from province to province in China with respect to different epidemic situation. Besides, uncertainties in meteorological prediction and chemical mechanism for secondary aerosol formation may also affect model predictions.

Table 1.

RIEMS-Chem performance statistics for hourly PM_2.5 concentrations from BASE in the 24 major cities of China during the study period from 1 January to 8 April 2020 (unit: μg m⁻³).

Cities	Long name	Location	PM2.5
			Obs	Sim	MB	NMB	R
BJ	Beijing	116.4°E, 40.0°N	51.22	53.71	2.48	4.9 %	0.64
TJ	Tianjin	117.3°E, 39.1°N	67.29	64.11	−3.18	−4.7 %	0.79
SJZ	Shijiazhuang	114.5°E, 38.0°N	89.31	73.74	−15.57	−17.4 %	0.74
ZZ	Zhengzhou	113.7°E, 34.8°N	75.00	64.83	−10.17	−13.6 %	0.74
TY	Taiyuan	112.5°E, 37.9°N	78.76	35.05	−43.71	−55.5 %	0.45
JN	Jinan	117.0°E, 36.6°N	65.63	60.41	−5.22	−8.0 %	0.70
SH	Shanghai	121.5°E,31.2°N	38.82	37.26	−1.55	−4.0 %	0.66
NJ	Nanjing	118.8°E, 32.0°N	43.90	59.16	15.26	34.8 %	0.60
HF	Hefei	117.2°E, 31.8°N	45.31	66.35	21.04	46.4 %	0.59
HZ	Hangzhou	120.1°E, 30.2°N	34.17	52.65	18.48	54.1 %	0.50
WH	Wuhan	114.2°E, 30.6°N	43.18	51.65	8.46	19.6 %	0.47
CS	Changsha	113.0°E, 28.2°N	50.87	48.87	−2.00	−3.9 %	0.30
NC	Nanchang	115.9°E, 28.7°N	40.39	47.83	7.44	18.4 %	0.36
FZ	Fuzhou	119.3°E, 26.1°N	25.06	27.30	2.24	9.0 %	0.29
GZ	Guangzhou	113.4°E, 23.2°N	24.32	40.95	16.62	68.3 %	0.39
NN	Nanning	108.3°E, 22.8°N	30.63	29.36	−1.27	−4.2 %	0.16
CD	Chengdu	104.0°E, 30.7°N	55.72	80.06	24.34	43.7 %	0.42
KM	Kunming	102.7°E, 25.0°N	31.87	14.13	−17.73	−55.7 %	0.36
CQ	Chongqing	106.5°E, 29.6°N	47.65	41.45	−6.20	−13.0 %	0.23
GY	Guiyang	106.7°E, 26.6°N	31.84	36.55	4.71	14.8 %	−0.06
XA	Xi'an	109.0°E, 34.3°N	75.27	45.75	−29.52	−39.2 %	0.56
HB	Harbin	126.6°E, 45.7°N	73.41	42.16	−31.25	−42.6 %	0.59
CC	Changchun	125.4°E, 43.8°N	66.46	39.19	−27.27	−41.0 %	0.49
SY	Shenyang	123.4°E, 41.8°N	64.87	35.22	−29.65	−45.7 %	0.59
ALL			52.03	47.82	−4.20	−8.1 %	0.53

Fig. 2 presents the observed and model simulated hourly concentrations of PM_2.5 and its chemical components in Beijing from 1 January to 8 April 2020. The model generally reproduces the temporal variations of aerosol components, but it tends to underpredict the peak values during the haze episode from 8 to 13 February, which could be due to uncertainties in chemical mechanism and in meteorological predictions. Table 2 presents the performance statistics for hourly PM_2.5 chemical components in Beijing during the study period. The model tends to underpredict sulfate and ammonium by approximately 25 %, with a smaller bias for nitrate. So far, sulfate concentration during haze period is often underpredicted in eastern China due to limited knowledge in sulfate formation mechanism (Gao et al., 2018). The low bias in OC (−14 %) could be due to underprediction of secondary organic aerosol, which is calculated by a two-product model. BC concentration is overpredicted by 20.8 %, which could be mainly associated with potential overestimation of BC emission, considering meteorological variables are predicted generally well in Beijing. Model performance for total PM_2.5 concentration is generally good in Beijing, with R of 0.64, and NMB of 4.9 %, respectively (Table 1).

Fig. 2.

The observed and model simulated (from the BASE case) hourly concentrations of PM_2.5 and its chemical components in Beijing from 1 January to 8 April 2020 (unit: μg m⁻³).

Table 2.

RIEMS-Chem performance statistics for hourly PM_2.5 chemical components from BASE in Beijing during the study period from 1 January to 8 April 2020 (unit: μg m⁻³).

Component	R	Sim	Obs	MB	NMB
BC	0.61	2.37	1.96	0.41	20.8 %
OC	0.50	13.58	15.82	−2.25	−14.2 %
SO42−	0.36	5.35	7.11	−1.76	−24.7 %
NO3−	0.58	13.00	13.65	−0.65	−4.8 %
NH4+	0.54	5.40	7.33	−1.93	−26.3 %

3.3. Gaseous component of the atmosphere

Fig. S2 presents the observed and model simulated hourly concentrations of O₃, NO₂, SO₂, and CO in Beijing during the study period. The model generally captures the temporal variation of gaseous species. The sharp peaks of SO₂ cannot be reproduced, probably due to relatively coarse model grid of the regional model and potential uncertainty in SO₂ emission. The model underpredicts O₃ levels (Fig. S2a) during 24–31 January 2020, which could be associated with the overprediction of NO_x (Fig. S2b) and thus more titration of O₃ during the Chinese Spring Festival. Table S2 shows the performance statistics for gas concentrations for all 24 megacities during the study period. In general, the model reproduces gas species concentrations fairly well, with Rs of 0.38, 0.46, 0.33, 0.44, and NMBs of −18.9 %, −16.1 %, −43.1 %, −34.7 % for O₃, NO₂, SO₂, and CO, respectively, for all cities.

3.4. Aerosol optical depth, shortwave radiation, and precipitation

Fig. 3 shows the model simulated and MODIS retrieved AOD at 550 nm at 13:30 LST averaged over the study period. The MODIS retrieval (Fig. 3a) exhibits higher AOD over wide areas from the North China Plain to the Yangtze River Delta and portions of the middle reaches of the Yangtze River. Model results generally agree with MODIS retrievals both in distribution and magnitude. The model-simulated AOD reaches 0.9 in Chongqing, where MODIS data is mostly absent mainly due to cloud effect there. The model underpredicts AOD in portions of south China adjacent to Vietnam, probably due to underestimation of biomass burning emission over Southeast Asia. The model simulated and observed daily AOD (at 550 nm) at the three sites of AERONET from 1 January to 8 April 2020 are shown in Fig. 4 . The model generally reproduces the day-to-day variations of AOD at the three sites, including the evident increase of AOD around 15 February.

Fig. 3.

AOD at 550 nm at 13:30 LST from (a) MODIS and (b) model simulation (from BASE), averaged over the period from 1 January to 8 April 2020.

Fig. 4.

The model simulated (from BASE) and observed daily AOD (at 550 nm) at the three sites of AERONET from 1 January to 8 April 2020.

Table 3 shows the model performance statistics for daily mean AOD at the three sites of AERONET around Beijing during the study period. The correlation coefficients between simulation and observation at the Beijing-CAMS and Beijing-RADI sites are both 0.66, with NMBs of 14–15 %. The statistics at the Xianghe site is better, with R of 0.75 and NMB of −8.8 %. Overall, the correlation coefficient is 0.69 and the NMB is −12.5 % for all sites, indicating the model is able to reasonably reproduce AOD in Beijing and surrounding areas.

Table 3.

Model performance statistics for daily AOD from BASE at the three sites of AERONET in China during the study period from 1 January to 8 April 2020.

AOD	Long Name	Location	Obs	Sim	MB	NMB	R
CAMS	Beijing-CAMS	116.3°E, 39.9°N	0.51	0.44	−0.08	−15.0 %	0.66
RADI	Beijing-RADI	116.4°E, 40.0°N	0.50	0.43	−0.07	−13.9 %	0.66
XH	Xianghe	117.0°E, 39.8°N	0.52	0.48	−0.05	−8.8 %	0.75
ALL			0.51	0.45	−0.06	−12.5 %	0.69

Fig. 5 shows the model simulated and CERES retrieved average net shortwave radiation fluxes at TOA during the study period. The model simulates higher shortwave radiation fluxes up to 180 W m⁻² over eastern Tibet, portions of south China and the western Pacific, generally consistent with CERES retrievals both in distribution and magnitude, although there are some overpredictions over the western Pacific.

Fig. 5.

(a) The CERES retrieved, (b) model simulated (from BASE) shortwave radiation fluxes at the top of the atmosphere (TOA) at 13:30 LST averaged over the study period from 1 January to 8 April 2020 (unit: W m⁻²).

Fig. 6 shows the model simulated and GPM retrieved total precipitation during the study period. It is impressive that the model well reproduces the spatial distribution of precipitation, with higher values mainly in southern China and portions of the western Pacific south of Japan, although the model tends to underpredict precipitation in southern China.

Fig. 6.

(a) The GPM retrieved, (b) model simulated (from BASE) accumulated precipitation from 1 January to 8 April 2020 (unit: mm).

In summary, RIEMS-Chem can generally reproduce the spatial distribution and temporal variation of meteorological variables (air temperature, wind speed, relative humidity, and precipitation), PM_2.5 and its chemical components, aerosol optical properties, as well as shortwave radiation and precipitation. The generally good model performance above lends confidence to the reliability of subsequent model results on aerosol radiative and feedback effects.

4. Model results

4.1. Aerosol Radiative Effects (ARE)

Fig. 7 a-c present the model simulated period-mean DRE, IRE, and TRE of anthropogenic aerosols during the study period from the BASE case (with real meteorology and anthropogenic emissions for year 2020). It clearly shows that DRE was higher in the Chongqing district and in the wide areas from the middle reaches of the Yangtze River (where Wuhan locates) to the North China Plain, with the maximum value up to −20 W m⁻² (Fig. 7a). DRE was larger in magnitude than IRE in most areas of North China, whereas IRE was remarkably stronger than DRE in South China. The maximum IRE reaching −50 W m⁻² occurred in the vicinity of Chongqing and portions of the western Pacific south of Japan. The distribution of TRE (sum of DRE and IRE) likely resembled that of IRE, with a stronger magnitude.

Fig. 7.

The model simulated surface (a) all-sky direct radiative effect (DRE), (b) indirect radiative effect (IRE), (c) total radiative effect (TRE) from the BASE case; (d) all-sky DRE, (e) IRE, (f) TRE from the EXP case; (g) all-sky DRE, (h) IRE, and (i) TRE from the difference between the BASE and EXP cases (BASE minus EXP). The numbers in the upper right corner of each panel denote averages over the entire domain during the study period.

Fig. 7 d-f shows the model simulated period-mean DRE, IRE, and TRE of anthropogenic aerosols during the study period from the EXP case (with meteorology for year 2020 but anthropogenic emissions for year 2019). It was striking that in EXP, DRE was evidently enhanced in Chongqing and from the middle reaches of the Yangtze River to the North China Plain, with the maximum value up to −35 W m⁻². This was mainly attributed to higher aerosol concentrations in the EXP case with 2019 emissions. Fig. S3 shows the average changes in near-surface aerosol concentrations due to the lockdown-induced emission reduction during the study period derived from the difference between BASE and EXP cases (BASE-EXP) without considering aerosol radiative feedback. It shows that emission reduction alone caused consistent decreases in PM_2.5 and its chemical components over East China, with the maximum decrease of PM_2.5 reaching 70 μg m⁻³ in Chongqing and the middle reaches of the Yangtze River (Fig. S3a). The domain-average PM_2.5 decrease was estimated to be 32 % over East China. However, it was noteworthy that IREs were almost the same between BASE and EXP cases. This was mainly attributed to the less sensitivity of cloud microphysical properties (cloud droplet radius, cloud optical depth) to increasing CCN at high CCN concentrations than that at lower CCN concentrations (Twomey, 1977; Liu et al., 2020). Due to increased magnitude of DRE, TRE was consequently enhanced, with the maximum increase in the vicinity of Chongqing in the EXP case.

Fig. 7 g-i shows the difference in the respective radiative effects between BASE and EXP cases (BASE minus EXP), which reflects the changes in radiative effects due to the COVID-19-induced emission reduction in 2020 relative to 2019 emissions. It was striking that a positive DRE anomaly up to 15 W m⁻² (43 % lower than in the EXP case) occurred in Chongqing and portions of the middle reaches of the Yangtze River (Fig. 7g). However, the change in IRE between cases was little as explained above, with a slightly weaker IRE in the BASE case in terms of domain average. Similar to the change in DRE, emission reduction in 2020 induced a positive TRE anomaly over entire East China, with the maximum value reaching as high as 15 W m⁻² in the same areas of maximum DRE change, which indicated an increase of solar radiation reaching the ground surface due to emission reductions.

Table 4 shows the model simulated period-domain averages of AOD, all-sky DRE, IRE, and TRE at the surface and at TOA in the typical regions during the study period from the BASE and EXP cases. The typical regions are classified as North China (NC), South China (SC), East China (EC), and East Asia (EA), described in detail in Table 4. It is noted that AOD was higher in North China (0.2) than that in South China (0.17), and AOD in East China (0.14) was higher than that in East Asia (0.09) in terms of domain average in the BASE case. It is clearly seen that the average AODs in the EXP case were higher than those in the BASE case for all regions, mainly due to higher emission amounts and aerosol levels.

Table 4.

The model simulated period-domain average AOD, all-sky DRE, IRE, TRE at the surface and TOA over specific domains during the study period from BASE and EXP (unit: W m⁻²).

Region	AOD	DRE		IRE		TRE
		Surface	TOA	Surface	TOA	Surface	TOA
BASE (with emission and meteorology in 2020):
NCa	0.20	−10.9	−6.0	−4.1	−4.1	−14.3	−9.4
SCb	0.17	−10.2	−3.6	−20.6	−20.5	−28.7	−21.6
ECc	0.14	−8.4	−3.6	−11.4	−11.4	−18.6	−13.6
EAd	0.09	−5.1	−2.4	−7.3	−7.4	−11.8	−9.0
EXP (with meteorology in 2020 but emission in 2019):
NCa	0.26	−13.6	−7.4	−4.1	−4.1	−16.7	−10.5
SCb	0.20	−11.8	−4.3	−20.7	−20.6	−30.0	−22.1
ECc	0.17	−9.9	−4.4	−11.5	−11.4	−19.9	−14.1
EAd	0.11	−6.0	−2.8	−7.4	−7.4	−12.5	−9.3

^aNC: North China, mainland China of 100.0°E ~ 123.0°E, 32.0°N ~ 42.0°N.

^bSC: South China, mainland China of 100.0°E ~ 123.0°E, 18.0°N ~ 32.0°N.

^cEC: East China, mainland China of 100.0°E ~ 135.0°E, 18.0°N ~ 55.0°N.

^dEA: East Asia, the entire study domain.

In both cases, DREs in North China were stronger than those in South China at the surface and at TOA. Intriguingly, DREs were 1.5–2.5 times the IREs in North China, whereas IREs were 2–5 times the DREs in South China. IREs in South China (approximately −21 W m⁻² both at the surface and TOA) were significantly stronger (about 5 times) than those in North China (−4.1 W m⁻²), mainly due to both higher cloud amounts and aerosol concentrations there. As a result, TRE in South China was larger than that in North China both at the surface and at TOA. Over East China and East Asia, DRE and IRE at the surface were comparable, whereas at TOA, IRE was approximately 3 times the DRE, dominating TRE.

Considering emission reductions due to the lockdown, DREs in the BASE case (in 2020) were consistently weaker than those in the EXP case (with emissions for 2019) both at the surface and at TOA for all the specific regions, and the percent decrease in surface DRE (20 %) from EXP to BASE cases in North China was larger than that (14 %) in South China. There was little difference in IRE between BASE and EXP cases for all the regions, although IRE was slightly stronger in the EXP case than that in the BASE case. Surface TRE in North China decreased from −16.7 W m⁻² in the EXP case to −14.3 W m⁻² in the BASE case (decreased by 14 %), while in South China, it decreased from −30.0 W m⁻² to −28.7 W m⁻² (decreased by 4 %). Over East China and East Asia, compared with the EXP case, surface DRE and TRE decreased by approximately 15 % and 7 % in the BASE case, respectively, indicating a considerable reduction in aerosol radiative effects due to emission control during the COVID-19 lockdown. The changes in TRE at TOA were somewhat smaller in magnitude than those at the surface between the two cases.

4.2. Radiative effects and feedbacks of anthropogenic aerosols on boundary layer meteorology, PM_2.5, and precipitation

4.2.1. Regional distributions

Fig. 8 presents the model simulated average changes in T2, WS10, PBLH, PM_2.5, and accumulated precipitation (PREP) induced by DRE, IRE, and TRE during the study period from the BASE case. It is found that DRE led to a decrease in T2 by up to 0.8 °C in Chongqing and the North China Plain. Correspondingly, WS10 and PBLH decreased by ~0.2 m s⁻¹ and ~ 60 m around those areas of temperature decrease. As a result, vertical diffusivity coefficient (K_z) tended to decrease by ~20 % over wide areas of East China (Fig. S4b). Consequently, PM_2.5 increased by up to 6 μg m⁻³ and 5 μg m⁻³ in the North China Plain and in Chongqing (Fig. 8d), respectively. It is noted that all the PM_2.5 components including primary aerosols like BC and secondary aerosols (inorganic aerosols and SOA) increased as well (Fig. S5). The increase in PM_2.5 components may result from combined changes in physical and chemical processes. The weakened diffusivity may enhance accumulation of both primary gas precursors and primary aerosols, while the increase in secondary aerosols could be also contributed by enhancement of chemical production processes. Fig. S6 shows the scatter plots of the ratio of total secondary inorganic aerosols (SNA: sulfate, nitrate, and ammonium) versus black carbon (BC) (SNA/BC) in the DRE and TRE cases versus the case without aerosol radiative feedback (CASE0). It is seen that the SNA/BC ratio in the presence of DRE is higher than that in CASE0 (Fig. S6 a-c), indicating enhanced secondary aerosol formation due to the direct aerosol radiative feedback, which is mainly attributed to the increased gas precursors and strengthened heterogeneous production due to increasing RH (Huang et al., 2020). It is noticed that the SNA/BC ratio with TRE is even higher than that in CASE0 especially in South China, suggesting a larger increase in RH and thus enhanced heterogeneous reactions while including IRE.

Fig. 8.

The model simulated average changes in (a, f, k) T2, (b, g, l) WS10, (c, h, m) PBLH, (d, i, n) PM_2.5, and (e, j, o) accumulated precipitation (PREP) induced by (a-e) DRE, (f-j) IRE, and (k-o) TRE during the study period from the BASE case (units, T2: °C, WS10: m s⁻¹, PBLH: m, PM_2.5: μg m⁻³, PREP: mm). The numbers in the upper right corner of each panel denote averages over the entire domain during the study period.

It is noteworthy that DRE caused an evident reduction in accumulated precipitation along the Yangtze River and in South China, with the maximum decrease up to 60 mm during the study period (Fig. 8e). Fig. S4c shows a general reduction in latent heat flux due to the DRE-induced cooling (Fig. 8a), and along with the reduced K_z, the upward transport of water vapor from the surface was weakened and consequently led to precipitation reduction.

It is remarkable that IRE caused consistent decreases in T2 by ~1.7 °C, PBLH by ~150 m, and WS10 by ~0.5 m s⁻¹ in South China, stronger than the changes by DRE (Fig. 8f-h). An evident divergence occurred in portions of South China due to the maximum cooling there. The divergence led to an anomalous northerly wind and drove water vapor to the southern coastal areas; meanwhile, wind speed over the sea increased and thus enhanced latent heat flux there (Fig. S4f). In response to the meteorological changes, PM_2.5 increased by up to 10 μg m⁻³ in portions of South China, which was mainly due to the weakened turbulent diffusivity (by ~40 %) (Fig. S4e) and enhanced chemical production discussed above. Compared with the precipitation change induced by DRE (Fig. 8e), IRE caused a more remarkable precipitation decrease in South China, with the maximum decrease up to 140 mm in the areas of maximum cooling and maximum PBLH decrease (Fig. 8j, f, h).

The decrease in precipitation by IRE was attributed to the combined effect of first and second indirect effects of aerosols. As shown in Fig. S7, the increasing CCN from CASE0 caused a decrease in cloud droplet effective radius (Fig. S7a) by up to 2 μm in portions of South China and the East China Sea, leading to an increase in cloud optical depth (cloud albedo) in the similar regions (Fig. S7b), and thus a negative surface radiative effect (Fig. 7b). The reduction in surface shortwave radiation resulted in a decrease in surface temperature, a divergence in the areas of maximum cooling (Fig. 8f, g), and an increase in water vapor in the coastal areas (Fig. S4f) as discussed above, which generally resulted in decreases in cloud liquid water path (due to weaker convection of water vapor) and precipitation in the vicinity of major cooling areas, and increases in cloud liquid water path and precipitation in the vicinity of coastal areas (Fig. S7c, S7d). The above processes are recognized as the first indirect effect. Fig. S8 shows the model simulated autoconversion rate from cloud water into rain water, cloud liquid water path, and precipitation in CASE0 and in the IRE case with only the second indirect effect (in which aerosol perturbation to cloud droplet effective radius, i.e., the first indirect effect is turned off). It is impressive that the autoconversion rate decreased greatly by up to twenty times over portions of South China and the western Pacific compared with that in CASE0 (Fig. S8d, S8a) due to weaker collision between smaller cloud droplets, consequently leading to increasing cloud liquid water path (Fig. S8e) and decreasing rain droplets and precipitation (Fig. S8f), which is regarded as the second indirect effect. In the coastal areas of South China, the precipitation increase induced by the first indirect effect (Fig. S7d) was comparable to the precipitation reduction by the second indirect effect (Fig. S8f), resulting in less changes in precipitation by IRE (Fig. 8j).

Considering both DRE and IRE, TRE induced a decrease in T2 by up to 2 °C in South China, where WS10 decreased by ~0.5 m s⁻¹, with an apparent divergence there as well (Fig. 8k, l). The PBLH decrease induced by TRE mainly occurred in South China and the North China Plain, with the maximum decrease as high as 160 m in portions of South China (Fig. 8m). The aerosol radiative feedback consistently increased PM_2.5 concentrations in East China, with the maximum increase up to 15 μg m⁻³ in the North China Plain, Chongqing, and the middle reaches of the Yangtze River (Fig. 8n). The PM_2.5 increase induced by TRE was mainly attributed to the weakened K_z (Fig. S4h) and enhanced chemical production of secondary aerosols (Fig. S6 d-f) as discussed above. The decrease in K_z induced by IRE was larger than that by DRE in South China, whereas in North China, the K_z decrease by DRE was larger (Fig. S4). TRE induced a precipitation decrease up to 180 mm over wide areas of South China (Fig. 8o). The total precipitation decrease induced by TRE was larger than that by either DRE or IRE, but it was not a linear sum of them.

Fig. S9 is the same as Fig. 8 except for the EXP case. In general, the changes in meteorological variables and PM_2.5 in the BASE case were smaller than those in the EXP case due to lower aerosol levels caused by the lockdown in 2020 compared to those under 2019 emission conditions. Fig. 9 shows the difference between BASE and EXP (BASE minus EXP) cases in the TRE-induced changes in meteorological variables and PM_2.5 concentration averaged over the study period. Because there was little change in IRE between BASE and EXP cases, the difference induced by TRE was mainly contributed by DRE. The TRE-induced T2 decrease in the BASE case was smaller than that in the EXP case over East China (Fig. 8k vs Fig. S9k), indicating the COVID-19 emission reduction generally induced an anomalous surface warming during the study period in 2020, compared to the EXP case with emissions for 2019 (Fig. 9a). The maximum warming was as high as 0.3 °C in Chongqing and in the middle reaches of the Yangtze River. Consequently, the emission reduction induced an anomalous PBLH increase up to 30 m and an anomalous PM_2.5 decrease up to 14 μg m⁻³ around those areas of maximum warming (Fig. 9b, c). This revealed a weaker radiative feedback effect by TRE on PM_2.5 in the BASE case than that in the EXP case due to emission reduction during the COVID-19 lockdown. It is interesting to note that emission reduction alone led to a decrease of PM_2.5 up to 70 μg m⁻³ in the middle reaches of the Yangtze River discussed above (Fig. S3a), and aerosol radiative feedback caused additional 16 μg m⁻³ PM_2.5 decrease in those regions (Fig. 9c), suggesting the radiative feedback enhanced the PM_2.5 decrease due to emission reduction by ~20 % in East China. The precipitation decrease induced by TRE in the BASE case was generally smaller than that in the EXP case, leading to a positive precipitation anomaly of ~20 mm over wide areas of South China, despite sporadic negative anomalies occurred in some parts of South China (Fig. 9d). The difference in the TRE-induced precipitation decreases between BASE and EXP cases were mainly contributed by DRE (Fig. 8e vs Fig. S9e). Fig. S10 shows the model simulated cloud water path with only DRE or IRE in the BASE and EXP cases, respectively. It shows that cloud water path considering only DRE (indirect effect was turned off) in the EXP case is lower than that in the BASE case, which could be explained that higher aerosol loadings in the EXP case induced a stronger DRE and prevented more water vapor from moving upward, and thus less precipitation than in the BASE case. However, there was little difference in cloud water path with only IRE (direct effect was turned off) between BASE and EXP cases because IREs were almost the same between the two cases.

Fig. 9.

The model simulated differences in the changes of (a) T2, (b) PBLH, (c) surface PM_2.5, and (d) accumulated precipitation (PREP) induced by TRE averaged over the study period between the BASE and EXP cases (BASE-EXP) (units, T2: °C, PBLH: m, PM_2.5: μg m⁻³, PREP: mm). The numbers in the upper right corner of each panel denote averages over East China during the study period.

4.2.2. Regional averages

Table 5 presents the model simulated period-domain average changes in T2, WS10, PBLH, PM_2.5, and PREP induced by DRE, IRE, and TRE over the specific domains during the study period from the BASE and EXP cases. In the BASE case, the changes in T2, WS10 and PBLH induced by DRE in North China were slightly larger than those in South China, consistent with the relative magnitude of DRE in North and South China. As a result, the direct radiative feedback caused a larger PM_2.5 increase in North China (2.1 μg m⁻³) than that in South China (1.9 μg m⁻³). In North China, the changes in meteorology and PM_2.5 induced by DRE were somewhat larger than those induced by IRE, whereas in South China, the effect of IRE was remarkably larger than that of DRE, leading to approximately 1.0 °C surface cooling, 72 m decrease in PBLH, 4 μg m⁻³ increase in PM_2.5, and 41 mm decrease in accumulated precipitation in terms of domain average. The TRE-induced change was not equal to a simple sum of individual changes by DRE and IRE due to nonlinear interactions among chemistry, radiation and meteorology. The changes in T2, WS10, PBLH, PM_2.5, and PREP induced by TRE were estimated to be −1.3 °C, −0.2 m s⁻¹, −89 m, 5.6 μg m⁻³ and -48.3 mm in South China, which were more evident than the changes in North China. For East China, TRE totally caused decreases in T2, PBLH and PREP by 0.7 °C, 52 m, and 18.9 mm, and an increase in PM_2.5 by 3.1 μg m⁻³, respectively, in the BASE case.

Table 5.

The model simulated period-domain average changes in T2, WS10, PBLH, PM_2.5, and PREP induced by DRE, IRE, and TRE over specific domains during the study period from BASE and EXP (units, T2: °C, WS10: m s⁻¹, PBLH: m, PM_2.5: μg m⁻³, PREP: mm).

Region	DRE-induced changes					IRE-induced changes					TRE-induced changes
	T2	WS10	PBLH	PM2.5	PREP	T2	WS10	PBLH	PM2.5	PREP	T2	WS10	PBLH	PM2.5	PREP
BASE (with emission and meteorology in 2020):
NCa	−0.30	−0.04	−31.6	2.10	−2.8	−0.18	−0.03	−13.6	1.02	−3.7	−0.44	−0.06	−41.8	2.9	−5.7
SCb	−0.38	−0.07	−33.0	1.91	−27.5	−1.08	−0.16	−71.5	4.44	−40.7	−1.26	−0.18	−89.4	5.6	−48.3
ECc	−0.31	−0.05	−28.0	1.50	−10.7	−0.44	−0.07	−30.2	1.92	−15.5	−0.67	−0.10	−51.7	3.1	−18.9
EAd	−0.15	−0.02	−13.1	0.61	−5.0	−0.15	−0.02	−10.0	0.55	−6.1	−0.27	−0.03	−21.2	1.1	−7.3
EXP (with meteorology in 2020 but emission in 2019):
NCa	−0.38	−0.05	−39.5	3.57	−3.5	−0.18	−0.03	−13.2	1.40	−3.7	−0.50	−0.07	−48.3	4.5	−6.1
SCb	−0.46	−0.07	−40.1	4.11	−29.6	−1.08	−0.16	−71.6	6.98	−40.9	−1.30	−0.19	−93.9	9.6	−49.6
ECc	−0.37	−0.05	−33.7	2.77	−11.6	−0.44	−0.07	−30.2	2.94	−15.6	−0.71	−0.10	−56.1	5.0	−19.6
EAd	−0.17	−0.02	−14.8	0.94	−5.3	−0.15	−0.02	−10.0	0.84	−6.1	−0.28	−0.04	−22.5	1.6	−7.5

Region definition is the same as that in Table 4.

In the EXP case with 2019 anthropogenic emissions, it is noteworthy that all the changes induced by DRE were generally stronger than those in the BASE case; however, the IRE-induced changes were almost the same between EXP and BASE cases, due to the similar magnitude of IRE between two cases. It is noticed that the IRE-induced PM_2.5 change in the EXP case was larger than that in the BASE case (2.9 μg m⁻³ vs 1.9 μg m⁻³ for East China), which could be explained that the radiative feedback-induced PM_2.5 change at high aerosol loadings (EXP) was larger than that at lower aerosol levels (BASE) although IREs were almost the same between the two cases. TRE caused approximately 0.5 °C and 1.3 °C surface cooling, 4.5 and 9.6 μg m⁻³ PM_2.5 increases, and 6.1 and 49.6 mm precipitation reductions in North China and South China, respectively. By comparing the BASE and EXP cases, it is noteworthy that the emission reduction resulted in 0.06 °C and 0.04 °C surface warming, 1.6 and 4 μg m⁻³ PM_2.5 decrease, and 0.4 and 1.3 mm precipitation increase during the COVID-19 lockdown in 2020 (BASE) by aerosol radiative and feedback effects in North China and South China, respectively, compared to the EXP case with 2019 emissions. In the same way, for East China, the COVID-19 lockdown caused 0.04 °C surface warming, 1.9 μg m⁻³ PM_2.5 decrease and 0.7 mm precipitation increase on average, whereas it caused negligible changes in boundary layer meteorology over East Asia.

5. Conclusions

An online coupled regional climate–chemistry–aerosol model (RIEMS-Chem) was applied to explore the direct, indirect, and feedback effects of anthropogenic aerosols during the COVID-19 lockdown period from 23 January to 8 April 2020 over China. The impacts of the lockdown-induced emission reduction on radiation, boundary layer meteorology, and PM_2.5 concentrations were quantified. Model performance was validated against a variety of observations for meteorological variables (air temperature, wind speed, relative humidity, and precipitation), PM_2.5 and its chemical components, aerosol optical properties, as well as shortwave radiation flux at the top of the atmosphere. The comparison demonstrated that RIEMS-Chem was able to reproduce the spatial distribution and temporal variation of the above variables reasonably well.

During the study period, in terms of domain average, AOD was higher in North China (0.2) than that in South China (0.17). DRE dominated over IRE in most regions north of the Yangtze River, whereas IRE was evidently stronger than DRE in the Yangtze River regions and South China. IREs in South China (approximately −21 W m⁻² both at the surface and TOA) were significantly stronger (about 5 times) than those in North China (−4.1 W m⁻²), mainly due to both higher cloud amounts and aerosol concentrations in South China. Over East China and East Asia, domain-average DRE and IRE were comparable at the surface, whereas IRE was approximately 3 times the DRE at TOA.

Due to the lockdown-induced emission reduction, DRE during the study period in 2020 (BASE case) was reduced by up to 15 W m⁻² (43 %) in Chongqing and portions of the middle reaches of the Yangtze River, compared to the EXP case with 2019 anthropogenic emissions. However, the difference in IRE between BASE and EXP cases was small due to little sensitivity of cloud microphysical properties to CCN increasing at high CCN (aerosol) concentrations, although IRE was slightly weaker in BASE than that in EXP cases. In terms of domain average, surface TRE in North China decreased from −16.7 W m⁻² in the EXP case to −14.3 W m⁻² in the BASE case (14 %), while in South China, it decreased from −30.0 W m⁻² to −28.7 W m⁻² (4 %). Over East China, surface TRE decreased by approximately 7 % on average in the BASE case relative to EXP case, indicating a considerable reduction in the magnitude of total aerosol radiative effects due to the COVID-19 emission reductions.

DRE led to decreases in T2, WS10, and PBLH by up to 0.8 °C, 0.2 m s⁻¹, and 60 m in the North China Plain, Chongqing, and portions of southwest China. Consequently, PM_2.5 increased by up to 6 μg m⁻³ in the above regions due to radiative feedback. DRE induced a larger precipitation decrease in South China than that in North China, with the maximum decrease reaching 60 mm during the study period. IRE caused consistent decreases in T2 by ~1.7 °C and in WS10 by ~0.5 m s⁻¹ in South China, with an evident divergence around the areas of maximum cooling. IRE caused a PBLH decrease by up to 150 m and a more remarkable precipitation decrease by up to 140 mm in portions of South China. The feedback of IRE increased PM_2.5 concentration by ~10 μg m⁻³ in South China.

The COVID-19 emission reduction generally induced an anomalous surface warming during the study period in 2020. The maximum warming was as high as 0.3 °C in Chongqing and in the middle reaches of the Yangtze River. Consequently, the emission reduction induced an anomalous PBLH increase up to 30 m and an anomalous PM_2.5 decrease up to 14 μg m⁻³ in the areas of maximum warming. The emission reduction generally resulted in an anomalous increase in precipitation exceeding 20 mm over wide areas of South China. Above results suggest the weather could be warmer, with more precipitation and declined PM_2.5 level due to emission reductions by the COVID-19 lockdown.

In terms of domain average, in the BASE case, the changes in T2, WS10, PBLH and PM_2.5 induced by DRE in North China were slightly larger than those in South China. In North China, DRE induced larger changes in meteorology and PM_2.5 than those induced by IRE, whereas in South China, the effect of IRE was remarkably larger than that of DRE. Averaged over East China, TRE totally caused decreases in T2, PBLH and precipitation by 0.7 °C, 52 m, and 18.9 mm, and an increase in PM_2.5 by 3.1 μg m⁻³, respectively, during the COVID-19 lockdown in 2020. The IRE-induced changes were almost the same between EXP and BASE cases. Through aerosol radiative and feedback effects, the COVID-19 emission reductions resulted in 0.06 °C and 0.04 °C surface warming, 1.6 and 4 μg m⁻³ PM_2.5 decrease, and 0.4 and 1.3 mm precipitation increase during the lockdown period in 2020 in North China and South China, respectively, compared to the EXP case, however, the lockdown caused negligible changes on average over East Asia.

While this study reveals considerable radiative and feedback effects of anthropogenic aerosols during the COVID-19 lockdown over China, there are still potential uncertainties which may be further addressed: 1.) emission inventory during the lockdown period needs further check because the degree of emission reduction varies considerably among provinces due to different epidemic situation in China, which could bias model predictions high or low; 2.) secondary aerosols are generally underpredicted, especially during haze events, which could underestimate direct radiative and feedback effects over China and requires further development of secondary aerosol formation mechanism; 3.) precipitation is generally underpredicted over China, which may lead to overpredictions of aerosol concentration and bias in indirect effect; 4.) this study focuses on the responses of radiation and meteorology to emission changes during a short-time lockdown period, with prescribed monthly mean sea surface temperature and without considering interaction between aerosols and sea surface flux. As the COVID-19 epidemic is ongoing, long-term impacts of the COVID-19 epidemic on air quality, radiation and climate change will be explored in the future.

CRediT authorship contribution statement

Mingjie Liang: Investigation, Formal analysis, Software, Validation, Writing – original draft.

Zhiwei Han: Methodology, Investigation, Writing – review & editing, Supervision.

Jiawei Li: Software, Investigation, Validation.

Yele Sun: Resources, Data curation.

Lin Liang: Validation, Data analysis.

Yue Li: Software, Data analysis.

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.

Editor: Jianmin Chen

Appendix A. Supplementary data

Supplementary material

Data availability

Data will be made available on request.