Why did air quality experience little improvement during the COVID-19 lockdown in megacities, northeast China?

Abstract

To inhibit the COVID-19 (Coronavirus disease 2019) outbreak, unprecedented nationwide lockdowns were implemented in China in early 2020, resulting in a marked reduction of anthropogenic emissions. However, reasons for the insignificant improvement in air quality in megacities of northeast China, including Shenyang, Changchun, Jilin, Harbin, and Daqing, were scarcely reported. We assessed the influences of meteorological conditions and changes in emissions on air quality in the five megacities during the COVID-19 lockdown (February 2020) using the WRF-CMAQ model. Modeling results indicated that meteorology contributed a 14.7% increment in Air Quality Index (AQI) averaged over the five megacities, thus, the local unfavorable meteorology was one of the causes to yield little improved air quality. In terms of emission changes, the increase in residential emissions (+15%) accompanied by declining industry emissions (−15%) and transportation (−90%) emissions resulted in a slight AQI decrease of 3.1%, demonstrating the decrease in emissions associated with the lockdown were largely offset by the increment in residential emissions. Also, residential emissions contributed 42.3% to PM_2.5 concentration on average based on the Integrated Source Apportionment tool. These results demonstrated the key role residential emissions played in determining air quality. The findings of this study provide a scenario that helps make appropriate emission mitigation measures for improving air quality in this part of China.

1. Introduction

The geographical northeast of China comprises three provinces, Heilongjiang, Jilin, and Liaoning. These provinces have been featured by heavy industries and long and cold winters, often experiencing severe air pollution in winter (January, February, November, and December). Thirty-two warnings for serious air pollution were issued by the three provincial capitals (Shenyang, Changchun, and Harbin) from 2018 to 2020. In recent years, persistent air pollution has been intermittently witnessed in megacities of northeast China, for example, a haze event lasted for seven days in January 2020, and the AQI (Air Quality Index) peaked at up to 500. Before 2017, air pollution in these provinces was gradually alleviated subject to emission control actions but got worse again in the subsequent three years, characterized by an increasing frequency of heavy haze events. Concerns were raised about the effectiveness of the emission control measures taken in previous years. Following the occurrence of COVID-19 (Coronavirus disease 2019) and nationwide lockdown in China from January to March 2020, many regions in China experienced considerable reductions in social and human activities, which significantly reduced anthropogenic emissions (Lu et al., 2021; Wang et al., 2021b; Zhao et al., 2020b). Surprisingly, during this period, the mean AQI of cities in northeast China was 79, which almost remained at the same level as that of 2018 (77) and 2019 (82), which brought attention to the scientific communities and policymakers.

Like northeast China, the Beijing-Tianjin-Hebei (BTH) region and Yangtze River Delta (YRD) also experienced abnormal haze events (Chang et al., 2020; Sulaymon et al., 2021; Zhao et al., 2020a). These abnormal air pollution events have been attributed to unfavorable meteorological conditions, such as low wind speed, low precipitation, and low planetary boundary layer height (PBLH), which offset the benefits of emission reductions (Gao et al., 2021; Sulaymon et al., 2021; Zhang et al., 2021a). Regional atmospheric transport of pollutants driven by meteorology and elevated secondary aerosol also played vital roles in air quality degradation during the lockdown. For example, Lv et al. (2020) demonstrated that increased secondary particles from a distance were transported in downtown Beijing, aggravating air pollution. Wang et al. (2021a) attributed air pollution in the Pearl River Delta (PRD) during the lockdown to the increments in secondary organic and inorganic aerosols resulting from regional transport. Other studies indicated that unexpectedly growing household coal burning, which was one of the major sources of anthropogenic emissions in northern China, also caused air pollution (Wu et al., 2018; Zhang et al., 2019). Referring to articles that have been published, we find studies relative to factors influencing air quality in northeast China during the lockdown are insufficient; few works discussed the response of air quality to meteorology and emission changes in China, including northeast China, but this part of China was not their main focus. Thus, more studies are needed to probe the reasons for the little improvement in air quality in northeast China.

In this study, extensive WRF-CMAQ (the Weather Research and Forecast model coupled with Community Multiscale Air Quality) model simulations were performed to address the concerns about the little improvement in air quality in five megacities (Shenyang, Changchun, Jilin, Harbin, and Daqing) in northeast China during the lockdown, and to discuss the factors impacting air quality. In particular, the contributions of meteorological conditions to the air quality of these megacities were considered. The influences of varied emissions on air quality were also addressed, especially residential and industry emissions. The low temperature (−15 °C) accompanied by extended stay-at-home due to the lockdown enhanced feed coal and biomass burning for heating. The increasing residential emission was supposed to be responsible for the little improvement in air quality. Contrarily, declining industry emissions resulting from the lockdown were discerned. It is known that northeast China is an important industrial base, with the stereotype regarding industry emissions dominating air quality in this part of China during the lockdown will be discussed. The present study aims to fill knowledge gaps in understanding how and to what extent the meteorology and emissions would otherwise contribute to deteriorated air quality during the COVID-19 lockdown, which is also helpful in adopting appropriate measures to mitigate air pollution in this part of China.

2. Data and methods

2.1. Study region and data

The study region is located in northeast China, with an area of 528,300 km² and 34.5 million residents. Five targeted megacities, Shenyang, Changchun, Harbin, Jilin, and Daqing, were selected, with the first three megacities being the capital of Liaoning province, Jilin province, and Heilongjiang province, respectively. The meteorological observation data were obtained from China Meteorological Data Service Centre (http://data.cma.cn). Measured concentrations of PM_2.5, PM₁₀, SO₂, NO₂, CO, and O₃ were collected from the China National Environmental Monitoring Center (http://106.37.208.233:20035/). Moreover, information related to COVID-19 in cities could be found in the National Health Commission of the People's Republic of China (http://www.nhc.gov.cn/), the official websites of municipal governments, and municipal health commissions.

2.2. WRF-CMAQ configurations and modeling scenarios

The Combined Weather Research and Forecasting (WRF v3.9)/model with the Community Multiscale Air Quality (CMAQ v5.02) model was employed to simulate meteorological variables and pollutant concentrations. Compared with the observation-based approach, the modeling approach features some advantages. The main target of this research is to quantify the influences of meteorological conditions and emission changes on air quality, and the modeling approach is available to explore contributions of meteorology and emissions to air quality by using scenario analysis. The modeling approach makes it more efficient and cost-effective to obtain meteorological variables in downtowns, with the available observed meteorological data (derived from http://www.nmic.cn, most stations located in the suburb) being unable to reflect urban meteorology. Here, two nested domains with spatial resolutions at 27 km (domain 1) and 9 km (domain 2) were set up, with domain 2 covering the five targeted megacities (Fig. 1 ). The simulation period spanned from 1 January to March 31, 2018 and 2020, respectively, with the first 3 day as spin-up. The Multi-resolution Emission Inventory for China in 2016 (MEIC, http://www.meicmodel.org) and the Model of Emissions of Gases and Aerosols from Nature (MEGAN, version 2.04) provided anthropogenic and biogenic emissions, respectively. Furthermore, a new land cover dataset in 2017 with 1 km resolution derived from Tsinghua University was incorporated into WRF to improve the accuracy of simulated meteorological variables (Fu et al., 2021b). Details of model configurations are presented in Table 1 .

Fig. 1.

The domains (left panel) and observational stations (right panel). The black dot on the left panel indicates the location of the city. The green and red dots on the right panel indicate the meteorology and air quality monitoring stations, respectively.

Table 1.

Model configuration.

Configuration	Description
Domain size	90 × 90 (domain 1), 82 × 82 (domain 2)
Horizontal resolution	27 km × 27 km, 9 km × 9 km
Vertical level	30 vertical levels
Spin-up time	Three days
Modeled episode	1 January to March 31, 2018 and 2020
Land surface scheme	Noah land surface model (Chen and Dudhia, 2001)
Planetary boundary layer scheme	YSU (Hong et al., 2006)
Longwave radiation scheme	RRTM (Mlawer et al., 1997)
Shortwave radiation scheme	Dudhia (Dudhia, 1989)
Initial and boundary conditions	NCEP-FNL (1° × 1°)
Gas-phase mechanism	CB05 (Sarwar et al., 2008)
Anthropogenic emission inventory	MEIC (http://www.meicmodel.org)
Biogenic emission inventory	MEGAN (Guenther et al., 2006)

Abbreviation: NCEP-FNL, National Centers for Environmental Prediction-Final Analysis dataset.

The five megacities were almost simultaneously locked (from 27 January to January 29, 2020) and unlocked (from 29 February to March 2, 2020). During the lockdown, social activities were almost at a standstill. We selected January (before lockdown), February (during lockdown), and March (after lockdown) in 2020 to elucidate the response of the air quality to meteorological conditions, while the influence of emission changes on air quality was focused on during the lockdown. Eleven model scenarios were designed (Table 2 ). January, February, and March 2018 were selected as the reference months in that their AQIs were the lowest in the last ten years, largely attributable to favorable meteorological conditions (Fu et al., 2021a). The air quality status is evaluated by AQI, and it is calculated according to China's national standards defined in the Technical Regulation on Ambient Air Quality Index. AQI has six levels, which include: “good” (0–50), “moderate” (51–100), “lightly polluted” (101–150), “moderately polluted” (151–200), “heavily polluted” (201–300), and “severely polluted” (>300).

Table 2.

Modeling scenarios.

Scenario	Period	Emission inventory
MET2018_JAN	January 2018	Default MEIC
MET2018_FEB	February 2018	Default MEIC
MET2018_MAR	March 2018	Default MEIC
MET2020_JAN	January 2020 (before lockdown)	Default MEIC
MET2020_FEB	February 2020 (during lockdown)	Default MEIC
MET2020_MAR	March 2020 (after lockdown)	Default MEIC
EMI2020_IND	February 2020	Industry −15%
EMI2020_TRA	February 2020	Transportation −90%
EMI2020_RES	February 2020	Residential +15%
EMI2020_ITR	February 2020	Industry −15%, Residential +15%, and Transportation −90%
EMI2020_MAR	March 2020	Residential +5%, Transportation −70%

Note: The changes in emissions, for example, “Industry −15%“, indicate a 15% reduction in industrial emissions, and the same as residential and transportation emissions.

Modeling scenarios are presented in Table 2. The contributions of meteorology and emissions to air quality are determined by changing a single factor. Specifically, MEIC inventory remains the same in scenarios while quantifying the contribution of meteorology to AQI. Also, meteorological conditions are uniformly employed in scenarios involving the contribution of emission changes. The configurations of the initial and boundary conditions, physical parameterization schemes, and sea-salt file were adopted uniformly in all scenarios. These measures could reduce the uncertainties of modeling runs. Following Dong et al. (2020) and Zhang et al. (2021b), the contribution ratio of meteorological conditions to AQI in domain 2 is estimated by the fractional change following formula (1). Formula (2) is adopted to obtain the contribution ratio by emission changes.

During the lockdown, the change in industrial emissions was based on National Development and Reform Commission (NDRC). The value added of industrial enterprises above the designated size of Liaoning, Jilin, and Heilongjiang reduced by 8.5%, 12.2%, and 8.2% in the first quarter of 2020, according to NDRC. Additionally, considering the impacts of the lockdown on small and medium-sized enterprises, the industry emission was reduced by 15%. According to AutoNavi, derived from Urban Traffic Analysis Report Transportation, the transportation emission was reduced by 90% during the lockdown. We referred to a 10% increase in residential emissions in Northern China (Sulaymon et al., 2021; Wang et al., 2020) influenced by the COVID-19 lockdown. The proportion rose by 15% in this study while considering the increasing heating demand induced by the “stay-at-home” orders. After the lockdown (March 2020), the heating demand decreased, and a 5% increase in residential emissions was used. In terms of a 70% decrease in transportation emissions, it was based on transportation recovery reports derived from government official websites.

$$\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{i}\mathrm{b}\mathrm{u}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}=\frac{\left(\mathrm{M}\mathrm{E}\mathrm{T}2020-\mathrm{M}\mathrm{E}\mathrm{T}2018\right)}{\mathrm{M}\mathrm{E}\mathrm{T}2018}\times 100\%$$ (1)

Where MET2020 and MET2018 are scenarios relative to the simulated AQI using meteorological conditions in January, February, and March of 2020 and 2018, including MET2018_JAN, MET2018_FEB, MET2018_MAR, MET2020_JAN, MET2020_FEB, and MET2020_MAR.

$$\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{i}\mathrm{b}\mathrm{u}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}=\frac{\left(\mathrm{E}\mathrm{M}\mathrm{I}2020-\mathrm{M}\mathrm{E}\mathrm{T}2020\right)}{\mathrm{M}\mathrm{E}\mathrm{T}2020}\times 100\%$$ (2)

The same as formula (1) but for contribution ratio by emission changes.

2.3. Model validation metrics

The simulated wind speed at 10 m (WS10), temperature at 2 m (T2), relative humidity at 2 m (RH2), and planetary boundary layer height (PBLH) were validated against observations. Six simulated air pollutants, including PM_2.5, PM₁₀, SO₂, NO₂, CO, and O₃, were used to calculate AQI. The simulated pollutant concentrations and AQI were validated against observational data. Mean bias (BIAS), mean absolute error (MAE), root-mean-square error (RMSE), and Pearson correlation coefficient (CC) were employed for the validation of modeling results.

$$\mathrm{B}\mathrm{I}\mathrm{A}\mathrm{S}=\frac{1}{\mathrm{n}}\sum _{\mathrm{i}=1}^{\mathrm{n}}\left({\mathrm{S}}_{\mathrm{i}}-{\mathrm{O}}_{\mathrm{i}}\right)$$

$$\mathrm{M}\mathrm{A}\mathrm{E}=\frac{1}{\mathrm{n}}\sum _{\mathrm{i}=1}^{\mathrm{n}}\left|{\mathrm{S}}_{\mathrm{i}}-{\mathrm{O}}_{\mathrm{i}}\right|$$

$$\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}=\sqrt{\frac{1}{\mathrm{n}}\sum _{\mathrm{i}=1}^{\mathrm{n}}{\left({\mathrm{S}}_{\mathrm{i}}-{\mathrm{O}}_{\mathrm{i}}\right)}^2}$$

$$\mathrm{C}\mathrm{C}=\frac{{\sum }_{\mathrm{i}=1}^{\mathrm{n}}\left({\mathrm{S}}_{\mathrm{i}}-{\mathrm{S}}_{\mathrm{a}}\right)\left({\mathrm{O}}_{\mathrm{i}}-{\mathrm{O}}_{\mathrm{a}}\right)}{\sqrt{{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\left({\mathrm{S}}_{\mathrm{i}}-{\mathrm{S}}_{\mathrm{a}}\right)}^2}\sqrt{{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\left({\mathrm{O}}_{\mathrm{i}}-{\mathrm{O}}_{\mathrm{a}}\right)}^2}}$$

Where S_i and O_i represent simulated data and observations, S_a and O_a are simulated data and observations on average. “N” is the frequency of data.

3. Results and discussion

3.1. Verification of simulated results

The simulated meteorological variables and pollutant concentrations were evaluated against measurements. Those simulated and measured meteorological variables were verified over the five megacities using statistical metrics. As shown in Fig. 2 , the BIAS and RMSE of simulated WS10 met the criteria of the European Environment Agency (EEA, 2011). The BIAS and RMSE of simulated T2 averaged by all the periods were -2.8°C and 3.7°C, while the two metrics were 9.5% and 13.9% for RH2, which are reasonable compared to other studies (Li et al., 2020; Liu et al., 2019; Tan et al., 2017; Wang et al., 2016; Zhang et al., 2017). The CC values between simulated WS10, T2, and RH2 and their corresponding measurement were all greater than 0.72, suggesting that the simulated results matched well with observations. The simulated results can reflect the relationships between meteorological variables and air quality. Generally, low WS10, T2, and PBLH paired with high RH2 are expected during persistent haze events. For example, in January 2020, a persistent haze event occurred over the five megacities with an AQI over-standard value of 118 on average (Fig. 3 ). Accordingly, the simulated WS10, T2, and PBLH were reduced by 0.7 m/s, 1.4 °C, and 142 m accompanied by RH2 increasing by 9.5% as compared with that in 2018.

Fig. 2.

Hourly time series of (A) WS10, (B) T2, (C) RH2, and (D) PBLH. The black dot and solid blue line in (A), (B), and (C) represent observations and simulations. The simulated hourly PBLH in 2018 and 2020 are shown in (D). All modeled and observed meteorological variables are averaged over the five megacities.

Fig. 3.

Time series of simulated and observed PM_2.5, PM₁₀, SO₂, NO₂, CO, and O₃ (μg/m³) at 3-hour intervals. The black dot stands for simulated values, and the solid red line represents measurements. Both simulated and measured concentrations are averaged over five megacities.

CMAQ tended to underestimate PM_2.5, PM₁₀, NO₂, CO, and O₃ concentrations in January, especially during the haze events but performed better in February and March. As shown in Fig. 3, the model captured fluctuations and magnitudes of concentrations of the six pollutants relative to observations, characterized by high CC values. The simulated PM_2.5 (the primary pollutant in winter) was better aligned with the measurement among the six criteria pollutants.

3.2. Overview of factors influencing air quality

Simulated pollutants (PM_2.5, PM₁₀, SO₂, NO₂, CO, and O₃) concentrations and AQI of grids within downtowns of the five megacities were extracted. Contributions of factors to AQI showed differences from January to March. As shown in Table 3 , the positive fraction indicates that factors yield AQI increment and air quality degradation, and vice versa. The results revealed that meteorology significantly contributed to the AQI changes (>30%) in the five megacities in January 2020 and then decreased to 14.7% on average in February 2020. In March 2020, the meteorology made a negative contribution, indicating that favorable meteorological conditions alleviated air pollution. The emission changes over northeast China yielded the AQI fractions of −8.6%–2.4% in February 2020, and the range was −3.5%–0.2% in March 2020. However, such a degree of change was insignificant because the contribution ratio presented in Table 3 is based on EMI2020_ITR and EMI2020_MAR, thus, the benefits of reduced emissions are largely offset by the increase in residential emissions.

Table 3.

AQI fractions (%) in the five megacities before, during, and after the COVID-19 lockdown influenced by meteorological conditions (MET) and emissions (EMI, based on EMI2020_ITR and EMI2020_MAR).

City	January 2020 (Before lockdown)		February 2020 (During lockdown)		March 2020 (After lockdown)
	MET	EMI	MET	EMI	MET	EMI
Shenyang	41.4	–	8.3	−8.6	−14.2	−3.5
Changchun	46.6	–	25.3	−3.5	−40.9	−1.8
Jilin	36.6	–	12.5	0.4	4.4	−0.5
Harbin	43.5	–	14.0	2.4	−9.6	0.2
Daqing	30.2	–	13.3	−6.4	9.7	−1.0

In this study, reference months (January, February, and March in 2018) instead of climate averages over multi-years were adopted to achieve the contribution of meteorology to AQI, and the reason was that the contribution by meteorology based on different reference years (2010, 2016, 2017, and 2019) varied considerably, in the range of −20%–10% for AQI and ranging from −20.1 μg/m³ to 12.6 μg/m³ for PM_2.5 concentration (Chen et al., 2019; Xu et al., 2020). Additionally, multi-year meteorological conditions were not representative to reflect certain periods spanning days or weeks.

3.3. Impact of meteorology

As shown in Table 3, the meteorology-induced changes in AQI increased from 8.3% to 25.3% during the lockdown (February 2020) in the downtowns of the five megacities, with the largest increase occurring in Changchun, manifesting that the meteorology could be considered a critical factor to induce little improvement in air quality. Spatially, unfavorable meteorology caused increasing AQI in downtowns and suburbs of the five megacities (Fig. 4 ). It is noticed that studies relative to the contribution of meteorology during the lockdown in northeast China are deficient. The contribution ratio here was compared with the other regions in China by selecting simulated PM_2.5 as an example. In this study, the simulated PM_2.5 increased by 17.4% (8.1 μg/m³) as contributed by meteorology. Changes in PM_2.5 are +3.4%, +9.8%, −16.1%, and −18.3% in Beijing, Wuhan, Shanghai, and Guangzhou (positive value indicates meteorology induces PM_2.5 increase and vice versa) achieved by Zhao et al. (2020b). Likewise, Song et al. (2021) find changes in PM_2.5 concentration in the sequence of Guangdong province (+1.1 μg/m³), Sichuan province (−0.6 μg/m³), Shanghai (−0.9 μg/m³), and Hubei province (−5.0 μg/m³). It was found that PM_2.5 experienced the highest increase in the study region, inferring that northeast China suffered from worse meteorological conditions during the lockdown as compared with other regions in China.

Fig. 4.

The contribution of meteorology to AQI changes in percentage during the lockdown.

The simulated meteorological variables during the lockdown were compared with that of 2018 (Table 4 ). The mean WS10 and PBLH reduced by 0.5 m/s and 107 m, and RH2 increased by 8.8% compared with that in 2018, averaged by the five megacities. The results suggested atmospheric conditions during the lockdown were more stable than in 2018, which favored the accumulations of air pollutants in the lower atmosphere. The Pearson correlation coefficients (in the bracket in Table 4) between AQI and simulated WS10 and PBLH were negative during the lockdown (P < 0.05), indicating that large AQI values correspond to declining WS10 and PBLH. T2 was negatively correlated with AQI in Changchun, Jilin, and Harbin (P < 0.05). No significant correlations between RH2 and AQI were identified, but stronger air pollution seemed to be associated with higher RH2, which was favorable to the growth of aerosol particles.

Table 4.

Comparisons of simulated WS10, T2, RH2, and PBLH in downtowns of the five megacities between February 2018 and February 2020.

City	WS10 (m/s)		T2 (°C)		RH2 (%)		PBLH (m)
	2018	2020	2018	2020	2018	2020	2018	2020
Shenyang	3.2	2.9 (−0.67*)	−7.4	−8.4 (−0.29)	54.4	72.1 (0.12)	438	315 (−0.71*)
Changchun	3.3	2.9 (−0.64*)	−11.7	−13.3 (−0.48*)	64.3	74.1 (−0.07)	376	269 (−0.68*)
Jilin	3.1	2.4 (−0.63*)	−13.2	−13.4 (−0.51*)	75.4	78.9 (0.05)	321	220 (−0.62*)
Harbin	3.2	2.6 (−0.59*)	−16.5	−16.5 (−0.36*)	75.5	79.2 (−0.04)	248	169 (−0.57*)
Daqing	3.2	2.7 (−0.42*)	−15.2	−16.6 (−0.03)	67.6	77.0 (0.25)	313	185 (−0.38*)

Note: Pearson correlation coefficients between the AQI and the four meteorological variables are shown in the bracket. * represents the significant level at p < 0.05.

3.4. Impact of emission changes

To highlight the response of air quality to changes in local emissions in the study region during the COVID-19 lockdown (February 2020), three simulations were conducted by letting the industry and transportation emissions drop by 15% (EMI2020_IND) and 90% (EMI2020_TRA) and a 15% increase in the residential emission (EMI2020_RES). The response of AQI to emission changes was different in megacities (Table 5 ). AQI experienced decreases in the range of 5.5% to 9.8% when industry emission was reduced by 15%. Similarly, a 90% reduction in transportation emissions yielded AQI decreases within the range of 3.0% to 5.4%. The considerable reduction in transportation emissions in the five megacities caused an obvious reduction in NO_X concentration in the range of 15.8%–22.3%. The degree of AQI increases presented as Shenyang (5.6%) < Daqing (5.9%) < Changchun (7.4%) < Jilin (8.5%) < Harbin (10.4%) when residential emissions increased by 15%. The increase in residential emissions exerted the greatest impact on air quality in Harbin, while air quality in Shenyang and Daqing was mainly determined by industry emissions. The difference could be ascribed to the contribution of emission sectors to major pollutants. Here, the CMAQ-ISAM (Integrated Source Apportionment) tool was employed to investigate the proportions of major pollutants contributed by different emission sectors based on EMI2020_ITR. As seen in Fig. 5 , PM_2.5 and SO₂ concentrations were dominated by industry and residential emissions in the five megacities. The sum of the proportion was in the range of 84.3%–88.2% for PM_2.5 and 81.7%–90.8% for SO₂. Residential emissions determined PM_2.5 concentration (the primary pollutant determining AQI in northeast China in winter) in Harbin, Changchun, and Jilin, contributed 50.5%, 46.2%, and 45.8%, respectively.

Table 5.

Changes in AQI (%) induced by emission changes.

City	EMI2020_IND	EMI2020_TRA	EMI2020_RES	EMI2020_ITR
Shenyang	−9.1	−5.4	5.6	−8.6
Changchun	−7.6	−4.0	7.4	−3.5
Jilin	−5.7	−3.0	8.5	0.4
Harbin	−5.5	−3.2	10.4	2.4
Daqing	−9.8	−3.1	5.9	−6.4

Fig. 5.

Contributions of different sectors to PM_2.5, SO₂, and NO_X concentrations in percentage for the five megacities.

Simulated results of EMI2020_ITR showed that the changes in emissions from industry (−15%), transportation (−90%), and residential (+15%) emissions led to changes in AQI of −8.6%−2.4%. The AQI decreased by 3.1% averaged by the five megacities. However, the magnitude of the AQI reduction seemed not to respond sufficiently to the considerable emission reductions. Residential emissions increment may be responsible for this phenomenon. In February 2020, the onset of lockdown in northeast China, followed by the end of the Spring Festival holiday, had brought migrant workers to stay in the suburb. Take Harbin as an example, during the lockdown, the number of residents staying in the suburb increased by about 0.34 million compared to 2019 according to reports by Chinese official media (https://news.cctv.com and http://www.people.com.cn). According to a report from Harbin Daily, the low temperature (−11.6 °C on average in February 2020) accompanied by the “stay-at-home” orders enhanced feed coal consumption for domestic heating and cooking in the suburb. It was speculated that the emissions resulting from feed coal burning were the main cause of air pollution in Harbin during the lockdown. The results of ISAM revealed that residential emissions dominated PM_2.5 concentration in Harbin. The increase in residential emissions (making AQI enhanced by 10.4%) would overwhelm the benefits of emission reductions (AQI dropped by 8.7%), yielding a 2.4% increase in AQI. These results provided support for the speculation by Harbin Daily. AQI experienced a slight increase of 0.4% in Jilin, meaning that the benefit brought by emission reduction was thoroughly offset by the increment in residential emissions. In Shenyang, Changchun, and Daqing, the declining AQI resulted from the reductions of industry and transportation emissions was largely offset by the increment in residential emissions.

From the above results, we found that residential emissions could be considered the main cause for AQI changes in megacities of northeast China instead of the stereotype regarding industrial emissions dominating air quality (these megacities have been known as major industrial bases in China). From a broad perspective, the dominating emission source switching was attributed to industrial emissions control. In the past decade, abatements in industrial emissions have been a national strategy to improve air quality. Perhaps more obviously, during the lockdown, the longer period of stay-at-home of residents enhanced the demand for feed coal and biomass fuels consumption. Thus, the important role played by residential emissions to AQI change was discerned. Residential emissions, especially rural household coal consumption without end-of-pipe controls, urgently need focus in future air pollution controlling works (Table 6 ). In addition, previous investigations of the factors influencing air quality during the lockdown focused on the whole of northeast China. However, the present study examined the responses of air quality to emission changes on the city scale over this part of China. More details relative to the influence of emission changes on air quality were achieved.

Table 6.

AQI in downtowns and suburbs of different megacities during the lockdown.

	Shenyang Downtown	Shenyang Suburb	Changchun Downtown	Changchun Suburb	Jilin Downtown	Jilin Suburb	Harbin Downtown	Harbin Suburb	Daqing Downtown	Daqing Suburb
EMI2020_ITR	57	40	69	51	74	46	72	59	37	28

4. Conclusion

Although the reduction of anthropogenic activities had been witnessed during the COVID-19 lockdown (February 2020) in northeast China, the air quality in the major megacities (Shenyang, Changchun, Jilin, Harbin, and Daqing) in this part of China did not show significant improvement. In probing the reasons for this abnormal phenomenon, the responses of the air quality of the five megacities to meteorological conditions and emission changes were explored via multiple model scenarios using the WRF-CMAQ model.

Simulated results revealed that the unfavorable meteorology was a cause, resulting in a 14.7% increment in AQI averaged over the five megacities during the lockdown. Low WS10 and PBLH accompanied by high RH2 relative to their baseline data during the same period in 2018 were observed. To mitigate the air pollution resulting from such unfavorable meteorological conditions, additional and emergency measures to further reduce emissions are essential. We found that the residential emission instead of the industry emission determined air quality in Harbin and Jilin during the lockdown. The increasing residential emission (+15%) accompanied by declining emissions from industry (−15%) and transportation (−90%) resulted in a slight decrease in AQI of 3.1% averaged by the five megacities, meaning that the benefits of reduced emissions were largely offset by increased residential emissions, which was considered another important reason contributing to little improvement in air quality. The results provide references for policymakers to control residential emissions for mitigating air pollution. More special provincial policies, like the Plan of Scattered Coal Control from 2020 to 2022 in Heilongjiang Province, are urgently needed for those megacities across northeast China that rely heavily on household coal heating in winter.

Author statement

Donglei Fu: Conceptualization, Data Curation, Formal analysis, Investigation, Methodology, Software, Validation, Writing - Original Draft. Xiaofei Shi: Conceptualization, Data Curation, Software, Validation. Jinxiang Zuo: Data Curation, Validation. Stephen Dauda Yabo: Formal analysis, Writing - Review & Editing. Jixiang Li: Data Curation, Resources. Bo Li: Conceptualization, Formal analysis, Methodology, Software. Haizhi Li: Data Curation, Resources. Lu Lu: Methodology, Resources. Bo Tang: Resources, Software. Hong Qi: Conceptualization, Formal analysis, Funding acquisition, Methodology, Resources, Supervision, Validation, Writing - Review & Editing. Jianmin Ma: Conceptualization, Formal analysis, Methodology, Software, Writing - Review & Editing.

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

Data will be made available on request.