Spatiotemporal characteristics and co-effects of air quality and carbon dioxide emissions changes during the COVID-19 epidemic lockdown measures in China

Abstract

The COVID-19 pandemic prompted several nations, including China, to enact unprecedented lockdown measures, leading to significant alterations in environmental conditions. Previous studies have solely analysed the impact of lockdown measures on air pollutants or carbon dioxide (CO₂) emissions during the COVID-19 pandemic in China, but few have focused on the spatio-temporal change characteristics and synergistic effects between the two. In this study, we constructed a methodological framework to examine the spatiotemporal characteristics and co-effects of air quality (PM_2.5, SO₂, and NO₂) and CO₂ changes in 324 prefecture-level cities in China due to the COVID-19 blockade measures from January 24 to April 30, 2020, using the regression discontinuity in time method and co-effect control coordinate system. The results show that a significant improvement in air quality and CO₂ emissions during the lockdown period, with notable north‒south heterogeneity. During the major lockdown period (January 24 to February 29), the measures resulted in respective reductions of 5.6%, 16.6%, and 25.1% in the concentrations of SO₂, NO₂, and CO₂ nationwide. The proportions of cities with negative treatment effects on PM_2.5, SO₂, NO₂, and CO₂ were 39.20%, 70.99%, 84.6%, and 99.38%, respectively. Provinces where concentrations of CO₂ and NO₂ declined by over 30% were primarily concentrated in southern areas of the ‘Yangtze River Defense Line’. Starting from March, the improvement effect of air quality and CO₂ has weakened, and the concentration of air pollutants has rebounded. This study offers crucial insights into the causal effects of lockdown measures on air quality changes, and reveals the synergy between air quality and CO₂, thereby providing a reference for devising effective air quality improvement and energy-saving emission reduction strategies.

Graphical abstract

1. Introduction

Air quality (AQ), carbon dioxide (CO₂) emissions, and the Corona Virus Disease 2019 (COVID-19) pandemic are three major issues that have a profound impact on human health, the environment, and the global economy (Wu et al., 2020). Air pollution is a critical public health issue that has been linked to numerous chronic diseases, including respiratory and cardiovascular diseases, as well as cancer. CO₂ emissions are the primary driver of climate change, and has far-reaching impacts on the environment and human health. Scholars have made attempts to mitigate CO₂ emissions and to improve air quality from multiple perspectives, including adjusting energy structures and increasing the efficiency of energy (Lin et al., 2010; Zou et al., 2023; Wang et al., 2021a), increasing the use of clean energy (solar, wind, hydrogen) (Mikulčić et al., 2022), developing environmentally friendly materials (Cao et al., 2021, 2022; Wang et al., 2021b; Rufford et al., 2012; Shamsabadi et al., 2017; Hou et al., 2015), promoting ecological carbon sequestration (Deng et al., 2023; Zomer et al., 2017), and accelerating the demonstration and application of carbon capture, utilization and storage (CCUS) technology (Sepehri and Sarrafzadeh, 2018; Sepehri et al., 2020). Furthermore, financial emission reduction measures such as carbon taxation and carbon trading (World Bank, 2017; Wang et al., 2022) and the implementation of new energy transportation vehicles (Li et al., 2016) have been widely advocated and implemented.

Air pollution and high CO₂ emissions exacerbate climate change and public health problems, making individuals more vulnerable to respiratory diseases, including COVID-19 (Sathe et al., 2021). The COVID-19 pandemic has caused unprecedented disruptions to societies and economies worldwide, and its impact on air quality and CO₂ emissions has become an emerging topic of research (Le Quéré et al., 2020). Studies have indicated that lockdown measures had a positive impact on optimizing air quality in the short term. An analysis of air quality data from 164 countries revealed that the concentrations of global NO₂ and PM_2.5 decreased by 5% and 4%, respectively, compared to prelockdown levels (Dang and Trinh, 2021). In Portugal and Spain, levels of NO₂ and PM₁₀ decreased in March, April, and May due to restrictions on maritime traffic to limit the spread of the pandemic (Silva et al., 2021). In India, a densely populated country, a dramatic reduction in air pollution in urban areas was observed when the emissions of pollutants produced by human activities were completely shut down or restricted (Singh and Nanda, 2020; Sathe et al., 2021). PM_2.5 levels in Mexico, South Africa, Italy, Spain, and Australia decreased in 2020 compared to 2019; however, some countries, such as Canada, the United States, and Russia experienced an increase in PM_2.5 levels (Das et al., 2020). Similarly, the coronavirus, as a destructive factor affecting socioeconomic activities, air quality, and urban mobility, has significantly affected CO₂ emissions. European Union countries and the United Kingdom experienced sudden decreases of 10.66% and 4.36% in their CO₂ emissions in 2020 and 2021, respectively (Sahraei and Ziaei, 2022; Adebayo et al., 2022).

As of 24:00 BST on June 10, 2020, China reported a cumulative total of 83,057 confirmed cases and a cumulative total of 4634 deaths (Xu and Cui, 2020). To quickly control the spread of the virus caused by population movement, all provinces in China launched a first-level response in late January 2020 and implemented the strictest lockdown measures. As China implemented a first-level public health emergency response to the COVID-19 pandemic, human activities across the country decreased significantly, and nearly all economic activities were suspended. These measures effectively controlled the spread of the virus but also had an impact on the economy. However, it is evident that the lockdown measures did have a positive impact on air quality/CO₂ emissions in the short term. Lu et al. (2021) found that the control of CO₂ in southern Chinese cities was significantly stronger than that in northern cities. In the meantime, compared to megacities, small and medium-sized cities had similar effects on the control of nitrogen dioxide and sulfur dioxide but a greater effect on PM_2.5 and PM₁₀. From January to March 2020, the concentration of NO₂ in Hubei Province continued to decrease, and the concentration of PM_2.5 decreased from 49.22 mg/cubic meter to 44.34 mg/cubic meter. The concentrations of CO, SO₂, O₃, PM₁₀, and AQI indices all decreased significantly during the lockdown period, but all air pollutants rebounded after the lockdown ended (Taoa et al., 2021). Nsabimana and Foday Jr (2020) believed that the decrease in coal consumption in China during the 63-day lockdown period after the Spring Festival resulted in 3458.175 tons of SO₂ and 1745.33 tons of CO₂ not being emitted into the atmosphere. The mandatory suspension of social activities inevitably had an impact on air quality and could lead to sudden changes in CO₂ and air pollutant emissions.

Previous studies only focus on the changes in air pollutants or CO₂ emissions in some areas or cities caused by a certain measure/event, ignoring the long-term effects after the measure was lifted, which may result in the taken-for-granted bias that the measure will definitely improve air quality (Shi et al., 2016; Li et al., 2017; Chen et al., 2018). The regression discontinuity in time (RDiT) method, as a classical quasi-experimental approach, allows for the ex-post evaluation of the long-term impacts of sudden events in the absence of a control group. This method has been widely used to assess the effectiveness of pollution control measures during large conferences or events in China (Li et al., 2017; Shi et al., 2016), as well as to evaluate the effectiveness of air quality alert programs (Chen et al., 2018; Neidell, 2010). Li et al. (2017) employed a quasi-experimental design based on pollution control projects during the APEC meetings and Victory Day Parade, utilizing the regression discontinuity in time (RDiT) method to assess the effectiveness of air pollution regulation in China. Shi et al. (2016) utilized difference-in-differences (DiD) and regression discontinuity (RD) designs to examine the impacts of temporary strengthened pollution regulation during the annual plenary sessions of the National People's Congress and the National Committee of the Chinese People's Political Consultative Conference (referred to as the “Two Sessions”) on air quality. Zeng and Wang (2022) found that the COVID-19 pandemic and the resulting lockdown measures have led to reduced traffic and industrial activity, which has resulted in improved air quality in some regions.

In addition, numerous analyses of the co-effects of air pollutants and CO₂ have been conducted in the past, exploring the potential of synergistic emission reduction from various perspectives such as heavy polluting industries, urban transportation, and regulatory policies (Wang et al., 2016; Zeng et al., 2017; Zhang et al., 2022). The transition to low-carbon alternatives in the power and heating sectors, such as the “coal-to-electricity” project, has made significant contributions to the coordinated control of CO₂ and regional air pollutants. The introduction of new energy vehicles and upgraded fuels has also shown positive synergistic effects on urban air pollution control (Gao et al., 2014; Alimujiang and Jiang, 2020). Other studies have investigated the effectiveness of national or provincial measures, such as energy reform, environmental policies, and import/export policies, in achieving coordinated control (He et al., 2010; Dong et al., 2015; Scovronick et al., 2019; Shao et al., 2020; Zhang et al., 2022). However, there is limited research available specifically on the synergistic reduction effects of COVID-19 pandemic control measures on AQ and CO₂ emissions.

Therefore, this study proposed a methodological framework for evaluating the impact of COVID-19 lockdown measures on AQ and CO₂ emissions. This paper adopts the regression discontinuity in time (RDiT) method and a co-effect control coordinate system evaluation method to give more reliable estimates of the causal effects between lockdown measures and air quality (AQ), taking 324 cities in mainland China as the study population and considering the changes in meteorological conditions and temporal trends. The RDiT design is employed to analyse and compare the temporal and spatial characteristics of AQ and CO₂ variations during different periods of lockdown measures in China. Subsequently, by utilizing the treatment effects of the lockdown measures, a co-effect control coordinate system is established to facilitate the assessment of the synergistic emission reduction effects of AQ and CO₂. This study will help policy-makers explore the upper limit potential for air pollution control and CO₂ reduction and recognize the effects of different socioeconomic activity intensities on each air pollutant and CO₂ emissions.

2. Materials and methods

2.1. Study area

Concerned about hazy weather and global climate change, China promulgated and implemented the “Ambient Air Quality Standards” in 2012. Through measures such as promoting clean energy, controlling coal combustion, and limiting automobile traffic, the average daily CO₂ emissions of Chinese provinces in 2020 are 0.81 MT/day, as shown in Fig. 1 , and the concentration of fine particulate matter (PM_2.5) is 33 μg/m³, down 8.3% year-on-year (CNEMC, 2021; Liu et al., 2020). CO₂ emissions per unit of GDP decreased by 1.0% compared to 2019 and by 18.8% compared to 2015, exceeding the target of an 18% reduction in the 13th Five-Year Plan (MEE, 2020). Future economic trends in China are conducive to achieving carbon neutrality, and creating multifaceted development opportunities for environmental protection. China's population has entered a low growth phase. The average annual growth rate of China's population in recent years has been less than 0.2%, significantly lower than the 0.7% growth rate from 1991 to 2019, and will enter a zero-growth phase for some time to come, which will drive down the growth rate of total population energy consumption and so on. In addition, China's economic growth rate is significantly lower. In 2019, China's average annual growth rate was approximately 5%, significantly lower than the 1991–2019 growth rate of 9.5% (Hu, 2021). By actively pushing forwards the reform of the energy consumption structure and promoting the energy technology revolution, China is expected to significantly reduce energy consumption per unit of GDP to reach or fall below the world average.

Fig. 1.

National average air pollutant concentrations and CO₂ emissions by province in 2020.

2.2. Methodology

2.2.1. Methodological framework

As an unprecedented global event in 2020, the COVID-19 pandemic presents a unique window of opportunity for investigating alterations in air quality. This study aims to explore the variations in air pollutant and CO₂ emissions in China during lockdown measures, ascertain the impact of socioeconomic stagnation on air quality, and propose recommendations for enhancing and refining current environmental policies based on the analysis of air quality changes in correlation with the implementation of lockdown measures. Furthermore, this research contributes to China's response to the call for global low-carbon development by reducing carbon emissions and supporting the nation's goals of reaching peak carbon emissions by 2030 and carbon neutrality by 2060. Previous studies have solely analysed air pollutants or CO₂ emissions during lockdown periods, neglecting a comprehensive assessment of the synergistic relationship between the two and the subsequent effects following policy relaxation. To address this challenge, this paper presents a methodological framework for holistically evaluating AQ and CO₂ changes during lockdown measures (Fig. 2 ).

Fig. 2.

The method framework for this study.

Initially, the study employs the regression discontinuity in time (RDiT) method to analyse the temporal and spatial characteristics of AQ and CO₂ changes in China under lockdown measures. By considering time trends, the lagged effects of climate on AQ, and unobservable variables, the treatment effects of lockdown measures on various cities are analysed during each observation period. Comparative analyses of multiple observation periods reveal the temporal trends of treatment effects, while multilevel geographic divisions enable horizontal comparisons of treatment effects across different regions, thereby determining the spatial distribution of treatment effects. Subsequently, we establish a co-effect control coordinate system using the treatment effect of AQ and CO₂ during each observation period, facilitating the assessment of the synergistic emission reduction effects of policies on air pollutants and CO₂. Ultimately, based on the performance of air pollution control efforts during distinct observation periods, this paper presents policy implications for nationwide AQ and CO₂ management.

2.2.2. Regression discontinuity design

In accordance with the methodology of prior research (Auffhammer and Kellogg, 2011; Davis, 2008; Viard and Fu, 2015; Zeng and Wang, 2022), a regression discontinuity in time (RDiT) model was utilized to estimate the causal effects of lockdown measures on AQ (PM_2.5, SO₂, and NO₂) and CO₂, as shown below.

$$\ln \left(p_t\right)={\beta }_0+{\beta }_{1}F{C}_t+W_t\cdot {\mathrm{\Gamma }}_t+f\left(Dat{e}_t\right)+{\epsilon }_t$$ (1)

where $p_t$ refers to the daily average pollutant concentrations or CO₂ emissions; ${\Gamma }_t$ is a vector of different variables, including time dummies and meteorological variables; and $W_t{\Gamma }_t$ is used to absorb the impact of the meteorological condition variation. $Dat{e}_t$ represents the date; $f\left(Dat{e}_t\right)$ is a Chebyshev polynomial in time; ${\epsilon }_t$ denotes the unobservable factors; $F{C}_t$ is the variable of interest, which is the treatment dummies of the lockdown measures set according to the rules below:

$$F{C}_t=\{\begin{array}{cc}0& beforeorafterthelockdownmethod\\ 1& duringthelockdownmethod\end{array}$$ (2)

More specifically, meteorological variables include the polynomials of daily average air temperature, daily average dew point temperature, daily sea level pressure, daily average wind speed rate, wind direction, daily liquid precipitation depth dimension and their lags. Season effect and $f\left(Dat{e}_t\right)$ are used to control the time-fixed effects and trends. Based on the results of previous studies (Auffhammer and Kellogg, 2011; Li et al., 2017), we used the 8th-order polynomials for $f\left(Dat{e}_t\right)$. ${\epsilon }_t$ is clustered at the season-of-year levels.

After controlling for weather variables and time trends, the treatment process needs to be based on the following assumptions:

$$E\left[F{C}_t\cdot {\epsilon }_t|F{C}_t,W_t,f\left(Dat{e}_t\right)\right]=0$$ (3)

This assumption is sound because, as of January 25th, all 31 provinces in the country have implemented a first-level public health emergency response. In this emergency situation, localities have implemented measures to restrict or cease fairs, rallies, performances, and other crowd-gathering activities, as well as work, business, and classes. These closure measures are mandatory and have been broadly implemented nationwide in accordance with the National Emergency Response Plan for Public Health Emergencies. As a result of the sudden shift in social and economic activities, there has been a substantial decrease in air pollutant indicators and CO₂, and this decrease has had a significant impact on AQ and CO₂.

Based on Zeng and Wang (2022), we set three assessment periods in 2020, from January 24 to February 29, March 1 to March 31, and April 1 to April 30, to assess the impact of the lockdown measures on AQ and CO₂ changes in China in 2020. A total of 324 prefecture-level cities in mainland China (Table S1), excluding the Tibet Autonomous Region and Hainan Province, were assessed to obtain an estimate for each city of ${\hat{\beta }}_1$, the percentage impact of a pandemic on the air pollutant concentration or CO₂ emissions. ${\hat{\beta }}_1$ is the chance effect of interest expressed as a percentage and is also referred to as the treatment effect. Negative or positive values of ${\hat{\beta }}_1$ indicate an improvement or deterioration in air quality, respectively.

2.2.3. Co-effects assessment methods

The co-effect control coordinate system evaluation method was used to analyse the degree of synergy between air pollutant and greenhouse gas emission reductions under various scenarios of epidemic control measures (Wang et al., 2016). The co-control effect coordinate system pertains to the utilization of diverse coordinate points within a two-dimensional or multidimensional spatial coordinate system to indicate the effect of the lockdown measure on different pollutants and greenhouse gases, as well as the “synergistic” state between them.

Upon obtaining the treatment efficacy for various air pollutants from the aforementioned RDiT model, this study regards the three research periods as distinct remedial strategies, employing the co-effect control coordinate system to estimate the collaborative emission reduction impact of these policies on atmospheric contaminants (Zhang et al., 2022). The emission reduction effect of a policy is defined as the treatment effect of lockdown measures on air pollutant concentration indices within the observed period.

$$E^{\left(i,j\right)}=-{{\hat{\beta }}_1}^{\left(i,j\right)}$$ (4)

where $E^{\left(i,j\right)}$ denotes the abatement effect of the policy i on the air pollutant j, and ${{\hat{\beta }}_1}^{\left(i,j\right)}$ denotes the treatment effect of the air pollutant j in the observation period i. If $E^{\left(i,j\right)}>0$, the i pollutant's emission is reduced; if $E^{\left(i,j\right)}<0$, the i pollutant's emission is increased; if $E^{\left(i,j\right)}=0$, there is no change in emission.

The co-control effect coordinate system can show the emission reduction of AQ and CO₂ at the same time. As illustrated in Fig. 3 , the two-dimensional coordinate system comprises the horizontal and vertical coordinates. The horizontal coordinate represents the reduction effect on greenhouse gases under a given control measure scenario, while the vertical coordinate indicates the reduction effect on air pollutants under the same scenario. The visualization of the horizontal and vertical coordinates of the points effectively demonstrates the co-effect of the measure on the reduction of both greenhouse gases and air pollutants.

Fig. 3.

Coordinate system for performance comparison of pollutant co-control effects.

The first quadrant in the coordinate system denotes that the control measure can simultaneously reduce both pollutants and greenhouse gases. The second quadrant indicates that the control measure can decrease pollutants, but it may increase CO₂ emissions. The third quadrant shows that the control measure can simultaneously increase pollutants and CO₂ emissions. Finally, the fourth quadrant shows that the control measure can reduce CO₂ emissions but simultaneously increase pollutants. As shown in Fig. 3, in the first quadrant, the larger the angle formed between the connecting line of a point and the origin, the better the reduction effect of the control measure on air pollutants when the measure exhibits the same reduction effect on CO₂ emissions. (Point B has a better co-effect than Point C, and the co-effects of the measures represented by points A and C are positively correlated.) When the connecting line forms the same angle as the abscissa, the farther the distance from the origin, the better the reduction effect of the control measure on both pollutants and CO₂ emissions. (Point A is better than Point C.)

2.3. Data sources

The daily average data of the air pollutant concentrations for 324 cities in China in 2020 were obtained from the daily air quality monitoring data provided by the China National Environmental Monitoring Center (CNEMC, http://www.cnemc.cn/). Hourly meteorological data were obtained from the public FTP server of the National Climatic Data Center (NCDC, https://www.ncei.noaa.gov), and the nearest meteorological station was selected as the representative station in each city's monitoring site, according to the “Basic Meteorological Element Observation Data for China Surface Meteorological Stations” provided by the China Meteorological Data Network (http://data.cma.cn/). The daily CO₂ emission data for 324 cities in China were obtained from the provincial carbon accounting emission inventory of Liu et al. (2020) (https://arxiv.org/abs/2004.13614) and the China City Greenhouse Gas Working Group (CCG, https://wxccg.cityghg.com/) dataset of CO₂ emissions for cities. Table S2 presents summary statistics of the variables involved for 324 cities from January 1st, 2020, to December 31st, 2020.

3. Results

3.1. Temporal characteristics of the impact of lockdown measures on AQ and CO₂

The changes in AQ (PM_2.5, SO₂ and NO₂) and CO₂ emissions in 2020 were compared to the same period in the previous year (Fig. 4 ). Specifically, the evaluation periods of January 24 to February 29, March 1 to March 31, and April 1 to April 30 are denoted as Feb, Mar, and Apr, respectively. The urban lockdowns implemented in China have had a considerable short-term impact on the air quality of cities, resulting in significant reductions in air pollutant concentrations and CO₂ emissions. In the Feb part, NO₂ obtained a decrease of about 15 μg/m³, and CO₂ emissions decreased by 27.4% on average. However, as the policy was gradually lifted and social activities resumed, abrupt measures such as city closures had an adverse effect on air quality after March. The indicators of each air pollutant rebounded, and even in April, they were frequently higher than those in the same period in 2019.

Fig. 4.

Temporal variations in AQ and CO₂ emissions in the three observation periods of 2019 and 2020

Notation: The green shade indicates a reduction in the concentration or emissions of that air pollutant in 2020 compared to 2019, while red indicates an increase. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

The overall treatment effects of AQ and CO₂ in 324 cities during different evaluation periods after removing outliers are shown in Fig. 5 . The urban lockdowns implemented in China have had a considerable short-term impact on the AQ of cities, resulting in significant reductions in air pollutant concentrations and CO₂ emissions. However, as the policy was gradually lifted and social activities resumed, abrupt measures such as city closures had an adverse effect on AQ after March.

Fig. 5.

Temporal treatment effects of AQ and CO₂ emissions in 324 cities

Notation: The solid black line represents the median treatment effect of the studied cities, while the dashed line indicates the quartiles of the treatment effect of all cities. Specifically, the evaluation periods of January 24 to February 29, March 1 to March 31, and April 1 to April 30 are denoted as Feb, Mar, and Apr, respectively. The subsequent figure is identical to this one.

From January 24 to February 29, the percentages of cities with negative treatment effects on PM_2.5, SO₂, NO₂ and CO₂ were 39.20%, 70.99, 84.6% and 99.38%, respectively. It was 33.85%, 43.85%, 77.2% and 66.98% in March and 35.55%, 40.7%, 40.33% and 38.63% in April, respectively. Meanwhile, the median line of treatment effectiveness of the lockdown measures for all types of pollutants showed an upwards trend over time. Similarly, the absolute value of the treatment effect was considered significant at 10% or more. The percentages of cities with significant negative treatment effects for PM_2.5, SO₂, NO₂ and CO₂ were 25.00%, 26.23%, 62.04% and 99.07% between January 24 and February 29, 26.77%, 28.57%, 24.76% and 47.53% in March, and 26.07%, 26.44%, 26.63% and 16.67% in April, respectively.

Among the 29 important cities included in the study (comprising 21 provincial capitals, 4 autonomous region capitals, and 4 municipalities), the average treatment effects on air pollutant concentrations decreased over time. From January 24 to February 29, the treatment effects for PM_2.5, SO₂, NO₂, and CO₂ were −3.72%, −9.05%, −17.4% and −24.15%, respectively; 17.16%, 0.8%, −1.98% and −10.74%, respectively, in March; and 31.71%, 2.28%, −3.67% and −3.12%, respectively, in April.

The average treatment effects and their 95% confidence intervals (CI) for CO₂ emissions and air pollutant values for the 324 cities are shown in Fig. 6 . From January 24 to February 29, the average treatment effects for CO₂, SO₂, and NO₂ were negative, while the treatment effects for SO₂ and NO₂ shifted to positive values in March and April, and the CO₂ treatment effects remained negative. Compared to the scenario without an outbreak, the lockdown measures resulted in a 5.6%, 16.6%, and 25.1% reduction in SO₂, NO₂, and CO₂ nationwide concentrations, respectively. Notably, only the treatment effect of PM_2.5 was positive in all three observation periods, with the highest value of 21.87% in March. Furthermore, the mean values of the treatment effects of the lockdown measures for all pollutants showed an upwards trend over time.

Fig. 6.

Average treatment effects with 95% CI of AQ and CO₂ emissions in 324 cities.

The aforementioned findings indicate that the air quality improvements from China's lockdown measures occurred mainly from January 24 to February 29, when the values of all pollutants except PM_2.5 declined. Commencing in March, the treatment effects of lockdown measures on air pollutants experienced a rebound. In comparison to January 24th through February 29th, the proportion of cities with negative treatment effects for CO₂, SO₂, and NO₂ diminished, while the proportion of cities with markedly negative treatment effects for CO₂ and NO₂ also dropped substantially. This suggests that although air quality improvements persisted in March, the enhancement effect had weakened compared to January 24th through February 29th.

3.2. Spatial characteristics of the impact of lockdown measures on AQ and CO₂

From February 21, 2020, China successively lowered the response level of major public health emergencies in each province, and there was strong spatial heterogeneity in the effects of control measures on each observation site due to the different economic and social development in each region. Based on the above analysis of the temporal trends of treatment effects, this paper concludes that January 24 to February 29 is the most important period for assessing the impact of lockdown measures on AQ and CO₂. During this period, the treatment effects were individually assessed for 324 cities, and the results were arithmetically averaged according to provincial divisions, as shown in Fig. 7 .

Fig. 7.

Average treatment effectiveness of AQ and CO₂ in each province

Notation: The 31st parallel passes through Shanghai, Jiangsu Province, Anhui Province, Hubei Province, Chongqing Municipality, Sichuan Province and the Tibet Autonomous Region of China.

The daily average CO₂ emissions and daily average NO₂ indicators show obvious north‒south heterogeneity, and the significant treatment effects are concentrated in southeastern China. The southern part of the comparison line contributes 64.55% of the national GDP and 10.57% of the national average GDP per capita and is the dividing line of economic growth and urban development, called the “Yangtze River defence line”, with the provinces passing through 31° north latitude as the boundary. As shown in Fig. 7, from January 24 to February 29, the provinces with CO₂ and NO₂ decreases of more than 30% are mainly concentrated in the southern part of the “Yangtze River Line”. This indicates that the control measures cause better treatment effects in areas with higher levels of economic and social development, i.e., lower air pollutant concentrations and lower CO₂ emissions.

The spatial heterogeneity of the treatment effect coefficients at the regional level is discussed, as shown in Table 1 , according to the division of urban agglomerations proposed in the 13th Five-Year Plan of China (Table S3). Positive values represent the positive effect of the city lockdown measures on AQ and CO₂ management, i.e., reduced AQ and CO₂ emissions.

Table 1.

Statistics on the effectiveness of AQ and CO₂ treatment by province.

Region	Statistical indicators	CO2	PM2.5	SO2	NO2
Pearl River Delta	Mean	0.40	0.13	0.03	0.37
	Standard deviation	0.00	−0.08	−0.06	−0.09
	Max	0.39	0.00	−0.11	0.18
	Min	0.40	0.26	0.12	0.54
Yangtze River Delta	Mean	0.30	−0.23	0.02	0.31
	Standard deviation	−0.07	−0.14	−0.05	−0.20
	Max	0.13	−0.45	−0.08	0.93
	Min	0.41	0.11	0.16	−0.06
BTH	Mean	0.20	0.00	0.08	0.08
	Standard deviation	−0.04	−0.29	−0.08	−0.07
	Max	0.09	−0.40	−0.03	−0.07
	Min	0.23	0.44	0.19	0.22
Middle reach of Yangtze	Mean	0.25	−0.09	0.09	0.32
	Standard deviation	−0.09	−0.10	−0.15	−0.18
	Max	0.13	−0.30	−0.04	0.11
	Min	0.34	0.06	0.63	0.82
Central Plain Area	Mean	0.22	−0.18	−0.01	0.15
	Standard deviation	−0.03	−0.17	−0.07	−0.11
	Max	0.16	−0.40	−0.10	−0.22
	Min	0.28	0.14	0.15	0.35
Cheng-Yu region	Mean	0.35	−0.14	0.06	0.18
	Standard deviation	−0.02	−0.11	−0.08	−0.10
	Max	0.35	−0.37	−0.02	0.35
	Min	0.42	0.07	0.27	0.15
Guanzhong region	Mean	0.18	−0.01	0.05	0.12
	Standard deviation	−0.01	−0.12	−0.07	−0.08
	Max	0.17	−0.20	−0.05	−0.04
	Min	0.20	0.30	0.21	0.22
West Coast of the Strait	Mean	0.34	0.05	0.09	0.46
	Standard deviation	−0.07	−0.16	−0.14	−0.23
	Max	0.11	−0.27	−0.05	0.03
	Min	0.41	0.35	0.63	0.93

In terms of CO₂, among the eight urban agglomerations of interest in this paper, the Pearl River Delta and the Cheng-Yu region have the largest decreases of 40.05% and 35.37%, respectively, and the West Coast of the Strait and the Yangtze River Delta also have treatment effects of more than 30%. However, the Guanzhong region has the smallest reduction rate of 18.41%. In terms of NO₂, the West Coast of the Strait and the Pearl River Delta had the largest reductions of 46.18% and 37.01%, respectively, and the Middle Reach of Yangtze and the Yangtze River Delta had more consistent improvements of 32.42% and 31.48%, respectively. For SO₂, the treatment effect of all eight urban agglomerations was below 10%, of which only the Central Plain area was positive, with the lockdown measure causing its SO₂ concentration to rise by 0.91%. For PM_2.5, three urban agglomerations, the Pearl River Delta, the West Coast of the Strait and BTH (Beijing-Tianjin-Hebei), obtained an improvement in PM_2.5 concentration through the lockdown measures. However, all other urban agglomerations had different degrees of upwards PM_2.5 concentration, especially the Yangtze River Delta, which reached 22.98%.

3.3. The impact of COVID-19 on the average AQ and CO₂ at national and regional levels

Since the State Council issued the Action Plan for Air Pollution Prevention and Control, China has included air pollution control and CO₂ emission intensity reduction targets in the government's target responsibility assessment. Based on the above analysis of the temporal and spatial characteristics of the impact of confinement measures on air quality, this paper estimates the impact of COVID-19 on the annual average values of air pollutants for 324 cities. Table 2 lists the ten provinces with the greatest impact of COVID-19 on annual average concentrations of CO₂ emissions in 2020. When the coefficient of impact is positive, the lockdown measures have demonstrated an effect of reducing both AQ and CO₂ emissions in the respective region. The full results are presented in Table S4.

Table 2.

Impact of COVID-19 on annual average concentrations of various pollutants in 2020.

Level	Region	CO2	PM2.5	SO2	NO2
National level	All cities	0.241	−0.007	0.058	0.158
Regional level	Chongqing	0.421	−0.006	0.046	0.309
	Guangdong	0.403	0.094	0.017	0.336
	Guizhou	0.364	−0.033	0.074	0.323
	Yunnan	0.360	−0.116	0.042	0.166
	Fujian	0.348	0.053	0.068	0.388
	Zhejiang	0.348	−0.105	0.024	0.447
	Sichuan	0.347	−0.124	0.052	0.152
	Hunan	0.327	−0.071	0.103	0.312
	Jiangxi	0.309	0.021	0.200	0.485
	Shanghai	0.289	−0.263	−0.014	0.113

Compared to the scenario without an outbreak, the lockdown measures resulted in a 5.8%, 15.8%, and 24.1% reduction in SO₂, NO₂, and CO₂, respectively, as well as a 0.7% growth in PM_2.5 nationwide. This means that PM_2.5 should have decreased in the no-epidemic scenario; however, the lockdown measures prevented it from decreasing. Considering the corresponding actual values, it is posited that in the absence of lockdown measures, PM_2.5, SO₂, NO₂, and CO₂ average levels should have been 33.4, 11.0, 29.5 and 0.1 Mt per day, respectively, nationwide (Fig. 8 ).

Fig. 8.

The impact of COVID-19 on AQ and CO₂ at national levels

Notation: Unit is 10² Mt for CO₂ and ug/m³ for AQ.

Correspondingly, the COVID-19 pandemic precipitated a reduction of 3.72%, 9.05%, 17.4%, and 24.15% in PM_2.5, SO₂, NO₂, and CO₂ levels, respectively, across 29 pivotal metropolises. These urban centers, characterized by heightened socioeconomic activity, dense populations, and well-developed transportation and industrial pollution sources, experienced a pronounced and conspicuous impact on air quality amidst the crisis. In the absence of lockdown measures, the average daily levels of PM_2.5, SO₂, NO₂, and CO₂ would have been 40.2, 10.9, 41.6, and 0.2 Mt, respectively (Fig. 9 ). The confinement measures used in this epidemic helped uncover the upper limit of the potential for air pollution control and carbon emission reduction and helped recognize the impact of different socioeconomic activity intensities on each air pollutant and CO₂.

Fig. 9.

The impact of COVID-19 on AQ and CO₂ at regional levels

Notation: Unit is 10² Mt for CO₂ and ug/m³ for AQ.

3.4. Co-effects of lockdown measures on AQ and CO₂

To investigate the maximum efficiency of reaching air pollution control and CO₂ reduction, a two-dimensional coordinate system (Table 3 ) is used to identify the reduction effects on air pollutants and CO₂ for each observation period, as shown in Fig. 10 , to help in the initial screening of policies.

Table 3.

Explanation of the co-coordinate system.

Point distribution	Implication
First quadrant	Reducing both air pollutants and CO2 emissions.
Second quadrant	Reducing air pollutant emissions while increasing CO2 emissions.
Third quadrant	Increasing both air pollutants and CO2 emissions.
Forth quadrant	Reducing CO2 emissions while increasing air pollutant emissions.
Origin	No effect on air pollutants or CO2.
Positive half-axis of horizontal coordinate	Reduces CO2 emissions while having no effect on air pollutant emissions.
Negative half-axis of horizontal coordinate	Increase CO2 emissions while having no effect on air pollutant emissions.
Positive half-axis of vertical coordinate	Reduces air pollutant emissions while having no effect on CO2 emissions.
Negative half-axis of vertical coordinate	Increase air pollutant emissions while having no effect on CO2 emissions.

Fig. 10.

Co-effects of air pollutants and CO₂.

According to Fig. 10, from January 24th to February 29th, the co-effect of China's lockdown measures on CO₂ and PM_2.5 was negative, indicating that the reduction in CO₂ was accompanied by an increase in PM_2.5 concentrations. Comparing Fig. 10(b) and (c), it can be observed that these measures not only reduced CO₂ but also decreased the concentrations of SO₂ and NO₂. Furthermore, the co-effect between CO₂ and NO₂ was better than that between CO₂ and SO₂. In March, CO₂ continued to decrease under the influence of relaxed lockdown measures, accompanied by an increase in AQ concentrations. During this time, the co-effect of SO₂ was slightly better than that of NO₂, while the synergistic reduction of PM_2.5 was the least effective. In April, both CO₂ emissions and AQ concentrations rebounded due to the influence of less stringent lockdown measures and widespread resumption of work and production. However, the trends in their co-effects remained consistent with March, with the co-effect between CO₂ and PM_2.5 being the weakest, followed by NO₂, and then SO₂.

Based on the above analysis, policy-makers need to be cautious about two issues. First, the increase in PM_2.5 caused by lockdown measures requires policy-makers to seek other solutions to reduce PM_2.5 concentrations while reducing CO₂ emissions. Additionally, after the lockdown measures were gradually lifted, there was a strong rebound in air pollutant and CO₂ emissions, which had a negative impact on AQ and CO₂ improvement.

4. Discussion

4.1. Optimization of pollution reduction policies

The relationship between air quality, CO₂ emissions, and the COVID-19 pandemic is complex and multifaceted. From an environmental perspective, the improvements in air quality and CO₂ emissions during the lockdown are noteworthy. Building upon previous research using the RDiT model to study short-term air quality variations (Li et al., 2017; Zeng and Wang, 2022), this study not only examines the impact of COVID-19 lockdown measures on air pollutant concentrations in 324 Chinese cities but also incorporates the study of CO₂ emissions and the co-effects between air pollutants and CO₂. From January 24th to February 29th, the PM_2.5 treatment effect showed a small upward trend overall, but the treatment effect in provincial capitals and other key cities was negative, i.e., the blockade measures hindered PM_2.5 emissions in large cities. from November 31st, 2019 to April 31st, 2020, PM_2.5 levels decreased by 4.8 μg/m³ compared to the same period of the previous year, with the largest decrease occurring in January and February, at 18% (Li et al., 2021). Therefore, although PM_2.5 concentrations continue to decline, this paper argues that China could obtain a more significant reduction in overall PM_2.5 concentrations in 2020 without blocking measures.

This study reveals a significant decline in the number of cities exhibiting noticeable improvements in carbon emissions, dropping sharply from 99.07% at the initiation of Level-1 response to 47.53%, further decreasing to 16.67% in April. These findings align with Liu et al.'s (2021) research, which suggests that by April, apart from Hubei Province, the CO₂ emissions of all provinces approached or even surpassed the levels of the same period in 2019, with some provinces experiencing growth compared to the previous year. Additional studies indicate a resurgence in PM_2.5, NO₂, and SO₂ concentrations starting from March (Li et al., 2020; Zhang et al., 2020). Wang et al. (2020b) argue that the Level-1 public health emergency response in February led to a 36–53% reduction in NO₂ concentrations in first-tier cities, but by late April, the reduction had decreased to below 10%. Therefore, both this study and previous research indicate that the effectiveness of lockdown measures in mitigating air pollution experienced a rebound due to the gradual relaxation of measures in March and April, with more pronounced effects observed in large cities, which may be closely linked to the orderly recovery of the Chinese economy.

China was the first major economy to recover after the new coronavirus pneumonia outbreak (Wang and Zhang, 2021). On February 27, the number of new confirmed cases dropped to single digits for the first time in both provinces other than Hubei province and in the prefecture-level cities in Hubei Province other than Wuhan City. As of March 28, the average start-up rate and reinstatement rate of industrial enterprises above the national scale reached 98.6% and 89.9%, up 15.5% and 38%, respectively, from February 23 and 3.6% and 22%, respectively, from March 1. At the end of April, the reinstatement rate of industrial enterprises above the national scale exceeded 99%, the reinstatement rate of small and medium-sized enterprises reached 88.4%, and the reinstatement rate of major projects exceeded 95% (ONSC, 2021). This paper suggests that the large-scale resumption of work and production has enabled the industrial and electricity sectors, which are the main sources of CO₂ emissions, to gradually resume operations, leading to a rebound in carbon emissions and air pollutant concentrations.

This expands the scope of the study beyond air quality changes and elucidates the combined effects of large-scale lockdown measures on air pollutant concentrations and carbon emissions over a specific period. It highlights the upper limits of both air pollution reduction and CO₂ emission reduction potential, providing policymakers with more comprehensive policy recommendations. Firstly, benefiting from the increased demand for contactless social interactions caused by the pandemic, online work, telemedicine, e-commerce, and contactless services have become conventional business models. Governments should further leverage the advantages of digital platforms to reduce the intensity of social activities while ensuring economic benefits. This approach reduces transportation demand, energy consumption, and contributes to a decrease in CO₂ emissions.

Furthermore, to address air pollution and CO₂ emissions in a coordinated manner, policymakers should optimize the energy structure by gradually reducing reliance on fossil fuels in the transportation and industrial sectors and increasing the proportion of clean energy. Firstly, industrial emissions regulation should be strengthened by implementing stricter emission standards for the industrial and transportation sectors, including measures such as suspending or restricting operations of highly polluting industries, enhancing supervision and enforcement, and reducing pollutant emissions. Additionally, policy measures such as tax incentives and subsidies can encourage businesses to adopt environmentally friendly production methods, improve resource recycling rates, and reduce pollutant emissions during the production process. Secondly, there should be a strong focus on developing clean energy sources such as solar power, wind energy, and hydroelectric power to reduce dependence on fossil fuels and lower CO₂ emissions. Promoting green transportation, sustainable supply chains, and expanding investments and support for renewable energy are essential steps. Thirdly, optimizing public transportation systems and encouraging greater use of public transport to reduce private vehicle usage and decrease traffic pollution are crucial (Wang et al., 2020a; Wang et al., 2023).

Concurrently, as air pollution exhibits transboundary characteristics, regional cooperation is vital for improving AQ. Provincial governments should intensify cross-regional environmental enforcement, undertake joint inspections of enterprises and projects within the area, and prevent “jurisdictional protectionism” in environmental law enforcement. Moreover, policy-makers should establish an AQ monitoring data sharing platform and fortify pollution source surveillance and early warning systems, ultimately achieving the objective of promptly apprehending neighbouring regions' AQ conditions. Consequently, relevant provincial departments can expeditiously implement measures to counter transboundary pollution dissemination, collaboratively resolving air pollutant and CO₂ emission reduction dilemmas.

4.2. Recommendations on emergency management

The COVID-19 pandemic offers an opportunity for reflection on how to further enhance China's emergency response capabilities in preventing and dealing with major sudden events and improve the government's emergency management and handling of such events.

Firstly, the emergency management system in China needs to be improved in several aspects. One area is the insufficient role and coordination mechanisms of the emergency management department (Chen, 2020). The emergency response to the current COVID-19 outbreak was primarily led by the Health Commission, rather than being centrally coordinated by the Ministry of Emergency Management. This arrangement significantly hampers the coordination and efficiency of the emergency management system's implementation. Secondly, there has been inadequate emphasis on the comprehensive planning system throughout the emergency management process. Due to a lack of timely modification and dynamic management, some local governments' emergency plans lacked practicality and operational effectiveness when faced with the COVID-19 outbreak (Ma, 2020). Thirdly, the number and status of laws and regulations regarding emergency management do not align, and the content related to the rule of law is incomplete. The existing laws do not clearly assign specific responsibilities, hindering the fulfillment of duties and subsequent accountability across departments. Lastly, the government's handling of public opinion in emergency management has revealed issues such as excessive emphasis on propaganda, insufficient communication, inconsistent messaging, and one-sided reporting.

To ensure the orderly conduct of emergency response, proactive measures are crucial. The government should establish and strengthen a comprehensive system for national and local-level emergency material reserves, ensuring swift allocation of resources in times of crisis. Simultaneously, it is important to improve the mechanisms for selecting and training emergency personnel and enhance cooperation with international organizations and other countries in emergency management. Furthermore, it is essential to enhance the coordination between environmental protection and emergency management. Environmental factors should be taken into full consideration to safeguard environmental interests while dealing with major emergencies. This can be achieved by improving the environmental monitoring and early warning systems, enabling early detection and mitigation of environmental risks, thus reducing the likelihood of emergencies triggered by environmental factors.

The technological innovations that have emerged during this pandemic deserve further application and integration into our regular social systems. Drawing lessons from the increased demand for contactless interactions, the government should continue to promote the development of e-commerce, remote work, and contactless services, aiming to reduce societal transportation needs and energy consumption. The application of big data and artificial intelligence can contribute to accurate information dissemination, enhancing resource utilization efficiency. To address concerns regarding user privacy and security issues, the government should establish standardized and actionable regulations for data security usage. Additionally, decision-makers should fully leverage the role of universities, research institutions, and businesses as key innovators. Measures such as increasing the openness and sharing of information resources, supporting fundamental research, improving mechanisms for integrating basic research with industrialization, and fostering collaboration among various innovation entities should be implemented to facilitate the development of a technological innovation industry chain to tackle major emergencies.

From a societal perspective, public awareness and participation in emergency management should be enhanced. This can be achieved through nationwide emergency education initiatives to disseminate basic knowledge and skills for responding to emergencies, thereby improving the public's awareness and abilities for self-help and mutual assistance. Regular national emergency drills should be conducted to enhance the public's emergency response capabilities through simulated exercises. Public involvement in emergency management should be encouraged, including the spontaneous formation of community-based emergency response teams, aiming to comprehensively strengthen society's overall capacity for responding to major emergencies.

In summary, by strengthening the formulation and implementation of emergency plans, enhancing public awareness and participation in emergency management, increasing the application of technological innovation in emergency management, optimizing the legal environment for emergency management, improving the emergency resource reserve and support system, and reinforcing the synergy between environmental protection and emergency management, we can further enhance China's emergency response capabilities in preventing and dealing with major sudden events and improve the government's emergency management and handling of such events. This will help to enhance the country's overall response capabilities and risk mitigation, laying a solid foundation for sustainable development and the well-being of the people.

5. Conclusions

This study proposed a methodological framework for evaluating the impact of COVID-19 lockdown measures on AQ and CO₂ emissions. First, the study employs the RDiT method to analyse the temporal and spatial characteristics of AQ and CO₂ changes in China under lockdown measures. By considering time trends, the lagged effects of climate on AQ, and unobservable variables, the treatment effects of lockdown measures on various cities are analysed during each observation period. The study indicates that during the lockdown period from January 24 to February 29, 2020, the overall AQ and CO₂ within the research area significantly improved, while the treatment effects exhibited prominent north‒south heterogeneity. The lockdown measures effectively advanced the decline in CO₂, SO₂, and NO₂, albeit resulting in a diminished reduction in PM_2.5 compared to the hypothetical scenario without lockdown measures. In the southern portion of the “Yangtze River Defence Line” — regions with advanced economic development — treatment outcomes are more favourable. The Pearl River Delta, the western coast of the strait, and other areas experienced the most substantial decreases in CO₂ and NO₂ levels, with a marked reduction in PM_2.5 concentrations in some urban agglomerations. Nevertheless, commencing in March, the AQ and CO₂ improvement effect diminishes, and air pollutant concentrations rebound. As the pandemic is controlled and industrial and power sectors, which are major CO₂ emitters, returned to operation, resulting in an upwards trend in carbon emissions.

Furthermore, we established a co-control effect coordinate system using the treatment effect of AQ and CO₂ during each observation period, facilitating the assessment of the synergistic emission reduction effects of policies on AQ and CO₂. From the perspective of emission reduction efficiency, during the main lockdown period, although CO₂ emissions have decreased, there has been an increase in PM_2.5 concentration while SO₂ and NO₂ concentrations have decreased. Therefore, the synergistic reduction effect between CO₂ and PM_2.5 is not ideal. Ultimately, based on the performance of air pollution control efforts during distinct observation periods, this paper presents policy implications for nationwide AQ and CO₂ management. Future research could leverage the upper limit of air quality improvement potential obtained from this study to conduct long-term tracking and comparison of the effects of short-term, small-scale lockdown measures on specific cities or provinces throughout 2021–2022, thereby further eliminating lag effects and climate interference. Concurrently, by examining the socio-economic impacts of lockdown measures, we can delve into each detailed sector, such as industry, transportation, commerce, and other emissions-related entities, to uncover their respective contributions to air quality improvement. Overall, while the lockdown measures initially enhanced air quality, their effects waned over time and with the easing of restrictions, even led to inhibiting AQ and CO₂ improvements. Policy-makers should focus on reducing PM_2.5 concentrations while curbing CO₂ emissions to address these challenges.

There are some potential limitations of this study. First, the chosen breakpoint in the article may not be the optimal selection. When analysing the impact of lockdown measures on AQ and CO₂, it may be necessary to consider multiple potential breakpoints, such as before and after the implementation of lockdown measures and varying degrees of lockdown enforcement. A comprehensive analysis of multiple results should be conducted. Relying solely on a single temporal breakpoint may lead to biased conclusions. Additionally, breakpoint regression methods face challenges in capturing dynamic effects when analysing the impact of lockdown measures on AQ and CO₂. The present study only introduces lagged climate and time variables. Consequently, further research could incorporate lagged pollutant concentrations and policy measures into the model to capture the dynamic relationship of air quality indicators over time and investigate the influence of lockdown measures on air quality during different time periods.

Notes

The authors declare no competing financial interest.

CRediT authorship contribution statement

Weiqi Liang: Conceptualization, Data curation, Methodology, Formal analysis, Writing – original draft. Huihui Wang: Conceptualization, Funding acquisition, Methodology, Writing – original draft. Hanyu Xue: Data curation, Formal analysis. Yidong Chen: Data curation, Formal analysis. Yuhao Zhong: Data curation, Formal 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.

Handling Editor: Giovanni Baiocchi

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

Data will be made available on request.