Study on Collaborative Emission Reduction in Green-House and Pollutant Gas Due to COVID-19 Lockdown in China

In recent years, as China’s peaking carbon dioxide emissions and air pollution control projects have converged, scholars have begun to focus on the synergistic mechanisms of greenhouse gas and pollution gas reduction. In 2020, the unprecedented coronavirus pandemic, which led to severe nationwide blockade measures, unexpectedly provided a valuable opportunity to study the synergistic reduction in greenhouse gases and polluting gases. This paper uses a combination of NO2, O3, and CO2 column concentration products from different satellites and surface concentrations from ground-based stations to investigate potential correlations between these monitoring indicators in four Chinese representative cities. We found that XCO2 decreased in March to varying degrees in different cities. It was witnessed that the largest decrease in CO2, −1.12 ppm, occurred in Wuhan, i.e., the first epicenter of COVID-19. We also analyzed the effects of NO2 and O3 concentrations on changes in XCO2. First, in 2020, we used a top-down approach to obtain the conclusion that the change amplitude of NO2 concentration in Beijing, Shanghai, Guangzhou, and Wuhan were −24%, −18%, −4%, and −39%, respectively. Furthermore, the O3 concentration increments were 5%, 14%, 12%, and 14%. Second, we used a bottom-up approach to obtain the conclusion that the monthly averaged NO2 concentrations in Beijing, Shanghai, and Wuhan in March had the largest changes, changing to −39%, −40%, and −61%, respectively. The corresponding amounts of changes in monthly averaged O3 concentrations were −14%, −2%, and 9%. However, the largest amount of change in monthly averaged NO2 concentration in Guangzhou was found in December 2020, with a value of −40%. The change in O3 concentration was −12% in December. Finally, we analyzed the relationship of NO2 and O3 concentrations with XCO2. Moreover, the results show that the effect of NO2 concentration on XCO2 is positively correlated from the point of the satellite (R = 0.4912) and the point of the ground monitoring stations (R = 0.3928). Surprisingly, we found a positive (in satellite observations and R = 0.2391) and negative correlation (in ground monitoring stations and R = 0.3333) between XCO2 and the O3 concentrations. During the epidemic period, some scholars based on model analysis found that Wuhan’s carbon emissions decreased by 16.2% on average. Combined with satellite data, we estimate that Wuhan’s XCO2 fell by about 1.12 ppm in February. At last, the government should consider reducing XCO2 and NO2 concentration at the same time to make a synergistic reduction.


Introduction
The emergence of the coronavirus disease 2019 (COVID-19) occurred in 30 December 2019 and was declared a global pandemic by the World Health Organization on 11 March 2020 [1]. In response to the COVID-19 outbreak, the China central government had gradually implemented severe nationwide lockdown measures since the end of January 2020. These measures placed the society of China on hold, significantly reducing the emissions of pollutants. In the past decade, China had experienced several social pauses, including routine ones, such as the Lunar New Year and special events, such as the Olympic game and G20. However, 2020 is a special year because, in the first half of the year, most cities in China were blockaded, and industrial, power, and transportation activities were minimized. In the second half of the year, China's social activities have gradually returned to normal. Hence, the outbreak of COVID-19 pandemic in 2020 provides an opportunity for the analysis of the carbon cycle and compositional relationship between gases (CO 2 , NO 2 , and O 3 ).
Owing to the outbreak of COVID-19, greenhouse gas emissions have changed significantly, especially in developing countries. Before the pandemic, the concentration of carbon dioxide (XCO 2 ) increased by 2 ppm per year over the previous decade, and the global XCO 2 increased from 278 ppm (prior to the industrialization era) to 410 ppm in 2020 [2]. However, during the pandemic, Le et al. [3] found that the daily global CO 2 emissions decreased by −17% by early April 2020 relative to the mean 2019 levels. Zhu et al. [4] found an abrupt 8.8% decrease in global CO 2 emissions (−1551 Mt CO 2 ) in the first half of 2020 compared to the same period in 2019. The International Energy Agency (IEA) used monthly projections of fossil fuel energy demand and estimated a −5% decline in global CO 2 emissions from January to April 2020, compared to the same period in 2019 [5]. Moreover, the impacts of lockdown in different regions on air quality were observed [6][7][8][9][10]. Dantas et al., considered the consequences of partial lockdown on air quality and obtained the conclusion that PM10 level was reduced to a low level, and NO 2 decreased because of the lockdowns [11]. Sahin et al., explored the consequences of weather on COVID-19 pandemic and concluded that wind showed a positive correlation with COVID-19 cases [12]. Ogen et al., studied the outcome of NO 2 on COVID-19 mortality and concluded that long-term exposure to NO 2 increased fatalities because of COVID-19 [13]. Xie and Zhu explored the effect of temperature on COVID-19 transmission and found a negative relationship between temperature and increased COVID-19 transmission when the temperature was between 16.8 • C and 27.4 • C [4]. Shi et al., examined the outcome of temperature on COVID-19 transmissions and showed that temperature had a positive linear association with COVID-19 cases when the temperature was lower than 3 • C [14]. These studies provided a scientific basis for the study of the relationship between air pollutants and polluting gases and effective ideas for the prevention and control of pollutants and polluting gases.
These studies analyzed changes in trace gases, emissions, and temperature during the epidemic period in 2020 from different perspectives, but the analysis of trace gas concentration changes in 2020 is rare. Few scholars have studied the XCO 2 changes [15][16][17][18][19][20][21][22], meanwhile, the government implemented lockdowns caused by the epidemic has provided us with a good opportunity for this analysis. We have accomplished several works in this article as follows. Firstly, we analyzed changes in monthly averaged concentrations of CO 2 , NO 2 , and O 3 in Beijing, Shanghai, Wuhan, and Guangzhou. For Beijing, Shanghai, Guangzhou, and Wuhan, the results suggested that the amounts of changes in XCO 2 were 0.13, −0.99, −0.89, and −1.12 ppm in February, respectively. The concentrations of NO 2 and O 3 changed by −24% and 5%, −18% and 14%, −4% and 12%, −39% and 14%, respectively, by comparing data in February 2019 with the Sentinel-Satellite-5 data. Furthermore, for Beijing, Shanghai, Guangzhou, and Wuhan, with regard to the ground monitoring network data, the largest amounts of changes in monthly averaged concentrations of NO 2 and O 3 were −39% and 14%, −40% and −2%, −61% and 9% in February, respectively. However, for Guangzhou, it is in December, and the decrease in NO 2 and O 3 is −40% and −12%, respectively. Secondly, we have evaluated the relationship between the variation of XCO 2 and the variation of NO 2 and O 3 . We found the effect of NO 2 concentration on XCO 2 is positively correlated from the point of the satellite (R = 0.4912) and the point of the ground monitoring stations (R = 0.3928). Furthermore, we found that XCO 2 and O 3 are correlated, with a correlation coefficient of 0.2391(from satellite observations) and −0.3333 (from ground monitoring stations). Therefore, from the point of promoting the reduction in XCO 2 in atmosphere, the contribution of NO 2 concentration is greater than that of O 3 concentration. Moreover, the government should consider reducing XCO 2 and NO 2 concentration at the same time to make a synergistic reduction. The remaining parts of this work are arranged as follows. The data and method we used are described in Section 2. The main results and discussions are demonstrated in Section 3. Finally, we conclude the whole study in Section 4.

Study Area
In this paper, four locations were selected: Beijing, Shanghai, Guangzhou, and Wuhan. The main reasons are as follows. First, the Beijing, Shanghai, and Guangzhou are international cities and are located at different latitudes [10], whereas Wuhan is the first epicenter of COVID-19 around the world and is the most important city in central China in terms of economy, education, innovation, and industry; the climatic characteristics of these metropolitan areas are quite different. Therefore, these cities provide good conditions for studying changes in CO 2 , NO 2 , and O 3 concentrations amid the COVID-19 lockdown in different regions. We show the spatial location distribution of the research area in detail in Figure 1.

Remotely Sensed Products
We analyzed changes in the monthly averaged concentrations of CO 2 , NO 2 , and O 3 in 2020. For NO 2 and O 3 data, we used a Sentinel-5 Precursor offline level 3 product [23], which was obtained from the Google Earth Engine platform [24]. The Sentinel-5 Precursor is a satellite launched on 13 October 2017 by the European Space Agency to monitor air pollution. Owing to the incomplete data of O 3 and NO 2 in 2018, we used data from January 2019 to December 2020.
As for CO 2 data, the GOSAT_FTS_L3_V2.95 data of bias-corrected [25] from January 2016 to December 2020 were selected. The Gosat and Gosat-2 satellites have been in orbit since 2009 and 2018, respectively, and their global coverages holds the potential to reveal new information about the carbon cycle by top-down atmospheric inversion methods combined with column-average CO 2 retrievals. The GOSAT satellite embarks the Thermal and Near-infrared Sensor for carbon Observation (TANSO) to characterize the column abundances of CO 2 and CH 4 . Six of the seven data channels of this sensor operate in the near-infrared part of the solar spectrum, so this device cannot be used whenever the planet reflects little solar light, i.e., in polar regions during winter. For additional information on that instrument, please consult this website [26]. Besides, a suite of passive remote sensing satellites can measure the reflected sunlight spectra in the infrared region over the globe. These measured spectra are used to retrieve the column-averaged CO 2 concentration (XCO 2 ) (e.g., GOSAT, GOSAT-2, OCO-2, OCO-3, and TanSat). However, the interference of clouds and aerosols in XCO 2 retrieval often results in sparse spatial and temporal coverage. Therefore, there is no data for these latitudinal regions in China, namely latitudes higher than 48.7 • N in December, higher than 51.2 • N in November and higher than 51.3 • N in January.

Ground Monitoring Data
To further explore the relationship of changes in monthly averaged NO 2 and O 3 concentrations with XCO 2 , a bottom-up approach was adopted for analysis, and ground monitoring data were used. The China Air Quality Network includes daily averaged gas concentration changes in China [27]. Hence, along with remotely sensed data, the ground monitoring data were also included in this work. In addition, to evaluate the accuracy of the monthly averaged CO 2 concentration data from our algorithm, we used the TCCON (Hefei site) data. The Total Carbon Column Observing Network (TCCON) [28,29] is a network of ground-based Fourier transform spectrometers that record direct solar spectra in the near-infrared spectral region. From these spectra, accurate and precise column-averaged abundances of CO 2 are retrieved and reported. For the period of this data and the purpose for which the data are used, we explain in detail in Table 1.

Analysis Method
This paper mainly adopted a comparative approach for analyzing the impact of the COVID-19 lockdown on the atmospheric environment in 2020. Hence, our work mainly focused on the following four points in detail. Firstly, given the bias-corrected GOSAT_FTS_L3_V2.95 productions are discrete point data, we used the empirical bayesian kriging (EBK) interpolation theory to fill the data gaps in the study area [30]. Additionally, the EBK method is a ground statistical interpolation method that automatically performs the most difficult steps in building an effective kriging model [31][32][33]. The EBK method differs from other kriging methods in illustrating the error introduced by estimating a basic semivariogram. Other kriging methods underestimate the standard error of prediction because they do not consider the uncertainty involved in this semivariogram estimation [33]. For NO 2 and O 3 planar product data, we directly used it. Second, for GOSAT and GOSAT-2 satellites and ground monitoring data in 2016, 2017, 2018, and 2019, we obtained standard curves using locally weighted regression (LWR). The LWR is a nonparametric method. In each prediction of new samples, the adjacent data will be retrained to obtain new parameter values, to avoid under-fitting and reduce the interference of distant data effectively. Third, we analyzed change in XCO 2 by comparing the standard curve data in 2020. To analyze the concentration influence of NO 2 and O 3 on the change in CO 2 concentration, we used the top-down method and bottom-up method, respectively. The investigation of local processes, construction of models, and extrapolation with spatial data to larger scales has been called the "bottom-up" approach. The "top-down" approach aims to obtain data values over a large area in a short period of time without knowing the underlying details of the data caused by subtle scale changes. Therefore, the top-down and bottom-up approaches are thus complementary, and we comprehensively analyzed the influences of NO 2 and O 3 concentrations on change in XCO 2 with satellite data and ground monitoring data in a "multiple constraint" approach.

Spatial Distribution of Remotely Sensed CO 2 Concentrations
GOSAT and GOSAT-2 provide passive inversion data and are particularly vulnerable to cloud, aerosol, latitude, and dark night constraints, which cause loss of satellite observation data. We utilized GOSAT_FTS_L3_V2.95 productions, as described in Section 2, and completed the conversion of the GOSAT and GOSAT-2 data points to the surfaces by adopting empirical bayesian kriging (EBK) theory. The data of vacant areas in the study area were filled. Figure 2 shows the interpolated data by using only EBK theory in 2020. Given that the atmospheric environment in China is complex and GOSAT and GOSAT-2 are particularly sensitive to aerosols and clouds, the data gaps in eastern and central China were large. Moreover, the Hainan and Taiwan Provinces as well as the South China Sea are not marked on the local maps in Figure 2. Fortunately, in this case, we were able to use the EBK theory to complete the monthly averaged spatial interpolation on the 0.25 • grid.
The Total Carbon Column Observing Network (TCCON) is a network of ground-based Fourier transform spectrometers recording direct solar spectra in the near-infrared spectral region [26]. From these spectra, accurate and precise column-averaged abundances of CO 2 are retrieved and reported. In addition, we evaluated the accuracy by combining TCCON site (Hefei [29]) data in 2016 (as shown in Figure 3), and the coefficient of determination (R 2 ) and root-mean-square error (RMSE) were 0.915 and 0.654, respectively.

Analysis of Changes in XCO 2
Multiple pixels in the sub-study areas were found because the spatial resolution of the products after interpolation was 0.25 • . To evaluate the influence of COVID-19 on the XCO 2 of each month in 2020, we matched the five-year XCO 2 curve to a scale from 2016 to 2020. First, we set the 2019 XCO 2 data as a baseline and then unified the 2016, 2017, 2018, and 2020 XCO 2 to the 2019 baseline by revising the annual XCO 2 curve increments. This increment was revised using the annual XCO 2 curve increments at Waligura stations from global atmosphere watch in China [34]. By performing local weighted regression on all grid XCO 2 changes in the current sub-study, the grid scale XCO 2 changes were converted to city scale. Because cities vary in development and geographical location, we added a city increment for each city relate to the annual XCO 2 curve increments. We use the standard deviation as the city increment, which is the difference between the XCO 2 for each month in 2020 and the trend line XCO 2 . We studied the change in the monthly averaged XCO 2 of the city. The image resolution was 0.25 degree after interpolation, and the average amount of change in several pixels contained in each city was obtained.
Then, we tested the revised XCO 2 data in 2020 using the Kruskal-Wallis test. The Kruskal-Wallis test is a nonparametric test of three or more datasets. It is used to test the consistency of the null hypothesis and its alternative hypothesis about differences between at least two samples of a population function distribution. A standard is usually used in evaluating significance. The P value is 0.05. If the P value is less than or equal to 0.05, difference between two samples are considered statistically significant; otherwise, they are considered insignificant. A significant difference between the two samples indicated that the XCO 2 data in 2020 and local weighted regression XCO 2 from 2016 to 2019 were independent. That is, when the P value in Figure 4 is less than 0.05, it indicates that the prevention and control measures during this epidemic period caused this change. For the period from February to May in Wuhan, the P value was less than 0.05. For the period from March to April in Beijing, the P value is also less than 0.05, and for Shanghai and Guangzhou in February, the P value is also less than 0.05. The above results show that the monthly averaged XCO 2 in 2020 changed relative to the trend line data; the variation in XCO 2 concentration is shown in Figure 5.  Due to the impact of COVID-19 in Beijing, Shanghai, Guangzhou, and Wuhan, Figure 5 shows the monthly averaged XCO 2 changes values in 2020. The Chinese government has taken lockdown measures in most Chinese cities from the end of January to early April. Therefore, the XCO 2 fluctuates relatively sharply during the period. We found the amounts of changes in XCO 2 were 0.13, −0.99, −0.90, and −1.12 ppm in February, for Beijing, Shanghai, Guangzhou, and Wuhan. Then XCO 2 continued to decrease from March to May in Beijing. In Shanghai and Guangzhou, both international cities, XCO 2 gradually returned to normal levels in April and March, respectively.
Notably, the trend of XCO 2 in June, July, and August varied between the coastal cities (Shanghai and Guangzhou) and the inland cities (Beijing and Wuhan). Furthermore, this difference is a combination of effects resulting from seasonal impact weights and the corresponding intensification of industrial emissions. The weight of the seasonal impact of XCO 2 gradually increased in June, July, and August. The weight of seasonal influence represents the enhancement of CO 2 absorption capacity from surface vegetation. Besides, by analyzing the change in NO 2 concentration from June to August in Figure 6, we found that the corresponding industrial emissions of Beijing and Wuhan gradually increased in the three months. Therefore, under the combined action of these two factors, the amounts of changes in XCO 2 in June, July, and August in Beijing were 1.54, −0.92, and 0.90 ppm, respectively. In addition, the amounts of changes in XCO 2 in Wuhan were 1.30, −0.70, and 1.1 ppm. For coastal cities of Shanghai and Guangzhou, there are some differences in the change in CO 2 concentration. Although Shanghai and Guangzhou are coastal cities, the latitude of Guangzhou is lower than that of Shanghai, and this condition increases the absorption effect of XCO 2 brought by season in Guangzhou. In addition, Guangzhou's geographical proximity to Zhaoqing, an industrial city, increases the dispersion of XCO 2 in Guangzhou. In this case, the amounts of changes in XCO 2 in June, July, and August were −0.29, −0.57, and 0.49 ppm. However, as the engine of China's economy, Shanghai's industrial recovery speed is extremely fast (XCO 2 was fully recovered in April, and the XCO 2 in May and June showed an upward trend), and industrial emission intensity is increasing. Therefore, under the action of increasing seasonal factors, the XCO 2 in June, July, and August showed a gradually increasing trend, with values of 0.41, 0.08, and 1.20 ppm.
In September and October, the amounts of changes in XCO 2 remained 0 ppm in the inland cities of Beijing and Wuhan. For the coastal cities of Guangzhou and Shanghai, the amounts of changes in XCO 2 remained 0 ppm in September, but the amounts of changes in XCO 2 were 0.57 and −0.35 ppm in October, respectively. This phenomenon is related to the prevention and control measures of the local government [35][36][37] and the industrial structure of the surrounding cities. In November and December, the Chinese government strengthened control measures to prevent the recurrence of the epidemic in winter. Second, the annual curve of XCO 2 are the interval of the increase in XCO 2 in November and December. Therefore, under the action of the above two factors, the amounts of changes in XCO 2 were 0 ppm in Beijing and Guangzhou. However, in Wuhan and Shanghai, the amounts of changes in XCO 2 fluctuated in November and December. The first outbreak of the epidemic in Wuhan, China lasted for a long time and had a significant effect on reducing the production vitality of some industrial companies in Wuhan [38]. The Shanghai's XCO 2 dropped in December because the Shanghai government had to strengthen the prevention and control of the epidemic after local cases emerged [37,39].
The epidemic affected Wuhan mainly in two periods, from January to May and from November to December. In Guangzhou and Shanghai, the effects were mainly from January to March and in December. Beijing is the political and economic center of China; its XCO 2 were affected by the outbreak from March to May, and the significant change occurred in March. In addition, the greater variation of XCO 2 was due to the seasonal effects in June, July, and August.

Analysis of Changes in the Concentration of Gases (NO 2 and O 3 ) from Top-Down and Down-Top, Respectively
Given that only 2019 and 2020 had complete data, we directly used Sentinel-5 Premonitory Offline Level 3 products. We obtained the monthly averaged NO 2 and O 3 data by determining the mean value in each study area. In this section, we compared the data detected in 2019. Figure 6 show the specific monthly averaged concentration (NO 2 and O 3 ) and increment changes in 2019 and 2020. The monthly averaged increment represents the increase in concentration from the previous month and to the current month.
Transport and industrial production are the main sources of NO X emissions, and thus the monthly averaged NO 2 concentration changes can reflect the cessation of productivity during the epidemic and the recovery of productivity after the epidemic. Figure 6 shows the concentration and increment change in NO 2 and O 3 in 2019 and 2020, respectively. In Figure 6, the curve of NO 2 concentration increment fluctuates mainly from February to May in 2019 and 2020, consistent with the duration of the epidemic released by the Chinese authorities [40]. During the epidemic outbreak, industrial emissions and traffic decreased to a large extent in all parts of China. In the atmosphere, compared with that in February 2019, the amplitude of NO 2 concentration in 2020 in Beijing, Shanghai, Guangzhou, and Wuhan decreased by 2.52 × 10 −5 , 2.32 × 10 −5 , 0.25 × 10 −5 , and 3.46 × 10 −5 mol/m 2 , respectively. The percentage decrease was 24%, 18%, 4%, and 39%, respectively. Then, in March, April, and May, the monthly averaged increase in NO 2 concentration in 2020 was smaller than that in 2019. Change in NO 2 concentration recovered to the 2019 level from June to December, indicating that the epidemic in China has been brought under control and industrial production has recovered [40]. From February to June 2020, the NO 2 concentration decreased, but the O 3 concentration showed an increasing trend ( Figure 6). Specifically, in February 2020, O 3 concentration increments in Beijing, Shanghai, Guangzhou, and Wuhan were 0.851 × 10 −2 , 1.55 × 10 −2 , 1.24 × 10 −2 , and 1.66 × 10 −2 mol/m 3 , respectively. The decrease percentage was 5%, 14%, 12%, and 14%, respectively. In addition, the increment change in O 3 concentration was extended to June, a month longer than the increment change time of NO 2 concentration. Similarly, in March, April, May, and June, the monthly averaged increase in O 3 concentration in 2020 was smaller than that in 2019. From July to December in 2020, the O 3 concentration increment in each region gradually returned to the normal state.
To further analyze the effect of NO 2 and O 3 concentration on the change in XCO 2 , we analyzed the daily averaged data of monitoring stations in the ground cities. Given that China's big events are held according to traditional holidays (Spring Festival and National Day), we matched the 2020 and 2019 time series according to the Lunar New Year. We obtained monthly averaged ground station data by averaging the daily averaged data of monitoring stations in the ground cities. Figure 7 shows the daily averaged and monthly averaged concentration change in NO 2 and O 3 in 2019 and 2020, respectively. We learned that the daily averaged NO 2 concentration in Beijing decreased in 11 February. The daily averaged and monthly averaged concentrations in 2020 were lower than those in 2019 from 11 February to 31 October because the prevention and control measures were strict in Beijing. The monthly averaged distribution of NO 2 concentrations began to decline in February, briefly returned to normal in November, and began to decline again in December. The largest annual decline in March was 15.56 µg/m 3 in Beijing. The percentage decline was 39%. NO 2 concentration decreased, whereas O 3 concentration increased by 4.23 µg/m 3 in February. The percentage increase was 5%. The monthly averaged NO 2 concentration decreased in March and April, and the corresponding monthly averaged O 3 concentration in the surface cities showed an obvious trend of decreasing initially and then increasing.
In the monthly averaged NO 2 concentration of Shanghai, the trends in change in NO 2 concentration in the ground monitoring stations and satellite data were similar. NO 2 concentrations at the ground monitoring stations declined from February to May, then the decline rate reached its maximum value in March, with a maximum value of 20.64 µg/m 3 . The percentage decline was 40%. Subsequently, the NO 2 concentrations returned to normal levels but fluctuated from June to November. Finally, in December, NO 2 concentrations dropped again in 2020 compared with those in 2019 because the Shanghai government introduced management measures to prevent the recurrence of the epidemic [37,39]. Regarding the monthly averaged O 3 concentration, the rising months in 2020 were February, April, September, and November. The daily averaged NO 2 concentration decreased from 15 February to 28 February, from 1 April to 22 April, from 1 September to 11 September, and from 1 November to 15 November.
For the monthly averaged NO 2 concentration in Guangzhou, the interval of decrease in 2020 can be divided into three parts: from February to March, from July to August, and from October to December. The maximum decrease was found in December, and the maximum value was 27.16 µg/m 3 . The percentage decline was 40%. The monthly averaged O 3 concentration in Guangzhou increased in February and May, and the corresponding daily averaged NO 2 concentration decreased from 6 February to 28 February and from 10 May to 21 May. These effects may have been due to the latitude and typhoon because the latitude in Guangzhou is extremely low. Wuhan is the first Chinese city to have an epidemic case and, thus, has a long shutdown time. We further analyzed the impact of NO 2 and O 3 on XCO 2 in each month in 2020. Two periods of decrease in the monthly averaged NO 2 concentration were found at the ground monitoring stations, namely, from February to May and from December to December. The maximum decrease was found in March, with a maximum value of 33.02 µg/m 3 . The percentage decline was 61%. The monthly averaged O 3 concentration in Wuhan increased from February to March and from November to November. The daily averaged NO 2 concentration first decreased and then increased from 14 February to 27 March. During this period, the monthly averaged O 3 concentration was higher than that in the same period in 2019, and the rate of increase in the monthly averaged O 3 concentration in March gradually slowed down compared with that in February. We found that the monthly averaged O 3 concentration in Wuhan in November 2020 decreased more slowly than that in 2019 and higher than that in 2019. In addition, during this period, the daily averaged NO 2 concentration decreased rapidly from 21 November to 30 November, with a value of 32.90 µg/m 3 . Figure 8 shows the percentage and concentration changes for O 3 and NO 2 for the four study areas from the satellite observation and the ground monitoring stations observation, respectively. In Figure 8a,b, the maximum monthly averaged concentration changes of NO 2 and O 3 are shown form satellite observation, respectively. In Figure 8c,d, the maximum monthly averaged concentration changes of NO 2 and O 3 are shown from ground monitoring stations, respectively. Then, we show the different changes for four cities in each subgraph. For each of the subgraphs, the gray and blue bars represent percentage data and concentration change data, respectively. In general, for the satellite observation estimates, the maximum monthly averaged NO 2

The Relation of NO 2 and O 3 Concentration on XCO 2
From the perspective of satellite and ground monitoring stations observations, we quantitatively show the effect of NO 2 concentration and O 3 on XCO 2 in Figure 9 from top-down and bottom-up. Figure 9 shows the relationship between the monthly averaged variation of XCO 2 observed by satellite and the monthly averaged concentration of NO 2 and O 3 observed by satellite (Figure 9c,d) and ground monitoring stations observations (Figure 9a,b) in February 2020 relative to 2019, respectively. To avoid the error influence of the interpolation theory of XCO2, and to accurately reflect the interaction relationship of XCO 2 on NO 2 and O 3 concentration, the XCO 2 data were adopted from the original point data of GOSAT_FTS_L3_V2.95 in this chapter, and the data processing theory has not changed. Similarly, we obtained monthly averaged differences in the concentrations of NO 2 , O 3 , and CO 2 in February 2020 relative to 2019. We have evaluated a correlation between the variation of XCO 2 and the variation of NO 2 and O 3 in Figure 9. CO 2 and NOx emissions derive from industries and traffic. Therefore, a reduction in emissions corresponds to a reduction in concentrations [38,39]. The variations of XCO 2 and NO 2 should be correlated (as shown in Figure 9a,c), and their correlation coefficients are 0.4912 (from satellite observation) and 0.3928 (from ground station observation), respectively. On the other hand, tropospheric O 3 derives from nonlinear photochemical reactions of the NOx and VOC precursors. Thus, a reduction in NO 2 can even cause an increase in O 3 . For this reason, it is possible that XCO 2 and O 3 (ground stations) are correlated, with a correlation coefficient of −0.3333.  From the point of promoting the reduction in XCO 2 in atmosphere, the contribution of NO 2 concentration is greater than that of O 3 concentration. However, the relationship between O 3 concentration and aerosols was strong because of photochemical reactions [41,42]. Therefore, we can regulate the concentration of O 3 by improving the aerosol. In general, from both top-down and bottom-up perspectives, we conclude that we can further improve carbon emissions by regulating NO 2 concentrations, but O 3 has a relatively small impact on the improvement of XCO 2 .

Conclusions
In this work, we analyzed abnormal XCO 2 changes in Beijing, Shanghai, Guangzhou, and Wuhan in each month in 2020, and combined the Sentinel-5 Precursor Offline Level 3 product and the ground monitoring network to analyze the effects of NO 2 and O 3 concentrations on changes in XCO 2 with top-down and bottom-up methods, respectively. We found significant differences in the monthly averaged concentrations of CO 2 , NO 2 , and O 3 . For Beijing, Shanghai, Guangzhou, and Wuhan, the results suggested that the amounts of changes in XCO 2 were 0.13, −0.99, −0.89, and −1.12 ppm in February. Furthermore, the concentrations of NO 2 and O 3 changed by −24% and 5%, −18% and 14%, −4% and 12%, −39% and 14% by comparing data in February 2019 with the Sentinel-Satellite-5 data. Moreover, for Beijing, Shanghai, Guangzhou, and Wuhan, with regard to the ground monitoring network data, the largest amounts of changes in monthly averaged concentration of NO 2 and O 3 were −39% and 14%, −40% and −2%, and −61% and 9% in February. However, the largest monthly averaged NO 2 concentration change in Guangzhou occurred in December, with a value of −40%. Thus, the O 3 concentration change was −12% in December.
In addition, we evaluated the relationship between the variation of XCO 2 and the variation of NO 2 and O 3 in Figure 9. CO 2 and NOx emissions derive from industries and traffic. Therefore, a reduction in emissions corresponds to a reduction in concentration [40]. The variations of XCO 2 and NO 2 should be correlated [43]. Additionally, the results also show that the effect of NO 2 concentration on XCO 2 is positively correlated from the point of the satellite (R = 0.4912) and the point of the ground monitoring stations (R = 0.3928). On the other hand, tropospheric O 3 derives from nonlinear photochemical reactions of the NOx and VOC precursors. Thus, a reduction in NO 2 can even cause an increase in O3. We also found that XCO 2 and O 3 are correlated, with a correlation coefficient of 0.2391(from satellite observations) and −0.3333 (from ground monitoring stations). Therefore, from the point of promoting the reduction in XCO 2 in atmosphere, the contribution of NO 2 concentration is greater than that of O 3 concentration. Nevertheless, the relationship between O 3 concentration and aerosols was strong because of photochemical reactions [41,42]. We can regulate the concentration of O 3 by improving the aerosol. These effects improved air quality and the status of the natural environment. At last, the government should consider reducing XCO 2 and NO 2 concentration at the same time to make a synergistic reduction.