Dramatic decline of observed atmospheric CO2 and CH4 during the COVID-19 lockdown over the Yangtze River Delta of China.

The temporal variation of greenhouse gas concentrations in China during the COVID-19 lockdown in China is analyzed in this work using high resolution measurements of near surface △CO2, △CH4 and △CO concentrations above the background conditions at Lin'an station (LAN), a regional background station in the Yangtze River Delta region. During the pre-lockdown observational period (IOP-1), both △CO2 and △CH4 exhibited a significant increasing trend relative to the 2011-2019 climatological mean. The reduction of △CO2, △CH4 and △CO during the lockdown observational period (IOP-2) (which also coincided with the Chinese New Year Holiday) reached up to 15.0 ppm, 14.2 ppb and 146.8 ppb, respectively, and a reduction of △CO2/△CO probably due to a dramatic reduction from industrial emissions. △CO2, △CH4 and △CO were observed to keep declining during the post-lockdown easing phase (IOP-3), which is the synthetic result of lower than normal CO2 emissions from rural regions around LAN coupled with strong uptake of the terrestrial ecosystem. Interestingly, the trend reversed to gradual increase for all species during the later easing phase (IOP-4), with △CO2/△CO constantly increasing from IOP-2 to IOP-3 and finally IOP-4, consistent with recovery in industrial emissions associated with the staged resumption of economic activity. On average, △CO2 declined sharply throughout the days during IOP-2 but increased gradually throughout the days during IOP-4. The findings showcase the significant role of emission reduction in accounting for the dramatic changes in measured atmospheric △CO2 and △CH4 associated with the COVID-19 lockdown and recovery.

Introduction

The concentration of CO2 in the atmosphere has increased from approximately 277 ppm in 1750 (Joos and Spahni, 2008) at the beginning of the industrial era, to 410.5 ppm in 2019 (WMO, 2020). The increase in anthropogenic emission of greenhouse gases (GHGs), in combination with other anthropogenic drivers such as reduction of CO2 uptake by the biosphere, are extremely likely to be the dominant cause of the global climate change since the mid-20th century (IPCC, 2014). It was reported that the CO2 emissions in China reached up to 11,255.88 million tons in 2018, accounting for roughly 30% of the total global emissions of CO2 (Crippa et al., 2020), in spite of the great efforts made by Chinese government to reduce emissions in China (Guan et al., 2009).

Unprecedentedly, the outbreak of the 2019 Coronavirus epidemic (COVID-19) led to strict lockdown measures implemented economy-wide for the first time on January 23, 2020 in China (Wu et al., 2020). Such lockdown policies caused significant reduction in global fossil fuel consumption (Le Quéré et al., 2020; Liu et al., 2020), which provides a unique opportunity to unravel the potential impact of reduced fossil fuel emissions on the concentrations of air pollutant and greenhouse gases (Le Quéré et al., 2020; Liu et al., 2020; Myllyvirta et al., 2020; Zheng et al., 2018). Wide-spread improvement in air quality has been widely reported following lockdown in the world's most polluted cities (Shrestha et al., 2020). Nevertheless, sporadic air pollution episodes frequently occurred in eastern China during COVID-19 (Chang et al., 2020; Su et al., 2020; Wang et al., 2020), which were attributed to a wide range of causes, including much shallower boundary layer (Su et al., 2020), regional transboundary transport (Huang et al., 2020), the enhanced conversion of NOx to particulate nitrate (Chang et al., 2020), aerosol heterogeneous chemistry promoted by high humidity (Le et al., 2020), and enhanced biomass burning (Wang et al., 2021).

Even with these known changes, previous research focusing on the changes in near-surface CO2 and other long-lived GHGs concentrations remains limited. On a global scale, a reduction of 0.08–0.23 ppm in annual CO2 concentration was estimated by the Global Carbon Project (GCP), due largely to the impact of COVID-19 lockdown (WMO, 2020). The similar conclusion was reached by Carbon Brief (Betts et al., 2020). In the Northern Hemisphere, a 0.25 ppm decrease of CO2 was expected at the end of April (Zeng et al., 2020). One finding has demonstrated that the reduction of CO2 was mainly limited to local sources (Chevallier et al., 2020). However, no significant changes in CO2 were observed in other places, such as near Gartow in Germany, within the time period constrained by their own local COVID-19 lockdown (Kutsch et al., 2020). As such, it is imperative to see whether there is a reduction of CO2 from synergistic analysis on more geographic scales, or if this is merely a localized phenomenon.

The purpose of this present study is to unravel the potential impact of COVID-related shutdown measures on atmospheric CO2 concentrations at a remote site, in specific Lin'an station (LAN, 119.72°E, 30.3°N, 138.6 m a.s.l.), which is a regional background station located at the western edge of the Yangtze River Delta (YRD) conurbation in Eastern China.

Data and methods

The present study is based on continuous measurements of atmospheric CO2, CH4 and CO mole fractions measured from January 2011 to April 2020 at LAN, about 50 km west of Hangzhou, 200 km southwest of Shanghai and 200 km South of Nanjing, three nearby cities bounding the conurbation where major CO2 emissions in this part of China occur (Fig. 1 ). The atmospheric CO2, CH4 and CO mole fractions are simultaneously measured by a Cavity Ring-Down Spectrometer (CRDS; Picarro Inc. USA). This instrument has been calibrated with a well-established system (Fang et al., 2013), where the CO2, CH4 and CO data measured are referenced to the WMO X2007 scale (Zhao et al., 2006; Zhao et al., 1997), WMO NOAA 04 scale and 2004 scale (Dlugokencky et al., 2005., Novelli et al., 2012), respectively. Additional data processing and quality control methods are extensively documented in the previous study by Fang et al. (2015).

Fig. 1

(a) The spatial distribution of CO2 emission for 2014 (data from Global Atmospheric Research, EDGAR v4.3.2). (b) A regional map shows the location of the station (filled red pentagram), the adjacent megacities (yellow hollow circle) and surrounding area.

To obtain the contribution of regional emissions to the observed concentrations of CO2, CH4 and CO, this work first extracts the background concentrations using the meteorological method and robust extraction of the baseline signal (Fang et al., 2015; Ruckstuhl et al., 2010). Second the best-fit smooth curves to the background data are obtained using the methods proposed by Thoning et al (Thoning et al., 1989). Finally, the background signal is subtracted from the matching original hourly means in an attempt to compute the difference between the measurement and the background value, herein defined as △CO2 △CH4 and △CO respectively (Mitchell et al., 2018). Scatter plots of △CO2 vs. △CO using the hourly average values during every individual study period are made, with the slope and intercept of the best fit △CO2/△CO linear regression derived and subsequently used for analysis.

Given that CO is often co-emitted along with CO2 due to combustion (Turnbull et al., 2006; Bakwin et al., 1998; Lin et al., 2020), it has been used as an excellent tracer. In specific, the observed ratio of △CO2/△CO is well recognized to be able to identify the relative contribution of fossil fuel combustion emissions and biospheric activities related to the uptake of atmospheric CO2 (Zhang et al., 2013).

The intensive observational period (IOP) covers the period from 3 January to 20 March 2020 and is divided into 4 subperiods for the sake of subsequent analysis: pre-lockdown (IOP-1), lockdown and Chinese New Year (IOP-2), post-lockdown easing phase 1 (IOP-3), and post-lockdown easing phase 2 (IOP-4). For details of the exact days of the four IOPs, refer to Table S1. During the reference time period 2011–2019, the time origins in all figures, unless noted otherwise, are set as the Chinese Lunar New Year (CNY) to form a valid comparison with the time used when strict lockdown measures were implemented for the COVID-19 lockdown period in 2020. This is because CNY is known to be a period of time when many businesses suspend work for a 1 to 2 week long holiday and many people leave heavily developed urban areas like the YRD and head to their family homes/villages elsewhere, leading to a known decrease in emissions in large conurbations (Huang et al., 2012).

Results and discussion

Temporal evolution of △CO2, △CH4 and △CO

Fig. 2 shows the time series of daily average △CO2, △CO, CO2/CO and △CH4 during the four IOPs computed using all data from 2011–2019 and data only from 2020 respectively. As a whole, the day-by-day variation dominated for both △CO2 and △CO in 2020, which showed an almost completely in-phase temporal variation pattern. Nevertheless, the temporal variation of the ratio of △CO2/△CO was found to be out of phase with the variations of both △CO2 and △CO, due to the larger vibration in △CO. During IOP-1, elevated △CO2, occurred in more days in 2020 as compared with the period from 2011–2019, suggesting the year-on-year increase in emission was not different from the other year-on-year changes. This can be partly supported by provincial CO2 emission inventories for China published by Shan et al. (2018; 2020), which followed the Intergovernmental Panel on Climate Change (IPCC) emissions accounting method (Fig. S1 in the supplementary material). When summing up the CO2 emissions for all sectors over Anhui, Zhejiang, Jiangsu province and Shanghai, the amount keep increasing from 2011 to 2017, except the decline in 2016. Higher △CO were observed in nearly half of the days in IOP-1, partly attributed to the reduced or negative growth rate in CO emissions due to improved energy efficiency which has been widely reported (Zheng et al., 2018). For △CH4, higher values were observed in most days in 2020 than the historic period, but not as many as △CO2. The CH4 emissions provided by EDGARv6.0 (www.https://edgar.jrc.ec.europa.eu/dataset_ghg60#p1) shows the increasing trend in CH4 emissions. The gridded data were extracted for the area of 27.5°-34°N by 118.5°-122°E, which basically covers Yangtze River Delta of China. Furthermore, notable is the stronger fluctuations of △CO2 and △CO during IOP-1 than during the subsequent IOPs, which were due to a combination of meteorological variability and an ever increasing and more variable emissions profile due to the deepening and constant change of the economy as it continued to both increase and deepen year-on-year, as compared to the lockdown, which forced many industrial and commercial emissions sources to go offline or scale back significantly. For instance, persistent low-level clouds observed above LAN (not shown) during the last four days (20–23 January 2020) of IOP-1 could at least account for the evident spikes of △CO2, △CO and △CH4 from a perspective of the cloud based reduction of solar radiation reaching the ground surface during daytime, thereby suppressing the planetary boundary layer (PBL) and leading to more accumulation of air pollution and CO2. Additionally, another non-negligible factor could be the enhanced emissions induced from more intensive traffic as well as a flurry of last-minute activities in the industrial space as many final orders are rushed just before CNY.

Fig. 2

Time series of △CO2, △CO and △CH4 daily means observed at LAN station from IOP-1 to IOP-4 in 2011–2019 and 2020. (a) Bottom: the daily average of △CO2 in 2011–2019 (green dot) and 2020 (blue bar). Top: the △CO2/△CO ratio in 2011–2019 (orange dot) and 2020 (grey bar). (b) The daily average of △CO in 2011–2019 (green dot) and 2020 (blue bar). (c) The daily average of △CH4 in 2011–2019 (green dot) and 2020 (blue bar). Error bands of all species in 2011–2019 indicate 95% confidence intervals of each day.

During IOP-2, sharp drops in △CO2 and △CO compared to IOP-1 were observed in both 2011–2019 and 2020, with the latter period experiencing an even larger magnitude of reduction. Significant differences exist between IOP-1 and IOP-2 at the 95% confidence level when applying the t-test for both 2011–2019 and 2020. The decrease in 2011–2019 was mainly due to decreased economic activity and travel that both occur during the CNY holiday (Ding et al., 2013), while the lower than climatological mean concentration observed in 2020 reflected a larger than normal reduction in fossil fuel activity consistent with the lockdown associated with COVID-19. Among them, evident reduction occurred in the later stage when forced confinement and lockdown protocols were strictly implemented, including social distancing and home quarantine, the shutting down of all unnecessary industries, and restriction of transportation both inter-city and intra-city.

During IOP-3, gradual recovery of △CO2 and △CO were observed during the 2011–2019 period, associated with the end of the CNY holiday and a slow return to economic activity. There is significant difference between IOP-2 and IOP-3 in 2011–2019. By comparison in 2020 low △CO2 and △CO continued in the first 13 days of IOP-3 as the lockdown measures continued. During this period, most industries were still closed, even though a staged resumption policy was deployed. However, △CO2 and △CO were observed to increase gradually during IOP-4 of 2020 with the successful implementation of the staged resumption of economic activity policy.

Box plots of △CO2, △CO, △CH4, △CO2/△CO slope and intercept during the four IOPs are shown in Fig. 3 . During 2011–2019, the average concentrations of △CO2, △CO and △CH4 decreased from 3.0 ppm, 160.6 ppb and 12.5 ppb in IOP-1 to 0.8 ppm, 68.1 ppb and 0 ppb in IOP-2, then increased to 6.2 ppm, 114.7 ppb and 18.7 ppb in IOP-3, and ultimately reduced to 5.6 ppm, 64.8 ppb and 16.1 ppb in IOP-4, respectively. The one-trough mode was mainly due to the significant reduction in anthropogenic activities during the CNY holiday (IOP-2) (Ding et al., 2013), and the slight variation in △CO2, △CO and △CH4 as the time transitioned from IOP-3 to IOP-4 was probably due to high frequency changes in synoptic events, short-term changes in industrial and transportation emissions, or changes in the uptake of terrestrial ecosystems (Wang et al., 2007; Peters et al., 2017; Wang et al., 2021; Deng et al., 2021). Furthermore, the lower △CO2/△CO slope and higher △CO during IOP-1 relative to post-CNY period (i.e., IOP-3 and IOP-4) was partly attributed to the high transportation emissions from “spring travel rush” in the beginning of the CNY holiday, as emissions sources from transportation often lead to lower △CO2/△CO compared to large, efficient power plants (Turnbull et al., 2011; Wang et al., 2010). This can be further corroborated by the lower △CO2/△CO ratio during IOP-2, which could be caused by the drop of emission from industry-sector during CNY (Liu et al., 2020; Zheng et al., 2018). Besides the contribution of emission recovery from industry-sector, the increase in △CO2/△CO slope from IOP-2 to IOP-4 was also partly attributed to the reduction of low efficiency domestic heating sources with the growing temperature, as well as less efficient smaller or individual industrial workshops which are also less efficient in general, and which had a greater tendency to not re-open, or to upgrade after the end of the holiday period.

Fig. 3

Box plots of △CO2 (a), △CO (b), and △CH4 (d) and slope and intercept of △CO2/△CO linear regression (c) for the four IOPs in 2001–2019 (blank box) and 2020 (blue or red box). The minimum, 25th percentile, median, 75th percentile and maximum, as well as 1% percentile (fork), mean (hollow square) and 99% percentile (fork) are shown in the box plot, respectively.

On average, the △CO2 (17.6 ppm), △CO (189.7 ppb) and △CH4 (29.0 ppb) during IOP-1 of 2020 were higher than the climatological means for the period 2011–2019, indicative of the constantly increasing CO2 emissions over the YRD as economy continued to grow (Wang et al., 2010; Zeng et al., 2008; Shan et al. 2018; Shan et al. 2020; Zheng et al., 2020). Likewise, the much higher △CO2/△CO slope (59.9) in IOP-1 compared to 29.7 during 2011–2019 suggested much higher combustion efficiency, which to some extent reflected the Chinese government's efforts on the adjustment of energy structures and novel technology application for improving combustion efficiency (Demirbas, et al., 2009; Zheng et al., 2018). This is supported by the recent implementation of China's Stage VI emissions standard–the most stringent vehicles emission standards–which became effective as of July 2019 in Hangzhou (Xinhua Net. 2020), leading to a consistent result with the observed △CO2/△CO ratios (Bishop and Stedman, 2008). The higher △CO2/△CO intercept compared to IOP-2, IOP-3 and IOP-4 implied the higher net effect of biogenic sources and sinks in the YRD region, which was attributed to enhanced ecosystem respiration due to exceptionally high temperature seen in January of 2020 (Rustad et al., 2001; Bond-Lamberty and Thomson., 2010). During IOP-2, larger reductions in △CO2 (15.0ppm), △CO (146.8 ppb) and △CH4 (14.2 ppb) were observed in 2020 compared to last nine-year mean (Grey shading in Fig. 3), concurrent with the implementation of strict confinement which led to the reduced emissions from industry, power supplies, transportation and residential living. The reduced △CO2/△CO slope (42.2) indicated evident emission drop from power generation and industry. During IOP-3, even lower △CO2 (2.2 ppm), △CO (36.1 ppb) and △CH4 (-0.5 ppb) are observed. The higher △CO2/△CO slope (67.6) reflects the synthesis impact of gradual recovery of production and power plants concerning residential energy use and persistent drop of emission from transportation (Liu et al., 2020). The significantly low △CO2/△CO intercept suggested the strong net sink of terrestrial ecosystem through photosynthesis. During IOP-4, all enterprises necessary for life were resumed, △CO2 (6.5 ppm), △CO (72.4 ppb) and △CH4 (10.8 ppb) gradually increased to the levels similar to the last nine-year climatology mean. The increase in △CO2/△CO slope suggests the rebound of industries with the ease of quarantine controls in China.

Diurnal variations of △CO2 during four IOPs

Fig. 4 shows the diurnal cycle of △CO2 in 2011–2019 and 2020 during the four different IOPs. The first observation is that this time cycle is mainly driven by the diurnally varying local sources/sinks and dynamics of the PBL (Bakwin et al., 1998). Overall, the trough in terms of △CO2 values occurred at 1300–1500 LST when the PBL grew to the maximum depth by strong turbulent transport processes (Li et al., 2017; Guo et al., 2020), thereby leading to the greatest dampening of surface CO2 fluxes. Meanwhile, the net uptake of CO2 occurred at noon when the removal of CO2 by photosynthesis that became active exceeded respiration from the terrestrial ecosystem. These findings are consistent with basic theory and add support to the idea that the measurements are valid and consistent. Henceforth any significant changes as discussed were likely not due to the placement of the measurement sites or other non-linear effects associated with an improper coverage.

Fig. 4

Hourly mean of △CO2 in 2011–2019 (blue) and 2020 (red) during IOP-1 (a), IOP-2 (b), IOP-3 (c) and IOP-4 (d). The bands are 95% confidence intervals of each hour.

Interestingly, the diurnal cycles of △CO2 exhibited a pattern with two-peaks and one-trough, irrespective of 2011–2019 and 2020. One peak occurred around sunrise (0600–0900 LST), which corresponded to rush hour when intensive traffic emissions occurred. The other peak happened in the midnight (2400–0300 LST), when the net sources from vegetation respiration happened at roughtly the same time as the most stable and shallow PBL. Between the first peak and the trough, △CO2 gradually decreased from sunrise to midday when the rapid growth of PBL diluted CO2 by drawing in fresh air from aloft. Between the trough and the second peak, △CO2 gradually increased after sunset as emissions filled the shallow nighttime PBL.

As expected, during IOP-1 of 2020, hourly mean △CO2 throughout the course of day was much higher than in the same period of 2011-2019. During IOP-2, hourly mean △CO2 decreased in both 2011–2019 and 2020, compared to IOP-1, with the largest decline occurring in 2020, reflecting the dramatic impact of lockdown measures. During IOP-3, an uptick in △CO2 was observed for the diurnal readings in 2011–2019, suggesting a recovery of CO2 emission from fossil fuels combustion with the end of CNY holiday. While in 2020, △CO2 kept declining during most of time, much lower in general than as observed in 2011–2019, indicating the persistent impact of low CO2 emission due to quarantine measurements. During IOP-4, average CO2 increased throughout all hours in 2020, increasing to similar levels as the climatological mean, suggesting the impact of alleviation of quarantine measures in China, although possibly on a slightly lower planetary background due to increasing reductions in economic activity elsewhere.

Potential impact of meteorological variables

It is well documented that meteorological conditions have effects on the surface CO2 through diffusion and mixing processes (Bischof et al., 1980; Haszpra et al., 2012). Here, the meteorological variables related to the vertical satiability and horizontal diffusion were used, including boundary layer height (BLH), lower tropospheric stability (LTS) and wind speed (WS). BLH, to some extent, affects the dilution of near-surface pollutants mainly through vertical convection and turbulence mixture (Nair et al., 2018; Lou et al., 2019). LTS is defined as the difference in potential temperature between 700 hPa and the surface (Slingo, 2007), which can be used to determine the thermodynamic state of the lower troposphere (Guo et al., 2016). WS is recognized to be able to dictate the horizontal advection of air masses. The association of daily mean BLH, LTS and WS with △CO2 was analyzed only during two time periods each day: from 1200 to 1600 LST (the time period when the strongest convection and turbulence occurred) and from 2300 to 0300 LST (the time period when the stable PBL dominated), respectively. The strongest correlations were found in the afternoon (1200–1600 LST). Time series of daily mean of BLH, LTS and WS during 1200–1600 LST and △CO2 anomalies relative to 2011–2019 are plotted in Fig. 5 . Among all factors, BLH was most correlated with △CO2 anomalies (R = −0.40), followed by WS (R = −0.27) and LTS (R = 0.19). Averaged WS (R = −0.26) and LTS (R = 0.15) during 2300-0300 LST were also found correlated with △CO2 anomalies (data not shown). The anti-correlations between BLH/WS and △CO2 suggested that higher BLH and WS reinforced vertical dilution and horizontal dispersion of CO2, respectively, and consequently led to reduced CO2 (Yi et al., 2000). The positive relationship between LTS and △CO2 indicated the stable lower troposphere dampened the diffusion of surface CO2 efflux. However, the individual correlation values were considered relatively small, warranting further discussion of the real-world application of the general theory to high-frequency measurements.

Fig. 5

Time series of daily averaged △CO2 anomalies relative to 2011–2019 (black bar), BLH (green solid line), LTS (blue solid line), and WS (red solid line) during 1200–1600 LST. R(CO2BLH), R(CO2LTS), R(CO2,WS) denotes the correlation coefficients between △CO2 anomalies and BLH, LTS and WS, respectively.

From IOP-1 to IOP-4, the average LTS gradually decreased, while BLH and WS generally increased except for a slight decrease in BLH from IOP-2 to IOP-3. In such case, one explanation for the evidently higher △CO2, △CO and △CH4 during IOP-1 is the weaker atmospheric advection and entrainment indicated by low BLH/WS and high LTS. The wind field at 850 hPa (Fig. S1 in Appendix A) shows that the low-level winds are mostly from southwest, bringing air pollution from southern China and Continential Southeast Asia. During IOP-2, the reduced CO2 was partly attributed to more diffusion by entrainment of fresh air from above with rise in WS/BLH and decline in LTS. Contrarily, there were other atmospheric circulation tends that tended to increase CO2 at the receptor site as due to the following aspects. Firstly, the decrease in geopotential height at 500 hPa (GH500hPa) suggested stabilizing lower tropospheric circulation. Secondly, the wind field at 850 hPa indicates that the air masses that passed over this region in large part derived from the sea region to the northeast of Shanghai and subsequently passed over the background measurement station, providing ample ability to sample the impacts of emissions changes from the megacity core. During IOP-3, WS and LTS were favorable for mixing, which could result in the decrease in CO2, as well as the possible mixing in of enhanced CO from above due to long-range transport of sources from Southern China and/or Southeast Asia. This is further consistent with the increase in GH500hPa also indicating strong local convection. Similarly, the wind field indicated the influence of long-range transport from central China with slow speed. By contrast, a slight decrease in BLH could lead to suppressed vertical mixing and the consequent buildup of CO2. During IOP-4, increased BLH/WS and reduced LTS were generally linked to growing vertical advection. Besides, enhanced GH500hPa also supported strong convection. Nevertheless, the observed CO2 increased, suggesting that the recovery in CO2 was dominated by the emission increase from socioeconomic activities recovery. Moreover, the wind field showed more influence from Northern China, a region with intensive industries. These results are all consistent with the observed economic recovery in northern China.

The △CO2 segregated by horizontal wind direction were also studied (Fig. S2 in Appendix A). The impact of the control measures can be found from the different patterns between 2020 and 2011–2019. During the period 2011–2019, the winds from the ENE-E-ESE sectors were generally accompanied with higher CO2 during all periods, due to abundant emission sources in the mega-conurbation existing from Shanghai through Hangzhou and Nanjing, all located to the east through northeast. From IOP-1 to IOP-2, reduced CO2 were observed in all directions, indicating reduced CO2 emissions nearby due to limited human activities during CNY holiday which are known to occur throughout China. From IOP-3 to IOP-4, CO2 from most directions were enhanced when CNY holiday ended. In 2020, higher CO2 were observed for all wind sectors during IOP-1 compared to 2011–2019, mainly due to increasing trend of CO2 emissions. During IOP-2, CO2 from all wind directions largely decreased to the similar level with the climatological mean under deployment of the lockdown policy, except for the higher load under the prevailing winds from the SSW-S-SSE-SE that corresponded to the downtown areas of LAN. The weak COVID signal in small town reflected the emissions reductions are concentrated mainly in industrialized regions and areas, which is consistent with the baseline of required emissions for power generation, water distribution, medical use, etc. were not as impacted as industrial and transport emissions. During IOP-3, reduced CO2 was observed in all directions except for E-ESE, generally corresponding to the Hangzhou conurbation, compared to both IOP-2 in 2020 and climatological mean, reflecting the recovery of factories in Hangzhou under staged resumption policy. During IOP-4, the CO2 from all directions were observed similar to climatological mean, suggesting that socioeconomic activities surrounding LAN gradually recovered to the normal state in 2011–2019.

Summary and conclusions

In this study, the influence of emission reductions during the COVID-19 lockdown on atmospheric CO2, CH4 and CO in China was comprehensively investigated. Larger reductions in △CO2, △CO and △CH4 than 2011–2019 were observed from IOP-1 to IOP-2, consist with the remarkable emission reduction in fossil fuel combustion and industrial sources due to the confinement measures. Large reduction in △CO2/△CO slope during IOP-2 indicates the prohibition of high efficiency combustion like industries combined with a larger CO2 uptake connected to the enhanced temperature and less anthropogenic disturbance associated with the lockdown, and possibly higher CO sources due to long-range transport from activities occuring in Southern China or Southeast Asia. The increase in △CO2/△CO slope during IOP-3 suggests gradual recovery of CO2 emission from industries and delayed drop CO2 emission from transportation. The continuous decrease in △CO2, △CO and △CH4 during IOP-3 were attributed in part to a strong net sink of terrestrial ecosystem uptake through photosynthesis (which was enhanced in part due to the lockdown) and continuous decrease in CO2 emissions from fossil fuel combustion in the vicinity of LAN except for the large conurbation from Hangzhou to Shanghai and Nanjing. During IOP-4, △CO2, △CO and △CH4 increased to the level similar to the climatological means, implying the more robust and widespread recovery of anthropogenic activities under relaxed public health policies in China. The diurnal variability was also affected during the lockdown period. The △CO2 hourly mean declined throughout the day during IOP-2 and during most of time in IOP-3, and subsequently recovered during IOP-4. CO2 was found negatively correlated with BLH/WS and positively correlated with LTS, respectively. GH500hPa and wind direction were shown to have an influence on the near-surface CO2.