Introduction

The outbreak and spread of the novel coronavirus (COVID-19) has become a major public health event threatening the safety of the population and social stability (Liu et al. 2020a; Wang et al. 2020a). On December 9, 2020, a total of 67,780,361 confirmed cases of COVID-19 have been reported worldwide, and 94,500 confirmed cases in China (https://covid19.who.int/table), and Wuhan City in Hubei Province reported 68,149 confirmed cases in total (https://voice.baidu.com/act/newpneumonia/newpneumonia). In order to prevent the virus from spreading rapidly with the movement of the population, Wuhan City was completely shut down from January 23 until April 7, thus lasting 76 days.

A decline of atmospheric pollution has been observed in many cities and nations worldwide as a consequence of lower vehicle traffic, industrial activities and pollutant emissions (Bao and Zhang 2020; He et al. 2020; Lian et al. 2020; Zheng et al. 2020). For instance, the Royal Belgian Institute for Space Aeronomy studied the effect of the COVID-19 on air quality and found that tropospheric NO2 above major cities in China decreased by 30–50% early 2020 compared to 2019 (https://www.aeronomie.be/en/news/2020/tropomi-observes-impact-corona-virus-air-quality-china). Air pollution has highly decreased during city lockdown, and air quality improvements were larger in colder, richer and more industrialized cities (He et al. 2020). A study of 44 cities in northern China showed an average 7.80% decline of the air quality index (AQI, Bao and Zhang 2020). Specifically, concentrations of airborne particulate matter (PM2.5) decreased by 35% and concentrations of nitrogen dioxide (NO2) declined by 60% in January 23–February 29 versus January 1–22, 2020 (Shi and Brasseur 2020). Lower SO2, NOx and PM2.5 emissions over the Yangtze River Delta region were explained by reduced industrial operations, constructions and vehicle traffic (Li et al. 2020). Similar improvement of air quality was observed in central China, including Wuhan (Xu et al. 2020).

Wuhan is one of the most densely populated Chinese cities with a permanent resident population of 11.212 million in 2019. The major sources of air pollution in Wuhan are emissions from coal-fired enterprises, vehicle emissions and pollution from biomass burning from the surrounding countryside (Liu et al. 2020b). During lockdown, local public transportation was suspended, and the movement of people into and out of the city was interrupted. The vehicle numbers in Wuhan dropped at least 63.31%, and the dropping percentages of ring road and high-level road traffic were 84.81% and 80.22%, respectively (Wu et al. 2020). In mid-February, Wuhan lockdown measures were further strengthened with management of all residential communities (Pan et al. 2020). While global reduction of air pollutants has been evidenced in most cities worldwide, there is few detailed knowledge on the spatial evolution of pollution at the local scale. Therefore, we present here a high resolution mapping of human activities using nighttime light radiance, and the temporal variations of air pollutants PM2.5, PM10, SO2, CO, NO2 and O3 during the COVID-19 lockdown in Wuhan.

Experimental

Study area

As the capital of Hubei Province, Wuhan is a central city in central China and in the Yangtze River Economic Zone. Belonging to subtropical monsoon humid climate, Wuhan has abundant sunshine and rainfall with an annual average temperature of 15.8–17.5 ℃; the rainfall being predominant in June–August. Wuhan has developed roads, railway, water and air transportation, and it is a key industrial city in China. The whole Wuhan covers an area of 8,569.15 km2. Up to the end of 2019, the whole city had realized gross domestic product of CNY 1,622.321 billion with permanent resident population of 11,212,000, and the urbanization rate in 2019 was 80.49%.

Wuhan has 13 administrative districts, including seven central districts including Jiangan, Jianghan, Qiaokou, Hanyang, Wuchang, Qingshan and Hongshan, and six new urban districts: Caidian, Jiangxia, Huangpi, Xinzhou, Dongxihu and Hannan. Jianghan and Jiangan districts are core commercial development areas in Wuhan, and all kinds of business centers are mostly concentrated in these areas. The commercial development is quite balanced in Hongshan and Wuchang districts. Liu et al. (2020c) chose 91 sub-districts, i.e., at the block or the street scale, within central districts, and found that 63% of subdistricts of Jiangan, Jianghan and Qiaokou are commercial, and 78% of educational sub-districts are in Wuchang, Qingshan and Hongshan. However, the development of business centers in peripheral city areas and suburban areas in Wuhan relatively lags behind with partially low agglomeration degree and smaller scale as well as single function types. Except for Jiangxia and Dongxihu districts, no business centers conforming to planning expectations are formed (Kang et al. 2018; Xu et al. 2019). Besides, Wuhan embraces over 80 colleges and universities, which are also mainly concentrated in Wuchang, Hongshan and Qingshan districts. There are 10 state-controlled air pollution monitoring sites located in Wuhan City (Fig. 1, Table S1).

Fig. 1
figure 1

Spatial distribution of the 10 state-controlled air pollution monitoring sites located in Wuhan City

Full size image

Data sources

Air quality data in the first four months of 2018, 2019 and 2020 in the 10 state-controlled monitoring sites of Wuhan were acquired from the China Environmental Monitoring Centre (CNEMC) at http://www.cnemc.cn. In addition, the “mean” defined in this study was the arithmetic mean value of hourly values. The mass concentration of one air pollutant in Wuhan was the mean value of the concentrations of this pollutant measured at nine state-controlled monitoring stations. Chenhu Qihao, as the urban background site, was not included into the mean value of the whole city.

Hourly meteorological parameters including temperature in ℃, relative humidity in %, and 10-min average wind speed in m/s were acquired from Hubei Meteorological Service at http://hb.cma.gov.cn. The nighttime light (NTL) data were taken from Visible Infrared Imaging Radiometer Suite (VIIRS) onboard the Suomi National Polar-orbiting Partnership (Suomi-NPP, Zhang et al. 2017). The NPP-VIIRS Day/Night Band (DNB) month composite data in December 2019 and February 2020 were collected. The monthly average of NPP-VIRRS data were obtained from the website of Earth Observations Group (https://eogdata.mines.edu/download_dnb_composites.html).

Nighttime light

Nighttime light (NTL) data have been extensively used to investigate scenarios related to human social activities, such as dynamic monitoring of urban expansion (Ma et al. 2015), and estimation of energy consumption (Sun et al. 2020). In order to mitigate the pandemic, people were required to reduce unnecessary outings and gatherings. NTL data can thus be used as an auxiliary means to study the changes of production and life during the COVID-19 in Wuhan. The spatial resolution of the monthly NTL data from the Earth Observations Group is 15 arc-seconds. Each pixel of the NTL data contains information on the mean detected radiance, with unit of nW/cm2/sr.

Air quality index

According to the Technical Regulation on Ambient Air Quality Index (on trial), AQI is the maximum among air quality sub-indices of the following pollutants: SO2, NO2, CO, O3, PM10 and PM2.5 (Zheng et al. 2014), shown in Eq. (1, 2).

$${\text{IAQI}}_{{\text{P}}} = \frac{{{\text{IAQI}}_{{{\text{Hi}}}} - {\text{IAQI}}_{{{\text{Lo}}}} }}{{{\text{BP}}_{{{\text{Hi}}}} - {\text{BP}}_{{{\text{Lo}}}} }}\left( {C_{{\text{P}}} - {\text{BP}}_{{{\text{Lo}}}} } \right) + {\text{IAQI}}_{{{\text{Lo}}}}$$
(1)
$${\text{AQI = max (IAQI}}_{{1}} {\text{, IAQI}}_{{2}} {\text{, IAQI}}_{{3}} {,} \ldots {\text{, IAQI}}_{{\text{n}}} {)}$$
(2)

where IAQIP refers to the air quality sub-index of pollutant P; CP represents the mass concentration of pollutant P; BPHi is the breakpoint that is greater than or equal to CP; BPLo is the breakpoint that is less than or equal to CP; IAQIHi is the air quality sub-index corresponding to BPHi; IAQILo is the air quality sub-index corresponding to BPLo; n is the number of air pollutants. IAQI and corresponding thresholds of each pollutants is shown in Table S2. The AQI value is divided into six grades:

  1. 1.

    0–50: superior air quality;

  2. 2.

    51–100: good air quality;

  3. 3.

    101–150: mild air pollution;

  4. 4.

    151–200: moderate air pollution;

  5. 5.

    201–300: heavy air pollution;

  6. 6.

    More than 300: severe air pollution.

The air pollutant having the maximum IAQI value when AQI is larger than 50 is defined as the primary pollutant. In this study, the primary pollutant was identified in order to determine which pollutant provided the largest contribution to air quality degradation. The daily AQI in Wuhan was calculated as the average of the AQI at nine state-controlled monitoring stations.

Results and discussion

Nighttime light radiance to map human activity

We studied the effect of lockdown on human activities by mapping variations of nighttime light radiance before and during lockdown in Wuhan, green indicating light decline and red denoting light increase (Fig. 2a, b). The sum of radiance variations for each district is given in Fig. 2c and Table S3. Results show first a decrease of radiance for eight districts located mostly in the center, which was expected from the decrease of human activities during lockdown. Yet data also show an increase of radiance in five surrounding districts, which was less expected. Noteworthy, our findings reveal that radiance variation are not uniform and that a high-resolution mapping discloses heterogeneities such as radiance rises and declines at small scale in the same district. We interpret these finding as follows.

Fig. 2
figure 2

a, b Variations of nighttime light (NTL) radiance in Wuhan from December 2019 to February 2020, as monthly averages. Green indicates less radiance whereas red denotes more radiance. c Sum of NTL radiance variations in each district in Wuhan

Full size image

The shift of human activities revealed by variation of nighttime light globally agree with changes in human activities. For instance, during the epidemic, the Wuhan government adopted the most rigorous blocking and quarantine policies, and strictly restricted personnel flow and economic activities. The passages from and to Wuhan were cut off, people were required to stop working and stay at home, all unnecessary gatherings and commercial activities were forbidden, only supermarkets in nearly all malls were allowed to open and daily opening hours were greatly shortened (Liu et al. 2020a). As a consequence, the malls and stores in the main commercial districts in the city center were closed, and personnel flow and traffic flow were reduced to a great extent. In addition, most college students did not return to schools due both to holidays and the epidemic in February, and also most schools were blocked.

This explains well our results showing a higher radiance decrease in center districts versus surrounding districts (Fig. 2). Indeed, the central districts are highly urbanized areas with business and bristling colleges and universities. For example, the higher radiance decrease of − 2570 nW/cm2/sr was observed in the Hongshan district, which contains prosperous business centers in the Optics Valley and 38 higher learning institutions. On the contrary, the surrounding, more rural districts showed an increase of radiance, up to 1516 nW/cm2/sr for the Jiangxia district. This could be explained by the presence of residential areas where citizens stayed at home during the epidemic instead of moving daily to the city center, in agreement with a previous report (Liu et al. 2020a).

Overall, variations of nighttime light radiance show both expected declines in the city center and less expected rises in surrounding districts. High resolution mapping also reveals heterogeneous variations at low scale in the same district, e.g., radiance increase in some areas and radiance decrease in other areas. Radiance decrease is explained by decreasing activities in the center, whereas radiance increase denotes staying-home people in surrounding districts.

Global air quality during the lockdown

We assessed air quality in Wuhan during lockdown using the air quality index (AQI), which is the highest value among air quality sub-indices of the following pollutants: SO2, NO2, CO, O3, PM10 and PM2.5 (Fig. 3, Table S4). Results showed that almost all data are within the two best air quality categories: I for superior air quality and II for good air quality. These findings are explained by the strict blocking policy in Wuhan which induced an extreme reduction of pollutant emissions, as expected. Yet, surprisingly, a mild pollution event still took place after lockdown, as previously observed by Lian et al. (2020). Specifically, air pollution increased from January 26 with mean AQI of 43 to February 5 with mean AQI of 126. On February 5, AQI rised from 79 at 00h00 to 183 at 11h00, then declined to 79 at 18h00, then rised again to 169 at 21h00, then declined. In addition, the primary pollutant on February 5 was PM2.5, meaning that PM2.5 provided the largest contribution to air quality degradation. Overall, we observed a slight pollution event peaking on February 5 during lockdown, which is likely due to agricultural production, domestic heating and solid waste combustion.

Fig. 3
figure 3

Daily air quality index (AQI) in Wuhan during the lockdown period from January 23 to April 7, 2020. Arrow shows a mild pollution episode on February 5. AQI is the maximum among air quality sub-indices of the following pollutants: SO2, NO2, CO, O3, PM10 and PM2.5. The daily AQI in Wuhan was obtained by averaging the AQI at 9 monitoring sites

Full size image

Correlation of air pollutants and meteorological factors

We correlated levels of hourly air pollutants during February 4–6 (Table S5 and S6). Results show that particulate matter levels, e.g., PM2.5 and PM10, increased with SO2, NO2 and CO concentrations with correlation coefficients ranging from 0.44 for NO2 to 0.92 for CO, with p lower than 0.01. This finding could be explained by emissions from sources generating both particulate matter and gases, such as fuel combustion, industrial and agricultural production. CO sources include motor vehicle exhausts, power plants, heating boiler, domestic stove and solid waste combustion. Since vehicle traffic and industrial activities have decreased during lockdown, the most likely sources are power plants, agricultural production, domestic heating and solid waste combustion. On the other hand, ozone levels were not correlated with particulate matter concentration, and ozone levels increased with decreasing NO2 levels, with a − 0.44 coefficient. Similar trends have been observed in Hangzhou during lockdown and are explained by chemical titration of ozone by nitrogen oxides, e.g., NO (Wang et al. 2020b).

We also correlated levels of hourly air pollutants and meteorological factors during February 4–6 (Table S6). Temperature was positively correlated with the concentrations of all pollutants except for O3, with a maximum coefficient of 0.68 for PM2.5, followed 0.57 for PM10 and 0.54 for CO. This is probably due to the fact that temperature favors most chemical reactions in the atmosphere. Relative humidity was negatively correlated with the concentrations of all pollutants except for O3, with a maximum correlation of − 0.40 with PM2.5. This is likely due to the removal of particles and gaseous pollutants by adsorption on water droplets followed by fall down and deposition at high humidity. For instance relative humidity was higher than 90% on February 6, thus explaining the low AQI at this date, of 57 (Fig. 3). Last, the slight positive correlations observed between wind speed and particulate matter (Table S6) suggests that wind transport contribute to the accumulation of particulate matter in some places.

Overall, particulate matter increases with SO2, NO2 and CO levels, suggesting a common source, but is not correlated with ozone, suggesting a different source. Ozone increases with decreasing NO2 levels, which is explained by chemical titration. Except for ozone, pollutant levels decrease with relative humidity, which suggests removal of pollutants by precipitation.

Temporal evolution of air pollutants

We compared the evolution of mean monthly concentration of PM2.5, PM10, SO2, CO, NO2 and O3 in Wuhan during January–April in 2018, 2019 and 2020 (Fig. 4, Table S7). In 2020, mean concentrations of PM2.5, PM10, NO2 decreased by 31.9%, 31.5%, 37.9%, respectively, versus 2019, and by 34.3%, 46%, 45.6%, respectively, versus 2018, similarly to a previous study (Xu et al. 2020). These findings are explained by blocking and shutdown policies adopted by the Wuhan government during the epidemic. The decrease of PM10 and PM2.5 concentrations is explained by less motor vehicle exhausts, industrial activity and biomass burning (Fig. 4a,b). Indeed, the sources of PM2.5 in Wuhan are 27.1% from motor vehicles, 26.8% from secondary sulfate and nitrate, 26.4% from the industry and 19.6% from biomass burning according to Cheng et al. (2012).

Fig. 4
figure 4

Monthly concentration of PM2.5 (a), PM10 (b), SO2 (c), CO (d), NO2 (e) and O3 (f). The monthly mean concentrations of PM2.5, PM10 and NO2 were decreased obviously, in comparison with those in the same period of 2019 and 2018. The monthly CO concentration also reduced in comparison with those in the same period of 2019 and 2018, while the decrease amplitude was small. High concentration of O3 was found during the lockdown

Full size image

While SO2 levels were much lower in January 2020 versus 2018 and 2019, SO2 levels increased with time in 2020 to exceed 2018–19 levels in April (Fig. 4c). This finding is explained by the fact that the fire coal industry never stopped operation during lockdown, because coal combustion generates gaseous SO2. The rising SO2 in 2020 may also result from the increase in coal consumption for domestic heating (Wang et al. 2020b).

The decline of mean CO levels in 2020 was rather lower than most pollutants, of 17.0% versus 2018 and 18.7% versus 2019 (Fig. 4d). This is explained by the fact that shutdown policies have been specifically applied to labor-intensive light industry, whereas thermal power generation, steel and coking industries were not stopped. Mean NO2 levels decreased sharply in 2020 with a 45.6% drop versus 2018, and a 37.9% drop versus 2019 (Fig. 4e). This is due to the cessation of vehicle traffic and, in turn, to the sharp reduction of traffic emissions (Huang et al. 2020; Xu et al. 2020).

Mean ozone levels in 2020 increased by 32.1% versus 2018 and by 43.6% versus 2019 (Fig. 4f). Ozone levels also increased from January to April. Ozone rise during lockdown have been previously also observed in Wuhan and other cities (Huang et al. 2020; Wang et al. 2020b; Xu et al. 2020). Both ozone rises in 2020 versus 2018–19 and from January to April are explained by lower NOx pollution. Indeed, near-ground ozone is mainly produced under light conditions by precursors such as CO, NOx and volatile organic compounds (VOCs) (Huang et al. 2020). This explains the ozone increase from January to April due to rising daylight time period (https://www.timeanddate.com/sun/china/wuhan). The increase of ozone during lockdown is explained by the decrease in NOx. Indeed, the production of ozone from CO is initiated by free radicals such as the hydroxyl radical as follows; a similar reaction occurs by replacing CO by VOCs (https://en.wikipedia.org/wiki/Tropospheric_ozone):

$$\cdot{\text{OH }} + {\text{ CO }} \to \, \cdot{\text{HOCO}}$$
$$\cdot{\text{HOCO }} + {\text{ O}}_{{2}} \to {\text{ HO}}_{{2}}\cdot \, + {\text{ CO}}_{{2}}$$
$${\text{HO}}_{{2}}\cdot + {\text{ NO }} \to \cdot {\text{OH }} + {\text{ NO}}_{{2}}$$
$${\text{NO}}_{{2}} + {\text{ h}}\nu \to {\text{ NO }} + {\text{ O}}\left( {^{{3}} {\text{P}}} \right)$$
$${\text{O}}\left( {^{{3}} {\text{P}}} \right) \, + {\text{ O}}_{{2}} \to {\text{ O}}_{{3}}$$

During periods of high pollution, ozone production is moderated by the excess of NOx, because NOx reacts with radicals and thus decrease ozone yield. Whereas during lockdown, less radicals are scavenged by NOx because NOx levels are lower, and, in turn ozone production is higher under sunlight. High ozone levels are also associated with low levels of particulate matter and strong solar radiation as a result of the “weekend effect” (Murphy et al. 2007). Reports of ozone pollution in other cities during the lockdown have reached similar conclusions (Li and Tartarini 2020; Wang et al. 2020b).

Overall, PM2.5, PM10 and NO2 were highly reduced in the 2020 lockdown versus 2018 and 2019, which is explained by the halt of most activities. CO levels decreased to a lesser extent, which is probably due to the fact that steel and coking power industries did not stop. SO2 levels displayed a high reduction in January 2020 but increased and exceeded 2018–19 values in April 2020, which is attributed to the ongoing coal industry and in coal consumption for domestic heating. Ozone levels were higher in 2020 versus 2018–19, which is explained by less NOx pollution because NOx reacts with and inhibits ozone precursors.

Conclusion

We have studied the spatial and temporal variations of nighttime light and air pollutants during the COVID-19 lockdown in Wuhan. The spatial variations of nighttime light radiance are heterogenous at local scale. The most darkened area during the lockdown was Wuhan downtown due the halt of most companies, services and associated human labour. Yet, an increase of radiance in the five surrounding districts by home-staying people was observed. The daily variation of AQI in Wuhan during the lockdown period showed that almost all data were within the two best air quality categories. Analysis of the February 5 pollution event revealed positive correlations of particulate matter, SO2, NO2 and CO, and a negative correlation between O3 and NO2. Ozone increased with decreasing NO2 levels, which is explained by chemical titration. Regarding the relationship between air pollutants and meteorological factors during the pollution event, temperature was positively correlated with the concentrations of all pollutants except for O3. Particulate matter, SO2, NO2 and CO decreased with relative humidity, suggesting removal by precipitation. The comparison of 2020 mean data during January–April with previous years shows that PM2.5, PM10 and NO2 declined sharply in 2020; the decrease of CO was small; SO2 displayed a high reduction in January 2020 but increased and exceeded 2018–19 values in April 2020; due to less NOx pollution, ozone levels were higher in 2020 versus 2018–19.