Study on the variation of air pollutant concentration and its formation mechanism during the COVID-19 period in Wuhan

Abstract

To prevent the spread of COVID-19 (2019 novel coronavirus), from January 23 to April 8 in 2020, the highest Class 1 Response was ordered in Wuhan, requiring all residents to stay at home unless absolutely necessary. This action was implemented to cut down all unnecessary human activities, including industry, agriculture and transportation. Reducing these activities to a very low level during these hard times meant that some unprecedented naturally occurring measures of controlling emissions were executed. Ironically, however, after these measures were implemented, ozone levels increased by 43.9%. Also worthy of note, PM_2.5 decreased 31.7%, which was found by comparing the observation data in Wuhan during the epidemic from 8th Feb. to 8th Apr. in 2020 with the same periods in 2019. Utilizing CMAQ (The Community Multiscale Air Quality modeling system), this article investigated the reason for these phenomena based on four sets of numerical simulations with different schemes of emission reduction. Comparing the four sets of simulations with observation, it was deduced that the emissions should decrease to approximately 20% from the typical industrial output, and 10% from agriculture and transportation sources, attributed to the COVID-19 lockdown in Wuhan. More importantly, through the CMAQ process analysis, this study quantitatively analyzed differences of the physical and chemical processes that were affected by the COVID-19 lockdown. It then examined the differences of the COVID-19 lockdown impact and determined the physical and chemical processes between when the pollution increased and decreased, determining the most affected period of the day. As a result, this paper found that (1) PM_2.5 decreased mainly due to the reduction of emission and the contrary contribution of aerosol processes. The North-East wind was also in favor of the decreasing of PM_2.5. (2) O₃ increased mainly due to the slowing down of chemical consumption processes, which made the concentration change of O₃ pollution higher at about 4 p.m.–7 p.m. of the day, while increasing the concentration of O₃ at night during the COVID-19 lockdown in Wuhan. The higher O₃ concentration in the North-East of the main urban area also contributed to the increasing of O₃ with unfavorable wind direction.

1. Introduction

The COVID-19 pandemic has significantly challenged our daily life (Lonergan et al., 2020; Shereen et al., 2020). For public health, the Chinese Government ordered its highest Class 1 Response (Hubei Provincial People's Government, 2020), which is explained in The National Emergency Plan for Public Health Emergencies (The Central People's Government of the People's Republic of China, 2006). With this order, all the unnecessary transportation in and around Wuhan was shut down. All the unnecessary human activities were reduced to the minimum to reduce transmission and avoid cross-infection, including closing down local businesses, schools, colleges and universities (Zhou et al., 2020; Tang et al., 2020).

In another way, the COVID-19 lockdown in Wuhan is also an unprecedented emission mitigation measure that represents opportunities to understand air pollution in extreme cases. It has been widely accepted that PM_2.5 reduced about 40% in the lockdown conditions in Wuhan compared with the last few years (IQAir, 2020; Le et al., 2020; Ministry of Ecology and Environment of the People's Republic of China, 2020). In addition to the changes to the PM_2.5, O₃ was found to have about a 30% increase during the lockdown (Le et al., 2020, Ministry of Ecology and Environment of the People’s Republic of China, 2020). As for NO₂, Le et al. (2020) found a −93% decrease from satellite data, while the Ministry of Ecology and Environment of the People's Republic of China (2020) shows about a 40% decrease from the national ground station.

As described in previous studies, one of the main reasons causing the reduction in PM_2.5 might be the reduction of emissions. Vieno et al. (2015) used the EMEP4UK (UK -scale chemistry-transport model) atmospheric chemistry transport model to investigate the impact of the reductions in PM_2.5 anthropogenic emissions and found that the reductions of primary PM_2.5 emissions might be the most effective single-component control on PM_2.5. Liang et al. (2016) summarized the previous studies and concluded that industrial emission which induced secondary inorganic aerosols were the most dominant sources of PM_2.5 in urban areas in China. Many emission reduction campaigns were conducted in China to avoid air pollution, with Wang et al. (2009) and Li et al. (2011) having done a study about the emission reduction campaigns during the “Olympics Blue” in 2008. Sun et al. (2016), Huang et al. (2015) completed studies about the “APEC Blue” in 2014 and Han et al. (2016), Chu et al. (2018) and Ren et al. (2019) conducted a study about the “Parade Blue” in 2015. In these cases, China closed factories, industrial plants, construction sites, gas stations and kept vehicles off of the road in order to avoid air pollution, with the emission reduction campaigns proving to be effective. The aerosol extinction coefficient decreased to about 42.3% during the Beijing Olympic Games in 2008 compared with that in 2007. This study indicated the effectiveness of local air pollution control measures in Beijing areas under almost the same meteorological conditions (Yang et al., 2010). During the APEC periods in 2014, air pollutant concentrations had shown significant decreases over the North China Plain, especially over the Jing-Jin-Ji region, with NO₂ VCD (vertical column densities), AOD (aerosol optical depth), and AAOD (absorption aerosol optical depth) mostly reduced in Beijing, resulting in percentages of 47%, 34%, and 17% compared with that in the previous three years (Huang, 2015). Lin et al. (2017) found that daily PM_2.5 concentrations decreased from 98.57 μg/m³ to 47.53 μg/m³ during “APEC Blue”, and from 59.15 μg/m³ to 17.07 μg/m³ during the “Parade Blue”, using the same dates from the prior year as a reference. Recently, Wang et al. (2020) used CMAQv5.0.1 to simulate air pollution in China during January 1 to February 12, 2020 and found that the decrease of PM_2.5 in Wuhan was 30.79 μg/m³ when the emissions of transportation, industry and agriculture decreased to 20%.

More specifically, the literature indicates that when O₃ production is in a NO_x-saturated state (NO_x = NO + NO₂), a reduction in NO_x leads to an increase in ozone and the lack of NO emissions alleviates ozone titration (Le et al., 2020; Levy et al., 2014; Atkinson et al., 2000). Surface O₃ is normally low at night when NO emissions are high. However, during the daytime, the significant removal of ozone via reaction ($NO+O_3\to N{O}_2$) occurs in the vicinity of large NO emission sources (Kleinman et al., 2000; Lin et al., 1998). In a recent study, the lack of NO_x due to the reduction of emissions during the COVID lockdown in China led to substantial increases in O₃ (Huang et al., 2021). Given that PM_2.5 decreased and O₃ increased during the COVID-19 lockdown period in Wuhan, several questions come to mind. What are the differences and formation mechanisms of their chemical and physical processes in the atmosphere between the pandemic and those of normal years? How can we quantify the impacts of the COVID-19 lockdown on the chemical and physical processes? What are the differences of the COVID-19 lockdown impacts resulting in the pollution increasing or decreasing? How to find the most affected period of the day instigated by the COVID-19 lockdown?

To fill a literature gap in studying the influence of the COVID-19 lockdown (Huang et al., 2021; Le et al., 2020; Wang et al., 2020), this study focused on emission reduction ratio through examining partial differentiations in individual species chemical and physical processes. The study utilized the CMAQv5.3.1 (Byun and Schere, 2006) to conduct sensitivity simulation tests with four sets of emission data collected in Wuhan during the lockdown period. The study, which the individual pollutant actions show is more complicated than the previous studies, was demonstrated through implementing the PROCAN (Process Analysis Preprocessor) module (Byun and Ching, 1999). This module further revealed the quantitative effects of the individual chemical and physical processes, which instigated the mechanism for the changes in PM_2.5 and O₃ in Wuhan during the COVID-19 lockdown time. This paper found that PM_2.5 decreased mainly due to the reduction of emission and the positive contribution of aerosol processes. O₃ increased mainly due to the slowing down of chemical consumption processes, which made the concentration change of O₃ pollution higher at about 4 p.m.–7 p.m. of the day, and increased the concentration of O₃ at night during the COVID-19 lockdown in Wuhan.

2. Data and method

2.1. Data

This study used data from the 10 air quality stations in Wuhan from China National Environmental Monitoring Center. The FNL (Final Operational Global Analyses) 1 ° × 1 ° data that was produced by the National Centers for Environmental Prediction (NCEP), was used in this simulation to initialize the WRF (The Weather Research and Forecasting) model, which can be downloaded from the website in https://rda.ucar.edu/datasets/ds083.2/. The MEIC (Multi-resolution Emission Inventory for China) was developed and maintained by Tsinghua University (Zhang et al., 2009), which can be downloaded from this website in https://meicmodel.org.

3. Method

In this study, a WRF-CMAQ modeling system was applied to simulate the pollution. The Weather Research and Forecasting (WRF) Model was developed at the National Center for Atmospheric Research (NCAR), which is operated by the University Corporation for Atmospheric Research (UCAR) (Chen et al., 2007). The WRFv3.7.1 was used to generate the meteorological background for air quality simulation. The Community Multiscale Air Quality (CMAQ) modeling system is an active open-source development project of the U.S. EPA that consists of a suite of programs for conducting air quality model simulations (United States Environmental Protection Agency, 2020; Byun and Schere, 2006; Pye et al., 2017; Fahey et al., 2017). The CMAQv5.3.1 was used to simulate the spatial distribution and temporal variation of pollution within the study region from Dec. 13, 2019 to Jan. 15, 2020 and from Feb. 5, 2020 to Apr. 8, 2020. A twenty-days gap from 15th Jan. to Feb. 5, 2020, around Jan. 23, 2020 when the lockdown occurred, was left to better analyze the influence of the lockdown.

The PROCAN (Process Analysis Preprocessor) module was implemented in CMAQ by Byun and Ching (1999) which is an accounting system that tracks the quantitative effects of the individual chemical and physical processes, which were combined to explain the predicted hourly species concentrations from a simulation. The PROCAN module helped calculate integrated process rates and integrated reaction rates, which can then be used for diagnosing the physical and chemical behavior of these pollution processes. PROCAN has two components: Integrated Process Rate (IPR) analysis, and Integrated Reaction Rate (IRR) analysis, IPR was mainly used in this study.

The detailed WRF-CMAQ model configurations can be found in Table 1 (a) and Table 1 (b). The model simulation domain and topography height are shown in Fig. 1 . The scheme we used to calculate PM_2.5 followed the CMAQv5.3.1 mechanism cb6r3_ae7_aq rules, can be found at https://github.com/USEPA/CMAQ/blob/master/CCTM/src/MECHS/cb6r3_ae7_aq/SpecDef_cb6r3_ae7_aq.txt. There are 9 p.m._2.5 chemical or physical processes assigned by PROCAN, including HADV(horizontal advection), ZADV(vertical advection), HDIF(horizontal diffusion), VDIF(vertical diffusion), DDEP(dry deposition of species), CLDS(change due to cloud processes; includes aqueous reaction and removal by clouds and rain), AERO(change due to aerosol processes), CHEM(net sum of all chemical processes for species over output step) and EMIS(emissions contribution to concentration). The CHEM of PM_2.5 process analysis in the CMAQ model represents the heterogeneous reactions. The aerosol species involved in reactions in the mechanism definition file are listed here: https://github.com/USEPA/CMAQ/blob/master/CCTM/src/MECHS/cb6r3_ae7_aq/mech_cb6r3_ae7_aq.def. The production and loss of each of these reactions can be found in https://github.com/USEPA/CMAQ/blob/master/CCTM/src/gas/ebi_cb6r3_ae7_aq/hrprodloss.F. There are seven O₃ chemical or physical processes signed by PROCAN, including HADV, ZADV, HDIF, VDIF, DDEP, CLDS, CHEM.

Table 1(a).

WRF model configurations.

WRFv3.7.1
Vertical resolution	33 vertical levels
Microphysics scheme	WSM 3-class simple ice scheme
Boundary layer scheme	YSU scheme
Surface layer scheme	MM5 scheme
Land-surface scheme	Unified Noah land-surface model
Longwave radiation scheme	rrtm scheme
Shortwave radiation scheme	Dudhia scheme
Grid-nudging fdda	on
Domain center	31.0°N,112.5°E
Domain id	1	2	3
Domain size	57 × 57	91 × 67	64 × 70
Starting IJ-indices from the parent domain	×	(13,19)	(57,13)
Horizontal resolution	27 km	9 km	3 km

Table 1(b).

CMAQ model configurations.

CMAQv5.3.1
Horizontal advection	Yamo
Vertical advection	WRF
Horizontal diffusion	Multiscale
Vertical diffusion	ACM2
Deposition	M3Dry
Chemistry solver	EBI
Aerosol module	AERO7
Cloud module	ACM
Mechanism	cb6r3_ae7_aq
Domain id	1	2	3
Domain size	54 × 54	93 × 93	61 × 67

Fig. 1.

Simulation domain and topography height (a. Domain 1- central region of China, Domain 2-Hubei Province and Domain 3-Wuhan; b. Domain 3 and location of 10 air quality stations).

4. Observation analysis and model assessment

4.1. Observation analysis

Literature was reviewed to report the studies on the behavior of air pollution when the lockdowns were implemented (Table 2 ). Please note that all the observation data came from the national ground station except no.6, which used the data from TROPOMI (The Tropospheric Monitoring Instrument). As shown in Table 2, these studies reached a consensus that PM_2.5 reduced 30%–50% in the lockdown conditions compared with the preceding years. In contrast to the changes to the PM_2.5, Le et al. (2020) and Ministry of Ecology and Environment of the People’s Republic of China (2020) found an approximately 30% increase in O₃ during the lockdown. As for NO₂, Le et al. (2020) found a −93% decrease using the satellite, and other reports (Ministry of Ecology and Environment of the People's Republic of China, 2020) which show about a 40% decrease from the national ground station.

Table 2.

The observation results of correlational studies.

	Pollutant	Reduction	lockdown dates,2020	Compared dates, same time periods	Author
1	PM2.5	−44%	03th Feb.~ 24th Feb.	2019	IQAir (2020)
2	PM2.5	−50%		Prior 4 years average
3	PM2.5	−32.4%	23rd Jan. ~ 13th Feb.	Prior 5 years average, Chinese lunar calendar	Le et al. (2020)
4	PM2.5	−43.5%		Prior 5 years average, Georgian Calendar
5	O3	+25.1%		Prior 5 years average, Chinese lunar calendar
6	NO2(TROPOMI data)	−93%		2019
7	PM2.5	−38%	24th Jan. ~ 10th Feb.	2019	Ministry of Ecology and Environment of the People’s Republic of China, 2020
8	NO2	−45%
9	O3	+34.3%	20th Jan.~ 20th Feb.	2017–2019	Ministry of Ecology and Environment of the People’s Republic of China, 2020
10	NO2	−31.1%
11	NO	−50%

In this paper, the observations of ten air quality stations average over space and time were compared in Wuhan from the China National Environmental Monitoring Center. Considering that there must have been some changes in emissions between 2019 and 2020 in Wuhan, we separated the observations to four time periods. Then the observations periods: Dec. 15, 2018 to 15th; Jan. 2019 and Feb. 08, 2019; Apr. 08, 2019, were used to do comparative analysis. The observations from Dec. 15, 2019 to Jan. 15, 2020 were considered, however, as there was no emission influence from COVID-19 lockdown which had not yet begun, we only noted the differences between 2020 and 2019. For the other observation periods, were reviewed through the lens that there was causality from the COVID-19 lockdown compared to the observations in the same period in 2019. Table 3 shows the differences of PM_2.5, PM₁₀(inhalable particles), the ratio of PM_2.5/PM₁₀, O₃ and NO₂ between these 4 time periods. The calculate method is used as follows:

$$D=\frac{\left(\frac{\sum _1^{N2}Data2}{N2}-\frac{\sum _1^{N1}Data1}{N1}\right)}{\frac{\sum _1^{N1}Data1}{N1}}\times 100\%$$ (1)

where Data1, Data2 represents the observations (PM_2.5, PM₁₀, the ratio of PM_2.5/PM₁₀, O₃ and NO₂) from different time periods, N1, N2 represents the number of samples in Data1, Data2.

Table 3.

The differences of observations.

	Time periods of Data1	Time periods of Data2	DPM2.5	DPM10	DPM2.5/PM10	DO3	DNO2
A1	Feb. 08, 2019~ Apr. 09, 2019	Feb. 08, 2020~ Apr. 08, 2020	−31.7%	−31.8%	0.2%	43.9%	−51.8%
A2	Dec. 15, 2018~ Jan. 15, 2019	Dec. 15, 2019~ Jan. 15, 2020	−26.7%	−5.3%	−22.5%	−8.8%	−0.4%
B1	Dec. 15, 2019~ Jan. 15, 2020	Feb. 08, 2020~ Apr. 08, 2020	−42.2%	−35.9%	−9.7%	245.4%	−53.1%
B2	Dec. 15, 2018~ Jan. 15, 2019	Feb. 08, 2019~ Apr. 09, 2019	−37.9%	−11.1%	−30.2%	119.0%	−3.0%

Table 3 shows the differences of observations, A1 represents the differences of the air pollutant concentration during the COVID-19 lockdown in 2020 and the same time period in 2019. A2 represented the differences before the COVID-19 lockdown in 2020 and the same time period in 2019. B1 represented the differences before and during the COVID-19 lockdown in 2020. B2 represented the differences of the same time period as B1 in 2019. PM_2.5 decreased 31.7% from Feb. 08, 2020 to Apr. 08, 2020 compared with 2019, and O₃ increased 43.9%, NO₂ decreased 51.8%. The concentration of PM_2.5, PM₁₀ and NO₂ were lower in lockdown time period A2 than before the lockdown time period A1, meanwhile O₃ was higher. The results above show that the lockdown in Wuhan had a great influence on air pollution. Compared with B1 and B2, it's clear that PM_2.5 and PM₁₀ continually had a downward trend from December to April, and it's going down more in 2020. O₃ continually had an upward trend from December to April, but it's going up much more in 2020. NO₂ should not have changed much from December to April, but it decreased 53.1% in 2020. Our results are in alignment with the previous studies that PM_2.5 reduced 30%–50%. Beyond that, O₃ in our results increased 43.9%, much more than previous studies. The reasons might be that our observation periods lasted much longer until the lockdown was over in Wuhan. The results of observations show that the influence to O₃ of the lockdown in Wuhan was more significant when the concentration of O₃ is higher. There is an upward trend of O₃ from February to April in both 2019 and 2020, with the concentration of O₃ being higher in April compared to February, therefore there is a larger inferred increase in O3 levels when a longer observational period was compared.

Fig. 2 shows the concentrations of the four species pollutants from December 2018–April 2019 and December 2019–April 2020(considering that 2020 is a leap year, we used the Julian date here). It is clear that PM_2.5 and PM₁₀ had a continuing downward trend and O₃ had a continuing upward trend from December to April in both 2019 and 2020. The concentration of PM_2.5, PM₁₀ and NO₂ was lower in 2020 after February. Meanwhile the concentration of O₃ was much higher.

Fig. 2.

Comparison of observations of PM_2.5 (a), PM₁₀(b), O₃(c) and NO₂(d) in 10 stations average in Wuhan from Dec. 2018 to Apr. 2019 and Dec. 2019 to Apr. 2020.

4.2. Model assessment

In order to assess the performance of the CMAQ simulation and make this study more convincing, the Eva experiment scenario was carried out to simulate the pollution from Dec. 15, 2019 to Jan. 15, 2020, before the COVID-19 lockdown. This experiment used an adjusted anthropogenic emission inventory that is based on MEIC 2016. It was multiplied by coefficients compared with observations to be an alternative solution when the real-time emission inventory is unavailable. The simulated pollutants concentrations of the Eva experiment were compared to the observations of the ten air quality stations. We cannot directly evaluate our simulation during the COVID-19 lockdown in Wuhan because we cannot get the accurate emission inventory during that time. Therefore, the Eva experiment was used to evaluate the simulation of the CMAQ model and find out the suitable emission inventory, which confirmed that the simulation presented a reliable performance before the COVID-19 lockdown in Wuhan. The Emission Control module in CMAQv5.3.1 was used here to adjust the emission inventories in 2020. First of all, NO_x and SO₂ was adjusted in the emission inventory and compared to the simulation results with the observation of NO₂ and SO₂ (Formula 2~5 were used here).Then, the total amount of VOC and PM_2,5 was adjusted. Table S1 listed the adjust ratios and Table S2 listed the performance of EVA0 (without adjustment) and EVA6 (the emission inventory we used in the following paper) simulation from Nov. 15, 2019 to Jan. 14, 2020 compared with the observation in the average of the ten stations. The correlation coefficient (COR), root mean squared error (RMSE), normalized mean bias (NMB) and normalized mean error (NME) were calculated as follows:

$$COR=\frac{Cov(C_m-C_0)}{\sqrt{D\left(C_m\right)}\sqrt{D\left(C_0\right)}}$$ (2)

$$RMSE=\sqrt{\frac{\sum _1^N{(C_m-C_0)}^2}{N-1}}$$ (3)

$$NMB=\frac{\sum _1^N(C_m-C_0)}{\sum _1^N{C}_0}\times 100\%$$ (4)

$$NME=\frac{\sum _1^N\left|C_m-C_0\right|}{\sum _1^N{C}_0}\times 100\%$$ (5)

where C _m is the simulated concentration, C ₀ is the observed data, N represents the number of samples, Cov(x) means the covariation of x and D(x) means the variance of x.

All the data in this study we simulated by the WRF-CMAQ model was in the surface layer and analyzed by the daily average of the ten stations’ in Wuhan. Table 4 shows the detailed model assessment. Fig. 3 compared the observation and simulation of the Eva experiment of PM_2.5, PM₁₀, O₃ and NO₂ from Dec. 15, 2019 to Jan. 15, 2020. Compared with the observations, our simulation presented a strong performance.

Table 4.

Model assessment.

	PM2.5	PM10	O3	NO2
COR	0.49	0.57	0.72	0.58
RMSE	33.47	35.91	8.16	13.37
NMB	23.8%	16.6%	−13.8%	11.1%
NME	43.1%	36.3%	35.8%	20.2%

Fig. 3.

Comparison of the observation and Eva experiment of PM_2.5 (a), PM10(b), O₃(c) and NO₂(d) in 10 stations average in Wuhan from Dec. 15, 2019 to Jan. 15, 2020.

5. Simulation and results analysis

5.1. The average characterics of meteorological factors during the COVID-19 lockdown

Fig. 4 shows the average temperature, sea level pressure and relative humidity with wind in Wuhan from Feb. 08, 2020 to Apr. 08, 2020 simulated using CMAQ model. The average wind direction during the COVID-19 lockdown was from North-East to South-West. The temperature was approximately 12$°C$ in the main urban area of Wuhan, and the temperature was lower where the surface is water (including the Yangtze River, Hanjiang River and lakes in Wuhan). The sea level pressure was approximately 1020.2hpa in the main urban area of Wuhan. The relative humidity was approximately 67% in the main urban area of Wuhan, and was higher where the surface is water.

Fig. 4.

The simulated average temperature(a), sea level pressure(b) and relative humidity(c) with wind in Wuhan from Feb. 08, 2020 to Apr. 08, 2020.

5.2. Simulation schemes and results for emission control

5.2.1. Simulation schemes for emission control

In order to investigate the greatest possible emissions reduction ratio caused by the COVID-19 lockdown, Eva and the other four simulation scenarios were performed and compared. As seen in Table 5 , the Base (stand for baseline) experiment was regarded as the benchmark experiment, and Exp1 (stand for experiment 1), Exp2 (stand for experiment 2), Exp3 (stand for experiment 3) were regarded as the sensitivity experiments with different decreasing ratios in different emission categories. The MEIC emission inventory was used here and the ratios of PM_2.5, PM₁₀, NO_X, and VOC in MEIC emission inventory can be found in Fig. S1.

Table 5.

Scheme for emission control (the percentages mean remaining emissions in simulations).

Experiment	Time period	Agriculture	Industry	Power	Residential	Transportation
Eva	Dec. 15, 2019~ Jan. 15, 2020	100%	100%	100%	100%	100%
Base	Feb. 08, 2020~ Apr. 08, 2020	100%	100%	100%	100%	100%
Exp1	Feb. 08, 2020~ Apr. 08, 2020	50%	50%	100%	100%	50%
Exp2	Feb. 08, 2020~ Apr. 08, 2020	20%	20%	100%	100%	20%
Exp3	Feb. 08, 2020~ Apr. 08, 2020	10%	20%	100%	100%	10%

First, the Base experiment scenario used the same emission inventory as the Eva experiment in section 3.2 but simulated the time period in COVID-19 lockdown from Feb. 8, 2020 to Apr. 8, 2020, assuming that COVID-19 lockdown had no effect on the emissions in Wuhan. As seen in Fig. 5 , the concentration of PM_2.5, PM₁₀, and NO₂ was much higher than observation, meanwhile the concentration of O₃ was much lower.

Fig. 5.

Comparison of the observation and four experiments of PM_2.5 (a), PM₁₀ (b), O₃ (c) and NO₂ (d) in 10 stations average in Wuhan from Feb. 08, 2020 to Apr. 08, 2020.

Then, Exp1 was carried out and the emissions were decreased to 50% in agriculture, industry and transportation in Wuhan compared with the Base experiment. Considering that with the highest Class 1 Response, all the unnecessary transportation were shut down and all the unnecessary human activities were reduced to the minimum, including closing down local business, schools, colleges, universities, restricting the movement of people (Hubei Provincial People's Government, 2020) and carrying out agricultural production in different periods and batches (Agricultural and rural Bureau of Wuhan, 2020). The results in Fig. 5 shows that the decreasing ratio of agriculture, industry and transportation was still low.

Next, Exp2 was carried out, the emissions were decreased to 20% in agriculture, industry and transportation in Wuhan compared with the Base experiment. It is obvious that the concentration of PM_2.5 and PM₁₀ was similar to observation, but the concentration of NO₂ was still a little high, while O₃ was still a little low. Wang et al. (2020) did the similar experiment but cut the emissions in Hubei.

Finally, in order to get a better performance in simulating O₃, Exp3 was carried out and the emissions were decreased to 20% in industry and 10% in transportation and agriculture in Wuhan compared with the Base experiment. Considering that PM_2.5 and PM₁₀ were well simulated and the differences of ratios in the MEIC emission inventory between PM_2.5, PM₁₀ and NO_x mainly came from transportation and agriculture, Exp3 cut more in transportation and agriculture compared with Exp2.

The results of emission control experiments demonstrated that Exp3 had the best performance on both PM_2.5 and O₃. The COR, RMSE, NMB and NME of PM_2.5 were 0.28, 17.07(ug/m³), 1.9% and 38.1%. The COR, RMSE, NMB and NME of O₃ were 0.62, 14.45(ug/m³), −7.1% and 18.1% compared with observation. Our results were better than the previous study (Wang et al., 2020). As the results show, Exp3 presented great performance in simulating PM_2.5 and O₃ during the COVID-19 lockdown in Wuhan. Therefore, in our study, the greatest possible emissions were decreased to 20% in industry and 10% in agriculture and transportation during the COVID-19 lockdown period compared with the emissions before the lockdown period in Wuhan.

5.2.2. Variety characterics of pollutant concentration for emission control

The horizontal distribution of the average simulated PM_2.5, PM₁₀, O₃, O₃–8H (max 8-h average O₃) and NO₂ concentration from Feb. 08, 2020 to Apr. 08, 2020 in Base, Exp3 and their difference was shown in Fig. 6 . It can be found that the biggest differences of all the pollutants (as Fig. 6(c), (f), Fig. 6(i), (o)) occurred in the main urban area, which can be represented by the location of the national air quality stations in Wuhan. The results reflect that the COVID-19 lockdown influenced the pollution more in urban areas in Wuhan. The spatial and concentration differences of O₃ and O₃–8H revealed that the COVID-19 lockdown influenced more on the low level of O₃ in the urban area of Wuhan. As the max 8-h average O₃ usually occurs in the afternoon of the day, and the low level of O₃ always occurs at night, it can be deduced that COVID-19 had more influence on the low level of O₃ at night. The further analysis was carried out in section 4.2.5.

Fig. 6.

The averaged simulated concentration horizontal distribution during Feb. 08, 2020 to Apr. 08, 2020. PM_2.5 of Base (a), Exp3 (b) and differences (c). PM₁₀ of Base (d), Exp3 (e) and differences (f). O₃ of Base (g), Exp3 (h) and differences (i). O₃–8H of Base (j), Exp3 (k) and differences (l). NO₂ of Base (m), Exp3 (n) and differences (o). The differences were defined as Exp3 minus Base. The vector indicate wind, the black points means the location of 10 air quality stations.

5.3. Formation mechanism study baselined on processes analysis

5.3.1. Formula of processes analysis

In order to find out the formation mechanism that caused significant changes in PM_2.5 and O₃ during the lockdown time period, the Process Analysis module was used in CMAQv5.3.1. Formula 6~9 shows how to calculate the concentration of PM_2.5 and O₃ using process analysis. Table S3 shows the species index used in process analysis of PM_2.5 and O_3.

$$CH{A}_t=\left\{\begin{array}{c}{P}_{ZADV,t}+P_{HADV,t}+P_{HDIF,t}+P_{VDIF,t}+P_{DDEP,t}\\ +P_{CLDS,t}+P_{CHEM,t}+P_{EMIS,t}+P_{AERO,t}\left(P{M}_{2.5}\right)\\ P_{ZADV,t}+P_{HADV,t}+P_{HDIF,t}+P_{VDIF,t}+P_{DDEP,t}\\ +P_{CLDS,t}+P_{CHEM,t}\left(O_3\right)\end{array}\right\}$$ (6)

$$CH{A}_d=\sum _{t=24\times (d-1)}^{24\times d}CH{A}_t$$ (7)

$$C_t=C_{t-1}+CH{A}_t=C_0+\sum _1^{t}CH{A}_t$$ (8)

$$C_d=\frac{\sum _{t=24\times \left(d-1\right)}^{24\times d}{C}_t}{24}=\frac{\sum _{t=24\times \left(d-1\right)}^{24\times d}\left[C_{24\times \left(d-1\right)}+\sum _{\text{t}=24\times \left(d-1\right)}^{t}CH{A}_t\right]}{24}$$ (9)

where CHA _t means the total concentration change of all processes in hour t, P _process,t means the change due to process in hour t. CHA _d means the total daily concentration change of all processes in day d,$C_t$ means the concentration of PM_2.5 or O₃ in hour t,$C_d$ means the daily average concentration of PM_2.5 or O₃.

5.3.2. Daily average processes analysis

Fig. 7 shows the daily average process analysis of PM_2.5 and O₃ of Base and Exp3. As declared in section 2.2, for example, ZADV_PM_2.5 means the daily summation of PM_2.5 changes due to vertical advection in every hour, whereas, CHA_PM_2.5 means the daily change of PM_2.5, which is the summation of each hour and each process, where PM_2.5 means the daily average PM_2.5. It should be noted that the hourly concentration of PM_2.5 is equal to the summation of the initial concentration and changes of all the processes in this hour, which can be seen in Fig. S2.

Fig. 7.

Daily change of each process and their summation and daily average PM_2.5 and O₃ in the average over 10 stations in Wuhan from Feb. 08, 2020 to Apr. 08, 2020 (a.PM_2.5 from Base, b.PM_2.5 from Exp3, c.O₃ from Base, d.O₃ from Exp3).

As for PM_2.5, it's clear that emissions contributions always play an important role in increasing the concentration (Vieno et al., 2015) and had a stable performance during the whole simulation. Vertical diffusion, horizontal advection and vertical advection dominated the decreasing of PM_2.5, and it is clear that pollution always occurs while VIDF or HADV slows down (Hua et al., 2016). ZADV often had the opposite tendency compared with VDIF and HADV, and it influenced much more in Exp3 compared to Base. Aerosol processes increased pollution in the Base experiment, but on the contrary, reduced the pollution in Exp3 on average, and AERO was lower on the decreasing time of PM_2.5 and even reduced the pollution in Exp3. HDIF, DDEP, CLDS and CHEM gave a little contribution to PM_2.5. HDIF decreased the concentration a little while PM_2.5 increased and almost equal 0 while PM_2.5 decreased. The DDEP was higher, though it's similar in the Base and Exp3. CLDS is always equal to 0, but it decreased the concentration while there was rain or high humidity. The CHEM always increased the concentration and higher while PM_2.5 increased, and it's similar in the Base and Exp3. The CHEM of PM_2.5 is connected with heterogeneous reactions. The heterogeneous reactions can increase the concentration of PM_2.5.

Comparing Exp3 with the Base experiment, there is a conspicuous decrease in EMIS contributions that then lead to the decreasing of all processes. Under the influence of all the decreasing processes, the concentration of PM_2.5 stayed low during the lockdown in Wuhan.

As for other findings of note, in terms of O₃, vertical diffusion and horizontal advection dominated the increase of O₃, while the chemical processes dominated the decrease, as Hogrefe et al. (2018) and Tzella (1983) described in their studies. CHEM always had the similar tendency but had an opposite contribution compared with the summation of VDIF and HADV. Also, dry deposition always decreased the concentration of O₃. Strikingly, the cloud processes and horizontal diffusion gave little contribution. As seen in Fig. 7, the concentration change of O₃ mainly connects with the summation of HADV, VDIF and CHEM processes. However, compared with the Base experiment, although the emissions were cut down in Exp3, the concentration of O₃ was higher. It is important to note that the results were disproportionate in the decreasing CHEM and increasing HADV and VDIF.

5.3.3. Total contribution of each process

In order to quantify the impacts of the COVID-19 lockdown on the chemical and physical processes, analysis was conducted to examine the total contribution of each process for PM_2.5 and O₃ during Feb. 08, 2020 to Apr. 08, 2020, including concentration changes and their proportion as summarized in Table 6 TOT means the total concentration contribution of all the processes. The proportion of each process were calculated as:

$$S_{j,t}=\sum _{t=1}^t{C}_{j,t}$$ (10)

$$P_{j,t}=\frac{{\text{S}}_{j,t}}{\sum _{j=1}^n\left|S_{j,t}\right|}\begin{array}{ll}\hfill & \hfill \end{array}\left\{\begin{array}{c}n=9\left({\text{PM}}_{2.5}\right)\\ n=7\left(O_3\right)\end{array}\right\}$$ (11)

where $C_{j,t}$means the changes due to process j in hour t, and $S_{j,t}$means the sum of concentration changes of process j in hour t from t=1 to t, while$P_{j,t}$ means the proportion of process j in time t.

Table 6.

Total contribution of each process in the average over 10 stations in Wuhan from Feb. 08, 2020 to Apr. 08, 2020 (a. PM_2.5, b. O₃). (The symbol + means the tendency that increase the concentration, - means the tendency that decrease the concentration.)

S	TOT	ZADV	HADV	HDIF	VDIF	EMIS	DDEP	CLDS	CHEM	AERO
P
a.PM2.5(ug/m3)
Base	123.5	−5511	−19132	−36	−16860	37341	−1171	−21	71	5442
		−6%	−22%	0%	−20%	44%	−1%	0%	0%	6%
Exp3	50.5	−2619	−3927	−14	−7176	15501	−629	−450	26	−661
		−8%	−13%	0%	−23%	50%	−2%	−1%	0%	−2%
Exp3-Base	−74.0	2891	15204	22	9684	−21840	542	−429	−45	−6103
		5%	27%	0%	17%	−38%	1%	−1%	0%	−11%
Exp3/Base		48%	21%	39%	43%	42%	54%	2173%	36%	−12%

S	TOT	CHEM	DDEP	CLDS	ZADV	HADV	HDIF	VDIF
P
b.O3(ug/m3)
Base	5.8	−32282	−3573	−380	2259	9547	33	24402
		−45%	−5%	−1%	3%	13%	0%	34%
Exp3	21.4	−11922	−5318	−80	928	5476	16	10922
		−34%	−15%	0%	3%	16%	0%	32%
Exp3-Base	15.5	20360	−1745	301	−1332	−4072	−17	−13481
		49%	−4%	1%	−3%	−10%	0%	−33%
Exp3/Base		37%	149%	21%	41%	57%	48%	45%

The total contribution of each process for PM_2.5 in Base, Exp3, as well as the differences between Exp3 and Base (Exp3-Base), and the decline ratio from Base to Exp3(Exp3/Base) were shown in Table 6(a). The summation of all the processes, demonstrating the differences of concentration between the first hour and last hour in the simulation, was 124 μg/m³ in Base, 50 μg/m³ in Exp3. The difference was 32.5 μg/m³ in observation, with Exp3 much closer to the observation comparatively to Base, so Exp3 can better reflect the authentic situation for PM_2.5.

The proportion of each process in the differences between Exp3 and Base (Exp3-Base) shows us that EMIS and AERO dominate the decreasing changes, attributed to the COVID-19 lockdown, and VDIF, HADV dominate the increasing changes. The data indicated significant differences in each process as follows; EMIS is a dominant physical increasing process and accounts for approximately 44% of the contribution for all the processes in Base and 50% in Exp3. The changes attributed to the COVID-19 lockdown were about 42%. AERO increased pollution (6%) in the Base experiment, but on the contrary reduced the pollution (−2%) in the Exp3. The changes attributed to the COVID-19 lockdown were about −12% which can be explained through analysis of the condensation processes that changed aerosol species in different emission levels (Zhang, 2017). VDIF is a dominant physical decreasing process and accounts for approximately 20% of the contribution of all the processes in Base and 23% in Exp3, with the changes attributed to the COVID-19 lockdown being about 43%. HADV also contributed significantly and accounts for approximately 22% of the contribution of all the processes, including 13% in Exp3, with the changes attributed to the COVID-19 lockdown being about 21%.

In our simulation, we conclude that the PM_2.5 reduced during the COVID-19 lockdown period in Wuhan because of the contributions of EMIS and AERO decreased more than HADV and VDIF. The contribution of EMIS decreased to 42%, which always increased the concentration. And AERO had an opposite trend from increasing concentration to decreasing concentration and changed to −12%. HADV and VDIF decreased to 21% and 43% respectfully, which always decreased the concentration.

The total contribution of each process for O₃ in Base, Exp3, the differences between Exp3 and Base (Exp3-Base), as well as the decline ratio from Base to Exp3(Exp3/Base) were shown in Table 6(b). The summation of all the processes, demonstrated by the differences of concentration between the first hour and last hour in the simulation was 5.8 μg/m³ in Base, 21.4 μg/m³ in Exp3. The difference was 19.2 μg/m³ in observation, so Exp3 better reflects the real situation for O₃.

The proportion of each process in the differences between Exp3 and Base (Exp3-Base) shows that VDIF and HADV dominate the decreasing changes, while CHEM dominates the increasing changes. The data indicated significant differences in each process as follows; VDIF is a dominant physical increasing process and accounts for approximately 34% of all the processes in Base and 32% in Exp3, as the changes attributed to the COVID-19 lockdown is about 45%. HADV also contributed significantly and accounts for approximately 13% of all the processes in Base and 16% in Exp3, with the changes attributed to the COVID-19 lockdown about 57%. CHEM dominated the decreasing of O₃ and accounted for approximately 45% of all the processes in Base and 34% in Exp3, with the changes attributed to the COVID-19 lockdown being about 37%.

In our simulation, the reason why O₃ increased during the COVID-19 lockdown period in Wuhan was that the contribution of CHEM decreased more than HADV and VDIF. The contribution of CHEM decreased to 37%, which always decreased the concentration. And the contribution of HADV and VDIF decreased to 57% and 45%, which always increased the concentration.

The two dominated physical processes HADV and VDIF in the CMAQ model were highly connected with the meteorological parameters. The HADV and VDIF were solved in the science algorithms of CMAQ model (Byun et al., 1999) and shown as:

$$\frac{\partial \left(\sqrt{\hat{\gamma }}{\bar{\phi }}_i\right)}{\partial t}=-\nabla •\left(\sqrt{\hat{\gamma }}{\bar{\phi }}_i{\widehat{\stackrel{ˆ}{V}}}_{\xi }\right)\left(\text{HADV}\right)$$ (12)

$$\frac{\partial \left({\bar{q}}_i\right)}{\partial t}=-\frac{\partial \left({\hat{F}}_{q_i}^3\right)}{\partial \xi }+\left(\frac{Q_{{\phi }_i}}{\bar{\rho }\left(T\right)}\right)-{\hat{F}}_{q_i}^3\frac{\partial \left[\ln \left(\sqrt{\hat{\gamma }}\bar{\rho }\left(T\right)\right)\right]}{\partial \xi }\left(\text{VDIF}\right)$$ (13)

where $\hat{\gamma }$is the Jacobian of coordinate transformation, ${\phi }_i$is the trace species concentration in density units, t is the time step, ${\hat{V}}_{\xi }$is horizontal wind components in the coordinates $\xi$, $\xi$ is the terrain-influenced vertical coordinate, whose value is increasing monotonically with height, $q_i=\frac{\phi }{\rho }$is the species mass mixing ratio, ρ(T) is air density connected with temperature T, ${\hat{F}}_{q_i}^3$represents frictional forcing terms, $Q_{{\phi }_i}$is the source or sink term.

In formula 12, the HADV is mainly connected with the horizontal wind. As shown in Fig. 6, the average wind direction during the COVID-19 lockdown in Wuhan is North-East wind, and in Fig. 6(a) and (b), the concentration of PM_2.5 in the North-East of Wuhan was much lower than the main urban area. The favorable wind direction led to lower concentration in PM_2.5. In Fig. 6(g) and (h), it is different for O₃ that the concentration was always lower in the main urban area. Therefore, instead of decreasing, HADV increased the concentration of O₃ in the main urban area.

In formula 13, the formula used the eddy diffusion concept (K-theory) (Sutton, 1932), and as the theory proved, the species mass mixing ratio is highly connected with temperature (ρ(T)) and the trace species concentration in different vertical coordinate (${\phi }_i$). The simulation of PM_2.5 and O₃ had the same temperature gradient but different concentration gradient. The recent observation results in China has proved that the concentration of PM_2.5 generally decreased with height and the concentration of O₃ increased with height (Li et al., 2020). Therefore, the different tendency of VDIF between PM_2.5 and O₃ in our study can be explained by that VDIF decreased the surface concentration of PM_2.5 as a flux from the surface to upper layers and increased the surface concentration of O₃ as the flux from upper layers to the surface, in consistent with the study of Li et al. (2016) and Hogrefe et al. (2018).

In order to investigate why the AERO of PM_2.5 had an opposite trend in Base and Exp3 and why the HADV of O3 was always positive in the average of ten stations in Wuhan, the horizontal distribution of the average rates of AERO of PM_2.5 from Feb. 08, 2020 to Apr. 08, 2020 in Base, Exp3 and their difference was shown in Fig. 8 . In Fig. 8, the AERO of PM_2.5 had positive impact on the main urban area and negative impact on the other areas in Wuhan. The COVID-19 lockdown slowed down the positive rates of AERO but speed up the negative rates. Fig. S3 shows the rates of divided aerosol processes. The results in Fig. S3 revealed that the changes in aerosol species due to condensation was the main cause. The low emissions in the north-east of the main urban area in Wuhan became lower due to the COVID-19 lockdown. The AERO of PM_2.5 around the stations in the North-East of the urban area increased the concentration of PM_2.5 in Base but changed to decrease the concentration in Exp3. Then led to the opposite trend of the AERO of PM_2.5 in Base and Exp3.

Fig. 8.

The horizontal distribution of the average rates of AERO of PM_2.5 from Feb. 08, 2020 to Apr. 08, 2020, the black points means the location of 10 air quality stations. (a. Base, b. Exp3, c. differences as Exp3 minus Base).

5.3.4. Performances of each process while the pollution increased or decreased

In order to determine the impact of the COVID-19 lockdown regarding the pollution metrics during periods of increasing and decreasing pollution rates, analysis was conducted to examine the average contribution of each process and whether the concentration of pollutants increased or decreased from the previous hour (increasing and decreasing stages) as summarized in Table 7 . The positive value of increasing and decreasing stages means the concentration increased in the average of the whole simulation. The negative value of increasing and decreasing stages means the concentration decreased in the average of the whole simulation. The ratio means the concentration changed while the pollution increased compared with the concentration changed while the pollution decreased. The ratio reflects the difference between the periods of increasing and decreasing pollution in Base and Exp3, providing a direct target to find out the influences of COVID-19 lockdown to air pollution in Wuhan. As for PM_2.5, the ratio of EMIS was 1.07 in Base and 1.08 in Exp3. The results proved that the EMIS plays a similar role in increasing and decreasing stages. It decreased in almost equal proportion with the increasing and decreasing stages during the COVID-19 lockdown. The ratio of HADV was 1.01 in Base and 2.63 in Exp3. The results show that HADV played a similar role in the increasing and decreasing stages in Base but it contributed more in increasing stages than decreasing stages in Exp3.

Table 7.

Average contribution of each process while the pollution increased or decreased (a.PM_2.5, b.O₃). (The symbol + means the tendency that increase the concentration, - means the tendency that decrease the concentration.)

a.PM2.5(ug/m3)		Increasing stages	Decreasing stages	Ratio (Increasing stages/Decreasing stages)
Base	Hours	783h	657h	1.19
	ZADV	−4.47	−3.06	1.46
	HADV	−13.36	−13.20	1.01
	HDIF	−0.05	+0.00	0
	VDIF	−5.46	−19.15	0.29
	EMIS	+26.69	+25.03	1.07
	DDEP	−0.68	−0.97	0.70
	CLDS	+0.28	−0.36	−0.78
	CHEM	+0.04	+0.06	0.67
	AERO	+5.44	+1.80	3.02
	Total	+8.42	−9.85	−0.85
Exp3	Hours	812h	628h	1.29
	ZADV	−2.10	−1.46	1.44
	HADV	−3.74	−1.42	2.63
	HDIF	−0.02	+0.00	0
	VDIF	−2.11	−8.71	0.24
	EMIS	+11.13	+10.29	1.08
	DDEP	−0.34	−0.57	0.60
	CLDS	−0.13	−0.55	0.24
	CHEM	+0.01	+0.02	0.50
	AERO	+1.34	−2.78	−0.48
	Total	+4.06	−5.16	−0.79

b.O3(ug/m3)		Increasing stages	Decreasing stages	Ratio (Increasing stages/Decreasing stages)
Base	Hours	604h	836h	0.72
	CHEM	−17.87	−25.70	0.70
	DDEP	−3.14	−2.00	1.57
	CLDS	−0.33	−0.21	1.57
	ZADV	+1.40	+1.69	0.83
	HADV	+5.56	+7.40	0.75
	HDIF	+0.00	+0.04	0.00
	VDIF	+22.76	+12.75	1.79
	Total	+8.36	−6.04	−1.38
Exp3	Hours	616h	824h	0.75
	CHEM	−3.36	−11.96	0.28
	DDEP	−4.14	−3.36	1.23
	CLDS	−0.07	−0.04	1.75
	ZADV	+0.55	+0.71	0.77
	HADV	+3.48	+4.05	0.86
	HDIF	+0.00	+0.02	0.00
	VDIF	+10.56	+5.36	1.97
	Total	+7.02	−5.22	−1.34

It is clear that the COVID-19 lockdown had more impact on the decreasing stages of HADV. The ratio of VDIF was 0.29 in Base and 0.24 in Exp3. The results show that VDIF contributed more in dissipating stages, and it decreased in almost equal proportion in increasing and decreasing stages during the COVID-19 lockdown. The ratio of AERO was 3.02 in Base and −0.48 in Exp3. The results also show that AERO contributed more in increasing the concentration during increasing stages in Base, then changed to contribute more in decreasing the concentration during decreasing stages in Exp3. It is also evident that the COVID-19 lockdown had a significant impact on AERO in decreasing stages. As for other findings of note, in terms of O₃, the ratio of HADV was 0.75 in Base and 0.86 in Exp3. The results show that HADV contributed more in decreasing stages. In addition, the COVID-19 lockdown had a little more impact on the decreasing stages of HADV. The ratio of VDIF was 1.79 in Base and 1.97 in Exp3. The results show that VDIF contributed more in increasing stages, and the COVID-19 lockdown had a little more impact on the decreasing stages of VDIF. The ratio was 0.70 in Base and 0.28 in Exp3. The results show that VDIF contributed more in decreasing stages, but the COVID-19 had a significant impact on the increasing stages of CHEM.

In conclusion, the COVID-19 lockdown had a greater impact on the decreasing stages of horizontal advection and aerosol processes of PM_2.5, and the increasing stages of chemical processes of O₃, which might be caused by the lack of NO₂ and NO and explained in section 4.2.5.

5.3.5. Daily change of O₃

In order to find the most affected period of the day during the COVID-19 lockdown, Fig. 9 was carried out and shows the daily change of O₃ in observation, Base, Exp3 and the differences of each process between Exp3 and Base from the average of Feb. 08, 2020–Apr. 08, 2020. Fig. 9 represents the average of the whole COVID-19 lockdown period. TOT means total contribution of all the processes, while OBS means the observations. There is no doubt that Exp3 had a better performance in simulating daily change of O₃. In our simulation, it is apparent that Base had a lower O₃ concentration at night compared to Exp3. As seen in Fig. 9, for most of the day, the total contribution of all processes was a bit lower except 4 p.m.–7 p.m. The significant total differences of all the processes during the time of 4 p.m.–7 p.m. between Base and Exp3 resulted in higher O₃ levels at night during the COVID-19 lockdown period in Wuhan. The difference between the maximum and minimum concentration of the day change from the average during Feb. 08, 2020–Apr. 08, 2020, which was 63.38 μg/m³ in Exp3 and 77.23 μg/m³ in Base. The results here were in alignment with the results in section 4.2.4. These results reached an agreement that the COVID-19 lockdown made the differences smaller between the high concentration and low concentration of O₃. The COVID-19 lockdown had a greater impact on the increasing stages of chemical processes in O₃ as an hourly average. Beyond that, the results in Fig. 9 shows that the biggest difference of chemical processes attributed to the COVID-19 lockdown was at about 4 p.m.–7 p.m. in the decreasing stages.

Fig. 9.

The daily change of O₃ in observation, Base and Exp3(line) and the differences of each process between Exp3 and Base(bar) in the average over 10 stations in Wuhan from the average of Feb. 08, 2020–Apr. 08, 2020.

For future consideration, what causes the significant differences between Base and Exp3? Fig. 9 shows that the declining rates of decreasing the concentration that are mainly due to chemical processes, were much higher than the declining rates of increasing the concentration by vertical diffusion and horizontal advection decreased. The differences in Base and Exp3 around 6 p.m. lead to the different rate of the consumption of O₃, and then made O₃ higher at night. Chemical processes at about 4 p.m.–7 p.m. were the main factors that led to the high O₃ pollution during the COVID-19 lockdown in Wuhan. As described in the previous study, the main chemical processes of O₃ is that O₃ produced directly by photolysis of NO₂ (R1), where the oxygen atom (O) rapidly recombines with molecular oxygen to produce ozone (O₃). Normally, this reaction is counterbalanced by the reaction of NO with ozone(R2):

$$N{O}_2+hv\to NO+O$$ (R1)

$$NO+O_3\to N{O}_2$$ (R2)

There is always net removal of ozone at nighttime. Surface O₃ is normally low when NO emissions are high. The significant daytime removal of ozone via reaction (R2) occurs in the vicinity of large NO emission sources (Kleinman et al., 2000; Lin et al., 1998). In our situation, NO_x had decreased in the lockdown, leading to lower rates of ozone titration, which led to higher O₃ at night, leading to a higher daily concentration of O₃. One could easily hypothesize then, that the causes of the high O₃ conditions during the lockdown might be that the combination of the NO_x being reduced as a result of the COVID-19 lockdown and the reduction of fresh NO emissions alleviates ozone titration. As shown in Fig. 9, the TOT_Base (yellow dotted line) and TOT_Exp3 (green dotted line) was similar and the TOT_Exp3-Base was close to zero from 8 p.m. to 3 p.m. The increase of CHEM and decrease of VDIF, HADV, ZADV was balanced throughout the day by comparing the Exp3 with Base. The decreasing NO titration effect during the lockdown period made significant differences of all the processes during the time of 4 p.m.–7 p.m. between Base and Exp3, which then directly caused higher O₃ levels at night during the COVID-19 lockdown period in Wuhan. When the rates of CHEM during the time of 4 p.m.–7 p.m. slowed down in Exp3, the concentration of O₃ increased rapidly at night, leading to a high concentration during the day. For the decreasing of O₃ concentration, a good solution might be to enhance the ozone titration at about 4 p.m.–7 p.m. of the day, speeding up the consumption of O₃ by chemical processes.

6. Discussion and conclusion

In order to study the variation of air pollutant concentration and its formation mechanism through chemical reactions and physical processes in the atmosphere during the COVID-19 lockdown, in this paper the WRF-CMAQ model and Process Analysis module were used to simulate pollutant concentration and the contribution of processes. The air pollutant ratio changes during the COVID-19 lockdown period in Wuhan were studied too. The following are the main conclusions.

First, the observation results show that PM_2.5 reduced 31.7% and O₃ increased 43.9% from Feb. 8, 2020 to Apr. 8, 2020 compared to the same time period in 2019. The observation results from this study were in alignment with other scientist's studies (IQAir, 2020; Le et al., 2020; Ministry of Ecology and Environment of the People's Republic of China, 2020), adding validity to this data.

Secondly, there were five simulation experiments carried out. In order to properly calibrate the initial emission source of the model, the Eva experiment was used to assess the performance of the CMAQ simulation. The Base experiment was designed as a baseline without considering the COVID-19 lockdown in Wuhan. The other three experiments which considered the COVID-19 lockdown in Wuhan show that Exp3 had the best performances in simulating PM_2.5 and O₃, by a decrease of 20% in industry and 10% in agriculture and transportation. Therefore, Exp3 might reflect the most possible reduction of emissions caused by the COVID-19 lockdown in Wuhan.

Thirdly, analysis was conducted to examine the impacts of the COVID-19 lockdown on the chemical and physical processes as summarized in Table 6. While comparing the results of process analysis of PM_2.5 and O₃ between Exp3 and Base, it was found that (1) the reduction of PM_2.5 was mainly due to the reduction of emissions, which dropped to 42% in Exp3 compared with Base. The concentration contribution of the aerosol process in Exp3 changed to −12%, meaning that the aerosol process had an opposite tendency in changing PM_2.5 concentration in Exp3 compared with Base. In addition, the rates of concentration contribution in Exp3 are reduced to 43% for vertical diffusion and 21% for horizontal advection. (2) The increase of O₃ was mainly due to the weakening of the chemical process (weaken to 37% compared with the base), which was unfavorable to consumption, although the rates of diffusion and advection increasing O₃ concentration were reduced to 45% and 57% respectively.

As for the horizontal advection, it is mainly connected with the horizontal wind. The average wind direction during the COVID-19 lockdown in Wuhan is North-East wind, the concentration of PM_2.5 in the North-East of Wuhan was much lower than the main urban area. The favorable wind direction led to lower concentration in PM_2.5. It is different for O₃ that the concentration was always lower in the main urban area. Therefore, instead of decreasing, the horizontal advection increased the concentration of O₃ in the main urban area. The North-East wind was in favor of the decreasing of PM_2.5.The higher O₃ concentration in the North-East of the main urban area contributed to the increasing of O₃ with unfavorable wind direction. As for the vertical diffusion, the eddy diffusion concept (K-theory) (Sutton, 1932) has proved that the species mass mixing ratio is highly connected with temperature and the trace species concentration in different vertical coordinate. The simulation of PM_2.5 and O₃ had the same temperature gradient but different concentration gradient. The recent observation results in China has proved that the concentration of PM_2.5 generally decreased with height and the concentration of O₃ increased with height (Li et al., 2020). Therefore, the different tendency of VDIF of PM_2.5 and O₃ in our study can be explained by that VDIF decreased the concentration of PM_2.5 as a flux from the surface to upper layers and increased the concentration of O₃ as the flux from upper layers to the surface, in consistent with the study of Li et al. (2016) and Hogrefe et al. (2018). As for the aerosol process of PM_2.5, it had an opposite tendency during the COVID-19 lockdown. The aerosol process had positive impact on the main urban area and negative impact on the other areas in Wuhan. The COVID-19 lockdown slowed down the positive rates of aerosol processes but speed up the negative rates, changed the rates from increasing the concentration of PM_2.5 to decrease in the North-East of the urban area. The results also revealed that the changes in aerosol species due to condensation was the main cause.

The results of this study quantified the impacts of the COVID-19 lockdown to air pollution, and offered an answer in what and how an unprecedented emission mitigation measure can be done to prevent air pollution. The results show that although an unprecedented emission mitigation measure was carried out, the concentration of PM2.5 decreased significantly, the concentration of O₃ increased on the contrary. This result pointed out that to prevent O₃ and PM_2.5 pollution at the same time is more difficult than existing proposed measures (Wang et al., 2009; Li et al., 2011) in the field, that primarily were emission reduction campaigns including closing factories, industrial plants, construction sites, gas stations and keeping vehicles off of the road.

Fourth, in order to find out the differences of the COVID-19 lockdown impacts between the pollution that increased and decreased, analysis was conducted to examine the performances of processes changed while pollution increased or decreased due to the COVID-19 lockdown in Wuhan as summarized in Table 7. The concentration contribution ratios were used here. Through dividing the average process rates of increasing periods by decreasing periods in Base and Exp3, the ratios reflect the difference between the periods of increasing and decreasing pollution in Base and Exp3, providing a directly target to find the influences of COVID-19 lockdown to air pollution in Wuhan. (1) As for PM_2.5, the concentration contribution ratios of horizontal advection were 2.63 in Exp3 and 1.01 in Base while the pollution increase compared to decrease. The ratio shows that horizontal advection demonstrated more differences between the pollution which increased and decreased in Exp3. (2) The ratio of aerosol processes was −0.48 in Exp3 and 3.02 in Base, indicating that aerosol processes were more likely to decrease the concentration while the pollution decreased in Exp3, rather than increasing the concentration. (3) As for O_3, the ratio of chemical processes was 0.28 in Exp3 and 0.75 in Base, showing that chemical processes demonstrated more differences between the pollution which increased or decreased in Exp3.

The COVID-19 lockdown had a greater impact on the decreasing stages of horizontal advection and aerosol processes in PM_2.5, and on the increasing stages of chemical processes in O₃. The results can help better understand the differences of pollution increasing stages and pollution decreasing stages when an emission mitigation measure is carried out. This can further help the government to consider how to prevent O₃ and PM_2.5 pollution at the same time. According to the results of this paper, focusing on the decreasing stages of PM_2.5 and increasing stages of O₃ might be more effective when air pollution needs to be avoided.

Finally, in order to find the most affected period of the day by the COVID-19 lockdown, Fig. 9 was carried out and found the significant differences at about 4 p.m.–7 p.m. between Base and Exp3, which might be caused by the O₃ being higher at night during the COVID-19 lockdown period in Wuhan. The causes of the significant differences might be that the restriction of traffic leads to the reduction of NO concentration, which weakens the reaction between O₃ and NO, NO₂ an NO, stimulating a reduction of fresh NO emissions and alleviating ozone titration.

This study further pointed out the exact time period of the day that was most affected in addition to its similar results to previous studies (Huang et al., 2021; Le et al., 2020) that the reduction of fresh NO emissions alleviates ozone titration leading to the higher O₃ during the COVID-19 lockdown. Beyond that this result further confirms the previous conclusions through quantitative analysis and provides a new possible way that enhances ozone titration at about 4 p.m.–7 p.m. of the day to avoid O₃ pollution.

CRediT authorship contribution statement

Congwu Huang: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Visualization, Writing – original draft. Tijian Wang: Supervision, Funding acquisition, Project administration. Tao Niu: Conceptualization, Resources, Writing – review & editing, Supervision, Funding acquisition, Project administration. Mengmeng Li: Supervision. Hongli Liu: Data curation. Chaoqun Ma: Data curation.

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.

Abbreviations

COVID-19

2019 novel coronavirus

CMAQ

The Community Multiscale Air Quality modeling system

PM_2.5

fine particulate matter

PM₁₀

inhalable particles

PROCAN

Process Analysis Preprocessor

HADV

horizontal advection

ZADV

vertical advection

HDIF

horizontal diffusion

VDIF

vertical diffusion

DDEP

dry deposition of species

CLDS

change due to cloud processes; includes aqueous reaction and removal by clouds and rain

AERO

change due to aerosol processes

CHEM

net sum of all chemical processes for species over output step

EMIS

emissions contribution to concentration

COR

correlation coefficient

RMSE

root mean squared error

NMB

normalized mean bias

NME

normalized mean error

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Fig S1.

Fig S2.

Fig S3.