Unexpected Impact of COVID-19 Lockdown on the Air Quality in the Metro Atlanta, USA Using Ground-based and Satellite Observations

We studied the impact of COVID-19 (coronavirus disease 2019) lockdown on the air quality over the Atlanta area using satellite and ground-based observations, meteorological reanalysis data and traffic information. Unlike other cities, we found the air quality has improved slightly over the Atlanta area during the 2020 COVID-19 lockdown period (March 14–April 30, 2020), compared to the analogous period of 2019 (March 14–April 30, 2019). Ground NO 2 concentrations have decreased slightly 10.8% and 8.2% over the near-road (NR) and urban ambient (UA) stations, respectively. Tropospheric NO 2 columns have reduced 13%–49% over the Atlanta area from space-borne observations of TROPOspheric Monitoring Instrument (TROPOMI). Ground ozone and PM 2.5 have decreased 15.7% and ~ 5%, respectively. This slight air quality improvement is primarily caused by the reduced human activities, as COVID-19 lockdowns have reduced ~ 50% human activities, measured by traffic volume. Higher wind speed and precipitations also make the meteorological conditions favorable to this slight air quality improvement. We have not found a significant improvement in air quality over Atlanta amid the lockdown when human activities have reduced ~ 50%. Further studies are needed to understand the impacts of reduced human activities on atmospheric chemistry. We also found TROPOMI and ground measurements have disagreements on NO 2 reductions, as collocated TROPOMI observations revealed ~ 23% and ~ 21% reductions of tropospheric NO 2 columns over NR and UA stations, respectively. Several factors may explain this disagreement: First, tropospheric NO 2 columns and ground NO 2 concentrations are not necessarily the same, although they are highly correlated in the afternoon; Second, meteorological conditions may have different impacts on TROPMI and ground measurements. Third, TROPOMI may underestimate tropospheric NO 2 due to uncertainties from air mass factors. Fourth, the uncertainties of chemiluminescence NO 2 measurements used by ground stations. Consequently, studies using space-borne tropospheric NO 2 column and ground NO 2 measurements should take these factors into account.


INTRODUCTION
As the outbreak of COVID-19 (coronavirus disease 2019) and its pandemic spread worldwide, many countries have implemented measures to slow down and prevent the spread of COVID-19.
These COVID-19 lockdowns have caused significant reductions in human activities and traffic worldwide. As human activities have reduced, air quality has improved significantly worldwide (Bauwens et al., 2020;Goldberg et al., 2020;Huang and Sun, 2020;Jin et al., 2020;Koo et al., 2020;Liu et al., 2020;Morales-Solís et al., 2021;Naeger and Murphy, 2020;Oo et al., 2021;Represa et al., 2021). However, some studies found air quality has not improved as we expected or even worsened in some cities due to the complicated atmospheric chemistry, meteorological conditions, and emissions, even anthropogenic emissions from human activities and traffic have reduced significantly (Bekbulat et al., 2021;Chen et al., 2020;Le et al., 2020;Schiermeier, 2020;Shi and Brasseur, 2020).
The variation of ground air pollutants is controlled by complicated and non-linear physical and chemical processes (Kroll et al., 2020;Seinfeld and Pandis, 2016). Nitrogen oxides (NOx = NO + NO2) emitted into the atmosphere from anthropogenic sources are a major driver of ambient NO2 concentrations over populous areas in the world (Lamsal et al., 2013(Lamsal et al., , 2008. It is not surprising that ambient NO2 has reduced significantly in many areas due to the reduced human activities by COVID-19 lockdowns (Huang and Sun, 2020;Liu et al., 2020;Naeger and Murphy, 2020). The chemistry of ozone and aerosols (e.g., Particulate Matter 2.5 (PM2.5)) are more complicated than NO2. The production of ozone depends on volatile organic compounds (VOCs) and NOx with the presence of sunlight. When VOCs are high and NOx is low, it is the "NOx" controlled situation. In this situation, more NOx means more ozone. When VOCs are low and NOx is high, it is the "VOCs" limited situation. More NOx does not produce more ozone or even decrease ozone (Kroll et al., 2020). The chemistry of PM2.5 is even more complicated. PM2.5 can be emitted directly from combustion and produced through gas-phase chemical reactions to form products of low enough volatility to condense into the particle phase. In addition, secondary organic aerosols (SOAs) makes it even more complicated (Kroll et al., 2020;Seinfeld and Pandis, 2016). As a result, ozone and PM2.5 have had more diverse responses during the COVID-19 lockdowns (Bekbulat et al., 2021;Chen et al., 2020). Many studies have shown ozone increased, whereas PM2.5 decreased slightly, as the increase of government stringency (Gkatzelis et al., 2021). However, cities with slightly better or even worse air quality during COVID-19 lockdowns may need further analysis due to their unexpected responses to reduced human activities. COVID-19 lockdowns provide us a special scenario to understand the mechanisms that controlling air quality under reduced human activities (Kroll et al., 2020). In addition, analysis on meteorology is essential to interpretate the relationship between COVID-19 lockdowns and air quality (Goldberg et al., 2020). Unfortunately, many previous studies had not taken metrological conditions into account (Gkatzelis et al., 2021).
Atlanta is the largest and most populous metropolitan area in the southeastern United States. People's activities heavily depend on private vehicles, as traffic information is a good indicator of human activities. Geographically, Atlanta is an isolated non-attainment area and its major source of air pollutants are from local emissions (Diem, 2009). As one of the most diverse major cities in the U.S., Atlanta has been suffering air quality disparities and environmental justice for a long time (Park and Kwan, 2020;Servadio et al., 2019). These environmental justice issues have worsened worldwide during the COVID-19 pandemic (Kerr et al., 2021). Consequently, it is essential to analyze the impact of COVID-19 lockdown on the air quality in the Atlanta area for the local air quality management and environmental justice. Unfortunately, no previous studies have done such analysis in the Atlanta area.
This study aims to assess the impact of COVID-19 lockdown on the air quality in the Atlanta metropolitan using measurements of NO2, ozone and PM2.5 from multiple ground stations and spacebased tropospheric NO2 observations in conjunction with meteorological and traffic information.

Study Areas and COVID-19 Control Measures in Atlanta
The Atlanta metropolitan area in the state of Georgia is one of the major metropolitans in the United States that experience high concentrations of air pollution due to a range of factors, including urbanization and anthropogenic emissions (Diem, 2009;Henneman et al., 2015Henneman et al., , 2017. Most population of Atlanta concentrates in the area inside of Interstate Loop Highway (I-285) including the City of Atlanta, other cities, and unincorporated areas. In this study, we focus on of the total interference is variable, and depends not only on the relative fraction of actual NO2 to total reactive nitrogen compounds (NOy), but also on the characteristics of individual monitors (Lamsal et al., 2015). Consequently, ground NO2 measurements may overestimate NO2 concentrations, with minimum biases in the morning and maximum biases in the early afternoon, close to the overpass time of TROPOMI (Dunlea et al., 2007;Lamsal et al., 2015).
Hourly PM2.5 and ozone measurements are obtained from South DeKalb and United Avenue stations, both urban UA stations. The hourly data at South DeKalb station are from a Teledyne T640 PM Mass Monitor which an optical aerosol spectrometer for measurement with 0.1-10000 µg m -3 measurement range and 0.5 µg m -3 (1-h average) (Georgia Department of Natural Resources, 2019). The United Avenue station uses a tapered element oscillating microbalance (TEOM) which collects particles on a filter with a measurement range of 0-5 g m -3 , and 1.5 µg m -3 precision (1hour average) (Georgia Department of Natural Resources, 2020). The hourly ozone measurements of both stations are obtained from Model 49i Ozone Analyzer manufactured by Thermo Fisher that is based on continuous ultraviolet photometric method. The measurement range is from 0.05 ppb to 200 ppm with 1.0 ppb precision (Georgia Department of Natural Resources, 2020).

Meteorological Data
We use meteorological data from MERRA-2 (Modern-Era Retrospective analysis for Research and Applications, Version 2) (Gelaro et al., 2017). These meteorological data include 2-meter (2 m) wind speed, 2 m wind direction, 2 m temperature and 2 m relative humidity (RH), 850 hPa and 500 hPa wind speed and wind direction, planetary boundary layer (PBL) top pressure and cloud cover. MERRA-2 is an atmospheric reanalysis model produced by NASA Global Modeling and Assimilation Office (GMAO) with a spatial resolution of 0.5° × 0.625°. Due to the spatial resolution of our meteorological data, microscale and fine mesoscale features will not be accounted for, but these effects should be minor in this study. In addition, we obtain precipitation measurements from the weather station located in Atlanta Hartsfield Jackson International Airport as shown in Fig. 1.

TROPOMI NO2 Observations
TROPOMI is the single payload aboard the Sentinel 5 Precursor (S5P) satellite that has a sunsynchronous orbit with a local overpass time of approximately 13:30 with a near-daily global coverage since April 2018 (Veefkind et al., 2012). The TROPOMI NO2 retrieval algorithm is developed by the Royal Netherlands Meteorological Institute and based on the NO2 DOMINO algorithm with significant improvements including the improved retrieval of slant column density and spectral fitting (Boersma et al., 2018;Lorente et al., 2017;Van Geffen et al., 2015). TROPOMI retrieves tropospheric NO2 with a pixel size of 7 km × 3.5 km at nadir, and the resolution has been improved to 5.5 km × 3.5 km with a change in the S5P operation scenario since August 6, 2019 (orbit 9388) (Eskes and Eichmann, 2019). We only use TROPOMI offline observations with cloud coverage less than 0.3, and quality assurance greater than 0.75 (Eskes and Eichmann, 2019).

Traffic Data
We use traffic count data as the indicator of human activities in this study, as private vehicles are the major commuting tool in the Atlanta area (Atlanta Regional Commision, 2020). The traffic count data are collected from the Georgia Department of Transportation (GDOT) using the (GDOT)'s Traffic Analysis and Data Application (TADA) (https://gdottrafficdata.drakewell.com/pu blicmultinodemap.asp). GDOT collects traffic information through Continuous Count Stations (CCSs) installed on all types of public roads in Georgia. Each CCS is designed to provide volume counts continuously (24 hours per day, 7 days per week, 365 days per year) except during maintenance periods. We select three CCSs on major highways in the downtown and midtown areas of Atlanta to capture the traffic variation during the COVID-19 lockdown as shown in Fig. 1.

Conversion of TROPOMI Level 2 Data to Level 3
We produce level 3 data in the 2020 COVID-19 lockdown period (March 14-April 30, 2020) and its analogous period of 2019 (March 14-April 30, 2019) by averaging level 2 data to a common, regular 0.01° × 0.01° grid and using a physical oversampling method to the regridded level 3 tropospheric NO2 data. The physical oversampling method by Sun et al. (2018) temporally and spatially averages the level 2 satellite observations to level 3 grids. Two-dimensional super Gaussian functions are used to represent the spatial response functions of satellite sensors. In comparison with conventional approaches that only consider the pixel corners, this method gives considerable advantages of visualizing the distribution and local gradients of trace gases and for scenarios with short temporal windows (Huang and Sun, 2020;Sun et al., 2018).
The wind speed enhancement and wind direction change in 2020 may have strong and different impacts on the air pollution at ground level. 2-m wind direction has changed from 233° in 2019 to 273° in 2020, while the wind speed is similar. The dominant west wind in 2020 may transport air pollutants from downtown and midtown areas to the west of Atlanta, where all ground stations are, to increase ground stations' measurements. The impact of wind direction change may have a larger impact on the NR station. The dominant southwest wind at 2 m height in 2019 blows air pollutants away from the NR-285 station located in the south of I-285, while the dominant west wind in 2020 increases the concentrations at NR-285 station (Richmond-Bryant et al., 2017;Venkatram et al., 2013). As a result, the wind direction change may reduce the NO2 difference between 2020 and 2019 measured by both ground stations.
The higher wind speed at 850 hPa in 2020 may improve air quality through rapid transport and dispersion of air pollutants from local emission sources, even 2-m wind speed are similar in 2019 and 2020 (Goldberg et al., 2020;Naeger and Murphy, 2020). The larger enhancement of wind speed at 500 hPa in 2020 may cause more reductions in tropospheric NO2 columns observed by TROPOMI than ground NO2 measurements.
Temperature in 2020 COVID-19 lockdown period is slightly higher than the analogous period of 2019, while relative humidity, atmospheric boundary layer heights and cloud cover are similar in 2019 and 2020. On the other hand, the precipitation in 2020 is 28% higher than 2019. The considerably larger precipitation in 2020 is expected to washout air pollutants to improve the air quality in 2020 (Goldberg et al., 2020). In sum, meteorological conditions have contributed to improved air quality in the Atlanta area in the 2020 COVID-19 lockdown period. The meteorological conditions have different impacts on concentrations of air pollutants measured by ground stations and satellites.

Traffic Volume
Traffic volume has reduced 48.7% in 2020 relative to 2019 in the downtown and midtown areas of Atlanta due to the COVID-19 lockdown. Fig. 2(a) shows the mean hourly traffic volume from February 1 to April 30 in both 2019 and 2020. We have applied a 7-day moving average to smooth out the weekend's effect on traffic volume. The traffic volume does not have a strong variation in 2019 and the traffic volume before the lockdown in 2020 is similar to 2019. The traffic volume has decreased dramatically as the declaration of public health emergency on March 14, 2020, and reached the minimum on April 7, 2020, approximately 60% less than in the same period of 2019.

NO2
Tropospheric NO2 columns from TROPOMI in 2020 COVID-19 lockdown and the analogous period in 2019 are shown in Figs. 3(a) and 3(b). Tropospheric NO2 columns have remained high (> 0.1 Dobson Unit (DU)) over the entire Atlanta area in 2019, with tropospheric NO2 hotspots along major highways, over the downtown and midtown of Atlanta (~0.15 DU), and Atlanta International Airport (> 0.15 DU) as shown in Fig. 3(a). These tropospheric NO2 hotspots in downtown and midtown areas are associated with human activities and ground traffic; the hotspot near Atlanta International Airport is associate with air traffic (Rissman et al., 2013).
During the COVID-19 lockdown, tropospheric NO2 columns have reduced 28.1% (0.03DU) over the entire Atlanta metropolitan area as shown in Fig. 3(c). However, some tropospheric NO2 hotspots with fewer intensities still exist in 2020 in junctions of major highways located in the northwest, northeast and southwest corners of I-285 likely due to the remaining traffic. Goldberg et al. (2020) has shown a 35.2% reduction of tropospheric NO2 columns over the Atlanta area in 2020 lockdown period, compared to the analogous period of 2019. Our results are comparable with this study.
Mean ground NO2 measurements in 2020 COVID-19 lockdown and the analogous period of 2019 are listed in Table 2. Ground NO2 measurements have also decreased in 2020 COVID-19 lockdown period. NR NO2 has decreased from 14.99 ppb (± 5.67 ppb) in 2019 to 13.37 ppb (± 4.42 ppb) in 2020 with 10.8% reduction. UA NO2 has decreased from 8.25 ppb (± 3.23 ppb) in 2019 to 7.57 ppb (± 3.02 ppb) in 2020 with 8.2% reduction. Fig. 4 plots the hourly variation of NO2 in 2020 COVID-19 lockdown and the analogous period of 2019. Although daily mean NO2 has decreased in 2020, hourly mean NO2 in 2020 are not consistently lower than 2019 as shown in Fig. 4(a). 2020 NR NO2 during 19:00 and 8:00 is up to ~25% lower than 2019 NR NO2 at the same hours, while 2020 NR NO2 during 11:00 and 18:00 is up to 35% higher than 2019. On the other hand, UA NO2 in 2020 is ~10% higher than 2019 during 22:00 to 1:00, while 2020 UA NO2 become up to 30% lower than 2019 from 3:00 to 10:00. UA NO2 from 11:00 to 20:00 in 2020 is slightly less than 2019.   and 2020 (marked by orange) measured by near-road (NR, marked by triangles) and background (UA, marked by rounds), respectively. Panel (b) plots ratios of hourly 2020 NO2 to 2019 NO2 measured by NR and UA ground stations marked by blue triangles and rounds, respectively. The red triangle and round represent ratios of 2020 to 2019 tropospheric NO2 columns measured by TROPOMI collocated with NR and UA ground stations, respectively.
Collocated TROPOMI observations have indicated more significant reductions of tropospheric NO2. These collocated tropospheric NO2 columns are mean values of tropospheric NO2 columns of 3 × 3 grids centered at each station. TROPOMI observations have shown tropospheric NO2 over NR and UA stations have decreased ~23% and 21% at the overpass time of TROPOMI (~13:30 local time), respectively. These reductions of tropospheric NO2 columns observed by TROPOMI are higher than the reductions observed by ground stations. There are several factors may cause this disagreement between TROPOMI and ground observations. First, the tropospheric NO2 columns measured by TROPOMI and NO2 measured by ground stations are not the same, although tropospheric NO2 columns and ground NO2 may have high correlations (Cersosimo et al., 2020;Demetillo et al., 2020). The reductions in traffic and human activities may cause different percentage reductions for ground NO2 and tropospheric NO2 columns. Second, meteorological conditions may have different impacts on TROPOMI and ground measurements as discussed in Section 4.1. Third, TROPOMI may underestimate tropospheric NO2 due to uncertainties from air mass factors (Griffin et al., 2019). Forth, the uncertainties of chemiluminescence NO2 measurement may also contribute to this disagreement as these uncertainties reach maximum in the early afternoon and minimum in the morning (Dunlea et al., 2007;Lamsal et al., 2015). Ground NO2 measurements in 4:00-8:00 show ~18-30% reductions of NO2 in both NR and UA stations as shown in Fig. 4(b). These NO2 reduction measured by ground stations agree well with TROPOMI observations. Consequently, studies using satellite and ground NO2 measurements should take these factors into account.
Ground ozone during 2020 COVID-19 lockdowns has reduced consistently relevant to the analogous period in 2019 as shown in Fig. 5. Ground ozone has decreased 15-20% during daytime, and 20-30% during nighttime. Ground ozone has indicated the smallest reductions as 5-10% during morning and evening transition time. The Atlanta area is a NOx-controlled scenario for ozone production (Jin et al., 2020;McDonald et al., 2018;Wolff et al., 2013). The combination of slightly increased Formaldehyde (HCHO) (Sun et al., 2021) and reduced NOx emissions have developed a more NOx-controlled scenario for ozone production in the Atlanta area. As a result, this significant ozone reduction in the Atlanta area is primarily caused by the reduced NOx emission of COVID-19 lockdowns and increased HCHO emission of high temperature.

PM2.5
Mean PM2.5 during 2020 COVID-19 lockdowns is 9.98 µg m -3 (±1.58 µg m -3 ), with 5% less than 10.44 µg m -3 (±1.78 µg m -3 ) in the analogous period in 2019 as listed in Table 2. Hourly mean PM2.5 concentrations during 2020 COVID-19 lockdown are slightly lower than the analogous period in 2019 at most time as shown in Fig. 6. PM2.5 in 2020 is up to ~17% lower than 2019 most time but up to ~8% higher during 13:00-16:00. The PM2.5 chemistry is more complicated than ozone.  Some PM2.5 is primarily emitted directly from vehicles, which may be associated with the reduced traffic volumes due to the COVID-19 lockdowns. On the other hand, PM2.5 is also produced through complex reactions of gases (Kroll et al., 2020). In sum, the reduced traffic may cause the reduction of PM2.5, but more studies are needed to understand the non-linear chemistry behind the reductions.

CONCLUSIONS
We have utilized satellite and ground measurements, and reanalysis data in conjunction with traffic information to understand how NO2, a major air pollutant, responses to the decreased human activities in the Atlanta metropolitan area during the 2020 COVID-19 lockdown (March 14-April 30, 2020). Human activities, measured by traffic volume, in 2020 COVID-19 lockdown period has decreased 48.7% relevance to the analogous period of 2019. Furthermore, the human activities in 2020 have reduced 40%-50% during peak hours (6:00-19:00), and 50%-63% during off-peak hours (20:00-5:00).
Air quality in the Atlanta area during COVID-19 lockdown period has improved slightly, although human activities in 2020 are only approximately half of 2019. Ground NO2 in 2020 has reduced 10.8% and 8.2% by NR and UA stations, respectively. The reductions of ground NO2 indicate strong temporal variation in NR and UA stations. Tropospheric NO2 columns are observed 28.1% reductions by TROPOMI over the Atlanta area in 2020 COVID-19 lockdown period, compared to the analogous period of 2019. TROPOMI observations have indicated that tropospheric NO2 columns have decreased ~23% and ~21% over NR and UA stations, respectively, while collocated ground NO2 has increased ~20% by the NR station and decreased ~10% by the UA station on the overpass time of TROPOMI. Some factors may cause the disagreement on NO2 changes between ground and satellite observations. First, tropospheric NO2 columns measured by TROPOMI and ground NO2 concentrations measured ground stations are not necessarily the same, although they are highly correlated in the afternoon (Cersosimo et al., 2020;Demetillo et al., 2020). The decreased ground NOx emission may cause different percentage changes in tropospheric NO2 columns and ground NO2 concentrations. Second, meteorological conditions may have different impacts on TROPMI and ground measurements. Third, TROPOMI may underestimate tropospheric NO2 due to uncertainties from air mass factors (Griffin et al., 2019). Forth, the uncertainties of chemiluminescence NO2 measurement may also contribute to this disagreement as these uncertainties reach maximum in the early afternoon and minimum in the morning (Dunlea et al., 2007;Lamsal et al., 2015). Consequently, studies using satellite and ground NO2 measurements should take these factors into account.
Ground ozone and PM2.5 both have decreased over the Atlanta area during the 2020 COVID-19 lockdowns. Ground ozone in 2020 has decreased 15.7% relevance to 2019. The Atlanta area is NOx-controlled for ozone photochemistry due to the local NOx emission control (Jin et al., 2020;McDonald et al., 2018;Wolff et al., 2013). In 2020, the slightly increased HCHO concentrations over the southeast U.S. due to the warmer temperature made the Atlanta area a more NOx-controlled scenario for ozone photochemistry.
PM2.5 in the 2020 COVID-19 period has decreased ~5% relevance to the analogous period of 2019. Hourly PM2.5 concentrations in 2020 are slightly lower than 2019 in most time, but up to ~8% higher during 13:00-16:00. The reduced traffic may cause the slight decrease of ground PM2.5, but further studies are needed to understand the non-linear chemistry of PM2.5.
Meteorological conditions have also contributed to the improvement of air quality over the Atlanta area in 2020 COVID-19 period. Higher wind speed and precipitations in 2020 are favorable for better air quality (Goldberg et al., 2020).
We conclude that air quality has improved slightly in the Atlanta area during 2020 COVID-19 lockdown (March 14-April 20, 2020) compared to the same period in 2019, although human activities has reduced ~50%. This slight improvement is primarily caused by reduced human activities induced by COVID-19 lockdowns. Meteorological conditions have also contributed to this improvement of air quality. Due to the complex chemistry, further studies are needed to quantitatively understand the impacts of reduced vehicle emissions on atmospheric chemistry as consequences of the COVID-19 pandemic.

ACKNOWLEDGEMENT
We acknowledge the TROPOMI science team for TROPOMI level 2 data, Georgia Department of Natural Resources for ground air quality data, and Georgia Department of Transportation for traffic data. R. Ponder acknowledges the support from the Center of Excellence for Minority Women in STEM's Department of Defense Research Scholars Program. We also thank DeAnna Oser of Georgia Department of Natural Resources for her valuable comments.

FUNDING
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.