key: cord-0767704-3n5f4nno
authors: Campbell, Patrick C.; Tong, Daniel; Tang, Youhua; Baker, Barry; Lee, Pius; Saylor, Rick; Stein, Ariel; Ma, Siqi; Lamsal, Lok; Qu, Zhen
title: Impacts of the COVID-19 economic slowdown on ozone pollution in the U.S.
date: 2021-11-01
journal: Atmos Environ (1994)
DOI: 10.1016/j.atmosenv.2021.118713
sha: 729f982b381dd202c7c44b5c804baa130632da73
doc_id: 767704
cord_uid: 3n5f4nno

In this work, we use observations and experimental emissions in a version of NOAA's National Air Quality Forecasting Capability to show that the COVID-19 economic slowdown led to disproportionate impacts on near-surface ozone concentrations across the contiguous U.S. (CONUS). The data-fusion methodology used here includes both U.S. EPA Air Quality System ground and the NASA Aura satellite Ozone Monitoring Instrument (OMI) NO(2) observations to infer the representative emissions changes due to the COVID-19 economic slowdown in the U.S. Results show that there were widespread decreases in anthropogenic (e.g., NO(x)) emissions in the U.S. during March–June 2020, which led to widespread decreases in ozone concentrations in the rural regions that are NO(x)-limited, but also some localized increases near urban centers that are VOC-limited. Later in June–September, there were smaller decreases, and potentially some relative increases in NO(x) emissions for many areas of the U.S. (e.g., south-southeast) that led to more extensive increases in ozone concentrations that are partly in agreement with observations. The widespread NO(x) emissions changes also alters the O(3) photochemical formation regimes, most notably the NO(x) emissions decreases in March–April, which can enhance (mitigate) the NO(x)-limited (VOC-limited) regimes in different regions of CONUS. The average of all AirNow hourly O(3) changes for 2020–2019 range from about +1 to −4 ppb during March–September, and are associated with predominantly urban monitoring sites that demonstrate considerable spatiotemporal variability for the 2020 ozone changes compared to the previous five years individually (2015–2019). The simulated maximum values of the average O(3) changes for March–September range from about +8 to −4 ppb (or +40 to −10%). Results of this work have implications for the use of widespread controls of anthropogenic emissions, particularly those from mobile sources, used to curb ozone pollution under the current meteorological and climate conditions in the U.S.

The global confinement due to rising COVID-19 cases, in particular, travel restrictions and quarantining shelter-in-place orders are associated with significant decreases in surface transportation activity by about 50% globally (Le Quéré et al., 2020) . Consequently, the COVID-19 economic slowdown led to a "natural air pollution control experiment" due to widespread anthropogenic emissions reductions for atmospheric pollutants and their precursor gases, such as decreased oxides of nitrogen (NO x = NO + NO 2 ) emissions of up to ~ 50% (Zhang et al., 2020; EEA, 2020; U.S. EIA, 2020) . The COVID-19 impacts on air pollution occurred on a scale impossible to reproduce outside of such a global health emergency.

The emission reductions during the lockdown measures led to varying changes in air quality conditions across the world. During February-March 2020 in China, surface fine particulate matter (PM 2.5 ) and carbon monoxide (CO) concentrations decreased by about 25% and 17%, respectively (Yue et al., 2020) . It was approximated that although air pollutant emissions decreased by 40% in the Northern China Plain, there were high levels of PM 2.5 (hourly levels >200 μg m − 3 ) that persisted from late January to mid-February, while surface ozone (O 3 ) also increased sharply by 84% (Li et al., 2020) . In other cities, there were substantial decreases in NO x (56%) and PM 2.5 (42% in Wuhan and 8% in Europe), but surface ozone (O 3 ) increased in all cities (36% in Wuhan and 17% in China) (Sicard et al., 2020) . In fact, analyses of COVID-19 related air pollution changes show that ground level O 3 increased by 2-30% for 11 cities globally (Shi et al., 2021) . Ground-level O 3 is formed by chemical reactions of volatile organic compounds (VOCs) and NO x in the presence of sunlight, and its production rate can be altered via different chemical regimes and environments around the world (Sillman et al., 1990; Sillman 1995 Sillman , 1999 Pusede et al., 2012; Jin et al., 2013; Walaszek et al., 2018; Li et al., 2019) . High O 3 concentrations can lead to decreased lung function and cause impaired respiratory symptoms, which are particularly dangerous for young children, the elderly, and those with preexisting conditions including asthma, chronic obstructive pulmonary disease (COPD), lung cancer, and respiratory infection (Kar Kurt et al., 2016) .

The COVID-19 economic slowdown and related traffic decreases that reached up to ~50% during March-May in the U.S. (INRIX; U.S. News, 2020) led to notable changes in pollutants in urban areas of the country, particularly for NO 2 concentrations that declined significantly as recorded from both satellite-and ground-based observations. Ground-based monitors indicated statistically significant declines for NO 2 of 26.0% or 5.4 ppb in urban counties (Berman et al., 2020) , while the TROPOspheric Monitoring Instrument (TROPOMI) satellite observations showed that the NO 2 decreases ranged between 9.2% and 43.4% among 20 cities in North America (Goldberg et al., 2020) . Qu et al. (2021) finds that surface and satellite NO 2 measurements at U.S. EPA ground based sites with the 5% highest concentrations show consistent reductions of 22-26% in March-April and 8-13% in May-June 2020 compared to the same months in 2019.

It was clear by mid-May in the U.S, however, that surface O 3 levels had not fallen in coincidence with decreasing NO x levels across much of the U.S. (NPR, 2020) , where various analyses showed variable O 3 changes across different U.S. regions (Arunachalam et al., 2020; Ivey et al., 2020; Kang et al., 2020; Van Haasen, 2020; Yang et al., 2020) . The high spatial O 3 variability may have been exacerbated by relaxed shelter-in-place orders for many U.S. states in early-to mid-June that led to rebounding anthropogenic emissions in some regions.

Motivated by the apparent COVID-19 related O 3 variability in different global regions and in portions of the U.S., we hypothesize that there are regional differences in the impact of COVID-19 lockdowns on O 3 formation across the U.S. This study combines the use of both ground and satellite-based observations of NO 2 to infer changes in precursor emissions due to the COVID-19 economic slowdown, and then uses the derived emission changes and a chemical transport model to quantify the related changes in O 3 concentrations across the entire contiguous U. S. (CONUS). Here we focus on O 3 as an indicator of pollution changes due to the COVID-19 economic slowdown in the U.S. because 1) changes in traffic NO x emissions strongly impact O 3 formation, 2) O 3 is the dominating pollutant contributing to non-attainment zones in the warm summer months (Zhang et al., 2019; U.S. EPA, 2020) , and 3) because O 3 has well-defined health impacts (Anenberg et al., 2009; Fann et al., 2012) .

The National Air Quality Forecasting Capability (NAQFC) used in this work is a well-documented and evaluated air quality modeling system (Eder et al., 2006 (Eder et al., , 2009 Mathur et al., 2008; Stajner et al., 2011; Lee et al., 2017) . The NAQFC used in this work is based on the offline-coupled North American Mesoscale Model Forecast System on the B-Grid (NMMB) (Black, 1994; Janjic and Gall, 2012) , which provides the driving weather data to the Community Multiscale Air Quality (CMAQ) model, version 5.0.2 (U.S. EPA, 2012). CMAQ simulates the formation, transport, and fate for a suite of atmospheric composition parameters. The NAQFC has provided real-time air quality forecast guidance over the past decade for different EPA-defined criteria pollutants, including near-surface O 3 at a horizontal resolution of 12 × 12 km centered over CONUS. The main chemical configurations for the NAQFC includes the CB05-TuCl/Aero6 gas-aerosol phase mechanism with simple aqueous phase chemistry reactions (see Lee et al., 2017 for further details). The NAQFC chemical initial conditions are started from a "warm-start" off of previous operational model output ending on March 01, 2020 and incur a two-week spin-up time (i.e., March 01-14 is removed from analyses). The time-dependent chemical lateral boundary conditions below 7 km altitude are provided by the Goddard Earth Observing System-Chemistry (GEOS-Chem; http://acmg.seas.harvard. edu/geos/) model output for 2006 that are mapped to appropriate CMAQ species and tuned for the operational NAQFC (Bey et al., 2001; Lam and Fu, 2009; Tang et al., 2009) . We note that the version of the NAQFC used here has some slight differences compared to the operational NAQFC products (i.e., BASE simulation: see Section 2.2 below), which includes use of a single cycle 12Z forecast (as opposed to 4-cycle 00, 06, 12, and 18Z runs), and omission of wildfire emissions. These NAQFC configuration differences are not expected to make appreciable impacts on the O 3 changes under the COVID-19 scenarios investigated here.

The U.S. EPA AirNow observation network data (https://www.ai rnow.gov/) are ideally suited for studying the ongoing impacts of the COVID-19 economic slowdown on O 3 as they provide real-time air quality information across the U.S. for over 2,000 monitoring stations via the AirNow Application Programming Interface (https://docs.airno wapi.org/), while also maintaining a consistent historical data record and format to readily compare the average conditions of air pollution for different U.S. regions. The AirNow observation data are downloaded, processed/analyzed, and paired in space and time with the NAQFC simulation grid cells for the same hourly time periods using NOAA's Model and Observation Evaluation Toolkit (MONET) (Baker and Pan, 2017) . The AirNow observations are used to assess the spatial changes in O 3 and to compare and evaluate the NAQFC simulations.

The NAQFC 2020 simulations are based on the U.S. EPA National Emissions Inventory 2014v2 (NEI2014v2; U.S. EPA, 2018), and form the "BASE" case. We quantify the COVID-19 effect using the difference between a "business-as-usual" (BAU) and COVID-19 (C19) case, both based on the NEI2014v2 emissions (the BASE case). The BASE case is not projected into the forecast year, with the time lag being a long-recognized issue in NAQFC (e.g., Tong et al., 2012) . In both BAU and C19 cases, however, the emissions are projected from the BASE year (2014) to 2020. BAU and C19 use the same observation-based trend adjustment for the period of 2014-2019, with the only difference being the change from 2019 to 2020.

In the BAU case, the 2019 to 2020 emissions (mobile and area sectors) change is assumed the same as the 2014-2019 period, so the oneyear change is the mean changing rate (% per year) from the pre-COVID-19 period. This is the best estimate if the COVID-19 pandemic did not happen. In the C19 case, the observed satellite and ground-based observed NO 2 changing rate from 2019 to 2020 is used to represent the actual emission progression under the pandemic. Compared to the annual emission changes in the pre-COVID-19 period, there was a much larger change from 2019 to 2020 in many states of the CONUS.

For both cases, we derive the NO 2 trend data using the approach developed by Tong et al. (2015 Tong et al. ( , 2016 . In this approach, the emission-changing rate for each state is derived according to the following equation:

where AF is the emission adjustment factor (rate of emission change), ΔS and N S are the rate of change and the number of satellite data, respectively. ΔG and N G are the rate of change and the number of groundbased data, respectively. f S and f G are two weighting factors applied to the satellite and ground data, respectively. The values of ΔS and ΔG are calculated for each state using the vertical column density from the Ozone Monitoring Instrument (OMI) aboard the Aura satellite (Levelt et al., 2006 (Levelt et al., , 2018 , and the U.S. EPA Air Quality System (AQS) ground network (https://www.epa.gov/aqs), respectively. Here the value of f S is set to be 1 and f G to be 100 to place more weight in the ground-based AQS observations, and the OMI data is filtered with a low-value cutoff value (0.7 × 10 15 ) to further remove influence from retrieval noise and background sources of NO 2 (Tong et al., 2015 (Tong et al., , 2016 . Furthermore, the AQS NO 2 changes are selected from the early morning rush hours (0600, 0700, and 0800 local time) that are more representative of transportation emission sector changes during the COVID-19 economic slowdown (Tong et al., 2016) . Further details on the data processing and quality control procedures are provided in Tong et al. (2015) . We use OMI as opposed to other newer satellite observations (e.g., TROPOMI) because of the well validated, and relatively long data record of OMI dating back to 2004 (Lamsal et al., 2020) , which can more appropriately predict the emission trends used in deriving the BAU and COVID-19 cases.

The percent change in emission AFs between March-September (monthly averages) indicate widespread decreases in C19 emissions for March-May, with more states shifting to lesser decreases, or some relative increases by June-September (Fig. 1) . The emission AFs are applied to all mobile and area source sectors and emission species, which impact their associated chemical species and consequential O 3 formation in the NAQFC simulations under each scenario. The difference in the predicted O 3 concentrations between the BAU and C19 cases is attributed to the impact of the pandemic. Emissions from point/energy generating units are not adjusted in this study. The full simulation period analyzed in this study is March 01 -September 30, 2020 and is focused on the CONUS region.

The COVID-19 related economic slowdown generally led to decreased NO x emissions, but there is also state-to-state variability (Fig. 2) .

Between March-June ( Fig. 2a-c) , many states in the east (e.g., New York) showed a relatively large decrease in NO x emissions (up to ~ 50-65%; Supporting Fig. S1 ), while states in the west experienced smaller decreases (<25%) and some increases (e.g., Montana and South Dakota). During May-June ( Fig. 2c and d) and the following months, states in the east show progressively smaller emission decreases, with some states shifting to relative emission increases compared to BAU. Interestingly, some states in the west show larger emission decreases during and after June.

The calculated emission trends in the Central and Eastern U.S. mainly indicate lower NO x emissions for C19 compared to BAU for March-April, which then predominantly increase over time and become more similar to the BAU emissions by June-July. In fact, the C19 NO x emissions become higher than BAU for the Southeast and Upper Midwest U.S. in July (Fig. 3 ).

The regional and state-level variability in BAU and C19 emissions changes are impacted by differences in state-mandated shelter-in-place and partial recovery activities that occurred later in the spring and summer months during the pandemic, and largely from the impact that these protocols had on emissions from on-road passenger vehicles across the U.S. Mobility data from Apple Inc. (Apple Inc., 2020) indicates that there were widespread driving decreases and relatively high variability in March-May, which shifts to general driving increases and lower spatial variability in May-September (Supporting Fig. S2 ). This does not fully account for different transportation modes, and excludes much of the national and state-level truck traffic used in large-scale domestic shipping (i.e., Federal Highway Administration vehicle classes 5-13 or vehicle length >23 feet), which did not show similar trends (htt ps://www.ms2soft.com/traffic-dashboard/).

Relative to passenger vehicle emissions, the emissions from commercial vehicles and electricity demand sources are relatively unchanged and impacts the changes in secondary air pollution formation during the COVID-19 lockdown (Archer et al., 2020) . This affects the use of OMI-AQS NO 2 changes that are inclusive of all emission sector changes during COVID-19. Other emissions studies also agree with the timing of emissions changes estimated here and indicate that there were 13% decreases in transportation-related emissions early in March-May 2020 (Shilling et al., 2020) , and then rebounding emission trends from May-July, particularly by mid-June through September when states (1)) for the a)-g) "Business-As-Usual" (BAU) and h)-n) COVID-19 (C19) cases. began lifting restrictions (Le Quéré et al., 2020) . Furthermore, the shift to cleaner vehicle fleets in the BAU trend may also have led to fewer emissions compared to the C19 shift to more domestic, heavy-duty truck traffic (typically using diesel fuel) during the morning rush hours compared to less passenger traffic during the lockdown.

The shift to smaller decreases or increases for the OMI-AQS inferred NO x emissions in the summer months may also be impacted by an increasing background NO 2 concentration that partly limits the use of OMI observations in this work (discussed more below). The western U. S., which is characterized by relatively lower NO x emissions, indicates smaller decreases for C19 compared to BAU during all months, except September where C19 becomes slightly higher than BAU.

Evaluation of the BASE O 3 simulations (using NEI2014v2 emissions) against the U.S. EPA AirNow network for April-September 2020 shows that the NAQFC model performance is acceptable, while consistently falling within statistical criteria ranges for O 3 in Emery et al. (2017) . The exceptions are a slightly high normalized mean bias (NMB) and low correlation coefficient (R) in the Northeast (Supporting Table S1a ), and slightly high NMB in the Upper Midwest in May (Table S1e) , Central Plains U.S. regions in May, and Northwest U.S. in April-May (Table S1j ). The O 3 simulations for the BAU and C19 scenarios demonstrate mainly increases in the correlation, R, and Index of Agreement (IOA), and Regional NO x emissions (mol s − 1 ) trends ending on April 15 -September 15, 2020 for the BASE (black; NEI2014v2), "Business-As-Usual (BAU) (green), and COVID-19 (C19) (red) cases in the different U.S. EPA regions (https://www.epa.gov/aboutepa/regional-and-geographic-offices). Note the change in emissions scales on the y-axis. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) decreases in Normalized Mean Error (NME) compared to the BASE case.

Analysis of the 2020-2019 changes in observed and simulated BASE maximum daily 8-h average (MDA8) O 3 at 25 of the highly polluted cities for O 3 based on the American Lung Association "State of the Air® 2019" report (American Lung Association, 2020) provides important spatial information on the impacts of the lockdown on local air quality in areas of already elevated pollution (Fig. 4) . A similar analysis for the 2020 changes against the previous 5-year average (2015-2019) are shown in Supporting Fig. S3 , and are in good agreement with the changes shown in Fig. 4 for 2020-2019.

During the initial COVID-19 lockdown period in March-April, there are predominantly MDA8 O 3 decreases at all sites, with changes ranging from +10% to − 35% (e.g., Houston, TX) at the highly polluted cities. Rather expectedly, the BASE simulation does not show as large of an O 3 decrease in March-April due to the use of NEI2014v2 emissions that do not take into account COVID-19 related changes. Later in May and the O 3 -season months of June-August in the U.S., there are many more polluted cities that experience MDA8 O 3 increases, particularly in southern California (e.g., El Centro, CA) and eastward (with changes up to ~ +40 to +55%). There is also more observed site-to-site variability during the summer ozone season months in the U.S. The BASE simulation again does not capture the higher variability in ozone changes during the summer months of June-August. Later in September, mainly all sites in California show increases in MDA8 O 3 , while all sites in the Mid-Atlantic and the northeast U.S. show decreases. This is in better agreement with the BASE simulations and suggests both rebounding mobile emissions and a potentially strong role from natural, i.e., meteorological variability impacts on the ozone changes.

Overall, the MDA8 O 3 increases are mainly <30%, except for Houston TX in May and August, El Centro CA in July, and Los Angeles and Redding CA in September, which have increases >40%. In the eastern Mid-Atlantic U.S. cities, such as in Arlington, VA, there were relatively small decreases in MDA8 O 3 from March-August (≤15%), but in September there was a larger decrease of ~35%. In the northeast cities, including Hartford, CT, there are moderate MDA8 O 3 changes on the order of about ±15%. The largest decrease (increase) in MDA8 O 3 based on these cities was ~35% (55%) observed at Dallas TX (Houston TX) in March (August), and demonstrate the significant spatiotemporal variability in ozone changes during the COVID-19 economic slowdown.

The evaluation results (Supporting Table S1a ) and differences in AirNow site comparisons (Fig. 4) against the BASE simulation confirm the importance in projecting the emissions using available observations and the BAU and C19 trend scenarios to quantify the spatial changes in O 3 across the CONUS.

Analysis of the observed changes for 2020-2019 shows that there are widespread, moderate decreases in the MDA8 O 3 (Fig. 5a ; average ~ − 5.8 ppb) and hourly O 3 ( Fig. S4a ; average ~ − 4.3 ppb) across the AirNow sites in March-April, which shifts to smaller decreases and more widespread increases at the sites later in April-May.

Later in May-September there are more variability in the regional changes, where aside from the predominant ozone decreases at the sites in the southeast, there are many regions that show increases such as the upper Midwest and North Central Plains, and much of the western U.S. including California in August-September.

Comparing the observed AirNow ozone conditions between 2020 and 2019 are somewhat limited by the year-to-year variability regardless of the pandemic, and can underestimate the true impact of the COVID-19 economic slowdown. Thus, Table 1 There are predominantly decreases in hourly and MDA8 observed O 3 for 2020 compared to all previous 5 years, where the largest decreases (Table 1 ; 5-yr average hourly/MDA8 O 3 ~ − 3 ppb/-4 ppb). There are, however, year-to-year variability for the 2020 O 3 changes that demonstrates the role of non-linear formation from varying emissions and meteorological conditions. There are only a few instances of increased 2020 ozone, mostly during April-May, which demonstrated the smallest 5-year average AirNow observed decreases (~− 0.3 ppb/-1.3 ppb). There are predominantly decreases in hourly (MDA8) O 3 during June-September that ranged from − 1 to − 1.5 ppb (− 1.1 to − 2.4 ppb) for 2020 compared to the 5-year average. The majority of AirNow sites are situated in highly populated, urban/city locations, and are not as representative of the surrounding rural regions in CONUS. Overall, the AirNow observed 2020 O 3 changes in Figs. 4 and 5 show significant spatiotemporal variability during the pandemic period of March-September compared to the previous 5 years (2015-2019), and thus the experimental NAQFC C19 and BAU simulations are further used to quantify the O 3 changes in both urban and rural areas throughout CONUS.

The differences in the NAQFC sensitivity simulations (C19-BAU) qualitatively agree with AirNow observations and show widespread decreases in monthly average absolute (Fig. 6 ) and percent O 3 changes (Fig. S6 ) due to the COVID-19-related mobility changes and reduced NO x emissions in March-June, which are strongly controlled by the NO xlimited photochemistry across the rural regions of the U.S.

The robust photochemical indicator O 3 /NO y (Sillman et al., 1997; Liang et al., 2006; Zhang et al., 2009; Campbell et al., 2015) is used here to show the regions of NO x -limited and VOC-limited chemistry for the BAU and C19 cases in the U.S. (Supporting Figures S8a-S8b ). In the major urban regions, there are increases in surface O 3 associated with the NO x emission decreases. There are also local enhancements in the ozone decreases near urban regions in states with increased NO x emissions. These regions are characterized by VOC-limited photochemistry, where the decreases (increases) in NO x emissions lead to increases (decreases) in O 3 formation, as the change in NO x emissions are in proportion to the change in VOC emissions (not shown). During June-September, there is a shift to more widespread, but relatively smaller increases in O 3 from the south and southeast to western parts of the U.S., which are associated with increases in NO x emissions across the rural, NO x -limited regions (Fig. 2) , but not near the major urban centers that the model suggests are VOC-limited during this time. The areas of decreased C19 NO x emissions (more prolific in March-June) tend to enhance the NO x -limited conditions in the rural areas, and strongly mitigates the VOC-limited conditions in urban areas due to the additional increases in O 3 . The areas of increased C19 NO x emissions (more prolific in June-September) tends to mitigate NO x -limited conditions near the rural regions, and moderately enhances the already VOC-limited regions in urban regions (Supporting Fig. S8c ). We note that uncertainties in the biogenic VOC emissions in the NAQFC simulations will affect the calculation of possible summertime shifts in more NO x -limited regimes in some locations, particularly in the southeast CONUS where there is significant vegetation.

These results are also in qualitative agreement with the AirNow observations that show more areas of increased O 3 during the summer 

Average of all AirNow site hourly (MDA8) O 3 changes in ppb for the 2020 COVID-19 economic slowdown periods compared to the previous five years (2015-2019). Fig. 7 . Same as in Fig. 6 , but for the median value of MDA8 O 3 . months, particularly in the central and southern U.S. during June-August, and in California during August-September. The model agreement in the different CONUS regions tends to have dependence on the specific previous year compared to 2020 AirNow observations, where the simulated widespread enhanced ozone increases during summer has arguably the best spatial agreement with the observed 2020-2016 and 2020-2015 changes (Supporting Fig. 4d and e) . The simulated shift to more widespread increases in O 3 across the southern U.S. also agrees with NASA's global atmospheric composition model (GEOS-CF) simulations of O 3 that used sustained reductions in global anthropogenic emissions of NO x , carbon monoxide (CO) and VOCs compared to a BAU scenario. This demonstrates the high non-linearity of O 3 chemistry, which leads to widespread O 3 increases in the southern portions of the U.S. in July-August (Keller et al., 2020) . There are also localized increases in simulated ozone during all months in the urban coastline regions of the Long Island Sound (LIS) that span across the New York, New Jersey, and Connecticut state borders. This is in good agreement with the AirNow observations, where there are either hourly ozone increases or weaker decreases in the LIS region for the previous years compared to 2020 (Supporting Figs. S4a-4e) . The spatial patterns for the monthly median absolute (Fig. 7) and percent MDA8 O 3 changes (Supporting Fig. S7 ) are similar to hourly O 3 , but there are indications of larger decreases and increases for MDA8 O 3 in the enhancement regions.

Increases in the peak of daily O 3 values has consequence for the regions close to or in non-attainment for ozone standards. This is particularly true for the south-southeast states and other localized urban areas of the U.S. where increases in MDA8 O 3 of up to ~3-7 ppb may have affected the total number of O 3 exceedance days during the 2020 summer season.

There are also prominent decreases in MDA8 ozone in the northeast and western U.S., most notably in New York State during May-August, and in California, Oregon, and Washington during June-August. Decreases of up to ~3-5 ppb of MDA8 ozone are also important, and can result in alteration of the atmospheric oxidation capacity that can effectively change the production rate of secondary particles during the warm season (not shown). Co-impacts of O 3 and secondary particulate matter formation changes during the COVID-19 pandemic should be more closely studied in future work.

Using both observations and NAQFC modeling (based on CMAQv5.0.2) we show that the COVID-19 lockdown caused variable impacts on anthropogenic emissions and both hourly and maximum daily (MDA8) ozone concentrations across the U.S. There were widespread decreases in NO x emissions in the U.S. during March-June 2020, which led to widespread decreases in O 3 concentrations in the rural regions that are NO x -limited, but also localized increases near some of the highly populated urban centers that are VOC-limited. Later in June-September 2020, the data-fused AQS and OMI NO 2 changes suggested that many areas in the U.S. showed relative increases in NO x emissions for the C19 compared to BAU scenarios, and consequently the simulations suggest widespread increases in O 3 concentrations. The widespread NO x emissions changes also alters O 3 photochemical formation regimes, most notably the NO x emissions decreases in March--April, which can enhance (mitigate) the NO x -limited (VOC-limited) regimes in different regions of CONUS.

The average of all AirNow hourly (MDA8) O 3 changes for 2020-2019 range from about +1 to − 4 ppb (+1 to − 6 ppb) during March--September, and are associated with predominantly urban monitoring sites that demonstrate considerable spatiotemporal variability for the 2020 ozone changes compared to the previous five years (2015-2019). The simulated maximum values of the average hourly (MDA8) O 3 changes for March-September range from about +8 to − 4 ppb (+7 to − 5 ppb), which correspond to relative changes that range from about +40 to − 10% (+15 to − 10%). Overall, the variable spatial patterns in modeled ozone changes are in good agreement with the changes at the predominantly urban-limited AirNow sites; however, there are differences that vary depending on the specific previous year compared to 2020 during the COVID-19 pandemic.

To place our results into context, recent international results also show high spatiotemporal COVID-19 related O 3 changes both globally and regionally. Venter et al. (2020) showed that compared to the previous 3-year period, the COVID-19 lockdown led to marginal increases in observed O 3 of about 4% averaged across 34 countries during lockdown dates up until May 15, but that there were both increases and decreases at specific ground observation sites across the U.S. in January-May. This is in qualitative agreement with our results of spatial variability in the year-to-year changes for ground-site observations ( Fig. 5 and S4 -S5) and model-based results with both increased and decreased O 3 for different CONUS regions by May 15. Chossiere et al. (2021) found that of the 146 of the 252 regions analyzed showed O 3 decreases in response to the COVID-19 lockdowns, with 45 of those regions being statistically significant. However, they further indicate there was no statistically significant change in O 3 concentrations on average across in the U.S., with a range of − 5.6 to 4% (mean ~ 0.8%). The contrasting results of Chossiere et al. with our work may be due to different analysis methods, and because we more closely analyzed model-grid cell (12 × 12 km) resolved MDA8 O 3 changes later in the U.S. ozone season, which showed relatively higher local O 3 changes (~− 40 to +10%) due to the COVID-19 lockdown. Bray et al. (2021) showed widespread and rather prolific increases in observed ground level O 3 at AirNow ground-based sites during the peak March-April lockdown period in the U.S. The Bray et al. analysis is only in partial agreement with our results of observed AirNow changes in O 3 concentrations in some U.S. regions and specific year-to-year changes, particularly for the regions of increased O 3 in April-May (Table 1, Fig. 5 , and Supporting Figs. S4-S5). Indeed, there are some discrepancies between studies for unknown reasons at this time. Our results here extend the quantitative analysis of COVID-19 driven O 3 impacts further into the warmer "ozone season" months in CONUS (i.e., May-September 2020), which have implications for widespread emission changes during the most prolific ozone non-attainment months in the U.S.

Use of the satellite and ground-based NO 2 observations to reflect the actual COVID-19 related emission changes in 2020 has some limitations. The ground-based changes in NO 2 are heavily weighted towards the predominantly urban site locations, where the meteorological conditions and natural NO 2 emission sources, i.e., natural variability, plays an important role in accurately attributing the observed NO 2 changes to NO x emissions and consequently the NAQFC simulated O 3 concentration changes. Goldberg et al. (2020) found that the NO 2 was relatively low in 2020 (compared to 2019) based on the meteorological conditions in many regions, particularly the prevailing wind directions that transported predominantly clean air masses to major cities areas such as Miami, FL, and Washington D.C. The numerous natural sources of NO 2 (e.g., lighting, soil, and wildfires) that heavily impact background concentrations further confound the use of OMI NO 2 observations to infer the actual COVID-19 related emissions changes, especially during the summer months (Qu et al., 2021) .

The COVID-19 economic slowdown and natural experiment of widespread reductions in precursor emissions in different ozone formation regimes in the U.S. has implications for emission control strategies used to curb elevated pollution near major cities and remaining non-attainment areas in the U.S. The impact of the COVID-19 pandemic sheds further light onto the importance of significantly controlling VOC emissions in and around urban areas, such that widespread controls on NO x emissions are not efficient at reducing ozone levels in regions that experience similar reductions in VOC emissions. This suggests that VOC emission reductions may need to be even greater than NO x reductions in VOC-limited cities in the spring/summer. This work shows that there were widespread reductions in ozone levels during the initial period of COVID-19 lockdown, but there are indications of predominantly increased ozone near major cities and possibly more widespread increases later in the summer. It is these regions of elevated ozone that are of concern for those suffering with preexisting health issues, where the increased exposure to air pollution due to the COVID-19 related emissions changes may exacerbate the human susceptibility, health impacts, and spread of the COVID-19 virus itself (Chakrabarty et al., 2020; Wu et al., 2020) . Undoubtedly, more research is needed on the interconnections between air quality, exposure and susceptibility, and public health management during widespread pandemics and the many unintended consequences that follow.

For emissions adjustment factor preparation, the OMI/Aura satellite data are available at https://aura.gesdisc.eosdis.nasa.gov/data/Aura _OMI_Level2/OMNO2.003/ and the U.S. EPA AQS ground observation data are available at the AQS Application Programming Interface (API) at https://aqs.epa.gov/aqsweb/documents/data_api.html. The U.S. EPA AirNow observations are also available via the AirNow API at https://do cs.airnowapi.org/. The National Air Quality Forecasting Capability (NAQFC) code used in this work, which is based on CMAQv5.0.2, is published on Zenodo at http://doi.org/10.5281/zenodo.1079888 and is also available for download on GitHub at https://github.com/US EPA/CMAQ/tree/5.0.2. The raw, gridded emissions and NAQFC simulation output are very large (multiple Terabytes) and are freely available in two different ways: (1) on the local NOAA/NCEP repositories that can be directly transferred via SFTP or SCP in pieces, or manually copied in their entirety and provided via external hard drives, and (2) on a publicly available repository/transfer in pieces or via another high-speed file transfer service such as Globus (https://www.globus.org/datat ransfer).

The scientific results and conclusions, as well as any views or opinions expressed herein, are those of the author(s) and do not necessarily reflect the views of NOAA or the Department of Commerce. 

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.

State of the Air

Intercontinental impacts of ozone pollution on human mortality

reportApple COVID-19 Community Mobility Reports

Changes in air quality and human mobility in the USA during the COVID-19 pandemic

Impacts of COVID-19 related shutdown on onroad and air transportation emissions-related O3, NO2 and PM2.5 in the U.S. Using sensitivity modeling techniques

Can We Compare the COVID-19 and 2008 Crises?

Overview of the model and observation evaluation Toolkit (MONET) version 1.0 for evaluating atmospheric transport models

Changes in U.S. air pollution during the COVID-19 pandemic

Global modeling of tropospheric chemistry with assimilated meteorology: model description and evaluation

The new NMC meso-scale Eta Model: description and forecast examples

Impact of lockdown during the COVID-19 outbreak on multi-scale air quality

A multimodel assessment for the 2006 and 2010 simulations under the air quality model evaluation international initiative (AQMEII) phase 2 over North America: Part I. Indicators of the sensitivity of O3 and PM2.5 formation regimes

Ambient PM2.5 Exposure and Rapid Spread of COVID-19 in the United States

Air pollution impacts of COVID-19-related containment measures

An operational evaluation of the Eta-CMAQ air quality forecast model

A performance evaluation of the national air quality forecast capability for the summer of

Recommendations on statistics and benchmarks to assess photochemical model performance

Air Quality and COVID-19

Estimating the national public health burden associated with exposure to ambient PM2.5 and ozone

Disentangling the impact of the COVID-19 lockdowns on urban NO2 from natural variability

Traffic, precursor emissions, and ozone in the south coast air basin during California's COVID-19 shutdown

Scientific Documentation of the NCEP Nonhydrostatic Multiscale Model on the B Grid (NMMB). Part 1 Dynamics

Role of meteorological processes in ozone responses to emission controls in California's San Joaquin Valley

Air quality changes under COVID-19 social distancing in the United States: observational analysis and modeling sensitivity study

Pulmonary health effects of air pollution

Global impact of COVID-19 restrictions on the surface concentrations of nitrogen dioxide and ozone

A novel downscaling technique for the linkage of global and regional air quality modeling

OMI/Aura nitrogen dioxide standard product with improved surface and cloud treatments

Temporary reduction in daily global CO2 emissions during the COVID-19 forced confinement

NAQFC developmental forecast guidance for fine particulate matter (PM 2.5)

The ozone monitoring instrument

The ozone monitoring instrument: overview of 14 years in space

Anthropogenic drivers of 2013-2017 trends in summer surface ozone in China

Drivers for the poor air quality conditions in North China Plain during the COVID-19 outbreak

Evaluation of the ability of indicator species ratios to determine the sensitivity of ozone to reductions in emissions of volatile organic compounds and oxides of nitrogen in northern California

Assessment of the wintertime performance of developmental particulate matter forecasts with the Eta-Community Multiscale Air Quality modeling system

Traffic Is Way Down Because of Lockdown, but Air Pollution? Not So Much

On the observed response of ozone to NOx and VOC reactivity reductions in San Joaquin Valley California 1995-present

US COVID-19 Shutdown shows Importance of Background NO2 in Inferring Nitrogen Oxide (NOx) Emissions from Satellite NO 2 Observations

Abrupt but smaller than expected changes in surface air quality attributable to COVID-19 lockdowns

Special Report 3: Impact of COVID19 Mitigation on Traffic, Fuel Use and Climate Change

Amplified ozone pollution in cities during the COVID-19 lockdown

The use of NOy, H2O2, and HNO3 as indicators for ozone-NOxhydrocarbon sensitivity in urban locations

The relation between ozone, NOx and hydrocarbons in urban and polluted rural environments. Millenial Review series

The sensitivity of ozone to nitrogen oxides and hydrocarbons in regional ozone episodes

The use of photochemical indicators to evaluate ozone-NOx-hydrocarbon sensitivity: case studies from Atlanta

US National Air Quality Forecast Capability: Expanding Coverage to Include Particulate Matter

The impact of lateral boundary conditions on CMAQ predictions over the continental United States

New Direction: the need to develop process-based emission forecasting models

Long-term NOx trends over large cities in the United States during the 2008 Recession: intercomparison of satellite retrievals, ground observations, and emission inventories

Short-Term Energy Outlook

Office of Research and Development

Office of Air Quality Planning and Standards

Office of Air Quality Planning and Standards

Nonattainment Areas for Criteria Pollutants (Green Book)

Traffic Levels 'Unprecedented' amid Coronavirus

Effects of COVID-19 on transportation and air quality in the north central Texas region

The role of precursor emissions on ground level ozone concentration during summer season in Poland

Air pollution and COVID-19 mortality in the United States, 2020. Strengths and limitations of an ecological regression analysis

Improvement of the air quality in the LA basin during the COVID 19 outbreak based on real time data and CMAQ

Changes of anthropogenic carbon emissions and air pollutants during the COVID-19 epidemic in China (in Chinese)

Ozone pollution: a major health hazard worldwide

Probing into regional ozone and particulate matter pollution in the United States: 2. An examination of formation mechanisms through a process analysis technique and sensitivity study

NOx emission reduction and recovery during COVID-19 in east China

This study was co-funded by the National Oceanic and Atmospheric Administration's (NOAA's) Weather Program Office (Grant/Contract Number NA19OAR4590082) and the NASA Health and Air Quality Program (Grant/Contract Number 80NSSC21K0512). Support for this work was also provided by NOAA's Office of Oceanic and Atmospheric Research and the Cooperative Institute for Satellite Earth System Studies (CISESS).

Supplementary data to this article can be found online at https://doi. org/10. 1016/j.atmosenv.2021.118713.