Substantial decreases in NO2 emissions from reduced transportation volumes in US cities during COVID-19 shutdowns reveal health vulnerabilities of urban populations

,

has resulted in instituting stay-at-home orders around the world, which has many negative externalities associated with it, but one positive one has been a marked decrease in many criteria air pollutants due to decreases in transportation volumes and industrial production (Nakada & Urban, 2020;Sharma et al., 2020), including reduced emissions of nitrogen dioxide (NO 2 ).This change has also been quantified via satellite imagery which indicates more than 10% decline in tropospheric pollutant measurements over inhabited regions around the globe (Liu et al., in review;Venter et al., 2020).
As anthropogenic activities far surpass natural emissions (Walters et al., 2015) they have resulted in a three-to six-fold increase in nitrogen oxide (NOx = NO + NO 2 ) emissions since the pre-industrial era (Jaeglé et al., 2005).Sources of NOx include fossil fuel/biofuel combustion, industry, and transport category constituting of vehicles, ships, and aircraft, while as natural sources of NOx include soil nitrification-denitrification processes, wild fires and lightning (Walters et al., 2015).Road emissions from tail pipe emissions, resuspended dust particles and friction processes result in NOx, carbon monoxides and volatile organic compounds (VOC's) which has profound and measurable health implications in populations (Cesaroni Giulia et al., 2013;Krzyżanowski et al., 2005;Peel et al., 2005).Besides increasing acidification, global climate changes, decrease in visibility, and ozone and aerosol increases in the troposphere (Bermejo-Orduna et al., 2014), NOx also increases small particle formation (Galloway et al., 2003).
The onset of COVID-19 has posed a unique opportunity to quantify changes in vehicular NO 2 emission as a result of reduction of vehicle volume in the U.S. Due to its adverse health impacts, NO 2 emission, a precursor to ground-level ozone and particulate matter concentration, has resulted in its usage as a marker for combustion emissions over regions (Bechle et al., 2011).To examine changes in NO 2 in cities and how that relates to vehicular traffic and health status of the population during the COVID-19 pandemic, we examine the impact of stay-at-home orders  Bureau, May2020) Seven of the 15 cities traffic volume data was also accessed as reported by Department of Transportation's continuous sensor on a roadway segment in the respective cities.To get a uniform scale of vehicle usage, aggregate VMT data for the 15 counties was accessed from StreetLight Data (https://www.streetlightdata.com/our-data/)which run over 100 billion location data into an algorithm to aggregate and normalize travel patterns by region.

Tropospheric NO 2 data
For monthly averaged NO 2 tropospheric data (January through April of 2020), we acknowledge the free use of tropospheric NO 2 column data from the Global Ozone Monitoring Experiment-2 (GOME-2 (METOP-B)) satellite from www.temis.nl.This sun-synchronous satellite processes NO 2 concentrations from the ground up to about 10 Km and has a geometric pixel resolution of 60 x 30 km 2 .Due to environmental uncertainties and the density of the slant column retrieved by the sensor this data can accurately estimate tropospheric column with 35-60% precision (Boersma et al., 2004).Major chemical and transport processes related to NO 2 , along with cloud cover, also play a role in the retrieval process and uncertainty in these values.However, the integrated tropospheric column of NO 2 data is dominated by lower tropospheric amounts of NO 2 (Ma et al., 2006), which makes it a useful variable to incorporate for such studies.
Fifteen locations of the continuous NO 2 sensors were used to extract pixel values of the tropospheric NO 2 data utilizing ESRI's ArcGIS Desktop 10.8, which was then rescaled on a scale of 0-100 for visual comparison.

Health Data
To assess the health status of the cities studied, Estimated Prevalence and Incidence of Lung Disease data from American Lung Association (ALA) was accessed.This data estimation is available at a county level and is based on a Behavioral Risk Factor Surveillance Survey conducted in 2017 and 2018 Centers for Disease Control's (CDC) joint report with other state and national registries (https://www.lung.org/research/trends-in-lung-disease/prevalenceincidence-lung-disease).Additionally, county-level COVID-19 cases and death data was accessed from USAFacts (https://usafacts.org/), a not-for-profit organization providing U.S. government data.

Results
A sharp reduction in NO 2 was observed in 13 of the 15 cities examined, based on the same ~1month window from 2019 to shut-down conditions in 2020 (Table 2).All cities except Jacksonville, Florida showed a decline in the continuous NO 2 sensors, from -4% to -63%, with an average across the cities of -26% for weekdays and -24% on weekends (Table 2).Averaged monthly values during the study period show a decline in NO 2 values ranging from -5.89% to -59.7% (Table 3).  3

. Monthly average of NO 2 from continuous sensors on a road segment
The stay-at-home order resulted in a significant drop in VMT in the 15 counties in this study (Fig 1).For the seven cities where traffic volume data was available, the drop in weekday volume correlates with the decrease in NO 2 -an expected but nevertheless significant finding (Fig. 2), with the exception of San Jose, California.The relationship between traffic volume and NO 2 on the weekends is weaker (Fig. 3).
Most of the tropospheric NO 2 data from the 15 cities shows a decline in March and April 2020 as compared to 2019 (Table 4), with the month of March resulting in the highest aggregated decline of 34.5% (Fig. 4).New York showed the highest unit decline in March, and Houston showed the highest unit decline in April (Table 4; Figs. 5, 6)-both cities experienced the greatest unit decline during these months.Raster images from these two cities visualize this decline (Fig. 7).Health data at the county level from USAFacts and ALA indicates that in 2018, Marion County (the City of Indianapolis consolidated the entire county, thus county health data is at the same population scale as city data) had the highest percent of COPD and lung cancer cases, which are highly correlated at 0.84 (Table 6 and Table S1).Marion County ranked 4 th highest in percent asthma cases and 2 nd highest in percent pediatric asthma cases (Table S2).Percent adult asthma has a high correlation of 0.57 with percent COVID-19 cases and 0.42 with normalized VMT (Table S3).The five highest correlations among the health and data were percent COVID-19 cases, percent COVID-19 deaths, VMT normalized by area of the counties, asthma, and tropospheric NO 2 extracted from GOME-2 pixel values (Table S4, S5, S6).

County
COVID-19 data indicates that during our study period, COVID-19 case rates and death rates in Marion County were both 3 rd highest (Table S7, S8).The percentage of people with adult asthma shows the highest correlation with cases and death related to COVID-19 (Table S9), and correlations of mean NO 2 values indicate that it has the second highest positive correlation (0.32) with asthma cases in the study region (Table S10).

Discussion
High vehicular emissions can result in corridors of heavy pollution (Redling et al., 2013) in rural and urban regions.If left un-examined this can have increased adverse health effects on the population in the region, thus worsening conditions like respiratory disease, cardiovascular disease, and cancers, and even causing premature mortality (Lamsal et al., 2013;Filippelli et al., 2020).Findings in our study are consistent with other research which shows that NO 2 pollution is linked with increased asthma events in predominantly urban areas (Achakulwisut et al., 2019).Despite uncertainties from co-pollutants, short term exposure to NO 2 results in a likely causal relationship between it and ischemic heart disease (IHD) (Cesaroni Giulia et al., 2013;Stieb et al., 2020), and a 20 ppb increase in NO 2 results in increase in chronic obstructive pulmonary disease (COPD) hospital visits, cardiovascular disease, lung cancer in adults, and respiratory mortality (Cesaroni Giulia et al., 2013;Peel et al., 2005).
Most states in the U.S. started their stay-at-home order close to the third week in March of 2020.All the cities in this study except Jacksonville, FL significantly declined in NO 2 emission data from the continuous sensors in the road segments.The Jacksonville case was likely a result of a delayed start to the stay-at-home order in Florida, or perhaps too great of a mismatch between the location of the NO 2 monitor and the traffic volume sensor.When ground level data lacks consistency, tropospheric NO 2 satellite data, even with a geometric pixel resolution of 60 x 30 km 2 , can be utilized in a meaningful way to examine various regions.GOME-2 data here also shows Jacksonville with the highest increase of 18% in tropospheric NO 2 column in March 2020, but in April 2020 it decreases down to almost the same levels as 2019, when most cities in the U.S. followed the stay-at-home orders.It is important to keep the difference in spatial resolution in mind when comparing ground level sensor data to satellite measurements (Drosoglou et al., 2017).
In comparing traffic volume and NO 2 emissions in 7 of the 15 cities we find traffic volume reduction and NO 2 emissions following a similar trend of substantial declines during weekdays, with the exception of San Jose.Since the traffic volume sensors and the NO 2 sensor are not colocated, we need to be careful in pairing the two sets of data.For a comprehensive examination, VMT can also be used as a proxy to NO 2 emissions or in conjunction with ground level sensor.It is important to note that meteorological conditions like temperature, wind speed, relative humidity, and precipitation which play a role in transport of atmospheric gases (Tobías et al., 2020) and particles were not considered in this analysis.
Overlaying available health data from ALA and USAFacts, we find that Queens (NY) had the highest case rate and death rate from COVID-19.Marion County (Indianapolis) was third in place for both at 0.47% and 4.74%.VMT (normalized by area of each county) has the highest correlation of 0.71 with COVID-19 case rate and 4 th highest at 0.42 with percent adult asthma cases (Table S4) which in turn has a 0.58 correlation with percent COVID-19 cases (Table S9).
Correlations of percent COVID-19 case rate, death rate, and VMT normalized by area, and ground level NO 2 all include asthma and tropospheric NO 2 values.Disability Adjusted Life Years (DALYs) can be calculated based on population exposure to a number of pollutants (e.g., Landrigan et al., 2017), including criteria air pollutants such as ozone and PM2.5.For the purposes of this study using NO2 only, and for a short window of time during which NO 2 decreases, DALY calculations are not appropriate.We assume that the decreases would have to be substantial and long-lived to yield a life-time health benefit, but our results do point to a future for many US cities where improved population health due to a decrease in air pollution is achieved through electrifying vehicular fleets and improving industrial emission controls.

Conclusion
These results reveal a number of critical relationships between traffic volume, local emissions of NO 2 , and the pre-existing health conditions of those most heavily impacted by air pollution, which may make them more susceptible to the more severe presentation of COVID-19 disease: 1.A substantial decline in NO 2 can be driven largely by policy-in this case, crisis policy involving virtually locking down vehicular traffic in cities.
2. Many urban areas have substantial percentages of the population with pre-existing conditions, potentially linked to air pollution exposure, which may make them more susceptible to severe COVID19 disease.
3. Linking NO 2 data derived from ground-based and satellite-borne sensors is useful for filling in key spatial data gaps and for contextualizing the sparse ground-level data with more spatially integrative satellite observations.The silver lining provided by shut-down related air quality improvements are likely temporary, but lay bare the reality that air pollution likely makes inhabitants of some cities quite vulnerable to those very co-morbidities that exacerbate COVID-19 disease.

Acknowledgements
For monthly averaged NO 2 tropospheric data (January through April of 2020), we acknowledge the free use of tropospheric NO 2 column data from the Global Ozone Monitoring Experiment-2 (GOME-2 (METOP-B)) satellite from www.temis.nl.This work was partially supported by the Environmental Resilience Institute, funded by Indiana University's Prepared for Environmental Change Grand Challenge Initiative, and by National Science Foundation award ICER-1701132 to Filippelli.
(https://www.streetlightdata.com/our-data/).Health data is publicly availability via the Estimated Prevalence and Incidence of Lung Disease data from American Lung Association (ALA) was accessed.This data estimation is available at a county level and is based on a Behavioral Risk Factor Surveillance Survey conducted in 2017 and 2018 Centers for Disease Control's (CDC) joint report with other state and national registries (https://www.lung.org/research/trends-in-lung-disease/prevalence-incidence-lung-disease).
Tropospheric NO 2 column data from the Global Ozone Monitoring Experiment-2 (GOME-2 (METOP-B)) satellite is available upon request from www.temis.nl.

Contributions to this work
Heitzelman and Filippelli conceptualized this work, performed analyses, and wrote the manuscript.Lulla assisted with statistical analysis and tropospheric NO 2 satellite pixel analysis.

Table S9 .
Correlation of diseases and COVID-19 for all 15 cities studied.

Table S10 .
Correlation of NO 2 with disease in all 15 cities measured pollution and health outcomes can be achieved quickly with local and state policy directives, perhaps leading to more population-level health resilience in the face of future pandemics.
. were accessed through the respective state agencies for our study period.Due to the recent nature of this data, the 2020 NO 2 values had not been validated at the time of retrieval.
from March 23 -April 24, 2020, as compared with March 25 -April 26, 2019.We utilize calibrated high-quality daily data for NO 2 from EPA grade sensors in cities around the US.NO 2 emissions in 15 of the top 17 most populous cities in the U.S. (Table1) are assessed, and compared to satellite results.We also examined traffic comparative traffic volumes, and assessed the health status of inhabitations of cities to project potential theoretical health benefits of NO 2 reductions and vulnerabilities to severe forms of COVID-19 disease due to asthma, COPD and lung cancer2.Methodology2.1 NO 2 and Vehicle Miles travelled (VMT) dataTo examine the impact of the stay-at-home orders, NO 2 daily averaged data from continuous ground level sensors from road segments in 15 of the top 17 populated (Table1) cities in the U.S