Global Air Quality and COVID-19 Pandemic: Do We Breathe Cleaner Air?

The global spread of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) has challenged most countries worldwide. It was quickly recognized that reduced activities (lockdowns) during the Coronavirus Disease of 2019 (COVID-19) pandemic produced major changes in air quality. Our objective was to assess the impacts of COVID-19 lockdowns on groundlevel PM2.5, NO2, and O3 concentrations on a global scale. We obtained data from 34 countries, 141 cities, and 458 air monitoring stations on 5 continents (few data from Africa). On a global average basis, a 34.0% reduction in NO2 concentration and a 15.0% reduction in PM2.5 were estimated during the strict lockdown period (until April 30, 2020). Global average O3 concentration increased by 86.0% during this same period. Individual country and continent-wise comparisons have been made between lockdown and business-as-usual periods. Universally, NO2 was the pollutant most affected by the COVID-19 pandemic. These effects were likely because its emissions were from sources that were typically restricted (i.e., surface traffic and non-essential industries) by the lockdowns and its short lifetime in the atmosphere. Our results indicate that lockdown measures and resulting reduced emissions reduced exposure to most harmful pollutants and could provide global-scale health benefits. However, the increased O3 may have


ABSTRACT
The global spread of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) has challenged most countries worldwide. It was quickly recognized that reduced activities (lockdowns) during the Coronavirus Disease of 2019 (COVID-19) pandemic produced major changes in air quality. Our objective was to assess the impacts of COVID-19 lockdowns on groundlevel PM2.5, NO2, and O3 concentrations on a global scale. We obtained data from 34 countries, 141 cities, and 458 air monitoring stations on 5 continents (few data from Africa). On a global average basis, a 34.0% reduction in NO2 concentration and a 15.0% reduction in PM2.5 were estimated during the strict lockdown period (until April 30, 2020). Global average O3 concentration increased by 86.0% during this same period. Individual country and continent-wise comparisons have been made between lockdown and business-as-usual periods. Universally, NO2 was the pollutant most affected by the COVID-19 pandemic. These effects were likely because its emissions were from sources that were typically restricted (i.e., surface traffic and non-essential industries) by the lockdowns and its short lifetime in the atmosphere. Our results indicate that lockdown measures and resulting reduced emissions reduced exposure to most harmful pollutants and could provide global-scale health benefits. However, the increased O3 may have

INTRODUCTION
The novel coronavirus was renamed by the International Committee on Taxonomy of Viruses as severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) (Lai et al., 2020;Secon et al., 2020). On January 11, 2020, the first death of Coronavirus Disease of 2019  in China was reported. The first case of COVID-19 outside China was registered in Thailand on January 13, 2020. Since then, COVID-19 has rapidly spread around the globe. On January 20, February 19, and February 20, 2020, the first cases of COVID-19 were observed in the US, Iran, and Italy, respectively (Johns Hopkins Coronavirus Resource Center https://coronavirus.jhu.edu/map.html). The number of confirmed cases has grown exponentially in most affected countries creating a global concern such that on March 11, 2020, WHO declared the outbreak as a pandemic situation (Gupta, 2020).
As the number of confirmed cases continues to rise, not only global health but also the global economy, social networks, global transportation, and air quality are being profoundly affected. Stories in national and international media reported improvements in ambient air quality. They raised the question for the general public and the governments as to whether local and global air quality improvements should be expected. Researchers have begun to report the impact of the COVID-19 pandemic on air quality. To date, published studies have addressed air quality during the COVID-19 outbreak on a country-level scale in multiple countries, including Europe (Andree, 2020;Collivignarelli et al., 2020;Nakada and Urban, 2020;Tobías et al., 2020;Velders et al., 2020;Zoran et al., 2020), Asia (Mahato et al., 2020;Sharma et al., 2020;Wang and Su, 2020;Xu et al., 2020), the Americas (Adhikari and Yin, 2020;Berman and Ebisu, 2020;Nakada and Urban, 2020;Siciliano et al., 2020). Few studies exist in the literature that addressed air quality at the global scale during the pandemic (Bauwens et al., 2020;Lal et al., 2020;Le Quéré et al., 2020), and a clear assessment of its global impact has not yet been provided. Daily global CO2 emissions were reduced by 17% (Le Quéré et al., 2020). Satellite data showed reductions in CO, NO2, and Aerosol Optical Depth (AOD) globally (Bauwens et al., 2020;Lal et al., 2020). However, satellite data reflect the total air column values making it difficult to conclude the health-relevant air pollution near the ground.
Nitrogen dioxide (NO2), a short-lived gas, O3, a major atmospheric oxidant, and PM2.5 are important pollutants with climate, ecosystem, and health impacts. Exposure to these pollutants can induce oxidative stress and systemic inflammation that can reduce the host's immunity against bacterial and viral infections (Ciencewicki and Jaspers, 2007;Carugno et al., 2016). Thus, it is imperative to understand the changes and trends of these pollutants in the atmosphere during the COVID-19 pandemic period.
This study aimed to assess the impact of the COVID-19 pandemic on ambient air quality.

Sampling
Three main criteria air pollutants that are widely monitored across the globe, including PM2.5, O3, and NO2 were analyzed to represent the global air quality. The validated concentration data were extracted from local stations of the air-monitoring organizations and/or their official websites as well as from open sources for the period from January 1, 2015 until April 30, 2020. All data presented in this work are average daily data that were mostly converted from 1-hour or 1-minute resolution depending on the data source. More details about data validation are presented in SI.
Countries were selected based on the availability of the air quality data, the severity of the COVID-19 outbreak, and the level of air pollution. Briefly, five continents, including 34 countries, 141 cities, and 458 air monitoring stations, were investigated. The name of the countries, cities, types of monitoring stations and air pollutants, and chosen calendar period studied in this article are presented in Table S1. The detailed sampling information regarding each city is found in the Methodology section of the Supplementary Materials.

Data Validation
The validation process may vary in different countries. However, typically outliers and suspected data are flagged. Some of these flagged data are evaluated monthly by experts. Evaluations consist of checking the flags and comparing data with other nearby stations. If there is a valid reason to reject outliers or other flagged data, the data are rejected. An additional annual evaluation might be conducted for the whole data. PM data are calibrated against reference measurements and subsequently corrected.

Statistical Analysis
To understand the impact of COVID-19 on the air quality, two main methods, including Pairwise Comparison and Time Series Decomposition were implemented.

Method IA
This section aims to compare the mean monthly values of PM2.5, O3 and NO2 concentrations between 2020 and the respective average of 2017-2019 for all cities (if data were available) from January till the end of April. Firstly, a normality test was carried out for each given month. If it failed, the Signed Rank test was used; otherwise, the paired t-test was carried out. The threshold for a significant change was set to α ≤ 0.05. If data for a given city was missing for the period 2017-2019, then 2018-2019 was used or only 2019 (depending on the availability of the data); otherwise, this method was skipped for the given city. This method minimizes seasonal effects. However, this method might not truly examine the effect of COVID-19 if significant changes in the sources of emissions occur in 2020 compared to previous years.

Method IB
In this method, only 2020 data were used. We first identified the date and measures the government took to prevent the spread of the COVID-19 (Table S2), including the date for observing the first case of COVID-19, national/local recommendation, national/local lockdown, and easing lockdown. Then, the 2020 period was divided into five sub-periods including (before the first case (1), first case until the beginning of the local or national recommendation (whatever comes earlier) (2), local or national recommendation period (3), local or national lockdown period (whatever comes earlier) (4) Eased lockdown (5). ANOVA on Ranks test was carried out among each pair of the time period to understand how the governmental policy could impact the air quality during the COVID-19 spread in the given country. Although the lockdown dates are available for the countries in this study, we do not know the exact date of the recommendation or lockdown action for the given city. Thus, the dates for the local recommendation and local lockdown specified in Table S2 for a given country were considered for all cities of that country in this study.
This method ignores the seasonal variations of the pollutant concentrations. In the northern hemisphere, the reductions in the concentration over time in 2020 could occur due to the higher planetary boundary layer as temperature increases from January to March. The situation is reverse in the southern hemisphere.

Time Series Decomposition
We performed the statistical analyses in R language, and we employed STL function from "seasonal" package for the decomposition. Method of STL, Seasonal and Trend decomposition using Loess, is an adaptable and robust method for the time series decomposition (Cleveland et al., 1990). Time series data usually contain different types of patterns, and in order to capture the underlying patterns of our time series data, we first need to decompose it into several components, each of which represents a different structure in the data (Karavoulias, 2020). Time Series Decomposition is a statistical task in which we deconstruct the data into several parts. Assuming an additive decomposition, we can consider our time series constructed from three components of trend, seasonal and irregular elements. So, we can describe our time series at time t(yt) as: Description of each element in our time series decomposition is as follows: • Tt: The trend component at time t represents the long-term evolution of the time series. This element exists in the data when there is a persistent increase or decrease. • St: The seasonal component at time t describes the seasonal variations of our time series, which happens over a fixed and known period. Our time series consists of yearly seasonality. • It: The irregular component at time t reflects randomness in our data and characterizes the remainder of our time series after removing the trend and seasonality components. To conclude if the governmental preventive measures due to the COVID-19 led to a certain change in the pollutant concentrations, all the three statistical methods have to demonstrate the impact of the COVID-19. Firstly, the average concentration of the given pollutant for a given month in 2020 has to be statistically significantly different than the average concentration of that pollutant in the same month during 2019 and earlier. Secondly, the pollutant concentration in the month with different levels of the concentration compared to the previous years has to be statistically different compared to the concentration before observing the first case of the COVID-19. Finally, the trend analysis has to show a decreasing trend from at least 2019 till 2020.

RESULTS AND DISCUSSION
Figs. S1(a)-S1(d) presents a map showing COVID-19 affected zones starting from January 20, 2020, according to the situation reports of the World Health Organization (WHO) (WHO, 2020). In less than one month since the first case of pneumonia of unknown etiology had been detected in Wuhan City, Hubei Province, China, the novel virus spread to three neighboring countries: Japan (1), South Korea (1), and Thailand (2) ( Fig. S1(a)). Fig. S1(b) shows the global situation about two weeks later on February 1, when the COVID-19 had expanded to the Western Pacific, South-East Asia, United States of America, Canada, Europe, and Eastern Mediterranean region, while the COVID-19 outbreak was still largely centered around Hubei Province. Data available from WHO showed that the main mode of transmission was from symptomatic cases.
As reported on March 1 (Fig. S1(c)), the number of globally confirmed cases increased 7.3 times within a 1-month period. There were several new states including Iran, Iraq, UAE, Saudi Arabia, Pakistan, Afghanistan, Mexico, Brazil, Georgia, Azerbaijan, Ecuador, Ireland, Monaco, Norway, Ireland, Estonia, Lithuania, Belarus, the Netherlands, Belgium, Luxemburg, Romania, Greece, North Macedonia, Austria, Croatia, Switzerland, Algeria, Oman, Egypt, Qatar, Nigeria, and New Zealand reporting rising numbers of cases. The world map in Fig. S1(d) displays the pandemic spread as of April 1, 2020.
Based on the experience from the 2003 SARS-Co-V and MERS-Co-V outbreaks, countries around the globe implemented anti-pandemic measures to retard the spread of the virus. Actions enforceable by law such as closing schools and kindergartens, closing non-essential shops (shops and stores apart from food, doctors, and drug stores), closing non-essential industrial production, closing recreational venues and public parks, curfews, social distancing, stay-at-home orders, and total movement control were taken. Such preventive measures lead to reductions in emissions of certain air pollutants and consequently affect global air quality.  measured in January to April 2017-2019. In Asia, Europe, the Americas, and Africa, PM2.5 and NO2 concentrations decreased from January to April (as the arrows go from upper right to lower left), while urban O3 concentration increased. In Oceania that was influenced by the meteorology of the southern hemisphere, PM2.5 and O3 concentrations decreased from January to April while NO2 concentrations increased. Fig. 1(A) shows that the monthly average NO2 concentrations from January to April 2017-2019 were 8.0, 7.7, 6.1, 5.0 ppb (the Americas), 31.0, 30.1, 29.5, 26.6 ppb (Asia), 20.6, 20.6, 18.8, 17.4 ppb (Europe), and 5.3, 6.0, 6.8, 8.3 ppb (Oceania), respectively. While in 2020, due to the COVID-19 preventive measures and the meteorological conditions, the average NO2 concentration dropped to 10.5, 9.9, 7.9, 7.0 ppb (the Americas), 21.3, 23.1, 18.8, 17.1 ppb (Asia), 16.7, 14.1, 11.5, 8.5 ppb (Europe) and 5.4, 4.8, 5.8, 6.4 ppb (Oceania), respectively. Among the four continents, Asia had the highest NO2 concentration from 2017 to 2020. The lowest NO2 concentration over this period was measured in Oceania.

Global Air Quality
Tables S9-S11 present the percentage change and the statistical analyses of the NO2, O3, and PM2.5 variations between 2020 (January to April) and 2017-2019 (January-April) averages, respectively. The NO2 concentrations were significantly smaller in 2020 in all months compared to the 2017-2019 averages across all four continents (as the arrows of Fig. 1(A) are located in the lower right below the dotted line) except for January in Oceania (Table S9). Among the four continents, Europe experienced the highest NO2 reduction from February to April 2020 compared to the previous years due to the drastic reduction of on-road diesel vehicles and traffic during the lockdown that are major sources of NO2. The highest NO2 reductions during January were observed in Asia. This ranking among the four continents is consistent with the chronology of the COVID-19 outbreak among the different continents. Since COVID-19 was first observed in Asia very early in 2020, the highest reduction of NO2 in January was observed in Asia. The sharpest European declines in NO2 were in February 2020. Although the COVID-19 outbreak evolved in the Americas after Asia and Europe, controls in the Americas were not as effective as Asia and Europe in reducing the outbreak. Oceania already had the lowest NO2 background compared to other continents. Thus, the change in NO2 in 2020 compared to 2017-2019 was the smallest compared to the other continents. Fig. 1 (Table S10). Fig. 1(C) shows that the average PM2.5 concentrations from January to April 2017-2019 were 12.9, 12.8, 11.0, 12.9 µg m -3 (the Americas), 77.5, 58.4, 47.8, 39.0 µg m -3 (Asia), 22.9, 19.8, 16.8, 14.2 µg m -3 (Europe), 40.2, 35.8, 32.6, 27.0 (Africa) and 7.0, 6.3, 6.3, 6.7 µg m -3 (Oceania), respectively. PM2.5 concentrations from January to April 2020 in the Americas, Asia, Europe (except January) and Oceania (except January) dropped to 10.5, 10.1, 10.4, 10.3 µg m -3 (the Americas), 61.9, 51.9, 38.5, 29.7 µg m -3 (Asia), 36.1, 16.2, 13.2, 13.0 µg m -3 (Europe), and 13.9, 5.9, 5.1, 6.2 µg m -3 (Oceania) while it increased to 49.1, 42.3, 39.1, 26.6 µg m -3 in Africa (except in April). Among the five continents, Asia and Africa experienced the highest PM2.5 concentrations, followed by Europe, the Americas, and Oceania in 2017-2020. In Asia, Europe, and Africa, PM2.5 concentrations decreased from January to April, while in the Americas and Oceania, there were fluctuating values. Table S11 shows that in Asia and the Americas, the average PM2.5 concentration in 2020 declined significantly from January to April 2020 compared to the corresponding months during 2017-2019. Similar significant reductions were observed for Europe (except in January and April) and Oceania (except in January and April). Alternatively, PM2.5 concentrations increased in Africa in 2020 compared to the 2017-2019 averages in all months (except April) and in Europe and Oceania for January. No statistically significant changes in PM2.5 occurred in April 2020 compared to April 2017-2019 in Africa, Europe, and Oceania.
The observed changes in Fig. 1 result from changes in emission rates caused by the COVID-19 preventive measures, changes in the global economy led to reduced industrial emissions (excluding power plants and oil/gas and petrochemical industries), goods transport, and a variety of personal activities such as commuting to work and by differences in meteorological conditions such as temperature, solar irradiance, wind speeds, and planetary boundary layer heights. To focus on the impact of COVID-19 on global air quality, Figs. 2 to 4 present the global maps for the changes in NO2, O3, and PM2.5 concentrations only during the strict lockdown period (excluding periods of restriction easing (Table S2)) compared to the average concentrations before the first observed case of COVID-19 (for the northern hemisphere) or compared to the average concentration during the same time period during 2017-2020 (for the southern hemisphere). Globally, the NO2 concentration decreased due to the lockdown except in some countries in Western Europe (the Netherlands and Germany), Istanbul, Nur-Sultan, Osaka, and Sydney, where no significant changes in NO2 concentrations occurred. On average, a global 34.0% reduction in NO2 concentration during the strict lockdown (until April 30, 2020) period was estimated. Major sources of NO2 include transport, including vehicular emissions, especially diesel vehicles, aviation, and ships that were substantially reduced during the lockdown resulting in reduced NO2 concentrations. Fig. 3 shows that the global average O3 concentration increased by 86.0% during the strict lockdown period. The O3 concentration increase was expected for three reasons. First, during this period, there were increased temperatures and solar irradiance in the northern hemisphere. Second, the global reduction in nitrogen oxide (NO) emissions reduced the titration of O3 and the formation of NO2. Third, the VOC-limited O3 formation regime (larger NOx concentration (the sum of NO and NO2) compared to VOCs) can yield higher O3 formation when NOx was reduced. However, outside of cities and especially in suburban areas where urban air pollution plumes often affect air quality, the reduction of the NOx and VOCs emissions in the urban core might result in O3 decreases since they are often NOx-limited regimes. The impact of the global pandemic on the O3 concentrations in rural regions is uncertain and requires investigation.
Based on the five years of the data collected in this study (refer to the figures presented in the Supplementary Information), most of the cities included in these analyses were VOC-limited. Thus, increased O3 concentrations were expected as a result of decreased NOx. However, some strongly NOx-limited cities with biogenic or industrial VOC (oil refinery) sources showed no changes in O3 or even decreased O3 concentrations, such as Ahmedabad, India, and Abu Dhabi, United Arab Emirates. Fig. 4 shows that PM2.5 concentrations were significantly reduced during the lockdown in Asia and South America while it did not change in New York State (NYS), Mexico City, New Zealand, and much of Europe. Increased PM2.5 concentrations were observed in northern France, central Germany, London, and the United Arab Emirates (UAE). Globally, a 15.0% reduction in the Fig. 2. The global map of NO2 changes during lockdown period in 2020-The percentage values show the changes in the average NO2 concentration of the given pollutant during the lockdown period in 2020 compared the average concentration before observing the first case of COVID-19 in 2020 (for the northern hemisphere) or compared to the average NO2 concentration during the same time period during 2017-2020 (for the southern hemisphere). average PM2.5 concentration was observed. The increased or constant PM2.5 during the lockdown preventive measures might be attributed to three major causes. First, increased O3 might produce increased secondary organic aerosol (SOA) formation through reactions with olefins (Tobías et al., 2020). Second, new particle formation produces particles more effectively during the lockdown resulting in increased particulate surfaces that enhances heterogeneous chemistry and particle mass (Yao et al., 2018). This mechanism might explain why reductions in transportation and industrial emissions during the lockdown in China did not eliminate severe air pollution episodes  and haze formation .
Third, in several regions, the lockdown forced people to stay at home in areas where biomass or coal-burning stoves are used for heating and cooking, leading to increased emissions (Amouei Torkmahalleh et al., 2017;Thakur et al., 2020). In many cities, traffic accounts for only 20-35% (Thakur et al., 2020) of the ambient PM2.5. If the main emissions reductions were caused by major traffic decreases, the PM2.5 reductions would be much less than in NO2. Three other seasonal causes could increase PM2.5 during early to mid-spring compared to the late winter that occur annually, distinct from the 2020 lockdown measures. First, in urban areas, around 70% of PM2.5 tends to be secondary in origin (Amato et al., 2015;Hopke et al., 2020). Complex heterogeneous chemistry and meteorological conditions in China might have favored enhanced formation of PM2.5 as reported by  during the COVID19 lockdown. However, increased residential coal and biomass combustion also with spring festival emissions are also likely causes . Second, this study assumes similar meteorological conditions between 2020 and the previous years. However, in some cities, for example, in the Netherlands, the wind direction might have favored increased PM2.5, NO2, and O3 concentrations during the 2020 lockdown compared to prior years (Velders et al., 2020). Third, long-range continental transport of atmospheric dust from North Africa to Central Europe and the Middle East in spring (Moulin et al., 1997;Varga, 2020) and inner continental long-range transport of wildfire particles in spring, typically from Russia and Belarus to Finland, could result in increased PM2.5 concentrations (Niemi et al., 2009). Our analyses showed that NO2 emissions were substantially reduced due to measures taken to prevent COVID-19 and the spreading of the SARS-CoV-2. This reduction also seem to have increased the concentrations of secondary pollutants such as O3, as well as the formation of SOA reflected as a changed composition of PM2.5. Since NO2 seems to be the pollutant highly impacted by the traffic emissions in many cities and because traffic was markedly reduced during the lockdown, NO2 experienced the largest decreases in ambient concentrations.

CONCLUSION
The unprecedented reduction in global economic activity caused by the COVID-19 related restrictive measures led to noticeable changes in emissions and consequently in ground-level air pollution. Changes in NO2 and O3 surface concentrations are more pronounced and were retained their sign (reduced NO2 and increased O3) globally. Long-range transport of particles, the presence ORIGINAL RESEARCH Special Issue on COVID-19 Aerosol Drivers, Impacts and Mitigation (XII) https://doi.org/10. 4209/aaqr.200567 of natural particulate matter, increased residential emissions, and modified particle formation processes during this period of the generally reduced emissions resulted in compositional changes in PM2.5 as well as an overall decrease in concentrations. However, these reductions came with significant economic (Cutler and Summers, 2020) and other costs (Chandan et al., 2020;Mortazavi et al., 2020;Rosenbaum, 2020). Thus, the air quality improvements would need to be obtained without the extensive other consequences resulting from the imposed reductions in activities.

ACKNOWLEDGMENTS
The authors of this study truly appreciate the funding provided by Nazarbayev University through the Collaborative Research Grant (grant number: 091019CRP2104). MAT, MN, EA, and NE are the members of the Chemical and Aerosol Research Team (CART) and also the Environment and Resource Efficiency Cluster (EREC) at Nazarbayev University, and they would like to acknowledge the resources provided by CART and EREC to conduct this research. For India, except for Delhi, daily and hourly averaged air quality data were obtained from the Central Pollution Control Board (CPCB) CCR, OpenAQ or Air quality historical data platform, and location of stations were obtained from the OpenAQ website. CPCB is gratefully acknowledged for making the data available. Further, the CPCB data were consolidated and curated by Mr Prem Maheshwarkar, Earth and Environmental Sciences, IISER Bhopal whose contribution is also acknowledged. In Chile, the contributions made by Mr. Roberto Martinez, Head of the Planning and the Standards, Department of the Air Quality and Climate Change Division, Ministry for the Environment to provide data to this study, are greatly appreciated. In Brazil, CETESB (Environmental Company of the State of Sao Paulo) is acknowledged for its contribution to data preparation. In Cyprus, the authors would like to thank the Department of Labor Inspection of the Ministry of Environment of the Republic of Cyprus for providing the observational data for their monitoring network. For New Zealand, authors acknowledge the Environment Canterbury as the source for the Christchurch data. For Spain, Barcelona, the authors would like to thank "Department of Environmental Quality, Generalitat de Catalunya" as the data source. Turkish Ministry of Environment and Urbanization and Turkish State of Meteorological Services provided air quality and meteorological data, respectively, that are greatly appreciated. In the Netherlands, data were obtained from the National Institute for Public Health and the Environment (RIVM) and contained additional data from the GGD Amsterdam (provided by Dave de Jonge) and the DCMR Environmental Protection Agency (provided by Ed van der Gaag) and National Institute for Public Health and the Environment (RIVM) that are greatly appreciated. The authors would like to thank Tehran Air Quality Control Company for providing the data for the city of Tehran.