Spread of COVID-19, Meteorological Conditions and Air Quality in the City of Buenos Aires, Argentina: Two Facets Observed during Its Pandemic Lockdown

: This work studied the spread of COVID-19, the meteorological conditions and the air quality in a megacity from two viewpoints: (1) the correlation between meteorological and air quality (PM 10 and NO 2 ) variables with infections and deaths due COVID-19, and (2) the improvement in air quality. Both analyses were performed for the pandemic lockdown due to COVID-19 in the City of Buenos Aires (CABA), the capital and the largest city in Argentina. Daily data from temperature, rainfall, average relative humidity, wind speed, PM 10 , NO 2 , new cases and deaths due COVID-19 were analyzed. Our ﬁndings showed a signiﬁcant correlation of meteorological and air quality variables with COVID-19 cases. The highest temperature correlation occurred before the conﬁrmation day of new cases. PM 10 presented the highest correlation within 13 to 15 days lag, while NO 2 within 3 to 6 days lag. Also, reductions in PM 10 and NO 2 were observed. This study shows that exposure to air pollution was signiﬁcantly correlated with an increased risk of becoming infected and dying due to COVID-19. Thus, these results show that the NO 2 and PM 10 levels in CABA can serve as one of the indicators to assess vulnerability to COVID-19. In addition, decision-makers can use this information to adopt strategies to restrict human mobility during the COVID-19 pandemic and future outbreaks of similar diseases in CABA. shows that exposure to air pollution was signiﬁcantly correlated with an increased risk of becoming infected and dying from COVID-19. Therefore, this study shows that air pollution in CABA can serve as one of the indicators to assess vulnerability to COVID-19. In addition, it can serve decision-makers for the adoption of strategies to restrict human mobility during the COVID-19 pandemic and future outbreaks of similar diseases.


Introduction
The Coronavirus disease 2019 (COVID-19) is identified as an infectious disease caused by severe acute respiratory syndrome novel coronavirus 2 (SARS-CoV-2) [1]. November 2019 was the date of the world's first case of coronavirus . Patient zero could be a person living in Hubei-Wuhan (China). On December 2019, China alerted the World Health Organization (WHO) of several cases of unusual pneumonia in Wuhan, therefore officially identifying the cause of the COVID-19 outbreak

Study Area
CABA is the capital and the largest city of Argentina. As shown in Figure 1, the city is located on the western shore of the estuary La Plata River, on the South American continent's southeastern coast (34 • 36 S, 58 • 22 W). The area of CABA is 203 km 2 and its population in 2020 based on projections of results of the 2010 Population Census is 3,075,646 inhabitants, with a population growth rate of 9.58% per year [27].

Study Area
CABA is the capital and the largest city of Argentina. As shown in Figure 1, the city is located on the western shore of the estuary La Plata River, on the South American continent's southeastern coast (34°36′ S, 58°22′ W). The area of CABA is 203 km 2 and its population in 2020 based on projections of results of the 2010 Population Census is 3,075,646 inhabitants, with a population growth rate of 9.58% per year [27].

Data Collection
The data set used in this investigation was from 5 March to 31 May 2020 ( Figure 2). It was obtained from the Ministry of Health for new cases (daily), total cases (accumulated), and mortality (daily) due to COVID-19 [21]. Daily meteorological data were obtained from Argentina's National Weather Service [28]. The data consist of minimum temperature ( • C), maximum temperature ( • C), average temperature ( • C), humidity (%), and accumulated Rainfall (mm).

Data Collection
The data set used in this investigation was from 5 March to 31 May 2020 ( Figure 2). It was obtained from the Ministry of Health for new cases (daily), total cases (accumulated), and mortality (daily) due to COVID-19 [21]. Daily meteorological data were obtained from Argentina's National Weather Service [28]. The data consist of minimum temperature (°C), maximum temperature (°C), average temperature (°C), humidity (%), and accumulated Rainfall (mm). The Spearman rank correlation tests were used to examine the correlation between variables, typically used for a non-normal distribution dataset, as shown in other studies [15,16,18,20]. Nonnormal distribution was verified previously by the Shapiro-Wilk normality test application as shown in Table A1 (Appendix A). The correlations were done for new cases, total cases and mortality due to COVID-19, with meteorological and air quality variables, using lag up to 15 days over CABA.
Two commonly reported pollutants (PM10 and NO2) were obtained from the air quality network of this city. This data was obtained from hourly records from its three representative stations (see The Spearman rank correlation tests were used to examine the correlation between variables, typically used for a non-normal distribution dataset, as shown in other studies [15,16,18,20]. Non-normal distribution was verified previously by the Shapiro-Wilk normality test application as shown in Table A1. The correlations were done for new cases, total cases and mortality due to COVID-19, with meteorological and air quality variables, using lag up to 15 days over CABA. Two commonly reported pollutants (PM 10 and NO 2 ) were obtained from the air quality network of this city. This data was obtained from hourly records from its three representative stations (see Figure 1) [29]. Then, we estimated the daily mean values of the recorded data, considering those data with a temporal representation greater than 75% of the time during the study period. Also, PM 10 and NO 2 pollutants data measured by the air quality network of this city were compared with data measured by the S5p/TROPOMI-ESA [30,31], both at the same time in 2019 and 2020.

Satellite Data Processing and Analysis
The European Space Agency (ESA) Sentinel-5 Precursor (S5p) is a low Earth orbit polar satellite which provides information and services on air quality, climate, and the ozone layer. The payload of the mission is the TROPOspheric Monitoring Instrument (TROPOMI) that measures key atmospheric constituents including ozone, NO 2 , SO 2 , CO, CH 4 , CH 2 O and aerosol properties [30]. The level 2 product of NO 2 tropospheric column gives the total atmospheric NO 2 column between the surface and the tropopause with a spatial resolution of 3.5 km × 7 km. The quality of the product observations depends mainly on cloud cover, surface albedo, and presence of snow, among other factors. A quality assurance variable (qa_value) ranges from 0 to 1. The recommended qa_value = 0.75 removes cloud-cover scenes, partially snow/ice-covered scenes, errors, and problematic retrievals [32]. The data provided by the satellite is given with an orthogonal scanline to the flight direction of circa 2600 km. Each observational data has a temporal dimension referenced to the orbit start time and spatial dimension in scanline and flight direction, which are georeferenced with latitude and longitude coordinates. Then, in order to represent the satellite data for the periods and study area in CABA, the TROPOMI NO 2 level 2 product data (molecules/cm 2 ) were remapped in a cylindrical projection with a resolution of 0.05 • × 0.05 • . Each observation remapped on the same output pixel is averaged, obtaining a single georeferenced output for the whole dataset. The air quality network of CABA does not measure PM 2.5 . Therefore, PM 2.5 levels have been retrieved from satellite images obtained on 8 March, 20 March and 19 April (2018, 2019 and 2020), using Copernicus' Earth online viewer [33], as shown in Figure A1.

Results and Discussions
CABA is the head political, financial, tourist, and cultural metropolis of Argentina. It is also the most densely populated Argentine city, with 14,450.8 pop/km 2 [34]. Figure 2 shows the daily and accumulated evolution of transmission cases and deaths by COVID-19 in this city, which deserves special attention due to its relevance, as well as its high population density. Previous studies indicated the importance of the analysis of meteorological conditions in the spread of COVID-19 in highly densely populated areas [15,35]. On 5 March, the first national case of COVID-19 was confirmed in CABA [21]. In a few days, as cases rapidly multiplied, the Argentine national government established a lockdown on 20 March to minimize and control the spread of this pandemic disease [22]. Nevertheless, infections and deaths continued to grow rapidly ( Figure 2). Because of the lockdown, observations indicated that NO 2 concentration levels over the city decreased [30,31] (Figure 3), as well as PM 2.5 levels, as is shown in . This circumstance is analyzed in the following subsections.  Figure 4A shows a decrease in maximum, minimum, and average temperature since the months analyzed correspond to the austral fall season. The rainfall had variations between 10 and 40 mm/day, while relative humidity presented an average of 74.5% ( Figure 4C). The average daily wind speed ( Figure 4B) presented peaks above 8 m/s in March, decreasing in April to 7 m/s and 4.4 m/s in May, 2020. This situation is due to the weakening of the winds in fall and winter [36][37][38]. Figure 5 presents the statistical correlation coefficient for meteorological variables with new cases, total cases and mortality. New cases ( Figure 5A) had a higher negative correlation at 8 days lag (r = −0.74, p < 0.01), 7 days lag (r = −0.74, p < 0.01) and 15 days lag (r = −0.68, p < 0.01) for average temperature, minimum temperature and maximum temperature, respectively. Humidity and rainfall with, new cases showed a higher positive correlation at 10 days lag (r = 0.19, p < 0.05) and 0 day (r = −0.35, p < 0.01), respectively. Wind speed had a higher negative correlation at 2 days lag (r = −0.33, p < 0.01). Total cases ( Figure 5B) showed a higher negative correlation at 4 days lag (r = −0.82 and r = −0.72, p < 0.01) for average and minimum temperature, respectively, while this occurred with 15 days lag (r = −0.70, p < 0.01) for maximum temperature.  Figure 4A shows a decrease in maximum, minimum, and average temperature since the months analyzed correspond to the austral fall season. The rainfall had variations between 10 and 40 mm/day, while relative humidity presented an average of 74.5% ( Figure 4C). The average daily wind speed ( Figure 4B) presented peaks above 8 m/s in March, decreasing in April to 7 m/s and 4.4 m/s in May, 2020. This situation is due to the weakening of the winds in fall and winter [36][37][38]. Figure 5 presents the statistical correlation coefficient for meteorological variables with new cases, total cases and mortality. New cases ( Figure 5A) had a higher negative correlation at 8 days lag (r = −0.74, p < 0.01), 7 days lag (r = −0.74, p < 0.01) and 15 days lag (r = −0.68, p < 0.01) for average temperature, minimum temperature and maximum temperature, respectively. Humidity and rainfall with, new cases showed a higher positive correlation at 10 days lag (r = 0.19, p < 0.05) and 0 day (r = −0.35, p < 0.01), respectively. Wind speed had a higher negative correlation at 2 days lag (r = −0.33, p < 0.01). Total cases ( Figure 5B) showed a higher negative correlation at 4 days lag (r = −0.82 and r = −0.72, p < 0.01) for average and minimum temperature, respectively, while this occurred with 15 days lag (r = −0.70, p < 0.01) for maximum temperature.   Humidity showed a higher positive correlation with total cases at 15 days lag (r = 0.18, p < 0.05), and rainfall showed significant correlations at day 0 (r = −0.32, p < 0.01). Also, wind speed had a higher negative correlation at 2 days lag (r = −0.27, p < 0.01) in total cases. Mortality ( Figure 5C) presented higher negative correlation at 7 days lag (r = −0.65, p < 0.01) for average temperature, at 2 days lag (r = −0.68, p < 0.01) for minimum temperature and 1 day lag for maximum temperature (r = −0.59, p < 0.01). Humidity and mortality cases did not show a significant correlation, but rainfall showed a high negative correlation at 2 days lag (r = −0.33, p < 0.01). Also, wind speed displayed a high negative correlation at 2 days lag (r = −0.32, p < 0.01). Few studies have investigated the relationship of SARS-CoV-2 virus infections with meteorological variables using days lag to consider their incubation time. Humidity showed a higher positive correlation with total cases at 15 days lag (r = 0.18, p < 0.05), and rainfall showed significant correlations at day 0 (r = −0.32, p < 0.01). Also, wind speed had a higher negative correlation at 2 days lag (r = −0.27, p < 0.01) in total cases. Mortality ( Figure 5C) presented higher negative correlation at 7 days lag (r = −0.65, p < 0.01) for average temperature, at 2 days lag (r = −0.68, p < 0.01) for minimum temperature and 1 day lag for maximum temperature (r = −0.59, p < 0.01). Humidity and mortality cases did not show a significant correlation, but rainfall showed a high negative correlation at 2 days lag (r = −0.33, p < 0.01). Also, wind speed displayed a high negative correlation at 2 days lag (r = −0.32, p < 0.01). Few studies have investigated the relationship of SARS-CoV-2 virus infections with meteorological variables using days lag to consider their incubation time. A recent study investigated the effects of temperature and humidity on new daily cases and new COVID-19 deaths in 166 countries, using day lag up to 3 days, and showed negative correlations with temperature and humidity [39]. These results are consistent with what has been found in this study for temperatures but not for humidity, where positive correlations were found. Another study conducted in Hong Kong showed humidity and rainfall had a positive correlation with virus diseases [40].

Correlation SARS-CoV-2 Infections with Meteorological and Air Quality Variables
Air quality as PM 10 and NO 2 variables (shown in Figure 6) demonstrated a significant correlation with new cases, total cases, and mortality. Correlation with PM 10 was noted the highest correlation in new cases (1 day lag, r = 0.20, p < 0.1), total cases (15 days lag, r = −0.25, p < 0.05), and mortality (15 days lag, r = −0.25, p < 0.05). However, PM 10 correlation with total cases also showed a positive relationship (0 day, r = 0.18, p < 0.1). NO 2 did not show significant correlation with new and total cases (p < 0.1), but it showed a negative correlation with mortality (4 days lag, r = −0.29, p < 0.01). Recent studies conducted in New York City showed that average air quality is significant for new cases, total cases, and mortality by COVID-19 [16]. Additionally, Table 1 displays a comparison of our data with previous studies that demonstrate a relationship between COVID-19 infection and deaths with air pollutants. These results agree with the findings presented here. Overall, these findings are consistent with other research showing that wind speed, humidity, temperature and air quality have a significant correlation with the transmission of infectious diseases and their associated deaths [16,41,42]. In that direction, results showed significant correlations (p < 0.01) using 7 to 15 days lag for meteorological conditions, 12 to 15 days lag for PM 10 Figure 7, the greatest reduction in the concentration of PM 10 and NO 2 in 2020 occurred from the start of the lockdown (March 20) to April 20. Then, concentrations showed behavior similar to 2019, possibly related to the easing of the lockdown, made to minimize the negative impact on the economy [22]. Arkouli et al. [36] showed that CABA in the cold season presents the lowest values of the ventilation coefficient. Therefore, it indicated higher probabilities of poor air quality, and that was confirmed by the higher concentrations of PM 10 and PM 2.5 measured by them. Other studies showed that 45.5% of NOx emissions in this city were from mobile sources, such as cars, taxis, buses and trucks [38].

Impacts on the Air Quality
Atmosphere 2020, 11, x FOR PEER REVIEW 11 of 23  Figure 7, the greatest reduction in the concentration of PM10 and NO2 in 2020 occurred from the start of the lockdown (March 20) to April 20. Then, concentrations showed behavior similar to 2019, possibly related to the easing of the lockdown, made to minimize the negative impact on the economy [22]. Arkouli et at. [36] showed that CABA in the cold season presents the lowest values of the ventilation coefficient. Therefore, it indicated higher probabilities of poor air quality, and that was confirmed by the higher concentrations of PM10 and PM2.5 measured by them. Other studies showed that 45.5% of NOx emissions in this city were from mobile sources, such as cars, taxis, buses and trucks [38].  During the COVID-19 pandemic lockdown, several studies around the world have studied its impact on air quality. Ghahremanloo et al. [51] reported NO 2 reductions of up to 83% in East Asia, while in Southeast Asia, region reductions were observed in PM 10 (26-31%), PM 2.5 (23-32%), NO 2 (63-64%), SO 2 (9-20%), and CO (25-31%), in urban areas from Malaysia [52]. In India, AOD reductions of up to 50% were perceived in New Delhi [5]. Otmani et al. [53] found reductions of 75%, 49% and 96% for PM 10 , SO 2 and NO 2 in Salé City (Morocco). Also, emissions of NO 2 have been reduced up to 40% in Iraq compared to the pre-lockdown [54]. The United Kingdom showed average reductions of NO, NO 2 and NOx between 32% and 50% at roadsides on lockdown [55]. Changes in the United States' air pollution showed declining NO 2 of 25.5% compared to historical years [56]. Additionally, Muhammad et al. [10] using satellite data showed NO 2 reductions of 20-30% in Italy, France, and Spain. Thus, in line with our findings, this pollutant reduction is related to the lockdown aimed to stop the spread of the SARS-CoV-2 virus. Especially, after 20 March (start of lockdown), a decrease was observed in the average daily concentrations measured by the air quality network of CABA, as shown with a red line in Figure 7. Tropospheric NO 2 measured by the S5p/TROPOMI-ESA (Figure 8

The Two Facets Observed on COVID-19 Pandemic Lockdown
Pandemic lockdown is a critical time due to infected people with SARS-CoV-2, increasing cases of death, and economic damage [7]. However, it has generated a window to analyze the correlation between rapidly spreading viral diseases like COVID-19 with weather and air quality indicators. Our findings show correlations between meteorological and air quality variables several days before positive identification or death (shown in Figures 5 and 6). Thus, this study allows us to expand knowledge about the meteorological and air quality variables analyzed that could be used for decision-makers to consider designing measures to reduce the risk of COVID-19 and death, and also to better understand how the analyzed variables can vary in the different climate change scenarios and therefore in the spread of viral diseases [57,58].
Additionally, Table 2 shows reductions in mean NO 2 concentrations, according to the data from the air quality network of CABA and the S5p/TROPOMI-ESA satellite [30,31]. A decreasing trend in the data measured was observed from the S5p/TROPOMI-ESA satellite too. There was only an increase during the S1 and S2 situations (see S1/S2 definition in Table 2) as shown in , but it was considered as a regional effect also observed in other recent studies carried out in South America [3,9]. Situations S2, S3, and S4 provide insight into some measures that could be taken to control polluting emissions in the city. Studies conducted by Vrekoussisy et al. [59] in Athens, Greece, showed that the NO 2 columns over Athens have been significantly reduced in the range of 30-40% during the economic crisis of 2008. They also found strong correlations between pollutant concentrations and economic indicators, showing that the economic recession resulted in proportionally lower levels of pollutants in large parts of Greece. In addition, Tong et al. [60] studied the implications for surface ozone levels to changes in NOx emissions in the United States during the 2008 global recession and they observed large national reductions in NO x emissions of up to 21%. Recent studies investigated 419 episodes of financial crises in more than 150 countries during the period 1970-2014, analyzing the impact of financial crises on air pollutant emissions. The results showed that, in the short term, as a consequence of the financial crises, emissions decrease for CO 2 , SO 2 and NO x by 2.6, 1.8, and 1.7%, respectively, but, in the medium-term, financial crises have an insignificant effect on emissions, or in some cases lead to a 1-2% increase, cancelling out the initial benefit [61]. These studies showed that the decreases in pollutants are related to a financial crisis; similarly, in our study, it could be generated by the economic crisis of the COVID-19 pandemic. However, the reactivation after this pandemic and financial crisis should serve to establish more environmentally friendly measures, trying to transform the temporary reductions into permanent lower levels.
Our study shows that exposure to air pollution is significantly correlated with an increased risk of infection and death due to COVID-19 (as shown in Table 1). In addition, human mobility restriction measures provide the greatest benefit for COVID-19 mitigation [62,63], because prevention is actually more cost-effective than cure [64][65][66] or death [67]. Therefore, the results of this study show that air pollution in CABA can serve as one of the indicators to assess vulnerability to COVID-19. Moreover, strategies to restrict human mobility during the COVID-19 pandemic and future outbreaks of similar diseases seem to be adequate.
This study has delivered strong evidence regarding the association of COVID-19 expansion with various meteorological and pollutants indicators, and improvement in air quality in CABA. However, it has some limitations. Firstly, COVID-19 is an infectious disease that is related to additional variables that must be considered in a comprehensive study. Also, future research should include air pollutants such as PM 2.5 , SO 2 , black carbon and SARS-CoV-2 in PM 2.5 particles. Moreover, socio-economic aspects such as measures of social distancing, full and partial shutdown, personal hygiene, among others, should be explored.

Conclusions
Meteorological and air quality variables were important factors in determining the incidence rate of COVID-19 in CABA. We also observed that air quality in terms of NO 2 and PM 10 improved during the most restrictive time of the COVID-19 pandemic lockdown in CABA. Our study shows that exposure to air pollution was significantly correlated with an increased risk of becoming infected and dying from COVID-19. Therefore, this study shows that air pollution in CABA can serve as one of the indicators to assess vulnerability to COVID-19. In addition, it can serve decision-makers for the adoption of strategies to restrict human mobility during the COVID-19 pandemic and future outbreaks of similar diseases.

Abbreviations
The following abbreviations are used in this manuscript: