key: cord-0893471-2yppf4c5
authors: Peralta, Oscar; Ortínez-Alvarez, Abraham; Torres-Jardón, Ricardo; Lastra, Manuel Suárez; Castro, Telma; Ruíz-Suárez, Luis Gerardo
title: Ozone over Mexico City during the COVID-19 pandemic
date: 2020-10-22
journal: Sci Total Environ
DOI: 10.1016/j.scitotenv.2020.143183
sha: e848cbba9d1233661f528e275b0e08ad2b7780cd
doc_id: 893471
cord_uid: 2yppf4c5

During the COVID-19 pandemic lockdown, emissions of primary criteria pollutants in the Mexico City Metropolitan Area (MCMA) were substantially reduced, as in many other cities in the world. Unexpectedly, the daily average ozone concentration profile was practically indistinguishable from that of the past two years for the same time span in the calendar. So, we compared surface meteorology data, CO, NOx and O3 hourly concentrations in the MCMA from the ozone season (from March 1 to May 31) for the years 2018, 2019, and 2020. Also, TROPOMI satellite data on column count of CO, NO2 and HCHO, above a sparse grid of surface points in the MCMA, were also compared for March, April, and May 2020 with those from 2019. Population density used as a background variable to increase understanding of the observed differences allowed us to propose that reductions in NOx were so drastic that ozone formation moved rapidly from a VOC sensitive region towards a NOx sensitive region. The relevance of that unplanned policy provides impacts of contingent short-term emissions control actions during very high pollution episodes. Further analysis of these and other related data concerning VOC speciation and emissions patterns during the coronavirus lockdown may provide guidelines to enhance emission control policies in the post-COVID-19 times to come.

Air pollution makes people more vulnerable to respiratory infections. Around 95% of the planet's population is breathing unhealthy air, according to the annual State of Global Air Report (Health Effects Institute, 2020) , and Mexico City is not the exception. It has a long history of air quality management challenges. For instance, the Mexican ozone standards were breached in spring 2016 in almost all monitoring stations, both the 1-hour and 8-hour limits (0.095 ppm and 0.070 ppm, respectively) were surpassed. The ozone 1-hour average exceeded the threshold and triggered the local Atmospheric Environmental Contingency Plan (AECP) on several occasions, leading to the activation of the ACEP which had not been activated in more than a decade (INECC, 2016) . The AECP mandates restrictions on mobility and on some business and industrial activities. A year later, in spring 2017, a week of ozone contingencies, from May 15 -21, and May 22 -24, forced the authorities to reduce mobility in the city. In spring 2018, there were only two contingencies, but the next year officially established (SEDEMA, 2020) . However, ozone levels did not show an appreciable decrease. It is not possible to know if VOC levels also decreased, due to the fact that measurements of those ozone precursors are not common, since they are not criteria air pollutants.

Ozone sensitivity in the MCMA region has been recognized as VOC-sensitive, and several authors have explained the lowering in the rate of titration of O 3 by NO and the decrease of NOx in a VOC-limited environment (Dantas et al., 2020; Sicard et al., 2020; Tobías et al, 2020; Gaffney and Marley, 2020) . So, a deeper analysis of ozone chemistry sensitivity during the lockdown, based on the principles of the photochemical indicators approach could be used to explain this effect (Sillman and West, 2009; Torres-Jardón et al, 2009; Ying et al, 2009 ).

The SARS-COV-2 coronavirus appeared in China in December 2019, causing an epidemic in Wuhan. It then spread to all continents and soon after was declared a global pandemic by the World Health Organization (WHO). Mexican authorities suspended school classes in March 23, following the WHO recommendations. Later, on March 30, they declared a health emergency and issued stricter rules aimed at containing the spread of COVID-19 after the number of cases increased. Consequently, vehicular activity decreased, and many people stayed at home. However, people on low incomes and the myriad of informal workers followed those directives with an uneven degree of compliance. As of May 7, the authorities reported 2,961 deaths, 29,616 vehicles surely decreased, whereas those related to domestic energy use may have increased. Also, the lockdown happened during the Mexico City's dry season, when forest fires and agricultural burning in the central region of Mexico contributed several ozone precursors.

The use of photochemical models could be of great help in understanding the environmental situation, but one limitation of attempting to explain the ozone formation through modeling is that the exact percentage of potential reductions in both NOx and VOC during the lockdown is unknown. So, satellite-retrieved formaldehyde (HCHO) and nitrogen dioxide (NO 2 ) column densities have been employed as indicators to investigate spatial and temporal changes in emissions of VOC and NOx around the world as a consequence of the impact of the SARS-CoV-2 pandemic (Muhammad et al, 2020) . Besides the typical primary anthropogenic and biogenic HCHO emissions, including those from sporadic biomass burning, the atmospheric oxidation of several volatile organic compounds produces secondary formaldehyde that can be observed spatially and temporally with the Visible Infrared Imaging Radiometer Suite (VIIRS) on board the Sumo-NPP satellite (Wang et al, 2020; Kajino et al, 2019) . Thus, the column density of HCHO can be used as a surrogate for the emissions of regional VOC, and the NO 2 column measurements represent both primary NO 2 and oxidation of NO emissions, including NO 2 from the decomposition of PAN transported to the study area. Although natural emissions of NOx may contribute to the burden of these pollutants, anthropogenic emissions surpass those occurring naturally.

Hourly average ozone, NOx, NO 2 , and CO data were obtained from the RAMA monitoring stations from the period of March 1 -May 31, 2018, 2019, and 2020.

Information on measurement principles is available from the RAMA website (RAMA, 2020). All monitoring equipment fulfills the requirements of the US EPA related to network monitoring methods. In summary, ozone is measured by the UV photometric method, NOx by the gas phase chemiluminescence method, and CO by non-dispersive infrared photometry (NDIR).

In order to identify whether or not there were changes in emissions of NOx between the two previous years and the year 2020, as well as to observe whether those changes influenced ozone levels, the monitoring data were statistically processed considering five representative monitoring stations located in the northeast (TLA), northwest (SAG), center (MER), southeast (CCA) and southwest (UAX) sectors of the urban area (Figure 1) , as well as a whole considering all the measurement sites. In particular, we focused on both the peaks of CO and NOx during the period 06:00 to 09:00 as indicators of changes in mobile emissions of these precursors group and the afternoon ozone peaks as surrogates for the extent of the photochemical production in MCMA.

On the other hand, in absence of VOC monitoring data during COVID-19, we economic activities where combustion is involved. All combustion processes release NOx and a fraction of incomplete burned fuel as CO and VOC. Thus, it could be expected that a change in CO could be related to a proportional change in VOC emissions. The 06:00 to 09:00 monitoring data period corresponds to the peak in urban traffic; besides that, it is the time period when photochemical reactions and the surface mixing layer are still restricted (Fujita et al., 1992) . The ambient measurements in the 06:00 to 09:00 period reflect reasonably well on-road vehicular emissions because of the higher contribution of this source category to total emissions. Also, emissions are more uniformly distributed across the urban area.

According to the most recent emissions inventory for MCMA, vehicular emissions contribute around 89% of CO, 82% of NOx, and 18% of VOC to the total burden (SEDEMA, 2017). However, area sources, mainly evaporative from storage and racking of liquid fuels, leaks of L.P. gas tanks, and use of solvents and other industrial sources, account for nearly 65% of total VOC emissions (SEDEMA, 2017).

We did a t-test (two-tail test assuming equal or unequal variances, depending on the F-test of two variances and a p-value of 0.05) to test the null hypothesis that the means of the medians of two populations of selected indicators obtained from all the monitoring stations in MCMA for 2020 with respect to the 2018 and 2019 years were not different. The alternative hypothesis was that the means of indicators for the two populations under test were different. The indicators tested were hourly maximum ozone and NOx and the medians of the 06:00 to 09:00. CO and NOx averages, respectively.

Ozone formation is driven by complex nonlinear photochemistry processes and is J o u r n a l P r e -p r o o f mainly controlled by VOC and NOx emissions (Sillman and He, 2002; Duncan et al 2014; Souri et al 2020) . Actions to reduce NOx emission of the main sources will decrease ground-level ozone concentrations in NOx-limited areas but increase ozone concentrations in VOC-limited areas. On the other hand, control strategies to reduce VOC emissions will decrease ozone concentrations in VOC-limited regimens but increase its formation and concentration in NOx-limited areas (Chang et al, 2016) .

In order to explain if the processes associated with the maximum average ozone concentrations observed during the three years have changed, an evaluation of the correlation between O 3 and an assumed equivalent NOy was performed based on the concept of photochemical indicators (PI). According to Torres-Jardón et al (2009) Given that all combustion processes release NOx and a fraction of incomplete burned fuel as CO and VOC, we assumed that in the absence of VOC time series of monitoring data, including during COVID-19, the CO-NOx correlation may indicate the direction change of the VOC/NOx ratio due to a reduction of transport and other economic activities where combustion is involved.

TROPOMI is onboard the Sentinel-5 Precursor (S5P) satellite that orbits along the nearpolar sun-synchronous orbit at 824 km above the ground, with a 17 days repeat cycle and equator-crossing time of 13:30 local solar time (LST) on the ascending node. A scanning swath of TROPOMI covers 2,600 km wide, providing daily global coverage. It has four two-dimensional spectrometers covering wavelengths from 270 to 2385 nm, which is divided into eight wavelength bands. The third wavelength band, from 320 to 405 nm, is used for HCHO retrieval. The spectral resolution of this wavelength band is about 0.5 nm (FWHM). The spatial resolution of the instrument is 7 × 7 km 2 at nadir.

We also compared data from the Visible Infrared Imaging Radiometer Suite (VIIRS) i.e. NO 2 and HCHO, can be used to understand the chemical sensitivity and ozone regimen, i.e. VOC-limited, NOx-limited.

Data are aggregated in approximately 10 km 2 cells. Population data were obtained by aggregating census tract centroids that fell within each cell. Data correspond to the 2010 census (INEGI, 2010) , the most recent population data available at that level.

Population data were mapped against CO, NO 2 and HCHO concentrations. Values were obtained by IDW interpolation of the grid centroid values.

The Mexico City Metropolitan Area (MCMA) is among the 10 largest cities in the world with people exposed daily to high pollution outdoor conditions. The city's location in the tropics (19º N) and its altitude (2240 masl) create a high-radiation environment, a condition that drives the chemistry leading to the formation of photochemical ozone. One of the key reactions in ozone formation in the troposphere is the photodissociation of NO 2 . The photolysis frequency depends on the solar actinic flux that depends on local optical conditions, including aerosols and gases present in the air above the city (Castro et al, 1997; 2001) . Temperature and UVA radiation data from all stations was used, and both parameters are directly linked to the photolysis of NO 2 and O 3 formation atmosphere.

In the MCMA, ozone concentrations reach risk thresholds very often, between 01:00 for all the combination of years. The peak for O 3 medians of all stations in the three years was 80 ppb, although the maximum for the 2020 three-month period appeared one hour earlier than for the peaks for 2018 and 2019. There was also a slight decrease in the 2020 O 3 peaks for TLA and SAG located in the north of the city (representative sites) indicating that upwind regional contributions of ozone were reduced also. The small increases in nocturnal O 3 for 2020 were consistent with the effect of the lower nighttime emissions of NOx during the breakdown, which limited the titration reaction of remaining O 3 by NO. Figure 4 shows the 06:00 to 09:00 scatter plots of NOx vs. CO correlations for medians of the hourly averages from all monitoring sites in the MCMA, from March to May for the three years. The hourly ambient CO data available on the SEDEMA database for the MCMA come from instruments whose equivalent EPA method code is less than 500. That implies that they did not have enough resolution to observe changes below 1 ppm. Those instruments measure CO concentrations at only 0.1 ppm and have in general detection limits of 0.5 ppm which does not represent a problem for evaluating compliance of air quality standards. In fact, data are reported to just one decimal place.

In order for the CO data to be useful for our evaluation, we obtained a correction equation from a statistical correlation analysis between hourly data registered with a CO NDIR analyzer typical of the monitoring network and a PICARRO analyzer which measures CO at parts-per-billion (ppb) sensitivity with negligible drift using the Cavity Ring-Down Spectroscopy (CRDS) method, at the same site (CCA station). Hourly data from February to the middle of March 2020 were used for the analysis. CO data from the CRDS analyzer were partitioned into hourly CO intervals of 0.1 ppm according the records from the CO NDIR instrument. The average of each group of CO data for each J o u r n a l P r e -p r o o f Journal Pre-proof instrument was calculated resulting seven bins of 0.1 ppm each, plus one additional which included all the CO NDIR concentrations higher than 0.7 ppm. For the correlation analysis, we set a CO dummy value as an equivalent to CO background concentration of 0.1 ppm as the intercept (CO axis) to ensure the background condition.

The best coefficient of determination (R 2 = 0.97) was obtained with a polynomial regression model (y = 0.184 + 1.004x -0.004x). With this equation, the CO 06:00 -09:00 averages of the medians were corrected for all years. We assumed that the correction could represent the CO averages reasonably well. Figure 4 shows that the span of data points in 2020 was shorter than previous years, and that trend was stronger from March to May as restrictions on mobility and non-essential activities became stronger while the epidemic progressed. The slopes (∆CO/∆NOx) of the correlations for each month in the years 2020 and 2019 were relatively similar between them. Interestingly, the slopes for the months of March and April 2020 were within the range of the slopes for the three months in previous years, but the slope for May 2020 was higher than any of the other months in all years.

The t-test performed to identify differences between the 06:00 to 09:00 means of the medians for CO for all data among the three years showed that they were not different to each other, although the difference between 2018 and 2019 was small.

However, the t-test for the respective 06:00 -09:00 means derived from the NOx hourly medians showed strong significant differences between the years 2018 -2020 and 2019

However, the 2020 columns are 19% lower on average than 2019. Figure 7c shows a similar pattern for HCHO, but it is only reduced by 7%. Reducing mobility and economic activities to avoid COVID-19 infections decreased NO2 emissions, and did affect HCHO (VOC) emissions too, but to a minor degree; therefore, the HCHO/NO2 ratios for 2020 were larger than those for 2019 as seen in Figure 5d . If HCHO/NO2 < 1.5, it is within the VOC limited region, 1.5 < HCHO/NO2 < 2.3 corresponds to the transition region, and HCHO/NO2 > 2.3 to the NOx limited region (Chang et al, 2016) .

On average, the ratio was 1.49 in 2019, whereas it has grown to 1.61 in 2020. This slight change is just enough to displace the system to a transition region.

The MCMA is not uniform in its population density and economic activities, nor in emissions. TROPOMI CO column density data (figure 8), representing combustion energy use, showed a larger drop in the industrial (northwest) and business (downtown) quarters than in the densely populated residential areas (downtown to east and northeast). A somewhat similar pattern is observed in the NO 2 column map and table 3.

The more affluent parts of the city (downtown to west and southwest) those more able to abide by the stay-at-home directive, show stronger drops in CO and NO 2 column density. In 2019, the HCHO column density map, as a surrogate for VOC, shows higher column densities in the more densely populated areas, in the industrial quarters to the northwest, the financial services quarters from downtown to Santa Fe on the west side and in the forested mountains and the greener and wealthier parts of the city to the west and southwest. The strong reductions in VOC emissions during 2020 show a more uniform distribution but with a stronger color where two or three characteristics of population density, green areas, and wealth tend to coincide. Coarse gradients are products of the sparse cells grid coupled with the interpolation method used.

It is very likely that the stay-at-home directive, the lockdown of non-essential activities to reduce coronavirus propagation in the MCMA, which included a ban issued on private car use according to the last digit in a car's license plate to further reduce mobility, have displaced the average VOC/NOx ratio to an ozone chemistry region between a transition region or a region sensitive to NOx , as shown in figures 5. Also, using CH 2 O column density as surrogate for ambient VOC, supports a displacement to a transition to a NOx sensitive regime. Both analyses are consistent in the conclusion that a considerable part of the MCMA is no longer VOC sensitive in this ozone season under lockdown due to the COVID-19 epidemic.

To further understand this unplanned air quality experiment, we assume that the ozone formation chemistry of the MCMA may follow the idealized Empirical Kinetic Modeling Approach isopleths plot as proposed by Sillman and He (2002) , which is based on a typical scenario of morning NOx and VOC emissions normalized to the respective extension of the urban area (figure 9). Data point A in figure 9 represents the assumed average NOx and VOC emissions spatially distributed in the MCMA according to the emissions inventory for the year 2016 (SEDEMA, 2018) and the estimated urban surface area for the same base year. Coincidentally, the data points fall on the 80 ppbv maximum ozone isopleth, about the same as maximum ozone in figure   3 . To stay in the same ozone isopleth after strong NOx emissions reductions and crossing the transition line to the NOx sensitive region, an increase in VOC may be needed. However, VOC from incomplete combustion should follow the same decrease as that observed for CO. With current data, we cannot distinguish between contributions from domestic, commercial, and mobile sources, but these category sources may not be the origin of the required VOC increase. Indeed, any additional VOC needs first to J o u r n a l P r e -p r o o f replace the losses linked to CO reductions and add more to increase the VOC/NOx ratio.

Where may these necessary VOCs come from? It may be assumed that domestic use of energy for cooking at home did increase, but not enough to compensate for the overall decrease in other sources as observed in the CO map (figure 7). VOC from incomplete combustion of GLP in domestic appliances may have increased, whereas those from incomplete combustion of gasoline did decrease. Diesel may have changed less, as in Mexico is mostly used in heavy duty public transport and supplies. However, sanitation guidelines during the lockdown and beyond may have induced an increased use of cleaning products, many of them containing reactive VOC, alcohol, and chloride.

Additionally, evaporative VOC emissions need to be taken into account as March and

April have been warmer in recent years. The same may be said for biogenic VOC.

Following the approach of the basic graphic tool, we hypothesize that the atmospheric chemistry associated with the unexpected sustainment of a constant maximum ozone average in the MCMA followed the same isopleths observed in previous years.

To develop effective strategies for reducing ozone, we need to understand the complex relationship between ozone, VOC and NOx in rural and urban emissions. The persistent evidence.

The analysis we share in this paper falls short of fully explaining this unplanned air quality experiment in MCMA. The analysis needs to be expanded to the complete Megalopolis of Mexico. Available data on mobility from apps or services need to be transferred to modeling-ready emissions inventories to reproduce the observed pollutant concentrations. The same needs to be done for energy consumption and supply, as well as for cleaning and sanitation products. So, lessons need to be learned quickly to contribute to the return to a more sustainable future normality.

Most of Mexico City lives under a volatile organic compound (VOC) limited system; that is, the formation of O 3 has been determined to be in the VOC-sensitive regime, where an increase of VOC leads to a rise in O 3 while an increase in NOx leads to a decrease of O 3 . The VOC limited regime frequently occurs in densely populated urban atmospheres. So, the usually high concentration of ambient ozone in Mexico City no longer can be attributed only to vehicles, nor to gasoline composition. A paradigm shift to extend emission control policies to other sources or activities is needed. J o u r n a l P r e -p r o o f 

CCA de la UNAM

Investigating ambient ozone formation regimes in neighboring cities of shale plays in the Northeast United States using photochemical modeling and satellite retrievals

The impact of COVID-19 partial lockdown on the air quality of the city of Rio de Janeiro

Satellite data of atmospheric pollution for U.S. air quality applications: Examples of applications, summary of data end-user resources, answers to FAQs, and common mistakes to avoid

Comparison of Emission Inventory and Ambient Concentration Ratios of CO, NMOG, and NOx in California's South Coast Air Basin

Chemistry of environmental systems: fundamental principles and analytical methods

Sentinel-5 precursor/TROPOMI Level 2 Product User Manual Nitrogen Dioxide

Censo de Población y Vivienda

Detectability assessment of a satellite sensor for lower tropospheric ozone responses to its precursors emission changes in East Asian summer

Red automática de monitoreo atmosférico

Calidad del aire en la Cd

Secretaría de Medio Ambiente de la Ciudad de México. Movilidad durante la contingencia por COVID-19 en la ZMVM

Amplified ozone pollution in cities during the COVID-19 lockdown

Some theoretical results concerning O 3 -NOx-VOC chemistry and NOx-VOC indicators

Revisiting the effectiveness of HCHO/NO2 ratios for inferring ozone sensitivity to its precursors using high resolution airborne remote sensing observations in a high ozone episode during the KORUS-AQ campaign

Changes in air quality during the lockdown in Barcelona (Spain) one month into the SARS-CoV-2 epidemic

Assessment of the ozone-nitrogen oxide-volatile organic compound sensitivity of Mexico City through an indicatorbased approach: measurements and numerical …

Development of a nighttime shortwave radiative transfer model for remote sensing of nocturnal aerosols and fires from VIIRS

The authors are grateful to Ana Rosa Rosales Tapia, from the Institute of Geography at UNAM, and Miguel Angel Flores Roman, from the National Institute of Ecology and