Impact of COVID-19 Lockdowns on Air Quality and Health. Association between Concentrations of Tropospheric Ozone and Infection Cases and Deaths in Spain


 BackgroundThis work describes the changes of the air quality and the health implications caused by the lockdown of the first-wave provoked by the SARS-CoV-2 pandemic. Air pollutants were studied in 83 locations in Southern Spain. The study covered urban and industrial gases, NO2, CO, SO2, H2S and O3, and also PM10 and PM2.5 particles.MethodsIt was evaluated the increase and decrease of concentrations during the state of alarm declared on 14th March. Pearson correlations for air pollutants, meteorological factors, vehicular traffic densities (VTDs) and data of infections and deaths caused by the COVID-19 disease were also assessed.ResultsIt was found a clear reduction in carbon monoxide (-25% to -83%), particulate matter (-21% to -42%) and mainly nitrogen dioxide (-55% to -81%) in trafficked areas during the lockdown, reducing cardiovascular and respiratory problems. CO, SO2 and H2S increased (+26 to 34%, +68 to +85% and +32 to +84%) at industrial locations. O3 increased along the lockdown period coinciding with reductions in NO2 and CO (r = -0.90 and -0.81). This ozone rising constitute the ozone lockdown effect (OLE), increasing the risk of pneumonia hospital admissions. Regarding traffic, Pearson coefficients between ozone and VTDs were higher during lockdown than pre-lockdown period, and in the most trafficked areas a reduction in PM10 and PM2.5 levels was observed, contributing this also to the OLE.ConclusionsEffects of ozone on COVID-19 disease was revealed by the graphic associations and correlations found between O3 levels and infection cases and deaths, which were remarkable, constituting in this case the ozone COVID effect (OCE): when concentrations of O3 increase, the incidence of the disease is higher; when O3 falls, infection cases are reduced.


Introduction
The crisis generated by COVID-19 pandemic is consolidating itself as a phenomenon unprecedented with unpredictable consequences. In the hope of nding a de nitive vaccine, most of the governments induced their population to lockdown in their homes or cities. Considering the number of deaths and infection cases, Spain has been from the beginning one of the countries most affected worldwide and in Europe by this crisis [ 1 ].
In the rst wave of the pandemic the Spanish Government declared the 'state of alarm' on Saturday 14 th March 2020 [ 2 ]. As the days go by, the lockdown limited signi cantly the activity of the population. Once the trend of infections decreased, on 4 th May the Government started the deescalation process, probably too soon. Consequently, a rst outbreak occurred on 17 th May with people infected by a ight from abroad. At the end of July, Spain suffered numerous outbreaks that increased the spread exponentially to more than 1000 infections/day, more than 6000 cases in August/September and more than 20,000 (4,000 in Andalusia) in October/November [ 3 ].
End of September Spain declared the second wave with the highest rate of daily cases of COVID-19 with 38,273 new cases (data of 16 th November), which was signi cantly higher in comparison with the new cases in Germany, 14,580, Italy, 27,352, United Kingdom, 21,363 or France with 9,072 cases [ 4 ]. The Region of Madrid (839 cases/100,000 inhab.) had many cities with incidences over 1000 cases, that is why nally on 5 th October it was declared the second lockdown here. As a second "state of Alarm" was declared on 25th October for 6 months [ 5 ], the rst objective of the present work is the need to provide data on the rst lockdown, focusing toward the lockdowns in late 2020 and early 2021, since health experts predict a third wave after Christmas holidays in terms of the COVID-19 disease and other health effects. For instance, some authors demonstrated a 24%-35% reduction before and after lockdown in hospital admissions for other health effects, such as acute myocardial infarctions in France [ 6 ]. Besides, according health experts, it is expected.
The number of infections is different for different cities, so in addition to the determinant factors, such as the number of inhabitants and population density, it is necessary to know if other factors can in uence the incidence of the disease. So, the second objective is the search of possible relationships between the SARS-CoV-2 and some air pollutant or meteorological variable. For this aim we will quantify by means of graphical and statistical tools the variations produced by the lockdown on atmospheric composition and the relationships of air pollutants and meteorological factors with the COVID infection cases and deaths.

Page 3/13
The region of study in Spain. Characteristics of the Southern Europe The Southern Europe studied was the Region of Andalusia (8.4 million inhab., 87,268 Km 2 ) and with a particular emphasis on the cities of Sevilla, Barcelona and Madrid. Andalusia can be differentiated climatically into two areas by the Strait of Gibraltar. The Western zone of the Atlantic Ocean includes the provinces of Huelva https://goo.gl/maps/ooQL5u1XgTK2nwQ48, Cadiz https://goo.gl/maps/f8q3TrfRR2ApQFrd9, Sevilla https://goo.gl/maps/g9juFtEyEh1ucBZR7 and Cordoba https://goo.gl/maps/5sVSpxdwGfebPJQH7, and the Eastern area of the Mediterranean Sea includes Jaen https://goo.gl/maps/Y12NQsXtXeC5ftTv5, Malaga https://goo.gl/maps/iShggpNQAbMX8tMy5, Granada https://goo.gl/maps/zeCvBPS6Y6jGHbWt5 and Almeria https://goo.gl/maps/Jjwk1wK152TtuGym7.
In the Western zone the Guadalquivir River Valley forms a triangle of about 15,000 Km 2 ( Figure 1) through which the Atlantic winds enter from S-SW direction crossing the Valley. Winters are short and mild, so domestic heating systems are not a source of air pollution. There are numerous anticyclone events and frequent thermal inversions. Saharan dust intrusions are also carried toward the region. The Eastern part is characterised by four microclimatic subzones: two adjacent and very different areas, 'Sierra Nevada' (Granada) and the 'Costa del Sol' (Malaga), and two more arid and drier areas: Jaen, widely dedicated to the olive cultivar, and Almeria, with a desert climate.
The results obtained in the Region of Andalusia can be extrapolated to the rest of the Iberian Peninsula, to the Mediterranean countries of Southern Europe and to other European countries or the world.

Characteristics of the Air Monitoring network
The Andalusian Government has an air quality monitoring network with more than eighty stations distributed throughout the eight provinces of the region to measure air pollutants ( Figure 1, Table 1). The corresponding measurements can be downloaded from the website of the Department of Environment of the Government of Andalusia [ 7 ], however validated concentrations were provided by the same Department through o cial requests. The monitoring stations are classi ed in relation to the emission source (background, industrial, tra c) and the type of area (rural, urban, suburban). Table 1 shows the name of each station, the city council and the province where they are located, as well as the type of area and emission source, mentioned above. Data of the region of Madrid and its capital was provided by the Department of Environment of the Community of Madrid from its Air Quality Network [ 8 ]. The data of the city of Barcelona was provided by the Department of Territory and Sustainability of the Government of Catalonia from its Air Pollution Monitoring and Forecasting Network [ 9 ].
Air pollutants, meteorological factors, tra c densities and COVID data In this work we have processed concentrations of gas pollutants, NO 2 , CO, SO 2 , H 2 S and O 3 and suspended particles, PM 10 and PM 2.5 . In order to study whether there was signi cant variation in the magnitude of these vectors caused by the lockdown period, daily medians were gathered from all stations for four months, from January 1 st to April 30 th 2020. In parallel, daily records of meteorological parameters for the same period were obtained from the Spanish Meteorological Agency [ 10 ]. They were wind speed (WS), rainfall (RF), air temperature (AT), atmospheric pressure (AP) and air humidity (AH).
The rst consequence of a lockdown is a drastic reduction in vehicle tra c through the cities and roads. In the rst days of the lockdown, some press [ 11 ], blogs [ 12 , 13 ] and preprints [ 14 , 15 ] reported that the restrictions of economic activities and population movements resulted in a decrease in the level of air pollution. Thus, in order to study the in uence of the lockdown on the vehicle transport, the data of tra c densities were available in the largest city in Andalusia, the city of Sevilla (37º23'N, 5º58'W, 141.3 Km 2 , 7 m a.s.l.), the most representative city in terms of air pollution by urban tra c. Daily data of thirty entry/exit routes from the city were provided by the Mobility Management Centre [ 16 ], the tra c control centre of the City of Sevilla.
The "state of alarm" did not stop the progress of the number of daily virus infections and deaths until the end of April 2020. The mechanism of the virus transmission is not yet fully controlled, but it is also not clear if there are other factors that may contribute to the incidence of the disease. For this reason, our study investigated the relationships between meteorological and pollutants data with the number of daily infections and deaths [ 17 ] caused by the SARS-CoV-2.
Positive-negative effects of lockdown. Strategy for the cause-effect patterns recognition The methodology was divided into three stages: the rst consisted of detecting the change of slope from 14 th March in each concentration-time plot. Then differences before-after 14 th March were estimated and quanti ed in percentage. Thirdly it was performed a study on correlations including meteorological variables, air pollutants, tra c densities and data of infection cases and deaths. The Pearson high coe cients were validated con rming with its graphic pro les [ 18 ].
Alternatively we could evaluate the lockdown effects by comparing the 2020 data with those of previous years, such as the 2019 year; however it is not clear that these comparisons truly re ected the impact of lockdown on environmental concentrations as was declared by Jia et al. [ 19 ]. In consequence we have considered more effective to compare two periods: a) A pre-lockdown period, 1 st January/13 th March and b) a duringlockdown period, 14 th March/30 th April.

Results And Discussion
Air pollutants in pre-lockdown and during-lockdown periods. Health consequences We have evaluated the behaviour of NO 2 , CO, SO 2 , H 2 S, O3, PM10 and PM2.5 along the four months of the study at each station of the network to determinate graphically how the lockdown has in uenced the air quality. Besides, we have calculated the change in slope and quanti ed the percentage of variation of the concentrations before and after 14 th March according with the type of area and source.

Nitrogen dioxide and carbon monoxide
Nitrogen dioxide is the most clearly reduced its concentration (global mean of -53.7%) in all stations [ 20 ]. Only three nearby industrial areas increased the level. NO 2 is the typical pollutant emitted by car exhausts and energy production, in both NO and NO 2 species, i.e. NO x [ 21 ]. Carbon monoxide is originated from similar sources [ 22 ] and its concentrations down by -27.0%. . The rest of urban and suburban stations decreased over 60.4%. rural stations the percentages were -24.2% in Gerena and -37.9% in Sierra Norte, both located more than 50 km from the city, so the rate of reduction decreases the more distance from the big city. The behaviour of NO 2 in the rest of cities is shown in Figure 3b. Some representative reduction rates were -66.1% (in Pozo Dulce, the centre of Huelva), -68.4% (in Marbella, a city of Malaga Province) and -81.3% (in San Fernando, a city of Cadiz Province), having all of them high tra c densities and nearby industrial activities (Huelva and Cadiz), as it was also reported by other studies [ 24 , 25 , 26 ).
Due to the direct relationship between the concentration of NO 2 in the air and hospital admissions for cardiovascular disease [ 27 ], the health consequences of the NO 2 reduction were extremely bene cial, such as the lower number of acute hospital admissions for cardiovascular disease in the short-term (acute ischemic syndromes, atrial brillation, and decompensated heart failure [ 28 ], and possibly a lower cardiovascular mortality due to environmental causes in the long run. Regarding carbon monoxide, stations decreased its concentration after lockdown and others increased it, as it was reported in USA [19], Italy [ 29 ] and by interdisciplinary studies developed for 40 cities [ 30 ]. The highest decreases starting from the pre-lockdown values were observed in stations of Sevilla (-45.0% from 477 µg m -3 ), Jaen (-48.7% from 432 µg m -3 ), Cadiz (-52.6% from 522 µg m -3 ) and Huelva (-58.2% from 705 µg m -3 ), with values two times higher than in Cordoba (-32.7% from 370 µg m -3 ) and Granada (-33.0% from 363 µg m -3 ). Pre-lockdown levels in the period 1 st January-13 March 2020 were similar to the 2019 values for the same stations and period. Thus, the highest decreases were also related to high pre-lockdown concentrations, as in the case of NO 2 . In conclusion, for both NO 2 and CO the highest reduction tases were in stations where pre-lockdown levels are elevated. These stations were near to the two most important industrial areas, the Chemical Park of Huelva, and the Campo de Gibraltar of Cadiz, but also in cities with high rate of vehicles circulation, the province of Sevilla. These results are observed by the SENTINEL-5P satellite measurements of CO concentrations when comparing both periods (Figure 2, bottom): pre-lockdown values were between 106 and 129 µmol m -2 , and between 60 and 83 µmol m -2 during lockdown.
In the case of carbon monoxide the health consequences of the CO reduction are not clear because no direct relation between CO at low-medium concentrations and hospital admissions are usually found by researchers [ 31 ]. Some authors found direct association with negative effects of short-term CO exposure only over 2,000 µg m -3 [ 32 ].
The reduction rates of CO in Sevilla were from -24.8% in Bermejales (-63.1% NO 2 ) to -50.0% in Torneo (-55.4% NO 2 ) (Figure 3c). Curiously, within the lockdown period we have observed a small and unexpected period of increase inside the decreasing pro le. This coincides with the Easter holiday period (6 th April), which suggest the departures of private vehicles from the city despite of being forbidden for citizen. When the behaviour of CO in the rest of cities was decreasing ( Figure 3d) the stations with the most reduction rates had high tra c densities, such as Pozo Dulce (in On the other hand, if a lockdown consists of a total braking of all vital activities, an increase in CO should indicate an increase in 'essential' economic activities that cannot be stopped. In the province of Sevilla, the increases corresponded to stations located near major industrial areas (see above google maps links): Gerena (+35.8% CO, -24.2% NO 2 ) is next to a big copper mine. Similarly, three stations close to each other, Ranilla (+25.0% CO, -69.2% NO 2 ), Alcala (+16.5% CO, -71.8% NO 2 ) and Dos Hermanas (+19.4% CO, -71.4% NO 2 ) receive the in uence of a big stainless steel factory, two detergent plants and a cement factory. In addition, the CO pro le presented a continuous increase in the form of 'sawtooth cycles' (Figures 3e and 3f). This type of variation was also observed for aerosol particles [ 33 ] and for particles and nitrogen dioxide. In our case, observing these cycles we found that they occured coinciding with rainfall episodes as it can see in Figure 3e.
In the rest of cities the increases of CO corresponded to stations Joya, Cartuja, Barrios and Marconi (Figure 3f). Joya (Almeria, +43.2% CO, +16.6% NO 2 ) and Barrios (Cadiz, +27.7% CO, -37.0% NO 2 ) are next to coal thermal power plants, and Barrios is near the rst major stainless steel factory of Andalusia (see above google maps links). Cartuja (in Jerez de la Frontera, Cadiz, +26.6% CO, -79.6% NO 2 ) is next to the industrial zone of the city and Marconi (Cadiz, +25.1% CO, -85.0% NO 2 ) is next to the 'Zona Franca' industrial area and the 'Bay of Cadiz' Port Area. Consequently, differences between decreases and increases of carbon monoxide concentrations were caused by the increasing activity in power generating industries and others essential industries during the lockdown period. Ejido (+190.5% in SO 2 , -79.0% in NO 2 ) has no relevant industries, but it contains the largest extension of plastic greenhouses in Spain ('sea of plastic'), and H 2 S also increased +103.2%. Greenhouse plastics are usually degraded by elemental sulfur, which is used to disinfect greenhouses, which is why commercial plastics for greenhouses are manufactured protected against sulfur. So, these extreme values are due to burning of used plastics mixed with organic residues. Also, in Moguer, next to the petrol re nery in Huelva Chemical Park, an increase of +84.7% in SO 2 was observed (-56.2% in NO 2 ), similar to those reported in UK [ 37 , 38 ]. Obejo and Villaharta are near a coal thermal power plant, and the values increased too much for SO 2 (+82.6% and +92.3%) and for H 2 S (+84.3% and +32.5%) after 14 th March. Villaharta also receive NO 2 emissions (+34.7%) probably from the chimney of the power plant (see above google maps link). Regarding Gerena, the mine zone showed an increase of +67.9% in SO 2 in comparison with a decrease of -24.2% in NO 2 . All these extreme increases in SO 2 levels must be attributed to the increase in burning of coal [ 39 ] or other fossil fuels by thermal power plants [ 40 ] or other industrial activities, such as mines or re neries. According Wang et al. [25], concentrations of SO 2 and other pollutants such as CO and NO 2 appeared to have positive effect on hospitalizations, especially at high concentrations, for instance by bronchiectasis. Also other authors [ 41 ] reported that only sulfur dioxide was associated with out-patient visits.
Associations of SO 2 and CO with pneumonia were reported [ 42 ]. Consequently, it would be advisable to observe especially these unusual patterns to alert the population, although they are less frequent than the general decreases observed for the studied air pollutants.
Ozone and Particulate matter (PM 10 and PM 2.5 ) In 2018, 1,206 exceedances of threshold limits [ 43 ] were reported, which 979 were due to tropospheric ozone, 224 to particulate matter, 3 to nitrogen dioxide and none to sulphur dioxide.

Ozone
The atmospheric conditions most suitable for achieving high levels of this secondary pollutant are the result of the reactions of nitrogen oxides (NO x ) with carbon monoxide, methane and non-methane volatile organic compounds (NMVOCs), i.e. their precursors, under high solar radiation and temperatures [ 44 ]. Excessive ozone in the air can have a marked effect on human health. It can cause breathing problems, trigger asthma, reduce lung function and cause lung diseases [ 45 ].
In connection with our study, we observed that the concentration of ozone increased during the lockdown in all monitoring stations. In this sense, the increase was related to the fall of NO 2 and CO, because of O 3 concentrations were negatively correlated with NO 2 , CO and also PM 2.5 and PM 10 . The increase is attributed in almost all studies to the decrease of NO 2 (NO x or NO) and VOCs [ 46 ]. The highest increases occur in tra c stations where the pre-lockdown levels of ozone are low. It is certainly not normal that whilst in most stations nitrogen oxides decreased, they increased in parallel ozone levels. Therefore we propose an explanation: if under certain conditions the reduction of NO 2 emissions can involve an increase in ozone levels, a process well known as 'Ozone Weekend Effect (OWE)' [ 47 , 48 ], we asked ourselves, Could the weekend effect explains similarly the increase in ozone during lockdown? Our hypothesis a rm: "despite the decrease in the levels of its main precursors, NO 2 and CO, the increase in ozone levels during lockdown is explained by an effect that we will be calling the 'Ozone Lockdown Effect (OLE)', similar to the 'weekend effect' but longer in time". Other similar study developed in four cities from Southern Europe, including Valencia (Spain), also reported a similar concept, the 'lockdown effect on O 3 production' as the increase of ozone during lockdown for a longer period [ 49 ].
In terms of health effects, some authors declare that no signi cant relationship between tropospheric ozone and hospital admissions was found for respiratory diseases [ 50 ]. However, the majority of studies found associations with respiratory and cardiovascular affections [27, 51 , 52 ], such as pneumonia and COPD, chronic obstructive pulmonary disease [ 53 ].
If we make a joint study for CO, NO 2 , O 3 before/after 14 th March in Sevilla, the levels of NO 2 were reduced by up to -19 µg m -3 in Centro (historical Centre of Sevilla), Torneo and Ranilla, while the maximum reduction of CO reached -401 µg m -3 in Torneo, Ranilla and also Santa Clara (in the border of Sevilla). As ozone was negatively correlated with NO 2 and CO (Figures 4a and 4b) [52] where the reduction of road tra c is more signi cant than in those areas where the reduction occurs near industrial activities.
These results con rm our hypothesis and thus, despite the decrease of NO 2 and CO, the increase of tropospheric ozone during the lockdown constitutes the 'ozone lockdown effect, OLE'. In addition, in the 4-month period studied, the 'ozone weekend effect, OWE' was observed in most of the weekly periods of all stations. PM 10 and PM 2.5 As mentioned above, the poor air quality was mainly due to particles and ozone. Nowadays, particulate matter is a primary pollutant that has a negative correlation with ozone [45] because of aerosols can have a radiation effect through both absorption and scattering [ 55 , 56 ]. Furthermore, other authors propose that aerosols reduced the actinic ux photolysis rates under absorption and scattering, so the surface ozone production decreases [ 57 , 58 ]. Li et al. [ 59 ] demonstrated that O 3 was strongly inhibited under the condition of high PM 2.5 concentration. In addition, as in the case of CO and NO 2 , our correlation study shows also high negative correlation coe cient between O 3 and PM 10 PM 2.5 (Figures 4c and 4d).
In the province of Sevilla, particle concentrations decreased in all the areas, particularly in stations located inside the city, Torneo and Principes, with higher PM 10 and PM 2.5 decreases than stations far from the city (Figure 5c levels (-57%) in lockdown were higher than in PM 10 (-42.0%), suggesting and corroborating the much more anthropogenic contribution (Figure 5d). stations with a high tra c, such as Jerez (-42.5%) and Marconi (-28.1%). Rural and background areas, such as the Alcornocales National Park, and the 'Arcos de la Frontera' city, show low decreases, -7.3%, while in industrial areas PM 10 seems to decline very lightly, such as in Rinconcillo (+5%), Linea and Palmones (-5%) or even rising in Zabal (+14%), all them near a big stainless steel factory and a big petrol re nery of the Bay of Algeciras. In Huelva, the same perceptions were observed. The greatest decreases in PM 10 occur within the city of Huelva, where the most intense urban tra c was stopped at 14 th March. Titan (-32%), Rosales (-29%), Pozo Dulce (-25%) and Carmen (the University Campus, -21%) were the representative stations, whereas inside industrial areas, Palos and Moguer (-14%), San Juan and Rabida (-11%, -9%) and the coastal areas, Punta Umbria (-10%), Mazagon (0%) and Matalascañas (+8%) had lower decreases or even increased the particle concentrations.
In conclusion, these results in Sevilla, Cadiz and Huelva reinforce the idea that since ozone and PM 10 /PM 2.5 are negatively correlated and tra c zones suffer the greatest reductions, we must consider PMs as further variables involved in the 'ozone lockdown effect OLE' like CO and NO 2 .
Some authors found that during lockdown the ozone concentrations were higher than in normal situations, being this related with the increase of organic carbon and decrease of elemental carbon in aerosols [ 60 ].
Nevertheless, if ozone is the most irritating gas pollutant, it may be assumed that its presence could in uence the incidence of the COVID-19 disease. Therefore, can it be assumed that there is an important relationship between airborne ozone levels and the number of COVID-19 infections or deaths?
Correlation study for air pollutants: relationship with the incidence of COVID-19. Meteorological and tra c implications As mentioned above, correlation studies were done between atmospheric and meteorological variables, tra c densities and also with number of infections and deaths due to COVID-19 disease. The study was done for all representative locations in the region. In the case of tra c densities the study was cantered in Sevilla, the most tra cked city.

Ozone and number of infections and deaths by COVID-19
On the other hand, we have detected an interesting relation between daily infections/daily deaths and ozone concentration. The correlation coe cient between ozone and number of infections was r = + 0.71 (Figure 6a): when ozone is high the number of infections is also elevated and when ozone decreases the infection cases also was reduced.
In addition, Figure 6b shows how infections cases in Sevilla increased 1-2 days after the ozone concentration rise. The possibility of a coincidental parallel increase of ozone and infection cases is not plausible because the pro le also shows a period from 1 st April where decreasing infections is preceded by decreasing ozone. In order to check this behaviour in other important cities outside the Southern Spain, we also studied this relationship in the cities of Madrid and Barcelona (Figures 6c and 6d), corroborating the behaviour observed in Sevilla: in graphics b), c) and d), every peak/valley of the infections black lines (INF) is preceded by a peak/valley of the ozone red lines (O 3 ) (see grey lines), even in the both rising and decreasing periods. This fact con rms our hypothesis, which states that the presence of high levels of ozone is a determinant factor on the incidence of COVID-19 disease, further reducing the infection rate when the ozone level decreases. This behaviour constitutes an effect that we will be calling the 'Ozone COVIDEffect (OCE)'.
The number of deaths was also correlated with the ozone level (r = + 0.59), so that, despite the high rate of person-to-person transmission of the new COVID-19 virus, the high oxidizing character of ozone constitutes the rst ambient factor that enhances the incidence of the COVID-19. As a result, we propose a mechanism for infected people, in which the ozone entering into the lungs accelerates the occurrence of the virus symptoms, even in healthy people, increasing the seriousness of the disease.

Air pollutants and meteorological parameters
Firstly, regarding Pearson correlations studied (Table 2), we con rmed the well known relationships along air pollutants, such as mentioned above for ozone, O 3 with NO 2 , O 3 with CO, ozone with PM 10 , ozone with PM 2.5 and for nitrogen dioxide, NO 2 with CO, NO 2 with PM 10 , NO 2 with PM 2.5 or CO with PM 10 . No signi cant correlations were found for SO 2 (r < ± 0.32). Other well known correlations were also found between meteorological parameters and air pollutants, for instance those for atmospheric pressure (AP), such as AP with O 3 (Figure 7a), AP with NO 2 , AP with CO or AP with PM 10 , those for wind speed (WS), such as WS with O 3 , WS with NO 2 or WS with CO (Figure 7b), or those for air temperature (AT), such as AT with O 3 or AT with NO 2 (Figure 7c).
Correlations with rain should be performed in a different way: a signi cant value of precipitation (RF > 0.4 mm) implies an air-cleaning effect the next day, which is why concentrations of one day need to be correlated with that of the following. So, correlations found (Table 2) for the rainfall (RF) were for CO, RF with SO 2 , RF with NO 2 and RF with PM 10 (Figure 7d), but no correlation was found for ozone. This way, these correlations indicate a gas/particles-cleaning effect, being possible a dissolving effect in the raindrops and thus obtaining rainwater with nitrate or sulphate in solution [ 61 ]. Thus, we have seen above that increases of CO in sawtooth form of Figure 3e were related to the correlation with rain events ( Table  2). A particle-cleaning effect is also produced by rain events [ 62 ]. Regarding infection cases and deaths, atmospheric pressure showed negative correlations with cases of infections and deaths, AP with INF (r = -0.88), AP with DEA (r = -0.71), and positives with temperature, AT with INF (r = + 0.73), AP with DEA (r = + 0.51).

Air pollutants and tra c of vehicles
The correlation study also included the vehicular tra c densities (VTDs) in the city of Sevilla, which should be correlated with air pollutants.
Thirty-three street and roads routes were included. Since the main consequence of the lockdown was the reduction of tra c densities, the correlation study compared Pearson coe cients before 14 th March and after 13 th March, as was studied by Hashim et al. [24].
The results showed that Pearson coe cient for VTD/NO 2 was higher before the lockdown than after it (r = + 0.72 < 14 th March, r = + 0.59 > 13 th March, NO 2 , Figures 7e, 7f) but for ozone it was lower before (r = -0.47 < 14 th March, r = -0.63 > 13 th March, O 3 , Figures 7g, 7h). This fact con rms the regular and major transit of cars into pre-lockdown, and during lockdown, the much minor tra c and the ozone lockdown effect. What is more, the result is different in the case of carbon monoxide, where Pearson coe cient for VTD/CO is positive (r = + 0.53 < 14 th March) within prelockdown, as NO 2 , but negative and higher (r = -0.62 > 13 th March) into lockdown, as O 3 , con rming the rise of road transport along industrial areas. No correlation was found for SO 2 .

Conclusions
According to the results obtained, we have succeeded to describe the effects of COVID-19 lockdown on the change of atmospheric composition.
In general, we have demonstrated the clear decrease in carbon monoxide, particulate matter and nitrogen dioxide due to the tra c of vehicles. On the other hand, we detected the cases in which carbon monoxide and sulfur dioxide increased concentrations in stations located near large industrial areas, such as stainless steel or thermal power plants. Besides we have concluded that some extremely high values of sulfur dioxide and hydrogen sul de observed were caused by the frequent burning activities of greenhouse plastics. Studying the whole of region, it is concluded that there are no differences by geographical factors, only by type of emission. Additionally, the major reduction rates in air pollutants occurred, in general, when the pre-lockdown concentrations were high.
In particular, the most interesting conclusions of this study were related to ozone and airborne particles. Regarding ozone, which is a secondary pollutant negatively correlated with the rest of pollutants, the increase in concentration during the lockdown is attributed to the 'ozone lockdown effect' OLE, similar to the 'ozone weekend effect' OWE, but over the longer term. With regard to airborne particles, PM 10 and PM 2.5 , which were also negatively correlated with ozone, the reduction of levels in areas strongly in uenced by tra c contributes signi cantly to the 'ozone lockdown effect' as CO and NO 2 . Regarding tra c intensities (VTDs), we have con rmed the evidence of the ozone lockdown effect through the fact that the O 3 /VTDs correlations were higher during the lockdown than in the pre-lockdown, in contrast to the VTDs/NO 2 correlations. As we studied more than 80 monitoring stations in the Southern Spain, it is expected that the results and conclusions can be extrapolated to the rest of the country and probably to other Mediterranean countries, as we already demonstrated the same ozone OCE effect in Sevilla, Barcelona and Madrid.
Finally, as a result of the high positive correlation found between ozone and the number of COVID-19 cases, which was graphically con rmed, we can a rm that the presence of signi cant concentrations of tropospheric ozone in the breathing air constitutes a critical factor on the major incidence of the disease. This effect was con rmed on the cities of Madrid and Barcelona with the same results of Sevilla. Consequently, we can stated that ozone is those air pollutant that more health risk present for the population in relation to harmfull respiratory effects and the SARS-

Declarations Declaration of competing interest
The authors declare that they have no known competing nancial interests or personal relationships that could have appeared to in uence the work reported in this paper.