Air Pollution and Covid-19: An Ecological Study in Mainland Portugal

Air quality stands out as an important determinant of health, as its degradation was associated with around 4.2 million premature deaths in 2019, primarily due to heart and respiratory problems. It is shown that the elderly, the children, and individuals with pre-existing health conditions are simultaneously more sensitive to the impacts of air pollution and Covid-19, due to their fragile immune systems. Scientific evidence has shown the consequences of exposure to air pollutants to respiratory system diseases, emphasizing that it could be an important factor in explaining the spatial pattern of Covid-19 incidence and mortality. The aim of this study is to analyze the spatial association between air pollutant PM 2.5 and the incidence and mortality of Covid-19 during March and December of 2020 in mainland Portugal. Weighted geographical models (GWR), were applied to identify and understand patterns, as well as explanatory factors in this relationship. The results obtained through the GWR models reveal that the pollutant PM 2.5 has an association that varies in space. The incidence rate is higher in the southern, central, and northern regions of the country. The results of this study contribute to the analysis and assessment of the impact of air pollutants on human health, specifically in relation to health outcomes associated with Covid-19. It became evident that the concentration of PM 2.5 is an important factor in explaining Covid-19 incidence rate in Portugal.


Introduction
The health status of a population in a given place depends on a vast and complex set of biological, cultural, social, economic, and environmental factors, referred to as health determinants (Marmot & Wilkinson, 2005;Marmot et al., 2008;Santana, 2014).Among the dimensions of determinants that influence health, the natural environment (air, water, and land) stands out: air quality is a significant determinant of health as its deterioration threatens the health of communities and natural ecosystems (Santana, 2014).
Air pollution has contr ibuted to several problems affecting the health and well-being of the population, especially in urban areas (Lelieveld et al., 2019).According to the EEA (2013), air pollution is an important risk factor for several diseases, such as respiratory infections, heart diseases, and lung cancer; consequently, it is associated with increased medication use, as well as medical consultations or episodes of hospital emergencies.Prolonged exposure to certain pollutants can even result in premature death due to their toxicity (Guo et al., 2017).
Studies have been suggesting that exposure to air pollution effects populations unequally (Verbeek, 2019;Shen et al., 2020;Fergunson et al., 2021;Brazil, 2022).In the literature the concept of environmental injustice is used to describe how unequal environmental burdens is influenced by socioeconomic status: socioeconomically vulnerable individuals tend to live in areas with higher levels of road traffic and industrial activity due to the affordable cost of housing, as well as working in locations more exposed to harmful environmental conditions (WHO, 2010;Ortiz et al., 2019;Banzhaf et al., 2019;Ferguson et al., 2021;Frolick et al., 2022;Brazil, 2022).
According to the World Health Organization (WHO) (2020), Covid-19 is an extremely contagious disease that primarily affects the respiratory system.
It is recognized in the literature that environmental, social, demographic, as well as genetic factors play an important role in the risk of contagion and the severity of the Covid-19 disease (Costa & Costa, 2020;Dias et al., 2022;Banik et al., 2020).Among these, air pollutants such as PM 2,5 , NO 2 and PM 10 are considered important factors despite the mechanisms of its impacts are still not fully understood (Ali and Islam, 2020;Sheppard et al., 2023).The common interpretation is that air pollutants exposure impacts the inflammatory and immune response in the respiratory system, which, due to prolonged exposure, becomes fragile and vulnerable to viruses and diseases, including Covid-19 (Wu et al., 2020;Wang et al., 2020).
Thus, the aim of this study is to analyze the spatial association between long-term exposure to the air pollutant particular matter with a diameter of 2.5 micrometers (PM 2.5 ) and the incidence and mortality of Covid-19 during March and December 2020 in mainland Portugal.As far as the authors know, this study represents the first research in Portugal to specifically examine the geographical impact of air pollution on the incidence and mortality rates associated with Covid-19 and it build up from the knowledge produced by several works such as Costa & Costa, 2021;Almendra et al., 2021b;Almendra et al., 2021c;Costa & Costa, 2020;Nogueira et al., 2020;Vieira et al., 2020;Azevedo et al., 2020;Marques et al., 2021;Sousa et al., 2021;Silva et al., 2022;Barbosa et al., 2022.

Demographic data
The resident population and the aging index for 2020 were collected from the Portuguese National Institute of Statistics (INE).

Study Area
The study area is the mainland territory of Portugal (hereinafter referred as Portugal), and the analysis is conducted at the municipal level, which is the most disaggregated administrative level with epidemiological information provided by the Directorate General of Health (DGH).

Association between Air Pollutants and Covid-19
Both Ordinary Least Square (OLS) and Geographically Weighted Regression (GWR) models were applied, using ArcGIS software, in order to analyze the geographical pattern of the association between Covid-19 incidence and mortality and the air pollutant Furthermore, Moran's Index was applied to the residuals of the OLS model to evaluate their spatial heterogeneity.
Both simple and adjusted GWR models were developed.This adjustment by ageing index was made due to significant disparities observed in various phases of the pandemic, with the elderly population being the most affected, especially those residing in nursing homes and senior residences (Almendra et al., 2021a).
GWR was developed to address possible spatial variations in the regression coefficients between variables.Consequently, GWR can essentially be seen as an extension of linear regression models that add a level of modeling sophistication by allowing the relationships between independent variables and the dependent variable to vary depending on location (Faka et al., 2023).The spatial structure of the model was defined based on the AIC criterion (i.e., the number of neighbors and the optimal bandwidth).
Essentially, the main difference between the OLS and GWR techniques is that the parameters in the global model are constant, assuming that the effects are evenly distributed across the country,

Geografia
while in the GWR model, the coefficients have location-dependent variability (Isazade et al., 2023).

Mean Concentration of PM 2.5
Analyzing the variability of PM 2.5 concentrations in mainland Portugal (Figure 1), it is possible to observe that the concentration of this pollutant is not evenly distributed between the portuguese municipalities.The municipalities that surround and belong to the metropolitan areas have high concentrations of PM 2.5 , often exceeding the levels considered acceptable by the World Health Organization (WHO), which states that annual exposure levels above 5 µg/m³ are associated with adverse health effects (WHO, 2021).The same happens in numerous municipalities in the Central and Algarve region.Municipalities bordering Spain also stand out, such as: Vimioso (7302,9/100 000 inhabitants), Chaves (6320,8/100 000 inhabitants) and Marvão (6322,4/100 000 inhabitants).

Covid-19 Mortality Rate in Portugal Mainland (2020)
Similarly to the incidence rate, the spatial pattern of the Covid-19 mortality rate does not follow a heterogeneous pattern (Figure 3).despite not being statistically significant.Even so, the OLS residuals, after applying Moran's I, present a significant spatial autocorrelation.

OLS Models
The spatial pattern of the residuals, confirmed by Moran's I, shows that the OLS model cannot fully explain the spatial heterogeneity of the influencing PM 2.5 supporting the need for developing GWR models.

Geographically Weighted Association Between Covid-19 and PM 2.5 Concentration (GWR)
Figure 4 presents the coefficients resulting from the application of GWR models to assess the association between health outcomes and the pollutant PM 2.5 .It is possible to observe a positive association in most of the country (Fig. 4A); also the association between Covid-19 incidence and PM 2.5 tends to be more intense in municipalities with a higher concentration of this pollutant.
The coefficients showing a positive association were higher along the northern region, while in the central inland region, municipalities bordering Spain displayed negative coefficients, suggesting an inverse relationship.However, the models adjusted by ageing

Cadernos de
Geografia index (Fig. 4B), show a slightly different pattern, as the coefficients resulting from the association between the incidence rate and PM 2.5 are higher in municipalities of the southern coastal and northern interior.
Figure 4C shows the coefficients resulting from the association between Covid-19 mortality rate and PM 2.5 .Most of the country's municipalities, exhibit negative PM 2.5 coefficients, revealing an inverse relationship, with the exception of municipalities from the Lisbon metropolitan area, northern coastal municipalities, the coastal Alentejo, and the western Algarve.However, after adjusting the model for the aging index (Fig. 4D), the pattern presents important differences, it can be observed that, despite the municipalities in the northern and central regions near the border still displaying negative coefficients, southern municipalities have high positive coefficients.

Discussion
This study analyzes the spatial association between PM 2.5 and the incidence and mortality of Covid-19, during March and 2020, in mainland Portugal, using OLS and GWR, to identify, and understand patterns, as well as explanatory factors in this relationship.OLS models showed a significant positive association between Covid-19 incidence and PM 2.5 and a negative non-significant association between Covid-19 mortality and PM 2.5 .
After adjusting by aging index, the association between the incidence and PM 2.5 remained positive and significant and the association between mortality changed direction.The OLS residuals show a significant spatial correlation in both models (simple and adjusted).
GWR models showed that the regions of the north and the Lisbon Metropolitan Area displayed the strongest associations between PM 2.5 and the incidence rate.In addition, observing the results from GWR for the mortality rate, most of the country's municipalities reveal an inverse relationship that partly shifted direction after adjusting for the aging index, highlighting the role of ageing in explaining the patterns of mortality due to Covid-19.
According to Wu et al. (2020), prolonged exposure to PM 2.5 impacts the respiratory system, Nevertheless, this trend is not applicable to all countries, some countries present lower PM 2.5 concentrations but higher incidence and mortality rates (i.e., Russia, Spain, the Uk, and Iran).Another study conducted by Middya & Roy, at district-level, in India, using GWR models to study the association between Covid-19 mortality rate and PM 2.5 adjusted by total population, age, education, and households with at least 9 persons, concluded that in the western districts of India there is a strong positive association between PM 2.5 and Covid-19 mortality, whereas in the other districts, there is no such strong association.
According to the authors, this heterogeneous distribution is related to many underlying factors, such as demographic, socioeconomic, and environmental pollution variations between different districts of India.Similarly, the results obtained in our study also point out spatial differences and spatial inequa- the disease tends to be severe in the older population, due to the weakening of the immune system throughout life (Chen et al., 2021).
The study of Zhou et al. (2020)

Conclusion
The Covid-19 pandemic has brought various consequences for society at different levels.Its PM 2.5 .OLS is a regression analysis technique and according to the fundamental theoretical assumptions of this technique, the relationship between the dependent variable and the independent variables can be defined as linear where the values of Covid-19 Incidence or Mortality Rate are estimated by the values of PM 2.5 and Aging Index (Chwialkowski et al., 2023).
By assessing the spatial pattern of the Covid-19 incidence rate (Figure2), it is possible to observe that the distribution does not follow a homogeneous pattern.The metropolitan areas (Lisbon and Porto) and the municipalities surrounding them have a high incidence rate.In the case of Lisbon Metropolitan Area, it is possible to highlight the municipalities:Loures which has an incidence rate of 4745,27 cases per 100 000 inhabitants, Amadora (4518,41/100 000 inhabitants), Lisboa (4159,42 /100 000 inhabitants).And, for Porto Metropolitan Area it is possible to highlight the municipalities of Póvoa de Varzim (7392,42/10 0 0 0 0 inhabitants), Vila do Conde ( 7 270,0 2/10 0 0 0 0 in ha bit a nt s), a n d Valo n g o (6078,88/100 000 inhabitants).
making it fragile and vulnerable to viruses and diseases, including Covid-19.An exploratory study conducted by Yu et al. (2021), at country level for the entire globe, also used GWR to analyze the association between Covid-19 incidence and mortality rate with PM 2.5 concentration.They concluded that PM 2.5 concentrations influence the spatial patterns of C o v i d -19 o u t c o m e s a c r o s s t h e s t u d y a r e a .
impact and long-term repercussions are still the subject of ongoing studies.Identifying spatial patterns, trends, and explanatory factors in the spread of Covid-19 is important for understanding the geographical nature of the disease, allowing the formulation of strategies based on evidence.The results obtained with the methods used, reveal that long-term exposure to air pollutants is significantly associated with the incidence rate of Covid-19 at the national level.However, this is not clearly observed for the mortality rate, suggesting that municipalities with higher mortality rates from Covid-19 may be less exposed to air pollutants.Nonetheless, other relevant Covid-19 determinants display relevant role in the outcome of this disease and are out of the scope of this text.After adjusting for the aging index, it is possible to identify sets of municipalities where the association is positive.The results obtained should be interpreted with caution, as they do not indicate a protective effect associated with exposure to air pollutants.Further research is needed to understand the internal factors that influence both pollution and mortality rates in these municipalities.This work provides valuable information for policymakers and local intervention stakeholders to adequately identify risk factors that influence Covid-19, including air pollution.However, further research and an expansion of the variables analyzed are needed to build knowledge about the various nº 49 -2024 factors that modify the association between exposure to pollutants and the incidence and mortality of Covid-19.

Table 2
OLS Model Adjusted by the Aging Index