Comparison of Metropolitan Cities for Mortality Rates Attributed to Ambient Air Pollution Within the Context of SDGs Using the AirQ Model

In the present study, the air pollution dynamics of the metropolitan cities of Balıkesir, Bursa, İstanbul, Kocaeli, Sakarya and Tekirdağ in the Marmara Region, which is the geographical region with the highest urban and industrial activity in Turkey, were examined for the time period between 2016 and 2019. Annual changes in the cities in terms of air pollution, which was examined with a focus on the PM 2.5 parameter as indicated by United Nations (UN) Sustainable Development Goals (SDGs), differences in the cities by years, and the seasonal changes in air pollution in the cities were investigated. Additionally, mortality rates attributed to air pollution were calculated with the AirQ+ software based on integrated exposure-response function recommended by the World Health Organization (WHO) and the UN using city-scale statistics of fatal disease cases that can be attributed to air pollution. It was determined that all cities in the Marmara Region study area exceeded the limit PM 2,5 values specied by the European Union (EU) in the years 2016, 2017 and 2018 while only Kocaeli and Tekirdağ were below the limit values in 2019. The limit values specied by the WHO were exceeded in all cities in each year. A total of 46.920 premature deaths attributed to the exceedance of WHO limit values were calculated for the years 2016, 2017, 2018 and 2019 with 11.895, 13.853, 11.748 and 9.429, respectively. Determining national limit values for the PM 2.5 parameter, which is among the most important factors of air pollution, and monitoring it in a sustainable manner using a sucient number of well-equipped stations is of great importance. This way, national, regional and urban action plans regarding the impact of air pollution on human health, as indicated by UN SDGs, can be prepared.


Introduction
Air pollution is of vital importance for many areas including energy, climate, transportation, trade, agriculture and biological diversity. In addition to posing risks for the environment, economy and food safety, its impact on human health is also critical (Pope et al., 1995;Rafaj et al., 2018). It has been revealed through the epidemiological studies conducted since the 1980s that the relationship between particulate pollution and mortality is consistent. In these studies, it was reported that air pollution at high concentrations has caused the death of hundreds of thousands of individuals in Europe and the United States (Ware et al., 1981;Katsouyanni et al., 1997;Dominici et al., 2003;Samoli et al., 2008;Fattore et al., 2011;Ghanbari Ghozikali et al., 2016). The negative effect of particulate matters, especially those with a grain diameter of less than 2.5 mm (PM 2.5 ), on human health was determined not only at high concentrations but also at low concentrations, and such matters are linked to respiratory morbidity and mortality (Pope et al., 1995;Sun et al., 2004;Bell, Peng and Dominici, 2006;Downs et al., 2007).
According to previous epidemiological studies, PM 2.5 has a stronger impact in terms of premature deaths and potential health effects from prolonged exposure compared to PM 10 (Hamanaka and Mutlu, 2018). It was reported that particulate matter of this size bypasses the defence mechanisms of the upper respiratory tract and affects all body systems by rst reaching the pulmonary alveoli and then the systemic circulation (Lee, Kim and Lee, 2014).
One of the important risk factors for particularly lung cancer (Vena, 1982) stroke (Hong et al., 2002), type II diabetes (Rao et al., 2015) and cardiovascular diseases (Yamamoto, Phalkey and Malik, 2014) is air pollution. The risk factor of air pollution varies at different levels of socio-economic development. For example, urban air pollution poses a greater risk factor in middle-income countries compared to highincome countries because signi cant improvements have been achieved in high-income countries in terms of controlling this risk with public health policies (WHO, 2009). Current calculations attribute 7 million premature deaths around the world to air pollution, and approximately 4.2 million of these result from outdoor air pollution (WHO, 2021). The economic cost of premature deaths attributed to outdoor particulate matter pollution and indoor air pollution to the European Nation in 2010 was approximately 1.5 trillion dollars (Prüss-Üstün et al., 2016). In a study conducted by the WHO on deaths attributed to outdoor air pollution for the year 2012, it was reported that approximately 36% of the deaths resulted from stroke while this rate was 36% for ischemic heart diseases, 14% for lung cancer, 8% for chronic obstructive lung diseases and 6% for acute lower respiratory tract infections (WHO, 2016a).
Findings obtained from epidemiological studies show that exposure to air pollution is essentially related to ve diseases depending on the age group. These can be listed as: i) Acute low respiratory infections (ALRI) in young children, ii) Cerebrovascular diseases (stroke) in adults, iii) Ischaemic heart diseases (IHD) in adults, iv) Chronic obstructive pulmonary disease (COPD) in adults and v) Lung cancer in adults (WHO, 2016c). Various epidemiological studies determined a relationship between the daily increase in suspended particulate matter concentration and the increase in mortality rate, while also revealing the relationship between mortality and the number of hospitalizations due to respiratory and cardiovascular diseases (Miri et al., 2016).
For Turkey, which is a party to the SDGs, the Marmara Region is the most signi cant geographical region in terms of both urban and industrial activities. While the six metropolitan municipalities surrounding the Sea of Marmara host some of the most important industrial production points in Turkey, they also represent settlements with the highest urbanization rate and population density. Despite the fact that air pollution resulting from the use of coal for heating purposes in the Marmara Region has decreased over the years, the increase in population-related urban activities and industrial activities keep air pollution among signi cant health threats.
In the present study, the change in mortality rate attributed to particulate matter air pollution in the Marmara Region over the years (2016-2019) has been examined and the cities were compared. In this comparison, it was decided to use the AirQ + model proposed by the WHO. The results obtained were compared with those gathered using the same model for different cities around the world, and interpreted.

Materials And Methods
In the present study on the average level of air pollution and the impact of this level on human health, the years 2016, 2017, 2018 and 2019 are evaluated. 2020 data have been excluded in order to avoid the misleading interferences of the impact of the COVID-19 pandemic on air quality. The ow diagram in Figure 1 shows the steps followed in the study and the path followed in the assessment of the ndings.

Study area and air quality monitoring stations
The Marmara Region consists of 11 cities located around the Sea of Marmara in the northwest of Turkey where approximately 31% of the country's population lives. There are 6 metropolitan municipalities in the Marmara Region where urban and industrial activity is relatively denser. By population, these cities can be listed as İstanbul (15.462.452 (TURKSTAT, 2021). The surface area of the 6 metropolitan municipalities selected as the study area is approximately 45.000 km 2 and these cities represent the region as they account for approximately 68% of its terrestrial area. Air quality measurements in Turkey are conducted regionally by the Clean Air Centres a liated with the Ministry of Environment and Urbanization. As of 2019, there are 5 air quality measurement stations a liated with the Marmara Clean Air Center in Balıkesir, 6 in Bursa, 12 in İstanbul, 12 in Kocaeli, 4 in Sakarya and 5 in Tekirdağ. Figure 2 shows the location of the study area and the air quality measurement stations located in the area.
The data of the stations that made measurements for more than 75% of the days in a calendar year were regarded as suitable within the scope of the present study. The daily average PM 2.5 concentrations of the stations that met these conditions were averaged, and the annual average PM 2.5 concentration was calculated by considering all of the resulting daily average PM 2.5 concentrations of each city. Within the scope of the present study, the PM 2.5 concentrations for the air quality measurement stations where no PM 2.5 measurement was performed were calculated by converting the PM 10 concentrations measured using the conversion coe cient (the conversion coe cient speci ed for Turkey by the WHO is 0,666128). Table 1 shows the number of stations that performed direct PM 2.5 measurements in the cities included in the present study, and the number of stations where the PM 2.5 concentration was calculated by converting PM 10, by years.  (UN, 1991(UN, , 2001(UN, , 2021WHO, 2016c). In the present study, the multi-year natural death rates for the cities were calculated to cover the period between 2009 and 2018. Deaths that could not technically be associated with air pollution, such as tra c accidents, murders and suicides, were not included in the total number of deaths (Revised Mortality Underlying Cause Instructions, 2014). Age-speci c mortality rates and statistics related to the causes of death were obtained from the Turkish Statistical Institute by city.
Formula 1 is used to calculate the age-speci c mortality rate (ASMR) for adult individuals aged over 30 in each city.
For this calculation, the age-speci c mortality rate statistics published by the Turkish Statistical Institute and the Address-Based Population Registration System were used. N represents the number of natural deaths over the age of 30 while P represents the population in the same age group. It was acknowledged that all individuals aged over 30 were exposed to the average level of the measured or converted PM 2.5 concentrations.

Air Q model and calculation of mortality rate attributed air pollution
Version 2.0 o the AirQ+ software, which was developed by the World Health Organization (WHO) Regional O ce for Europe and the Bonn O ce of the European Centre for Environment & Human Health to calculate the mortality and morbidity of air pollution at a given settlement with reference to the General Principles of Health Risk Assessment of Air Pollution (WHO, 2016b), was used in the present study. Using AirQ+, the number of deaths attributed to long-term PM 2.5 exposure exceeding the ambient air level of 10 mg/m 3 as speci ed in air quality guidelines was calculated.
The Integrated Exposure-Response (IER) function is used in the prediction of the burden of disease attributed to long-term exposure to PM 2.5 in ambient air. (Burnett et al., 2014). Despite the presence of certain limitations, the IER approach provides a bene cial framework to evaluate the consistency of ndings and estimate the burden of disease attributed to air pollution (Pope, Cohen and Burnett, 2018).
The AirQ+ tool was selected as it uses this function, which enables the effects of air pollution originating from different sources to be evaluated together.
The measurement of the health effects of exposure to air-pollutant materials is based on the concept of Attributable Proportion (AP) and the calculation of Years of Life Loss (YLL). AP, i.e. the fraction of the health outcome, exposure of a speci c population within a certain time period can be calculated using Formula 2 (Ezzati et al., 2002): Within this equation, RR(c) represents the relative risk for the health outcome of exposure in category c while P(c) represents the population rate in category c of exposure.
Relative Risk (RR) is a function used to predict the size of the relationship between exposure and disease. Relative risks originating from air pollution are generally modelled with log-linear function (Formula 3) (Ezzati et al., 2002;Lim et al., 2012).
In the equation, X represents the pollutant concentration in mg/m 3 . X 0 shows the cut-off or counterfactual concentration. In the present study, the annual average cut-off value of 10mg/m 3 (as speci ed in the WHO air quality guidelines) was used for X 0 in the AirQ+ model. Finally, b represents the change in Relative Risk (RR) for a change of one unit at a concentration of X.

Evaluation of air quality measurements
In the cities included in the present study, data was obtained from a total of 32 stations in 2016 while this number was 42 in 2019. Table 2 shows the annual PM 2.5 concentrations recorded in the cities and the descriptive statistics parameters Mean±Standard Deviation, median and Interquartile Range -IQR values. occurring in Bursa, particularly in winter months, essentially originates from industrialization, and the signi cant pollution factor is coal, which is used for domestic heating and as an industrial fuel. It is known that the ring road passing through the north of the city creates tra c-related pollution based on the prevailing wind direction. The compaction of the particulate matter carried from the north through this type of wind also affects the air quality in Bursa negatively (MoEU, 2020a). Agricultural areas are located in the north of the city and the dust originating from agricultural activities is regarded as among the important factors of inner-city air pollution. The use of coal with low calori c value is the leading air pollution factor in Sakarya. In relation to this, the low air circulation in the city centre, particularly during winter months, causes pollutants to get concentrated in the low parts of the city. Additionally, the number of facilities that were granted environmental permits in terms of emission control was 94 in the year 2016 while this number has increased to 220 in the year 2020, which is an important fact in terms of industrial air pollution (MoEU, 2020b).
The average PM 2.5 concentrations of the cities, recorded each day of the year between 2016 and 2019, were examined and Figure 3 shows the daily concentration accumulations. In all cities, accumulations were observed in the second quarter in the daily PM 2.5 concentrations recorded.
Since the cities included in the present study were subjected to multiple comparisons within themselves (4 years), whether there was a statistically signi cant difference in pairwise comparisons between the years (two years) was examined using the Bonferroni Correction. According to this, signi cant differences were determined in all years except the dual years of in Balıkesir, 2016in Bursa, 2016in İstanbul, 2016in Kocaeli, 2016in Sakarya, and 2018 in Tekirdağ. Table 3 shows the Bonferroni Correction results regarding the pairwise comparisons of the cities' annual PM 2.5 concentrations.  According to this, it was observed that the annual seasonal differences between the sunny/cloudy periods in all cities were statistically signi cant (p<0.05) with the exception of the sunny and cloudy periods in Tekirdağ in 2018 (t: 1.846, p:0.066), and that the air quality during the sunny periods was better compared to the cloudy period. Based on this, Figure 5 shows the seasonal differences in the sunny/cloudy periods of the cities by year.
When the change in the average PM 2.5 concentrations of the cities over 48 months is examined, the result obtained is in parallel with the Unpaired T-Test performed for seasonal change. Figure 6 shows the change in the monthly average PM 2.5 values of the cities recorded over 48 months and the increase in the PM 2.5 concentration during cloudy months.
When the monthly average PM 2.5 concentrations of the cities included in the present study are considered, the highest monthly average in 2016 was measured during the cloudy months while the lowest average was measured during the sunny months with the exception of Balıkesir and Tekirdağ. In 2017, the highest monthly average was measured during the cloudy months with the exception of Tekirdağ while the lowest monthly average was measured during the sunny months in all cities. In 2018, the highest monthly average was measured during the cloudy months with the exception of İstanbul and Kocaeli while the lowest monthly average was measured, once again, during the cloudy months with the exception of Bursa and Sakarya. In 2019, the highest monthly average was measured during the cloudy months while the lowest monthly average was measured during the sunny months. Table 4 shows the months in which the highest and lowest PM 2.5 values were recorded in the cities between 2016 and 2019, along with mean±sd values. Meteorological factors are important in terms of understanding the change in air pollutant concentrations and the change in their distribution. In addition to other air pollutants, the tendencies of PM are also affected by topography and meteorological variables (Rodríguez, Dupont-Courtade and Oueslati, 2016;Mohtar et al., 2018;Liang and Gong, 2020). In the study conducted by Bodor et al. (2020) in the Transylvania region of Romania, the highest monthly average PM concentration was recorded in February at four times the annual average. Additionally, a negative correlation was observed between PM and temperature (Bodor et al., 2020).
It was revealed by Cichowichz et al. (2017) that air pollutant concentrations in Poland are high during the winter months and heating seasons and low during the summer months. Mohtar et al. (2018) revealed that PM concentrations in Malaysia were high during these seasons in relation to burning biomass and coal with low calori c value. In the present study, statistically signi cant differences were determined in the average PM 2.5 concentrations during the seasons categories by the number of sunny and cloudy days, and it was revealed that PM-related air pollution was highest in all cities during November, December, January and February.

Natural mortality rates and air pollution attributable deaths
In the cities included in the present study, the natural mortality rates in adults aged over 30 were calculated in units of 100 thousand, taking into account all mortality cases except those in relation to external injuries and poisoning. Figure 7 shows the mortality rates in the cities.
The natural mortality rate in adults (30+) during the years 2013-2014 was recorded as 876 in Paris, 871 in Santiago, the capital of Chile, during 2013-2016, and 934 in the Lombardia region of Italy during 2013(WHO, 2018. In the present study, the city with the highest mortality rate was recorded as Balıkesir while Istanbul had the lowest mortality rate.
As a fraction of the health outcome, the exposure of the population aged over 30 in the cities was calculated in the form of attributable proportion and shown in Figure 8.
In the attributable proportion, upper and lower rates were calculated by considering a con dence interval of 95% (95% CI). Based on this, among the four years considered, a maximum attributable proportion of 26.32% is calculated in Bursa for the year 2017. In the study conducted by Toscano et al. (2021) covering the period between 2003 and 2019, it is reported that the sharp decrease in the NO 2 concentration in Italy's Campania region and the relative decrease in the PM 10 and PM 2.5 concentrations also caused a decrease in the attributable proportion parameter.
The AirQ model was operated with the Integrated Exposure-Response Function reference using the agebased (30+) natural mortality rates and annual average PM 2.5 concentrations in the cities. Based on this, Figure 9 shows the premature deaths caused by long-term exposure to PM 2.5.
In relation to the statistically signi cant decrease (p<0.05) in PM 2.5 concentrations between 2016 and 2019, it is thought that the decreasing trend in the number of premature deaths in all cities is statistically signi cant. It is thought that premature deaths may be prevented if air quality is improved to a point where the annual average PM 2.5 concentration in cities is reduced to the annual average PM 2.5 concentration threshold speci ed in the WHO guidelines. Pope & Dockery (2006) revealed that an increase of 10 μg/m 3 in the 24-hour average PM 2.5 concentration increases the relative risk of daily cardiovascular mortality by 0.6-1.1%. Additionally, Garcia et al. (2016) revealed a signi cant relationship between PM 2.5 exposure and deaths caused by cardiovascular diseases, ischemic heart diseases and cardiopulmonary diseases in California. Approximately 39% of all deaths in Turkey in the years 2016, 2017 and 2018 resulted from vascular diseases. Approximately 65% of all vascular diseases in Turkey result from ischemic heart disease and other heart diseases (TURKSTAT, 2019). In the examination performed within the scope of the present study, it was calculated that 46% of the deaths in Balıkesir resulted from vascular diseases while this rate was 44% in Bursa and 45% in Sakarya. This rate was calculated as 34% for İstanbul, 34% for Kocaeli and 40% for Tekirdağ. It is thought that the average PM 2.5 concentrations measured, the mortality rates in the cities and national cause of death statistics are in a signi cant relationship. Figure 10 shows the mortality rate (per 100 thousand) of preventable premature deaths in similar to attributable proportion. This indicator is of importance as it allows for a depopulated assessment on the health effects of air pollution. Although there was a decrease in the mortality rates in Bursa, Balıkesir and Sakarya between 2016 and 2019, it is observed that mortality rate in these cities is clearly higher compared to other cities.
In In a study conducted in 11 of the largest cities in Poland, which is the country with the highest annual average PM 2.5 concentration among European Union countries, it was reported that the annual average PM 2.5 concentration ranged between 14.3 and 52.5 μg/m 3 . The number of deaths attributed to long-term exposure to PM 2.5 concentration was estimated to be 9.6 to 22.8 of every 100.000 lung cancer deaths and 48.6 to 136.6 in cases resulting from heart and lung diseases (Badyda, Grellier and Dabrowiecki, 2017). It is estimated that in 25 cities in Iran where PM 2.5 pollution is common, the annual average PM 2.5 concentrations are 1.5-6.1 times higher than the WHO threshold value and, consequently, 13.321 of all annual deaths recorded in these cities are classi ed as premature death attributed to air pollution (Hadei et al., 2020). In the present study, a decrease was observed in the overall natural mortality rates between the years 2016 and 2019. In 2019, the cases in Balıkesir, Bursa and Sakarya were calculated as above 100 (per 100 thousand), higher than İstanbul, Kocaeli and Tekirdağ.
The impact of sharp increases and decreases in air pollutant concentrations on morbidity and mortality has been subject to many studies. In the study conducted by Turner et al. (2011), it is stated that an increase of 10 μg/m 3 in the PM 2.5 level has caused an increase of 15-27% in deaths from cancer. Mahesvaran et al. (2010) examined the impact of air pollution on post-stroke survival rates with 3.320 patients and 1.856 death cases in London. Based on this, it was revealed that an increase of 10 μg/m 3 in the concentration of particulate matter with a diameter of less than 10 μm increased the probability of poststroke mortality by 52%. In the study conducted by Brook et al. (2010) on the relationship between cardiovascular diseases and air pollution, it was reported that a decrease of 10 μg/m 3 in PM 2.5 concentrations increased life expectancy by 0.61 ± 0.20 years. In the study conducted by Pala et al. (2021) in all of Turkey, it was stated that 44.617 cases of premature death in 2018 could have been prevented had the air quality been at the annual average PM 2.5 concentration level speci ed by the WHO.
In this context, in the present study, 11.895 (95% CI,13.853 (95% CI,11.748 (95% CI, 7.838-15.243) and 9.429 (95% CI, 6.250-12.303) premature deaths were attributed to excessive air pollution values in 2016, 2017, 2018 and 2019, respectively. It is thought that such premature deaths can be prevented if a level of air quality that is in accordance with the WHO limits can be established.

Conclusion
With the modern industrialization and motorization around the world, the emission of anthropogenic aerosols is sharply increasing, which brings along not only regional but also global effects by means of long-range transportation. Aerosols are globally regarded for their potential effects on climate change and human health (Sun et al., 2004).  Lee et al. (2014) emphasized that PM 2.5 affected the whole body system. For this reason, directly measuring the PM 2.5 parameter rather than calculating with any conversion coe cient is of importance in terms of monitoring this parameter in a sustainable manner. Additionally, the relatively small number of stations that directly measure the PM 2.5 parameter is thought to be related to the fact that no national-scale limit values were speci ed regarding the PM 2.5 parameter and that it was not included in air quality regulations. In this respect, specifying limit values for PM 2.5 in the Regulation on Air Quality Assessment and Management within the framework of WHO documents and EU regulations is thought to be of importance for the effective monitoring of its effects on human health.
It is known that almost all parameters related to air pollution originate from incineration processes, industrial activities and tra c. It is also known that atmospheric convection and meteorological conditions play a role in the distribution and effects of air pollutants. For this reason, it is necessary to limit the use of coal with low calori c value in order to reduce air pollution caused by heating. It may be effective for gas distribution rms and local governments to cooperate in providing incentives and support mechanisms for the transition from coal to natural gas. It would be bene cial to conduct studies towards adopting sustainable production and consumption principles with the aim of reducing industrial air pollution. These studies can focus on the main subjects of technological developments, sustainability of administrative sanctions and the effective use of economic tools. In order to reduce tra c-based emissions in urban areas, it would be bene cial to promote walking and cycling short distances through infrastructure improvements. It is also thought that it would be effective to introduce administrative regulations promoting the use of electric or hybrid vehicles rather than those using fossil fuels. Additionally, providing digital communication tools that can recommend alternative routes to drivers and notify pedestrians on what time to go out in order to reduce exposure to tra c-based emissions as a public service is thought to be bene cial from many aspects. Finally, it is of importance to plan residential areas in consideration of air pollution dynamics, meteorological conditions and topographical factors.

Acknowledgement
We would like to thank Mr. Emre Yaşar who supported us in checking the necessary statistical analyses within the study, the Marmara Clean Air Centre who provided the location information of the Air Quality Measurement Stations in the Marmara Region. Also, the authors special thank to Prof. Dr. Burcu Onat and Prof. Dr. Ülkü Alver Şahin for their valuable assessments and comments.
Authors' contributions ACK: literature search, study design, data acquisition, data analysis, statistical analysis, AirQ model results, manuscript preparation, corresponding author NS: concept, design, data interpretation, de nition of intellectual content, manuscript editing, 'guarantor'

Funding
There is not any funding body in the design of the study and collection, analysis, and interpretation of data.

Data availability
The datasets generated and/or analysed during the current study are available in the;  Number of days exceeding the WHO threshold for daily average PM2.5 Seasonal (winter/summer) differences for mean PM2,5 in cities by year Changes in the estimated number of attributable cases by years Figure 10