Investigating the urban fabric - air pollution and local climate relationships in the Marmara Region and Istanbul

More than half of the world's rapidly growing population lives in cities, leading to the rapid expansion of urban areas. Urban expansion not only increases environmental pressures, but also creates areas where the effects of climate change are concentrated. This situation makes the effects of extreme weather events more pronounced. Heavy rainfall, heat waves and storms pose significant risks, especially in large cities, and require the reorganisation of cities to adapt to growing populations and environmental threats. It is important to consider human thermal comfort when designing urban areas, to improve air quality and to create cities that are resilient to sudden temperature changes. Factors such as population density, transport networks and industrial activities have a direct impact on urban climate dynamics, leading to problems such as the heat island effect and air pollution. While these problems make sustainable and equitable urban planning difficult, the impacts of poor air quality and thermal stress negatively affect the lives of city dwellers. In this context, planning studies on air pollution and urban climate change are crucial for more liveable cities. This thesis presents a systematic analysis of the relationship between land use and air pollution, starting from an upper scale investigation of the relationship between land use and air pollution, including seasonal effects, to a lower scale investigation of the relationship between local climate, urban texture and air pollutants In the first stage, within the scope of Istanbul-Gebze, Kocaeli-Yalova, Bursa and Çanakkale provinces in the Marmara Region, the differentiation of the effects of CORINE land use level-1 classes of artificial surfaces, agricultural areas and forest areas on aerosol index, NO2 and SO2 air pollutants obtained from Sentinel 5P/TROPOMI satellite data were examined by post-hoc ANOVA tests. The analyses were carried out separately at 5 km and 1 km resolution and as a result of the resolution comparisons, the 1 km scale was found appropriate for the analyses. In the second stage, a more detailed investigation was carried out in the Istanbul-Gebze region and the relationships between air pollutants and subclasses of artificial surfaces (continuous city, discontinuous city and industrial/commercial areas) were investigated. The results provided a more comprehensive understanding of the dynamics between local climate, urban structure and air pollutants. In the final stage of the study, the relationship between surface temperature and factors such as urban texture, air pollutants, NDVI, NDBI, elevation, distance to coast and distance to forest were analysed within the urban areas of Istanbul. In these analyses, correlation tests, decision trees, linear regression analysis and XGBoost regression model were used to evaluate the relationships in detail. According to the results of the thesis study, in the first stage, it was determined that the effects of different land use types such as artificial surfaces, agricultural areas and forest areas on air pollutants differ in Istanbul-Gebze, Kocaeli-Yalova, Bursa and Çanakkale regions. It was found that this situation may change periodically and seasonally. In urban areas in the Istanbul-Gebze Region, it was found that aerosol index and NO2 data showed significant differences between areas with continuous urban fabric and areas with discontinuous urban fabric for both years in all seasons except autumn 2021. Then, in the second stage, the relationship between air pollutants and urban texture was investigated with correlation tests by including road data. The results showed that aerosol index and NO2 air pollutants had the highest correlation with the weighted urban texture-road class. This class was followed by urban texture and road data respectively. The relationship between air pollutants and urban texture classes (discrete low-rise, discrete mid-rise, compact low/mid-rise) was analysed by taking road density into account. Post-hoc test results revealed that aerosol index and NO2 air pollution data showed statistically significant differences in compact low/mid-rise and high road density texture classes compared to discrete low-rise and discrete mid-rise texture classes. In the third stage, the effect of independent variables (season, NDVI, NDBI, urban texture, aerosol index, NO2, SO2, distance to coast, distance to forest, elevation) on surface temperature was modelled using XGBoost machine learning technique without any period distinction, and the degree and direction of the effect of these variables were evaluated by SHAP analysis. The findings obtained at this stage show that season has the greatest effect on surface temperature (importance ranking value, 0.6). The season is followed by NDBI with a value of 0.35, NDVI with a value of 0.23 and urban texture. When comparing the accuracy of the linear regression and XGBoost regression models used to model the local climate, the accuracy value (R2) of the linear regression model was calculated as 0.701, while this value was found to be 0.804 in the XGBoost regression model. These results show that XGBoost is more successful in modelling complex data relationships. Finally, as a result of analysing the relationships between local climate, urban texture and air pollutants, it was found that densification in the urban texture increases the surface temperature and also causes an increase in NO2 and aerosol index values. In addition, it was revealed that the increase in aerosol index and NO2 also increases the surface temperature. These findings clearly show that the effects of air pollutants and urban fabric should be taken into account in urban planning. The TROPOMI satellite used in this study has a higher spatial resolution than other air quality satellites, allowing it to be used in and contribute to urban climate and air pollution research studies. This thesis analyses the relationships between urban climate, urban texture and air pollutants, which have not been extensively researched. The findings of this study make an important contribution to the literature by revealing that the effects of air pollutants and urban fabric should be taken into account in the planning of sustainable and climate change-resilient cities.