3-D AIR POLLUTION ESTIMATION USING A HYBRID SPATIAL MODEL: A CASE STUDY OF ZHUNAN-MIAOLI AREA, TAIWAN

Abstract

The rapid global urbanization has transformed cityscapes, giving rise to iconic skyscrapers that define modern cities. However, alongside this urban evolution, a pressing concern arises –the air quality within these towering urban environments. Fine particulate matter, known as PM2.5, poses a grave threat to human health and the environment. These tiny particles, measuring 2.5 micrometers or less, can penetrate deep into the human respiratory system, posing severe health risks. Due to the limitations of traditional land-use regression models in estimating the variation of air pollution with altitude, this study employs a novel hybrid spatial model to assess the three-dimensional distribution of PM2.5 in the atmosphere.We employ a comprehensive methodology, integrating diverse datasets and advanced modelling techniques, to uncover significant findings. Our analysis reveals the non-uniform nature of PM2.5 distribution, both horizontally and vertically. Variable selection identifies key factors influencing PM2.5 levels, including Broadleaf Forest, Carbon Monoxide (CO), and Height. Our ensemble model demonstrates robust performance, with Gradient Boosting Regression (GBR) and Random Forest Regression (RFR) exhibiting superior predictive capabilities. This study provides valuable insights into the complex interplay of environmental factors affecting PM2.5 concentrations in high-rise urban environments, emphasizing the need for targeted air quality management strategies considering both horizontal and vertical variations.