2:00 PM - 2:15 PM
[013] Analysis of the influence of urban forms on the seasonal variation of particulate matter using machine learning approach: A case study of Seoul, Korea
Keywords:Particulate matter, Machine Learning, Smart Seoul Data of Things (S-DoT), Seasonal Variation
Efforts to mitigate and understand the public health effects of air pollution have an interesting history. Among various pollutants, atmospheric fine particulate matter with diameter less than 2.5 μm (PM 2.5) and 10 μm (PM 10) are known to endanger human health that causes respiratory, cardiovascular, and metabolic diseases. The aim of this research is to generate a high-performance machine learning model to predict PM 2.5 and PM 10 levels in Seoul, South Korea. The analysis was conducted through developing a random forest model that integrates 849 air quality sensor data across the city, using two- year period (April 2020- May 2022) hourly data of PM 2.5 and PM 10 measurements, urban form factors such as transportation, land use, density, and other geographical features. A spatially and temporally varying predictors were included in the random forest model to predict the concentration of fine particles. Results show that seasonal spatial patterns in particulate matter are more apparent in winter and spring than in summer and fall seasons. The dominant PM concertation in winter was significantly influenced by various topographic varibles such as elevation. Different emission sources such as highway and traffic volume were also found to be associated with poor air quality specially in colder period. However, spatial lag term and meteorological conditions related to the site vegetation determined PM concentration for fall and summer seasons. This research demonstrates the importance of using machine learning algorithms in predicting seasonal air quality and the typology developed from this study can help identify different strategies by season to reduce PM concentration in Seoul.