Assessing Personal PM2.5 Exposure: A Method Leveraging Movement Routes and Activity Space Information

Abstract

Previous studies have consistently shown a significant correlation between air pollution, particularly PM2.5, and various diseases, as well as increased mortality rates. This study introduces a novel approach for predicting time-specific indoor PM2.5 exposure by incorporating individual movement routes and activity spaces using GPS tracking data and a time–activity diary. The models were trained separately for each hour of the day (e.g., 0:00–0:59, 1:00–1:59) with a total of 24 models. Their applicability was demonstrated with data gathered from actual participants. Additionally, automated machine learning (AutoML) was utilized to optimize prediction performance. The results revealed that the proposed model effectively accounted for the influence of outdoor PM2.5 concentrations and meteorological factors. The performance varied across different indoor environments, with the subway station model showing the highest prediction accuracy. Future research should address these uncertainties, adopt more advanced modeling techniques, and consider diverse indoor variables for a comprehensive understanding. The insights from this study could significantly enhance health risk assessments associated with fine particulate matter exposure.