Integrating statistical and agent-based modelling for activity-based ambient air pollution exposure assessment

Assessment of long-term human exposure to spatiotemporally highly variable air pollution requires accounting for human space-time activity. Individual exposure and space-time track data are not available over large populations and for long periods and a modelling approach is required. However, activity-based exposure models face here challenges in setting up the model and overly-large computations. Aiming for long-term and large-population simulations, we propose an activity model which integrates statistical and agent-based modelling by treating mobility-related variables as random variables. Probability distributions for these variables are estimated or derived from mobility datasets containing observed activities. On top of the activity model, we implemented an exposure model. A case-study of exposure assessment was developed using hourly air pollution maps. The activity model can potentially integrate any mobility data and is thus applicable when limited time activity data is available at the individual level.