A novel approach for dynamic population activity in urban-scale exposure estimates

Population exposure estimates are used in epidemiological studies to evaluate health risks associated with impacts of air pollution on human health. Traditional approaches in exposure modelling assume that air pollutants' concentrations at the residential address of the study population are representative of overall exposure. This approach is acknowledged to introducing bias in the quantification of human health effects, as individual and population-level mobility is non-existent. To sufficiently model population numbers for exposure estimates, the dynamic population activity (DPA) must be known. Information on DPA is mostly derived from national or municipal surveys and is scarce.

We developed a generic approach to model DPA integrating the Copernicus Urban Atlas land use and land cover product with literature based and microenvironment-specific diurnal activity data (Ramacher et al. 2019) and moreover taking into account gridded monthly day- and night-time populations as derived in the ENACT project (https://ghsl.jrc.ec.europa.eu/enact.php), while also considering indoor and outdoor environments. This approach produces maps with distribution of citizens in various microenvironments (MEs) and hours of the day. These maps can consequently be used to calculate population-level outdoor exposure when paired with consistent spatio-temporal air pollution concentration fields. In this study, we applied the generic DPA approach to the cities of Hamburg (DE) and Athens (GR). Hourly, urban-scale pollutant concentrations were produced by the Chemistry Transport Model (CTM) EPISODE-CityChem (Karl et al. 2019) driven by detailed local emission inventories, 4D meteorological fields and regional pollutant boundary conditions for 2015. Both the concentrations for NO_x, O₃ and PM_2.5 as well as the DPA maps for Hamburg and Athens were simulated on a 100 m horizontal resolution grid, and were then combined to calculate population exposure. We additionally used gridded population densities (based on residential addresses) to calculate population exposure, i.e. following a static population approach. We compared the exposures from the two approaches to capture the effect of a population moving in space and time. 

The presented approach to account for dynamic instead of static population activity in urban population exposure calculations is beneficial for cities in European regions where relevant population data are missing. It is found that by taking into account movement of population through different urban environments as well as commuting per se, the overall exposure estimates are elevated when compared to a static approach. Furthermore, we have shown that by considering infiltration of outdoor concentrations to indoor environments there are substantial decreases in population exposure estimates. This approach and the implications on exposure estimates is believed to be of interest to air pollution health studies.