Beyond residence: Rethinking environmental exposure assessments through human daily mobility

Accurate environmental exposure assessment is essential for understanding environment–health relationships. Epidemiological studies typically rely on home-based assessments, yet such assessments overlook individuals’ daily mobility and may therefore introduce exposure misclassification. Mobility-based assessments have been proposed to address this limitation by incorporating individuals’ daily movements across diverse environments. Despite their conceptual appeal, mobility-based assessments are still supported by limited and inconclusive evidence regarding their practical advantages over home-based assessments. This calls for further investigation into their operationalization, validity, and implementation. Accordingly, this thesis aims to advance mobility-based assessments toward more accurate and human-centered approaches. The first part of this thesis examined methodological uncertainties in the operationalization of mobility-based assessments. Chapter 2 evaluates whether the use of global positioning system (GPS) data introduces bias in environment–health associations through selective daily mobility bias (SDMB). By comparing absolute and relative exposures to snack and soft drink outlets along GPS-tracked and modeled commuting routes with dietary intakes, this study finds no evidence of SDMB, suggesting that the use of GPS data in this context is unlikely to introduce bias in environment-health associations. Chapter 3 assesses whether incorporating contextual information (e.g., travel modes and distance decay effects) affects exposure assessments across multiple environmental exposures. Travel modes substantially influenced PM2.5 and blue space exposures, while distance decay effects affect green space and noise exposures, underscoring that contextual information beyond GPS trajectories can meaningfully alter exposure estimates. The second part evaluated the validity of mobility-based assessments. Chapter 4 compares mobility-based and home-based estimates against personal measurements of air pollution. Mobility-based approaches improve assessment validity by 64% for PM2.5 and 21% for black carbon; however, these gains stem primarily from accounting for indoor–outdoor concentration differences rather than mobility data alone. These findings indicate that improving validity requires integrating contextual information rather than relying solely on mobility data. The third part of this thesis explored practical strategies for implementing mobility-based assessments. Chapter 5 demonstrated that short-term GPS data (7–14 days) can sufficiently represent long-term mobility (over months) for long-term exposure assessment. However, exposure estimates based on yearly and monthly maps differed substantially, highlighting that the temporal resolution of exposure surfaces should be carefully considered when assessing exposures over specific periods. Chapter 6 showed that low-frequency GPS data (1-25 min) are feasible and sufficient for long-term assessments. For passive exposures (air pollution, green space, noise), >80% of exposure is attributable to two main activity locations. In contrast, fast food exposure was mainly associated with in-transit periods. Chapter 7 synthesizes these findings, emphasizing that mobility data alone provides limited improvement in exposure assessment validity unless combined with contextual information. Exposures are shaped jointly by individuals’ spatial locations, behavioral contexts, and environmental interactions. Accordingly, this thesis advocates shifting from mobility-based toward context-based exposure assessments to enhance assessment accuracy. Practically, the results show that smartphone-based, short-term, low-frequency GPS data are sufficient for long-term assessments, and that activity location–based approaches offer a scalable and efficient pathway for advancing person-centered exposure assessment to population-level assessments.