Current assessments link poor air quality to around seven million premature deaths worldwide annually. However, exposure studies, often utilising measurements from stationary outdoor instruments from sparse monitoring networks, cannot capture spatial heterogeneity or the fact that people spend significant fractions of their time indoors. This failure to assess the actual pollution exposure individuals receive leads to inaccuracies in pollution-health associations, potentially masking the factors that drive the observed health responses, resulting in misinformed policies.

To address these limitations, a portable personal air quality monitor (PAM) was developed, allowing for the assessment of actual personal exposure to key pollutants: CO, NO, NO₂, O₃, and PM_2.5, as well as providing location (GPS) and other parameters for time-activity assessment.

The work in this thesis develops a framework, which, when applied to large scale fieldwork studies, is capable of disaggregating personal exposure by source and linking it to health parameters for hundreds of participants. At the core of the framework is a methodology for apportioning personal exposure into pollution generated by indoor sources and pollution generated by outdoor sources.

This apportionment is achieved by employing a mass-balance model and estimating values of ventilation rates, indoor loss rates and indoor source characteristics, collectively referred to as “exposure determinants”.

The framework was applied to data from the AIRLESS project, which involved the deployment of PAMs to 250 residents of Beijing and the surrounding area. Personal exposure to NO₂, O₃ and PM_2.5 was found to be lower than that inferred from measurements from stationary outdoor reference instruments, suggestive of indoor losses for these pollutants.

The results show differences between indoor-generated and outdoor-generated exposures, for example, 55% of participants’ exposure to CO was from indoor sources, compared with 30% of PM_2.5. Apportioned exposure metrics, for example indoor- and outdoor- generated CO, while the same molecule, may be proxies for different mixtures of pollutants, which may have different health impacts.

As expected, home ventilation rates were higher in the summer than in the winter, and the overall mean ventilation rate was estimated to be 3.12 hr^-1, which is comparable to values found in the literature. Knowledge of the seasonal and demographic variability of exposure determinants will be crucial in the future modelling of total personal exposure at the population scale.

This thesis concludes with the construction of a Linear Mixed Effects Model (LMEM), linking the novel exposure metrics and estimated exposure determinants to a health marker, in this case Peak Expiratory Flow (PEF). While the associations with personal exposure and PEF appear minimal in this study (concerns about the accuracy of self-reported PEF as an indicator are raised), it is expected that this framework will be of significant value when extended to directly examine the effects of the novel exposure metrics and estimated exposure determinants on other health parameters. This will provide insights into the source-related health effects of air pollution to drive more effective environmental policy.