Effect of workplace mobility on air pollution exposure and its inequality in the UK
A large number of epidemiological studies have identified air pollution as a major risk to human health. Short-term and long-term exposures to air pollutants such as PM2.5, NO2 and O3 cause cardiovascular and respiratory diseases, cancer and other adverse health effects. These lead to decreased quality of life, increased hospital visits and premature mortality. Due to a high spatial and temporal variability of air pollution exposure, exposure inequalities exist within the society. Published studies suggest that it is often the most deprived and susceptible who are disproportionately exposed to the highest concentrations of some of the most ubiquitous air pollutants. However, most epidemiological and exposure studies do not take into account the spatio-temporal variability of air pollutants and population mobility within their assessments which is likely to lead to exposure misclassification and, consequently, a bias in the associated health effects. Several modelling approaches have been developed to improve estimates of population exposure. Either statistical or deterministic models are now commonly used to predict air pollution concentrations. Deterministic models include atmospheric chemistry transport models (ACTMs), which tend to be used for larger study areas on a regional or higher scale, and Gaussian dispersion models, which on a local – urban – scale are able to predict concentrations at very high spatial resolutions. The aim of this thesis is to investigate how workplace-related population mobility and spatio-temporal variability of air pollution affect population exposure and its inequality in the UK. Firstly, the effect of exposure to ambient air pollution at the place of work or study on overall population exposure in the UK is examined using publicly available data from Census 2011. The analysis is conducted for the whole of the UK (England, Wales, Scotland and Northern Ireland), and separately for Scotland only. The residential population distribution and daytime population distribution data are combined with concentration fields of key air pollutants (PM2.5, NO2 and O3) generated by the EMEP4UK atmospheric chemistry transport model at relatively high spatial (approximately 1.5 km × 2 km) and temporal (hourly) resolutions to calculate population exposure of stay-at-home ‘static’ population and a ‘dynamic’ population which spends a proportion of time on weekdays at the place of work or study. The calculated exposures of static and dynamic populations are compared, and sensitivity studies of different working hours of the dynamic population are conducted. The highest difference between dynamic and static population exposures is observed for NO2 (0.28 µg m-3 or 2.0% increase in the UK, 0.29 µg m-3 or 23.1% increase in Scotland) for working hours between 08:00 and 18:00. The calculated differences for PM2.5 and O3 are much smaller. Whilst at the population level the exposure difference is small, a case study using virtual individuals suggests a potential large variation between individuals. Secondly, the exposure of dynamic population and population subgroups to air pollution is examined in a case study of the Central Belt of Scotland region. Additionally, the two largest and demographically contrasting urban areas within the region – Glasgow and Edinburgh – are considered separately. For the analysis, anonymised personal data of the participants of the Scottish Longitudinal Study (SLS), which is a representative sample of the Scottish population, are linked at the postcode unit level with air pollution concentrations generated by EMEP4UK (approximately 0.8 km × 1.4 km spatial resolution). The SLS participants are stratified by age, ethnicity and socio-economic status (SES) for the population subgroup exposure assessment. Exposures at residential address and the place of work or study are considered using three different work pattern scenarios and the results are compared with exposures of the ‘static’ population. Exposure gradients are observed across all demographic characteristics. Young people between 21 and 30 years of age tend to have the highest exposure to NO2 and PM2.5, and lowest to O3; however, those aged 31 to 50 tend to be most affected by inclusion of exposure at workplace. The patterns for SES and ethnicity are complex and study area specific; however, people in the two least deprived deciles consistently have the lowest residential and residential-workplace exposure to NO2 and PM2.5 but tend to see the highest increase in exposure due to workplace mobility. Overall, including exposure at place of work in exposure estimates tends to alleviate some of the exposure inequalities observed in the static population exposure assessments. Thirdly, the effect of using different air pollution models on ‘dynamic’ population exposure estimates and its inequalities is investigated. The city of Edinburgh is chosen as the study area. Two models are considered: EMEP4UK and a Gaussian plume dispersion model, ADMS Urban. Detailed traffic emission data for all major and some minor roads in the city, and gridded emissions from other sources, are used in the ADMS Urban modelling. The model output is verified against available monitoring data. Differences in modelled output are observed between the models which are subsequently translated into differences in population exposure estimates. The effect of workplace exposure on overall population exposure to NO2 is larger for the ADMS-Urban model than for the EMEP4UK model; however, the magnitude is still very small (≤ 1.6%). For O¬3 and PM2.5, the effects are smaller and largely comparable between the models. With some notable exceptions, both models show similar patterns in both exposure inequality and the influence of workplace mobility on it. Lastly, the overall findings and their implication for assessment of exposure and exposure inequality are discussed. This work suggests that inclusion of the place of work or study in exposure assessments makes only a small difference for population-scale burden assessment, particularly for those pollutants with secondary contributions such as O¬3 and PM2.5. This conclusion is not particularly sensitive to the atmospheric chemistry/dispersion model used.