Abstract:
These days air pollution has become a critical environmental issue all over the world. Among
all population groups, schoolchildren are a vulnerable cohort due to their immature defence
mechanisms and relatively high inhalation rate. Early studies of schoolchildren and air
pollution concentrated mainly on their exposure at home or at school. There is an urgent need
to estimate schoolchildren’s exposure during their daily commutes from home to school that
comprises the main exposure in their daily routine. However, this estimation is still challenging
at population scales because of the difficulty in predicting the local variation of air pollutants
and modelling students’ commute routes from home to school. Privacy concerns force the use
of simulated home addresses that may introduce bias to the estimated dose/exposure. It remains
unknown whether exposure injustice exists during walking commutes or not. This doctoral
thesis aims to address these research gaps and concerns by developing an adaptable modelling
methodology to quantify population-scale students’ dose of ambient NO2 (largely from vehicle
emissions in Auckland) during their walking to school at different scales, and develop a piece
of open-source software for automating land use regression (LUR) modelling and air pollution
mapping.
The results showed that the developed software PyLUR was efficient and versatile in LUR
modelling and mapping, and could be used elsewhere so long as the required input data are
available. The newly developed multi-scale LUR model (R2: 0.85) performed slightly better
than the UK model (R2: 0.83) and the standard LUR model (R2: 0.80), and significantly better
than the IDW interpolation (R2: 0.65) and the OK interpolation (R2: 0.69). No single predictor
variable was common to all the scale models, and this revealed the varying importance of the
same predictor variables at different scales. Of all the walking students studied, only 17.48%
of them in the whole Auckland could find an alternative lowest-dose route. The portion was
higher (26%) in central Auckland because of its better road network connectivity. For only
about 30% of the students, a 1% increase in route length was associated with a > 1% reduction
in dosage if using the alternative lowest-dose route. Greater benefits by walking the lowestdose
routes were gained in suburban Auckland (a less-polluted area) than in central Auckland,
which highlighted the importance of taking the alternative lowest-dose route, especially for
those students whose shortest-distance routes overlapped with or ran parallel to an arterial road.
The use of simulated home addresses underestimated route length and reduced dosage of the
alternative routes by up to a quarter in comparison with the results derived from the observed home addresses. Exposure inequality among the studied students existed at a minor level, but
patterns of environmental justice (EJ) in central Auckland were opposite to those in suburban
Auckland. This thesis quantified students’ dose of NO2 during walking commutes using rich
observed data at population scales for the first time, and revealed the bias introduced to the
modelled dose by using simulated home addresses. The findings of this doctoral thesis could
provide policymakers with scientific evidence for air pollution prevention and choices for
schoolchildren to minimize their daily commute exposure.