Abstract
Exposure to air pollution in the form of fine particulate matter (PM2.5) is known to cause diseases and cancers. Consequently, the public are increasingly seeking health warnings associated with levels of PM2.5 using mobile phone applications and websites. Often, these existing platforms provide one-size-fits-all guidance, not incorporating user specific personal preferences. This study demonstrates an innovative approach using Bayesian methods to support personalised decision making for air quality. We present a novel hierarchical spatio-temporal model for city air quality that includes buildings as barriers and captures covariate information. Detailed high resolution PM2.5 data from a single mobile air quality sensor is used to train the model, which is fit using R-INLA to facilitate computation at operational timescales. A method for eliciting multi-attribute utility for individual journeys within a city is then given, providing the user with Bayes-optimal journey decision support. As a proof-of-concept, the methodology is demonstrated using a set of journeys and air quality data collected in Brisbane city centre, Australia.
Highlights
Air pollution is a major environmental risk to health, estimated to cause 4.2 million premature deaths worldwide annually (World Health Organization, Accessed: 27-11-2018), and 3000 deaths each year in Australia (Australian Institute of Health and Welfare, 2016)
62 A Bayesian Decision Framework for Personalised Cycle Route Selection applications. These websites and applications provide generic, one-sizefits-all air quality measurements or warnings for a given location, either based on a single nearby air quality monitoring station, a numerical weather prediction model, or a network of stationary monitoring stations used in combination with a dispersion model, land use regression model and/or satellite data
Rather than assuming a stationary spatial field our Integrated Nested Laplace Approximation (INLA) air quality model incorporates an adaptation of the barrier model of Bakka et al (2018), originally developed to model coastlines, to better represent the way in which high-rise buildings block the flow of air within the city, causing non-stationarity
Summary
In response to the growing recognition of mobile air quality monitors, Del Sarto et al (2016) developed a hierarchical Bayesian spatiotemporal modelling approach for representing such data This model uses a three level hierarchy in which the mobile observations are modelled in the first stage of the hierarchy, and a latent spatio-temporal process, defined on a discrete space-time grid, is considered in the following stages, capturing covariate information. Rather than assuming a stationary spatial field our INLA air quality model incorporates an adaptation of the barrier model of Bakka et al (2018), originally developed to model coastlines, to better represent the way in which high-rise buildings block the flow of air within the city, causing non-stationarity Following this we explore approaches for equating exposure to PM2.5 along a given route to the associated impact on health, the first decision-relevant journey attribute.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
