Mapping Spatio-Temporal Patterns of Disabled People in emergencies: A Bayesian approach
Emergency management can greatly benefit from understanding the spatiotemporal
distribution of individual population groups as this will optimise the allocation of
resources and personnel needed in case of an emergency caused by a disaster. This is especially true for people with a disability as they tend to be overlooked by emergency officials. This is generally approached statically using census data,not taking into account the dynamics of disabled peoples concentrations throughout space-time as exhibited in large metropolitan areas such as London. Transport data collected by automatic fare collection methods (such as Transport for London's Oyster card scheme) combined with accessibility covariates (number of opportunities/destinations within an areal unit) have the potential of being a good source for describing the distribution of this concentration. The aim of this study is to explore these datasetsfor use within the scope as described above. The paper attempts to model the distribution using discrete spatio-temporal variation methods. More specifically, it uses Poisson spatio-temporal generalised linear models built within a Bayesian hierarchical modelling framework, ranging from simple to more complexones, while taking into account the spatio-temporal
interactions that emerge between the space-time units. The performance of the resulting models in terms of their ability to explain the effects of the covariates as well as predicting future disabled peoples counts were compared relative to each other using the deviance information criterion and posterior predictive check criterion. Analysis of the results revealed a distinct spatiotemporal pattern of disabled users for Oyster card datasets, which deviates from the transportation habits of the rest of population. The effect of the chosen covariates diminishes as model's complexity increases, giving rise to patterns that could potentially be explained by including sociological aspects in the models.