The geographical data science of population flows

Abstract

This research seeks to address the issues of spatial structure and counter intuitive findings in modelling population movements. Previous research based on interaction data for census administrative boundaries indicates urban – rural patterns in the propensity to travel which are exasperated in peripheral areas. These findings are considered to be an effect of the spatial structure of zonings and thus an unsuitability of the census geography in reflecting the phenomena of movement. This research, through the example of the case study of Northern Ireland, aims to create a geography through regionalization which is better suited to the geography of commuting and to apply such functional fit regions to spatial interaction modelling to better understand and evaluate regional and local commuter patterns with respect to key demographic variables. A five stage open approach is proposed with geovisualization, demographic data linkage, regionalization, spatial interaction modelling, and factor analysis to produce robust results which draw out the underlying tendencies of journey to work movements using fit for purpose zones.

Based on functional fit regions, findings indicate a number of key demographic tendencies in the propensity to travel. These include the effect of affluence on the willingness to travel further; the effect of urban dwelling as a deterrent to travel further for work; the influence of rurality on the propensity to travel greater distances; the impact of a stronger employment market area on the reluctance to travel and finally the effect of increased religious differences between home and workplace on a reduced willingness to travel further. The spatial structure of the geography of modelling has a profound influence on the modelled outputs. This research demonstrates that functional fit for purpose regions created through regionalization provide a superior fit for modelling the phenomena of commuting or wider movements. The methodological framework for this research is to provide an open source, transparent and reproducible approach to geovisualization, region building and modelling through Python notably using PySAL library and Max-p regionalization and R scripting through the Factanal package. An open approach allows for applicability and development in wider fields such as transport planning, demographic policy and mobile data intelligence.

Date of Award	Jul 2020
Original language	English
Awarding Institution	Queen's University Belfast
Sponsors	Queen's University Belfast & Northern Ireland Department for Infrastructure
Supervisor	Ian Shuttleworth (Supervisor) & Jennifer McKinley (Supervisor)