Automated zone design for the spatial representation of population

Any mapping or analysis involving population data aggregated to geographical areas (zones) is subject to the modifiable areal unit problem (MAUP), namely that observed patterns are influenced by both the scale (size) and aggregation (boundary placement) of zones. This has important implications, not only for researchers undertaking analyses, but also for national statistical organisations needing to decide what zoning system(s) (sets of zones) to employ to release population data. This thesis explores, enhances and extends automated zone design techniques for the spatial representation of population. It addresses three key themes: the use of automated zone design to explore the modifiable areal unit problem; automated maintenance and adaptation of existing zoning systems; and the importance of building blocks in automated zone design. It shows that administrative geographies are not necessarily the most appropriate zones for exploring health and environment relationships and that automated zone design can be used to explore sensitivity of results to the MAUP. It demonstrates that automated procedures can be used to update existing zoning systems which have become unfit for purpose due to population change, and also to modify existing zones to make them suitable for representation of other phenomena such as workplace statistics. It provides evidence that building blocks are a crucial, but under-rated, component of the zone design process and concludes that all zone design should be based on sound theoretical reasoning and a clear conceptualisation of the phenomena and processes being represented. The thesis has had a demonstrable impact on the fields of environment and health, and official population statistics; its concepts and methods have been employed by a diverse range of researchers, as well as by the Office for National Statistics to create 2011 Census output areas and workplace zones for England and Wales