Settlement classification from multi-scale spatial patterns using georeferenced administrative data on buildings in low and middle-income countries

Model-based population estimates continue to improve with more refined input datasets and more
sophisticated modelling techniques to better reflect population distributions. Urban environments,
however, remain a challenge for population models. The heterogeneous landscape of cities, from highrises
to industrial estates, and mixed use buildings makes population predictions difficult. Most
population data from a census or survey reflect the residential population, or a "nighttime" population
when people are at home, yet for many cities, accurate classifications identifying where these residential
areas are remain lacking.
During the day, we know that city population swell with workers, students, shoppers, and other
commuters, yet these populations are generally not accounted for in population estimates. Therefore,
improved settlement classification models and building usage are a critical input for improving
population estimates. Accurate urban population models are more important than ever as the majority of
global population and population growth are occurring in urban areas.
This research is to improve on the existing modelling technique of settlement classification using building
footprint and other building characteristics e.g. use, height etc. with a focus on low and middle income
settings which typically have growing urban centres and often lack up-to-date information on settlement
types or neighbourhoods. This has been identified as very helpful in identifying slums and informal
settlements, areas of potential health risk, and population density. Also, as urban settlements continue to
grow, understanding their morphology, both within and between cities, becomes key for planning,
delivering, and monitoring projects in support of sustainable development.