WP8 - Mapping/modelling
of population
distribution
Population density in Liège computed using a census unit disaggregation method based on the 3D geometry and on the nature of buildings
What?
The determination of the population spatial distribution is essential to understand and solve several environmental problems. This is especially true in urban contexts where the population density can greatly vary from place to place according to the urban morphology and land use linked to socio-economic and historic specificities of urban zones. In combination with census data, medium and high resolution remote sensing classifications and post-classifications can be used improve the spatial resolution of population density estimation.
Why?
The accurate determination of the present-day and future population spatial distributions is essential to organize sewage network, to enhance the accessibility to commercial, industrial and residential area and make it possible a sustainable mobility, to design coherent emergency plans, to organize the health systems …
How?
Traditionally, the modelling of population spatial distribution is performed using both census data and remote sensing products (Donnay & Unwin 2001, Harvey, 2002, Herold et al., 2002 and 2003, Liu & Herold, 2007) . Land Cover classifications from WP 1 and 2 and Land Use maps from WP 4 and 5 of MAMUD will be used in this purpose. Several methodological approaches can be exploited to disaggregate the population of census units. Traditional methods are based on the proportion of each LC or LU class and the corresponding population. Generally a spatially constant population density per class is used in the cross product rule (Rule of Three). Another global approach uses a multiple linear regression approach to express the population in function of the proportion of the area covered by each class. Recent solutions take in consideration the spatial variation of the relationship between population and RS information using Geographically Weighted Regressions. Some researchers have also expressed the population density in function of spatial metrics.
Results
The challenge of this part of the MAMUD project is the application of these methods in very heterogeneous urban contexts which are more realistic contexts than the ones analysed in the majority of reference papers. Furthermore, a special attention will be paid on the usefulness of 3D information in the disaggregation process.
The output of this WP will be a model adjusted on the past data linking the population to LC/LU. This model will then be used to extrapolate the population distribution from the prediction of the MOLAND urban growth modelling calibrated and validated in WP 6.
