Giving dimension to point locations: urban density profiling using population surface models

Recently, there has been a proliferation of `point datasets' recording demographic and socio-economic conditions at point locations in geographical space. These datasets have yielded useful information about population densities and land use at micro-scales although their analysis requires dimension to be ascribed to point data when density is defined as a quantity of population per area of land. Density estimation techniques applied to point datasets ascribe area to point locations but the amount of area given is modifiable. In this paper we consider population surface modelling (PSM) as a density estimation technique that transforms vector point data to a near-continuous raster surface of cells. The search window is revealed as the parameter that most controls the model output and extent of the space-filling from points. The effects are shown of altering this (and other) PSM parameters on estimates of the total residential areas of four UK settlements at various threshold densities and using Ordnance Survey® (UK) Code-Point® data. The outputs from PSM are shown to be modified and determined by the scale of analysis. Yet, it is the ability to consider in combination the measures of population density at multiple scales that offers potential for modelling different types of land use.