Enhancing areal interpolation frameworks through dasymetric refinement to create consistent population estimates across censuses

Abstract

ABSTRACTTo assess micro-scale population dynamics effectively, demographic variables should be available over temporally consistent small area units. However, fine-resolution census boundaries often change between survey years. This research advances areal interpolation methods with dasymetric refinement to create accurate consistent population estimates in 1990 and 2000 (source zones) within tract boundaries of the 2010 census (target zones) for five demographically distinct counties in the US. Three levels of dasymetric refinement of source and target zones are evaluated. First, residential parcels are used as a binary ancillary variable prior to regular areal interpolation methods. Second, Expectation Maximization (EM) and its data-extended version leverage housing types of residential parcels as a related ancillary variable. Finally, a third refinement strategy to mitigate the overestimation effect of large residential parcels in rural areas uses road buffers and developed land cover classes. Results suggest the effectiveness of all three levels of dasymetric refinement in reducing estimation errors. They provide a first insight into the potential accuracy improvement achievable in varying geographic and demographic settings but also through the combination of different refinement strategies in parts of a study area. Such improved consistent population estimates are the basis for advanced spatio-temporal demographic research.