Multi-layer multi-class dasymetric mapping to estimate population distribution

Abstract

The spatial patterns of population distribution are very important information for most regional planning and management decisions. But the socioeconomic data are usually published in areal aggregated format due to privacy concerns. Although choropleth maps are used extensively to display spatial distributions of these areal aggregated data, patterns may be distorted due to assumptions of homogeneous distributions and the modifiable areal unit problem. Most human activity, including population distribution, is spatially heterogeneous due to variations in topography and regional development. A multi-layer multi-class dasymetric (MLMCD) framework was proposed in this study to better redistribute the regionally aggregated population statistics into smaller areal units and reveal more realistic spatial population distribution pattern. The Taipei metropolitan area in Taiwan was used as a case study area to demonstrate the disaggregation ability of the proposed framework and the improvements to the traditional binary or multi-class dasymetric method. Assorted data, including remote sensing images, land use zoning, topography, transportation and accessibility to facilities were introduced in different layers to improve the redistribution of aggregated regional population data. The concept of multi-layer multi-class dasymetric modeling is both useful and flexible. Different levels of accuracy in this population redistribution process can be achieved depending on data and budget availabilities and the needs for different data usage purposes.