Establishing relationships between parcel data and land cover for demographic small area estimation

Abstract

Dasymetric and other small area estimation methods often use land-cover data in order to refine the spatial resolution of population data. The attribute tables of these land-cover data, however, are often related only weakly to population distribution. Recent studies have examined the use of parcel data, but parcel data are not available in all places. Thus, it becomes useful to identify the links between parcel data and land-cover data so that land-cover data can be used where parcel data are not available. This article identifies and validates the relationships between land-cover and parcel data to improve small area estimation. Establishing this link between parcel data and land cover makes it possible to estimate the distribution of building types within each land-cover type. This article develops a method to do this and illustrates its general use with a case study for Boulder County, CO. A ground truth layer combines census block group data, individual parcel records, and land cover. Target zones constructed by homogeneous patches of land cover found in census block group units permit identification of the distribution of building types within small areas. Land cover is enriched with a simple pattern metric called the inner dimension metric in order to indicate how far inside of a developed region each developed land pixel is located. Poisson generalized linear models establish the relationship between parcel building type and land-cover type. The results suggest strong and significant relationships between residential building counts and land-cover data. This research will improve selection of related variables for dasymetric models to create small area population estimates of census housing attributes.