| Abstract: | Mismatching sets of boundaries present a persistent problem in spatial analysis for many different applications. Dasymetric mapping techniques can be employed to estimate population characteristics of small areas that do not correspond to census enumeration boundaries. Several types of ancillary data have been used in dasymetric mapping but performance is often limited by their relatively coarse resolution and moderate correspondence to actual population counts. The current research examines the performance of using high resolution ancillary data in the form of individual address point datasets which represent the locations of all addressable units within a jurisdiction. The performance of address points was compared with several other techniques, including areal weighting, land cover, imperviousness, road density and nighttime lights. Datasets from 16 counties in Ohio were used in the analysis, reflecting a range of different population densities. For each technique the ancillary data sources were employed to estimate census block group population counts using census tracts as source zones, and the results were compared with the known block group population counts. Results indicate that address points perform significantly better compared with other types of ancillary data. The overall error for all block groups (n = 683) using address points is 4.9% compared with 10.8% for imperviousness, 11.6% for land cover, 13.3% for road density, 18.6% for nighttime lights and 21.2% for areal weighting. Using only residential address points rather than all types of locations further reduces this error to 4.2%. Analysis of the spatial patterns in the relative performance of the various techniques revealed that address points perform particularly well in low density rural areas, which typically present challenges for traditional dasymetric mapping techniques using land cover datasets. These results provide very strong support for the use of address points for small area population estimates. Current developments in the growing availability of address point datasets and the implications for spatial demographic analyses are discussed. |
