High-resolution population grid in the Conus using Microsoft building footprints: A feasibility study

Abstract

Better knowledge of where people live is of great importance for a wide range of studies, including disaster responses, public health, resource management, and urban planning. Given the increasing demand for population grid with improved quality, this study explores the feasibility of generating high-resolution (100m) population grids in the Conterminous U.S. (CONUS) using a total of 125 million building footprints recently released by Microsoft. Those building footprints were used to disaggregate census tract population of the latest ACS 5-year estimates (2013-2017). Land use dataset from the OpenStreetMap (OSM) was applied to trim raw buildings footprints by removing those that are not likely residential. Weighting scenarios were designed, with which a dasymetric model was applied to disaggregate the ACS census tract estimates into a 100m population grid product. The results suggest that building footprints as a weighting layer, particularly footprint size after trimming, outperforms other commonly used weighting layers and is able to capture the great heterogeneity of population distribution at the micro-level. This study provides valuable experience in developing high-resolution population grid products that can benefit a wide range of studies in need of spatially explicit population data.