Authors: Nathan Trombley*, Oak Ridge Institute for Science and Education, Jessica Moehl, Oak Ridge National Laboratory, Eric Weber, Oak Ridge National Laboratory
Topics: Land Use, Geographic Information Science and Systems, Population Geography
Keywords: Small-area mapping, Dasymetric, Statistical modeling
Session Type: Paper
Start / End Time: 5:20 PM / 7:00 PM
Room: Napoleon C2, Sheraton 3rd Floor
Presentation File: No File Uploaded
Dasymetric mapping seeks to distribute population more accurately by spatially restricting populations to inhabited/inhabitable portions of observational units and/or by varying population density among different land classes. LandScan USA uses this approach by restricting population to building area detected from remotely sensed imagery, but also goes a step further by classifying each cell of building area in accordance with ancillary land use information from national parcel data. Modeling population density according to land use is critical. For instance, apartment buildings would likely have a higher density of residents than single family homes for any given area of building detection. This paper presents a modeling approach which assigns different densities to different land uses. For areas where the parcel data is insufficient, National Land Cover Database (NLCD) data is used to define the land use type of building detection. Furthermore, LiDAR data is incorporated for many cities, allowing the independent variables to be updated from two-dimensional building detection area to total building floor space. In the end, four regression models were created to explain the effect of land use on resident distribution reflecting each pairing of two vs three-dimensional building area and parcel vs NLCD based land use classes. By and large, the resultant coefficients followed intuition, but importantly quantify the relationships between different land uses. For instance, in the two-dimensional model, apartment building area had a density nearly 4 times greater than single family residences. These coefficients define the ratios at which population is distributed in a residential dasymetric model.