Impacts of Defining Residential and Non-Residential Areas for Global Population Modeling

Keywords: population, residential, land use maps, points of interest (POI), LandScan
Session Type: Virtual Paper Abstract
Day: Sunday
Session Start / End Time: 2/27/2022 02:00 PM (Eastern Time (US & Canada)) - 2/27/2022 03:20 PM (Eastern Time (US & Canada))
Room: Virtual 68

Abstract

High-resolution global population distribution models are traditionally developed using a top-down approach where underlying ancillary data influence the disseminations across a multitude of different spaces. With the ever increasing availability of spatially explicit data, like building footprints and open source land use maps, many of these models are now incorporating both, top-down and bottom-up methods. However, some high-resolution spatial delineation data still lack the information necessary to apply proper weighting techniques to distribute populations to the correct spaces in and around built up areas.
This paper explores the impacts of defining residential and non-residential areas in 3 major cities and why knowing what type of activities occupy individual spaces is just as important as having attribute-less spatial fidelity. Specifically, a global Points of Interest (POI) dataset was used to create a binary output raster grid of two very specific activity spaces that have unique temporal activity signatures. These rasters were then compared against LandScan Global's 2020 dataset. Together, these data identified areas where traditional methods could be improved simply by applying different expected densities per spatial delineations. While there were caveats in the fact that LandScan Global is an ambient population raster (average over 24 hours) and the global POI dataset is user-generated content and not fully representative of all actual POIs, this approach still proves useful for scaling to the globe and flagging over and under represented areas for more accurate distributions of population dynamics.