Effectiveness of Remotely Snsed Built Areas in Constraining Gridded Population Estimates

Authors: Philip Reed*, University of Louisville, Forrest Stevens , Advisor, Andrea Gaughan, Advisor
Topics: Population Geography, Spatial Analysis & Modeling, Geographic Information Science and Systems
Keywords: Population Geography, GIS, Population Modeling, WorldPop, Facebook, Malawi, Rwanda, Haiti, Thailand, Nepal, Madagascar
Session Type: Paper
Day: 4/14/2018
Start / End Time: 8:00 AM / 9:40 AM
Room: Napoleon A3, Sheraton, 3rd Floor
Presentation File: No File Uploaded


Human population distribution is critical knowledge that informs many disciplines. Generally, high-resolution census data and detailed ancillary inputs are components of the most accurate models. However, how different remotely sensed built areas may be used to inform and constrain gridded population techniques has yet to be fully explored. This study assesses the effectiveness of the World Settlement Footprint, Global Human Settlement Layer, and Facebook delineated built areas to dasymetrically constrain gridded population estimates for six countries. The resulting models include binary dasymetric redistribution, random forest (RF) with dasymetric component, and a hybrid of the previous two approaches. While model accuracy regionally varies, RF and hybrid models outperformed binary dasymetric models. Of built covariates, Facebook contributed the most to RF modeling, and produced the lowest comparative error when applied as a distributional mask. The accuracy of each model may be explained by the underlying quality of census data, ancillary inputs, and built areas.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login