Abstract

Population Mapping for Humanitarian Relief in Haiti and Pakistan
Track: Census and Statistical
Authors: Derek Azar

Knowledge of population distribution is an essential part of human geography. Census data are typically only available at coarse resolution in the countries of greatest interest for humanitarian relief efforts and operational planning. The U.S. Census Bureau has developed a novel disaggregation method. The method was developed for a project to create a high-resolution population map of Haiti and further refined in a similar project covering Pakistan. We use image classification methods that rely on highly automated machine learning and on an observed linear relationship between built-up area and population density. For both countries, a gridded population map was produced with census population disaggregated down to 100-meter grid squares. The high resolution of the gridded map allows for neighborhood level analysis with population estimates comparable to or better than existing global population datasets. In cooperation with the NGA, Census created websites for data dissemination that provide easy public data access.