Using Classified and Unclassified Land Cover Data to Estimate the Footprint of Human Settlement

Accurate, up-to-date maps of and georeferenced data about human population distribution are essential for meeting the United Nations Sustainable Development Goals progress measures, for supporting real-time crisis mapping and response efforts, and for performing many demographic and economic analyses. In December 2014, Esri published the initial version of the World Population Estimate (WPE) image service to ArcGIS Online. The service represents a dasymetric footprint of human settlement at 250-meter resolution. It is global and contains an estimate of the 2013 population for each populated cell. In 2016 Esri published an additional image service representing the earth’s population in 2015 at 162-meter resolution. Esri’s WPE is produced by combining classified land cover data indicating predominantly built-up or agricultural locations with Landsat8 Panchromatic imagery, road intersections, and known populated places. The model detects where settlement is likely to exist beyond the areas classified as predominantly built up. The result is a global dasymetric raster surface of the footprint of settlement with a score of the likelihood of human settlement for each cell of the footprint. Population data are apportioned to this settlement likelihood surface by overlaying population counts in polygons representing census enumeration units or political units representing population surveys. This paper presents the method developed at Esri for producing the estimate of settlement likelihood.