Gianluca Boo, Edith Darin, and our own LCDS Doug Leasure together with colleagues from WorldPop, Columbia University, UCLA, and the DRC Census Bureau used building footprints from high resolution satellite imagery to produce up-to-date population estimates broken down by age and sex across thousands of 100m grid squares—not much larger than a football pitch—throughout five provinces of the DRC.
Up-to-date population figures are critical for planning government services, implementing public health campaigns, and even preparing for the next census. However, the Democratic Republic of the Congo (DRC) has not completed a national population and housing census since 1984, relying instead on demographic projections which are often imprecise when applied over a period of several decades.
The DigitizeAfrica project is mapping every building in sub-Saharan Africa using Ecopia.ai's artificial intelligence to extract building features from Maxar's high resolution satellite imagery. To infer demographic characteristics from these state-of-the-art building footprints, the authors conducted household surveys across five provinces of the DRC to collect basic information about the sizes and age-sex composition of households from a sample of locations. They linked these traditional survey data with satellite-derived building footprints and other geospatial data using a bespoke Bayesian statistical modelling framework to produce unbiased population estimates with realistic estimates of uncertainty.
Although no model-based estimation can replace a proper population census, this innovative paper provides a new pathway for producing reliable up-to-date population figures and highlights the value of space-based Earth observations for population estimation in locations where a census is challenging.
To read the full paper, see here.
Citation of paper:
Boo, G., E. Darin, D.R. Leasure, et al. (2022). High-resolution population estimation using household survey data and building footprints, Nature Communications, 13: 1330.
