Bayesian building count estimation using ground surveys and satellite-derived data

This study integrates ground surveys with satellite-derived datasets in a Bayesian hierarchical model to improve the accuracy of gridded building count data in Kasaï-Oriental, Democratic Republic of the Congo. Ground surveys enumerated freestanding structures in 213 small areas, which were fit using gridded building count from satellite-derived building footprints according to different settlement classes. The model substantially reduced bias, inaccuracy, and imprecision compared to the original building footprint data. Posterior estimates consisted of probabilistic gridded building count for the entire province. The study highlights the benefits of Bayesian assessments of satellite-derived datasets in data-scarce contexts.