ML-based geographic sampling frames miss transitory populations in fragile regions
A hybrid sampling-frame method that combines machine-learning building footprints with satellite imagery, tested in displacement-affected Iraqi communities, that reduces the systematic undercounting of displaced and mobile populations.