Openfootprint-HH-sampling: a reproducible building-based sampling framework for spatially explicit household surveys in data-scarce urban environments

Reliable household sampling frames are frequently unavailable in rapidly growing secondary cities of sub-Saharan Africa, complicating spatially explicit household surveys. Although satellite-assisted methods have improved field logistics, few approaches combine automated frame construction from open geospatial data with systematic validation of how realized samples adhere to the intended spatial design. We developed OpenFootprint-HH-Sampling, a parameterized QGIS workflow that transforms Google Open Building Footprints into spatially explicit sampling frames. Implemented in Ziniaré and Kombissiri, Burkina Faso, the approach combined proportional stratified random sampling by administrative sector with GPS-based mobile navigation. Sampling performance was evaluated through sector-level allocation, exclusive nearest-neighbor matching, nearest-neighbor distance analysis, and Ripley’s L functions. Direct differences between theoretical and realized L(r) functions were quantified using ΔL(r), MAD, MAE, RMSE, correlation coefficients, and R² values. A total of 723 household interviews were completed during August-October 2024. Median displacement between theoretical and realized interview locations was 89.4 m in Ziniaré and 90.3 m in Kombissiri, with larger deviations in some peripheral sectors. Nearest-neighbor analysis showed preserved local spatial structure in 5 of 10. Ripley-based comparisons indicated that both theoretical and realized samples retained clustered spatial configuration, while ΔL(r)-based metrics revealed stronger preservation at the city level than at the sector level. At the city level, theoretical and realized L(r) functions showed very high shape similarity, with R² values of 0.980 in Ziniaré and 0.988 in Kombissiri. Open building footprint data can support reproducible, spatially explicit household sampling where conventional address systems or complete household registers are unavailable. The results demonstrate operational feasibility and substantial fidelity of the realized sample to the intended spatial design, particularly at the city scale, rather than demographic representativeness of the underlying household population. The workflow may be adapted for health, nutrition, disease surveillance, and demographic surveys, provided that context-specific field verification and household-level inclusion checks are incorporated where population inference is required.