Key points are not available for this paper at this time.
Buildings play a critical role in understanding settlement patterns and are essential for crisis management, urban planning, energy efficiency, and multi-hazard risk assessment. To address the need for accessible global building data, we introduce a dataset containing 2.7 billion building footprints classified using the building taxonomy of the Global Earthquake Model. By conflating the AI-derived Google Open Buildings and the Microsoft Global ML Building Footprints datasets, and the crowd-sourced OpenStreetMap, we created the most detailed and extensive building dataset to date. This conflation helps balancing out the completeness bias in OpenStreetMap in which mapping is most complete in countries with high human developing index. We validated occupancy types and building height estimation using Kullback-Leibler divergence across specific cities, and through cadaster data from Slovenia and Greece, revealing that, while some misclassifications occur due to definitional differences or data limitations, the dataset overall provides reliable and valuable building information at global scale. Examples to use the data are to identify vulnerabilities of buildings for natural hazards or to model population distributions.
Oostwegel et al. (Mon,) studied this question.