Appearance
News
Coming up: v25.01 - a new version of the dataset
21 May 2026
We are working on a new version of OpenBuildingMap. The next release will include more information per building: level of urbanity, roof type, date of construction, and improved building coverage. Stay tuned!
Our paper is out
27 October 2025
Where are all the buildings on our planet? With all the satellite imagery and open data available, this seems like a straightforward question — but we don't have an answer yet.
Our new paper in Nature Scientific Data (Nature Portfolio) about OpenBuildingMap is our own attempt to answer this question: a global building footprint dataset, including height, floorspace and occupancy type. Its main sources are OpenStreetMap, GHSL, Open Buildings and the Global ML Building Footprints. You may find good use for the dataset as an urban planner, in disaster management, or for climate analysis.
Read the article: From Footprints to Functions: A Comprehensive Global and Semantic Building Footprint Dataset
Download the building footprints: https://doi.org/10.5880/GFZ.LKUT.2025.002
EGU General Assembly 2024 — From shelters to skyscrapers
19 April 2024
We are presenting the next step of our work at EGU in Vienna this week: combining OSM's rich tagging scheme with the sheer scale of AI-derived building datasets to get a global picture of not just where buildings are, but what they are.
OSM now has close to 600 million buildings, but the Google Open Buildings dataset has 1.8 billion and Microsoft's 1.3 billion — neither complete, but both enormous. The catch is that neither Google nor Microsoft include building attributes like occupancy type. OSM does, through land use tags, amenity tags and points of interest. So we use those OSM tags to classify building footprints from all three datasets and deduplicate them onto a ~100×100 m grid, giving priority to OSM, then Google, then Microsoft. The result: 3.7 billion buildings, with occupancy types derived wherever we can.
Why does this matter for disaster management? Knowing the location and type of buildings improves risk forecasting, helps estimate where people are during a disaster, and makes post-disaster damage assessments more actionable. Schools and hospitals become findable. Population distribution becomes estimable.
Full abstract: Oostwegel et al. (2024), EGU24-14888, DOI: 10.5194/egusphere-egu24-14888
State of the Map Europe 2023 — OSM building completeness
5 November 2023
We are presenting our work on global OSM building completeness at the OSM Science conference in Antwerp this week. OSM is growing fast — around five million buildings per month — but the coverage is far from uniform. Some regions are nearly complete, while others are barely touched. We want to map exactly that gap, and keep it up to date.
Our approach uses Google Open Buildings and the Global Human Settlement Layer (GHSL) as reference datasets. For each small tile on a global Quadtree grid (~100×100 m at European latitudes), we compare the built area in OSM against the reference. Google's data has the most accurate footprints where available; for the rest of the world we fall back to GHSL. The assessment is dynamic: every minute, changes to OSM buildings trigger a recalculation of the affected tiles, so the completeness map is always in sync with the latest edits.
The work points to something practical: knowing which tiles are incomplete — and which of those have little recent mapping activity — makes it much easier to prioritise where help is needed, whether for everyday OSM contributions or for humanitarian mapping after a disaster.
Full paper: Oostwegel et al. (2023), DOI: 10.5281/zenodo.10443307
EGU General Assembly 2023 — Global completeness from remote sensing
28 April 2023
We are presenting our work on global OSM building completeness at EGU in Vienna this week, together with colleagues from GFZ and LiveEO. OSM currently has around 530 million building footprints — somewhere between a quarter and a half of all buildings on Earth, depending on how you estimate the global total. A truly global completeness view does not yet exist, and that is what we are building.
We use GHSL to fill that gap: by comparing the built-up area in each grid cell to the OSM building footprints, we estimate how complete OSM is, calibrated against areas we have manually checked. It is not ground truth — GHSL picks up infrastructure and industrial areas as well as buildings, and has its own false positives and negatives — but it gives a useful, honest reflection of where OSM stands globally and where the most effort is still needed.
Full abstract: Oostwegel et al. (2023), EGU23-13160, DOI: 10.5194/egusphere-egu23-13160