What this measures
Sites come from OpenStreetMap: anything tagged telecom=data_center, building=data_center, lifecycle variants (construction:, proposed:, disused:, abandoned:, was:, demolished:), and construction or industrial land whose name says "data center" in several languages.
Status follows the tag: construction prefixes or a construction landuse mean under construction; proposed/planned prefixes or named brownfield/greenfield land mean planned; disused, abandoned, was, removed, demolished mean abandoned. Each site shows the exact tag that decided it.
Footprint is the geodesic area of the mapped outline. A building inside a mapped campus is shown but counted once, under the campus. Sites mapped only as a point have no measured area; the calculator can give them the median footprint of their status.
Share of land divides footprint by the region's area from Natural Earth 1:50m boundaries (inland lakes included, so shares are slightly low). The multiplier models land acquired beyond the buildings.
Green space uses Sentinel-2 Level-2A (10 m) from the Earth Search catalogue on AWS Open Data. Inside the footprint, in the growing season of each year, cloud and shadow pixels are removed with the scene classification layer, the Normalized Difference Vegetation Index is taken per pixel (median of up to two scenes), and the share of pixels at or above 0.3 is the green fraction. Green area = fraction × footprint. "Measure green space now" runs the same method in your browser for any site.
Imagery comparison swipes between yearly Sentinel-2 cloudless mosaics (EOX, 2016 onward) or Esri's latest high-resolution imagery.
Where SatDump fits
SatDump decodes weather and Earth-observation satellites you receive yourself (NOAA and MetOp AVHRR, Meteor MSU-MR, MODIS…). Those pixels are about 1 km wide, so they cannot outline a data center; they do show the vegetation state of the surrounding region on the day of your pass, independent of any company's servers. The repository ships a SatDump preset that writes a linear vegetation index as a GeoTIFF, and pipeline/satdump_ingest.py turns each pass into a map layer plus a per-site regional reading.
Known gaps
- OpenStreetMap coverage is uneven: well mapped in Europe and North America, sparse elsewhere; many hyperscale campuses are mapped late or only as points.
- "Abandoned" depends on mappers updating tags; cancelled projects that never broke ground are rarely mapped.
- The vegetation index sees living plants, not land-use designation: lawns and landscaping count as green; dry-season grassland may not.
- 2017 Level-2A coverage outside Europe and North America is patchy; pick a later "before" year if a site reports no scene.
Sources and licences
Sites © OpenStreetMap contributors, ODbL. Sentinel-2 data: Copernicus, via Element 84 Earth Search on AWS Open Data. Mosaics: Sentinel-2 cloudless by EOX IT Services GmbH (2016 CC BY 4.0, later years CC BY-NC-SA 4.0). Boundaries: Natural Earth (public domain). Libraries: Leaflet, Leaflet.markercluster, geotiff.js. Code: MIT.