GEDTM30: a global bare-earth Digital Terrain Model at ~30 m (1 arc-second) resolution, produced by the OpenGeoHub Foundation.
Source & method: A true Digital Terrain Model (DTM) — it represents the bare-earth surface, excluding buildings and vegetation. Unlike the Copernicus DEM GLO-30 surface model, GEDTM30 is a terrain product. It is built with a global-to-local machine-learning (Random Forest) data-fusion of existing global DEMs corrected against nearly 30 billion high-quality ICESat-2 and GEDI spaceborne-lidar elevation points. Two layers are published: the ensemble mean terrain height and the ensemble standard deviation (a per-pixel uncertainty map).
Resolution: 30 m (1 arc-second) nominal, on a single global mosaic (one item, not tiled).
Coordinate reference: Geographic WGS84 (EPSG:4326); elevations in meters on the EGM2008 geoid (EPSG:3855).
Hosting: Assets reference the public, anonymously readable Cloud-Optimized GeoTIFFs on s3.opengeohub.org in place — Darly does not re-host or copy the pixels.
Coverage: Global (excluding Antarctica).
Version: v1.2 (current). New non-breaking versions overwrite this mosaic in place; a breaking change (different resolution/CRS/datum) would get its own collection.
Documentation:
Access
The catalog is browsable without a key. Elevation lookups need a free API key, sent as an X-API-Key header; see the quickstart.
curl "https://api.darly.io/collections/glo-ogh-gedtm-30m/items?limit=5"curl -H "X-API-Key: $DARLY_API_KEY" \
"https://api.darly.io/v1/elevation/point?lat=48.8566&lon=2.3522&collections=glo-ogh-gedtm-30m"from pystac_client import Client
catalog = Client.open("https://api.darly.io")
search = catalog.search(collections=["glo-ogh-gedtm-30m"], bbox=[-180, -65, 180, 85], max_items=5)
for item in search.items():
print(item.id, item.assets["dtm"].href if "dtm" in item.assets else item.assets["dsm"].href)Providers
- OpenGeoHub Foundation — producer, processor, host, licensor
- Darly — processor
Citation
Ho, Y., Hengl, T., et al. (2025). GEDTM30: global ensemble digital terrain model at 30 m and derived multiscale terrain variables. PeerJ. Data: https://doi.org/10.5281/zenodo.14900181