Skip to content

Copernicus DEM — introduction#

earthlens ships a first-class dem backend for global land elevation data. It targets the public Copernicus DEM GLO-30 (~30 m) and GLO-90 (~90 m) grids on the anonymous AWS Open Data buckets — no account, no SDK login, no API key.

For the hands-on walkthrough see Usage, the two dataset ids on the Available datasets page, and the rendered API on the Reference page.

Why the dem backend exists#

Copernicus DEM is already reachable through several shipped backends — gee, stac (Planetary Computer / Earth Search / CDSE), earthdata — but every one of those paths needs an account (a Google Earth Engine service account, an Earthdata Login, or a STAC signer). This backend earns its place by offering the same tiles with zero credentials.

The Copernicus DEM AWS buckets are anonymous and requester-free:

  • s3://copernicus-dem-30m — GLO-30 (~30 m), region eu-central-1.
  • s3://copernicus-dem-90m — GLO-90 (~90 m), region eu-central-1.

Every design choice below protects that account-free path: no new SDK (reuses the shipped [s3] unsigned boto3 substrate — the same client goes / nwm / radar ride on), no signer, and no decode step that would drag in a heavy stack.

How it works#

A request maps to the 1° x 1° COG tiles whose SW corner lies inside the bounding box, one anonymous head_object per candidate, and one whole- file download_file per tile that exists. download() returns the list[Path] of downloaded COGs, in bbox row-major order.

from earthlens.earthlens import EarthLens

paths = EarthLens(
    data_source="dem",
    dataset="cop-dem-glo-30",
    lat_lim=[30.2, 30.8],
    lon_lim=[31.2, 31.8],
    path="dem_out",
).download()
# -> [PosixPath('dem_out/Copernicus_DSM_COG_10_N30_00_E031_00_DEM.tif')]

What earthlens does — and doesn't do#

  • Fetches raw COG tiles. earthlens hands the tiles back as they came from the bucket. Cropping the bbox down to the exact AOI, mosaicking neighbouring tiles into a single raster, reprojecting to a UTM zone, and rendering a hillshade are all pyramids' job — its COG reader (pyramids.Dataset.read_file) plus .crop / .to_crs and pyramids.dataset.merge.merge_rasters (for a multi-tile mosaic) already do this. earthlens does not import rasterio, gdal, osgeo, or xarray.

  • 1° tile granularity. The buckets serve one COG per 1° x 1° tile, so a coastal bbox spanning several tiles yields one file per tile. To merge them, hand the list of paths to pyramids.dataset.merge.merge_rasters(paths, out_path).

  • Ocean tiles are absent — logged and skipped. Copernicus DEM ships no tile over open ocean. A bbox that crosses a coast produces a ragged coverage; the missing tiles are logged at WARNING and the download continues. See the accompanying notebook for a Nile Delta walkthrough that surfaces this cleanly.

  • DEM is time-invariant, so aggregate= is rejected. A DEM has no time axis to reduce over. Passing a non-None aggregate= raises NotImplementedError.

  • Land only. For sea-floor bathymetry (GEBCO / ETOPO1) use the separate bathymetry backend.

Two ways to select the resolution#

The dataset= kwarg picks the bucket:

  • dataset="cop-dem-glo-30" (default) → the 30 m bucket, tile-name token 10.
  • dataset="cop-dem-glo-90" → the 90 m bucket, tile-name token 30.

Or use one of the discoverability aliases (copernicus-dem, cop-dem, elevation) — they all route to the same backend.

Attribution#

Copernicus DEM is free to use with attribution:

© ESA — produced from Copernicus data.

The per-tile INFO/eula_F.pdf (shipped alongside every tile in the bucket) is the ESA end-user licence. earthlens does not emit a runtime LicenseWarning for Copernicus DEM — please honour the attribution in downstream publications.