Dataset Viewer
The dataset viewer is not available for this dataset.
The JWT signature verification failed. Check the signing key and the algorithm.
Error code:   JWTInvalidSignature
Exception:    InvalidSignatureError
Message:      Signature verification failed
Traceback:    Traceback (most recent call last):
                File "/src/libs/libapi/src/libapi/jwt_token.py", line 286, in validate_jwt
                  decoded = jwt.decode(
                      jwt=token,
                  ...<2 lines>...
                      options=options,
                  )
                File "/usr/local/lib/python3.14/site-packages/jwt/api_jwt.py", line 368, in decode
                  decoded = self.decode_complete(
                      jwt,
                  ...<8 lines>...
                      leeway=leeway,
                  )
                File "/usr/local/lib/python3.14/site-packages/jwt/api_jwt.py", line 265, in decode_complete
                  decoded = self._jws.decode_complete(
                      jwt,
                  ...<3 lines>...
                      detached_payload=detached_payload,
                  )
                File "/usr/local/lib/python3.14/site-packages/jwt/api_jws.py", line 270, in decode_complete
                  self._verify_signature(
                  ~~~~~~~~~~~~~~~~~~~~~~^
                      signing_input,
                      ^^^^^^^^^^^^^^
                  ...<4 lines>...
                      options=merged_options,
                      ^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/jwt/api_jws.py", line 417, in _verify_signature
                  raise InvalidSignatureError("Signature verification failed")
              jwt.exceptions.InvalidSignatureError: Signature verification failed

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Copernicus GLO-30 Terrain — H3-Indexed Global Elevation

DOI

A global terrain-elevation dataset derived from the Copernicus GLO-30 Digital Elevation Model, re-published as H3-indexed Parquet so you can get terrain statistics with a hash lookup instead of managing raster tiles.

For any H3 cell you get the mean/min/max/stddev elevation of the ground inside it, how many source pixels backed that estimate, and the geoid undulation needed to convert to height-above-ground. Available at H3 resolutions 2 through 9 (continent-scale down to ~170 m hexes).

H3 resolution subdivision over the Colorado Rockies near Aspen

The H3 hierarchy in action — a cluster of res-4 cells over the Colorado Rockies subdividing to res 8; each hexagon splits into its 7 children per level. Aspen sits in the green valley near the center; heights are the dataset's elevation_mean.

▶ Fly around it in 3D — interactive viewer


What's in it

One row per H3 cell, at each published resolution. Schema is identical at every resolution:

Column Type Units Description
h3_cell uint64 H3 cell index (the primary key)
elevation_mean float32 metres Mean of all source pixels in the cell
elevation_max float32 metres Maximum source pixel (obstacle clearance)
elevation_min float32 metres Minimum source pixel (valleys, water)
elevation_stddev float32 metres Std-dev of source pixels (terrain ruggedness)
pixel_count uint32 Source pixels aggregated into this cell (confidence weight)
geoid_undulation float32 metres EGM2008 geoid height above the WGS84 ellipsoid (see Vertical datum)
  • Elevations are orthometric (height above the EGM2008 geoid — Copernicus GLO-30's native vertical datum), in metres.
  • Horizontal datum is WGS84. Cell geometry is standard H3.
  • Statistics are computed independently at each resolution directly from the ~30 m source pixels (not rolled up from a finer level), so each resolution is internally exact.

Resolutions

Res Hex edge Hex area Rows (land) Typical use
2 158 km 86,700 km² ~3.0 k Coarse regional
3 60 km 12,400 km² ~17 k Sub-state
4 22.6 km 1,770 km² ~110 k Metro region
5 8.5 km 253 km² ~740 k County-scale
6 3.2 km 36 km² ~5.1 M City district
7 1.22 km 5.16 km² ~35 M Airport area
8 460 m 0.74 km² ~246 M Within-airport / fine terrain
9 174 m 0.105 km² ~1.72 B Block / small parcel

How it's organized

res2/part-000.parquet                      # res 2–7: one file per resolution
res3/part-000.parquet
…
res7/part-000.parquet
res8/base=<0–121>/part-000.parquet         # res 8–9: sharded by H3 base cell
res9/base=<0–121>/part-000.parquet
  • One directory per resolution (res2/res9/).
  • Res 2–7 are a single Parquet file each.
  • Res 8 and 9 are sharded by H3 base cell (base=<0–121>/, ~111 shards). The shard key is pure integer math on the cell index — no H3 library call: base = (h3_cell >> 45) & 127. Read just the shards your cells fall in; it also prunes by row-group stats inside each shard (data is sorted by h3_cell). Routing works in plain SQL with no H3 extension: WHERE base = (h3_cell >> 45) & 127.
  • Rows within a file are ordered by h3_cell; files are Zstandard-compressed Parquet.
  • Provenance (source, Copernicus attribution, vertical/horizontal datum, partition key, schema version) is embedded in each Parquet footer.

Quick start

Query in place with DuckDB (no full download — reads only the file you need over HTTP):

INSTALL httpfs; LOAD httpfs;
INSTALL h3 FROM community; LOAD h3;

-- high terrain at res 6 (a single ~small file)
SELECT * FROM 'hf://datasets/whereyaheading/copernicus-glo30-h3-terrain/res6/part-000.parquet'
WHERE elevation_mean > 3000;

Look up the elevation at a point (Python):

import h3, duckdb

lat, lon, res = 39.2232, -106.8687, 8           # Aspen, CO
cell  = h3.latlng_to_cell(lat, lon, res)
shard = (cell >> 45) & 127                       # which shard — pure bit math, no H3 call
REPO  = "hf://datasets/whereyaheading/copernicus-glo30-h3-terrain"

row = duckdb.sql(f"""
    INSTALL httpfs; LOAD httpfs;
    SELECT * FROM '{REPO}/res8/base={shard}/*.parquet' WHERE h3_cell = {int(cell)}
""").fetchone()
print(row)   # (h3_cell, mean, max, min, stddev, pixel_count, geoid_undulation)

Download the whole thing (e.g. for bulk joins):

from huggingface_hub import snapshot_download
path = snapshot_download("whereyaheading/copernicus-glo30-h3-terrain", repo_type="dataset")

For res 8/9, filter on the res-1 parent first so DuckDB skips irrelevant partitions.


Vertical datum & height-above-ground

Elevations here are orthometric (referenced to the EGM2008 geoid). GNSS / ADS-B devices report ellipsoidal height (above the WGS84 ellipsoid). The two differ by the geoid undulation N, shipped here as geoid_undulation so you don't have to carry a geoid model:

geoid_undulation = N = h_ellipsoidal − H_orthometric   (negative in CONUS, ~ −15 to −36 m)

The EGM2008 geoid spinning, exaggerated

The EGM2008 geoid, exaggerated ~13,000× — red where it rises above the WGS84 ellipsoid, blue where it dips below (the undulation N, ±~100 m). The geoid_undulation column samples this field per cell.

To compute height above the terrain from an ellipsoidal-height source:

height_above_ground = h_ellipsoidal − elevation_mean − geoid_undulation     (all metres)

Worked example (an aircraft on the ground at Teterboro, KTEB): h_ellipsoidal ≈ −29 m, elevation_mean = 1 m, geoid_undulation ≈ −32 m−29 − 1 − (−32) ≈ +2 m above ground. ✓

geoid_undulation is evaluated at each cell's centroid from the EGM2008 model (EPSG:3855).


Coverage & caveats

  • Land only / sparse. Cells exist only where Copernicus has data (land + coast). A missing cell means open ocean → treat as sea level (0 m orthometric).
  • It's a surface model (DSM), not bare ground (DTM). Elevations include tree canopy and buildings — an upper bound on ground height in vegetated/built areas.
  • Sharp peaks read low. A 30 m DSM under-samples knife-edge summits — e.g. Everest reads ~8,738 m vs. the true 8,849 m. Check peaks against elevation_max, expecting the Copernicus value, not the survey height.
  • Water is flat-lined. Oceans and large water bodies read ~0 m; inland water sits at its surface elevation.
  • Polar latitudes carry the usual high-latitude DEM artifacts.
  • These are aggregate statistics, not the original raster — for sub-cell precision use the source DEM directly.

How it was built

  • Source: Copernicus DEM GLO-30 (2021 release), the global 30 m DSM, ~26,450 1°×1° Cloud-Optimized GeoTIFFs, from the AWS Open Data registry.
  • Aggregation: every source pixel is assigned to the H3 cell containing its centroid and folded into composable statistics (count, sum, sum-of-squares, min, max) — done independently per resolution. Cells straddling processing-tile seams are recombined exactly.
  • Geoid: geoid_undulation from EGM2008 (EPSG:3855), per-cell centroid.

The full build pipeline and a step-by-step reproduction guide live in the source code repository: github.com/whereyaheading/copernicus-h3.

Validation

Methodology and pass criteria are in VALIDATION.md; the latest run against this dataset is in VALIDATION_REPORT.md (currently 23/23 checks passing). The headline correctness signal: Σpixel_count is identical at every resolution (225,865,152,000) — each resolution independently aggregated the exact same source land pixels.


License & attribution

The data is a derivative of the Copernicus DEM and is redistributed under the Copernicus DEM licence. If you use this dataset you must carry the following attribution, and the same obligation passes to your downstream users:

Produced using Copernicus WorldDEM-30 © DLR e.V. 2010-2014 and © Airbus Defence and Space GmbH 2014-2018 provided under COPERNICUS by the European Union and ESA; all rights reserved.

See the LICENSE file for the full Copernicus DEM terms and disclaimer. By downloading the data you accept those terms.

Citation

If you use this dataset, please cite:

Haugen, Frances (2026). Copernicus GLO-30 Terrain, H3-Indexed (v1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.20518940. Derived from Copernicus DEM GLO-30.

@misc{haugen2026copernicush3,
  author       = {Haugen, Frances},
  title        = {Copernicus GLO-30 Terrain, H3-Indexed},
  year         = {2026},
  version      = {1.0},
  publisher    = {Zenodo},
  doi          = {10.5281/zenodo.20518940},
  url          = {https://doi.org/10.5281/zenodo.20518940},
  note         = {Derived from Copernicus DEM GLO-30}
}

Citing this dataset does not replace the mandatory Copernicus DEM attribution — see the LICENSE file.

Downloads last month
84