Bio-ORACLE v3.0 — Global Marine Environmental Layers

Global marine environmental rasters from Bio-ORACLE version 3.0, converted to cloud-optimized GeoTIFF and aggregated to H3 hexagonal cells at resolution 6.

This bucket holds the 2010-decade baseline, depth-mean realm: 16 requested layers, published as 18 datasets (current direction and aspect are each split into a sine and a cosine component, see Circular layers).

Citation

Both papers must be cited:

Assis, J., Fernandez Bejarano, S.J., Salazar, V.W., Schepers, L., Gouvea, L., Fragkopoulou, E., Leclercq, F., Vanhoorne, B., Tyberghein, L., Serrao, E.A., Verbruggen, H., De Clerck, O. (2024) Bio-ORACLE v3.0. Pushing marine data layers to the CMIP6 Earth system models of climate change research. Global Ecology and Biogeography. https://doi.org/10.1111/geb.13813

Tyberghein, L., Verbruggen, H., Pauly, K., Troupin, C., Mineur, F., De Clerck, O. (2012) Bio-ORACLE: A global environmental dataset for marine species distribution modelling. Global Ecology and Biogeography, 21, 272-281. https://doi.org/10.1111/j.1466-8238.2011.00656.x

License: Bio-ORACLE is released under the GNU General Public License (source). No version is stated upstream, so the STAC license field is other with a link to the provider's terms rather than a guessed SPDX id.

Layers

Water-column layers are the depth-mean realm (averaged over the water column), baseline experiment, decade 2010. Terrain layers are benthic and static.

Each dataset publishes two assets:

Query with DuckDB

Build a full SDM covariate table by joining the layers you need on h6. Every dataset shares the same cell set, so the joins are exact.

H3 resolution: this data has no h8 column

This is deliberate, not a defect. The source grid is 0.05 degrees (~5.5 km), so one pixel covers ~31 km2. An h6 cell is 36.1 km2, which matches the pixel. An h8 cell is 0.74 km2, which would replicate each measured value across ~42 identical cells and publish apparent detail the source does not have.

To join against an h8-native dataset (GBIF or OBIS occurrences, carbon, GHS-POP), roll that dataset up to h6:

Columns present: h6 (native), h5, h0 (partition key).

Aggregation

Every value is an area-weighted mean of the source pixels inside the cell. The columns are already averages, so combining cells means averaging again, weighted by cell area, never adding. There is no meaningful catalog-wide SUM of any column here.

These are raster reductions rather than feature conversions, so there is no _cng_fid column; every row is one distinct cell and no de-duplication is needed.

Circular layers: current direction and aspect

swd (current direction) and aspect are compass bearings in 0-360 degrees. A bearing cannot be averaged arithmetically — two north-facing cells at 1 and 359 degrees would average to 180 degrees, due south. Each is therefore stored as its sine and its cosine, averaged independently, and recombined on read:

directional_consistency is 1 where a single source pixel contributed and falls below 1 where pixels within the cell pointed in different directions.

The same applies to terrain/aspect-sin and terrain/aspect-cos.

No-data

The COGs use NaN as no-data rather than a numeric sentinel. bathymetry_mean reaches about -10700 m, so the conventional -9999 would fall inside the physically valid range of that layer and could be read back as a real depth. The hex has no NULL values: land and no-data pixels simply produce no cell.

Provenance

Source NetCDF files, as delivered by the Bio-ORACLE download service, are retained under public-bio-oracle/raw/. Conversion to COG and the sine/cosine decomposition are done by catalog/bio-oracle/scripts/netcdf_to_cog.py in the data-workflows repo.