public-bioclimate

Bioclimatic variables aggregated to H3 hexagonal cells, for climate-exposure and species-distribution work.

CHELSA 2.1 future bioclimate — SSP3-7.0, 2041–2070

Downscaled CMIP6 bioclimatic variables from CHELSA v2.1, aggregated to H3 resolution 8 (cells of about 0.74 km²) over the global land surface.

• 195,048,994 cells across 108 hive partitions
• Five climate models kept as separate columns, plus their median, minimum and maximum
• Seven bioclimatic variables, temperature in °C and precipitation in mm
• License: CC0-1.0

Variables

Seven bioclimatic variables, each with five climate model columns plus median, min and max (60 columns in all):

Column names follow <variable>_<model>, e.g. bio1_gfdl_esm4, plus bio1_median, bio1_min, bio1_max. The five models are GFDL-ESM4, IPSL-CM6A-LR, MPI-ESM1-2-HR, MRI-ESM2-0 and UKESM1-0-LL.

H3 index columns: h8 (native, the usual join key), h5 and h4 for coarser joins, and h0 as the partition key.

All five models are kept so any summary can be computed directly, and so the disagreement between models stays visible. For annual mean temperature the models differ by 2.0 °C on average, and by as much as 10.4 °C where they agree least.

⚠️ Annual precipitation is clipped at 6553.5 mm

The source stores precipitation as a 16-bit integer at 0.1 mm resolution, so bio12 cannot exceed 6553.5 mm — below the ~11,900 mm reached in the wettest places on Earth. Cells at that value are a lower bound, not a measurement: 247,217 cells hit it in the median and 348,969 in at least one model. Exclude them from any analysis of precipitation extremes:

Query with DuckDB

Restrict to a single partition when you know the region — it reads far less data:

Joining against other datasets

Join on h8, which is the resolution this dataset is stored at and the one most of the catalog shares:

Datasets held at coarser resolutions can be joined on h5 or h4 instead.

Visualising in MapLibre GL JS

This dataset ships as H3 hex parquet only — there is no PMTiles layer and therefore no source-layer to set. To draw it, query the cells you need and convert them to polygons with h3_cell_to_boundary_wkt(), then add the result as a GeoJSON source:

Important to know

• These are projections, not observations. Every value is conditional on the SSP3-7.0 scenario and on the model that produced it.
• Values are already means over the source pixels in each cell. Combining cells means averaging again, weighted by cell area — not adding. H3 cell area varies with latitude.
• Footprint is global land. The CHELSA source also covers the ocean; this dataset is restricted to cells intersecting the WWF Terrestrial Ecoregions of the World, which includes Antarctica.

Scenarios and periods

This collection is SSP3-7.0 for 2041–2070. Further scenario and period combinations (SSP1-2.6, SSP3-7.0 and SSP5-8.5 across 2011–2040, 2041–2070 and 2071–2100) are being added as sibling collections in this bucket.

Source and citation

Karger, D.N., Conrad, O., Böhner, J., Kawohl, T., Kreft, H., Soria-Auza, R.W., Zimmermann, N.E., Linder, H.P., Kessler, M. (2017). Climatologies at high resolution for the earth's land surface areas. Scientific Data 4, 170122. https://doi.org/10.1038/sdata.2017.122

Data: https://chelsa-climate.org/ — released under CC0 1.0.