Lab data

Files

Each file is a plain download; Parquet files can also be queried in place, for example from DuckDB. The labs also read critwire's species, countries and range-listings files under /v1/. The lab files sit outside /v1/ because they change as the course does: check manifest.json, which lists every file's columns, rows, size, sha256 and sources, and the /v1/ files' sha256 the labs were tested with.

Total: 95.9 MB (95,877,990 bytes). Shapes are GeoParquet 1.0.0: the outline is in a geom column as WKB, in longitude and latitude (WGS 84). Some outlines are invalid on purpose, for the lab on fixing geometry.

Reading the CSV files: Namibia's country code is NA, which some tools read as a missing value. An empty cell is the missing value. In pandas, pass keep_default_na=False, na_values=[""] to read_csv; in R, pass na.strings = "" to read.csv. DuckDB and the Parquet files keep it as the text NA.

PostGIS

learn_curriculum.sql builds every lab table in the schema learn_curriculum with nothing but psql and the extensions postgis and postgis_raster:

psql -X -v ON_ERROR_STOP=1 -d your_database -f learn_curriculum.sql

It runs as one transaction and checks its own row counts before it commits. Loading it again replaces the whole schema, including any tables you added to it; a failed load leaves the previous schema as it was. It refuses to run if anything outside the schema depends on it, such as one of your views. A database used only for the labs avoids both problems.

The GBIF sample

The occurrence files hold 40,395 records of critwire's species from 400 datasets shared through GBIF. They come from one GBIF download, so anyone can fetch the same records again: 10.15468/dl.fbpk9w.

Each record is open: its license, or its dataset's when that is stricter, is CC0 1.0 or CC BY 4.0, and the license column says which. Each species has at most 2,000 records, chosen by the lowest md5 of the GBIF id, so any database picks the same ones. These are records of where people saw or collected an animal, not maps of where it lives.

GBIF registered the sample itself as a derived dataset, so each dataset in it gets credit: 10.15468/dd.5sthtk. Cite it when you use these records:

Derived dataset GBIF.org (28 October 2026) Filtered export of GBIF occurrence data https://doi.org/10.15468/dd.5sthtk

Sources and licenses

Credit each source you use. Each file's sources are listed above and, ready for code, in manifest.json.

Natural Earth

Made with Natural Earth. Free vector and raster map data @ naturalearthdata.com.
License: Public domain.

RESOLVE Ecoregions 2017

Dinerstein, E., Olson, D., Joshi, A., et al. (2017). An Ecoregion-Based Approach to Protecting Half the Terrestrial Realm. BioScience 67(6), 534–545. https://doi.org/10.1093/biosci/bix014
License: CC BY 4.0. DOI: 10.1093/biosci/bix014.

Jung et al. 2020, habitat map v004

Jung, M., Dahal, P.R., Butchart, S.H.M., Donald, P.F., De Lamo, X., Lesiv, M., Kapos, V., Rondinini, C., Visconti, P. (2020). A global map of terrestrial habitat types. Scientific Data 7, 256. https://doi.org/10.1038/s41597-020-00599-8. Version 004, https://doi.org/10.5281/zenodo.4058819 (CC BY 4.0).
License: CC BY 4.0. DOI: 10.5281/zenodo.4058819.

GBIF occurrence download

GBIF.org (23 October 2026) GBIF Occurrence Download https://doi.org/10.15468/dl.fbpk9w
License: CC BY 4.0. DOI: 10.15468/dl.fbpk9w. Derived dataset: 10.15468/dd.5sthtk.

critwire

critwire (https://critwire.com), CC BY 4.0
License: CC BY 4.0.

Datasets

The 400 datasets in the sample, with the publisher that shared each, its DOI, its declared license and how many of its records the sample holds. The same list is in datasets.csv.