GBIF Species & Occurrence Scraper

Export biodiversity data from GBIF (the Global Biodiversity Information Facility) — the world's largest open index of species records, backed by more than 100,000 datasets from museums, herbaria, research institutions and citizen-science platforms.

This actor turns any search term into clean, flat JSON/CSV/Excel:

• Species mode — taxonomic name records: scientific name, canonical name, rank, taxonomic status and the full hierarchy (kingdom → phylum → class → order → family → genus → species) plus authorship.
• Occurrences mode — individual observation and specimen records including GPS latitude/longitude, country, event date, year, basis of record, institution, dataset and collector.

No API key. No login. No proxy required.

Features

• Two scrape modes: species (taxonomy) and occurrences (records with coordinates)
• Multiple search queries per run — scientific names, genera, families or common names
• Offset pagination handled automatically, with de-duplication across queries
• ISO-8601 normalisation of GBIF's messy eventDate values (date-only, datetime, ranges, plain years)
• Every field is nullable and every record is flat — no nested objects to unpack
• Browser-grade TLS fingerprint (curl_cffi, Chrome 124) with an automatic stdlib fallback
• Runs on 512 MB, datacenter-clean, no proxy costs

Input

Field	Type	Default	Description
mode	select	species	species for taxonomy records, occurrences for records with coordinates
queries	array	["Puma concolor", "Quercus"]	Free-text GBIF search terms
maxItems	integer	100	Total record cap across all queries (max 1000)

{
  "mode": "occurrences",
  "queries": ["Puma concolor", "Panthera onca"],
  "maxItems": 500
}

Output

Species record

{
  "type": "species",
  "key": 116892593,
  "scientific_name": "Puma concolor",
  "canonical_name": "Puma concolor",
  "rank": "SPECIES",
  "taxonomic_status": "ACCEPTED",
  "kingdom": "Animalia",
  "phylum": "Chordata",
  "class_name": "Mammalia",
  "order_name": "Carnivora",
  "family": "Felidae",
  "genus": "Puma",
  "species_name": "Puma concolor",
  "authorship": "(Linnaeus, 1771)",
  "num_descendants": 0,
  "url": "https://www.gbif.org/species/116892593",
  "source": "gbif",
  "scraped_at": "2026-07-28T11:37:02.247510+00:00"
}

Occurrence record

{
  "type": "occurrence",
  "key": 5015469523,
  "scientific_name": "Puma concolor (Linnaeus, 1771)",
  "latitude": -19.588499,
  "longitude": -56.087017,
  "country": "Brazil",
  "country_code": "BR",
  "event_date": "2016-09-09T00:00:00",
  "year": 2016,
  "basis_of_record": "HUMAN_OBSERVATION",
  "institution": "NABU|naturgucker",
  "dataset": null,
  "kingdom": "Animalia",
  "family": "Felidae",
  "genus": "Puma",
  "elevation": null,
  "recorded_by": "105162832",
  "url": "https://www.gbif.org/occurrence/5015469523",
  "source": "gbif",
  "scraped_at": "2026-07-28T11:37:03.422709+00:00"
}

Use cases

• Biodiversity research — assemble species checklists for a taxon or region and cross-reference them against your own field data.
• Ecology — pull georeferenced occurrence points to fit species distribution models (MaxEnt, ENMeval) or map ranges in QGIS.
• Species datasets — build training corpora for taxonomic name matching, fuzzy name resolution or ML classification of scientific names.
• Conservation — track where and when a threatened species has been recorded, by which institution, and how sampling effort changed over time.

Pricing

Roughly $0.5 per 1,000 records. GBIF is free and public, so there are no proxy or API costs baked in — you only pay for compute. A 1,000-record run finishes in about a minute on 512 MB.

Notes & limits

• GBIF's full-text search is fuzzy: broad terms like Quercus return every name usage across all source datasets, so the same binomial can appear from multiple checklists (each with its own key). De-duplication is by GBIF key, not by name.
• Not every occurrence has coordinates. GBIF stores records from herbarium sheets and old catalogues that were never georeferenced — latitude/longitude will be null for those. In a typical run 90 %+ of records carry coordinates.
• dataset, elevation, authorship and other optional Darwin Core terms are frequently empty upstream. That is source data, not scraper loss.
• Data is published by GBIF under open licences (mostly CC0 / CC-BY / CC-BY-NC). Check the licence of each contributing dataset before republishing, and cite GBIF in academic work.

Related actors

• iNaturalist Scraper — citizen-science observations with photos
• UniProt Scraper — protein sequences and annotations
• Earthquake Scraper — global seismic events with coordinates
• Air Quality Scraper — station-level pollution measurements

Development

python -m py_compile src/*.py
python tests/test_smoke.py     # offline smoke tests, no network needed