Accessing OBIS data

Finding & accessing OBIS data

Elizabeth Lawrence

2026-07-24

Getting data out of OBIS

Several ways in, depending on what you need:

• obis.org — browse, search, and cite datasets
• OBIS Mapper — visual exploration + quick downloads
• robis — programmatic access from R
• Parquet + robisdb (new) — fast, local access to the full dataset

This mirrors the options from Module 8, but we’re focusing on the four most relevant for this group. Full CSV exports and direct IPT downloads exist too, but the four above cover most real workflows. Good moment to ask the room what they’re trying to do (one species vs. a whole region vs. bulk analysis) since that determines which of these fits best — the next few slides map roughly to “small ad hoc pull” → “bulk/programmatic.”

The OBIS portal — obis.org

Your starting point for finding and understanding data:

• Search & browse by taxon, region, or dataset
• View dataset summary + citation info
• Find data provider’s contact details

Worth contrasting with the Mapper (next slide): downloads from the homepage/dataset page come as separate table files, while Mapper/full-export downloads are flattened into one occurrence+event table, with eMoF/DNA extensions available only if you check the box for them. This distinction trips people up, so flag it explicitly.

OBIS Mapper — mapper.obis.org

Visual, filter-driven exploration of OBIS data:

• Filter by taxon, geography, time, and extensions
• Combine multiple layers in one view
• Download individual layers as CSV
  • Better for smaller datasets <20,000,000 records
• Incldue eMoF and/or DNA-derived data extensions
  • (check the box — not on by default)

Great tool for a quick visual gut-check before doing anything programmatic — “does this region actually have eDNA data in it?” The Maptool naming collision is a real source of confusion, worth stating explicitly since Maptool came up earlier in the sensitive-data context.

robis — programmatic access from R

install.packages("robis")
library(robis)

records <- occurrence(scientificname = "Megaptera novaeangliae", fields="associatedSequences")
dna <- dataset(scientificname = "Megaptera novaeangliae", fields="associatedSequences")

• Query by species, taxon group, geometry (WKT), depth, time, quality flags…
• Built-in checklist() for taxon lists in an area
• dataset() — query for datasets instead of biological records

github.com/iobis/robis

This is the go-to for most “I need this species/region’s data in R” tasks — talks directly to the OBIS API under the hood, so no local storage needed. Good for small-to-medium pulls; for anything approaching “the whole OBIS dataset” or repeated heavy queries, the next slide’s Parquet approach is a better fit.

New: fast access via Parquet

• OBIS publishes the full global dataset as Parquet fields
• Brand-new R package, robisdb
  • (released this month — actively evolving)

pak::pak("iobis/robisdb")
library(robisdb)

con <- connect_opendata_local("~/data/obis-open-data")
occ <- con |> occurrence_db(scientificname = "Minuca rapax")

• Mirrors robis functions where possible (occurrence_db(), checklist_db()…)
• Runs on DuckDB — fast, local queries without loading everything into memory
• Built for scale: full-dataset or repeated analyses that would be slow via the API

This is very new (this month), so treat it as “here’s where things are heading” rather than a fully mature, documented workflow — check github.com/iobis/robisdb for the latest since details may shift. Good to mention this exists specifically because loading the entire OBIS dataset via the API/robis isn’t feasible on most laptops — Parquet + DuckDB is OBIS’s answer to that scale problem. Related datasets (speciesgrids, obistherm) are also queryable the same way.