Open Subsurface Data

A verified catalogue of open subsurface datasets (wells, seismic, horizons, production), shown on one map, retrievable with one manifest, and readable by Ophiolite and by any tool that reads GeoJSON or STAC.

Status: scaffold, 23 October 2026. Nine records, placeholder outlines, verifier at level L1.

What "verified" means

Every record carries a level produced by osd verify, never typed by hand:

Level	Meaning	Who establishes it
L0	Listed: a record exists with publisher, origin and licence as stated	a contributor, by pull request
L1	Reachable: the origin page answers	the verifier, on the network
L2	Checksummed: files hashed, licence text captured	the verifier (E02)
L3	Outlined: wellbores and survey outlines derived from the data itself	the outline extractor (E03)
L4	Converted: read into a canonical type with a fidelity record	the Ophiolite readers (E09)

A level never skips a rung. A placeholder outline drawn by hand keeps a record at L1.

Layout

records/<publisher>/<dataset>.yaml   facts about the dataset as published (schema/record.schema.json)
outlines/<publisher>__<dataset>.geojson  outlines in WGS84 (schema/outline.schema.json)
schema/                              the three contracts: record, verification, outline
src/osd/                             tools: validate, verify, build, status
site/                                the map (static, MapLibre, no server)
build/                               generated: catalog/ (published) and verification/ (per record)
docs/                                ARCHITECTURE.md, ENGINEERING.md, program/ (queue and charters)

Run it

uv sync
make check        # lint, tests, validate, build: the one CI check, offline
make verify       # network: reach every origin and write build/verification/
make site         # http://localhost:8765

Contribute a dataset

Copy records/_template.yaml to records/<publisher>/<dataset>.yaml, fill in what the publisher states, open a pull request. CI validates the record and verifies the origin. See CONTRIBUTING.md.

Licences

Code: Apache-2.0 (LICENSE). Records and outlines: CC0 1.0 (records/LICENSE), facts about public datasets. The datasets themselves keep their publishers' licences, named in each record.