linked-lcacommons

A Linked Data wrapper around the Federal LCA Commons, a federation of U.S. federal agency, national laboratory and academic life-cycle inventory publishers.

What is here, and what is not

The Commons publishes 27 repositories. All 27 are described here from its live metadata API. Dataset content is served for two of them, because the Commons has no release-download endpoint and only these publish their archives elsewhere. Ask for a dataset in one of the other 25 and you get a 404 that says so and links upstream — not one implying the record does not exist.

The two are not an arbitrary pair. Together they are the closure of the USLCI supply graph: all 3 730 of its supply links resolve within them, 358 crossing from USLCI into the electricity baseline, and the baseline names no suppliers of its own.

U.S. Life Cycle Inventory Database: 962 processes, 4 081 flows, 494 sources, 317 locations, 67 actors. Published by NREL; quarterly releases.
U.S. electricity baseline: 771 processes, 2 310 flows. Where a supply chain bottoms out into generation.

The supply graph

These are unit processes: a production step with its real material and energy inputs, and often an explicit link to the process supplying each one. There are 3 730 such links across the packaged corpus, and the graph is traversable in both directions.

A process with a lot of upstream: Hardboard, at hardboard plant — 119 exchanges, 46 of which name the process that supplies them. Follow lca:suppliedBy from any input.
The other direction, which the source data does not contain: Electricity; at user; consumption mix — US — named as a supplier by 264 other processes. olca records a supply relationship only on the consumer, so this direction is built by indexing the whole corpus at startup.
A supply link that crosses repositories: Vacuum infusion, rigid composites part, at plant — consumes electricity from the baseline repository. 358 links do this, which is why the two repositories are packaged together.
An aggregated result rather than a unit process: Phenol resorcinol formaldehyde hardener, at plant — typed lca:LciResult, not lca:UnitProcess. Its inputs are already rolled up, so it must not be double-counted into a chain that also contains its parts. 26 of USLCI's 962 are these.

Browsing and searching

Browse by category: category — 251 categories, built from the category strings the records carry, since olca has no category datasets. Try emissions to air, the largest at 1 167 flows.
Search all 27 repositories: search?q=concrete — q, as in every product wrapper of this family, which this wrapper sends upstream as query. Upstream itself ignores q and answers with all 360 224 records, which is why the translation happens here and an unknown parameter is a 400 rather than passed on. Narrow with group, which upstream does honour.
All 27 repositories: repository — each says whether this wrapper carries its datasets. Compare a packaged one with one that is only described.
Release history: release/…/USLCI_Database_Public — 11 quarterly releases. This wrapper carries one and says which; it cannot diff two, because no release can be downloaded from the Commons.

Things worth looking at

A flow with an identifier that links outside this corpus: Aluminum — CAS 7429-90-5, formula Al. 5 013 of the 6 391 flows carry a CAS number, which is how an elementary flow here joins up with a substance record anywhere else.
A place, and its boundary kept separately: United States of America — 2.8 KB, carrying the centroid and a locn:geometry link to the boundary, which is 775 KB of WKT. Inlining it would make the commonest small record in the corpus one of the largest.
An organisation, and something that is not one: EPS Industry Alliance is an org:Organization. But “NETL and USDA and EPA” is three bodies in one record, so it is a bare prov:Agent with a note saying why it was not split. Of USLCI's 67 actors, 42 are individual people and only 25 are organisations.
A record that states no terms of use: Flat Glass Bridge; USLCI to USEEIO — one of 45 that carry no restrictions statement. It gets lca:rightsUnstated and an explanation, never an invented licence.
Numbers that are not constants: Units of currency — upstream stores exchange rates as though they were unit conversions, with no date attached. Each carries a note saying so.

Formats

Every document is available as Turtle, N-Triples, JSON-LD and RDF/XML by content negotiation, or by suffix: append .ttl, .nt, .jsonld or .rdf. Appending .json returns the upstream olca-schema record verbatim — which is JSON and not JSON-LD however it is labelled, since those files ship with a null @context.

The terms this wrapper coins are at vocab, each with a comment saying what it means and why it exists. This description of the wrapper itself is at index.

Terms

The Commons hosts rather than authors. Each dataset carries the rights and restrictions its own submitting organisation declared, and those differ between records within one repository — 45 of the packaged processes state no terms at all, and this wrapper says so rather than inventing a licence for them. Every served document names its upstream source in prov:hadPrimarySource; dereference it for the authoritative version.

This wrapper serves a packaged release, not a live view of the Commons. Repository, release and search metadata are fetched live; dataset content is not, because the Commons publishes no endpoint for it.