The kinds of water

A water flow says four things. Until this taxonomy the list had somewhere to put three.

The last one is the new piece. Because the deduplication signs on every semantic field of a flow, a difference held in no field is a difference that disappears: EF 3.1’s Water use works as a pair — a positive factor when water is withdrawn, a negative one when it is returned — and when the two halves looked identical, one of them was merged away and the negative half went with it. In the case audited that inflated a water use result by roughly a factor of 434.

The kind of water is not the same question as the water body, and the two are not interchangeable. Water taken from a lake is lake water, from a lake. Water discharged to a lake is plain water, to a lake. For a withdrawal the two agree, which is why the withdrawal body can be read off the kind; for a discharge they are two separate facts, and nothing infers one from the other.

17 kinds, as a small hierarchy. The names are not ours: each one is tied to a published environmental vocabulary, so that sea water in this list means what sea water means elsewhere and can be checked by anyone. Two authorities, split by what they can actually type — ENVO for what the water physically is, and AGROVOC for the accounting category no OBO ontology carries. Where both publish the class, both are cited: cooling water is named by each of them, which is corroboration rather than redundancy, and it is the find that decided this design — EF’s Water to Cooling is not a modelling quirk to tolerate inside a name but a published class with a definition and a parent.

A kind with no flow object mints nothing and never appears on an inventory row: it exists so that the kinds under it have the parent the vocabulary gives them. A kind with an object but no flow is one this build did not reach — several are used by the Swiss BAFU list, so which of them are empty depends on which sources went into the build.

Most of what is above is what ENVO already states, checked automatically against a stored copy of what the vocabularies actually publish rather than taken on trust. 4 of 17 kinds state something of their own — an edge the authority declines to assert, a citation that is a close match rather than an equivalence, or an identifier nobody else publishes. Each says which, and why, so that none of it passes as somebody else’s.

Three smaller decisions, recorded so they are not re-litigated. Brine is the one kind carrying no CAS number: EF 3.1 gives Water, salt, sole water’s 7732-18-5, and that is treated as an error and dropped, because brine is a mixture rather than the molecule. Green water is deliberately not filed under rainwater: physically it is rain that infiltrated the soil, but it exists as a separate flow for water-footprint accounting rather than physical reasons, and treating it as a kind of rain would blur that — so it sits outside the tree and cites AGROVOC rather than ENVO. And water vapour is its own branch, because ENVO treats it as a gas rather than a form of liquid water; the unit does not tell the two apart, since vapour is filed in both kg and m³.

Every flow published on one of these kinds. The filters are read together: Cooling water with Return is the return half of the cooling water pair, not everything that is either. A withdrawal and a return of the same kind are the two halves EF characterises with opposite signs — seeing both on one row is the whole point of recording the kind.

19 of 52 flows.