How Hydrionis handles data

A water data platform is only as good as its discipline about where numbers come from. These are the rules Hydrionis operates under, enforced in code rather than by convention.

Every value is an observation

There is no second, looser path for a figure that someone simply knows. Every number is a structured observation carrying a metric, a subject, a period, a unit, a source, a basis, a confidence level and a retrieval date. A value without a valid source identifier fails the build. That single constraint is what makes the source panels on this site trustworthy rather than decorative.

Basis: how the number came to exist

Native geography

Every metric records the geographic level at which it is actually reported. Rainfall is a grid or station quantity. Leakage is a utility quantity. Drinking water compliance is a supply zone quantity. Tariffs are a tariff-jurisdiction quantity.

Values are stored at that native level and only transformed to another where it is scientifically legitimate — area weighting for rainfall, population weighting for access rates, simple summation for volumes. Where a transformation is applied, the observation carries a record of what was done and to what inputs.

When sources disagree

Two authoritative sources frequently give different values for the same apparent quantity. Hydrionis treats that as information rather than as a problem to resolve by picking one.

• Both values are retained, each with its own source and period
• The likely reason for the divergence is stated: different definitions, different methods, different reporting periods, or different geographic boundaries
• Hydrionis does not silently select the more convenient figure

The commonest causes are definitional. Leakage reported by a UK regulator and non-revenue water reported internationally are different quantities. Water stress under the SDG 6.4.2 definition subtracts environmental flow requirements and produces higher values than the older simple ratio. Nitrate reported as N and as NO₃ differ by a factor of 4.43.

Gaps

Transcription status

Sprint 1 values were transcribed from published compilations by hand rather than ingested automatically from publisher APIs. Every such observation carries the qualifier "transcribed, pending automated ingestion", which is shown in the source panel and exposed in the API. This is a genuine limitation and it is stated on every affected value rather than in a footnote.

Observation identifiers are deterministic, so ingested values will supersede transcribed ones in place, and the superseded record is retained rather than deleted.

Licensing

Every source carries structured rights: whether reuse, redistribution, derivative use and commercial reuse are permitted, whether attribution is required, and when the terms were last checked. Display rights, download rights and future commercial rights are modelled separately, because publicly accessible is not the same as commercially reusable and treating them as equivalent is how data platforms end up in trouble.