Where the data comes from

Providers

The provider mix is not fixed. Coverage expands region by region, and a region is only published once at least one provider with a redistributable licence covers it — an importer cannot run at all until it declares a licence, a licence URL, and an exact attribution string. There is no "licence unknown" state anywhere in the pipeline.

Each licence badge links to that licence's full text. That is not decoration: the map's attribution control credits these providers by name and links here rather than printing every credit string inside a control a few hundred pixels wide. Both CC BY 4.0 and CC BY-IGO permit attribution by link only when the linked page carries the complete required information — the verbatim credit, the licence, and a way to read it — so this page is what makes that form compliant.

The source_id values above are the exact strings returned in the observations array of GET /v1/areas/{id} and in the source_id field of GET /v1/population. Use them to join an API response back to the entry on this page.

This page is the human-readable summary. GET /v1/sources is the authoritative machine-readable registry: the same rows plus the per-country entries this page collapses into one card (for example geoboundaries-au), with the exact licence each carries.

Attribution requirements

Attribution is a condition of use, not a courtesy. If you publish, redistribute, or build a product on data returned by this API, reproduce the attribution string of every provider whose data you used, exactly as written above.

• Copy the attribution text verbatim. Do not shorten it, translate it, or replace it with a link to this page.
• Attribute per provider, not per API. Naming GeoAPI alone does not satisfy any of the licences below it.
• Determine which providers you actually used from the source_id field on each observation you consumed, rather than crediting all of them by default.
• Business listings served by /v1/places are first-party and proprietary. They may be queried through this API but not bulk-extracted or resold as a standalone dataset.

How observations are resolved

An area frequently carries several observations that disagree, because the providers behind them measured different things in different years. Every observation is kept and served; one is selected as the headline figure by the following order, applied top to bottom until a single value remains.

1. Method. census outranks projected, which outranks modelled.
2. Source priority. A per-provider rank configured at import time.
3. Recency. The most recent reference year wins.

An observation that falls far below the combined total of the area's own sub-divisions cannot be describing the same territory, so it is excluded from the resolved figure — though never deleted: it stays in the observations array, flagged with the reason, so the disagreement is inspectable rather than silently averaged away. An observation that sits far above its sub-divisions is reported instead of withheld, since that pattern usually indicates gaps in the children rather than an error in the parent. See the population endpoints for the response shape this produces.