Sources and standards

Where the data comes from.

80 organisations publish the data we hold. Here is every one of them, and the standards our data and our models are held to.

Publishers

80 publishers, named.

The geological surveys, mining registries and space agencies whose published data we hold, reconcile and put to work.

Australia · 12

The Americas · 26

Europe · 16

Africa and the Middle East · 15

Asia and the Pacific · 8

International · 3

We reconcile and interpret published data; the publishers own it. Names and logos identify the publishers of data in our holdings and belong to their owners; their use does not imply endorsement of, or partnership with, MineDSS. Image credits and licences are on the credits page.

Standards

How we keep ourselves honest.

Models tested on unseen ground

Every model is scored on ground held back from its training; for the national models that ground sits at least 10 km from anything they trained on. A model that fails is not used.

Mineral mapping tested blind

Benchmarked against published USGS reference mapping, and tested before the answer was read.

Every figure counted

Every number on this site is counted from our data, and every mineral we list is one we actually serve.

What our work is not

A resource or reserve estimate (JORC, NI 43-101)
A discovery claim: we rank ground, and the drill confirms
Legal advice on tenure or land access
A substitute for a live register query on the day you peg

Data standards

How the data is kept honest.

The value of a dataset this size is not its size. It is whether you can trust any single row of it.

Measured, never recalled

Every figure on this site is regenerated from the data before it is published. A number that cannot be re-derived from the holdings does not appear.

The publisher on every row

Each feature carries its publisher, the layer it came from and the date it was read, so any claim can be traced to its source.

Dated, not timeless

Registers move weekly. Every snapshot says when it was taken, and a pegging or transaction decision is always sent back to a live query.

Placeholders are read per publisher

One publisher's −1 is another's genuine zero. Sentinel and not-recorded codes are interpreted publisher by publisher, never assumed.

Duplicates collapsed, not counted twice

The same sample often appears more than once, mostly the same publisher's data held in two releases or in an earlier compilation. Those records are clustered and counted once: 110M distinct samples, not 200.5M records.

Audited, release by release

Every release of the geochemistry passes an automated battery of checks for impossible values, unit errors and placeholders. What fails is recorded and handled reversibly, never silently fixed.

Talk to us

Talk to us.

Exploration and mining companies interested in working with us, investors, and the publishers and researchers whose work the record is built on.

Talk to us [email protected]