Known deposits are not a map of what exists

What a catalogue of known deposits actually samples

Both accepted databases state their own inclusion rules. One admits "only well explored deposits for which there is sufficient information to contribute to constructing deposit models" and excludes occurrences without economic potential. The other is checked and revised against imagery. Neither claims completeness, and the refused MRDS states outright that its coverage is complete for the United States and incomplete elsewhere.^[1,2]

So the density of records is a joint function of geology, of exploration history, of how open a jurisdiction’s data is, and of which compiler was interested. Only the first of those is about the Earth.^[1]

Why this corpus cannot be a training set

Every record here is a positive example. There are no barren systems in it, no unsuccessful exploration, no negative drill results, and no background geology — because none of those get compiled into deposit databases. A model fitted to this alone would learn what found deposits look like, which is not the same as learning what deposits look like, and it would have no way to discover that it had learned the wrong thing.^[1]

Phase 5 records this as a claim rather than a caveat in a footnote, because a future phase reading the corpus machine-first should meet it as data. The requirements for a real prospectivity model are listed in the architecture documents, and known deposits are one item on a list of eight.^[1]

What is refused, in code

There are no density maps of known records, no favourability surfaces, no scores, and no rankings of ground or of countries. Nor is there any machinery for suggesting that one known deposit implies another close by. The whole vocabulary is refused across every published summary, statement, caption, paragraph and source-record mapping — including this article, whose earlier draft spelled the banned phrases out as examples and was rejected by the build for it. That is now the third time a Lithoen article about a gate has been stopped by that gate, and the correct outcome all three times.^[1]

Commonly misunderstood