Data Source vocabulary

[Markus Demleitner]

Hi Vandana,

Thanks for your input!

On Mon, Jul 20, 2026 at 04:52:43PM -0700, Vandana Desai via semantics wrote:
> The proposed term "Modeled" says :"This includes the current case of no
> attempt to simulated existing objects", and so seems to fit the
> OpenUniverse2024 mock galaxy catalogs and simulated images.
>
> However, the proposed term "Simulation" says "simulation of an instrument
> output", which seems like it it also applies.

Have you looked at the latest version of this vocabulary that I hope
also reflects IPAC input:
http://www.ivoa.net/rdf/data-source/2026-02-03/data-source.html

(aw, dang, I forgot to update the date tag... I need tooling support
for that it seems; but this is a draft, so I don't bother fixing it.
Meanwhile, the vocabulary in this state really is from 2026-06-30).

I think the current definitions make it rather clear that your
example would have abstract-model as data source.

> Aside from that, we'd like metadata that would make it easier for
> visualization tools to understand that I can overlay the OU2024 catalogs on
> the OU2024 images, but it would not make sense to overlay 2MASS on OU2024
> images.

I think that would be "only attempt overlays if something has
observation or concrete-model as data source".  If you find that's
not enough, what extra metadata would you need?

> *Rubin *(summarized from Gregory Dubois-Felsmann, cc'd)
>
> One could have a concept of a set of universes, one of which is the real
> one, while the others are theoretical ones that are not expected to align
> with the real one other than statistically.
>
> For example, Rubin avoids combining DP0.2 data and DP1 data on the same
> ObsTAP service, because the former is derived from a cosmological
> simulation, not at all aligned in detail with the real one, while DP1 is
> based on real on-sky data.

This is what largely inspired the current language.  If you have
suggestions on how to better reflect this test in the descriptions,
by all means let me know.

> *SPHEREx (also from Gregory)*
> In SPHEREx, we have real data but we also have an image simulation that is
> driven off of real on-sky catalogs of objects -- and the two look
> remarkably similar, to the point where it's meaningful to blink them.
>
> It would be helpful if the standards could steer users away from comparing
> data products that are incompatible.

Absolutely, and this is the baseline of what we're after here.

> I hope this input is helpful! Happy to discuss.

I'd be extremely grateful for more input on the current concept
descriptions.  If someone has less clunky words for "Abstract" and
"Concrete Model", I'd be grateful, too, but the central point for me
is to get the concepts right and their definitions as watertight as
natural language can get.

Thanks,

         Markus