time series with sparse em axes

[Markus Demleitner]

Hi Marco,

On Fri, Mar 21, 2025 at 06:26:59PM +0100, Molinaro, Marco via semantics wrote:
> for a "heritage data" project we have a set of observations
> that consist of time series of nearly monochromatic RHCP & LHCP
> radio fluxes of the full disk Sun.
>
> Unfortunately the files they are stored in (FITS bintables)
> contain:
> - the 6 individual observational frequencies, sparse from 237Mhz to 2.7GHz
> - for each of them 3 time series: LHCP, RHCP and the sum of them (full flux)
>
> This leads us to describe them, no matter if in obscore or epncore,
> as 6 separate time series records referencing the same FITS dataset,
> because we are unable to describe the sparse "em" axes.

Well, as long as the time axis is sampled uniformly, there is nothing
that would keep you from having just one time series with 18 columns,
and I believe clients like SPLAT would do a reasonable job displaying
them.

If the time axis is not sampled uniformly over all frequencies, I
think I'd go for per-frequency time series with one column each for
the three observation modes.

All that is based on what I think will make this data easy to work
with with generic VO tools, completely independently of the
underlying physics and in particular community practice.  Therefore:

> Or maybe Radio, HE, TD IGs have some insight on this.

I'd frankly ask Solar System first (and its vice chair already
confesses he can't help here).  I'd say *if* there's some dominant
practice in the community, whatever format that calls for is what I'd
serve up by default; and then I'd attach a datalink that yields
SDM-compliant tables with Ada blocks (:-)
<http://ivoa.net/documents/Notes/LightCurveTimeSeries/index.html>
for interoperability beyond that community.

Consider this as not much more than my 2 cents, though.

        -- Markus