Javier Jimenez Shaw via PROJ <[email protected]> writes:

> Is there a way to add the geoid model in the short id of a CRS? There is a
> way to put the epoch with an @. Maybe another way to set the geoid model?
>
> (dynamic) CRS with an epoch: EPSG:[email protected]
> Compound CRS: EPSG:6516+5703
>
> Many users want to set the geoid model used. Something like (but not
> necessarily)
> EPSG:6516+5703[GEOID12B]

This seems conceptually problematic.  I don't think epoch and geoid
model are at all the same kind of thing.  I think this isn't related to
compound CRS so much as the definition of a vertical CRS.

5703 is NAVD88.  That is still, as far as I know, defined by passive
controls.

People obtain estimates of NAVD88 heights by using HAE (surely mostly in
EPSG:6319) and then using a geoid model.  Of course, all measurements
are estimates and the traditional approach is leveling from passive
controls, also an estimate.

Presumably someone wanting to write EPSG:5703[GEOID12B] is saying that
the heights were obtained via HAE and then transformed to NAV88.  That's
an estimate of HAE combined with an estimate based on geoid model use.

I feel that this sort of information is source metadata and not part of
the CRS.

Is the purpose so that when the data is transformed back to HAE that the
same transform will be used?   If not, I wonder what the purpose is.

If this is ok, should we allow annotation of whether static carrier
phase or RTK was used, receiver and antenna brand, observation
intervals?  There might be biases that later become understood, and this
information is helpful in estimating the variance.  Where is the
boundary in terms of how information about how coordinates was obtained
between things that can be annotated in a CRS expression and things that
can not or should not be annotated?

An alternative view is that 5703[GEOID12B] is a *different CRS* from
5703.  I lean to this somewhat.  To deal with the various different CRSs
created by this view, it feels like an ensemble situation.  We should
then have 5703, NAVD88(original), and then a variety of NAVD88(GEOIDnn),
all within some NAVD88(ensemble).

It would be interesting to see what NGS says about whether it is
reasonable to consider NAVD88 obtained by a particular geoid model as a
distinct reference frame.  I suspect they'd say that no, there is one
frame, and they have published successive transformations that they
believe to be increasingly accurate.

I also wonder about the differences in coordinates obtained by varying
reasonably modern geoid models compared to the uncertainty in HAE.

Personally I have decided to store observations in EPSG:6319 (without a
compound orthometric height, so that HAE is stored), because I can
compute NAVD88 later when I want to, and it feels cleaner to store data
closer to what was observed.  Plus, I expect that NAPGD2022 will be
released within my lifetime, and 6319 HAE to NAPGD2022 seems more
straightforward then transforming from NAVD88.  However, my vague
perception is that my approach is odd.

It's interesting about 'many users' it would be interesting for them to
write here and explain their situation and how this annotation works,
with an analysis of estimated errors.


_______________________________________________
PROJ mailing list
[email protected]
https://lists.osgeo.org/mailman/listinfo/proj

Reply via email to