[PROJ] Problems with CGVD28 height

[Greg Troxel]

You said that you get the same results starting from NAD83.  I would
suggest debugging the transform NAD83 and not trying to use WGS84 until
the first step works right.  proj is known to mistransform WGS84 to
NAD83 due to ensemble fuzz.  My understanding is that NAD83(CSRS)v2,
being from 1998, will be pretty closely aligned with newer US NAD83
realizations and later Canadian ones, and thus not have the accuracy
issues of the 1986 version of NAD83, which is what causes 90%+ of the
issues.  So this is all likely a red herring, but still avoiding asking
proj to do it will avoid trouble; you don't need that trouble and height
trouble both at the same time.

NRCAN seems to publish transforms from ITRF to modern NAD83(CSRS).
So I'd assume your "WGS84" coordinates are ITRF, use the NRCAN
transforms, and then assume that those coordinates in some highish
version are equivalent to v2, as a first guess.

Looking at EPSG:32181, MTM zone 1, the bbox shown is enormous, but the
description says: Area of use: Canada - Newfoundland - onshore east of
54°30'W.

Looking at EPSG:5713, CGVD28, it seems that Newfoundland is not part
of it, and it starts at -59.73.   I have to wonder if that is intended
or a bug.  It is derived from leveling, and if they didn't have leveling
to Newfoundland, it just wouldn't be part of the datum.   It seems the
grid files are probably some modern estimation of what the heights would
have been had they been defined.

If you get datasheets for passive control in eastern Newfoundland, do
they publish CGVD28 heights?

I went to

  https://webapp.csrs-scrs.nrcan-rncan.gc.ca/geod/tools-outils/gpsh.php?locale=en

and entered rough coords for St. John's

  latitude 47.5825
  longitude 52.8580 (west)
  h 0

for CGVD28 HT2_2010v70 and NAD83(CSRS) [they don't specify versions
which is an accuracy clue], then I get -10.121 for H.  On the other hand
if I put in Montreal coordinates

  latitude 45.5095
  longitude 73.5768
  h 0 

I get H 30.78.

So they are returning results in their web tool.


Given all of that, the comments from Erixen Cruz make a lot of sense to
me.  If they've published a grid file covering newfoundland, and it has
sensible values, then it may make sense to use it.  The flip side is
that the area of validity of a datum is not a bbox in lat/lon.  The
shape can be much more complicated.  So it's not necessarily possible to
publish a grid file that only covers the valid area, unless it has
codepoints for invalid.  I bet the senior height expert at NRCAN knows
how they are trying to thread this needle.

I would write to NRCAN and ask them.  I have very little experience with
them, but whenever I have asked what I thought was a a hard question to
NGS, only a small number of times, I have gotten a really good answer
from someone whose name I recognize from reports.