[PROJ] [gdal-dev] Static/Dynamic datum problems

[Even Rouault]

> I believe that the problem is that WGS84 is being used in webmapping
> situations as if it were a static datum, locked to a fixed epoch, and in
> other cases it is treated as if it were a dynamic datum.
> 
> This has led to the situation we have in Australia where:
> 
> WGS94 -transform-> WGS2020 (leads to a coordinate shift of ~ 1.8 metres)
> WGS94 -null transform-> WGS84 (which was correct in 1994)
> WGS2020 -null transform-> WGS84 (which will be correct in 2020)

You likely meant GDA94 (instead of WGS94) and GDA2020 (instead of WGS2020).

Yes, indeed if you look in the EPSG dataset, and transform between GDA94 and 
EPSG:4326 (the "average" WGS84 CRS), you'll get a null transform, but 
advertized with a 3m accuracy only, whereas direct transformation between 
GDA94 and GDA2020 has 1 or 5cm accuracy depending on which method is taken. 
And same for GDA2020. So there is no transitivity in chaining transformations.

The thing is that EPSG:4326 is really a not super accurate CRS, given that its 
underlying datum is actually a "datum ensemble" in the modern geodetic 
terminology:
http://docs.opengeospatial.org/as/18-005r4/18-005r4.html#53

If you look at the candidate for the new CRS WKT revised standard
https://portal.opengeospatial.org/files/81176
you' ll see this

	ENSEMBLE["WGS 84 ensemble",
	  MEMBER["WGS 84 (TRANSIT)",ID["EPSG",1166]],
	  MEMBER["WGS 84 (G730)",ID["EPSG",1152]],
	  MEMBER["WGS 84 (G834)",ID["EPSG",1153]],
	  MEMBER["WGS 84 (G1150)",ID["EPSG",1154]],
	  MEMBER["WGS 84 (G1674)",ID["EPSG",1155]],
	  MEMBER["WGS 84 (G1762)",ID["EPSG",1156]],
	  ELLIPSOID["WGS 84",6378137,298.2572236,LENGTHUNIT["metre",1.0]],
	  ENSEMBLEACCURACY[2]
	]

And I've heard that EPSG might consider to revise their future dataset to use 
that datum ensemble as the base for CRS EPSG:4326 to better reflect this 
reality, instead of basing it on the datum EPSG:6326, which is not really well 
defined.
So mapping to EPSG:4326 makes only sense if you don't need accuracy lower than 
2 meters.

You might pick up one of the WGS 84 realization, and use the time-dependent 
transformations I mentionned in my previous email to go from GDA94 or GDA2020 
to that realization.

Even

-- 
Spatialys - Geospatial professional services
http://www.spatialys.com