Thanks Even. That content would fit perfectly in the talk... but it is only 20 min. I will see how much time I have to go into it.
On Thu, 9 May 2024 at 14:40, Even Rouault <[email protected]> wrote: > Javier, > > you might perhaps find some hints at > https://www.ga.gov.au/scientific-topics/positioning-navigation/geodesy/transform-convert > or > https://www.icsm.gov.au/sites/default/files/2021-10/GDA2020%20Technical%20Manual%20V1.7.pdf > . There are also a number of (closed) issues on PROJ tracker about that, > because that confuses everybody, but I believe the answer is that all and > none of the 3 transformations is correct, because there is no universal > answer. Mostly comes from the fact that GDA94 & GDA2020 are plate-anchored > and WGS 84 (as EPSG:4326) is a global datum ensemble with an accuracy of 2 > m. > > There are time-dependent transformations between GDA94 and WGS 84 (G1762), > and GDA2020 and WGS 84 (G1762), but for generic WGS 84, things are more > fuzzy, as one might consider that GDA94 = WGS 84 (at 1994.0) and GDA2020 = > WGS 84 (at 2020.0). > > So when you do WGS84 to GDA2020, if you are talking about coordinates with > an epoch of ~ 2020, then the null Helmert transformation is appropriate, > since by definition GDA2020 = ITRF2014 at 2020.0 ~= WGS84 at 2020 > > But if you are considering WGS84 coordinates with an epoch of ~ 1994.0, > then WGS84 ~= GDA94, and it makes sense then to propose the transformations > between GDA94 and GDA2020, hence (3) and (4). As far as I remember the > "standard" conformal transformation between GDA94 and GDA2020 is the > (non-null) 7-Helmert, and a NTv2 grid (GDA94_GDA2020_conformal) was just > derived from this 7-Helmert transformation because some software prefer > grids to Helmert transformation. The proposed WGS 84 to GDA2020 (4) > actually uses GDA94_GDA2020_conformal_and_distortion, which is a variation > of the conformal grid that takes into account some distortion (cf paragraph > 3.2.2 of the GDA2020 technical manual) > > I suspect the 3m accuracy was given so that none of the 3 transformations > particularly appear as better as the others, as ultimately the user needs > to decide what "type of WGS 84" he has. > > Just my non-authoritative guesses. > > Even > Le 09/05/2024 à 12:47, Javier Jimenez Shaw via PROJ a écrit : > > Hi > > Preparing a talk for FOSS4G Europe in Tartu (see you there if you come!), > I found that the 3 transformations in EPSG from WGS 84 to GDA2020 (2, 3 > and 4) they all have 3m accuracy. > That is not giving a clue to PROJ about which one is better. > > (2) is a null Helmert transformation. > (3) is a Helmert with 7 numbers > (4) uses a grid. > > Is there anybody from the Australian agency here? Is there any reason why > the three transformations have the same accuracy? I would expect different > values. I looks like a copy-paste issue. > > The same happens with the transformation from WGS 84 to GDA94. > > Thanks, > Javier > > .___ ._ ..._ .. . ._. .___ .. __ . _. . __.. ... .... ._ .__ > > _______________________________________________ > PROJ mailing > [email protected]https://lists.osgeo.org/mailman/listinfo/proj > > -- http://www.spatialys.com > My software is free, but my time generally not. > >
_______________________________________________ PROJ mailing list [email protected] https://lists.osgeo.org/mailman/listinfo/proj
