Australia’s geocentric datum is pretty boring, because Australia’s tectonics are pretty boring. We just don’t get large earthquakes. So the GDA94 → GDA2020 update is actually pretty minor.
If my understanding is correct, the 1.8m difference sometimes talked of between ITRF92 and GDA94 by 2020 is actually irrelevant so long as you do proper projection of your coordinates, which you should, but some things don’t—and so tweaking your reference point from time to time to compensate is pragmatic (in part because it lets people skip deformation models where a metre or so of accuracy is adequate, so temper my “should”). The GDA94 → GDA2020 update is more about certain changes in ITRF1992 → ITRF2014 (~9cm changes in ellipsoidal heights), and local crust deformations. Mass updating from GDA94 to GDA2020 is an easy reprojection.
In saying GDA94 is actually fine as far as continental drift is concerned: the thing most people don’t realise is that all of these coordinate systems are actually time-dependent transformations: they’ve got forecast continental drift baked in. So later datum updates just need to record what actually happened, if it was different enough (which it will be in places).
Now real earthquakes—they make things much more interesting. I like what https://www.linz.govt.nz/guidance/geodetic-system/coordinate... says:
> Some earthquakes, such as those of the Canterbury earthquake sequence starting in 2010, have caused metres of movement. Where this has happened we have updated the coordinates rather than simply include the movement in the deformation model. This is necessary as otherwise the coordinates will not be accurate enough for many applications. The deformation model still includes the earthquake deformation, but it is applied in reverse to transform coordinates for dates before the earthquake.
The deformation model page https://www.linz.govt.nz/guidance/geodetic-system/coordinate...> is also rather interesting, giving details of when they’ve released new deformation models, including both forward and reverse patches due to earthquakes, and it’s messy. I haven’t thought too deeply about the reverse patching technique (this isn’t a domain I work in) but I don’t think I like it (though it may be pragmatic) because it messes with epoch-era NZGD2000 points, seems to undermine the purpose of a datum unless you have recorded the coordinate times as well (rather than merely transforming to the epoch, which has now been spoiled). Did have a fun chat with a telecommunications field worker that I happened to meet when I was in New Zealand in 2021, who had a lot of interesting experience after the 2016 Kaikoura earthquake. Lots of stuff I’m not used to thinking about, coming from boring old Australia.