Check out a 2021 VERY comprehensive thesis: ""Investigation into the accuracy and practicality of methods for transforming coordinates between geodetic datums (PhD thesis, book-format edition)" by Andrew Ruffhead "
This has some really interesting analyses with included data sets. Good read. Clifford J. Mugnier, c.p., c.m.s. Chief of Geodesy, Emeritus LSU Center for GeoInformatics (ERAD 266) Dept. of Civil Engineering LOUISIANA STATE UNIVERSITY Baton Rouge, LA 70803 Research: (225) 578-4578 Cell: (225) 328-8975 honorary lifetime member, lsps fellow emeritus, asprs member, apsg ________________________________ From: PROJ <[email protected]> on behalf of Greg Troxel <[email protected]> Sent: Friday, November 26, 2021 11:35 AM To: Mathieu Poulin <[email protected]> Cc: [email protected] <[email protected]> Subject: Re: [PROJ] Change of datum not working Mathieu Poulin <[email protected]> writes: > When creating a transformation that requires a change of datum, it > does not work. For instance, going from Geographic coordinates in > WGS84 to UTM coordinates in NAD83CSRS. It does convert from geographic > lat,lon to UTM projection easting,northing but the datum remains > WGS84. I think you mean "the coordinates are those I'd expect if the datum transformation did not happen". This has been discussed a lot. Briefly, WGS84 is a name for a family of datums, including WGS84(TRANSIT), the original, and the original is considered equivalent to NAD83(1986). And, no one (outside the US DoD) has high-accuracy coordinates that are truly in a WGS84 realization, because there is no high-accuracy way to access the datum. The high-accuracy techniques (carrier phase vs CORS, RTK) are all relative to either national static datums, or something in someting like ITRF2014 or IGb14. There are lots of philosophical points about ensembles but my practical advice is: Surely your data is not actually in WGS84(ensemble). Beware that terminology is fuzzy with WGS84 the bare word as to whether people/programs mean: - the realization currently in use - the ensemble - the original member of the ensemble If your data was obtained from non-differential GPS observations, it would be in some specific realization of WGS84, depending on the date. If it used WAAS, it would be in the WAAS frame instead and not actually in WGS84. ITRF2008 is a good proxy for relatively recent WGS84 realizations, and I think what WAAS uses. (ITRF2014 is a good proxy for measuremnets since early 2021, in WGS84(G2139), but if you can tell the difference beteen actual measurements in ITRF2008 and ITRF2014 please tell us how :-)) Keep in mind that ITRF2008 coordinates of points in Canada have non-zero velocities. So you should be thinking about epoch if you want to be cm-level. If you are trying to avoid the meter-level shift and ignoring the ~2-3 cm/year over 10 years, that seems like an intermediate reasonable approach. (FWIW, I deal with coordinates in NAD83(2011) epoch 2010.0 obtained via RTK that are accurate at the 5-10 cm level, but once I move to any dynamic frame things get messier.) If your data is not accurate at the meter level, I would say it still makes sense to transform properly form NAD83 to ITRFfoo to match others and handle round-tripping, but the finer points probably don't make sense to worry about. You said NAD83(CSRS) but didn't mention which of the 7 realizations of that you are talking about. However I'm pretty sure those are intended to be static and very close the the US NAD83 realizations, and close to each other at more or less the 10 cm level. and thus my real advice is: relabel your "WGS84" data as being ITRF2008. Perhaps by a transform, which will be null, or perhaps just label it that way. Then, conversion to NAD83(CSRS) should do the datum transform. If you are still getting null transforms, transform to NAD83(CSRS)v7 instead of the bare CSRS word, which might be an ensesmble, and might be the original. A related issue is that with coordinates in Massachusetts State Plane, defined in terms of NAD83, converting to ITRF2014 can end up null, but if I convert first to NAD83(2011), which is a null transform, then the transform to ITRF2014 happens.
_______________________________________________ PROJ mailing list [email protected] https://lists.osgeo.org/mailman/listinfo/proj
