Issue with GDA94 and GDA2020

[kkhcivil1984]

Hello to all dear experts,

Please see the ZIP file in link below which contains one photo and 3 Agisoft projects.

I used one photo and imported it in Lat/Long WGS84, GDA94 Zone 56 and GDA2020 Zone 56. I also attached their *.prj file for your consideration.

https://www.dropbox.com/sh/u0jrjvrkf1uhbky/AAC94ifq7pAqf-sWvKAAtsgGa?dl=0

As you can see, the converted Easting and Northing for GDA94 and GDA2020 is THE SAME!!. I think both actually are in GDA2020 not GDA94. Please note that I used Geoids provided in https://www.agisoft.com/downloads/geoids/. AUSGeoid09 geoid model (EPSG::5111) for GDA94 and AUSGeoid20 geoid model (EPSG::5111) for GDA 2020.

Could you please assist me in rectifying it? I need Easting and Northing to be converted to GDA94.

Regards,

KKH

[Paulo]

Hi,

actually the Easting, Northing will be the same for both as the TOWGS84 entry in both projections are set to (0,0,0,0,0,0,0).... so basically your lat long will be equal in the 3 datums (WGS84, GDA2010, GDA94) as well as in same projected coordinates...

The only difference would be in height as you use different Geoid files for GDA94 and GDA2020.

[Paulo]

Please note that I found a 7 parameter similarity transformation (3 translations, 3 rotations, 1 scalefactor) definition for GDA94 to GDA2020 or WGS84 which is equivalent  in https://www.icsm.gov.au/sites/default/files/GDA2020TechnicalManualV1.1.1.pdf. These parameters are:

[𝒕𝒙, 𝒕𝒚, 𝒕𝒛, 𝒔𝒄, 𝒓𝒙, 𝒓𝒚,  𝒓𝒛] = [0.06155, -0.01087, -0.04019, -0.009994, -0.0394924, -0.0327221, -0.0328979]

Because of sign rotation convention the TOWGS84 entry should be set to
[0.06155, -0.01087, -0.04019, 0.0394924, 0.0327221, 0.0328979, -0.009994] in your prj file....

Attached is GDA94.prj that can be used in Metashape to do transformation from GDA94 to WGS84 or GDA2020

Following are coordinates (long,lat,h) of Point Alice Springs (ALIC) first in GDA94 with correct TOWGS84 parameters and then in WGS84:

Code: [Select]

chunk.crs
Out[7]: 2020-04-08 05:03:25 <CoordinateSystem 'GDA94 (EPSG::4283)'>
chunk.crs.towgs84
Out[8]: 2020-04-08 05:03:40 [0.06155, -0.01087, -0.04019, 0.0394924, 0.0327221, 0.0328979, -0.009994]
marker.label
Out[9]: 2020-04-08 05:03:56 'ALIC'
marker.reference.location
Out[10]: 2020-04-08 05:04:15 Vector([133.88551330000004, -23.670123899999997, 603.3465999991015]) i.e. (long, lat, h)
chunk.crs
Out[11]: 2020-04-08 05:04:40 <CoordinateSystem 'WGS 84 (EPSG::4326)'>
marker.reference.location
Out[12]: 2020-04-08 05:04:50 Vector([133.88552161855534, -23.670110143743194, 603.2488552965314]) i.e. (long, lat, h)

This corresponds to example transformation given in above cited Technical Manual (3.1.1)...

[kkhcivil1984]

Hi Paulo,

Thank you so much for your response. That is really good solution.

I used it to convert the example photo to GDA94 in Agisoft. The issue is you assumed that GDA2020 is equivalent to WGS84. But, it is not. Please see the "Differences.pdf" and "Converted GDA94.jpg".

As you can see, the difference between UTM and True GDA2020 is the same as difference between Converted GDA94 and True GDA94. It means that your provided 7 parameter similarity transformation is for converting GDA2020 to GDA94 which is exactly in line with https://www.icsm.gov.au/sites/default/files/2020-01/GDA2020%20Technical%20Manual%20V1.3.pdf. But, Drone photos have Lat&Long. Then, Agisoft in fact converts to its UTM if all 7 parameters left nil. But, with your provided parameters the coordinates will be converted to very close GDA94 which has a little bit differences to True GDA94.

Please note that I used https://www.binaryearth.net/AusDatumTool/ and https://www.icsm.gov.au/datum/gda-transformation-products-and-tools/software-and-plugins to calculate True GDAs and https://geodesyapps.ga.gov.au/geographic-to-grid to calculate UTM coordinates which is exactly in line with what Agisoft gave in "GDA94 coordinates.jpg" and "GDA2020 coordinates.jpg" as provided in the folder 1 of https://www.dropbox.com/sh/u0jrjvrkf1uhbky/AAC94ifq7pAqf-sWvKAAtsgGa?dl=0.

I think to have true GDA94, we need a 7 parameter similarity transformation to convert Lat&long or UTM coordinates to True GDA94. And also, 7 parameter similarity transformation to convert Lat&long or UTM coordinates to True GDA2020. I search the Internet but couldn't find anything. Could you please help me with that? Many thanks.

Best regards,

KKH

[Paulo]

Hi KKH,

this a typical geodetic problem...When we say latest realization of WGS84 is equivalent to GDA2020, it means somewhere at the cm level as datums are dynamic. From cited technical manual:

The WGS84 coordinates of tracking stations used to
compute the GPS broadcast orbit are adjusted annually for plate tectonic motion to an epoch
at the half year mark, e.g. WGS84 as used in the GPS broadcast orbit during calendar year
2014 was ITRF2008 at 2014.5. Consequently, differences between the ITRF2014 and WGS84
are negligible for most users.

and note that GDA2020 is a realization of  ITRF2014 datum at epoch 2020.0 (january 1st 2020)
So it is normal that you may see about a cm difference.

And note that the transformation tools or plugins you used have 2 options:
- conformal option grid which is same as 7 parameter similarity provided;
- conformal + distorsion grid option (more precise) which is not the same as 7 parameter.
From AusDatumTool

Converts between GDA94 and GDA2020 using either the 7-parameter similarity transformation described in the GDA2020 Technical Manual, or the conformal and distortion grid shift file for maximum accuracy.

Hence this could explain the few cm difference you encountered if you used the 2nd transformation option. Note that the 7 parameter similarity transformation will never be as accurate as combined conformal and distorsion grid but such grid transformation requires more elaborate implementation (NTV2 grid transformation) not typically present in most datum transformation. But in fact using a base station  whose reference coordinates are defined in GDA2020, would have your precise RTK image center coordinates automatically given also in GDA2020 and no problem from there. Same if you used a few GCPs also defined in GDA2020 to optimize your drone survey...

Maybe this clears up your doubts...

For example, here are the coordinates in Metashape of image 100_0006_0047.JPG first in WGS84 and then in GDA94 / MGA 56 with correct TOWGS84 entry:

Code: [Select]

chunk.crs
Out[16]: 2020-04-10 15:28:37 <CoordinateSystem 'WGS 84 (EPSG::4326)'> equivalent to <CoordinateSystem 'GDA2020 (EPSG::7844)'>
camera
Out[17]: 2020-04-10 15:28:41 <Camera '100_0006_0047.JPG'>
camera.reference.location
Out[18]: 2020-04-10 15:28:47 Vector([151.22995575, -33.6879153333333, 159.39]) i.e. (long, lat, h)
chunk.crs
Out[19]: 2020-04-10 15:29:51 <CoordinateSystem 'GDA94/ MGA zone 56'>
chunk.crs.towgs84
Out[20]: 2020-04-10 15:29:58 [0.06155, -0.01087, -0.04019, 0.0394924, 0.0327221, 0.0328979, -0.009994]
camera.reference.location
Out[21]: 2020-04-10 15:30:07 Vector([335934.3005454107, 6271039.124106831, 159.48522494253643]) i.e. (East, North, h)
And compare to results from AusDatumTool in attached screen copy... they are the same rounded to nearest mm (without using NTv2 grid files option)...