Add support for Canadian Geoid

[mbremner]

It would be helpful for Canadian users if DynAdjust had support for the geoid used in Canada.

I have converted the CGG2013an83 geoid to a .gsb file. Deflections of vertical are computed by taking the derivatives in the north and east directions of a bi-quadratic interpolation if each grid point and the surrounding 8 points. The first and last rows and columns have been removed as they do not have surrounding points.

The file is pretty large, around 200 mb. I am not sure if it should be submitted through a pull request. Is there a way to send you this file? If so, would it be possible to post it as an asset so that interested user could download it?

[rogerfraser]

In principle, geoid has been designed to support interpolation from NTv2 formatted geoid models regardless of location or source. However, very little testing beyond Australian geoid files has been undertaken. I would be glad to hear how geoid behaves with CGG2013an83.gsb (and also your EGM2008.gsb file mentioned here), and to know what changes might be needed to ensure universal geoid interpolation capability from any (NTv2-based) gsb file.

I suggest we keep this issue open until we're satisfied universal capability is provided.

On the subject of storage and access, our preference is to not store geoid files in this repo, but to direct users to 'point of truth' locations from where they can obtain the latest authoritative version(s). For example, Australian geoid files (AUSGeoid2020) can be found here: https://www.ga.gov.au/scientific-topics/positioning-navigation/positioning-australia/geodesy/ahdgm/ausgeoid2020

Is it possible for Canadian .gsb files to be made publicly available on an authoritative Canadian site?

[mbremner]

I have done some testing to compare the results of the DynAdjust Geoid software, using this geoid file, against the NRCan geoid interpolation software GPS-H. For this test I randomly generated coordinates for 50 points in and around North America, then computed the geoid undulation and deflections of vertical for each point with both tools. The largest difference in geoid undulation was 3mm and the largest difference is deflection of vertical was under 1 second. However, most stations had a differences 0mm for undulation and under 0.1 seconds for each deflection.

It is my understanding that GPS-H uses bi-quadratic interpolation for the undulations and computes the deflections from the derivatives of this bi-quadratic interpolation. DynAdjust Geoid uses bi-cubic interpolation of both the undulations and the deflections that I computed (using the GPS-H method).

Given this difference in methodology, it seems to me that the differences I observed are within reason. It seems to me that DynAdjust Geoid program is working correctly outside of Australia, at least with this geoid file.

I have also spoken with the team that currently hosts the current CGVD geoid files on NRCan's website and it seems that we should be able to also host this .gsb file there.

[rogerfraser]

Thanks @mbremner for this update and the testing you've undertaken to confirm expected behaviour. Thanks also to NRCan for offering to host the geoid file. It sounds like we have a solution to this issue.

Rather than close this as an issue, we'll convert this to a discussion so that we can monitor any further questions and experiences of others who might be interested in testing other geoid grid files and grid file formats (e.g. GeoTIFF and GGXF).