[Proj] Realigning velocity grid values by Helmert shift - is that normal?

[Kristian Evers]

All,

In the ongoing effort of improving geodetic capabilities in PROJ.4 I've run into what I think is an unusual situation. Before I dismiss it as such, I would like to check with the list if this is more common than I think and it should receive special attention in PROJ.4. The topic is realignment of a gridded velocity model with a Helmert transform.

I want to do transformations from ITRF2014 to ETRS89 in Scandinacia, using the transformations defined by the Nordic Geodetic Commision (NKG). The transformations are described by Häkli et al in [0]. For this transformation to work in PROJ.4 a velocity model operation is needed, which I intend to implement. Additionally, in this particular transformation a realignment of the velocity model is made with a Helmert shift. I am trying to determine the best way to handle this, hopefully the response to this post will help me do that. First a bit of background.

Because of the post glacial uplift in the Scandinavian countries the usual EUREF transformations between various ITRFs and ETRS89 is not accurate enough, and hence the NKG has defined its own transformations from the various ITRF's to the local Scandinavian/Nordic realizations of ETRS89. All transformations go through a common reference frame, NKG_ETRF00, making it easier to bring data from the different countries into the same reference frame.

The NKG_ETRF00 is defined as ETRF2000 at epoch 2000.0 adjusted for the post glacial uplift. The adjustment is done by applying a velocity model. This velocity model is a grid-based model originating from a GPS campaign in 2003, whereas the transformation described in [0] is based on a campaign from 2008. The results from the 2008 campaign reveals that the velocity model is not a good fit for the new data, and hence an adjustment is made to the model. In anticipation of a new velocity model being released soon, the original model is instead used by applying a Helmert shift to the gridded velocities. That is, the individual grid-values are realigned with a Helmert shift.

I have not come across a solution like this before, so it strikes me as being a bit odd. It's a clever solution to the problem, but not very easy to work with in a generic transformation software as PROJ.4. It works out fine mathematically and in custom scripts, but I find it a bit hard to incorporate into the transformation pipeline framework, since it is parameters that are being transformed, and not coordinates. The simplest solution is to just create a new gridded model where the correction has been applied, and using it with a "velocitygrid" operation. Another option would be to incorporate the Helmert shift into the velocitygrid operation so that the realignment can be performed when loading the grid. A third option is to introduces some a form of nested operation in the pipeline framework.

I don't want to make things more complicated than they need to be, so I will only consider option two and three above if the concept of "realigning gridded velocity models by Helmert shift" is a somewhat common operation. I have not come across a similar transformation before. Do other transformations that use a similar operation exist?

Best regards,
Kristian

[0] https://www.degruyter.com/downloadpdf/j/jogs.2016.6.issue-1/jogs-2016-0001/jogs-2016-0001.pdf
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/proj/attachments/20170914/1fabaf20/attachment.html>