Projecting "rotated pole" coordinate system (from netCDF)

[NathanElliott]

Hello-

Neither ArcGIS 9.3x or 10 seem to handle the importation of a netCDF file with a "rotated pole" coordinate system, listing the coordinate system as "unknown" in the data source tab.  Barring an automatic fix allowing the program to correctly read these data, how would I go about manually defining a projection so that they display properly?

The projection metadata provided in the netCDF file are as follows:
   char rotated_pole;
     :grid_mapping_name = "rotated_latitude_longitude";
     :grid_north_pole_latitude = 42.5; // double
     :grid_north_pole_longitude = 83.0; // double
     :_CoordinateTransformType = "Projection";
     :_CoordinateAxisTypes = "GeoX GeoY";

This is not enough information, in and of itself, to define a specific projection, which is probably why ArcGIS is having the problem.  Looking at the CF Metadata convention (specifically Appendix F), it appears that a third parameter of north_pole_grid_longitude defaults to 0 if not specified. 

I think these three parameters should be enough to manually define the projection, but I'm unsure of how to do so.  In the 'Define Projection' tool, what projected coordinate system should I select as a base?  There are no "rotated_pole" or "rotated_latitude_longitude" or "arctic somethingorother"...

Oh, and the purpose of defining the coordinate system is more so that I can project other things into the same coordinate system (for example a polygon representing the extent of the studyarea so that I can clip the netCDF data).

I apologize if this question has been asked/answered elsewhere, and thanks for your time!

Sincerely,
Nathan Elliott

[MelitaKennedy]

Hi Nathan,

ArcGIS doesn't support a geographic coordinate system with a rotated pole. We have one projection, Krovak oblique conic, which defines a new pole as part of the projection parameters, but that's not what you need.

It might be possible to use a custom datum transformation. The coordinate frame (7 parameter) transformation method works in XYZ space and supports 3 rotations.

I need to discuss this with a developer. What sphere or ellipsoid model is being used? That will affect the transformation values.

Edit: Actually it won't affect the transformation values per se, but the results. I'm now getting values that are in the right quadrant, but not matching some test points that I found. They may just be incorrect because I'm using WGS84 instead of a sphere. Can you identify some positions that I can use to verify my results?

Thanks,
Melita

[NathanElliott]

Hi Melita-

Thanks for your quick response!  I'm afraid there is no datum or spheroid information with the file metadata.  I would have assumed it's geographic, but I'll see if I can get in touch with the originators and find out.

The longitude values in the rotated grid range from -33.880 to 33.880 in increments of .440 degrees, symmetrically centered on 0.

The latitude values in the rotated grid range from -28.380 to 28.380, also in increments of .440 degrees and symmetrically centered on 0.

See attached for a table of congruent locations.  I had to pull them together manually, and though I double-checked them, I'm afraid it's still possible that I made some transcription errors.  Please let me know if you need anything else or come across something weird.

Thanks again for your help!

Nathan

[DouglasWardell-Johnson]

Sorry if its too late, but might be useful to others, http://gis.stackexchange.com/questions/10808/lon-lat-transformation/14445 gives a description on the process for rotating a rotated pole grid, using some trig functions, I used this (via a python script) to convert data from a netcdf as the OP was attempting to do from a rotated pole grid to its "unrotated" form, hope this helps!

[RHirner]

There's actually a way how to use rotated-pole raster data in ArcGIS using custom transformations: Importing rotated north-pole GRIB files into... | Dev001

[ZacharyKlaas]

Maybe someone can explain to me why a request was made about this eight years ago, and there still has been no action to integrate this into the Multidimension Tools?  NetCDF data uses this rotated pole projection a lot, and there's no capacity for bringing it in?

[MelitaKennedy]

It's been on my list of enhancements; has never made it to the top. I find the documentation on it confusing, but that's a personal fault. Someone else on my team could figure it out. I think it's being done incorrectly, as I think the rotated pole is assuming a sphere, not an ellipsoid, so we'd have to do some research on what that really means.

Melita

[KatyAppleton1]

Is there any improved capability in Pro?

I am trying to work with UKCP18 climate projection data, which uses a rotated pole projection for the finest-resolution data, and I'm just making zero progress. The blog linked above is 404, but accessing it via the internet archive shows that it relates to GRIB files (not NetCDF) and seems to refer to other workflows outside ArcGIS. In any case, I cannot get Define Projection to work on a NetCDF file, so being able to reproject those data with a custom transformation is a non-starter.

There's tons of potentially very useful data out there but it seems to be unusable in ArcGIS, which is highly frustrating.

[SergeyKozhevnikov]

Hello dear colleagues,
It is going to be very important, since ArcGIS Pro started to work with 3D voxel model. Most of block models from mining industry has rotation angle, and It is impossible to transform coordinates from local grid to one of international grids without rotation. So, the implementation of the projection of rotated pole in ArcGIS would be the best solution.