raster.reconstruct does not work on global netcdfs in 0-360 longitude format

[nickywright]

Reconstructing my raster does not work if my netcdf is in 0-360 format, GPlately will import the raster OK, and say that it has reconstructed it OK, but half the world (the 180-360 portion) will be missing.

I'm guessing this is because static polygons don't align (i.e., are these in -180 to 180 format), because my data ends up with NaNs in it?

i.e.:

checking the longitudes before saving out...

[michaelchin]

Thanks @nickywright for reporting this bug,

GPlately should be enhanced to support the netcdf grids in 0-360 longitude format. And when GPlately saves the reconstructed data into a new grid file, the longitude format should be preserved.

[michaelchin]

I will look into this issue after I have updated GWS for FK and made some progress with pyGMT.

[michaelchin]

@nickywright
If it is not too much trouble, could you please send me the netcdf file and jupyter notebook? I would like to do some debugging. Thank you very much. I appreciate your time and effort in supporting GPlately Dev team.

[nickywright]

Hi @michaelchin , notebook is here, and it uses the netcdf file downloadable from zenodo here.

[michaelchin]

Hi @michaelchin , notebook is here, and it uses the netcdf file downloadable from zenodo here.

Thank you very much. I will find some time to look into it asap. I am a bit busy recently. It may take longer than usual. Sorry about that.

[michaelchin]

@nickywright
Found a temporary workaround. Set the "realign=True".

raster = gplately.Raster( data=miocene_file, time=15, plate_reconstruction=r_model, realign=True )

[michaelchin]

@GPlates/gplately-dev

This issue can be fixed by setting the "realign=True". It is False by default. Anyone knows what is the reason that it is False by default? Can I set to True by default?

[jcannon-gplates]

It does sound like internally GPlately/pyGPlates would like to use [-180,180] (instead of [0,360]). So one idea is to realign internally regardless of the realign parameter (to [-180,180]). But respect the realign parameter (eg, go from [-180,180] back to [0,360]) in save_to_netcdf4?

Ah, but Nicky's example is writing with gplately.grids.write_netcdf_grid(..., grid=reconstructed_raster._data, ...). So that idea won't work.

@brmather might have a better idea.

[brmather]

Hi all,

The most painless way forward is probably to realign to [-180,180] before reconstructing the raster then un-realign it to [0,360] after it has been reconstructed.

[michaelchin]

This bug has been fixed in master branch.

Use this notebook to verify the fix.

Further improvement to the raster reconstruction function will be tracked with #310

Close this one now.