`GLORYS` field inpainting does not terminate because of bathymetry at the deepest sections

[xkykai]

When I run something like

T_GLORYS = Metadata(:temperature; dataset=GLORYSMonthly(), start_date, end_date)
T_GLORYS_data = FieldTimeSeries(T_GLORYS, grid)

the code does not terminate as some locations can never be inpainted. So I run with inpainting=NearestNeighborInpainting(200) to see where the fields cannot be inpainted.

Below I show a video of a sample field where I've painted locations where it is not inpainted (where value is exactly equal to zero) with magenta:

GLORYS_T_inpainting.mp4

Here we see that it is at very deep trenches where the data cannot be inpainted. A similar story is found when looking at the salinity fields as well:

GLORYS_S_inpainting.mp4

I'd guess that this is because a bathymetry issue. I don't have an idea of how to fix it yet (except maybe inpainting from above in the vertical under certain conditions).

[glwagner]

this seems like a bug in the function continue_downwards! which if I recall correctly is supposed to fill regions that are below the bottom, vertically, before horizontal inpainting is initiated.