Cube extents and importing other datasets

I know a related question has already been asked about deriving lat/lon values for the cube, but I’m encountering an issue with trying to aggregate other datasets to to the ESDL grid.

I’ve got access to a dataset containing the location of landforms globally at a 1/12 of a degree and so this should match the ESDL grid as both contain 4320 longitudinal values (i.e. 12*360 values). As far as I can tell however, the ESDL extent isn’t exactly -180 to 180, but rather is very close (e.g. -179.99 - 179.99). This problematic as rather than grid cells aligning perfectly, they’re misaligned (as the landform dataset I’m using is exactly -180 to 180), and because both datasets contain 4320 longitudinal values they’re not offset uniformly, but rather this varies (as the ESDL pixel size is therefore slightly smaller than exactly 1/12 of a degree). This is confirmed when exporting the ESDL cube as a .nc file and then viewing the co-ordinate transform parameters, which are listed as:

(0.08333299888121527, 0.0, -179.9999947465109, 0.0, -0.08333299998168098, 90.00000237645568)

I’ve included an image below to show this - the black pixel outline is an ESDL grid cell and the other colours are pixel values for the landform dataset. If anyone can shed any light on this it would be much appreciated! Particularly as to whether the ESDL grid was created from -180 to 180o and the coordinates have now been misaligned, or if aggregation to the grid was always using this slightly smaller pixel size/value extent as this has implications for including other variables in the cube.

Thanks in advance!

Hello Obaines

Yes, the numerical precision of the spatial grid is 10^-5 deg as you can also see in the tuple you have printed.
That is in your case seemingly insufficient. We shall consider your experience in the new cube version we have started to prepare.

If your issue is urgent, you can tell me what variables you are interested in and I run the cube generation with a sufficient precision and put it into the supplementary cube directory. Well, we might do that anyway.

Kind regards and Nga mihi


Hi Helge and Oliver,

this is actually only a problem of wrongly labelling the grid cells. I think the underlying data is actually processed in a way that it matches the grid by @obaines . The problem with the bad labels can be solved once ths PR is finished (sorry for taking so long), because it adresses a similar problem that @felixcremer had with his high resolution cubes.


Hi Helge and Fabian,

Thanks so much for your help - I thought maybe I was going wrong somewhere re grid extents! If it’s just a labelling issue and the underlying data were created properly then I’ll just recreate the datasets I was going to ingest using the full -180 to 180 grid.

Many thanks!