Thanks for taking some time on this. I have to check what the update does. I remember I spent some time in dealing with this special case (zoom crossing boundaries and interface separation), but I don’t recall now with what I actually came up in the code. If you plan to improve your experiment on the Agrif side, I can provide you a few different options than the existing ones.
Hi,
I checked in VORTEX where the update on parent is done, with interface separation and “fancy” grids. Actually, in your case (zoom crossing an E-W cyclic boundary), there is no problem. So good news, go ahead with it. However, in the case of a cyclic E-W grid, exactly matching parent E-W dimension, well, there’s a problem. See plot.
Thanks for double-checking!
Is this only a problem when the zoom starts and ends with the E-W cyclic boundary or in any “circumpolar” zoom? I mean, if the zonal dimension of the cyclic parent is let’s say 1000 is this only a problem if one defines the zoom from 1 to 1000 or also if it is defined from e.g. -500 to 500?
By circumpolar, I meant any zoom starting and ending at the parent E-W cyclic boundaries (and therefore having itself E-W cyclic boundaries too). E.g., something like this (taken from Patara et al, 2016), but with no discontinuity 
:
This is the only case that will imply discarding “interface separation”. Actually, It is not that simple to discard with AGRIF interface separation for selected boundaries (in that case indeed, we would like to keep interface separation near the northern and the southerm boundaries).
That’s exactly the configuraton I was thinking about; we’ve been discussing to close the discontinuity in an updated NEMO 5 version of it.