search

uwefladrich / rdy2cpl

A Python tool to create auxiliary files and remapping weights for the OASIS coupler
MIT License
3 stars 3 forks source link

Consistency between masks in Nemo (NE{T,U,V}M) and OpenIFS (IOLL) grids #20

Open jmrgonza opened 1 year ago

jmrgonza commented 1 year ago

Comparing the masks for the Nemo grids to that of the IFS, I noticed that the Black sea is masked in the OASIS definitions for Nemo, but not for IFS. I am not sure if this may cause some problems in the coupling.

Also, but I am not sure about that, it seems to me that the ifs mask does not follow the oasis convention (0 for unmasked values, 1 for masked), or maybe I am looking at the wrong grid.

I am getting the grid information in a python session, with a call to from_model_spec(grid_name). I have not checked the values written to the masks.nc file. It is possible that there is some arithmetic in between that I am unaware of.

jmrgonza commented 1 year ago

This pictures shows the differences in the sea masks for Nemo and OpenIFS. It was created using the corners information for each grid cell. Grey means that both models are unmasked. Blue means that Nemo is masked, but not OIFS. Red means that OIFS is masked, but not Nemo. I used the IOLL grid for IFS, but I had to reverse the IFS mask.

image

uwefladrich commented 1 year ago

Hi @jmrgonza, thanks for the report!

The definition of the couple grids is still under a lot of development. For example, the masking of the Black sea in the NEMO grids has recently been removed.

Also, I think that the masks you are looking at have been computed with a fall-back option in rdy2cpl, which is not what we are using in ECE4 and not well tested (i.e. probably buggy). In ECE4, we are reading the NEMO masks from the maskutil.nc file, which should have better masks. Please check this MR for the latest developments.