Effect of resolution when remapping Fesom from unstructured to regular grid

[chrisdane]

Hi

Remapping between grids is often done using cdo remap* in out. Thanks to the griddes.nc files (spheRlab), unstructured Fesom data can get easily remapped with cdo -remap* -setgrid,griddes.nc in out.

Below, the effect of the grid size of the target regular grid is shown. Every line is the difference between the global mean SST calculated on a regular grid minus the global mean SST calculated on the original unstructured grid (daily data on core mesh). The regular data was obtained with conservative remapping, e.g. cdo -remapycon,global_1 -setgrid,griddes.nc in out for a 1° regular resolution (following instructions from fesom.de).

The global mean SST is ~0.1 °C colder during winter on a 1° regular grid compared to the original unstructured data (green line). On 0.25° regular resolution, the difference reduces to ~0.03 °C (purple line). During summer, these two interpolations yield ~0.03 °C and ~0.01 °C warmer global mean SSTs compared to the original unstructured data. A regular resolution of at least 0.05° is necessary to limit the difference well below 0.01 °C.

Although the difference between original unstructured and remapped global mean SST is not large, it might be in regions of interest and/or for other meshes and variables. I think the effect of remapping should get communicated in a more transparent way, especially if Fesom data is shared with other colleagues, e.g. via ESGF.

Cheers,
Chris