Bug on linear interpolation

Olivier Marti

Normally, the wind stress is interpolated with the bilinear interpolation scheme of the ​Scrip library included in OASIS-MCT.

In IPSLCM6 at resolution ORCA1 x LMD 144x144, the bilinear interpolation of the wind stress is bugged. Two NEMO points around 0°N/180°W takes there wind stress values from LMDZ points around 0°N/0°E, and the weights, base on the distance between the source and destination points, are totally wrong..

The bilinear interpolation of OASIS-MCT compute weights and adresses that are written in a file : rmp_tlmd_to_uorc_BILINEAR.nc, rmp_tlmd_to_torc_BILINEAR.nc or rmp_tlmd_to_torc_BILINEAR.nc.

Resolutions which are know to have the bug :

• IPSLCM6 LR : eORCA1.2 x LMD144x144
• IPSLCM6 MR : eORCA1.2 x LMD 280x180

Resolutions which are know to be free of the bug the bug :

• IPSLCM5A2 VLR : ORCA2.3 x LMD96x95

A python program has been developed. It makes a few corrections of adresses and weights in a copy of the file.

The program will be available soon.

To run the correction, the steps are:

1. Run the model with bilinear interpolation of the wind stress ;

2. Check the wind stress fields on the NEMO grid ;

3. If they are wrong, get the weigths files rmp_tlmd_to_[uv]orc_BILINEAR.nc ;

4. Correct them with the program. You may need to adapt the program to the bug you observed :-(

5. Put them in the IGCM repository where the weights file are ;

6. Use MOZAIC interpolation for the wind stress, using these new weights files.

7. Check the result ;-)