Direkt zum Inhalt | Direkt zur Navigation

Personal tools

Sektionen
You are here: Home » FAQ & Forum » OASIS3-MCT forum » Transformations and interpolations with the SCRIP library in OASIS3-MCT » Conservative remapping for variable resolution unstructured grid

Conservative remapping for variable resolution unstructured grid

Up to Transformations and interpolations with the SCRIP library in OASIS3-MCT

Conservative remapping for variable resolution unstructured grid

Posted by Anonymous at March 22. 2018

Good afternoon,

I am coupling a reduced Gaussian grid atmosphere (OpenIFS) with an variable resolution unstructured grid ocean model (FESOM2). I would like use OASIS3-MCT conservative remapping for the atmospheric fluxes.

Does oasis_write_corner work with an unstructured mesh? I calculate the "corners" around a mesh point as the center points of adjacent scalar volumes. In the open ocean we have about 6 adjacent scalar volumes (can be different between grid points). Near the coast this can go down all the way to 3 in narrow inlets. In areas of resolution change we can have as many as 8 adjacent scalar volumes and subsequently 8 corners. I am giving oasis_write_corners the maximum number of corners (8) in the grid and the missing corners (if less than 8) are replaced by repeating the first corner. I have gotten no complains from oasis_write_corners after doing this.

I have however run into an issue where the flux values received by the ocean are many orders of magnitude bigger than those send by the atmosphere. I suspect there might be a connection with the corners messing up the conservative remapping? For a test case I send the value 1 everywhere and receive 2858036804.74055 on the ocean side.

My namcouple configuration is:

# --- Stresses for oce and ice --- [ifs fesom]

A_TauX_oce:A_TauY_oce:A_TauX_ice:A_TauY_ice taux_oce:tauy_oce:taux_ico:tauy_ico 1 2700 3 rstas.nc EXPORTED

A128 feom LAG=0

P 0 P 0

LOCTRANS SCRIPR CONSERV

AVERAGE

GAUSWGT D SCALAR LATITUDE 1 9 2.0

GLOBAL

Best regards, Jan Streffing

Re: Conservative remapping for variable resolution unstructured grid

Posted by Anonymous at March 22. 2018

Hi Jan,

- If you want to perform a conservative remapping you must use in your namcouple (see the doc pp. 37): CONSERV D SCALAR LATITUDE 1 FRACNNEI FIRST

- You must provide grids.nc, masks.nc and areas.nc (see the doc pp. 16)

- The option CONSERV after SCRIPR will redistribute on the target grid the difference of the fuxes integrated on the source and the target grids (see the doc pp. 39 for the options of CONSERV)

- When you do remapping with gaussian reduced grid as a source grid, you must put north_thresh = 2.00_dbl_kind and south_thresh =-2.00_dbl_kind in oasis3-mct/lib/scrip/scr/remap_conserv.F and recompile OASIS3-MCT

- You can use oasis_write_corner with an unstructured mesh - Your "volumes" must have no holes and cover the entire domain of simulation.

By the way how do you define the adjacent scalar volumes ? From the other points ?

Best regards, Laure

Re: Conservative remapping for variable resolution unstructured grid

Posted by Anonymous at March 22. 2018

Hey Laure, thank you for the input.

I changed north_thresh which was 1.45 before and replaced the scripr options. That seems to have done the trick. I have not yet verified that the fluxes are acutally conserved but the model no longer crashes due to erroneously large values. Progress!

I have one more question. Could you elaborate on the function of north_thresh / south_thresh? You said I must use +-2.00. I also work on EC-Earth, and the OASIS source code I used was taken from the EC-Earth Model. There we do conservative remapping with a reduced gaussian source grid and we have north_thresh = 1.45_dbl_kind and south_thresh =-2.00_dbl_kind. Is that something we should worried about in the EC-Earth conserv remapping?

Cheers, Jan

Re: Conservative remapping for variable resolution unstructured grid

Posted by Anonymous at March 22. 2018

Hi Jan,

north_thresh=+2 rad corresponds to lat=90° and south_thresh=-2 rad corresponds to lat=-90°

The Lambert projection used for latitudes above north_thresh  (south_thresh) give false results for the Gaussian reduced grid as the borders of the cells do not coincide in the projected space; this is because a corner of the cell is not necessarily a corner of the neighbour cell on a higher or lower latitude band, see attached picture CoinsGaussienneReduiteProjectionLambert.png. So you have to put north_thresh > 1.571 (=2) so to deactivate the projection in EC Earth.

Best regards, Laure

Re: Conservative remapping for variable resolution unstructured grid

Posted by Laure Coquart at March 22. 2018

Picture CoinsGaussienneReduiteProjectionLambert.png

Attachments
Reply to this
URLs will be automatically hyperlinked. Basic HTML tags are OK.
Powered by Ploneboard
© Copyright ENES Portal 2011