Croco and curvilinear grids

[soizicg]

Question

I use a ROMS forcing, where a variable "angle" keeps track of the grid's rotation for each hourly output. My question is do I need to provide Parcels with the rotated velocity vectors provided in the ROMS output or the zonal and meridional velocities?

The documentation gives a warning when writing kernels for vector fields on curvilinear grids, mentioning that Parcels automatically rotates the U and V field when necessary. How is Parcels able to rotate the velocity fields from curvilinear grid? Is it because it takes the f-nodes for the grid coordinates?

I have been playing with the new Fieldset.from_croco function to read my ROMS forcing, and I noticed that in the documentation example it takes the rho coordinates instead of the f-node. Does it makes a difference when the grid is rotated?

Thanks for the help.
Soizic

[erikvansebille]

Hi @soizicg, thanks for asking your question. Croco support is quite new, so please always double-check if the results make sense. But yes, in principle Parcels doesn't need the angle field because the rotation is done in the Curvilinear C-grid interpolation. See also section 2.1.2. of this paper: https://gmd.copernicus.org/articles/12/3571/2019/gmd-12-3571-2019.html

As for the f-points versus the rho-points; that may then indeed be a mistake. Parcels expects the locations of the southwest/lowerleft corners of the grid.

[soizicg]

Hi @erikvansebille, thanks for your answer, it makes sense that it is not any different for the croco implementation of Parcels.