CDO

Bjorn Stevens wrote:

I fear that this is a silly question, as a uniform tiling of the sphere is such a difficult problem, ut does CDO have an option to remap onto an equal area grid? Two possibilities I am interested in:

1. Specifying an ICON grid (R2B10) and map other models to that (as a nearly equal area)

if you have the target grid including cell bounds, CDO will be able to do any interpolation

2. Using reduced longitudes, i.e., ISCPP has long put its data on such a grid: https://isccp.giss.nasa.gov/docs/mapgridinfo.html

it depends in which format these grids appear. if coordinates AND bounds are given, CDO should be able to handle this.

The reason I ask is that for histograms of variables one would like to avoid over-sampling high latitudes.

in terms of grid area the R2B10 grid is quasi uniform - why not using the input data directly from the icon grid?

cheers
ralf

Thanks for the feedback. For ISCCP (not ISCCPP, my typo) they mostly worked with fortran binary and ascii and I have not looked to see how, or if, that has changed. Yes for ICON I am using the native data, but for other models I would like to sample on something similarly quasi-uniform, and I don't really understand how to handle grids. This also comes up in the DYAMOND processing... but that is another thread. Increasing the family (library) of grids that CDOs can work with seems like a good use of time, if (as I suppose) it is a one time sort of effort.

it all depends on how the grid description is done. I searched for something on the website you mentioned above, but could only find large asci files: https://isccp.giss.nasa.gov/pub/tables/ISCCP.D1GRID.0.GLOBAL.1983.99.99.9999.GPC. This is nothing I would like to teach CDO to understand, though. if you come up with some netcdf file it will be another story.

You could use the ICON grid as an intermediate grid for the other models if it is uniform enough for your purpose. would that be an options?

hth
ralf