(Notes by Philip Cameron-Smith (Unlicensed) based on information from Mark Taylor and Kate Evans (Unlicensed), 2014-10-31)
The SE dycore runs on cubed-sphere grids (eg, NE30, NE120). The most accurate output is obviously on those native grids. However, plotting and analysis programs often only work with rectangular lat-lon output. And ACME-CAM may not put all of the necessary information into the output files necessary for general plotting and analysis tools.
ACME is currently working on an interim solution that involves a regridding tool that regrids the ACME-CAM output files before reading them into UVCDAT: /wiki/spaces/WORKFLOW/pages/3245930
Another alternative is to tell ACME-CAM to regrid the output onto a rectangular lat-lon grid before writing it out to the output file. This uses bilinear interpolation, which isn't ideal for every purpose, but is good enough for many purposes.
This is setup in the user_nl_cam namelist, using the following namelist variables:
interpolate_analysis = .true.,.true.,.true.,.true.,.true.,.true.
interp_gridtype = 1 ! equal angle lat/lon grid, with point at pole (FV scalar grid)
interp_nlat=257 ! for a true equal angle grid, dimensions should be (N+1) x 2N
interp_nlon=512
This example will cause CAM to output onto a .5 deg grid.
The options for the output grid type are:
interp_gridtype = 1 equally spaced, including poles (FV scalars output grid)
interp_gridtype = 2 Gauss grid (CAM Eulerian)
interp_gridtype = 3 equally spaced, no poles (FV staggered velocity) Not Recommended
The interpolate_analysis options are for the different history tapes.
In the code there appears to also be an option for the type of regridding:
interp_type = 1 ! 0 = native high order, 1 = bilinear (default)
is included in Notes on tuning 1 degree ACME v0 model