Question about the global mean value using cdms2.regrid

[leiduan4320]

Dear all,

I have a question regarding the "regrid" function in cdms2.

When I regrid the global-scale climate variable (e.g., surface air temperature, precipitation) from a fine input grid to a coarse destination grid, how will the global mean value be treated?

I tested with some model outputs. The original grid resolution is 160° by latitude and 320° by longitude, and a series of Gaussian grid (64x128, 32x64, 10x20, 5x10) generated by cdms2. I compared results using "linear", "patch", and "conservative" using "regidTool = ESMF".

My understanding is that for each data point in the destination grid, the linear regression only makes use of the surrounding four data points and thus the global mean value should change, whereas the conservative method is able to keep the global mean value consistent with the original value.

However, it seems to me from the above results that the "linear" approach produces quite consistent results, both for spatial pattern and global mean value, especially for the last two cases (10x20 and 5x10).

I tried to search for documents (https://cdms.readthedocs.io/en/latest/manual/cdms_4.html) but haven't got much useful information yet. Can someone give me some hints or point me to the right link? Thanks very much!

Best,
Lei

[jypeter]

@jasonb5 this reminds me that there are lots and lots and lots of nice examples in the vacumm gallery , including many regridding examples

And vacumm relies on cdms2 and friends. I think it also relied on vcs a long time ago, but vcs become too hard/unreliable to use when it switched to vtk (that's also the reason why I haven't used vcs in a long time, even though it has so many nice and efficient defaults, because I don't like notebooks too much)

I think there are several issues in the many CDAT/* packages mentioning vacumm (e.g. #236)

If only the CDAT web site could be as complete as vacumm's!