esmpy integration documentation/notebook

[doutriaux1]

As @bonfils2 was telling me this week. It is primordial that users understand what is going on.
@bonfils2 for example needs to be able to get the weights back so that she can get exactly the same mean after regridding than before, even when masking.
@dnadeau4 I'm not 100% sure you're new/fast way to maskout data would allow for this, in any case we need a jupyter notebook that documents how to regrid in cdms/esmpy and what cdms does under the hood to prepare the data for esmpy.

[durack1]

@doutriaux1 @bonfils2 you get a second vote from me. I recall that there were some demos of the previous ESMF version findable at https://ice.txcorp.com/trac/modave/wiki/esmfLinear - these might make a good starting point

[jypeter]

I mentioned some seemingly nice ESMPy documentation in CDAT/cdat#1954

Now, it would also be interesting to know how the new stuff (ESMPy) compares to the older stuff (esmf), and the even older one I really appreciated for regular to regular grid (regrid2, especially regrid2.Horizontal) both in terms of accuracy and speed/memory usage (see #113)

The users may want to know how this compares to the scrip-based cdo

[doutriaux1]

@jypeter as far as regrid2 goes nothing changed.

[doutriaux1]

related to (thx @jypeter)
#113 #118

[jypeter]

I have tried to use ESMPy directly (conda install -n cdatm15 -c nesii/channel/dev-esmf -c conda-forge esmpy=7.1.0.dev32 in 2.10), but it's not easy to use, even if the available documentation is nice. @dnadeau4 is in the loop, with already a few mails exchanged with Ryan O'Kuinghttons 3 weeks ago. I hope I can find some time to experiment more

In the course of trying to locate alternative ESMPy resources, I have found the nice notebook below. Lots of information in there, but you have to read it over and over again, experiment, and read it again (and again)

https://github.com/nawendt/esmpy-tutorial/blob/master/esmpy_tutorial.ipynb

[doutriaux1]

@mattben we should link this from our website.

[doutriaux1]

@dnadeau4 IF you have time it would be awesome to do the exact same notebook but using cdms.