4DEnVar calibration

[rosiealice]

A note to keep track of the code from Tristan quaife's talk to the FATES meeting.

https://github.com/tquaife/4DEnVar_engine

This looks like it might be a good fit for our needs (being similar in principle to those of the JULES community).

If anyone can think of good site scale test cases for this then do chime in. I will try and check out this code when time allows.

[glemieux]

This was a really cool presentation to listen to as it brought back memories of my studies and (limited) work in optimal estimation and control theory. It looks like these techniques have there basis in Kalman filters application?

[JessicaNeedham]

NCAR's data assimilation team have a tutorial on data assimilation in CLM.
https://docs.dart.ucar.edu/en/latest/models/clm/readme.html

Is the thinking that Tristan Quaife's method would be preferable because the DA is separate from the model?

[rosiealice]

That was my understanding, yes. Historically the DART effort has been more focused on state estimation rather than parameter calibration, which is less relevant for land surface models. I don't know if CLM-DART includes this method, but my impression from a recent conversation was that it didn't.

[JessicaNeedham]

Got it, thanks @rosiealice. Data assimilation is all new to me but it sounds like cool stuff.