Discussion: Adding additional FATES parameters to PEcAn and extending supported model outputs

[serbinsh]

Hello All

Following up on my presentation to the modeling group a couple of weeks ago and our ongoing discussions about support for FATES in PEcAn I wanted to (finally) post this thread. First here is the ongoing PEcAn-FATES issue related to integration https://github.com/PecanProject/pecan/issues/1008 for reference

So the three main issues as I see it are: (1) how do we get more types of FATES output data supported by PEcAn, including outputs at different timesteps, which is supported by CL; (2) getting more FATES parameters updated by PEcAn, including priors and observational data constraints; (3) supporting the expanded PFT set and new parameter file (when released); (4) Starting PEcAn runs from inventory

For 1) We should settle on a set of outputs and desired timestep for each that folks would like to explore within PEcAn. In particular it would be good to add in height structured and other outputs that could be useful for understanding model uncertainties. I know we discussed some already but I lost my notes so we could just create a new draft list here to sort out which we will support in PEcAn

2) I think this should/could go with 3. But in general, what are people interested in seeing in PEcAn? All FATES params supported and the end user can create different versions of PFTs with/without specific params? We can start a list on that here as well; or should these be different threads?

num 4) In progress

[rgknox]

Thanks for starting this @serbinsh

As per my personal interests, I think my first use cases with PEcAn would certainly be the generation of parameter file priors.

My next interest would be parameter sensitivity analysis. I suppose I would try splitting that into a two part study, one performed in static stand structure mode looking at biophysical rates like GPP, respiration and energy fluxes. The other would perhaps be in either prescribed physiology mode or full mode, looking at classic inventory indicators like size structured growth increment and mortality, or recruitment.

Question, can one perform the sensitivity analysis you have been presenting to us, using non-scalar metrics (like size structured output)?

[rosiealice]

@serbinsh

Following up on the call, here's a link to the CLM/CTSM documentation on changing the output frequencies: https://escomp.github.io/ctsm-docs/doc/build/html/users_guide/setting-up-and-running-a-case/customizing-the-clm-namelist.html

The critical variables are: hist_nhtfrq: This is the frequency of the output timestamps. hist_mfilt: This is the number of those output timestamps for each file.

For hist_nhtfrq "A positive value means the output frequency is the number of model steps between output. Negative value means the output frequency is the absolute value in hours given (i.e -1 would mean an hour and -24 would mean a full day). Daily (-24) is the default value for all auxiliary files. Zero means the output frequency is monthly. This is the default for the primary history files."

For one set of annual files, with one value for each month, and another set of annual files, with one value for each day, one would do:

hist_nhtfrq = 0, -24 hist_mfilt = 12, 365

Then you need to set what variables are in the files other than the first/primary file (which is every default output). To do this you might do something like: hit_fincl2 = 'GPP', 'NPP', 'TLAI', 'ET'

to get a subset of the outputs. You can make diurnal files, also, of course, and so on.

Of course, as per this issue, we should think further about what sort of output we'd like to have in Pecan. But I wanted to make sure this functionality was under control first :)

[serbinsh]

Thanks @rosiealice this is super helpful!