Running spin-up simulation to use crop model with SCAM

[jaedynw]

What I am trying to do:
I am looking at single column CAM+CLM (SCAM) and sensitivity to surface cover. I will be running the model for 5 months at a time. I want the crop model to be active with predictive phenology. This link shows the forcing data that I am using. The ARM data is already processed into forcing files for SCAM. The iop datasets I am using run from May 1st to September 30th from 2012-2018.

Next steps I want to do:
I want to spin-up the model so that I can use the crop model with SCAM.

What version of the code are you using?
cesm2_3_beta17

Describe every step you took leading up to the problem:

This is the path for one of the first spin-ups I have already tried to do: /glade/derecho/scratch/jaedynw/archive/I2000Clm50BgcCrop.T42_T42.spinup_testrun_v5/rest

set CASETITLE=spinup_testrun_v5
set CASESET=I2000Clm50BgcCrop
set CASERES=T42_T42
set COMPILER=intel
set CESMROOT=/glade/work/$USER/cesm2_3_beta17
set IOP = arm97
set IOPNAME = scam_$IOP
./xmlchange STOP_OPTION=ndays
./xmlchange STOP_N=152
./xmlchange RUN_STARTDATE="2012-05-01"
./xmlchange CLM_ACCELERATED_SPINUP="on"
./xmlchange MOSART_IGNORE_WARNINGS="true"
./xmlchange JOB_WALLCLOCK_TIME=02:00:00
./xmlchange MOSART_MODE="NULL"

After building the case I used ./xmlchangeRESUBMIT=10

This is the path for the second spin-up: /glade/derecho/scratch/jaedynw/archive/I2000Clm50BgcCrop.T42_T42.spinup_testrun_final_v2/rest

set CASETITLE=spinup_testrun_final_v2
set CASESET=I2000Clm50BgcCrop
set CASERES=T42_T42
set COMPILER=intel
set CESMROOT=/glade/work/$USER/cesm2_3_beta17
./xmlchange STOP_OPTION=ndays
./xmlchange STOP_N=152
./xmlchange RUN_STARTDATE="2012-05-01"
./xmlchange RUN_TYPE=startup
./xmlchange MOSART_IGNORE_WARNINGS="true"
./xmlchange JOB_WALLCLOCK_TIME=02:00:00
./xmlchange MOSART_MODE="NULL"
./xmlchange CLM_BLDNML_OPTS="-ignore_warnings"
./xmlchange CLM_NAMELIST_OPTS="use_init_interp=.true. init_interp_method='general'"

After building the case I used ./xmlchangeRESUBMIT=10

Main issue:
After taking the last restart file and using it as the finidat file, I saw no differences in TG, QSOIL, etc. even after changing the types of crops in the land surface file: /glade/campaign/cesm/cesmdata/inputdata/lnd/clm2/surfdata_esmf/ctsm5.2.0/surfdata_T42_hist_2000_78pfts_c240425.nc

Questions:
1. Do I need to first run 200+ years in “-bgc_spinup on” mode even though I only want to look at growing seasons over 6 years?
2. Do I need to spinup again for at least 400+ years in the normal mode?
3. Which compset makes the most sense to use? IHISTClm50BgcCrop or I2000Clm50BgcCrop
4. Is the following resolution okay to use with either of those compsets? T42_T42

My main uncertainties are making sure the compset, resolution, and spin-up times make the most sense for what I want to do. Let me know if I need to clarify anything else. Thank you!

 

[oleson]

The first spinup you've setup is going to be a continous run, beginning 2012-05-01 and running a total of 10 X 152 = 1520 days. So it will run continously through 2016-11-29. That typically wouldn't really be adequate for spinning up the model and you'd end up with an initial file that reflects 11-29 conditions. Typically, as you mention, we spinup for 200 or so years in accelerated decomposition mode and then a few hundred years in normal mode.
However, those longer spinups are really mostly necessary for higher latitude grid cells so that soil carbon comes into equilibrium. Do you need to do a global spinup? I'm not that familiar with SCAM but it just simulates a single point/grid cell, right? So it seems like you could do a single point spinup for your site which would be much less expensive?
The I2000Clm50BgcCrop compset seems reasonable for the spinup you want to do, you'd just want to run a single point resolution I think. And it seems like you'd want to spinup over full years beginning 01-01 and save monthly or daily restart files at near the end of the spinup so that you would have one available for May 1st.

 

[oleson]

Sorry, bad math. 11 X 152 = 1672 days. I guess you'd need one more 152 day run to get you to beginning of May 1 of 2017? So that would be a 5 year spinup. Also, it doesn't look like you are using a COLDSTART which means you would be using a default Jan 1 initial file for a run that starts in May, not sure if that could result in some problems.
Regardless, I don't think that would be long enough in AD mode to get the model spunup.
Since you are running on Derecho, I'm happy to take this discussion offline if needed.

 

[jaedynw]

@oleson Thank you for your reply! I only want to run a single point/grid cell. When I use the I2000Clm50BgcCrop compset, the datm gives a fixed forcing but I want to use the forcing from this iop. Since I am not seeing changes in variables such as QVEGE, TG, etc., I am not sure if I am actually running SCAM. My goal is to see how the land surface changes (changing crop types) alter fluxes and then how those fluxes change the atmosphere (convection, rainfall, feedbacks in the atmosphere, etc.). That would be great. If you private messsage me, I can provide my email information.

 

[jaedynw]

Is there a specific compset that has worked with the crop model and SCAM?

 

[oleson]

Right, that compset (I2000Clm50BgcCrop) would not be invoking SCAM as it uses fixed atmospheric forcing (GSWP3V1), as you've noted.
Looking at the link you provided, I see a "FSCAM" compset which appears to invoke SCAM. The long name for this compset is (at least in release-cesm2.1.5):

2000_CAM60%SCAM_CLM50%SP_CICE%PRES_DOCN%DOM_SROF_SGLC_SWAV

This invokes the SP (satellite phenology) version of CLM. If you want to use the prognostic crop model you could change the compset to:

2000_CAM60%SCAM_CLM50%BGC-CROP_CICE%PRES_DOCN%DOM_SROF_SGLC_SWAV

I'm not sure if this has been tried before.