Questions about CLM5 BGC spinup and carbon cycle

[Gamling Dang]

Hello. Everyone.
I am carrying out the simulations of the carbon cycle and energy balance at single-point scale in subtropical forest. According to the tutorial (https://www2.cgd.ucar.edu/events/2019/ctsm/files/practical4-wieder.pdf), I spinup the case with AD mode and postAD. The atmospheric forcing data was from 2003-2009, 30-min. AD mode cycled the data for 400 years, and postAD cycled the same data for 200 years. Both output variables reached an equilibrium state by analysis from SpinupStability.ncl after postAD spinup.

Setting for AD spinup:
compset=I2000Clm50BgcCropGs
./xmlchange CLM_ACCELERATED_SPINUP="on"
./xmlchange CLM_FORCE_COLDSTART="on"
./xmlchange DATM_CLMNCEP_YR_START=2003
./xmlchange DATM_CLMNCEP_YR_END=2009
./xmlchange RUN_STARTDATE=0001-01-01
./xmlchange REST_OPTION=nyears
./xmlchange REST_N=400

Setting for exit spinup:
compset=I2000Clm50BgcCropGs
./xmlchange CLM_ACCELERATED_SPINUP="off"
./xmlchange CLM_FORCE_COLDSTART="off"
./xmlchange RUN_TYPE=startup
./xmlchange CONTINUE_RUN=FALSE
./xmlchange DATM_CLMNCEP_YR_START=2003
./xmlchange DATM_CLMNCEP_YR_END=2009
./xmlchange RUN_STARTDATE=0401-01-01
./xmlchange RUN_REFDATE=0401-01-01
./xmlchange REST_OPTION=nyears
./xmlchange REST_N=200
use_init_interp = .true.
finidat = '/path/to/2003-2009_400y.clm2.r.0401-01-01-00000.nc'

Normal case:
compset=I2000Clm50BgcCropGs
./xmlchange CLM_ACCELERATED_SPINUP="off"
./xmlchange CLM_FORCE_COLDSTART="off"
./xmlchange RUN_TYPE=startup
./xmlchange CONTINUE_RUN=FALSE
./xmlchange DATM_CLMNCEP_YR_START=2010
./xmlchange DATM_CLMNCEP_YR_ALIGN=2010
./xmlchange DATM_CLMNCEP_YR_END=2010
./xmlchange STOP_OPTION=nyears
./xmlchange STOP_N=1
./xmlchange RUN_STARTDATE=2010-01-01
./xmlchange CLM_CO2_TYPE='constant'
./xmlchange CCSM_CO2_PPMV=388.8
use_init_interp = .true.
finidat = '/path/to/2003-2009_200y.clm2.r.0601-01-01-00000.nc'


However, when I set the initial file from postAD spinup, and then ran a case from 2010-01-01, the simulations of GPP, latent heat, and sensible heat made sense. The NEE and soil respiration were higher than the observation data. I also compared all these data between spinup(400y-200y) and no spinup conditions. Even no spinup, NEE and soil respiration looked well with the monitoring data. After spinup, NEE became positive while it was negative in both no spinup and real case.

I guess the abnormal thing was caused by soil water because deep soil water simulated with spinup was much lower than that without no spinup. One thing that need to be noted, TWS reached equilibrium in post AD spinup, but could not reach stable even when I extended AD spinup time to 1000 years.

Can anyone help me or give me any suggestions on carbon cycle modeling using CLM5.0? Thanks a lot.

The results for AD spinup and postAD spinup using SpinupStability.ncl were attached. Also the variables related to carbon.

 

[oleson]

I can only make some general comments here. I'm surprised that the spinup script is showing equilibrium for TOTECOSYSC, TOTSOMC, and TOTVEGC in year 1 of AD. Usually this takes at least a couple of hundred years for those to come into equilibrium. Maybe you can attach the script you used and the plot it produced for both pAD and AD. That might also provide some clue as to why TWS is not in equilibrium. The output of the script notes that either TWS is not in equilibrium or that it is missing from the history files. Does it exist on the AD history files?
Based on your settings above, it seems like you are using the default GSWP forcing. I would not necessarily expect the model to perform well compared to site observations when forced by that data. It may perform better when forced by meteorological data measured at the site. But likely some site-specific parameter adjustments would need to be made even in that case.
The long-term mean of NEE may be negative, but could be positive for individual months/years, e.g., you are using years 2003-2009 for spinup and then subjecting the model to 2010 forcing. So there may be something about 2010 forcing that is causing the model to produce positive NEE.

 

[Gamling Dang]

Dear Oleson, many thanks for your reply.
Sorry for my unclear description. Yes, the default forcing for the I2000Clm50BgcCropGs is GSWP3. I had changed them with data from in-situ observations (30-min). For the NEE, some are positive or negative for the specific meteorological conditions. The confused thing occurred when I summed them up, the annual NEE was positive. As for the TWS, it was included in the history files.
I should add more information for my case. The domain region is about 10 km2, 0.03° x 0.03°. So I think some variables can reach equilibrium in a short time. The scripts for creating the spinup cases are attached. And the datm_in and lnd_in were also attached.
I also made another spinup using the "I1850" compset, both AD spinup and post spinup used the atmospheric forcing after 2000 (2003-2009). The simulation NEE using the initial file from "I1850" matched well with the monitoring data. So I think "I2000" made higher carbon accumulated in the ecosystem and caused positive NEE. But I1850 may not be suitable for the spinup after 2000. So do you have better ideas?
Looking forward to your suggestions.

 

[oleson]

I see you have daily average output for ADspinup and postSpinup. The SpinupStability.ncl script only works correctly on monthly history files or annual mean history files.
I wouldn't use an 1850 compset to spinup the model up and then use that to initialize an I2000. For one thing, the CO2 forcing will be quite different.

 

[Gamling Dang]

Yes, Dr. Oleson, I noticed. I processed the daily outputs into annual mean values using python. Their changes met the criteria of less than 3%. Most reached stable in more than a decade.
I am still confused about the initial state after spinup. The default initial file was clmi.I2000Clm50BgcCrop.2011-01-01.1.9x2.5_gx1v7_gl4_simyr2000_c190312.nc. I compared a case run with clmi.I2000Clm50BgcCrop.2011-01-01.1.9x2.5_gx1v7_gl4_simyr2000_c190312.nc and a case run with the initial file from (400 AD + 200 post) spinup, the result of the former one made sense, but the latter one had higher NEE and the annual total NEE indicated the subtropical forest was a carbon source.
I want to know if there are any methods to identify whether the initial state of the spinup file is right or not, except for the stable state. For example, I found the annual mean GPP in the stable spinup period (I2000) was higher than the GPP from remote sensing products. Thank you very much.

 

[slevis]

Unfortunately we do not have such a tool.