ask for help about spinup

[Ceil]

Hi,

I use F2010climo to run 65 years. I modify one equation in SourceMods/src.clm/SatellitePhenologyMod.F90 and LAI values in surfdata_0.9x1.25_hist_16pfts_Irrig_CMIP6_simyr2000_c190214.nc. Can I use the final 30 years for analyses (i.e., the first 35 years regarded as spinup)? It seems that SpinupStability_SP.ncl is only used for I-compsets, so I have no idea how to examine whether the spinup years for F-compsets is enough.

Also, for FHIST, I modify one equation in SourceMods/src.clm/SatellitePhenologyMod.F90 and PCT_CROP (and PCT_NATVEG, PCT_NAT_CFT) in landuse.timeseries_0.9x1.25_hist_16pfts_Irrig_CMIP6_simyr1850-2015_c170824.nc. Using the default FHIST, I only can get the simulations for the year 1979 to 2014. Can I use the final 20 or 30 years for analyses?

Thank you in advance!

 

[slevis]

@Ceil, decisions about spin-up and which years to use for analysis are of scientific significance and, therefore, up to each scientist to decide. If you lack experience in these matters, you may find insight in what others have done in the published literature.

Regardless, I think that SpinupStability_SP.ncl may provide similar insight for F-compsets as for I-compsets.

 

[Ceil]

Thanks for your reply! Do you mean that I can use the same thresholds defined in SpinupStability_SP.ncl that used for I-compsets, when spinuping F-compsets? Or, I should firstly run I-compsets and use SpinupStability_SP.ncl to judge whether the I-compsets is spunup. Then using the spunup results from I-compsets as initial forcing to run F-compsets? But I am not sure whether I can use all simulations outputed by F-compsets, or do I need to throw up the first several years?

 

[oleson]

I think for a satellite phenology run, you could just throw up the first several years. SP mode spins up fairly quickly. As noted in the User's Guide, SP will end up with fairly stable fluxes in about 10 years when initialized with an initial file (not cold start).

1.5.4. Spinning up the Satellite Phenology Model — ctsm release-clm5.0 documentation

The spinup script isn't really designed for F-compsets because the atmospheric forcing doesn't repeat.

 

[Ceil]

Thanks for your reply! Are there quantitative metrics or thresholds to measure whether the F2010climo or FHIST is already spun up? Because the about 10 years are kind of like a uncertain estimate, how can I know the accurate years needed? Can I use the same thresholds defined in SpinupStability_SP.ncl that used for I-compsets? I primarly focus on the outputed near-surface air temperture, albeo, energy fluxes (e.g., latent/sensible heat flux, ground heat flux), radiation variables (e.g., shortwave and longwave), and soil moisture.

 

[Ceil]

My another problem is, for FHIST, I only can get the simulations for the year 1979 to 2014 (36 years in total). If I want to do a analysis of 30 years (such as the final 30 years), I only have 6 years to throw up as spinup. The length of the year seems to be not enough. Perhaps using some quantitative metrics or thresholds can show that 6 years are enough to complete spinup. I am not sure, may be it is impossible. But I have no idea how to deal with this situation.

 

[oleson]

I think we've discussed these spinup issues before with you, i.e., this series of posts:

the out-of-box model shows no equilibrium after running SpinupStability.ncl

Hi, I used the out-of-box model as below to get 26-year monthly simulations (1979-2004), then I ran SpinupStability.ncl to see whether it's already spun up. But my figure shows no equilibrium. I only did these modifications in SpinupStability.ncl: subper = 1, filepath, and filename...

bb.cgd.ucar.edu

In it, we said:

"There are going to be trends in the various variables because you are running a FHIST case and the forcing (e.g., CO2) varies by year, so it's going to be difficult to determine if the model is spunup with your LAI changes. You might consider running an F2000 or F2000climo case for a number of years and then plotting the various variables and evaluate the trends. And then use the initial files from that to initialize your FHIST run."

"The stability script won't really work on that configuration because each atmospheric year is different. "

Perhaps you can do some smoothing on the time series of the relevant variables to minimize the year to year variablity and see what the trends look like. Sorry, but that is about all the advice we can offer at this point.

 

[Ceil]

Thanks for your reply! I wonder how to validate the model using observations, if using Fclimo2010? If using FHIST, I can use the observations (sush as ERA5, satellite or site observations) for the year 1979 to 2014? But if I got 30 years after running Fclimo2010, should I validate the model by comparing the obsevations in year 2010 with the 30-year average of the simulations?

 

[slevis]

@Ceil these are methodological questions that you as a scientist and your collaborators need to decide for yourselves.