Rainfed corn yield underestimation after 2000 and activation of potato crop in CLM.

What version of the code are you using?

• CTSM Version: alpha-ctsm5.2.mksrf.23_ctsm5.1.dev171
• Resolution: 0.05° over Lower Michigan
• Compset: IHISTCLM50_BGC_Crop
• Meterological Forcing: CONUS404

Dear Scientists,

I am running CLM5 to simulate rainfed corn, soybean, and potato. I have encountered a few issues that I would appreciate your guidance on:

1. Rainfed Corn Yield Underestimation
   The simulated rainfed corn yield tends to be underestimated after the year 2000. I have examined some key variables—such as BTRAN2, which suggests that water stress is not the cause (more than 0.7 in many random location in the growing period)—and also reviewed the phenology, which appears consistent with satellite-derived LAI (i.e., no phenology mismatch). I have attached examples for several representative locations. Could you please suggest additional diagnostics or possible reasons for this post-2000 underestimation?
2. Potato Crop Activation
   I noticed that potato is listed as inactive in CLM5. Could you please provide guidance or documentation on how to activate or parameterize this crop type for simulation?

Any advice or references would be greatly appreciated.

That's interesting about rainfed corn. Keep in mind that the model does not represent technological advances over time. Another consideration could be changing nitrogen availability.

In the CLM there's no potato parameterization to activate, so you would need to introduce such parameterization yourself.

I will ping @samrabin in case he's interested in your results.

Yes, technological advancement is always a possible issue in cases like this. That's discussed in Lombardozzi et al. (2020) as a reason for why our global trends aren't as steep as the observed. In this specific instance the adoption of Bt corn might be contributing.

Of course, it's also possible that something we do simulate isn't working well—perhaps something to do with total growing-season NPP and/or translocation of carbohydrates to the harvestable organ. Something is definitely odd with the stair-stepping behavior of CLM LAI in your timeseries graph.

Thanks a lot, dear dr. @slevis and dr. @samrabin, for your insightful and comprehensive responses.

I am beginning to calibrate the model now for improved crop county-level crop yield, and I would like to seek your kind guidance on a few points before proceeding.
1- I have identified a set of potentially sensitive parameters and already have their default values. However, I am uncertain where to find reliable ranges (minimum and maximum) for these parameters. Could you please advise on the best sources or approaches to determine them?

2- I plan to use the Multiobjective Adaptive Surrogate Modeling–based Optimization (MO-ASMO) approach for calibration. Would the parameters I have collected be sufficient for this purpose, or would you recommend including additional ones?

3- In my simulations, corn yield aligns well with USDA data before 2000 but is underestimated afterward (picture shared). Given limited computational resources, I intend to select a 5-year calibration window. Would you recommend focusing on a recent period (e.g., 2000–2005), and if so, might this choice negatively affect the earlier period’s performance?

4- I wanna consider RMSE as the objective function, what others would you suggest?

5- I would also like to calibrate the potato, which is currently inactive in the model. Do you have any suggestions on how to activate and parameterize it? I plan to use the same set of parameters for calibration initially.

I truly appreciate your time and expertise. Your advice would be invaluable as I move forward with this calibration work.


parameter Description Unit Default value

Hi Mohammad,

(1) I don't have much experience calibrating those parameters, so I'm not sure where to look—sorry.

(2) I also don't have experience with that method, so I can't say for sure, but my instinct is that those parameters are as good as any.

(3) Calibrating to just a subsection of the historical data may indeed negatively affect performance in other periods. However, depending on what you're calibrating, it's possible that you might get lucky and only affect the post-2000 period. I would lean towards including as much of the observation period as possible, unless you only care about that one sub-period.

(4) RMSE makes a lot of sense, but RMSE of what? Yield is the obvious target, but you might also consider things like evapotranspiration or GPP/NPP if you can find observational datasets for those. It depends on what you care most about, though.

(5) Potato would be great! Activating it is easy with a recent version of CTSM (including the most recent version you get if you check out a fresh copy), using the set_paramfile tool:


Then setting paramfile = '/path/to/new_clm_params_file.nc' in your user_nl_clm should give you a run with potato activated.

Dear Dr. @samrabin ,
Thanks for the invaluble response. I checked out a fresh copy of CLM using:
git clone GitHub - ESCOMP/CTSM: Community Terrestrial Systems Model (includes the Community Land Model of CESM)
cd CTSM
./bin/git-fleximod update
version: ctsm5.3.084

I have modified the paramfile using the param_utils tool and activated potatoes, plus I am also using your planting temperature window, howerver i get the following run error:
clm::TimeSetymd: error return from ESMF_TimeSet: setting get_calday

/glade/work/rahilmoh/CLM/Cases/MI85_check_newCLM_params : this is the case directory

I ran the case with and without your planting temperature window nc files, It couldnt be executed while running.
Its mentionable that i created the surface data and dyn lanuse timeseries.

The following is my user_nl_clm:
fsurdat='/glade/work/rahilmoh/CLM/Model/my_ctsm_for_surfdata_ctsm5.3.050/CTSM/tools/mksurfdata_esmf/surfdata_LM_Rahil_5km_hist_1980_78pfts_c250610.nc'
flanduse_timeseries ='/glade/work/rahilmoh/CLM/Model/my_ctsm_for_surfdata_ctsm5.3.050/CTSM/tools/mksurfdata_esmf/landuse.timeseries_LM_Rahil_5km_hist_1980-2022_78pfts_c250610.nc'
finidat = '/glade/derecho/scratch/rahilmoh/archive/MI46_141y135x_IHistClm50BgcCrop_0.05d_CTSM_DynLULC_Conus404_Spinup/rest/0036-01-01-00000/MI46_141y135x_IHistClm50BgcCrop_0.05d_CTSM_DynLULC_Conus404_Spinup.clm2.r.0036-01-01-00000.nc'
use_init_interp = .true.
check_finidat_year_consistency = .false.

paramfile = '/glade/work/rahilmoh/CLM/Params/potato_and_irrigated_potatoes_included.nc'


stream_fldFileName_swindow_start = '/glade/work/rahilmoh/Analysis/cropcal_manipulation/swindow_starts_ggcmi_crop_calendar_phase3_v1.01.2000-2000_c250611.nc'
stream_fldFileName_swindow_start = '/glade/work/rahilmoh/Analysis/cropcal_manipulation/swindow_ends_ggcmi_crop_calendar_phase3_v1.01.2000-2000_c250611.nc'
stream_meshfile_cropcal = '/glade/work/rahilmoh/Analysis/cropcal_manipulation/360x720_120830_ESMFmesh_c20210507_cdf5.nc'
cropcals_rx= .true.

limit_irrigation_if_rof_enabled = .true.
use_groundwater_irrigation = .false.

lower_boundary_condition = 3
use_bedrock = .false.
soilwater_movement_method = 1


hist_empty_htapes = .true.

! Monthly average output for both streams
hist_nhtfrq = 0, 0

! Stream 1: 1 file per year (grid-averaged output)
hist_mfilt = 12, 12

! Stream 1: grid-averaged output (.true.), Stream 2: subgrid/PFT-level output (.false.)
hist_dov2xy = .true., .false.

! Stream 1: all your standard variables (grid-averaged)
hist_fincl1 = 'TLAI'

! Stream 2: subgrid (PFT) level output for GRAINC_TO_FOOD,
hist_fincl2 = 'GRAINC_TO_FOOD'

Thank you so much for your support.
Sincerely,