Discrepancy in QIRRIG (surface, confined, and unconfined) between cesm2.2.2 and ctsm5.2 releases

[amans]

I have observed a stark contrast in irrigation coverage (mostly in the northern hemisphere) between cesm2.2.2 and ctsm5.2 CLM runs with identical case setup. Both runs used I2000Clm50BgcCrop compset and were simulated for 12 months starting 2000-01-01. Similar issue in all three irrigation variables (QIRRIG_FROM_GW_CONFINED, QIRRIG_FROM_GW_UNCONFINED, and QIRRIG_FROM_SURFACE). What is causing this difference? Has this behaviour been documented previously?

Is this issue related to my case setup? The case paths and output paths are as follows:

cesm2.2.2 case location: /glade/work/amans/cases/global_cesm2_2_2_I2000Clm50BgcCrop/
ctsm5.2 case location: /glade/work/amans/cases/global_ctsm5_2_I2000Clm50BgcCrop/

cesm2.2.2 output history files: /glade/derecho/scratch/amans/archive/global_cesm2_2_2_I2000Clm50BgcCrop/lnd/hist/
ctsm5.2 output history files: /glade/derecho/scratch/amans/archive/global_ctsm5_2_I2000Clm50BgcCrop/lnd/hist/

This issue is relevant to me because I am running regional Clm51BgcCrop simulations in the midwest US using surfaces subsetted for ctsm5.2. All QIRRIG values are zero for the midwest US simulation from 1980-2020.

Thank you!

 

[oleson]

@samrabin , do you have any insight on this problem? Does the user need to run another year to reset the crop calendars, or spinup from cold start, or ...?

 

[samrabin]

Hi @amans,

It looks like these two runs use different fsurdat files, which could cause the difference.

• cesm2.2.2 run: /glade/campaign/cesm/cesmdata/inputdata/lnd/clm2/surfdata_map/release-clm5.0.18/surfdata_0.9x1.25_hist_78pfts_CMIP6_simyr2000_c190214.nc
• ctsm5.2 run: /glade/campaign/cesm/cesmdata/inputdata/lnd/clm2/surfdata_esmf/ctsm5.2.0/surfdata_0.9x1.25_hist_2000_78pfts_c240216.nc

The cesm2.2.2 fsurdat file is much older, with ctsm5.2 having introduced many improvements in the surface dataset. At first I was going to say that if you're just doing offline land model runs, use ctsm5.2 and ignore the cesm2.2.2 results. But ctsm5.2 seems to be the one with the problem!

It looks like all the irrigation-related namelist settings are identical, so that's not the issue.

A few questions:

1. You say that "All QIRRIG values are zero for the midwest US simulation from 1980-2020." Could you give me the path to the case where you tested this?
2. Are yield values also zero for these irrigated crops?

 

[amans]

Hi @amans,

It looks like these two runs use different fsurdat files, which could cause the difference.

• cesm2.2.2 run: /glade/campaign/cesm/cesmdata/inputdata/lnd/clm2/surfdata_map/release-clm5.0.18/surfdata_0.9x1.25_hist_78pfts_CMIP6_simyr2000_c190214.nc
• ctsm5.2 run: /glade/campaign/cesm/cesmdata/inputdata/lnd/clm2/surfdata_esmf/ctsm5.2.0/surfdata_0.9x1.25_hist_2000_78pfts_c240216.nc

The cesm2.2.2 fsurdat file is much older, with ctsm5.2 having introduced many improvements in the surface dataset. At first I was going to say that if you're just doing offline land model runs, use ctsm5.2 and ignore the cesm2.2.2 results. But ctsm5.2 seems to be the one with the problem!

It looks like all the irrigation-related namelist settings are identical, so that's not the issue.

A few questions:

1. You say that "All QIRRIG values are zero for the midwest US simulation from 1980-2020." Could you give me the path to the case where you tested this?
2. Are yield values also zero for these irrigated crops?

Hi @samrabin,

Some background on my regional runs: Region of interest is the Mississippi River Basin (MRB). Surfaces were created and subsetted using ctsm5.2.mksurfdat_esmf tools. Forcing is CONUS404.

1. It seems I made a wrong statement. The case where I tested 1980-2020 is over the Mississippi River basin at 0.5deg. The path is /glade/work/amans/cases/I2000Clm51BgcCrop_MRB_0.5_mksurf with outputs in /glade/derecho/scratch/amans/archive/I2000Clm51BgcCrop_MRB_0.5_mksurf. Here, there are surface irrigation values (surprisingly, I missed this before) but no groundwater irrigation. Looking at my user_nl_clm file, I did not use "use_groundwater_irrigation=.true.", so this should be expected. However, I have another case at 0.05deg over the same domain for a shorter timeframe (2000-2009). Case: /glade/work/amans/cases/I2000Clm51BgcCrop_MRB_0.05_regMOSART_0.125. Outputs: /glade/derecho/scratch/amans/archive/I2000Clm51BgcCrop_MRB_0.05_regMOSART_0.125. Here, all QIRRIG values are zero. The 0.05deg case uses a mask mesh. 0.5deg does not. Also, 0.5deg surface has 5 nlevurb values compared to 10 nlevurb for 0.05deg. These are the only major differences that I can think of between the two cases.
2. I had not checked yield values previously. If GRAINC_TO_FOOD is the field for crop yield, then the 0.5deg case does have yield. The 0.05 case does not.

 

[amans]

After some digging, I found that the irrigation values are not zero, but very small. This may be why ncview prompt that the min and max are zero.
However, if I let it check all data, it does show irrigation values. Same thing happens for GRAINC_TO_FOOD as well.

So, I guess this issue was really just a quirk due to high resolution and ncview. The discrepancy between cesm2.2 and ctsm5.2 for the global case may or may not be an actual issue.

 

[samrabin]

Okay, great! I'll look into the global run discrepancy separately. Thanks!

 

[amans]

I ran the ctsm5.2 cases (global and regional) for 2 years. It seems that for the first year of my simulation there are no irrigation or crop values in the northern latitudes. From the second year, there are both irrigation and crop values.

Output path for global case: /glade/derecho/scratch/amans/archive/global_ctsm5_2_I2000Clm50BgcCrop
Simulation period: 2010-2011

Why would crop and irrigation outputs be zero in the first year of the run?

 

[samrabin]

Could you send me the path to your case?

 

[amans]

Could you send me the path to your case?

/glade/work/amans/cases/global_ctsm5_2_I2000Clm50BgcCrop

 

[samrabin]

Thanks. I'm not seeing anything in your setup that SHOULD be causing this, but could you re-try without these lines in user_nl_clm?

limit_irrigation_if_rof_enabled=.true.
use_groundwater_irrigation=.true.

Re-reading your message, I notice you said "northern latitudes." Could you post some plots showing what you mean? Also, what do you mean by "crop values"?

 

[amans]

could you re-try without these lines in user_nl_clm?
limit_irrigation_if_rof_enabled=.true.
use_groundwater_irrigation=.true.[/CODE]

I don't see any differences in the outputs after running without those lines.

I notice you said "northern latitudes." Could you post some plots showing what you mean?

I meant the differences between cesm run and ctsm run in the northern latitudes in the original post, which contains the plots. The northern latitudes are noticeably empty in the ctsm run.

Also, what do you mean by "crop values"?

That would be crop yield (GRAINC_TO_FOOD).

For reference, here is the comparison for year 1 and year 2 for the month of August for GRAINC_TO_FOOD for the run without limit_irrigation_if_rof_enabled and use_groundwater_irrigation in user_nl_clm.


Maybe the issue is related to initial condition or spinup? I have not performed spin up yet.

 

[samrabin]

Huh, that's really puzzling. I don't think you should need to do a spinup with a 2000 compset. @oleson and @slevis do y'all have any ideas? Is it worth looking at a HIST compset? Or maybe specifying a branch or hybrid run instead of startup?

 

[oleson]

My vague recollection is that the fact that crops in the northern hemisphere don't start growing until the second year is normal behavior, but I don't remember why. Something to do with accumulators maybe? I don't know why the southern hemisphere would be different though.

 

[samrabin]

The crop GDD accumulators (GDD0, GDD8, GDD10) are accumulated over different parts of the year in the northern and southern hemispheres. The southern hemisphere's includes January 1, so maybe that's why it takes an extra year there. However, those variables as well as their running 20-year means (GDD020 etc.) should be on the restart files, right? Unless I'm misunderstanding what happens in a startup run with a 2000 compset.

 

[oleson]

Right, maybe what I was remembering was associated with a cold start. We could ping Danica at some point, she might have an explanation.

 

[slevis]

@oleson and @samrabin what you wrote makes sense to me:
- In a cold start, crops may take a full year before the first planting
- In a run with a spun up finidat file, crops should start in the first year

...but maybe the second statement is incorrect for some reason. Did @amans try running long enough to confirm?

 

[amans]

Hi @slevis. I tested several multi-year regional runs without spinup for I2000 BgcCrop compset in ctsm5.2. There were no irrigation and crop yield value the entire first year.

I don't have a spun up file at the moment. But, using a restart file with CONTINUE_RUN=TRUE, crops start from the first year of the continue run.