Missing Values in CLM5-BGC Simulated Carbon Fluxes (NPP, NEP, NEE)

[657835429]

Dear all,

I have noticed a phenomenon when simulating carbon fluxes using CLM5-BGC, I am not getting any errors reported in my run.

When I look at the results using ncview, I find that other variables such as GPP, LAI, etc. are simulated without problems and there are no missing cases. But the simulations of carbon fluxes such as NPP, NEP, NBP, NEE, ER, etc. show exactly the same missing spatially. Even after replacing the default surface file and using a new surface file, the areas where these variables are missing are identical to the simulations under the default surface file.

To troubleshoot the problem, I used multiple sets of driver data (China meteorological forcing dataset and CRUNCEPv7), initialization files (default files: clmi.I2000Clm50BgcCrop.2011-01-01.1.9x2.5_gx1v7_gl4_ simyr2000_c180715.nc and a new initialization file based on the default initialization file Spinup again), but the results show that the regions where the missing occurs are all identical.

Is this normal for CLM simulations? Or is there a problem with my setup? I hope you can shed some light on this. In the file are my user_nl_clm ,user_nl_datm,and instructions for running the model, as well as a picture showing a normal GPP and a NEE with missing space.

Thanks,

Jiawei Bing

 

[oleson]

I don't recall having seen that behavior before, and it is not normal of course.
You could try a cold start to see if maybe the initial data is not being mapped correctly onto your grid for some reason.
I also wonder if your machine setup is such that the model is continuing to run even though you have floating point errors which creates NaNs in the output. You could try running with DEBUG=True.
I assume you've run an out of the box case with no problems...

 

[657835429]

I don't recall having seen that behavior before, and it is not normal of course.
You could try a cold start to see if maybe the initial data is not being mapped correctly onto your grid for some reason.
I also wonder if your machine setup is such that the model is continuing to run even though you have floating point errors which creates NaNs in the output. You could try running with DEBUG=True.
I assume you've run an out of the box case with no problems...

Thank you for your guidance. The model ran to completion without errors, but spatial missing values persist identically in carbon fluxes (NEE, NBP, NEP), while LAI and GPP appear normal—this discrepancy is puzzling. A subsequent two-year simulation reproduced the issue (log attached). I will execute a DEBUG=True run as suggested and report findings. Your continued input is appreciated.

 

[657835429]

I don't recall having seen that behavior before, and it is not normal of course.
You could try a cold start to see if maybe the initial data is not being mapped correctly onto your grid for some reason.
I also wonder if your machine setup is such that the model is continuing to run even though you have floating point errors which creates NaNs in the output. You could try running with DEBUG=True.
I assume you've run an out of the box case with no problems...

Hi oleson, I think I found the problem. The carbon flux simulation values are missing in the exact same location as PFT16 (Unmanaged Irrigated Crop). I then exported the carbon flux for PFT16 and the results also showed the missing values. To fix this, I set irrigate = .true. in user_nl_clm, but the output carbon flux is still missing. Can you tell me if my settings are wrong? I am using CLM5-BGC (not CLM5-BGC-Crop).

 

[oleson]

I'm not sure what you mean by "exporting the carbon flux for PFT16" or why/how you fixed this by setting irrigate = .true., sorry.
I don't see anything wrong with your settings. I ran a short global simulation with that compset and settings and the output looks normal. Maybe there is a problem with your surface dataset, although there are checks in the model to prevent running with problems like that. Maybe try a short global simulation to see if that works.