pft N balance error

[chen1718@umn_edu]

Hi,

I am new to the clm3.5 CN mode. I have been spin up the model for about 110 years, then I want to start a branch run with a new variable added to the histfile. But I met an N balance check error from CNBalanceCheckMod.F90:

nstep = 1877617 TS = 0.267429413266702E+03
pft nbalance error = -3.6226951892309600E-006 2807
begnb = 0.1091350123808753
endnb = 0.1091422703715759
delta store = 7.2579907005737576E-006
input mass = 5.9625604863640884E-006
output mass = 2.3272649750212913E-006
net flux = 3.6352955113427976E-006
veg type = 1

Without the cn mode turned on, the variable was added to the output file successfully (The variable I want to see is not calculated in the original model so I added a pointer in clmtype.F90, clmtypeInitMod.F90 and histFldsMod.F90 and made some change in CanopyFluxesMod.F90). Now I am not sure if the N balance problem is caused by the variable I added or the model itself? Can I just comment it as ckcastillo did to fpc_total? (Since I will not look into Nitrogen, just the carbon cycle and the C13 isotope part.)

Thanks for your attention and time!

Best,

Ming

Department of Soil, Water and Climate
UMN

 

[slevis]

I can only suggest that you return the code to its original state (ie, without any changes at all) and try again. Then you will know whether your changes caused the error.

Personally I hope that your changes caused the error because I would not know what to advise if the unchanged model gives the same error. I have never seen that error in my CN runs.

In this case I would definitely not advise to comment out the error because the nitrogen cycle plays an important role in CN's carbon cycle.

Sam Levis

 

[chen1718@umn_edu]

Dear Sam:

Thank you for your advice. I made the model to the state before any change (I even re-download the code to another directory on the cluster to make sure it is unchanged). But the error still comes out in the early stage of simulation. Sometimes it is pft C/N balance error, sometimes column C/N balance error.

It is very strange because if I continue the original run (cn mode) with the executable I complied before, the model runs well without any error message. I'm totally confused. If you have any suggestion, please let me know. Thanks very much!

Ming

 

[slevis]

You are right: this is very strange behavior and beyond my expertise. Considering that we have not encountered or heard of such a problem before, I suspect that you are having a hardware or compiler problem rather than a software problem.

Sam

 

[chen1718@umn_edu]

Thank you, Sam. I will check into the compiler and hardware and will update here if I find anything.

Ming

You are right: this is very strange behavior and beyond my expertise. Considering that we have not encountered or heard of such a problem before, I suspect that you are having a hardware or compiler problem rather than a software problem.

Sam

 

[chen1718@umn_edu]

The problem was just solved....to be honest, accidentally. I really don't know which part I did solved the problem. But if someone else meet the same problem I would like to discuss.
Now the clm-cn model runs well with the newly added output variables.

 

[slevis]

It's always good to be methodical about changing and testing code. It's always good to be able to return to the code that you started from in case things start going badly or you get inexplicable behavior. It helps to test changes one or a few at a time. These are only a few rules of thumb that help you keep your sanity while making changes to the model.

Sam

 

[chen1718@umn_edu]

Yes definitely, I'll learn from it :)

I have another question on single point run: Is the single point only represent for a single gridcell? If so, does it mean that I have to be careful when comparing the result to the tower observation? For example, the footprint of the tower has a radius of 2 km, but the grid cell is 1deg * 1 deg? - Can the single point be smaller than this?

 

[slevis]

You may use the mksrfdat tool to make a surface dataset for a single point, and you can choose the resolution of that point to be smaller than 1x1 degree.

Sam

 

[rrachmayani@marum_de]

Dear Sam Slevis,

I am trying to calculate GPP of the vegetation. I have doubt on which variable I should use to calculate the GPP, is it photosyntesis (FPSN) in (umol/m^2s)? and then I should convert the units into Giga ton or Kg C m^-2 y^-1 . And a summation of LCIND, RCIND, SCIND, HCIND, and NIND will get a land carbon storage amount? Thank you.

Best regards,
rima

 

[slevis]

In clm or clm-dgvm runs, look at FPSN. In clm-CN runs, look at GPP, because FPSN is the non N-limited GPP.

If you mean (LCIND+RCIND+SCIND+HCIND)*NIND, then I agree. Except that this does not include litter and soil carbon. I'm afraid I don't remember whether we keep track of soil carbon in the clm-dgvm...

Sam

 

[rrachmayani@marum_de]

Dear Sam,

I run with clm-dgvm, so I will take a look FPSN. Yes, I mean *IND multiply with NIND. As far as I know, these are not included with litter and soil carbon. Litter carbon is LITTERAG + LITTERBG summation? and soil carbon is CPOOL_FAST + CPOOL_SLOW?. So, should I also included litter and soil carbon for the land carbon storage amount? Thank you.

best,
rima

 

[slevis]

It depends how you define land carbon storage, and I will not interfere with your decision. Note that we never evaluated the clm-dgvm simulated litter and soil carbon...

Sam

 

[rrachmayani@marum_de]

Dear Sam,
I have question regarding NPP, I plot Net primary production (NPP) from clm-dgvm, and I get the maximum value is higher compare to other model (6 times higher). As I read in dgvm user's guide, NPP is net primary production (g C m-2 of plant functional type area; note different area unit). Is there something to do with the area unit? should I multiply NPP with a area unit or something? I am looking forward your suggestions from you. Thank you.

Best,
Rima

 

[slevis]

If you are comparing NPP data with different units, then it makes sense that you should convert the units to be consistent.

Sam