Creating a finidat file for a case

[aberg]

For a recent case I ran a baseline simulation using the default fsurdat and finidat files for a year. I also ran a case where I modified ONLY PFTs 0, 1, and 2 and the corresponding LAIs in a small part of western North America, in the fsurdat file. No other variables were changed. For this case, I had to first set finidat = ' ', run my case for a day to get a restart file, then run interpinic to get my finidat file (see thread "Running interpinic"). Then I ran this modified case for a year using my modified fsurdat file (containing the modified PFTs and LAIs) and the new finidat file I got from interpinic.

However, when I compared the model output between the two cases, I am getting significant differences over areas besides western North America (this is the only place I expect differences). For example, I am getting increases/decreases in VOCs over the Amazon, parts of Asia, etc. where the PFTs and LAIs were not changed at all. I am also getting plus or minus up to 12 K in ground temperature in the Siberia, Amazon, and other places. Those are huge differences.

I wondered if the problem could be with how I had to create the finidat file for the modified case. I only ran the case for a day to get the restart file for making the new finidat file. Would I need to run my modified case with finidat = ' ' to resolve these differences? I just can't think of another reason why my modified run has such huge differences in VOCs and temperature over places where the PFTs and LAIs were not modified (no other variables in my fsurdat files are different from each other). Any ideas?

 

[slevis]

Such large differences make me suspect that you are running the CLM coupled to CAM, right? If so, then check whether the differences are due to the different finidat files by creating two "I" cases (i.e., offline CLM) in all other ways the same as your coupled cases (e.g., use the same fsurdat and finidat files as in the coupled runs). Now history output after one or two time steps should look almost the same. If so, then the differences in the coupled runs are due to the added degrees of freedom when running an interactive atmosphere. If not, then there may be something wrong with the new finidat or fsurdat files...

Sam

 

[aberg]

Thanks for the suggestion, I'm going to give the 'I' runs a try and will let you know the outcome. I also have a few questions about interpinic and the finidat files.

Why is it necessary to run interpinic in this case? I was talking to my advisor, Colette Heald, and she said in previous versions of the model, vegetation could be specified and the model wouldn't complain about the need to run interpinic to create a new finidat file. So we were just wondering, what is interpinic doing in this case that then allows the model to run? Why doesn't the original finidat file work when I specify vegetation in the fsurdat file? Wouldn't we want the exact same initial conditions even when modifying some of the PFTs?

Also, I did get a chance to compare some variables between the finidat files. The differences in pft weights make sense (those are the changes I made) but I did find differences in the air temperature variables (though there was no difference between ground temperatures). Why would interpinic need to change the air temperature variables to make the finidat file consistent with the change in PFTs?

And the last question: we were wondering, what would happen if we just ran both cases for a full year with finidat = ' ' ? Are there consequences of running a case with finidat = ' ' and not actually having a finidat file specified?

Thank you so much, hopefully you can help clarify some things!

---
EDIT: Two I_2000 cases were run and the history output for both cases looks the same with the only differences being over western North America, which is where differences are expected (I'm comparing history output after 5 days, Jan 6th) . So I take that to mean there are no problems with my finidat and fsurdat files.

So are these differences due to added degrees of freedom unavoidable?

 

[slevis]

finidat files contain data in 1-dimensional (ie vector) form rather than in 2-dimensions. When you change the pft map, you change the dimension (ie vector length) of all the pft-level variables in the finidat file. By running interpinic you put data from a finidat file with a different vector length into the finidat file with the correct vector length needed for your run.

Setting finidat=' ' means that the model will use the initial conditions hardwired in the code, which are far from spun up. If spun up soil moisture is not a concern for you, then you could try finidat=' '.

Glad to hear that your files are ok.

Sam

 

[aberg]

When I looked more closely at my I_2000 case, I noticed there actually are a few gridcells scattered about globally that have slight differences. This would indicate something might be wrong with my fsurdat or finidat files.

I differenced my baseline and modified fsurdat files, and they are okay: the only changes are in PFTs and LAIs over western North America, where changes are expected. Outside of western NA, the differences in PFT and LAI are zero.

So this leaves the finidat file, which was created using interpinic as above. Could anything other than a faulty finidat file cause these slight differences in the I case? Could there be a problem with the way the finidat file was created? I can't figure out why I would see differences in the I case outside of the region I made modifications, since my fsurdat file is okay (and the finidat file I assume would be changed based on changes to the fsurdat file).

 

[slevis]

To confirm that the finidat file causes these slight differences, compare the two runs that (I think) you did with finidat=' '. I expect these two runs to be identical everywhere except where you changed the fsurdat file. This is the only way that I can think of that should give bit-for-bit identical answers in the areas that you did not change.

The interpinic tool maps data from one finidat file to another using a nearest neighbor method and may lead to slight differences due to issues of numerical precision, including in areas where you were not expecting such differences. We do not have a tool that will give bit-for-bit identical answers when you use different finidat files.

Sam

 

[aberg]

Thank you for the help, it looks like it was the finidat file causing small differences in the 'I' case. So by setting finidat = ' ' in my coupled runs I can eliminate those issues, and know that the differences that remain are not due to fsurdat or finidat.