Initialization of CLM with observed data

[paolino@cola_iges_org]

I'm trying to initialize CLM using observed snow cover and soil profile data, for the fully coupled CCSM. I've tried to reproduce a CLM initial condition file -- the "i" netCDF model output file -- by taking the "i" file output from a 10 day coupled run, and replacing soil and snow data with observed data.

I haven't been able to run past initialization using this IC, although all components do report successful initialization.

If I use the spun-up model CLM IC, I can run successfully. This seem to indicate that I've screwed up my constructed IC.

I've run the constructed IC through interpinic, although it isn't clear to me if that's necessary, since all data are on the same grid.

One clue: The print messages in the unsuccessful runs indicate that the RTM is being initialized, while the successful run using the CCSM output CLM IC don't report RTM initialization. The input RTMVOLR data should be the same for both runs.

Anybody have a suggestion?

 

[slevis]

First of all, to all who ask clm related questions, I would recommend submitting them under the clm links, because I don't often look at the non-clm pages...

Now starting from the bottom of paolino's email:

If RTMVOLR is supposed to be the same for both runs, and you haven't modified it, then I suspect that it is not related to your problem. Also it sounds as though running through interpinic didn't help, which doesn't surprise me, though I guess it was worth your try.

My suggestions:

Modify the original clm.i file more carefully. For example, look at all the variables in the clm.i file in case you have caused inconsistencies between variables. For example, in updating the snow data, have you updated the variable that indicates number of snow layers? If not, you have a bit more work, because you'll probably want to look in the clm3 technical note for the algorithm that determines the number of snow layers and make that variable consistent with your snow data.

If that's not the problem, then look at the existing data in every gridcell, landunit, or column where you intend to modify the data and see, for example, whether that column is supposed to have soil and snow data. Doing this manually could be time consuming, so try to automate the process by writing a script in fortran, ncl, idl, matlab, or another language that handles netcdf.

Good luck,
Sam Levis