How to adjust restart files for my simulations

[Yi Yao]

Dear,

I have some questions regarding the CMIP6 restart files. Here is the background:
1. I expanded the CFT column from 64 to 128, as I changed each irrigated column to three columns (drip, sprinkler and flood irrigated), then expanded the surface dataset and landuse timeseries file as well.
2. I want to do one historical simulations from 1980-2014, and later use the restart files to do future simulations.
3. I guess I also need to adjust the restart files, since there are only 64 CFTs there. For 2D variables (lat x lon), it is easy to do, while for 1D variables (gridcell), I am a bit confused.
So, my questions are:
1. Do we have to use the same surface dataset which is used for the cases which provide the restart files?
2. If there are mismatch between the restart files and the surface dataset, for example, even I did not do the development, there may still be mismatches (e.g. a new PFT), will it be a problem?
3. If I have to adjust the files, I wonder if there is a tool with which I can check which columns are in which gridcells?
4. If it is too hard to do, I guess I need to do a cold start. My study focus on temperature, humidity and precipitation. How many years do you think is appropriate?

Thanks in advance,
Cheers,
Yi

 

[nusbaume]

Hi Yi,

I am moving this post to the CTSM forum, as they have the most experience with modifying CFTs and PFTs, and thus will likely have the best chance to find answers to your questions.

Hope that helps, and have a great day!

Jesse

 

[oleson]

I'm not quite sure I understand your proposal but if you run a historical simulation 1980-2014 with surface and landuse datasets that have CFT=128, then I would think the restart file at the end of 2014 would have CFT=128 on it, if you've implemented it properly.
Have you checked that your historical simulation restarts properly? I would think that would be an indication that the restart files have CFT=128 on them.
Regarding the surface dataset, presumably if you are running 1980-2014 you would have a 1980 surface dataset and a landuse dataset for 1980-2014? The 1980 data on the surface dataset would need to match the 1980 data on the landuse dataset.
In general, a restart file can be interpolated to a different surface dataset (different number of pfts, columns, landunits, etc) using use_init_interp. But I don't know for sure if it would work for your configuration.