Settings Used for CCSM4 Control Run?

[wporter@mit_edu]

Hello,

I have CCSM4 up and running on our cluster, and am working on validating the port. Stability and perturbation tests have all passed fine, but long-term runs are showing some significant differences with the CAM control data at ESG.

I started my 30-year run using the B_1850 compset at f19_g16 resolution, and compared the results to the data at the link above using AMWG, but my averaged annual temperatures differed from the control set by up to 3.2 degrees:

Should I be using different starting parameters for this comparison, or should I be looking at other problems with the port?

Thanks!

Will Porter
Portland State University

 

[sirajkhan78@gmail_com]

Hi,

Sorry I cannot answer your question but I need help from you. Actually I am doing the same thing as you did but as I am new to this CCSM4 I am getting problem to start model for exact comparison with the data available at EGS. All my test for port validation are passed and model is running fine but how to start model (start and end date) for the comparison I am confused. I also need to run model for 30 years but what should be the control run setting in env_run.xml file. Can you guide me to run my model as you did for your graphs shown in your post. Or if possible can you send me your env_run.xml file?

Cheers
Siraj

 

[wporter@mit_edu]

Hi Siraj,

Sorry for the late reply -- we had some hardware issues that set us back temporarily, so my attention was focused elsewhere!

The good news is that I seem to have figured out the necessary settings myself, and our current averages are right on the money when compared to the control set. Here's how I set up our run:

- Create a new case using the B_1850_CN compset and the f19_g16 grid setting.

- Modify env_run.xml so that STOP_OPTION is "nyears" and STOP_N is "1".

- Change env_mach_pes as desired to make the best use of your cluster's resources. (This took some trial and error on our system.)

- Start the run with "./configure -case", followed by the build and queued run scripts. This should result in a single modeled year run.

- Once it is done, check for successful output. If everything looks as it should, edit env_run.xml again. Make sure CONTINUE_RUN is set to "TRUE" and change RESUBMIT to "29". This should make CCSM pick up right where it left off.

- After completion, you should have everything you need to use the diagnostics utility provided on the CAM pages. Just edit the script to point everything to the correct directories, and let it rip.

This is the procedure that worked for us -- I hope you find it useful!

Regards,
Will Porter

 

[eaton]

Hi Will,

Was the difference in results that you were concerned about in your initial post due solely to using the B_1850 compset rather than B_1850_CN ?

Thanks,
Brian

 

[wporter@mit_edu]

Hi Will,

Was the difference in results that you were concerned about in your initial post due solely to using the B_1850 compset rather than B_1850_CN ?

Thanks,
Brian

Hi Brian,

I believe so. Our brand-new cluster actually suffered a massive RAID failure right after my first post above and I lost everything related to that earlier B_1850 run, but I'm pretty sure that everything else about it was identical to this latest B_1850_CN run. After months of trial and error and technical difficulties, it's pretty satisfying to have a clean, stable port!

Will Porter

 

[wporter@mit_edu]

This does still leave me confused about the underlying question, though: is there any way to find out the settings used for the various diagnostic data sets available?

For example, the b40.1850_ramp.track1.2deg.001 datasets are labeled 2xCO2 and 4xCO2 for the 1909-1928 and 1978-1997 periods, respectively. If I want to duplicate these results, should I double/quadruple the CO2 levels manually in env_conf.xml, or will the run take care of this automatically? Where (or from whom) can I get this kind of information?

Thanks!

Will Porter