Same Standalone CAM3,4,5 initialization, different results.

[juniorphy@gmail_com]

Hi there,

Let me make a silly question about CAM initialization.

I just ran two simulations with exactly the same initial condition as the atmosphere, start in the same day using the same forcing. The results are different. They are dramatically different in the tropical region after a few days, especially around regions with convection.

Does make sense regarding the convection scheme of the model? Are there any random statistical assumption applied to the convection scheme to justify these changes for each simulation?

Many thanks for any answer.
JR

 

[nusbaume]

Hi JR,

There should be no random or stochastic processes in a regular CAM configuration, which means that if two model configurations are exactly the same then they should produce the same answers (at least within numerical error). We in fact regularly test that this is the case, so my guess is that there is probably some configuration or setting that is different between your two runs.

To start with, Are using the exact same version of CAM for each simulation? Are you also using the same compset and PE-layout? If so, then I would next make sure that the namelist files are the same between the two simulations. This can be done by going to the run directory and comparing each simulation's xxx_in file to the other, where xxx is the name of each model component (atm, lnd, ocn, etc.).

Finally, if your runs are on Cheyenne then feel free to post the paths here and I can take a look at them as well.

Thanks, and have a great day!

Jesse