Will two hybrid run give different results from the same restart file?

[qzg18@psu_edu]

Hi!

I started a hybrid run from the Large Ensemble 1, using BC5CN COMPOSET to create a 2000 control run.
My run accidentally stopped in 20160201, so I restarted it from the closest date which is 20151101. The restart was success so I had two sets of output data for 201512-201601.

However, when I compare the monthly vertically-integrated droplet concentration in 201601 for the two model output, the results are not same as shown in the following two figures.

Is that because of my model setting errors or because the hybrid run is not bit-by-bit restart? Thanks very much if anyone could give me any suggestions!

Cheers,
Qinxue

 

[stevenDH]

Hi

I've been looking into this as well a couple of weeks ago and this is what I understand from it (the model developers can correct me if nescessairy)
From the cime documentation page (5. Running a Case — CIME master documentation) I found the following regarding a hybrid run:

In a hybrid run, the model does not continue in a bit-for-bit fashion with respect to the reference case. The resulting climate, however, should be continuous provided that no model source code or namelists are changed in the hybrid run.

Hence I assume it makes sense that these results are not exactly the same, it is not bit for bit but rather the same climate, if you want it to be bit for bit you should use the branch setting instead of hybrid.

Cheers
Steven

 

[hannay]

It depends how you "restarted" your run:
- If you restarted as a hybrid, you should not get bit-to-bit results. The hybrid run is not bit-to-bit restart.
- If you did a restart or a branch, you should get bit-to-bit results.

An hybrid somewhat analogous to a branch run with relaxed restart constraints. I think that in a hybrid run, the ocean model only start at the second ocean coupling.

 

[qzg18@psu_edu]

Hi

I've been looking into this as well a couple of weeks ago and this is what I understand from it (the model developers can correct me if nescessairy)
From the cime documentation page (5. Running a Case — CIME master documentation) I found the following regarding a hybrid run:

In a hybrid run, the model does not continue in a bit-for-bit fashion with respect to the reference case. The resulting climate, however, should be continuous provided that no model source code or namelists are changed in the hybrid run.

Hence I assume it makes sense that these results are not exactly the same, it is not bit for bit but rather the same climate, if you want it to be bit for bit you should use the branch setting instead of hybrid.

Cheers
Steven

Hi Steven,

Cool! Thank you for the detailed explanation!
It helps a lot!

Cheers,
Qinxue

 

[qzg18@psu_edu]

It depends how you "restarted" your run:
- If you restarted as a hybrid, you should not get bit-to-bit results. The hybrid run is not bit-to-bit restart.
- If you did a restart or a branch, you should get bit-to-bit results.

An hybrid somewhat analogous to a branch run with relaxed restart constraints. I think that in a hybrid run, the ocean model only start at the second ocean coupling.

Hi Cecile,

Thank you very much for your warm help!
It solved my confusion.

Cheers,
Qinxue