CICE6 restart: scale_factor

[Dmitry Dukhovskoy]

Hello, I've written a python code to converte CICE4 to CICE6 restart. There is one variable that I cannot figure out how it should be saved in CICE6 restart. This is "scale_factor". According to icepack_shortwave.F90, scale_factor is the ratio of new to old net shortwave radiation. These fields look very different in CICE4 and CICE6. restarts. What is more confusing is that scale_factor in CICE6 output files look different from the restart saved for the same date (see figures).
The screenshots show scale_factor fromCICE6 history file (on the left) and CICE6 restart (on the right) for the same date. scale_factor from CICE4 looks similar to CICE6 history file. The field from the CICE6 restart does not make sense to me. Does someone know what is going on with scale_factor in CICE6 restart?
Thank you

 

[dbailey]

I don't see your pictures, but a couple of notes here. The scale_factor is kind of an artificial field and is biggest with night to day transitions. In earlier versions of CICE, we could compute the albedo and absorbed shortwave separately. So, we could compute the albedos at the end of the step and then the absorbed shortwave on the beginning of the next step. This assured synchronization with the atmospheric radiation. However, with the delta-Eddington radiation, the computation of albedo and absorbed shortwave are iteratively computed together and hence cannot be separated. So we do the following:

1. Estimate albedos at step 0. Send the albedos to the atmosphere.

2. Receive the atmospheric shortwave at step 1.

3. At the end of step one we compute new albedos and absorbed shortwave. This is technically using the incoming shortwave from step 1, but albedos from step 2. So, approximately Fswabs(1.5) = (1-a(2))*Fsw(1). Save this quantity for the next step as "scale_factor" which also goes on the restart.

4. Send albedos for step 2 to atmosphere.

5. Receive the atmospheric shortwave for step 2.

6. Adjust the absorbed shortwave from the end of last step Fswabs(1.5) to match the new atmospheric incoming shortwave. This gives you Fswabs(2) (1-a(2)*Fsw(2). We are actually doing this over the four bands and not just the broadband shortwave.

7. For a restart we have saved the incoming shortwave and scale_factor. We can use these to restart a run correctly.

The main thing to check is the shortwave diagnostics in the log file. These should be very small or identically zero. There is a big of a time lag issue in that the atmospheric radiation is only computed once an hour, but we couple every half hour. In standalone CICE we have interpolated the shortwave to whatever timestep is used in the model.

 

[Dmitry Dukhovskoy]

Thank you for quick response, let me go over your comments. Somehow the figures didn't get attached, let me try again (left is scale_factor from CICE6 history file, right is restart on a same date. scale_factor CICE4 restart looks similar to the left figure, i.e. the field from the history file)

 

[Dmitry Dukhovskoy]

Thank you for quick response, let me go over your comments. Somehow the figures didn't get attached, let me try again (left is scale_factor from CICE6 history file, right is restart on a same date. scale_factor CICE4 restart looks similar to the left figure, i.e. the field from the history file)

View attachment 4164View attachment 4165

OK, I think I know why the scale_factor fields look different in CICE6 history (left) and restart (right) outputs. History files are daily averaged. I'm still uncertain why scale_factor from CICE4 restart would look similar to the left and not right image. According to @dbailey comment, the scale_factor from the CICE6 restart makes sense and CICE4 restart does not look right. The question is can the scale_factor be reconstructed from the CICE4 restart fields (and how)? My restart reconstructed from CICE4 with CICE4 scale_factor works ok now but is there a better option for the scale_factor (e.g., make it = 1 everywhere)?

 

[dbailey]

Good question. Have a look at the way scale_factor is computed in CICE_RunMod.F90 in CICE4 versus CICE6. It might be that one was the total absorbed shortwave and the other was the ratio. I can't recall. However, you could certainly try just resetting to 1 in the restart file.