CICE Prescribed Mode Ice Thickness

[mg3662@columbia_edu]

I am seting up experiments looking at future changes in SST and their impact on the atmospheric circulation.  The component set I generated is: RCP8_CAM5_CLM40%CN_DICE%PRES_DOCN%DOM_RTM_SGLC_SWAV. and I am conducting hybrid runs from output from the CESM-LE at 2046.  My boundary conditions are created using the sea ice and sea surface temperatures from the CESM-LE (ensemble member 1).  So my results should begin from a very similar state. When comparing my hybrid run the day after the run starts to the original CESM-LE we see anomalies in surface temperature immediately over the sea ice with colder anomalies where sea ice should be thin and warm anomalies where it should be thick.  This seems to be caused by the prescribed mode assuming that the sea ice thickness is a constant at 2m.  I found the location in the source code (ice_prescribed_mod.F90 line 459-463) where 2m is prescribed and I have done test changing the value and the results support this idea.  These biases due to constant sea ice thickness, I believe, might be important sources for errors for my experiments so I am looking to use sea ice thickness instead.My question is whether this has ever been done before?  Is it feasible to replace the contant thickness value with the sea ice thickness at each location by loading in a new variable or might there be other potential issues?Thanks so much!Melissa For reference sake I have uploaded images of the model run with 2m (Cnt_to_CESM-LETS_20460102.png), 1m (Cnt_hi1_to_CESM-LETS_20460102.png), and 4m (Cnt_hi4_to_CESM-LETS_20460102.png) ice thickness.  The top figure is the CESM-LE the middle is my hybrid run and the bottom is the difference.  

 

[dbailey]

Let me clarify something first. Are you actually using DICE%PRES or CICE%PRES? I'm not sure what happens if you are using the former. If you are actually using the latter, the way to change the thickness is by modifying the module ice_prescribed_mod.F90. You would copy this into SourceMods/src.cice and edit it there. If you want to actually read in a spatially variable thickness dataset, this can be done but you also need to modify the ice_prescribed_mod.F90 module to do this. See what is done for "ice_cov" and you can add some analogous code for "ice_thk" for example. This is something that Clara Deser and Bob Thomas have done before.Dave

 

[mg3662@columbia_edu]

Hi Dave,Yes, it is CICE%PRES I made a typo earlier.  Ok, good it sounds like I am on the right track and that it is possible to make the change.  Thanks so much!!Melissa

 

[mg3662@columbia_edu]

Hi Dave,I have another question about the perscribed ice thickness.  I was able to alter the source code to include the thickness and it seems to have worked.  It got rid of a negative bias in surface air temperature around the sea ice edge (relative to the fully coupled model for the first day).  There is a bias that remains, a positive one over the central Arctic and I am trying to determine what would cause this.  In my experiments I am inputing daily sea ice and SST but when I ask for daily output from the cice model I only get daily resolution for the ice area and thickness but not any of the variables the model computes.  Is it possible that the prescribed mode is expecting monthly data only and so is only calculating the fluxes off of a single days worth of ice area and thickness?  The outputted daily ice area and thickness fields look correct.  Otherwise it is possible that this could be a signal of sea ice/atmosphere coupling but I don't have a clear idea what of what processes would be missing and would cause a positive bias on the order of +1K in the surface air temp in the prescribed mode.   Thanks!Melissa

 

[dbailey]

Technically the CICE prescribed mode is able to read whatever frequency. Check the ice.log files to see that the model is reading the data daily. Keep in mind that surface temperature and snow depth on the ice are still prognostic and will respond to the forcing from CAM. I don't think the process is missing per se, just that CAM likely has incoming shortwave and precipitation biases over the ice. You can try adjusting the "albedo" in the sea ice model to compensate for the shortwave somewhat. We've found big differences between coupled and specified ice/sst runs because of the snow cover in the Arctic. Have a look at the FAQ on albedo tuning in the sea ice:https://bb.cgd.ucar.edu/faq-albedoradiation-tuning-cesm-cice#newDave

 

[mg3662@columbia_edu]

Thank you so much for the input.  I looked into the log files and it does seem to ingest each day.  I think you are correct that snow is the issue here.  The differences between the fully coupled versus prescribed mode for the snow depth averaged over the first month is about 50% (lower in prescribed).  The negative anomalies in snow depth match up roughly to the locations of higher surface temperatures.  I'm surprised it is such a large difference between the coupled and prescribed model and that it occurs so quickly.  Unless this is a spin-up issue, where the hybrid run isn't taking the snow depth from the restart files (it does look like this).  Once I have run the model out for longer I can check to see what type of systematic bias exists.  Thanks again for all of your help!