Changing number of ice catregories?

[david_hebert@nrlssc_navy_mil]

I put this question in science, but it is more of a technical quesiton... We have been running CICE with the 'default' 5 ice categories since we started using CICE. I would like to explore using the WMO 7 categories, as that is becoming of interest to one of our model output users. Has anyone done this or something similiar, and if so could you share some suggestions? Off the top of my head I'm thinking about 1) Did you take a current restart file and manually change categories? or...2) Did you make a separate subroutine in CICE to read a 5 category restart, then call 'rebin' to let CICE redistribute to the 7 WMO categories?  Thank you for your suggestions!
David

 

[dbailey]

Hi David,Really good questions. In the CESM, we have not changed categories for many years. The main issue that I have with new restarts is the internal energy of the snow and sea ice. This also needs to be changed if you are changing the number of categories and is non-trivial getting the energy right for initialization of the thermodynamics. I personally think that a better approach is to start with ice_ic = 'default' or 'none' and run the model for 10 years or so to get a more realistic ice initial state for the new number of categories.Dave

 

[dbailey]

Hi David,Really good questions. In the CESM, we have not changed categories for many years. The main issue that I have with new restarts is the internal energy of the snow and sea ice. This also needs to be changed if you are changing the number of categories and is non-trivial getting the energy right for initialization of the thermodynamics. I personally think that a better approach is to start with ice_ic = 'default' or 'none' and run the model for 10 years or so to get a more realistic ice initial state for the new number of categories.Dave

 

[afroberts@lanl_gov]

Hi David,Elizabeth and I just conferred on this. Most groups stick with the standard 5 categories, but there is no reason not to use the 7 WMO configuration.  Beware, however, that you will not receive the same answer from the model in terms of hemispheric ice volume with the 7 WMO categories as with the standard 5, meaning that you will see a change in model climate, and I expect that to be more pronounced in a fully coupled model than in an ice-ocean case.- Andrew Roberts 

 

[afroberts@lanl_gov]

Hi David,Elizabeth and I just conferred on this. Most groups stick with the standard 5 categories, but there is no reason not to use the 7 WMO configuration.  Beware, however, that you will not receive the same answer from the model in terms of hemispheric ice volume with the 7 WMO categories as with the standard 5, meaning that you will see a change in model climate, and I expect that to be more pronounced in a fully coupled model than in an ice-ocean case.- Andrew Roberts 

 

[david_hebert@nrlssc_navy_mil]

Hi Dave, thanks for the speedy reply! I agree spinning up from a default conditoin would be best. Unfortunately, with the Navy Global model I'm working with, I won't have time to do an adequate long-term spinup. I was looking through the code, and see that rebin calls icepack_compute_tracers, I am hoping that would simply take care of the enthalpies.  If rebin doesn't work perhaps I'll try just re-initalizing the tracers with icepack_init_trcr for this one time. I think that should be okay here, as for my application this should only be a one-time recategorizing.Thank you!David 

 

[david_hebert@nrlssc_navy_mil]

Hi Dave, thanks for the speedy reply! I agree spinning up from a default conditoin would be best. Unfortunately, with the Navy Global model I'm working with, I won't have time to do an adequate long-term spinup. I was looking through the code, and see that rebin calls icepack_compute_tracers, I am hoping that would simply take care of the enthalpies.  If rebin doesn't work perhaps I'll try just re-initalizing the tracers with icepack_init_trcr for this one time. I think that should be okay here, as for my application this should only be a one-time recategorizing.Thank you!David 

 

[dbailey]

I agree that the spinup is not optimal. Sometimes, what has worked for me is initializing the ice state concentration, thickness, and surface temperature but then initialize the internal energy, salinity as is done in set_state_var for the 'default' case. 

 

[dbailey]

I agree that the spinup is not optimal. Sometimes, what has worked for me is initializing the ice state concentration, thickness, and surface temperature but then initialize the internal energy, salinity as is done in set_state_var for the 'default' case. 

 

[david_hebert@nrlssc_navy_mil]

Thank you both Dave and Andrew for your replies. Andrew, thank you for pointing out there would be a difference in hemishperic volume. Is the reason why the volume would be different with WMO settings the lack of categories above 2 meters? Would it be beneficial to add an 8th category with bound 4.67 meters like the original distribution? Thanks again!David

 

[david_hebert@nrlssc_navy_mil]

Thank you both Dave and Andrew for your replies. Andrew, thank you for pointing out there would be a difference in hemishperic volume. Is the reason why the volume would be different with WMO settings the lack of categories above 2 meters? Would it be beneficial to add an 8th category with bound 4.67 meters like the original distribution? Thanks again!David

 

[afroberts@lanl_gov]

 Hi David,Unfortunately you're venturing into a fundamental limitation of the current ice thickness distribution (ITD) physics and its numerical implementation.  No matter the number of thickness distirbution categories added, the solution does not converge.  It keeps changing.  We believe this may be a fundamental limitation of the current ITD theory, and of the calculation of compressive strength, which is why Elizabeth and I are keen to implement variational ridging in the model to see if it addresses this limitation. That doesn't help you right now, other than to warn you about venturing down the path of seeking convergence with an increasing number of categories. - Andrew

 

[afroberts@lanl_gov]

 Hi David,Unfortunately you're venturing into a fundamental limitation of the current ice thickness distribution (ITD) physics and its numerical implementation.  No matter the number of thickness distirbution categories added, the solution does not converge.  It keeps changing.  We believe this may be a fundamental limitation of the current ITD theory, and of the calculation of compressive strength, which is why Elizabeth and I are keen to implement variational ridging in the model to see if it addresses this limitation. That doesn't help you right now, other than to warn you about venturing down the path of seeking convergence with an increasing number of categories. - Andrew