Abnormal vertical profiles of ice temperature and salinity in the outputs of CESM2 fully coupled simulations

[jhkim]

Hello,

To check the sea ice mass balance of the model, I conducted simulations using the CICE5 stand-alone model and the CESM2 coupled model, and plotted vertical profiles of the temperature and salinity in the sea ice layer. However, in the CESM2 experiment, unphysical results were obtained for the temperature and salinity profiles of the sea ice. I've double-checked all and concluded this is definitely not a simple mistake. As no one seems to have reported this issue, I am posting about it. Both models used the “mushy” option for thermodynamics and the simulations were as follows:

- CICE5.1.2 with atmospheric and oceanic interannual forcings from JRA55-do (1958-2020)
- CESM2: B2000 atmosphere-land-ice-ocean coupled 100-yr simulations (ver. 2.1.3 and 2.2)

I examined the climatological mean and area-average north of 80N in the later part of the model integration. Figure shows the profiles of the first ice thickness category. There are definitely unphysical features in CESM2 outputs. These features also appear in other thickness categories at different layers (figure not shown).

Testing by changing the number of ice layers (7 vs. 8) and the CESM2 version (2.1.3 vs. 2.2) still led to the same issue. The model didn't blow up, but it seems that there is a bug when saving the outputs of sea ice temperature and salinity. Does anyone have experience in checking the sea ice internal temperature or salinity variables in your coupled model simulations?



Thanks,
Joohong

 

[dbailey]

Thank you for posting this. We have not looked very much at the vertical profiles of salinity and temperature. What is the time frequency of these plots? Monthly-mean, daily mean? What if you just pick a point instead of an area average? You have to be careful with spatial and temporal averaging of these when the ice in adjacent gridcells has a very different thickness.

 

[jhkim]

Thanks for quick response. The plot used Tinz and Sinz variables on monthly mean and areal mean (lat>80N, aice>0.01). Here are the comparison of profiles at several points, showing the abnormal internal profiles are at everywhere. Looking forward further comments. Thanks!