Icepack sea ice configuration
This section describes the implementation of the Icepack sie ice column physics package in the FESOM2 model. The scope of the following paragraphs is to provide a practical guide to users interested in detailed simulations of the sea ice system with FESOM2, and not to describe the scientific features of Icepack. A detailed description of the sea ice parameterizations here implemented can be found on the website of the CICE Consortium, which maintains and continuously develops this package.
Attention
The Icepack implementation in FESOM2 is still in a testing phase and we cannot guarantee a bugfree code nor good scientific results.
Note
To get more information regardng the implementation of Icepack in FESOM2, to report bugs, or to get advice regarding the model setup do not hesitate to open an issue on the FESOM2 GitHub repository or to contact Lorenzo Zampieri at lorenzo(dot)zampieri(at)awi(dot)de.
You are invited to update and develop further this documentation by pushing your changes to the FESOM2 Documentation repository on GitHub.
General information
Icepack–the column physics package of the sea-ice model CICE–is a collection of physical parameterizations that account for thermodynamic and mechanic sub-grid processes not explicitly resolved by the hosting sea-ice model, in our case FESOM2. The modular implementation of Icepack allows the users to vary substantially the complexity of the sea-ice model, with the possibility of choosing between several schemes and a broad set of active and passive tracers that describe the sea-ice state. Icepack v1.2.1 has been implemented in FESOM2 and can be used as an alternative to the standard FESIM thermodynamic module. As the standard FESIM implementation, the Icepack column-physics subroutines run every ocean time step. All the Icepack variables are defined directly on the nodes of the FESOM2 mesh, ensuring an optimal consistency between the ocean and the sea-ice components of the model. The inclusion of Icepack in FESOM2 required a revision of the calling sequence within the sea-ice model, which now follows that of the CICE model as illustrated in Fig. 1.
Fig. 1 Schematic describing the calling sequences of the Standard FESOM2 and FESOM2-Icepack implementations.
Icepack is licensed for use through the CICE Consortium. Therefore, we encourage the FESOM2 userbase interested in the Icepack features to be aware of the License when working with this model configuration. We report here a disclaimer from the Icepack website.
Important
Icepack releases are “functional releases” in the sense that the code runs, does not crash, passes various tests, and requires further work to establish its scientific validity. In general, users are not encouraged to use any of the CICE Consortium’s model configurations to obtain “scientific” results. The test configurations are useful for model development, but sea ice models must be evaluated from a physical standpoint in a coupled system because simplified configurations do not necessarily represent what is actually happening in the fully coupled system that includes interactive ocean and atmosphere components.
How to cite
The current Icepack version implemented in FESOM2 is Icepack 1.2.1. To acknowledge the development work behind the implementation of Icepack in FESOM2 please cite Zampieri et al. (2021), part of which used to compile this documentation, and Hunke et al. (2020), in addition to the usual FESOM2 papers.
Implementation
The implementation of Icepack in FESOM2 is fully modular, meaning that the users are free to vary the configuration via namelist parameters. When Icepack is used, namelist.icepack controls all settings related to the sea ice subgrid parameterizations, thus overriding the content of namelist.ice. The dynamics (EVP) and advection schemes are still controlled by the standard namelist.ice. Below we describe some of the most important namelist parameters, while we recommend consulting the official Icepack documentation for a more comprehensive description.
Namelist section &env_nml
nicecat Defines the number of sea ice thickness categories.
nfsdcat Defines the number of categories of the floe size distribution. This parameter should be set to 1 as the floe size distribution has never been tested in FESOM2.
nicelyr and nsnwlyr Defines the number of vertical layers in sea ice and snow.
Attention
Increasing substantially the number of thickness classes and vertical layers can lead to numerical instabilities (very thin vertical layers), memory issues, very large output files, and finally to a substantial slow down of the model because of the high number of tracers that need to be advected.
Namelist section &grid_nml
kcatbound Specifies which criteria is followed to discretize the Ice Thickness Distribution (ITD). Setting kcatbound equal to 0, 1, or 3 gives lower thickness boundaries for any number of thickness categories. Setting kcatbound=2 corresponds to the World Meteorological Organization ITD classification, and it is compatible only with nicecat=5,6,7.
Namelist section &tracer_nml
Logical parameters to specify parameterizations and passive tracers. Only tr_pond_cesm has been tested extensively.
Namelist section &nml_list_icepack
It regulates the type, frequency, and precision of the output for Icepack variables. Most of the Icepack variables can be defined as average over the grid cell (e.g. aice: average sea ice area fraction – 2D variable), or separately for each thickness class (e.g. aicen: sea ice area fraction in each thickness class – 3D variable), with the ITD information saved as a vertical dimension in the netCDF file. At the moment, variables defined over multiple vertical layers are output in separated files. For example, in a model configuration with n sea ice vertical layers, activating the qice output stream will lead to n files where qice_i.fesom.yyyy.nc contains the sea ice enthalpy of the i-th vertical layer averaged over the ITD. Similarly, activating the qicen output stream will lead to n files where qicen_i.fesom.yyyy.nc contains the sea ice enthalpy of the i-th sea ice vertical layer for each thickness class.
Compatibility with FESOM2 configurations
In Zampieri et al. (2020) the model was run with linear free surfaces (which_ALE=’linfs’), and other ALE coordinates have not been tested. In principle, Icepack should be independent of the scheme used to solve the sea ice dynamics. However, at the moment only the standard EVP is supported, while the mEVP and aEVP still show some strange behaviors. We are working on solving this issue as well as on testing further setups, and we will update this document as soon as progress is made.
Compilation
Compiling FESOM2 with Icepack is very easy if you are already used to the FESOM2 workflow. After cloning fesom2 from the GitHub repository, download the Icepack single column package:
cd src/icepack_drivers/
bash -l download_icepack.sh
The next step is to activate the Icepack flag in CMakeLists.txt by setting USE_ICEPACK from OFF to ON. At this point, you can proceed with the usual compilation via
bash -l configure.sh
The compilation of this FESOM2 version with the ESM Tools is not yet supported.
Running the model
Running FESOM2 with Icepack is not different from the standard case. Make sure to add the namelist.icepack file to your work directory. Two diagnostic files are generated in addition to the standard fesom2.0.out. icepack.diagnostics contains information about the Icepack configuration such as the value of some parameters, the tracers employed, and the boundaries of the ITD. icepack.errors possibly contains diagnostic information about errors in Icepack that can occur during the model run. Information about the running time are given in fesom2.0.out with the usual division in dynamics, thermodynamics, and advection.
The model output is saved in the result folder together with the standard ocean output. Note that outputting sea ice information using the standard FESIM variables (a_ice, m_ice, m_snow, etc.) is still possible also when using Icepack. These variables are consistent with the Icepack sea ice description (a_ice = aice, m_ice = vice, m_snow = vsno). An additional restart file is generated for Icepack, fesom.yyyy.icepack.restart.nc, and it is written with the same frequency as fesom.yyyy.oce.restart.nc and fesom.yyyy.ice.restart.nc.
Attention
Restarting the model after changing the number of ice thickness classes, the vertical discretization of ice and/or snow, and the number of passive tracers is currently not possible. Also, changing the thermodynamic and melt pond schemes during the run is not recommended. In these cases consider a cold start and repeat your spinup run.
Code structure
Icepack is a single column model and therefore its subroutines act on one grid cell. The Icepack code is downloaded from a separate repository (see instructions on how to compile the model) and is located in src/icepack_drivers/Icepack/columnphysics/. To integrate this code in a host General Circulation Model (GCM), in our case FESOM2, additional instructions are needed to define an interface between the two systems and to drive the Icepack subroutines. This interface is contained in the src/icepack_drivers/icedrv_*.F90 files, which are part of the FESOM2 repository, and will be briefly described in the following section.
Icepack drivers
icedrv_main.F90This file contains the main module of the Icepack drivers. All the variables are declared here, together with the interface of the subroutines contained in various submodules. If new variables or subroutines need to be added to the code, this is a good place to start. Try to maintain all the variables private to increase the modularity of the code, and use the transfer interface to exchange variables with FESOM2.icedrv_set.F90This file contains few subroutines that initialize the model parameters by reading the Icepack namelists or alternatively by extracting default values from the Icepack package. Furthermore,icepack.diagnosticsis written here, and the sea ice state is initialized in case of a cold start of the model.icedrv_allocate.F90This file contains subroutines that allocate the Icepack variables declared inicedrv_main.F90.icedrv_init.F90This file contains subroutines that initialize the Icepack variables declared inicedrv_main.F90and allocated inicedrv_allocate.F90.icedrv_step.F90This file contains few subroutines that describe the calling sequence of the sea ice model when Icepack is used in FESOM2.icedrv_advection.F90This file contains few subroutines that advect the Icepack tracers. If new parameterization or options are explored, you should check if the relative tracers are advected properly.icedrv_transfer.F90This file contains subroutines that describe the procedure to pass information between FESOM2 and Icepack.icedrv_io.F90This file contains subroutines that describe the I/O streams for the Icepack variables, including restart procedures. If new parameterization or options are explored, you should check if the relative tracers are restarted properly.icedrv_kinds.F90This file declares some standard types for variable declarations.icedrv_system.F90This file contains subroutines that handle model errors inside Icepack, possibly stopping the model run, and that output warning messages when appropriate.icedrv_constants.F90This file defines some constants that are used in the Icepack drivers.
Communication between Icepack and FESOM2
The Icepack environment is separated from the rest of FESOM2 and consists of a single big module with multiple submodules. Almost all the variables are private and are not visible by the FESOM2 code. The variables exchange between Icepack and FESOM2 takes place through the passing subroutines fesom_to_icepack and icepack_to_fesom.
Frequently asked questions
Should I use Icepack for my simulations?
It depends on your scientific questions. Icepack might be a good option if you are interested in sea ice processes in polar regions. In principle, the employment of Icepack should not negatively affect the ocean state but could make FESOM2 slower.
Is FESOM2 slower when run with Icepack?
Yes, the model integration is slower for two reasons: 1. The sea ice subgrid parameterizations are more complex compared to the standard FESIM. 2. Much more sea-ice tracers need to be advected. Overall, the sea ice component of FESOM2 becomes approximately four times slower with Icepack. Including additional output related to a more complex sea ice description can also contribute to deteriorating the model performances.
Which EVP scheme should I use with Icepack?
In principle, Icepack should be independent of the scheme used to solve the sea ice dynamics. However, at the moment only the standard EVP is supported, while the mEVP and aEVP still exhibit some strange behaviors. We are working on solving this issue and we will update this document as soon as progress is made.
Can Icepack be configured as the standard FESIM?
Yes, in principle it is possible to run Icepack with a single thickness class and with the 0-layer thermodynamics. However, the results obtained during the testing phase with this configuration were not very convincing and they seemed not compatible with the standard FESOM2 results. More investigations are needed to understand the cause of this behavior, which is likely related to a different implementation of the thermodynamic processes in the model.
Can Icepack be used in coupled configurations?
No, at the moment FESOM2 with Icepack has not been coupled with atmospheric models. A coupling with OpenIFS is planned and might be released in the upcoming months.
Can Icepack be used with data assimilation?
No, at the moment FESOM2 with Icepack has not been equipped with data assimilation capabilities.