ISMIP7 Protocol

ISMIP7 Protocol – In progress - Please visit page regularly for latest updates

The ISMIP7 protocol is currently under testing and builds upon the foundation of ISMIP6. Feedback and questions from modelers are welcome via our Github Discussions. You can also contact us via our contact -at- ismip.org  email.

ISMIP7 Protocol Overview

The ISMIP7 protocol overview spreadsheet below outlines the experimental framework, focus group structure, and modeling guidelines for the seventh phase of ISMIP. This overview provides a quick reference to the protocol's key components, with links to each focus group for more detailed information. 

For the full document, please read our current protocol "cheat sheet" and "Globus manual". 

What is an ISMIP7 submission for the 30th June?

• The README file linked to ISM submission can be downloaded here.

• Files for CESM2-WACCM SSP585 (C007) and your historical (C001) that have passed the ISM_SimulationChecker - these files can be revised later this summer. The experiment ID and file naming conventions are described here. The variable request for the filechecker is here and also explained in the cheat sheet and in the spreadsheet below. 

• When you are ready for your submission, please send an email to: ismip6 - at - gmail.com and include: 

1. Your globus email or id, 

2. Ice Sheet Modeled/domain_id (ex: AIS or GrIS): 

3. Modeling group name/source_id (ex: BCG): 

4. Ice Sheet Model Name/ism_id (ex: ISSM):

This will allow us to create an upload folder for you.

ISMIP7 variable request, filenaming and folder conventions

The spreadsheet  below lists the data request to be saved on the standard ISMIP7 grid that is closest to a model native grid.  This data request is the same as the one listed in a cvs file as part of the ISM_SImulationChecker.

The ISMIP7 data request reflect key variables that contain information on the simulated ice sheet state, ice sheet specific boundary conditions and a record of the climate forcing applied as a boundary condition to the ice sheet model, as these may be different from the input forcing file. The major difference compared to the ISMIP6 data request is the splitting of combined fluxes into their respective components. For example, we now ask for calving flux and ice front melt flux to be two separate files. In addition, if this is applicable to your model, we would like modelers to save depth average temperature (annually) and 3D temperature (at selected times). If you compute SMB from your own PDD or EMB,  you will be asked to save the data that you produce.  

Similar to ISMIP6, variables are requested as yearly snapshots corresponding to the end of year for state variables (ST, such as ice sheet altitude), and as yearly average for flux variables (FL, such as surface mass balance).  Flux variables are defined as positive when the process adds mass to the ice sheet and negative otherwise.

When 3D fields (x,y,z) are requested, such as ice temperature, please only report the selected times: initial state, beginning of projection, 2100, 2200, 2300 and for models with a back in time period, 1900.  Only for models that have 3D calculation. Output should be on the model native vertical coordinate, please follow the cfconvention.

A number of scalar output, or integrated measures, are also requested. To help prepare the integrated scalars based on 2d fields of a submission and also produces sea-level estimates in a format that is helpful for emulators, please use the scalar processing tool. This tool is still in progress and we welcome feedback.

Not all variables are mandatory, as indicated in the "yes/no" column. However, if your model can compute these variables, please include them in your submission. Variables in red or purple indicate a slight modification from the ISMIP6 data request  or a new variables for ISMIP7.

Guidance for file preparation 

• The ISMIP7 filename and convention document was used in generating the ISM_SimulationChecker. Please check this document as it will have more guidance. Example files that are compliant are available on Globus: AIS and GrIS. 

• In CMIP a full year is typically from first of January to the first of January of the following year. Time should be defined in “days since 1850-01-01“, using the "standard" Gregorian-Julian calendar. Guidance on how to generate the days and bounds can be found here: https://github.com/ismip/ismip7-time-encoding

• The end of the historical run should not be included at the beginning of the projections to avoid repetitions of the same values. This allows to merge together the different periods without having repeated entries. Models that do not have a historical period and start directly in January 2015 should still submit a historical run with one time step containing the model’s initial conditions.

• Single precision (float f4) should be used for all output

• All missing data should use the default netCDF4 fill value in attribute _FillValue. If missing_value is present in addition it has to have the same value (this is a general CF recommendation). Do not set missing values in ice masks, set to 0 instead.

ISMIP7 filenaming and folder conventions

The ISMIP7 filename and convention document will have more guidance and examples. The file naming convention and folder structure is inspired by CMIP.  Files should be submitted with the following directory structure and filenames.

Directory structure =

<domain_id>/

<source_id>/

<ism_id>/

<set_id>/

<set_counter>/

domain_id = ‘AIS’ or ‘GrIS’

source_id = Modelling group name

ism_id = Ice sheet model name and version

set_id = ‘CORE’ or ‘ESM’ or ‘PPE’

set_counter = ‘C001’, ‘C002’, ..., ‘E001’, ‘E002’, ...

Example directory

GrIS/NORCE/CISM3/CORE/C001

File name structure =

<variable_id>_<domain_id>_<source_id>_<ism_id>_<ISM_member_id>_<ESM_id>_<forcing_member_id>_<experiment_id >_<set_counter>_<time_range>.nc

variable_id = variable short name

domain_id = ‘AIS’ or ‘GrIS’

source_id = Modelling group

ism_id = Ice sheet model name and version

ISM_member_id = identifying ISM choices, e.g. m001

ESM_id = CMIP model name, e.g. CESM2-WACCM

forcing_member_id = identifying unique forcing choices, like different downscaling, PDD parameters, ..., e.g. f001

experiment_id = ‘historical’, ‘ctrl’, ‘ssp126’, ...

set_counter = unique index within a set C001, C002, E001, E002, P001, P002

time_range = start and end year of experiment, e.g. ‘2015-2300’

Instructions

variable_id: This is the variable short name as defined in the data request tables.

source_id: Name of the modelling group. No underscores “_”, dots “.” or special characters.

ism_id: ISM name and version. No underscores “_”, dots “.” or special characters.

ISM_member_id and forcing_member_id are in place to distinguish ensemble members in the PPE and ESM sets. In the CORE experiments they are typically set to m001 and f001.

ISM_member_id: mNNN identifies the ice sheet modelling choice variant. A different ISM_member_id is given for each new initial state and for different ice sheet model choices or parameters like input datasets, sliding law choice or parameters, GIA, calving, resolution ... .

forcing_member_id: fPPP identifies the forcing choice variant. A different forcing_member_id is given for a new choice of translating CMIP forcing to the ice sheet scale. This could be a different downscaling process or parameters within it, melt implementation or melt parameters.

The counters are three digit zero-padded numbers that are incremented for new variants/ensemble members. The historical experiment and the following scenarios share the same mNNN and fPPP numbers.

set_counter: Identifies the set by first letter (CORE, ESM, PPE), followed by a unique number that increments with each model run in a set: Cnnn, Ennn, Pnnn. The set_counter links to entries in a spreadsheet specifying parameter and modelling choices for this particular experiment. This means that all files with the same Cnnn, Ennn, Pnnn in the name will differ only by variable name. The counters are three digit zero-padded numbers.

Example names

iareafl_GrIS_NORCE_CISM3_m001_CESM2-WACCM_f001_historical_C001_xxxx-2014.nc

The complete set CORE files would look like this:

iareafl_GrIS_NORCE_CISM3_m001_CESM2-WACCM_f001_historical_C001_xxxx-2014.nc

iareafl_GrIS_NORCE_CISM3_m001_MRI-ESM2-0_f001_historical_C002_xxxx-2014.nc

iareafl_GrIS_NORCE_CISM3_m001_CESM2-WACCM_f001_ssp370_C003_2015-2300.nc

iareafl_GrIS_NORCE_CISM3_m001_MRI-ESM2-0_f001_ssp370_C004_2015-2300.nc

iareafl_GrIS_NORCE_CISM3_m001_CESM2-WACCM_f001_ssp126_C005_2015-2300.nc

iareafl_GrIS_NORCE_CISM3_m001_MRI-ESM2-0_f001_ssp126_C006_2015-2300.nc

iareafl_GrIS_NORCE_CISM3_m001_CESM2-WACCM_f001_ssp585_C007_2015-2300.nc

iareafl_GrIS_NORCE_CISM3_m001_MRI-ESM2-0_f001_ssp585_C008_2015-2300.nc

iareafl_GrIS_NORCE_CISM3_m001_CESM2-WACCM_f001_ctrl_C009_2015-2300.nc

iareafl_GrIS_NORCE_CISM3_m001_MRI-ESM2-0_f001_ctrl_C010_2015-2300.nc

iareafl_GrIS_NORCE_CISM3_m001_ERA5_f001_ocr_C011_2015-2300.nc

ESM files would typically look like this:

iareafl_GrIS_NORCE_CISM3_m001_xESMx_f001_historical_E001_xxxx-2014.nc

iareafl_GrIS_NORCE_CISM3_m001_xESMx-WACCM_f001_sspnnn_E002_2015-2300.nc

iareafl_GrIS_NORCE_CISM3_m001_xESMx_f002_historical_E001_xxxx-2014.nc

iareafl_GrIS_NORCE_CISM3_m002_xESMx-WACCM_f001_sspnnn_E002_2015-2300.nc

...

PPE files would typically look like this:

iareafl_GrIS_NORCE_CISM3_m001_xESMx_f001_historical_P001_xxxx-2014.nc

iareafl_GrIS_NORCE_CISM3_m001_xESMx-WACCM_f001_sspnnn_P002_2015-2300.nc

iareafl_GrIS_NORCE_CISM3_m002_xESMx_f001_historical_P001_xxxx-2014.nc

iareafl_GrIS_NORCE_CISM3_m002_xESMx-WACCM_f001_sspnnn_P002_2015-2300.nc

iareafl_GrIS_NORCE_CISM3_m002_xESMx_f002_historical_P001_xxxx-2014.nc

iareafl_GrIS_NORCE_CISM3_m002_xESMx-WACCM_f002_sspnnn_P002_2015-2300.nc

...

Guidance for reducing the size of files

The size of the model files on higher-resolution grids can be reduced by file compression, which will save space on the storage server. Example commands are given below. Using these commands, we can get 10x compression, and for the masks even more given that contiguous masks contain repeated data. NetCDF files have been designed with compression in mind. A NetCDF file can be compressed without changing the way it is read into Matlab or Python (or any other language that uses standard NetCDF read/write libraries).

The nccopy command copies an input netCDF file to an output netCDF file after compressing the file significantly. The ‘-d’ option stands for the deflation level, from 1 (faster but lower compression) to 9 (slower but more compression). The ‘-s’ option is the shuffling option to improve compression even more. We recommend using the ‘d1’ option, which seems to accomplish the desired compression.

Example of netcdf compression command:

nccopy -d1 -s sftgif_GIS_JPL_ISSMPALEO_historical.nc sftgif_GIS_JPL_ISSMPALEO_historical_c.nc

Example of compression variant, seems to work better for masks:

nccopy -d1 sftgif_GIS_JPL_ISSMPALEO_historical.nc sftgif_GIS_JPL_ISSMPALEO_historical_c.nc

ISMIP7 Focus Groups

More information on specific aspects of the protocol can be found on each of the Focus Group pages: