Getting started¶

Install¶

The package is not on conda-forge yet, so install from a checkout:

git clone https://github.com/ismip/ismip7-scalar-processing.git
cd ismip7-scalar-processing
conda env create -f ismip7_scalars_env.yml
conda activate ismip7-scalars
python -m pip install --no-deps --no-build-isolation .

You get three commands:

ismip7-scalars processes one experiment
ismip7-scalars-ensemble processes every experiment in a submission tree
ismip7-scalars-set-params writes the density file each model needs

Installation says what that installs and why the pip flags matter.

Lay out your files¶

Suppose your group is VUW, your model is PISM1, and you want to process the Antarctic ssp585 projection C007 against the historical run C001. The tools expect this layout, which is the same as an ISMIP7 submission:

Models/AIS/                    <-- this is --modelpath
└── VUW/
    └── PISM1/
        ├── params.nc
        └── CORE/
            ├── C001/
            │   ├── lithk_AIS_VUW_PISM1_m001_CESM2-WACCM_f001_historical_C001_1850-2014.nc
            │   ├── topg_AIS_VUW_PISM1_m001_CESM2-WACCM_f001_historical_C001_1850-2014.nc
            │   └── ...
            └── C007/
                ├── lithk_AIS_VUW_PISM1_m001_CESM2-WACCM_f001_ssp585_C007_2015-2300.nc
                ├── topg_AIS_VUW_PISM1_m001_CESM2-WACCM_f001_ssp585_C007_2015-2300.nc
                └── ...

Data/AIS/                      <-- this is --datapath
├── af2_AIS_16000m_v1.nc
├── maxmask1_AIS_16000m_v0.nc
├── iaf2_GIC_AIS_16000m_v0.nc
└── basins_regions_AIS_Rignot_extended_16000m_v1.nc

Two things to notice:

The model path is the ice sheet directory, Models/AIS here: the one holding a folder per group. The tools add the group and model themselves. If you point it at your own model’s folder, Models/AIS/VUW/PISM1, the tools look for files under Models/AIS/VUW/PISM1/VUW/PISM1 and put params.nc there too.
The data files are the area factors and masks for your grid resolution. Download them from the ISMIP Globus server, under Output-Processing.

Models/AIS and Data/AIS are the defaults, so if you run from the directory above them you can leave off --modelpath and --datapath.

Write params.nc¶

Each model needs a small file recording the ice, sea-water and fresh-water densities it was run with:

ismip7-scalars-set-params --region AIS --group VUW --model PISM1 \
    --rhoi 910 --rhow 1028 --rhof 1000 --modelpath Models/AIS

This writes Models/AIS/VUW/PISM1/params.nc. There is no default because the wrong densities put a systematic error into the sea-level numbers.

Run it¶

ismip7-scalars --region AIS \
    --group VUW --model PISM1 --modelid m001 \
    --esm CESM2-WACCM --forcingid f001 \
    --experiment ssp585 --configid C007 --hist-configid C001 \
    --modelpath Models/AIS --datapath Data/AIS

The projection is C007. Its sea-level change is measured from the end of the historical run, C001. The resolution is read from the model grid.

What comes out¶

Output/
├── nc/AIS/VUW/PISM1/CORE/C007/
│   ├── slvaf_AIS_VUW_PISM1_m001_CESM2-WACCM_f001_ssp585_C007_1850-2300.nc
│   ├── slg20_...nc   sla20_...nc
│   ├── lim_...nc     limnsw_...nc   iareagr_...nc   iareafl_...nc
│   └── tendacabf_...nc  ...
└── csv/
    ├── slvaf_AIS_VUW_PISM1_...csv
    └── slvaf-gic_AIS_VUW_PISM1_...csv

NetCDF files go in a tree that mirrors your model tree. CSVs go in one flat directory, one row per file with a column per year from 1850 to 2300, so that rows from many models can be joined into one table. Change the root with --outpath.

When something is missing¶

If a required input is missing, the run prints one line starting with SKIP: saying what it could not find, and exits with code 2. Exit 0 means it finished. Anything other than 0 or 2 is a bug; please report it.

Some inputs are only needed for part of the output. Without sftgrf and sftflf the state scalars are skipped and everything else is still written. Each flux variable is skipped on its own if its file is missing.

Next¶

Running the processing for every option, The output files for what the files contain, and Processing a whole ensemble for processing a whole submission at once.