misomip1analysis

A python package for analyzing MISOMIP1 and ISOMIP+ simulation results

Instructions for ISOMIP+ and MISOMIP1

1. One-time setup

1.1 install miniconda

If you don't already have an anaconda python environment:

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
/bin/bash Miniconda3-latest-Linux-x86_64.sh

At the end of the install procedure, it will ask you if it should add a line to your .bashrc. Either do that or manually source the file it says to add to your .bashrc.

Add the conda-forge channel and make sure packages come from there whenever possible:

conda config --add channels conda-forge
conda config --set channel_priority strict

1.2 create a conda environment

Create an env. for running the code:

conda create -n misomip -c conda-forge -c xylar python=3.8 misomip1analysis

1.4. download or link to results

Either download the results from OSF or make local symlinks to your own ocean results. First, make a directory somewhere for the simulation results and the analysis

mkdir isomip+
cd isomip+

Then, make one or more subdirectories for the model results:

mkdir POP2x
mkdir MPAS-Ocean
...

These directories need to have the same model name that appears in the NetCDF file name:

COCO, FVCOM, MITgcm_BAS, MITgcm_JPL, MOM6, MOM6_SIGMA_ZSTAR,
MPAS-Ocean, NEMO-CNRS, NEMO-UKESM1is, POP2x, ROMSUTAS

If your files don't have the expected name, you can always make symlinks.

1.5 Make a config file

vim config.Ocean0_COM

Put the following in the file:

[experiment]
## Options related to the experiment being analyzed

# The name of the experiment (one of Ocean0-4 or IceOcean1-2)
name = Ocean0

# The "setup" of the experiment (if any), either COM or TYP
setup = COM

[models]
## Options related to the models to analyze

# A comma-separated list of models whose results should be analyzed
names = POP2x, MPAS-Ocean

There are many other config options that you can copy and modify.

For MISOMIP1 analysis, I have separated out the IceOcean1r and IceOcean1ra experiments for all submitted results that didn't do this. If you are testing your own results and just have IceOcean1 not separated into retreat and readvance phases, you can give the experiment name simply as IceOcean1. As you can see in the example config files, setup is one of COM_ocean or COM_ice.

2. Each time you run the analysis

2.1 activate the environment

source ~/miniconda3/etc/profile.d/conda.sh
conda activate misomip

2.2 Modify the config file (if you need to)

See misomip1analysis/config.default for all of the possible config options you can change.

vim config.Ocean0_COM

Examples can be found in the configs directory in the GitHub repo.

2.3 Run the analysis

misomip1analysis config.Ocean0_COM

The results will appear in the analysis folder

3. Advanced: editing get the code

If you're ambitious and want to edit the code, this will get you started.

Clone from GitHub:

git clone git@github.com:xylar/misomip1analysis.git
cd misomip1analysis

Create an env. for running the code:

conda create -n misomip python=3.8 xarray dask netcdf4 numpy scipy \
    matplotlib progressbar2 ffmpeg cmocean

Make a symlink to the inner misomip1analysis directory (in the same directory as setup.py) where you have the results downloaded. Then, run the analysis with something like:

python -m misomip1analysis config.Ocean0_COM

where the config file is created in the same way as above.