ai2cm / physics_standalone

Standalone version of the FV3GFS physical parametrizations

FV3GFS Physical Parameterization Standalone Versions

This repo contains standalone versions of a selection of the FV3GFS physical parametrizations. They can be run using inputs that have been serialized from an actual c48 run of fv3gfs and are validated against serialized data from the same run right after the parameterization has been called.

Requirements

You need the following things installed if you want to run directly on the host you are currently logged on to or if you are working directly on your laptop.

• serialbox installation (branch savepoint_as_string from https://github.com/VulcanClimateModeling/serialbox2/)
• gfortran
• wget

Running

First time only, if you run inside a docker container:

./build.sh

After that:

./enter.sh # only if you are running using the Docker environment
cd <parameterization>
./get_data.sh
cd ../runfile
python physics.py <parameterization> <backend> <--data_dir> <--select_tile> <--select_sp>

By default, data_dir=../parameterization/data, select_tile=All, and select_sp=All.

For example, to validate microphysics using gtx86 backend for all tiles and savepoints:

python physics.py microph gtx86

To validate tile 0 and savepoint 0:

python physics.py microph gtx86 --select_tile=0 --select_sp=cloud_mp-in-000000

Note: You may have to adapt the Makefile to point to your serialbox installation if it is not installed under /usr/local/serialbox.

Running without docker

If you prefer to work without a Docker environment, you can ignore the ./build.sh and ./enter.sh commands.

Code coverage

In order to inspect the code coverage that FV3GFS actually has (and the data used to run the parameterizations as well) take a look at the code coverage report.

Generating serialized data

git clone fv3gfs-fortran
cd fv3gfs-fortran
git pull
git checkout serialize_physics
./phys_build.sh
[take any rundir and copy it here]
./phys_run.sh
cd /rundir
export SER_ENV=TURB
./submit_job.sh