ClimSim Online

Getting started

To evaluate online performance using Google Compute Engine Virtual Machine, follow these instructions.

The workflow can also be run on a local linux x86_64 system as well. Instructions for using local resources are in Local setup

Local setup

Clone the repo first

git clone [email protected]:leap-stc/climsim-online.git

Then update all the submodules

cd climsim-online
git submodule update --init --recursive

Building the container

Just run ./docker_build.sh

Running a control MMF simulation workflow

1. Ensure the right directories are created:

   a. mkdir inputdata to create the directory used for E3SM input data. This data will be auto-downloaded during the first E3SM run.

   b. mkdir scratch to use as the directory that hosts the binaries and all the output data from the run.

2. Launch the container using ./docker_launch.sh

3. Navigate to the climsim_scripts directory: cd /climsim/E3SM/climsim_scripts

4. Launch the control MMF simulation job using python example_job_submit_mmf.py

Running an online hybrid simulation workflow

1. Ensure the right directories are created:

   a. mkdir inputdata to create the directory used for E3SM input data. This data will be auto-downloaded during the first E3SM run.

   b. mkdir scratch to use as the directory that hosts the binaries and all the output data from the run.

   c. Extract the tarball downloaded from here: tar xvfz shared_e3sm.tar.gz. This will create a shared_e3sm that hosts all the NN models, normalization params, restart files, and reference e3sm simulation outputs needed.

2. Launch the container using ./docker_build.sh

3. Navigate to the climsim_scripts directory: cd /climsim/E3SM/climsim_scripts

4. Launch the control MMF simulation job using python example_job_submit_nnwrapper_v4_constrained.py

Evaluation

---

After you the hybrid simulation, to make the evaluation plot, you can go to the cd /climsim/E3SM/climsim_scripts directory and execute the following file:

python example_plot_monthly_rmse.py /storage/shared_e3sm/ /scratch/example_job_submit_nnwrapper_v4_constrained/run/

Change the first path to the shared_e3sm folder, and change the second path to the folder that contains the model output. This python script will generete a figure of monthly RMSE of temperature and moisture under climsim_scripts/figure.

To get access to this plot, you can copy it to the mounted scratch volume using

cp -r figure /scratch/example_job_submit_nnwrapper_v4_constrained/

License

climsim-online is provided under the Apache License 2.0, please see LICENSE.txt for full license text.