Source code of Surrogate modeling based History Matching for an Earth system model of intermediate complexity, accepted paper and poster at NeurIPS CCAI Workshop 2023.
This work is in continuity with the repository L96HistoryMatching, so you need to set up the same environment. To install the Python and R environment, use conda command:
conda env create -f environment.yml
For the rest of the setup, please read the instructions.md file.
- The DATA folder contains the outputs from the Experiments notebooks.
- The OUT_DATA folder contains the inputs used for iLOVECLIM experiments.
- The Experiments folder contains the experiment notebooks, as well as their parameters.
To reproduce an experiment, start with the HM-TuningILC notebook. Indicate the name of the experiment you want to run, and start running. If you want to create your own experiment, you can create a params.json file, on the model of the json in the Experiments/parameters/simulations folder. You will then need to run HM-TuningILC and Get-metrics alternatively starting with the former, as well as runnning iLOVECLIM. Don't forget to precise the name of the experiment you wanna run in the beginning of both notebooks (choose_exp variable).
You can contact the team on this website.