Author: Stephan Rasp - raspstephan@gmail.com - https://raspstephan.github.io
Hi, thanks for checking out this repository. This is a working repository, which means that the most corrent commit might not always be the most functional or documented.
A Guide for collaborators People hoping to collaborate with me on this project, please check out some guidelines here: https://github.com/raspstephan/CBRAIN-CAM/wiki/A-guide-for-collaborators
If you are looking for the version of the code that corresponds to the PNAS paper. Check out this release: https://github.com/raspstephan/CBRAIN-CAM/releases/tag/PNAS_final
The modified climate model code is available at https://gitlab.com/mspritch/spcam3.0-neural-net (branch: nn_fbp_engy_ess)
S. Rasp, M. Pritchard and P. Gentine, 2018. Deep learning to represent sub-grid processes in climate models https://arxiv.org/abs/1806.04731
P. Gentine, M. Pritchard, S. Rasp, G. Reinaudi and G. Yacalis, 2018. Could machine learning break the convection parameterization deadlock? Geophysical Research Letters. http://doi.wiley.com/10.1029/2018GL078202
The main components of the repository are:
cbrain: Contains the cbrain module with all code to preprocess the raw data, run the neural network experiments and analyze the data.pp_config: Contains configuration files and shell scripts to preprocess the climate model data to be used as neural network inputsnn_config: Contains neural network configuration files to be used withrun_experiment.py.notebooks: Contains Jupyter notebooks used to analyze data. All plotting and data analysis for the papers is done in the subfolderpresentation.devcontains development notebooks.wkspectra: Contains code to compute Wheeler-Kiladis figures. These were created by Mike S. Pritchardsave_weights.py: Saves the weights, biases and normalization vectors in text files. These are then used as input for the climate model.