A repository including the code necessary to reproduce the results present in Immorlano et al. 2023, “Transferring climate change knowledge” submitted to Nature on September 21, 2023.
- Related Publication
- Installation
- Run a demo version
- Run the full version
- Reproduce the results present in the paper
- Contributors
- Acknowledgements and References
- License
Immorlano, F., Eyring, V., le Monnier de Gouville, T., Accarino, G., Elia, D., Aloisio, G. & Gentine, P. Transferring climate change knowledge. arXiv preprint. DOI: arXiv.2309.14780 (2023). (in review)
Python version 3.11.0 or higher is needed.
A conda env containing all the packages and versions required to run the workflow and/or reproduce the figures present in the paper can be created by running the following command:
conda env create --file transferring_env.yml
This makes the installation easy and fast. The conda env was created and tested on a MacBook M2 Pro with MacOS Ventura 13.4.1 to enable the user to run a demo version of the workflow and reproduce the results present in the paper.
On the other hand, Tensorflow version 2.12.0 was used to run the entire workflow on Ginsburg (Columbia University) and Juno (CMCC Foundation) supercomputers.
architectures.py→ Definition and building of the Deep Neural Network (DNN) used in the Transfer Learning approacharea_cella.csv→ Area weighting factors for each gridpoint of the grid corresponding toCanESM5_CanOE_gridused in this workCanESM5_CanOE_grid→ Description of the grid used in this work (Grid from CanESM5-CanOE simulation)transferring_env.yml→ YAML file needed to create the conda environment to reproduce the experiments and the figureserf_estimates_with_aerosols_Zebedee_Nichols.csv→ Dataset of Effective Radiative Forcing values for each yearlib.py→ Routines called in several scriptsvariables.py→ Definition of variables used in the scriptsBEST_regridded_annual_1979-2022.nc→ BEST observational maps resulting from the preprocessing (i.e., regridding to CanESM5-CanOE grid, computaion of annual average maps, delete of years out of 1979-2022)gaussian_noise_5→ Directory containing noisy BEST observational maps to be used for Transfer Learning on observationsLand_and_Ocean_global_average_annual.txt→ Average annual uncertainties associated with BEST observational data. The column headers were manually removed for ease of use
The following scripts should be used to download and process CMIP6 and Berkeley Earth Surface Temperatures (BEST) data:
CMIP6_download_process.py→ Download of CMIP6 simulations from Climate Data Store and processingBEST_data_processing.py→ Processing of BEST observational mapsBEST_data_add_gaussian_noise.py→ Generation of noisy BEST observational maps to be used for Transfer Learning on observations
The following script should be used to perform Training, Transfer Learning on Simulations (leave-one-out cross validation) and the Transfer Learning on Observations:
First_Training.py→ Training of each DNN on the simulation of one of the 22 Earth System Models (ESMs) for one of the 3 SSP scenariosTransfer_learning_simulations.py→ Transfer Learning on ESMs simulations according to the Leave-One-Out cross validation approachTransfer_learning_observations.py→ Transfer Learning on BEST observational data
The following scripts must be used to reproduce the figures and compute the values for Supplementary Table 1 present in the paper:
Fig_1.pyFig_2.pyFig_3.pyFig_4.pyExt_Fig_1.pyExt_Fig_2.pySupp_Table_1.py
A demo version fo the entire workflow can be run. After having downloaded the GitHub repository, all the files needed to run the demo shall be organized as the following hierarchy:
├── root
│ ├── architectures.py
│ ├── area_cella.csv
│ ├── BEST_data_add_gaussian_noise.py
│ ├── BEST_data_processing.py
│ ├── CanESM5-CanOE_grid
│ ├── CMIP6_download_process.py
│ ├── Demo_download
│ │ ├── Data
│ │ │ ├── BEST_data
│ │ │ │ ├── Land_and_Ocean_global_average_annual.txt
│ ├── Demo_no_download
│ │ ├── Data
│ │ │ ├── BEST_data
│ │ │ │ ├── BEST_regridded_annual_1979-2022.nc
│ │ │ │ ├── gaussian_noise_5
│ │ │ │ ├── Land_and_Ocean_global_average_annual.txt
│ │ │ ├── CMIP6_data
│ ├── erf_estimates_with_aerosols_Zebedee_Nichols.csv
│ ├── Figures
│ ├── First_Training.py
│ ├── lib.py
│ ├── Transfer_learning_observations.py
│ ├── Transfer_learning_simulations.py
│ ├── variables.py
The demo version can be run with two options:
- download and build a small dataset and then train the DNNs on those data
- start directly with training the DNNs on the same small dataset already downloaded and processed
In both cases, the workflow will be executed for CNRM-ESM2-1, FGOALS-f3-L and MIROC6 models simulations and for the SSP2-4.5 scenario.
To this aim, the BEST observational maps used in this study (i.e., Global Monthly Land + Ocean — Average Temperature with Air Temperatures at Sea Ice (Recommended; 1850 – Recent) — 1º x 1º Latitude-Longitude Grid) should be gathered from the BEST archive (direct download: 1º x 1º Latitude-Longitude Grid (~400 MB)) . The file must be named BEST_2022.nc and placed in root/Demo_download/Data/BEST_data.
Since CMIP6 data were gathered from the Copernicus Climate Data Store, the CDS API key should be configured on your laptop according to the CDS API documentation before downloading them.
The following variables in variables.py should be set to:
demo_download = Truedemo_no_download = False
Now, the scripts should be executed in the following order:
CMIP6_download_process.pyto download and process (i.e., regridding to CanESM5-CanOE grid, computing annual average maps) CMIP6 simulation data for the three models and for SSP2-4.5. The resulting nc files will be saved inroot/Demo_download/Data/CMIP6_data/near_surface_air_temperature/Annual_uniform_remapped.BEST_data_processing.pyto process BEST observational maps (i.e., regridding to CanESM5-CanOE grid, computing annual average maps, delete years out of 1979-2022). The resulting file will be saved asroot/Demo_download/Data/BEST_data/BEST_regridded_annual_1979-2022.ncBEST_data_add_gaussian_noise.pyto add noise to the BEST observational maps. Specifically, the noise is sampled from a Gaussian distribution with 0 mean and stddev equal to the BEST observational data uncertainty. The addition of noise is repeated 5 times for each model and each scenario. The resulting file will be saved inroot/Demo_download/Data/BEST_data/gaussian_noise_5.First_Training.pyto train an individual DNN on the simulation of one of the three ESMs for the SSP2-4.5 scenario. A total of three DNNs are trained, each with the same architecture. The results will be saved inroot/Demo_download/Experiments/First_Training/First_Training_[date-time].Transfer_learning_simulations.pyto transfer learn the trained DNNs on the ESMs simulations according to the Leave-One-Out cross validation (LOO-CV) approach. The variableFIRST_TRAINING_DIRECTORYinTransfer_learning_simulations.pyshould be set to the directory name related to the first training (i.e.FIRST_TRAINING_DIRECTORY = First_Training_[date-time]). This is necessary to load the pre-trained models. The results will be saved inroot/Demo_download/Experiments/Transfer_Learning_on_Simulations/Transfer_learning_[date-time]/Shuffle_[number]. TheShuffle_[number]corresponds to an iteration of the LOO-CV approach.Transfer_learning_observations.pyto transfer learn the trained DNNs on the BEST observational data. The variableFIRST_TRAINING_DIRECTORYinTransfer_learning_simulations.pymust be set to the directory name related to the first training (i.e.FIRST_TRAINING_DIRECTORY = First_Training_[date-time]). This is necessary to load the pre-trained models. The results will be saved inroot/Demo_download/Experiments/Transfer_Learning_on_Observations/Transfer_learning_obs_[date-time].
A demo version can be run directly starting with training the DNNs on the CNRM-ESM2-1, FGOALS-f3-L and MIROC6 models simulations.
In this case, the following variables in variables.py should be set to:
demo_download = Falsedemo_no_download = True
Now, the scripts should be executed in the following order:
First_Training.pyto train an individual DNN on the simulation of one of the three ESMs for the SSP2-4.5 scenario. A total of three DNNs are trained, each with the same architecture. The results will be saved inroot/Demo_no_download/Experiments/First_Training/First_Training_[date-time].Transfer_learning_simulations.pyto transfer learn the trained DNNs on the ESMs simulations according to the Leave-One-Out cross validation (LOO-CV) approach. The variableFIRST_TRAINING_DIRECTORYinTransfer_learning_simulations.pymust be set to the directory name related to the first training (i.e.FIRST_TRAINING_DIRECTORY = First_Training_[date-time]). The results will be saved inroot/Demo_no_download/Experiments/Transfer_Learning_on_Simulations/Transfer_learning_[date-time]/Shuffle_[number]. This is necessary to load the pre-trained models. TheShuffle_[number]corresponds to an iteration of the LOO-CV approach.Transfer_learning_observations.pyto transfer learn the trained DNNs on the BEST observational data. The variableFIRST_TRAINING_DIRECTORYinTransfer_learning_simulations.pymust be set to the directory name related to the first training (i.e.FIRST_TRAINING_DIRECTORY = First_Training_[date-time]). This is necessary to load the pre-trained models. The results will be saved inroot/Demo_no_download/Experiments/Transfer_Learning_on_Observations/Transfer_learning_obs_[date-time].
The Demo software was tested on a MacBook M2 Pro equipped with MacOS Ventura 13.4.1. The expected run times are the following:
CMIP6_download_process.py: about 12 seconds (after the CMIP6 data download)BEST_data_processing.py: about 122 secondsBEST_data_add_gaussian_noise.py: about 7 secondsFirst_Training.py,Transfer_learning_simulations.py,Transfer_learning_observations.py: about 4.5 seconds per epoch
The full version of the entire workflow can be run. After having downloaded Source_data.zip from Zenodo, the files needed to run the entire workflow shall be organized as the following hierarchy:
├── root
│ ├── architectures.py
│ ├── area_cella.csv
│ ├── BEST_data_add_gaussian_noise.py
│ ├── BEST_data_processing.py
│ ├── CanESM5-CanOE_grid
│ ├── CMIP6_download_process.py
│ ├── erf_estimates_with_aerosols_Zebedee_Nichols.csv
│ ├── First_Training.py
│ ├── lib.py
│ ├── Source_data
│ │ ├── BEST_data
│ │ │ ├── BEST_regridded_annual_1979-2022.nc
│ │ │ ├── gaussian_noise_5
│ │ │ ├── Land_and_Ocean_global_average_annual.txt
│ │ ├── CMIP6_data
│ │ ├── Transfer_Learning_on_Observations
│ │ ├── Transfer_Learning_on_Simulations
│ ├── Transfer_learning_observations.py
│ ├── Transfer_learning_simulations.py
│ ├── variables.py
The full version will be run by training the DNNs on the 22 ESMs simulations for SSP2-4.5, 3-7.0 and 5-8.5 scenarios.
In this case, the following variables in variables.py must be set to:
demo_download = Falsedemo_no_download = False
Now, the scripts should be executed in the following order:
First_Training.pyto train an individual DNN on the simulation of one of the 22 ESMs for the SSP2-4.5, 3-7.0 and 5-8.5 scenarios. A total of 66 DNNs are trained, each with the same architecture. The results will be saved inroot/Experiments/First_Training/First_Training_[date-time].Transfer_learning_simulations.pyto transfer learn the trained DNNs on the ESMs simulations according to the Leave-One-Out cross validation (LOO-CV) approach. The variableFIRST_TRAINING_DIRECTORYinTransfer_learning_simulations.pyshould be set to the directory name related to the first training (i.e.FIRST_TRAINING_DIRECTORY = First_Training_[date-time]). The results will be saved inroot/Experiments/Transfer_Learning_on_Simulations/Transfer_learning_[date-time]/Shuffle_[number]. This is necessary to load the pre-trained models. TheShuffle_[number]corresponds to an iteration of the LOO-CV approach.Transfer_learning_observations.pyto transfer learn the trained DNNs on the BEST observational data. the variableFIRST_TRAINING_DIRECTORYinTransfer_learning_simulations.pyshould be set to the directory name of the first training (i.e.FIRST_TRAINING_DIRECTORY = First_Training_[date-time]). This is necessary to load the pre-trained models. The results will be saved inroot/Experiments/Transfer_Learning_on_Observations/Transfer_learning_obs_[date-time].
The models resulting from transfer learning on BEST observational data and used to predict temperature values up to 2098 can be downloaded from Hugging Face
The figures and the results present in the paper can be reproduced. After having downloaded Source_data from Zenodo, the files needed to reproduce the results shall be organized as the following hierarchy:
├── root
│ ├── area_cella.csv
│ ├── Figures
│ │ ├── Ext_Fig_1.py
│ │ ├── Ext_Fig_2.py
│ │ ├── Fig_1.py
│ │ ├── Fig_2.py
│ │ ├── Fig_3.py
│ │ ├── Fig_4.py
│ │ ├── Supp_Table_1.py
│ ├── Source_data
│ │ ├── BEST_data
│ │ ├── CMIP6_data
│ │ ├── Transfer_Learning_on_Observations
│ │ ├── Transfer_Learning_on_Simulations
The following scripts should be used to reproduce the figures and compute the values for Supplementary Table 1 present in the paper:
Fig_1.pyFig_2.pyFig_3.pyFig_4.pyExt_Fig_1.pyExt_Fig_2.pySupp_Table_1.py
- Francesco Immorlano (francesco.immorlano@cmcc.it)
- Veronika Eyring (veronika.eyring@dlr.de)
- Thomas le Monnier de Gouville (thomas.le-monnier-de-gouville@polytechnique.edu)
- Gabriele Accarino (gabriele.accarino@cmcc.it)
- Donatello Elia (donatello.elia@cmcc.it)
- Giovanni Aloisio (giovanni.aloisio@cmcc.it)
- Pierre Gentine (pg2328@columbia.edu)
If you use the resource in your research, please cite our paper. At the moment, we offer the bibliography of the arXiv preprint version.
@misc{immorlano2023transferring,
title={Transferring climate change knowledge},
author={Francesco Immorlano and Veronika Eyring and Thomas le Monnier de Gouville and Gabriele Accarino and Donatello Elia and Giovanni Aloisio and Pierre Gentine},
year={2023},
eprint={2309.14780},
archivePrefix={arXiv},
primaryClass={physics.ao-ph}
}
MIT License
Copyright (c) 2023 Francesco Immorlano
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.