Climate State Classifier

Software to train/evaluate models to classify labeled climate data based on a convolutional neural network (CNN).

Dependencies

• pytorch>=1.11.0
• tqdm>=4.64.0
• torchvision>=0.12.0
• torchmetrics>=0.11.2
• numpy>=1.21.6
• matplotlib>=3.5.1
• tensorboardX>=2.5
• tensorboard>=2.9.0
• xarray>=2022.3.0
• dask>=2022.7.0
• netcdf4>=1.5.8
• setuptools==59.5.0
• xesmf>=0.6.2
• cartopy>=0.20.2
• numba>=0.55.1

An Anaconda environment with all the required dependencies can be created using environment.yml:

conda env create -f environment.yml

To activate the environment, use:

conda activate climclass

environment-cuda.yml should be used when working with GPUs using CUDA.

Installation

climclass can be installed using pip in the current directory:

pip install .

Usage

The software can be used to:

• train a model (training)
• make predictions using a trained model (evaluation)

Input data

The input data samples must be given in the following naming convention, containing a single sample per file: <category_name><sample_name><data_type><class_label>.nc

Execution

Once installed, the package can be used as:

• a command line interface (CLI):
  • training:

  climclass-train

  • evaluation:

  climclass-evaluate

• a Python library:
  • training:

  from climatestateclassifier import train
  train()

  • evaluation:

  from climatestateclassifier import evaluate
  evaluate()

Example

An example application can be found in the directory demo. The instructions to run the example are given in the demo/README.md file.

License

Climate State Classifier is licensed under the terms of the BSD 3-Clause license.

Contributions

Climate State Classifier is maintained by the Data Analysis Department at DKRZ (Deutsches Klimarechenzentrum).

• Current contributing authors: Johannes Meuer, Claudia Timmreck, Shih-Wei Fang, Christopher Kadow.

References

@article{meuer2024fingerprints,
  title={Fingerprints of past volcanic eruptions can be detected in historical climate records using machine learning},
  author={Meuer, Johannes and Timmreck, Claudia and Fang, Shih-Wei and Kadow, Christopher},
  journal={Communications Earth \& Environment},
  volume={5},
  number={1},
  pages={455},
  year={2024},
  publisher={Nature Publishing Group UK London}
}