ClimateLearn

ClimateLearn is a Python library for accessing state-of-the-art climate data and machine learning models in a standardized, straightforward way. This library provides access to multiple datasets, a zoo of baseline approaches, and a suite of metrics and visualizations for large-scale benchmarking of statistical downscaling and temporal forecasting methods. For further context on our past motivation and future plans, check out our announcement blog post. Also, check out our arxiv preprint that showcases the flexibility of ClimateLearn in performing benchmarking and analysis on the robustness and transfer performance of deep learning models.

Usage

Python 3.8+ is required. The xESMF package has to be installed separately since one of its dependencies, ESMpy, is available only through Conda.

conda install -c conda-forge xesmf
pip install climate-learn

Quickstart

We have a quickstart notebook in the notebooks folder titled Quickstart.ipynb. It is intended for use in Google Colab and can be launched by clicking the Google Colab badge above or this link: https://colab.research.google.com/drive/1LcecQLgLtwaHOwbvJAxw9UjCxfM0RMrX?usp=sharing.

We also previewed some key features of ClimateLearn at a spotlight tutorial in the "Tackling Climate Change with Machine Learning" Workshop at the Neural Information Processing Systems 2022 Conference. The slides and recorded talk can be found on Climate Change AI's website.

Documentation

Find us on ReadTheDocs.

About Us

ClimateLearn is managed by the Machine Intelligence Group at UCLA, headed by Professor Aditya Grover.

Contributing

Contributions are welcome! See our contributing guide.

Citing ClimateLearn

If you use ClimateLearn in your research, please cite our paper:

@article{nguyen2023climatelearn,
  title={ClimateLearn: Benchmarking Machine Learning for Weather and Climate Modeling},
  author={Nguyen, Tung and Jewik, Jason and Bansal, Hritik and Sharma, Prakhar and Grover, Aditya},
  journal={arXiv preprint arXiv:2307.01909},
  year={2023}
}