What is 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.
This project is under active development. The API might undergo extensive changes in the near future.
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
We have a quickstart notebook in the
notebooks folder titled
Quickstart.ipynb that walks through an example usage of ClimateLearn for
weather forecasting from downloading the data through visualizing the
predictions of a trained model. It is intended for use in Google Colab and can
be launched by clicking
Why did we build ClimateLearn?
In recent years, there has been a growing interest in the application of ML-based methods for weather and climate modeling. While there are some leaderboard benchmarks, such as WeatherBench, ClimateBench, and FloodNet, that propose datasets and baselines for specific tasks in climate science, a holistic software ecosystem that encompasses the entire data, modeling, and evaluation pipeline across several tasks is lacking. Hence, we built ClimateLearn to standardize datasets, model implementations, and evaluation protocols for rigorous and reproducible data-driven climate science.
ClimateLearn is built and maintained by the Machine Intelligence Group at UCLA, headed by Professor Aditya Grover.