ClimateLearn - Climate Projection

← Back to all benchmarks

Date: 2023-07-19

Name: ClimateLearn - Climate Projection

Domain: Climate & Earth Science

Focus: ML for weather and climate modeling

Task Types: Forecasting

Metrics: RMSE, Anomaly correlation

Models: CNN baselines, ResNet variants

AI/ML Motif: Regression

Resources

Benchmark: Visit

Datasets: ClimateLearn GitHub Repository (data loaders and processing)

Results: ClimateLearn Paper (results section)

Keywords

medium-range forecasting ERA5 data-driven

Citation

• Tung Nguyen, Jason Jewik, Hritik Bansal, Prakhar Sharma, and Aditya Grover. Climatelearn: benchmarking machine learning for weather and climate modeling. 2023. URL: https://arxiv.org/abs/2307.01909, arXiv:2307.01909.

@misc{nguyen2023climatelearnbenchmarkingmachinelearning, 
  title={ClimateLearn: Benchmarking Machine Learning for Weather and Climate Modeling}, 
  author={Tung Nguyen and Jason Jewik and Hritik Bansal and Prakhar Sharma and Aditya Grover},
  year={2023}, eprint={2307.01909}, 
  archivePrefix={arXiv}, 
  primaryClass={cs.LG},
  url={https://arxiv.org/abs/2307.01909}
}

Ratings

CategoryRating

Software

5.00

Quickstart notebook makes for easy usage

Specification

5.00

Task framing (medium-range climate forecasting), input/output formats, and evaluation windows are clearly defined; benchmark supports both physical and learned models with detailed constraints.

Dataset

5.00

Provides standardized access to ERA5 and other reanalysis datasets, with ML-ready splits, metadata, and Xarray-compatible formats; versioned and fully FAIR-compliant.

Metrics

5.00

ACC and RMSE are standard, quantitative, and appropriate for climate forecasting; well-integrated into the benchmark, though interpretation across domains may vary.

Reference Solution

5.00

A Quickstart notebook is provided that uses ResNet as a baseline model

Documentation

5.00

Explained in the benchmark's paper.

Average rating: 5.00/5

Radar plot

Edit: edit this entry