CoastalBench: A Decade-Long High-Resolution Dataset to Emulate Complex Coastal Processes

Data Availability

Complete Dataset (Open Storage Network)

The full CoastalBench dataset is publicly available on the Open Storage Network (OSN), hosted in an S3-compatible bucket with anonymous access enabled.

Bucket Information:

• Endpoint: https://udel1.osn.mghpcc.org
• Bucket Name: coastalbench-data
• Access: Anonymous (no credentials required)

Download Using Python

We provide a ready-to-use Python script for downloading the dataset:

# Install required package
pip install boto3

# Download all files
python download.py

# Download to a specific directory
python download.py --output-dir /path/to/data

# List files only (without downloading)
python download.py --list-only

# Download specific files (e.g., only .nc files)
python download.py --file-pattern "*.nc"

See download.py for the complete implementation.

Download Using AWS CLI

First, install AWS CLI if you haven't already.

List all files:

aws s3 ls s3://coastalbench-data --endpoint-url https://udel1.osn.mghpcc.org --no-sign-request --recursive

Download all files:

aws s3 sync s3://coastalbench-data . --endpoint-url https://udel1.osn.mghpcc.org --no-sign-request

Download a specific file:

aws s3 cp s3://coastalbench-data/your-file.nc . --endpoint-url https://udel1.osn.mghpcc.org --no-sign-request

Preview Dataset (Hugging Face)

A 90-day preview subset is available on Hugging Face for quick exploration: https://huggingface.co/datasets/YupuZ/CoastalBench-preview

Citation

Please cite our paper if you find this code useful for your work:

@inproceedings{coastalbench,
  title={CoastalBench: A Decade-Long High-Resolution Dataset to Emulate Complex Coastal Processes},
  author={Xu, Zelin and Zhang, Yupu and Xiao, Tingsong and Lizaso, Maitane Olabarrieta and Gonzalez-Ondina, Jose M and Liu, Zibo and Chen, Shigang and Jiang, Zhe},
  booktitle={Forty-second International Conference on Machine Learning},
  year={2025},
}