Digital Typhoon: Long-Term Satellite Image Dataset of Tropical Cyclones

Provider
Asanobu Kitamoto et al., National Institute of Informatics (Digital Typhoon project)
License
CC-BY-4.0 (source)
Size
4 files
Stored at
Hugging Face (https://huggingface.co/datasets/torchgeo/digital_typhoon)
Access
Free and public. No account with the data provider is needed.
Category
Climate & Weather
Source
https://arxiv.org/abs/2311.02665

About Digital Typhoon: Long-Term Satellite Image Dataset of Tropical Cyclones

Digital Typhoon is the longest typhoon satellite image dataset, spanning nearly 50 years, built by the Digital Typhoon project at Japan's National Institute of Informatics to benchmark machine learning on long-term spatio-temporal data. It contains typhoon-centred satellite images, cropped from the original imagery with an azimuthal equidistant projection centred on the best-track location and calibrated across satellites into a long-term, homogeneous record, with best-track metadata for each storm. The authors propose machine learning tasks by inference type and target, including meteorological analysis, societal impact and climate change (NeurIPS 2023 Datasets and Benchmarks). This TorchGeo copy holds the Western North Pacific (WP) subset as a split archive, WP.tar.gzaa and WP.tar.gzab (about 57.9 GB measured 2026-11-14), concatenated and extracted before use and read with torchgeo.datasets.DigitalTyphoon. Version 3 with more basins is distributed by the project. The project provides the dataset under CC-BY-4.0; attribute Digital Typhoon Dataset (National Institute of Informatics).

How to load this dataset in Python

from huggingface_hub import hf_hub_download
path = hf_hub_download(repo_id="torchgeo/digital_typhoon", filename="WP.tar.gzaa", repo_type="dataset")
print(path)

Data structure

Files

Example file: WP.tar.gzaa

Analyze this dataset without downloading it

On Scigantic, Digital Typhoon: Long-Term Satellite Image Dataset of Tropical Cyclones opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.

This page is also available as Markdown and as JSON.

Related Climate & Weather datasets