A repository-hosted dataset of volcanic ash particle images and features
Volcanic ash is made of particles smaller than 2 mm and is produced during volcanic eruptions. Studying the properties of volcanic ash is key for a range of applications, including volcano monitoring. However, given the large variety of textures, colors, and shapes of particles from different eruptions and volcanoes it is challenging to classify them in a reproducible and systematic manner. Here, we present an open access repository-hosted dataset of 11,808 particles from 53 samples across 13 volcanoes. Particle data includes optical microscope images, physical characteristics of the shape, texture and color, and a classification label with the particle type (free-crystal, altered material, juvenile or lithic). The dataset has already been used to develop particle classifiers with Machine Learning. We plan to continue it to grow and encourage the volcano community to contribute towards standardizing volcanic ash classification.
Authors
- Claudia D’Oriano (ORCID: https://orcid.org/0000-0003-2237-5121)
- Hamdi Rifai (ORCID: https://orcid.org/0000-0002-3967-5019)
- Daniel W. J. Lee (ORCID: https://orcid.org/0000-0002-3251-6808)
- Dini Nurfiani (ORCID: https://orcid.org/0000-0002-3169-0978)
- Damià Benet (ORCID: https://orcid.org/0000-0002-5186-7048)
- Fidel Costa (ORCID: https://orcid.org/0000-0002-1409-5325)
- Massimo Pompilio (ORCID: https://orcid.org/0000-0002-0742-0679)
- Kévin Migadel
Institutions
- Institut de physique du globe de Paris (FR)
- Lamont-Doherty Earth Observatory (US)
- Nanyang Technological University (SG)
- Université Paris Cité (FR)
- State University of Padang (ID)
- Istituto Nazionale di Geofisica e Vulcanologia (IT)
- Earth Observatory of Singapore (SG)
- National Research and Innovation Agency (ID)
- Columbia University (US)
Publication Details
- Journal
- Scientific Data
- Published
- 2025-04-23
- DOI
- https://doi.org/10.1038/s41597-025-04942-9
- Primary Topic
- Geochemistry and Geologic Mapping
- Type
- article
- Field-Weighted Citation Impact
- 0.00