Technical note: Integrating generative AI with physics-based simulations for probabilistic modelling of volcanic clouds

Mingari, Leonardo; Folch, Arnau; Pascual, Heribert; Titos, Manuel

doi:10.5194/acp-26-15953-2026

Articles | Volume 26, issue 21

https://doi.org/10.5194/acp-26-15953-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/acp-26-15953-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 26, issue 21

Technical note

|

11 Nov 2026

Technical note |

| 11 Nov 2026

Technical note: Integrating generative AI with physics-based simulations for probabilistic modelling of volcanic clouds

Leonardo Mingari, Arnau Folch, Heribert Pascual, and Manuel Titos

Supplement

https://doi.org/10.5194/acp-26-15953-2026-supplement

Data sets

Supplementary material for egusphere-2025-6097: Synthetic experiment L. Mingari https://doi.org/10.5281/zenodo.22759580

Supplementary material for egusphere-2025-6097: Etna experiment L. Mingari https://doi.org/10.5281/zenodo.22759573

Model code and software

FALL3D A. Folch et al. https://doi.org/10.5281/zenodo.17737081

Supplementary material for egusphere-2025-6097: Etna experiment L. Mingari https://doi.org/10.5281/zenodo.22759573

Supplementary material for egusphere-2025-6097: Synthetic experiment L. Mingari https://doi.org/10.5281/zenodo.22759580

Download

Article (2528 KB)
Full-text XML

Short summary

Forecasting volcanic clouds is challenging because it requires many computationally demanding simulations to capture uncertainty. We use artificial intelligence to generate large numbers of realistic forecasts quickly and inexpensively. Tests show that this approach can reproduce the range of possible volcanic cloud behaviours and improve predictions when combined with observations. This could help scientists and decision-makers better understand volcanic hazards.