Axial Seamount experiment to test real-time eruption forecasts

Currently, scientists struggle to forecast volcano eruption events, as no universally reliable, real-time eruption forecasting framework is available. Instead, researchers often rely on retrospective analysis to evaluate eruptions. And although much has been learned from doing this, it can sometimes introduce biases, such as data snooping, hindsight reinterpretation, and post-eruption model adjustment.

As a potential remedy to this problem, a group of researchers working with the Geohazards Crisis Observatory have launched an ongoing experiment focused on developing a physics-based eruption forecasting framework. The findings are published on the arXiv preprint server.

A bias-proof experimental design

The experiment ultimately hopes to test out two hypotheses, the first being that volcanic eruptions can be forecasted in real time through the evaluation of certain patterns known to reveal the system’s “approach to catastrophic failure.” The second hypothesis is essentially that the timing of these volcanic eruptions can be forecasted reasonably well using probability.

However, a crucial part of the project involves reducing bias and creating a transparent and reproducible model. To do this, the team will create forecasts for Axial Seamount—an active submarine volcano off the coast of Oregon—each month before the next eruption.

These forecasts will be cryptographically hashed and archived before public release, ensuring integrity and preventing post-hoc changes. They plan to keep a master “meta-document” going, which includes a brief description of the methods and datasets, and then upload each version of the document to a publicly accessible archive, like arXiv.

The study authors write, “After the forecasted event has occurred, we will publish the complete set of forecast documents in an open-access repository (e.g. arXiv.org) and on the website of the Geohazards Crisis Observatory (www.geohazards-observatory.com). The corresponding hash digests will be cross-verified to demonstrate consistency with the original hashes recorded in the meta-document. A summary of all forecasts and their outcomes will also be included in the final release.”

They say that all forecasts will be released, regardless of whether or not they successfully predicted the event, and that selective reporting or modification of forecasts is not possible, because each version is permanently timestamped and publicly archived. Then, after the actual eruption occurs, they plan to go back and assess the predictions and improve the model accordingly.

Axial Seamount’s active past

The Axial Seamount erupted in 1998, 2011, and 2015, and another eruption is on the horizon, according to scientists. Initially, the eruption was predicted to occur in 2025, but the study authors predict in their recent analysis that the next eruption is more likely to happen in mid to late 2026. The change in predicted timing essentially came down to a slowing of the measured seafloor uplift—or inflation. Still, the researchers say the inflation is already higher than it was during the 2015 eruption.

The Axial Seamount is also one of the most highly-monitored volcanoes globally, making it an easy choice for developing prediction models. The site is regularly monitored by a network of four cabled bottom pressure recorders for measuring seafloor uplift and seismometers for recording earthquake activity. This data can be used to create a physics-based prediction model.

The study authors write, “This unique geophysical dataset provides an unprecedented opportunity to investigate pre-eruptive unrest processes and to test quantitative models for eruption forecasting. The 2015 eruption was previously anticipated about seven months in advance, within a one-year window, based on pattern recognition in the geodetic time series.

“This forecast was formulated empirically from the observed inflation pattern rather than derived from a physics-based model. However, subsequent applications of this technique have not produced reliable forecasts due to variable inflation rates, highlighting the need for physics-based forecasting frameworks.”

A hope for better future eruption predictions

The hope for this experiment is not so much to accurately predict the Axial Seamount eruption this time around, as it is to inform the development of more robust, physics-based forecasting tools, while also increasing public trust in scientific forecasts and creating a better understanding of the scientific process. Improved eruption forecasts could eventually enhance hazard preparedness for coastal communities and marine operations.

Study authors Bill Chadwick, from Oregon State University, and Scott Nooner from University of North Carolina at Wilmington, chronicle the experiment’s progress in their blog.

© 2025 Science X Network