Ergodic seismic precursors and transfer learning for short term eruption forecasting at data scarce volcanoes

Nature Communications, Journal Year: 2025, Volume and Issue: 16(1)

Published: Feb. 25, 2025

Abstract Seismic data recorded before volcanic eruptions provides important clues for forecasting. However, limited monitoring histories and infrequent restrict the available training forecasting models. We propose a transfer machine learning approach that identifies eruption precursors—signals consistently change eruptions—across multiple volcanoes. Using seismic from 41 at 24 volcanoes over 73 years, our forecasts unobserved (out-of-sample) Tested without target volcano, model demonstrated accuracy comparable to direct on exceeded benchmarks based amplitude. These results indicate precursors exhibit ergodicity, sharing common patterns allow observations one group of approximate behavior others. This addresses limitations individual sites useful tool support efforts volcano observatories, improving ability forecast mitigate risks.

Language: Английский

---

Sub-hourly forecasting of fire potential using machine learning on time series of surface weather variables

International Journal of Wildland Fire, Journal Year: 2025, Volume and Issue: 34(1)

Published: Jan. 29, 2025

Background Rapidly developing pre-fire weather conditions contributing to sudden fire outbreaks can have devastating consequences. Accurate short-term forecasting is important for timely evacuations and effective suppression measures. Aims This study aims introduce a novel machine learning-based approach potential test its performance in the Sunshine Coast region of Queensland, Australia, over period 15 years from 2002 2017. Methods By analysing real-time data local stations at sub-hourly temporal resolution, we aimed identify distinct patterns occurring hours days before fires. We trained random forest learning models classify conditions. Key results The achieved high out-of-sample accuracy, with 47% higher accuracy than standard danger index region. When simulating real conditions, model anticipated 75% fires (11 out 15). Conclusions method provides objective, quantifiable information, enhancing precision effectiveness warning systems. Implications proposed supports decision-makers implementing measures, ultimately reducing impact

Language: Английский

---

Eruption Forecasting Model for Copahue Volcano (Southern Andes) Using Seismic Data and Machine Learning: A Joint Interpretation with Geodetic Data (GNSS and InSAR)

Seismological Research Letters, Journal Year: 2024, Volume and Issue: 95(5), P. 2595 - 2610

Published: May 29, 2024

Abstract Anticipating volcanic eruptions remains a challenge despite significant scientific advancements, leading to substantial human and economic losses. Traditional approaches, like volcano alert levels, provide current states but do not always include eruption forecasts. Machine learning (ML) emerges as promising tool for forecasting, offering data-driven insights. We propose an ML pipeline using volcano-seismic data, integrating precursor extraction, classification modeling, decision-making alerts. Testing on six Copahue demonstrates our model’s ability identify precursors issue advanced warnings pseudoprospectively. Our model provides alerts 5–75 hr before achieving high true negative rate, indicating robust discriminatory power. Integrating short- long-term data reveals seismic sensitivity, emphasizing the need comprehensive monitoring. approach showcases ML’s potential enhance forecasting risk mitigation. In addition, we analyze geodetic (Interferometric Synthetic Aperture Radar Global Navigation Satellite System) assess deformation trends, in which notice absence of noteworthy signals associated with small eruptions, aligning their magnitude.

Language: Английский

---

WOVOdat web service data retrieval system for comprehensive volcano monitoring

Bulletin of Volcanology, Journal Year: 2025, Volume and Issue: 87(3)

Published: Feb. 26, 2025

Language: Английский

---

Forecasting Eruptions at Steamboat Geyser: Time Scales, Differentiability, and Detectability of Seismic Precursors Through Data‐Driven Methods

Journal of Geophysical Research Machine Learning and Computation, Journal Year: 2025, Volume and Issue: 2(2)

Published: April 12, 2025

Abstract Geyser eruptions provide a test bed for using geophysical data to forecast and understand heat mass transport in hydrothermal systems. We used time series analyses of seismic at Steamboat Geyser, Yellowstone National Park, identify short‐term precursors that are recurrent, detectable real time, distinctly identifiable, as well being rare during non‐eruptive periods. analyzed from March December 2018 patterns occurred before 31 eruptions. Four amplitude measures 700 time‐series features were computed the data. A template matching analysis identified an optimal 18‐hr window detecting precursors. applied random forest classify pre‐eruptive out‐of‐sample (eruptions not included model's training data), showing ability distinguish between two. This model performed better than simpler amplitude‐based approach. Seismic with most predictive power include autocorrelations, longest strike above mean, change quantiles, particularly within 4.5–16 Hz frequency range. isotonic regression, method converts raw outputs into calibrated probabilities, improve interpretability eruption forecasting outputs. The likelihood reaches 12.6% 18 hr prior event, representing marginal increase over static 8% probability derived solely intervals. Unlike interval‐based approach, our does rely on since last eruption, instead real‐time detect precursory signals. Our study advances Machine Learning methodologies by integrating estimation through which has advantages traditional approaches geysers highly irregular

Language: Английский

---

Probabilistic, Multi‐Sensor Eruption Forecasting

Geophysical Research Letters, Journal Year: 2025, Volume and Issue: 52(8)

Published: April 23, 2025

Abstract We developed an eruption forecasting model using data from multiple sensors or streams with the Bayesian network method. The generates probabilistic forecasts that are interpretable and resilient against sensor outage. applied at Whakaari/White Island, andesite island volcano off coast of New Zealand, seismic tremor recordings, earthquake rate, CO 2 , SO H S emission rates. At our shows increases in probability months to weeks prior three explosive eruptions were recorded between 2013 2019. Our outperforms use any sets alone as indicator for impending eruptions. Although can be easily adapted other volcanoes, complementing existing methods rely on single streams.

Language: Английский

---

Preface to the Focus Section on Volcano Monitoring in the Americas

Seismological Research Letters, Journal Year: 2024, Volume and Issue: 95(5), P. 2577 - 2579

Published: Aug. 2, 2024

Research Article| August 02, 2024 Early Publication Preface to the Focus Section on Volcano Monitoring in Americas Alicia J. Hotovec‐Ellis; Hotovec‐Ellis * 1U.S. Geological Survey, California Observatory, Moffett Field, California, U.S.A. *Corresponding author: [email protected] https://orcid.org/0000-0003-1917-0205 Search for other works by this author on: GSW Google Scholar Ricardo Garza‐Girón; Garza‐Girón 2Department of Geosciences, Warner College Natural Resources, Colorado State University, Fort Collins, Colorado, https://orcid.org/0000-0001-9775-9635 Gregory P. Waite; Waite 3Department and Mining Engineering Sciences, Michigan Technological Houghton, Michigan, Cristian Farías; Farías 4Departamento de Obras Civiles y Geología, Universidad Católica Temuco, Chile Susana Layana; Layana 5Millennium Institute Volcanic Risk – Ckelar Volcanoes, Antofagasta, https://orcid.org/0000-0002-0185-373X Matthew M. Haney 6U.S. Alaska Anchorage, Alaska, https://orcid.org/0000-0003-3317-7884 Author Article Information Publisher: Seismological Society America First Online: 02 Aug Online ISSN: 1938-2057 Print 0895-0695 © Letters (2024) https://doi.org/10.1785/0220240270 history Cite View This Citation Add Manager Share Icon Facebook Twitter LinkedIn Email Permissions Site Hotovec‐Ellis, Garza‐Girón, Waite, Farías, Layana, Haney; Americas. 2024; doi: Download citation file: Ris (Zotero) Refmanager EasyBib Bookends Mendeley Papers EndNote RefWorks BibTex toolbar search Dropdown Menu input auto suggest filter your All ContentBy SocietySeismological Advanced From Andes Aleutian Islands, are rich with volcanism that spans a diverse range tectonic settings, eruptive styles, levels activity, hazards. Over past 120 yr, have witnessed catastrophic volcanic eruptions significantly impacted nearby populations. Notable events include 8 May 1902, pyroclastic density current from Mount Pelée Martinique, which resulted loss over 28,000 lives Saint‐Pierre, only one or two survivors (Lacroix, 1904), 1985 eruption Nevado del Ruiz Colombia, triggered lahar left an estimated... You do not access content, please speak institutional administrator if you feel should access.

Language: Английский

---