Home
Projects
From space to summit: machine learning approaches for fusing diverse satellite and ground-based measurements of volcanic activity

From space to summit: machine learning approaches for fusing diverse satellite and ground-based measurements of volcanic activity

Year: 2026
Primary Supervisor: Susanna Ebmeier <[email protected]>
Institution: University of Leeds (School of Earth and Environment)
Possibility of becoming a CASE project: Yes
Collaborative Project: No
Academic Supervisors: David Ferguson <[email protected]> (University of Leeds, School of Earth and Environment), Geoff Kilgour, Ian Hamling
Research Themes: Tectonics, Volcanology and Solid Earth Geophysics
Project Partners: GNS New Zealand
Research Keywords: Satellite Data, Satellite Radar, Volcanic Eruptions, Volcano, Volcano Deformation
Relevant Degree Courses: Computer Science, Earth Science, Geological Science, Geophysics, Physics

Forecasting hazard associated with volcanic activity requires data – ideally multi-parameter measurements that capture different facets of volcanic activity. The most sensitive measurements of volcanic unrest typically come from local geophysical monitoring networks (seismicity, displacement and degassing), while a bigger picture understanding of processes within the Earth’s crust comes from analysis of erupted rocks. However, observations from space can now be made on a regional to global scale (deformation, thermal and gas emissions) and are providing new insight into volcanic unrest, especially in otherwise sparsely monitored settings. Interpreting such diverse data is challenging, and relies on both a conceptual understanding of the underlying volcanological processes, and robust approaches to statistical comparison.

This project will integrate satellite-derived time series of radar backscatter, displacement (InSAR), optical and infrared data with ground-based observations. The student will develop statistical and machine learning approaches for combining these time series in a way that is meaningful for volcanological questions. For example, (1) to detect transitions in unrest behaviour that may indicate an increase in hazard (e.g., Dualeh et al., 2023), (2) to connect impacts on surrounding ecosystems to eruption/unrest characteristics (e.g., Udy et al., 2024) and (3) to establish causal links between unrest in neighbouring systems (e.g. Reddin et al., 2023). The student will develop particular expertise in InSAR methods and in time series analysis using blind source separation methods and machine learning, skills highly valuable for future work in research and industry.

Case studies will include the Vanuatu volcanic arc, which stretches ~1200 km in the southwestern Pacific, includes fifteen major volcanoes, of which seven are currently active. Recent eruptions in Vanuatu (e.g., Hamling et al., 2019 ) have led to the evacuations of homes and damage to crops, water supplies and therefore livelihoods of surrounding communities.

The student will be supervised by scientists at the University of Leeds (Ebmeier, Ferguson) and at GNS, New Zealand (Hamling, Kilgour) and will part of an internationally leading group for satellite radar (InSAR) in Leeds. GNS (www.gns.cri.nz ) are CASE partners for this project and will host the PhD student for research visits during the project. In the UK the student will be part of the Centre for the Observations and Modelling of Earthquakes, Volcanoes and Tectonics (COMET), which will provide opportunities for training as well as a network of scientists using Earth Observations to research geohazards.