I am a postgraduate researcher at the Institute of Geophysics and Tectonics. My research interests cover volcanology, specifically submarine volcanology, and how we can use global satellite datasets and machine learning approaches to investigate the prevalence and characteristics of explosive submarine eruptions. I am part of the SENSE CDT (2022) – Centre for satellite data in environmental science, funded by NERC and the UK space agency. I am supervised by Dr. Susi Ebmeier and Dr. Tim Craig at the University of Leeds, Dr. Isobel Yeo at the National Oceanography Centre and Dr. Ryan Lloyd at Geollect.
SENSE website: https://eo-cdt.org/
I previously studied MPhys Physics at the University of Oxford, completing my Master’s thesis as part of the Atmospheric, Oceanic and Planetary Physics department. My Master’s thesis was titled: Investigating evidence of a latitudinal-dependent lunar regolith compaction from the Diviner measured thermal signals.
- Submarine Volcanism
- Satellite Imagery
- Machine Learning
The majority of the Earth’s volcanoes are on the ocean floor, but direct observations of submarine eruptions are very rare. This means that fundamental characteristics of submarine volcanism, including eruption repeat times, remain largely unknown. Although only a small subset of submarine events will result in changes at the ocean surface, many of these are detectable in satellite imagery. Localised ocean colour change occurs both when erupted material is sufficiently shallow, and in the period after an eruption, when volcanic material may stimulate algal blooms (e.g., Urai & Machida, 2005). Eruptions that breach the ocean surface can produce distinctive sub-aerial plumes dominated by steam (e.g., Carey et al., 2014), as well as new, often transient, land. The most distinctive satellite signals are produced by pumice rafts, which can persist for long periods of time, travel great distances and pose a hazard to shipping (Mantas et al., 2011). Past observations of submarine eruptions from satellite imagery have required the manual analysis of satellite images and are limited to individual case studies. This project will develop a systematic approach to detecting submarine volcanic events from global satellite data sets.
This will be done using freely available, global satellite datasets, especially Sentinel-2 and MODIS imagery, to characterise the signals produced by historical submarine eruptions.
Preliminary work has identified a set of ten historical eruptions observable in MODIS or Sentinel-2 imagery, all of which were associated with ocean colour change, while four produced subaerial plumes and three major pumice rafts. The student will use these past eruptions as the starting point for training data sets for supervised classification of imagery, initially to search for evidence of other eruptions at the same volcano. The broader aim is to develop a classification approach that can be applied over larger regions of the world’s oceans. This will be achieved using cloud computing where possible, starting with algorithms tested in Google Earth Engine. The CASE partner, Geollect, will provide input into the machine learning approaches developed, with the aim for maintaining some transferability for other applications relating to shipping and safety at sea. For the subset of events producing pumice rafts, we will compare to the properties of available hand specimens at NOC (from Metis Shoal 2019, Le Havre 2012 and other recent events) to investigate relationships between satellite signals and the texture or composition of the erupted material.
- MPhys Physics, University of Oxford
Research groups and institutes
- Institute of Geophysics and Tectonics