Space-borne technology has long been used to spot the tell-tale signs that can help to predict when volcanoes will reawaken. Now scientists have combined machine learning with more than half-a-million Copernicus Sentinel-1 images with the aim of forecasting eruptions on a global scale.
Some 800 million people across the world live within 100 km of a volcano and reliable information on precursor activity is vital to plan timely evacuations and other mitigation strategies.
Sentinel-1 of the EU’s Copernicus Earth Observation Programme is one of many remote sensing missions that monitor volcanic processes.
Both Copernicus Sentinel-1 satellites host a C-band synthetic aperture radar (SAR) instrument that continues to gather observations during the day and night, and under all weather conditions.
Imagery from this sensor can be manually inspected by experts to identify changes in the shape of volcanoes and the land surrounding them, which may be a sign of red-hot magma stirring beneath the surface.
Global distribution of deformation detections
Recent science has demonstrated the potential of machine learning techniques for spotting this volcanic deformation, but these methods have only been tested on relatively small Earth observation datasets covering a limited range of volcano types.
Building on this research, a team of scientists based at the University of Bristol in the UK used a huge global collection of Copernicus Sentinel-1 images collected between 2015 and 2020 to further explore the success and limitations of machine learning for detecting deformation signals.
A three-stage approach was employed to analyse the dataset.
First, Copernicus Sentinel-1 SAR images were processed into maps of surface displacement – called interferograms – that were grouped together by location, so that deformation could be observed over time.
Using coordinates provided by the Global Volcanism Programme, 592 224 interferograms were generated, detailing 1084 separate volcanoes. The dataset had good coverage of volcanoes in Europe, Latin America and Asia.
Next, a pre-existing deep learning algorithm was used to classify the images and flag volcanoes displaying deformation.
Given the far greater risk posed by false negatives – which could result in a potential eruption being overlooked – as appose to false positives, a conservative detection threshold was employed in the analysis.
Finally, the researchers used a range of tools to inspect each volcano flagged by the deep learning algorithm, to confirm the presence of deformation or volcanic activity.
Across the whole dataset, the detection threshold for deformation was 5.9 cm, equivalent to a rate of 1.2 cm per year over the five-year study period.
Sixteen volcanoes were flagged in the analysis and, of these, five erupted during the study period, six experienced slow deformation, two displayed non-volcanic deformation, and three were false positives caused by atmospheric artefacts linked to extreme topography, snow fall and other factors.
Among the incidences of true deformation, several new discoveries were reported, including a change in the deformation rate at Wolf Island in the Galapagos, and a reversal in deformation at Fernandina Island, also in the Galapagos, that coincided with a minor eruption.
Deformation of Galapagos volcanoes
The analysis showed that individual interferograms are best for detecting eruptions and intrusions, which are characterised by sudden, large deformation signals, although slow deformation can be observed provided high-quality data are maintained over a long enough timeframe.
In addition, the researchers noted the need to develop tailored processing strategies for regions with dense vegetation or snow cover, as well as atmospheric corrections using global weather models to reduce false positives, in future research.
Fabien Albino, remote sensing scientist and co-author of the research, said, “This study is the first to demonstrate the powerful combination of automatically processed satellite data and machine learning on a large global dataset, to detect volcanic deformation.
“Although this is a retrospective analysis, the ultimate aim is to develop a real-time monitoring and alert system that will be used in combination with other sources of information to support evacuations and mitigation efforts.”
The algorithms presented in this analysis have been adapted to run in an open-access web portal that enables volcanologists to search quickly through flagged deformation signals.
The project was supported by the UK Centre for the Observation and Modelling of Earthquakes, Volcanoes and Tectonics and the UK’s Natural Environment Research Council.
Sentinel-2 captures the Galapagos Islands
About the Copernicus Sentinels
The Copernicus Sentinels are a fleet of dedicated EU-owned satellites, designed to deliver the wealth of data and imagery that are central to the European Union's Copernicus environmental programme. The European Commission leads and coordinates this programme, to improve the management of the environment, safeguarding lives every day.
ESA is in charge of the space component, responsible for developing the family of Copernicus Sentinel satellites on behalf of the European Union and ensuring the flow of data for the Copernicus services, while the operations of the Copernicus Sentinels have been entrusted to ESA and EUMETSAT.
Did you know that?
Earth observation data from the Copernicus Sentinel satellites are fed into the Copernicus Services. First launched in 2012, with the Land Monitoring and Emergency Management services, these services provide free and open support, in six different thematic areas.
The Copernicus Emergency Management Service (CEMS) provides all operators involved in the management of Major Disasters, man-made emergency situations, and humanitarian crises with timely and accurate geo-spatial information derived from satellite remote sensing and completed by available in situ or open data sources.
The Copernicus Land Monitoring Service (CLMS) provides geographical information on land cover and its changes, land use, vegetation state, water cycle and Earth's surface energy variables, to a broad range of users in Europe and across the World in the field of environmental terrestrial applications. It supports applications in a variety of domains such as spatial and urban planning, forest management, water management, agriculture and food security, nature conservation and restoration, rural development, ecosystem accounting and mitigation/adaptation to climate change.
References:
Biggs, J., Anantrasirichai, N., Albino, F. et al. Large-scale demonstration of machine learning for the detection of volcanic deformation in Sentinel-1 satellite imagery. Bull Volcanol 84, 100 (2022). https://doi.org/10.1007/s00445-022-01608-x.