All posts by COMET2014

Iceland volcano collapse explained

COMET scientists have helped to shed new light on how volcanoes collapse during major eruptions, in new research published in Science.

The study, led by the University of Iceland, investigated a recent collapse at Bárdarbunga Volcano, Iceland, during the biggest volcanic eruption in Europe since the huge event at Laki in 1784.

Eruption column and lava flow from the air on 22 September 2014. Credit:Thórdís Högnadóttir
Eruption column and lava flow from the air on 22 September 2014. Credit: Thórdís Högnadóttir

The largest eruptions on Earth are commonly associated with collapse of the roof of a volcano into a magma chamber below. As they are infrequent, however – only five caldera collapses were recorded during the twentieth century – the processes involved are poorly understood.

The Bárdarbunga eruption, which lasted from August 2014 until February 2015, produced 1.5 km3 of basaltic lava.  In the course of the eruption, the top of the volcano caldera gradually sagged downwards, leaving an elongated bowl shaped depression over 13 km long and up to 65 m deep.

Bárdarbunga from the north, showing the main subsidence bowl within the caldera. Credit: Magnús T. Gudmundsson
Bárdarbunga from the north, showing the main subsidence bowl within the caldera. Credit: Magnús T. Gudmundsson

The total volume of the subsidence was 1.8 km3 – similar to the total volume of lava erupted and injected into the crust , implying a strong link between the two.

COMET scientist Marco Bagnardi said: “We saw the events at Bárdarbunga as an opportunity to better understand caldera collapse, and used multiple techniques to investigate.”

COMET researchers used radar data from satellites to measure ground deformation at Bárdarbunga’s caldera over a number of 24 hour periods.  As the topography continually changed, these data revealed movement of faults that reached to within a kilometre or so of the surface.

Combined with other techniques, including airborne altimetry, high precision GPS, seismology,  radio-echo soundings, and ice flow modelling, the results were used to create a detailed picture and timeline of how the caldera was collapsing and why.

From 16 August 2014 – before the eruption began – magma had been migrating out of a chamber 12 km below the ground, forming a fracture in the Earth’s crust.  Continued monitoring during the event showed that the magma was moving sideways from the volcano, beneath the surface, before finally erupting at Holuhraun, 47 km to the northeast, two weeks later.

Bardarbunga dike
The eruption in Holuhraun on its fourth day (3 September 2014). Credit: Thórdís Högnadóttir

Further analysis showed that a few days after the initial migration, the outflow of magma activated faults around the edge of the caldera leading to a series of earthquakes, which marked the beginning of the caldera collapse.  The collapsing roof then acted like a piston forcing even more magma out of the chamber below, which in turn led to further collapse.

COMET scientist Professor Andy Hooper explained: “Through modelling many different data sets, we were able to show that the caldera collapse was caused by magma leaving the reservoir, and this in turn squeezed more magma out of the reservoir, forming a positive feedback. This mechanism led to much more magma being erupted than would otherwise have been the case, which explains how eruptions on an even larger scale can occur.”

Overall, between 12 and 20% of the magma had left the magma chamber when the caldera collapse began.

Summarising the research, Prof Hooper added: “This work has given us real insight into caldera collapse, not only at Bárdarbunga but also at even larger eruptions.  What’s particularly interesting is how the collapse of the magma reservoir and the flow out of it clearly amplify each other.”

The paper, available now in Science, is Gudmundsson et al. (2016) Gradual caldera collapse at  Bárdarbunga volcano, Iceland, regulated by lateral magma outflow, doi:10.1126/science.aaf8988.

Coupling volcanic gas emission measurements to computational models of conduit gas flow

A new paper published in Geophysical Research Letters by Tom Pering and Andrew McGonigle has combined fluid dynamical modelling of gas flow in conduits with high time resolution measurements of volcanic gas discharge for the first time, revealing new insights into the dynamics of Stromboli volcano.

Their work is based on a recently developed approach using ultraviolet cameras which enable measurements of volcanic gas emission rates with unprecedented time resolution – around 1 Hz – such that gas release patterns associated with rapid explosive and non-explosive basaltic processes, can be resolved for the first time.

Data were captured on Stromboli, where an intriguing coda of lifetime on the order of 10s of seconds was identified following each explosion. Computational models were also developed to simulate the upward flow of conduit filling, so called “Taylor bubbles”, which are believed to be responsible for explosions on Stromboli when they burst at the surface.

The numerical models reveal the fissioning of smaller bubbles from the Taylor bubble bases to generate a train of “daughter bubbles”, thought to be responsible for generating the post-explosive coda upon arrival at the surface.

This process could play a primary yet hitherto unconsidered role in driving the dynamics of strombolian volcanism, both on Stromboli and other targets worldwide, with significant implications for the magnitude of resulting eruptions.

Combining models with field observations in this way shows considerable promise for improving our understanding of how gases drive volcanic activity.

The full reference is:  T.D Pering, A. J. S McGonigle, M. R James, G. Tamburello, A. Aiuppa, D. Delle Donne, M. Ripepe Conduit dynamics and post-explosion degassing on Stromboli: a combined UV camera and numerical modelling treatment, Geophysical Research Letters 2016 DOI: 10.1002/2016GL069001

Airborne volcanic ash detection using infrared spectral imaging

A new paper in Scientific Reports, co-authored by COMET’s Tamsin Mather, has demonstrated for the first time that airborne remote detection of volcanic ash is possible.

Airborne volcanic ash is a known hazard to aviation, but there are no current means to detect ash in-flight as the particles are too fine for on-board radar detection and, even in good visibility, ash clouds are difficult or impossible to detect by eye.

The economic cost and societal impact of the Icelandic eruption of Eyjafjallajökull generated renewed interest in finding ways to identify airborne volcanic ash in order to keep airspace open and avoid aircraft groundings.

The research, led by COMET Board Member Fred Prata, involved designing and building a bi-spectral, fast-sampling, uncooled infrared camera device (AVOID) to examine its ability to detect volcanic ash more than 50 km ahead of aircraft.

Experiments conducted over the Atlantic Ocean, off the coast of France involved an artificial ash cloud being created from a second aircraft, using ash from the Eyjafjallajökull eruption itself.

The measurements made by AVOID,  along with additional in situ sampling, confirmed the ability of the device to detect and quantify ash in an artificial ash cloud.  This is the first example of airborne remote detection of volcanic ash from a long-range flight test aircraft.

The full reference is Prata, A. J. et al. Artificial cloud test confirms volcanic ash detection using infrared spectral imaging. Sci. Rep. 6; doi: 10.1038/srep25620 (2016).

Detecting geohazards with GPS

COMET’s Marek Ziebart and Chris Atkins, both at UCL, have been using sidereal filtering to detect geohazards in near-real time.  This article describes how they have applied this to earthquakes in the Aleutian Islands, Alaska.

Geodetic quality GPS (Global Positioning System) data – fundamentally high precision range measurements made between orbiting spacecraft and tracking stations, often rigidly attached to bedrock – is one of the key observables available to us to analyse plate tectonics and the earthquake cycle. Such tracking stations now often make measurements at 1 Hz or higher. This enables us to calculate high rate position time series for the tracking stations, revealing various kinds of motion related to seismic waves and other forms of deformation.

 Figure 1: Deep braced GNSS tracking antenna at the Aleutian Islands, Cape Sarichef, Alaska (AV24-WestdahlNWAK2008). Photo:

Figure 1: Deep braced GNSS tracking antenna at the Aleutian Islands, Cape Sarichef, Alaska (AV24-WestdahlNWAK2008). Credit:

One such time series is shown below, calculated by a standard processing method called PPP (precise point positioning).  The plot shows the change in position in the north-south direction. When do you think the earthquake seismic waves begin to arrive?


Figure 2: 25 minutes of the 1Hz GPS time series for AV24 – a deep braced GNSS antenna on the Aleutian Islands, Cape Sarichef, Unimak Island, Alaska at the time of the 2011 Tohoku earthquake (WestdahlNWAK2008)

GPS Seismology

How can we use this type of data? The time series in figure 2 shows an estimate of the GPS antenna movement in the North-South direction during the 2011 magnitude 9.0 Tohoku earthquake over a period of 25 minutes. This is a measurement of the ground shaking – the data is used to understand the nature of the earthquake and to compare it to other earthquakes in different places around the world. Earthquakes often cause a change in the shape of the ground that is retained after the earthquake is over, and which can be determined from a time series. This kind of information can be used to estimate the size of the earthquake, how much energy was released and tells us about the earthquake’s epicentre location.

Another kind of instrument, a seismometer, can be used to determine similar information. The drawback of seismometers is that they can ‘clip’ when the seismic waves become too large to handle. A GPS antenna, on the other hand, has no such limitation – no matter how big the amplitude of the seismic waves, they can still be measured using GPS technology. Because of this, GPS antennas such as the ones shown in figures 1 and 6 below are sometimes called broadband GPS seismometers. In practice we combine information from both GPS antennas and seismometers – they are complimentary. There are now thousands of such GPS antennas installed around the world, connected to the internet and taking measurements 24 hours a day.


Figure 3: an example of a GPS tracking station network – this one is in the Gulf of Alaska and forms part of the US Plate Boundary Observatory (PBO)

So what’s the problem? The problem is Multipath – this is akin to radio interference where the signals transmitted by the orbiting satellites have more than one way of getting to the antenna. The strongest, direct signal – which is the one we want to measure – is distorted by signals bouncing off adjacent reflecting surfaces in the antenna’s vicinity. This causes spurious patterns of motion in the receiver time series. These can vary rapidly over a few seconds in a noise-like way (with amplitudes at the few mm level), or they can appear as a random walk – a slowly varying position error (with changes at the level of a few centimetres). Both these effects limit our ability to use GPS data for seismometry. To get some feel for the effect consider Figure 4.

On the day of the earthquake (shown as the black time series) the onset of the arrival of the largest seismic waves is shown clearly at around 6:03:00. A not dissimilar signature is also visible around 5:50:00. However, that same signature (an apparent ground movement of some 25mm over a few minutes, shaking and then reducing in magnitude) is visible at almost the same time on the previous day (in the grey time series). It doesn’t mean the earthquake was repeating – these are multipath distortions. They appear indistinguishable from seismic waves. Our problem, then, is how to remove the spurious, apparent motion from the time series, whilst retaining the actual signature of the earthquake. So what’s the solution?


Figure 4: the GPS time series (North-South component) for station AV24 on the day of the earthquake and on the day before. The time series on the vertical axis are offset by 60mm for clarity

Sidereal Filtering

Sidereal filtering is a technique used to reduce errors caused by multipath in the positioning of GPS receivers. It relies upon the receiver remaining static from one day to the next relative to its surrounding environment and takes advantage of the ground track repeat time of the GPS satellites, which is just less than a sidereal day (usually around 23 hours, 55 minutes, 55 seconds). The repeating multipath error can thus be identified and largely removed from the following day by subtraction.

A conventional position-domain sidereal filter (PDSF) identifies and removes this repeating pattern from a position time series. However, the ground track repeat time of individual GPS satellites can differ from each other by a few seconds, whereas a PDSF has to assume that all satellites have the same repeat time. This can cause problems for the PDSF, especially when the oscillations caused by multipath interference are particularly high. We have developed an observation-domain sidereal filter (ODSF) that identifies and removes multipath errors from the GPS phase measurements themselves.

Unlike the PDSF, it can account for the fact that the ground track repeat times of the GPS satellites differ by a few seconds from one another. This means that it is more effective at removing the effects of high-frequency multipath error and is less sensitive to satellite outages. For each phase measurement, the ODSF algorithm searches for an appropriate correction based on the azimuth and elevation of the relevant satellite. That correction is derived from the measurement residuals on the previous day that most closely correspond to that azimuth and elevation. The precision of these measures of azimuth and elevation needs to be high – a hundredth of a degree or better. This is readily achievable in our algorithm.


Figure 5: The same GPS time series for station AV24, but with the time series resulting from the two types of sidereal filter also shown. The time series on the vertical axis are offset by 50mm for clarity

Figure 5 shows the same 25-minute time series of 1 Hz displacements in northing as shown in Figure 4. Also plotted in red and green are the time series that result after applying the two types of sidereal filter: the PDSF and ODSF respectively. Notice that both the PDSF and the ODSF are largely successful at removing these oscillating errors. They reveal a surface wave arrival time of about 06:00 with the largest Love waves arriving at around 06:03. These events would not have been so easy to distinguish in the original unfiltered time series.


Figure 6: Deep braced GNSS tracking antenna on the Aleutian Islands, Cape Sarichef, Unimak Island, Alaska (AV27-WestdahlSWAK2008). Credit:


Figure 7: GPS time series for station AV27 including those resulting from the two types of sidereal filter. The time series on the vertical axis are offset by 50mm for clarity

Figure 7 similarly shows 1 Hz displacements at station AV27 which is only 11 km from AV24. One would therefore expect a very similar displacement signal. However, it is clear that the standard PPP time series, plotted in black and grey, are severely affected by strong short-period (~11 s) multipath error. In this case, the PDSF was unable to remove such a high-frequency error and instead increased the amplitude of these oscillations. The ODSF on the other hand was far more effective because it could take into account the differing repeat times of the satellites.


The precision of GPS technology has improved dramatically since 1995 when the system reached FOC (full operational capability). It is an effective tool for measuring geohazards and other geophysical phenomena in near real-time. However, data processing and interpretation must be handled carefully. Multipath effects on position time series can introduce spurious signatures that resemble seismic events and hamper our ability to exploit the technology. Sidereal filtering offers a very effective way to clean up GPS positioning time series to reveal more clearly geophysical events, effectively in real-time.

Over what distances do volcanoes interact?

In the geological past, large eruptions have often occurred simultaneously at nearby volcanoes. Now, a team of COMET scientists from the University of Bristol uses satellite imagery to investigate the distances over which restless magmatic plumbing systems interact.

In a study published in the journal Nature Geoscience, the scientists use deformation maps from the Kenyan Rift to monitor pressure changes in a sequence of small magma lenses beneath a single volcano. Importantly, they find that active magma systems were not disturbed beneath neighboring volcanoes less than 15 km away.

The lead author, Dr Juliet Biggs, explained: “Our satellite data shows that unrest in Kenya was restricted to an individual system. Inter-bedded ash layers at these same volcanoes, however, tell us that they have erupted synchronously in the geological past. This was our first hint to compare observations of lateral interactions based on recent geophysical measurements with those from petrological analyses of much older eruptions.

The team, which includes a recently graduated PhD student Elspeth Robertson and Bristol’s Head of Volcanology Prof. Kathy Cashman, took this opportunity to compare observations from around the world with simple scaling laws based on potential interaction mechanisms. They found that stress changes from very large eruptions could influence volcanoes over distances of up to 50 km, but that smaller pressure changes associated with unrest require a different mechanism to explain the interactions.

Prof Cashman explained ‘Volcanology is undergoing a scientific revolution right now – the concept of a large vat of liquid magma beneath a volcano is being replaced by that of a crystalline mush that contains a network of melt or gas lenses. The interactions patterns observed in Kenya support this view, and help to constrain the geometry and location of individual melt and gas lenses.”

The study was funded by two major NERC projects: COMET, a world-leading research centre focusing on tectonic and volcanic processes using Earth observation techniques; and RiftVolc, which is studying the past, present and future behavior of volcanoes in the East African Rift.

The research paper, The lateral extent of volcanic interactions during unrest and eruption, was published online in Nature Geoscience on 15th February 2016.

Sentinel-1 satellite captures volcanic surface changes that reveal the flow below

Pablo Gonzalez’s work on the 2014 Pico do Fogo eruption has been featured in the AGU’s Eos magazine.

Pico do Fogo. Credit: Nicole Richter
Pico do Fogo. Credit: Nicole Richter

The research uses a new satellite imaging system to model the subsurface path of the magma that fed the eruption, and shows that Sentinel-1’s TOPS InSAR technique has the potential to be used to study other natural hazards, including earthquakes and landslides.

Read the full article at EoS

COMET heads to Santiaguito, Guatemala

Scientists from COMET are participating in a workshop, funded by the US National Science Foundation, to study Santiaguito volcano, Guatemala.

SUAS shot of an explosion at Santiaguito volcano,  January 9 2016
SUAS shot of an explosion at Santiaguito volcano, January 9 2016

Matt Watson, Luke Western and Kate Wilkins have been in Guatemala since January 2nd, acquiring data with a range of instruments including three UV camera systems, three additional mini-UV spectrometers, a multispectral infrared camera, a lightweight infrared camera and a Phantom II Small Unmanned Aerial System (SUAS).

<a href=" site acheter”>SUAS selfie showing an array of IR and UV cameras
SUAS selfie showing an array of IR and UV cameras

The meeting, the first in a series of scientific and educational workshops to be held at an active laboratory volcano every two to three years, is lead by a selection of principal scientists who have different field-­based data collection expertise.

They, along with students and local scientists, conducted fieldwork just prior to the formal workshop.  During the main phase of the workshop additional participants, including other students and professionals, arrived and had an opportunity to both observe the field installations and participate in data collection.

Formal lectures, on both measurement techniques and recent findings on shallow conduit processes at Santiaguito, were given by the principal scientists on January 5th. Different groups then headed out for four days to make various observations and measurements of the volcano.

The focus for the following two days, the last of the workshop, was on breakout groups and hands‐on analysis of the multiple data types that were collected concurrently. Everyone participating in the workshop, including students and principal scientists, shared and received all the data products at the end of the workshop.

Following the workshop, analytical results, tools, and integrated products will be delivered to the participants and published electronically for the broader community.

Watson, Western and Wilkins, with Helen Thomas from Nicarnica Aviation, are now heading to both Pacaya and Fuego volcanoes. At Pacaya they hope to undertake a drone survey of the crater by adapting the Phantom II to fly the lightweight IR camera. At Fuego, they will acquire more imaging and spectral measurements of the volcano’s emissions and investigate installation of the multispectral infrared camera.

Kink in fault line below Nepal explains why Himalayas keep growing


An international team of scientists, led by COMET’s John Elliott, has shed new light on the earthquake that devastated Nepal in April 2015, killing more than 8,000 people.

In a study published in the journal Nature Geoscience, the scientists show that a kink in the regional fault line below Nepal explains why the highest mountains in the Himalayas are seen to grow between earthquakes.


Dr Elliott explained: “We have shown that the fault beneath Nepal has a kink in it, creating a ramp 20km underground. Material is continually being pushed up this ramp, which explains why the mountains were seen to be growing in the decades before the earthquake. The earthquake itself then reversed this, dropping the mountains back down again when the pressure was released as the crust suddenly snapped in April 2015.

“Using the latest satellite technology, we have been able to precisely measure the land height changes across the entire eastern half of Nepal. The highest peaks dropped by up to 60cm in the first seconds of the earthquake.”

Mount Everest, at more than 50km east of the earthquake zone, was too far away to be affected by the subsidence in this event.

The team, which included academics from the USA and France, also demonstrate that the rupture on the fault stopped 11km below Kathmandu. This leaves an upper portion that remains unbroken and will build up more pressure over time as India continues to collide with Nepal, and indicates that another major earthquake could take place within a shorter timeframe than the centuries that might be expected for the area.

Dr Elliott added: “As this part of the fault is nearer the surface, the future rupture of this upper portion has the potential for a much greater impact on Kathmandu if it were to break in one go in a similar sized event to that of April 2015.

“Work on other earthquakes has suggested that when a rupture stops like this, it can be years or decades before it resumes, rather than the centuries that might usually be expected.”

Study co-author Dr Pablo González, from the School of Earth and Environment at the University of Leeds and also a COMET member, said: “We successfully mapped the earthquake motion using satellite technology on a very difficult mountainous terrain. We developed newly processing algorithms to obtain clearer displacement maps, which revealed the most likely fault geometry at depth to make sense of the puzzling geological observations.”

The research paper, Himalayan megathrust geometry and relation to topography revealed by the Gorkha earthquake, was published online in Nature Geoscience on 11 January 2016.


Tracking the Etna eruption

On the evening of December 2 2015, Sicily’s Mount Etna began to erupt for the first time in over two years, reaching a brief but violent climax in the early hours of December 3 which included lava fountains as well as a column of gas and ash several kilometres high. The event was among the most violent seen at Etna over the last twenty years.

Ash cloud from Mount Etna’s Voragine crater lights up the sky. Credit: Marco Restivo/Demotix/Corbis
Ash cloud from Mount Etna’s Voragine crater lights up the sky. Credit: Marco Restivo/Demotix/Corbis

Luckily, good weather meant that the eruption could be monitored with visual and thermal cameras from the Istituto Nazionale di Geofisica e Vulcanologia (INGV) Etna Observatory.  According to INGV reports, activity peaked between 02:20 and 03:10 GMT when a continuous lava fountain reached heights well above 1km; with some jets of volcanic material reaching 3km into the sky.  Although the eruption had more or less ceased by dawn, the volcanic cloud had blown northeast, causing ash to be deposited on the nearby towns of Taormina, Milazzo, Messina and Reggio Calabria.

The eruption has so far continued, repeating the behaviour seen earlier with tall lava fountains and eruption columns many kilometers high.  Updates can be found on the INGV webpage.

COMET scientists at the University of Oxford have been tracking the volcanic plume’s progress using data from the Infrared Atmospheric Sounding Instruments (IASI) on board ESA’s MetOp-A and MetOp-B satellite platforms.  These instruments can detect the presence of volcanic SO2 in the atmosphere, using methods developed by the University’s Earth Observation Data Group.

The results, which can be found on the IASI NRT web page, showed that by Friday 4 December the plume had reached an area between Crete and Iraq, containing 0.06 Tg (1012g) SO2.

Estimate of SO2 amount from IASI-A overpass on the morning of 3 and 4 December 2015, assuming the SO2 between 9 and 10 km altitude
Estimate of SO2 amount from IASI-A overpass on the morning of 3 and 4 December 2015, assuming the SO2 between 9 and 10 km altitude

By the morning of 7 December, the plume had travelled from Sicily to Asia, reaching as far as Japan and the Pacific Ocean.

Screenshot from IASI NRT webpage 7 December 2015
Screenshot from IASI NRT webpage 7 December 2015

Dr Elisa Carboni, a COMET researcher based at the University of Oxford, said: “This is a great example of how we can track volcanic plume using the near real time IASI service. ”

You can follow the volcanic plume on the IASI NRT web page.



Volcanoes of the Ethiopian Rift Valley

Are the volcanoes of the Ethiopian Rift Valley now peaceful, or do they continue to pose a threat to the tens of millions of people who live and work the land across this vast region?

View into the Main Ethiopian Rift Valley, on the descent from Butajira to Ziway. Aluto volcano in the centre distance.
View into the Main Ethiopian Rift Valley, on the descent from Butajira to Ziway. Aluto volcano in the centre distance.

Read COMET scientist David Pyle’s blog post on the work of the NERC-funded RiftVolc consortium, which is carrying out a broad-scale investigation of the past eruptive histories, present status and potential for future activity of the volcanoes of the Central Main Ethiopian Rift.