February 2017 Issue Table of Contents
Currently, it is only possible to look inside an active volcano using indirect geophysical methods. One such method is to measure surface deformation, which results from subsurface magmatic or hydrothermal processes. Modern satellite data allows deformation to be measured at hundreds of volcanoes without relying on limited ground instrumentation. As a result, the number of known deforming volcanoes has increased from 44 in 1997 to over 220 in 2016. This article reviews the diverse ways by which volcanoes can deform, the typical rates and durations of such deformations, and the processes that drive deformation.
Keywords: volcano deformation, eruption, geodesy, InSAR, GPS
Whereas geological mapping allows scientists to see the inner plumbing of old volcanoes that have been exhumed by erosion, it is geophysical methods that allow us to study the internal structure of active volcanoes and the processes that may ultimately lead to eruption. Geophysical techniques make measurements near the surface to image the structure beneath a volcano, recording the stress and strain associated with magmatic processes (e.g. Pritchard and Gregg 2016). A common limitation is that multiple sensors must be placed around the volcano in order to make detailed inferences about the subsurface, and the majority of the world’s 1,400 subaerial volcanoes do not have such networks (e.g. Loughlin et al. 2015). Observations of surface deformation (‘geodesy’) can now be made using satellites and can be made globally (e.g. Fournier et al. 2010; Biggs et al. 2014). The number of known deforming volcanoes has quintupled over the last 20 years as traditional ground-based survey methods have been complemented with satellite technology. The ability to observe how different types of volcanoes deform offers a unique perspective on the behaviour of magmatic systems throughout the eruption cycle.
Despite the explosion of observations, our understanding of the significance of deformation (or the lack of deformation) at volcanoes is still in its infancy. The challenge for 21st century volcanologists is to link the new observations of surface deformation to volcanic processes, within the framework of other geological and geophysical observations of magmatic systems. In the coming decades, much more will be learned about complex magmatic systems, from the generation of melt to its interaction with the surface, atmosphere – or shallow crust – and there will be improved monitoring and forecasting of volcanic hazards.
This article will first review the geodetic methods that allow us to study subsurface processes in active volcanoes, and then summarize the broad range of volcanic behaviour that has been observed geodetically.
In 1997, deformation had been reported at 44 volcanoes (Dvorak and Dzurisin 1997); by 2010, there were 118 (Fournier et al. 2010) and; at the writing of this article, there are over 220, which we document in the supplementary table. This rapid increase is not the result of a rise in volcanic activity but is a consequence of improved observation and reporting, particularly in the developing world. Technological advances have been vital: a fleet of international satellites (10 radar satellites in 2017 with more planned; e.g. Pinel et al. 2014) make routine, global observations, and global positioning system (GPS) networks make ground-based geodetic observations routine and affordable. However, the list of deforming volcanoes is still incomplete because some volcanoes have never been studied or have only been studied incompletely due to inadequate data.
This review focuses on interferometric synthetic aperture radar (InSAR), the main satellite-based tool that is used to measure surface deformation at volcanoes. Radar (a name itself that derives from ‘radio detection and ranging’) involves the transmission and reception of microwave electromagnetic radiation (roughly 108–1011 Hz, or wavelengths of 1 mm to 1 m) (e.g. Pinel et al. 2014). At these wavelengths, radar systems can see through clouds and most types of precipitation and do not rely on the sun’s illumination, giving them a unique all-weather, day–night capability. Each pixel in a synthetic aperture radar (SAR) image is represented by a complex number, with the amplitude corresponding to the intensity of the returned radar energy and the phase equalling a fraction of the complete wavelength (having a value between 0 and 2π). When the phases from two images are combined to form an interferogram, the phase difference reveals variations in the distance between the ground and the satellite that appear as coloured fringes (Fig. 1). Each fringe corresponds to half the radar wavelength. The rate of ground displacement can be measured by taking the observed change in distance divided by the time interval between SAR images – this interval usually varies between a day and a few weeks, with rates of deformation ranging from mm/y to several m per day (Fig. 2). There are two notable limitations to the InSAR technique. First, decorrelation, which occurs when characteristics of the ground and its ability to reflect radar waves change rapidly, for example in heavily vegetated or agricultural regions. Second, atmospheric delays caused by water vapour in the troposphere. Time-series methods that can combine hundreds or thousands of images are increasingly important in overcoming these limitations (e.g. Dzurisin et al. 2006; Pinel et al. 2014).
The revolution in volcano deformation studies as a result of InSAR comes from the ability to routinely image deformation at nearly all of the world’s volcanoes with an accuracy of a few millimetres to a few centimetres. Even so, volcanoes can deform very rapidly (e.g. Fig. 2) and the repeat time between satellite overpasses may be too long to capture these temporal changes, particularly as the amount of data available for a given volcano varies widely. In some areas, images are collected during every overpass, while in others, data may never have been acquired by certain satellites. Furthermore, data from some satellites are available at no cost (such as the European Space Agency’s Sentinel mission), whereas imagery from other satellites can cost thousands of dollars. One way to overcome this limitation is to use all available satellites in the international constellation. Although the data from the different missions cannot be directly combined, using more satellites increases the frequency of ground observations and, thus, the ability to detect transient events.
Ground sensors complement InSAR observations and can be used to evaluate or correct uncertainties. The most commonly applied is the continuous global navigation satellite system (GNSS), which uses signals from the GPS and other navigation satellites to measure 3-D changes in co-ordinates. Rather than relying on the transmitted code used by cell phones and other domestic systems, geodetic GNSS receivers use the phase of the transmitted signal to achieve sub-millimetre accuracy (Dzurisin 2006). Other ground-based systems (tiltmeters, strainmeters, ground-based radar, levelling and triangulation surveys) (Dzurisin 2006) are useful where available, but their use is limited to a handful of volcanoes, and it is not possible to tell how widespread the observed processes are. Ground observations that are collected continuously are especially valuable because they overcome the gaps between satellite InSAR measurements.
Diversity of deforming volcanoes
A variety of processes can cause ground deformation at volcanoes: magma movements, landslides, faults, hydrothermal systems, and thermal or thermodynamic volume changes from heating, cooling, melting or crystallization (e.g. Dzurisin et al. 2006). Nevertheless, certain patterns of ground deformation can, to a limited extent, be diagnostic of specific physical process, e.g. a dyke intrusion, cooling lava/pyroclastic flow, or pressurizing ‘magma reservoir’, all of which have distinct patterns of ground displacement viewed with InSAR (Fig. 1A-E). Complexities arise depending on the direction the satellite is looking relative to the ground displacement (e.g. Dzurisin et al. 2006) and when multiple processes occur at nearly the same time (Figs. 1F, 1G). In this section, we describe some commonalities between deforming volcanoes and global patterns in the types of processes that produce ground deformation.
The Classic Volcano Deformation Cycle
Prior to an eruption, according to the classic model of the ‘volcano deformation cycle’, magma gradually inflates a magma chamber directly beneath the volcanic edifice until a threshold is reached at which point the chamber ruptures and an eruption rapidly empties and deflates the chamber (e.g. Dzurisin et al. 2006). The inflation phase causes uplift of the ground surface and large numbers of small earthquakes (volcano-tectonic seismicity), while eruption is accompanied by rapid subsidence (Fig. 3A). To a first approximation, this pattern of co-eruptive subsidence and inter-eruptive uplift has been observed at a number of volcanoes with different characteristic length scales and timescales and has been used to provide eruption forecasting in a number of cases (e.g. references in Dzurisin et al. 2006; Fournier et al. 2010). However, the recent explosion in volcano monitoring data has demonstrated that the classic model is an oversimplification. Many volcanoes just do not behave in this way. Furthermore, a simple, large, liquid-filled magma chamber is not compatible with modern geological or petrological observations. Nonetheless, the classic model of the volcano deformation cycle remains a useful concept from which to start.
Long-lived eruptions and intrusions provide an alternative opportunity to observe repeated cycles of behaviour. For example, the multi-decadal eruptions of Soufrière Hills Volcano (SHV) on Montserrat (an andesitic stratovolcano in the Caribbean) and Kilauea (Hawai’i USA) (a basaltic shield volcano) have served as test-beds for new ideas and new monitoring systems. Global positioning system instruments at SHV observed a simple first-order pattern: when lava is erupting rapidly, the surface subsides; during periods of no extrusion, it inflates (Wadge et al. 2014). Similar patterns have been observed at volcanoes with multiple, distinct eruptions, such as Fernandina in the Galapagos (e.g. Pinel et al. 2014) and Okmok (Alaska, USA) (e.g. Lu and Dzurisin 2014). However, close inspection of the time series shows that inter-eruptive uplift is interrupted by short reversals that are not associated with any magma output at the surface. At Okmok, these reversals have been attributed to pulses of gas loss or re-absorption (e.g. Caricchi et al. 2014).
Most magma does not reach the surface – but sometimes, the intrusion of dykes and sills and the growth of plutons are visible in the geodetic, as well as the geological, record. Earthquake swarms at Eyjfallajökull (Iceland) in 1994 and 1999 were associated with tens of centimetres of surface uplift but no eruption (Sigmundsson et al. 2010 and references therein). Similar patterns of seismicity and deformation in 2010 culminated in the eruption that disrupted Europe’s air traffic, causing huge economic losses. Magma is supplied in batches, some of which ‘stall’ and form an intrusive complex at the roots of the volcano, while others, as happened in the 2010 eruption, reach the surface, triggering large eruptions which can tap multiple reservoirs (Sigmundsson et al. 2010). Intrusions also occur within volcanic edifices. The orientation of these intrusions is controlled by the stress field within the edifice, and, if they reach the surface, they can determine the geometry of the resulting eruptive vents/fissures (e.g. Bagnardi et al. 2013).
The deformation pattern associated with a dyke intrusion is usually distinctive, having two lobes of lateral displacement on either side of a zone of subsidence (Fig. 1B). The lobes may appear asymmetric in the satellite line of sight, but the pattern is clearly different from the circular or elongated bulls-eye patterns of uplift typical of a sill (Fig. 1A). Some dykes stall before reaching the surface, while others feed eruptions. One recent example of this was the eruption of 1.4 km3 of lava at Holuhraun (Iceland), which was fed by a dyke originating at Bárðarbunga caldera 45 km away (Sigmundsson et al. 2015). Another example was the intrusion of an 80 km long dyke in Afar (Ethiopia), which intersected two small silicic centres, triggering minor eruptions and a number of shorter dykes that failed to reach the surface (Pinel et al. 2014 and references therein).
Continental growth also occurs by the formation of plutons and associated ductile deformation, which are major geological features but are harder to observe geodetically because they are thought to grow slowly over millions of years. Possible candidates for these plutonic processes, which could be on the order of mm/y and span 10–100 km, include Uturuncu volcano in the Central Andes of Bolivia, coincident with the mid-crustal Altiplano–Puna magma or mush body (e.g. Pritchard and Gregg 2016), and the uplift that has lasted for more than 100 years at Socorro (New Mexico, USA) (Pearse and Fialko 2010).
Volcanic conduits are thought to be extremely narrow (<10 m radius) and are, therefore, difficult to observe geophysically. Nonetheless, clues to their behaviour are given by sub-daily deformation cycles that are typically observed in association with Strombolian or Vulcanian eruptions. Tiltmeters installed on the crater rim at Soufriere Hills volcano (Montserrat) in 1996–1997 detected deformation patterns correlated with seismicity and explosions (Voight et al. 1999). The debate continues as to whether these observations should be interpreted as pressurization of shallow reservoirs, shear along the conduit wall, or the passage of volatiles (e.g. Nishimura 2009).
Lava lakes are essentially magma-filled conduits open at the surface, and, in a simple conceptual model, changes in the reservoir pressure would result in changes in lake level rather than surface deformation. Many of the world’s lava lakes happen to be in remote or inaccessible locations, but the best studied is undoubtedly Halemaumau (Hawai’i), where Patrick et al. (2015) showed that deformation and lake levels occur simultaneously – evidence for an open conduit linking the lava lake to the magma reservoir.
Caldera systems have long repose periods between very large eruptions, but do not remain quiescent – many calderas have frequent, small eruptions, known as resurgent volcanism. In the case of Santorini (Greece) the volume of erupted lava is directly proportional to the time since the previous eruption, evidence that magma supply from depth is continuous (Parks et al. 2012). However, in 2010, a short period of rapid uplift occurred with an equivalent volume to that anticipated for the next eruption, indicating that the magma supply to the shallow reservoir is pulsed rather than continuous (Parks et al. 2012).
Some caldera systems have been known to deform for decades without erupting. At Campi Flegrei (Italy) gradual subsidence over centuries has caused the Roman market at Pozzuoli to become submerged. The subsidence is occasionally interrupted by pulses of uplift but it remains unclear whether the cause is magmatic or hydrothermal (e.g. Chiodini et al. 2010). At Yellowstone (USA), deformation has a spatially and temporally variable pattern and is attributed to hydrothermal fluids moving between reservoirs as well as magma intrusion (e.g. Chang et al. 2007).
More recently, satellite observations have shown that deformation is occurring at many caldera systems that are not erupting, that have no record of historical volcanism, and that have no real-time monitoring. This includes many volcanoes along the densely populated East African Rift (e.g. Biggs et al. 2014 and references therein) and raises the question of how to interpret caldera deformation in terms of volcanic hazards. The fastest uplift rate seems to be at Laguna del Maule (Chile), which has been uplifting at a rate of 280 mm/y since 2007 (e.g. Fournier et al. 2010). However, the implications for the shallow magma body underneath remain unclear.
Is the behaviour of restless calderas related to external influences? The passage of seismic waves associated with large earthquakes has been shown to influence hydrothermal systems, probably as a result of interactions with gas bubbles. For example, both the 2010 Maule (Chile) and 2011 Tohoku (Japan) earthquakes caused subsidence at several nearby volcanoes, but the subsidence mechanism could be due either to changes in the hydrothermal system or to stress changes acting on the weaker rheological properties of a pluton (e.g. Pritchard et al. 2013).
During and After an Eruption
While co-eruptive subsidence associated with subsurface magma withdrawal is a common feature of eruptions, in many cases it is masked by local surface changes, including both the destruction and creation of topography. Explosive eruptions typically form new craters and vents, they can alter the morphology of existing structures, and they may trigger large collapses. Eruptive products include lava flows, pyroclastic flows and lahars and these can fill low topography. A new dome may grow near the vent (e.g. Pinel et al. 2014). Ash fall, while rarely thick enough to dramatically alter topography, can alter the appearance of the ground surface making it incoherent to satellite radar and so reducing the efficiency of ground-based monitoring that relies on solar panels. After an eruption, the new topography is typically oversteep, and processes such as landsliding and gravitational spreading (e.g. Ebmeier et al. 2014) act to restabilise the landscape. Subsidence of cooling and compacting lava flows can last for decades: Parícutin (Mexico) erupted in 1952 and is still subsiding today (e.g. Fournier et al. 2010).
Magmatic systems can respond to eruption in multiple ways. In the classic model of the volcano deformation cycle, the magma chamber begins to refill within days of the eruption [e.g. Okmok volcano, described in Lu and Dzurisin (2014)]. But sometimes the volcano just continues to subside [e.g. Kiska volcano in the Aleutian Islands of Alaska, described in Lu and Dzurisin (2014)]. For decades after an intrusion, the subsurface magma body will continue to cool, crystallize and degas: at Medicine Lake Volcano (California, USA), modern geodetic observations were combined with levelling surveys from the 1950s to demonstrate that subsidence has continued at ~10 mm/y for at least the last 65 years (Parker et al. 2014).
Limits of Detectability
Several eruptions appear to have taken place at volcanoes with no known deformation (e.g. Fournier et al. 2010; Lu and Dzurisin 2014). Even when co-eruptive subsidence is observed, the volume is usually less than the dense rock equivalent of the erupted products. The classic model of the volcano deformation cycle assumes that there is a constant volume flux of magma being supplied to the magma chamber, producing a linear rate of uplift between eruptions. Altering the boundary conditions such that magma is supplied from a deeper chamber at constant pressure modifies this model such that eruption is followed by exponentially decaying uplift. This may explain observations of eruptions that occur with negligible uplift prior to eruption and very rapid uplift following eruption, such as was observed at Westdahl (Alaska) (e.g. Lu and Dzurisin 2014).
The simplest explanation for a lack of co-eruptive subsidence is that deformation is occurring at a magnitude or resolution beneath our current ability to observe: too fast or too slow, too shallow or too deep, or obscured by atmospheric effects. Already, new satellites with higher resolution and faster orbital repeats (e.g. TerraSAR-X and CosmoSkyMed) have been used to observe previously undetectable processes (e.g. Salzer et al. 2014). Long-term monitoring missions, such as the European Space Agency’s Sentinels programme, have the potential to revolutionise detection capabilities. An alternative explanation is that we ought to consider mass balance, rather than volume balance, because density is far from constant in a three-phase magmatic system undergoing changes in temperature and pressure. Gas bubbles, after all, do make magmas very compressible (e.g. Caricchi et al. 2014).
The rapid increase in the number of geodetically studied volcanoes means that it is now feasible to treat the observations statistically. Biggs et al. (2014) showed that there is a strong link between deformation and eruption for over 500 systematically studied volcanoes, and also between non-deformation and non-eruption. Further, this relationship varies with volcanic parameters: shield volcanoes have a strong link between deformation and eruption and come closest to the classic volcano deformation cycle, while calderas frequently deform without erupting, suggesting that large volumes of magma are stored in the upper crust and that the deformation is caused by gas or hydrothermal fluids. Stratovolcanoes are the most likely to erupt without observable deformation, perhaps because the mass changes are accommodated in a way not detectable by current observation satellite systems, which have revisit times of weeks.
How do we know when a given deformation event may lead to eruption? One way to answer this question is to compare the duration of deformation events to their magnitude and see whether they led to an eruption (Fig. 2). As expected, deformation events that had large magnitudes usually lasted for a short amount of time and typically led to an eruption (e.g. Fournier et al. 2010). Yet there doesn’t seem to be a simple threshold at which deformation duration and/or magnitude should cause concern. On the other hand, there are many volcanoes that can deform at rates of 1–1,000 mm/y without causing eruption in the short term, and so these types of deformation events are not always hazardous. Figure 2 and the online supplementary table with this article are incomplete in several respects. Some volcanoes have never been studied. At others, the temporal sampling is inadequate to resolve pulses of rapid deformation, the deforming area being too small to resolve or had occurred within a data gap. In some cases, deformation was ongoing at the start or end of the available observations, making the specified duration a minimum estimate. There is a cluster of deformation events with a one day duration that likely correspond to even shorter time periods, but these short events could not be constrained because observations were not sufficiently frequent.
Although deformation has been reported at over 200 volcanoes, there remain several eruption styles that have been observed rarely, or not at all. The only geodetically observed rhyolite eruption occurred at Chaitén (Chile) in 2008, with only a few hours of pre-eruptive warning (e.g. Wicks et al. 2011). With a sample size of one, it is impossible to state whether the pattern observed is representative. The latest monogenetic eruption occurred at Parícutin during the period 1943–1953, which was before routine satellite observations. Perhaps most alarming is the lack of observation of the very largest eruptions. The most recent magnitude 7 eruption (i.e. erupting a volume in dense rock equivalent of 1011–1012 m3) occurred at Tambora (Indonesia) in 1815. Even recent magnitude 5–6 eruptions, with erupted volumes of 109–1011 m3 [such as Mount St. Helens (Washington, USA), and Mount Pinatubo (Philippines)] have limited geodetic observations. The only available example is the relatively small eruption of Campi Flegrei (Italy) in 1538, where historical records suggest that several metres of deformation occurred in the years before the eruption (Guidoboni and Ciuccarelli 2011). Calderas have the potential for extremely large eruptions, but could we distinguish between the semi-continuous unrest seen at so many calderas and the precursors to a major eruption?
Summary and Perspectives
Satellite techniques are rapidly improving our ability to monitor volcanoes on a global basis and have provided insight into the spatial and temporal changes in the subsurface stress fields around volcanoes and intrusions. The classic model of the volcano deformation cycle – co-eruptive deflation and inter-eruptive inflation – is seen at many volcanoes, but the rapid increase in geodetic monitoring has demonstrated that such a simple model is not always applicable. Deformation can be attributed to the movement and phase transitions of magma, volatiles and hydrothermal fluids and to the intrusions of dykes, sills and the growth of plutons, each of which have their own characteristic geodetic signatures. Surface processes during and after eruptions, such as the formation of volcanic flows and domes as well as edifice instability, may obscure subsurface processes. Alongside developments in observational ability, modelling capabilities have evolved from analytic solutions for point sources and other simplified geometries, to finite element models that can incorporate more complex rheological and structural information, to physics-based models capable of integrating geophysical monitoring with observations of degassing and petrology (e.g. Masterlark 2007; Anderson and Segall 2011).
Satellite imagery has proved a remarkable reconnaissance tool to discover new phenomena. As new satellites and constellations of satellites are launched in the coming years (e.g. Pinel et al. 2014), even more new discoveries will be made. Yet, the biggest questions remain: “Which of the deforming volcanoes are a threat?”, and, “Under what circumstances does deformation lead to eruption?”
We acknowledge the NERC-BGS Centre for the Observation and Modelling of Earthquakes, Volcanoes and Tectonics, a Benjamin Meaker Visiting Professorship to M.E.P from the Institute for Advanced Study, University of Bristol, and NASA grants NNX12AO31G and NNX12AM24G issued through the Science Mission Directorate’s Earth Science Division. We thank David Arnold, Susanna Ebmeier, Maria Furtney, Scott Henderson, James Hickey, Jennifer Jay, Amy Parker, Mike Poland and Tim Wright for contributing to the supplemental material and figures, and Jeff Freymueller, Estelle Chaussard, Keith Putirka, Francisco Delgado, and an anonymous reviewer for comments.
Anderson K, Segall P (2011) Physics-based models of ground deformation and extrusion rate at effusively erupting volcanoes. Journal of Geophysical Research: Solid Earth 116, doi: 10.1029/2010JB007939
Bagnardi M, Amelung F, Poland MP (2013) A new model for the growth of basaltic shields based on deformation of Fernandina volcano, Galápagos Islands. Earth and Planetary Science Letters 377-378: 358-366
Biggs J and 6 coauthors (2014) Global link between deformation and volcanic eruption quantified by satellite imagery. Nature Communications 5, doi: 10.1038/ncomms4471
Caricchi L, Biggs J, Annen C, Ebmeier S (2014) The influence of cooling, crystallisation and re-melting on the interpretation of geodetic signals in volcanic systems. Earth and Planetary Science Letters 388: 166-174
Chang W-L, Smith RB, Wicks C, Farrell JM, Puskas CM (2007) Accelerated uplift and magmatic intrusion of the Yellowstone caldera, 2004 to 2006. Science 318: 952-956
Chiodini G and 7 coauthors (2010) Long-term variations of the Campi Flegrei, Italy, volcanic system as revealed by the monitoring of hydrothermal activity. Journal of Geophysical Research: Solid Earth 115, doi: 10.1029/2008JB006258
Dvorak JJ, Dzurisin D (1997) Volcano geodesy: The search for magma reservoirs and the formation of eruptive vents. Reviews of Geophysics 35: 343-384
Dzurisin D (2006) Volcano Deformation: Geodetic Monitoring Techniques. Springer Science Business Media, 442 pp
Dzurisin D, Lisowski M, Wicks CW, Poland MP, Endo ET (2006) Geodetic observations and modeling of magmatic inflation at the Three Sisters volcanic center, central Oregon Cascade Range, USA. Journal of Volcanology and Geothermal Research 150: 35-54
Ebmeier SK, Biggs J, Muller C, Avard G (2014) Thin-skinned mass-wasting responsible for widespread deformation at Arenal volcano. Frontiers in Earth Science 2, doi: 10.3389/feart.2014.00035
Fournier TJ, Pritchard ME, Riddick SN (2010) Duration, magnitude, and frequency of subaerial volcano deformation events: New results from Latin America using InSAR and a global synthesis. Geochemistry, Geophysics, Geosystems 11: doi: 10.1029/2009GC002558
Guidoboni E, Ciuccarelli C (2011) The Campi Flegrei caldera: historical revision and new data on seismic crises, bradyseisms, the Monte Nuovo eruption and ensuing earthquakes (twelfth century 1582 AD). Bulletin of Volcanology 73: 655-677
Hamling IJ and 8 coauthors (2009) Geodetic observations of the ongoing Dabbahu rifting episode: New dyke intrusions in 2006 and 2007. Geophysical Journal International 178: 989-1003
Loughlin SC, Sparks S, Brown SK, Jenkins SF, Vye-Brown C (eds) (2015) Global Volcanic Hazards and Risk. Cambridge University Press, 408 pp
Lu Z, Dzurisin D (2014) InSAR Imaging of Aleutian Volcanoes: Monitoring A Volcanic Arc From Space. Springer-Praxis, Berlin Heidelberg, 390 pp
Masterlark T (2007) Magma intrusion and deformation predictions: sensitivities to the Mogi assumptions. Journal of Geophysical Research: Solid Earth 112, doi: 10.1029/2006JB004860
Nishimura T (2009) Ground deformation caused by magma ascent in an open conduit. Journal of Volcanology and Geothermal Research 187: 178-192
Parker AL, Biggs J, Lu Z (2014) Investigating long-term subsidence at Medicine Lake Volcano, CA, using multitemporal InSAR. Geophysical Journal International 199: 844-859
Parks MM and 11 coauthors (2012) Evolution of Santorini Volcano dominated by episodic and rapid fluxes of melt from depth. Nature Geoscience 5: 749-754
Patrick MR, Anderson KR, Poland MP, Orr TR, Swanson DA (2015) Lava lake level as a gauge of magma reservoir pressure and eruptive hazard. Geology 43: 831-834
Pearse J, Fialko Y (2010) Mechanics of active magmatic intraplating in the Rio Grande Rift near Socorro, New Mexico. Journal of Geophysical Research: Solid Earth 115, doi: 10.1029/2009JB006592
Pinel V, Poland MP, Hooper A (2014) Volcanology: Lessons learned from synthetic aperture radar imagery. Journal of Volcanology and Geothermal Research 289: 81-113
Pritchard ME, Gregg PM (2016) Geophysical evidence for silicic crustal melt in the continents: where, what kind, and how much? Elements 12:121-127
Pritchard ME, Jay JA, Aron F, Henderson ST, Lara LE (2013) Subsidence at southern Andes volcanoes induced by the 2010 Maule, Chile earthquake. Nature Geoscience 6: 632-636
Salzer JT and 6 coauthors (2014) Satellite radar data reveal short-term pre-explosive displacements and a complex conduit system at Volcán de Colima, Mexico. Frontiers in Earth Science 2, doi: 10.3389/feart.2014.00012
Sigmundsson F and 36 coauthors (2015) Segmented lateral dyke growth in a rifting event at Bárðarbunga volcanic system, Iceland. Nature 517: 191-195
Sigmundsson F and 15 coauthors (2010) Intrusion triggering of the 2010 Eyjafjallajökull explosive eruption. Nature 468: 426-430
Voight B and 24 coauthors (1999) Magma flow instability and cyclic activity at Soufriere Hills volcano, Montserrat, British West Indies. Science 283: 1138-1142
Wadge G and 5 coauthors (2014) An overview of the eruption of Soufriere Hills Volcano, Montserrat from 2000 to 2010. In: Wadge G, Robertson REA, Voight B (eds) The Eruption of Soufriére Hills Volcano, Montserrat from 2000 to 2010. Geological Society, London, Memoir 39: pp 1-40
Wicks C, de la Llera JC, Lara LE, Lowenstern J (2011) The role of dyking and fault control in the rapid onset of eruption at Chaitén volcano, Chile. Nature 478: 374-377