February 2017 Issue Table of Contents
Magma systems that supply volcanoes can extend throughout the crust and consist of mush (melt within a crystalline framework) together with ephemeral magma accumulations. Within a crystal-rich mush, slow processes of melt segregation and heat loss alternate with fast processes of destablisation and magma transport. Magma chambers form by two mechanisms: incremental magma intrusion into sub-solidus rocks or the segregation and rapid merging of melt-rich layers within mush regions. Three volcanic states reflect alternations of slow and fast processes: dormancy, unrest and eruption. Monitoring needs to detect processes of melt and fluid movements in the lower and middle crust during destabilisation to improve forecasting.
Keywords: volcano, magma chamber, monitoring, forecasting, volcanic eruptions
A major societal application of volcanology is giving an early warning of volcanic activity to reduce risks to communities. Forecasting is enabled by monitoring geophysical signals and volcanic emissions, and by observing volcanic phenomena. Forecasts are informed by historical records, geological studies and empirical knowledge about past patterns of activity. Underpinning these endeavours are ideas about the igneous processes that control volcanism, which are often viewed through the concept of an upper crustal magma chamber. Ideas are changing, however, about how the magmatic systems beneath volcanoes work. Here, we summarise emerging concepts and explore what they mean for monitoring and forecasting.
Observations Of Magma Systems Beneath Active Volcanoes
The concept of an upper crustal (typically 3–10 km) magma chamber explains many phenomena associated with volcanic unrest and eruptions, such as shallow earthquakes, ground deformation and gas emissions. These three phenomena can be attributed to the internal processes within a magma chamber, to the movement of magma and magmatic fluids (supercritical exsolved volatiles) from the chamber to the surface, and to the transfer of heat and magmatic fluids to hydrothermal systems (Fig. 1). Many petrological and geochemical characteristics of volcanic products can also be explained by processes within a magma chamber that as itself at least transiently connected to a magma supply at depth. For example, replenishment by new magmas is widely postulated to trigger eruptions and explain various petrological characteristics of volcanic materials.
Increasingly, however, more than one magma chamber is invoked to explain geochemical and petrological observations of magmatic systems. The replenishment model implies a deeper magma source, while geochemical, petrological, and geophysical observations suggest that eruptions can tap multiple magma bodies that might be either vertically or laterally distributed. As examples, the eruption of the Soufrière Hills volcano (Montserrat) tapped four different magma bodies extending from 5 km to at least 15 km depth (Christopher et al. 2014); eruptions of Mount St Helens (Washington, USA) involved a magmatic system extending from about 5 km to at least 30 km depth (Blundy et al. 2008); eruptions from the Taupo Volcanic Zone (New Zealand), tapped at least two different magma bodies from vents 25–30 km apart (Cooper et al. 2012); and eruptions of Eyjafjallajökull (Iceland), tapped melt bodies throughout the crust (Tarasewicz et al. 2012). Geobarometry on phenocrysts indicate magma depths of origin extending into the middle and lower crust (e.g. Putirka 2017 this issue). These findings lead to the concept of a transcrustal magmatic system in which the shallow magma chamber may be only a volumetrically minor part of the entire volcanic plumbing and volcanic behaviour is governed by the whole system.
“Magma” is here defined specifically as “eruptible melt with or without suspended crystals and volatiles”. Magmatic mush has enough crystals to form a touching framework, and is synonymous with a partially molten (super-solidus) rock. The transition from magma to mush occurs over a narrow crystal content (typically between 55% and 65%), with many orders of magnitude increase in viscosity (Fig. 2). Magma chambers are regions of magma embedded within mush and, commonly, fully solidified plutons. Magma chambers can supply volcanic eruptions. A transcrustal magmatic system refers to the entire network of magma and mush and sub-solidus plutons embedded in older crustal rocks: it is the integrated result of magmatism in a given location. Within an active magmatic system, there can be multiple regions of mush and magma at different levels. The term “magma reservoir” was originally synonymous with the term “magma chamber”, but is increasingly used to describe the entire transcrustal magmatic system. Bachmann and Huber (2016) review these notions in detail for silicic systems and provide their historical roots.
Seismic tomography supports the concept that magmatic systems extend throughout the crust below many volcanoes (e.g. Huang et al. 2015). Importantly, these studies fail to find large coherent regions of magma, and seismic velocity anomalies are commonly interpreted as near-solidus mush (<10% melt). This suggests that either large magma bodies are not present or are too small to be resolved seismically, except where discrete melt-rich lenses are recognised from strong seismic reflections (e.g. Sinton and Detrick 1992).
Mush-dominated sub-volcanic systems are also indicated by geochemical and geochronological analysis of igneous rocks and their constituent minerals, which add key information on timescales. A first-order observation relates to the complex growth histories and diverse crystal origins of minerals. Importantly, widespread disequilibrium textures and diverse crystal compositions record mixing of different magmas, disruption of mush, entrainment of refractory plutonic materials, and xenocrysts incorporated from surrounding country rock (Davidson et al. 2007). Trace element zoning patterns in crystals commonly indicate short magma residence times (years to centuries) in upper crustal chambers (Morgan and Costa 2010). In contrast, U-series studies of zircon indicate very long residence times (>103–105 years) for fractionation and storage of melts (Hawkesworth et al. 2004). Reconciling these timescales requires long residence of fractionated melts (slow processes) and rapid crystallization when magmas are transported and assembled in the upper crust (fast processes). From this perspective, the long times represent the timescales of chemical evolution, while the short times record rapid physical rearrangement of magmatic systems that can lead to volcanic eruptions.
Transcrustal Magma plumbing: Dynamical Considerations
We now discuss the dynamic processes that form transcrustal magma systems. We define the base of the crust as the petrological Mohorovicic discontinuity (Moho), or true boundary between the crust and mantle. This is distinct from the seismically defined Moho, which may lie above crust-derived ultramafic cumulate rocks. The crust acts as a physical barrier to magma ascent, and primitive basaltic magmas rarely erupt at the Earth’s surface. Instead, most magmas stall and cool, de-gas, crystallize and evolve to lower density and chemically evolved melts. Heat transfer with, and volatile (notably water) fluxes into, surrounding crust can cause partial melting and produce magma from older rocks or remobilise earlier intrusions (Bachmann and Huber 2016). Within transcrustal magma systems, we distinguish three types of igneous processes: first, vertical (upward) transport of magma; second, stalling in the crust (intrusion); third, internal (super-solidus) magma and mush processes.
Vertical magma transport can occur by fast or slow mechanisms. Flow of magma and magmatic fluids along fractures enables fast transport (Rubin 1995). The transporting fracture system is preserved in solidified rocks as dykes (magma) or veins (fluids) in brittle cold crust and is manifested in active volcanic systems as earthquakes, tremor and deformation. In more ductile environments, especially mush systems, magma and fluids likely ascend through ephemeral fractures that may not be preserved in the geological record and need not be associated with measurable geophysical signals. Propagation speeds for transient fractures are commonly ~101–10-3 m/s, so timescales for flow and emplacement are relatively short (minutes to years).
Slower transport mechanisms relate to ductile processes on different scales. At small scales, buoyant melt or fluid can segregate from partially molten rock to form melt-rich layers by compaction (Solano et al. 2012) or shear strain (Holtzman et al. 2003). Compaction-driven segregation is both rapid (101–103 years) and efficient for low viscosity melts (basalt to andesite) and their associated fluids. Segregation creates evolved melt lenses and refractory mafic and ultramafic cumulate rocks. Segregated fluids can also separate easily from their parent melts. For high viscosity melts (e.g. rhyolitic magma), segregation timescales are much longer (104 to >106 years) and segregation is correspondingly less efficient.
Stalling Magma Movement: Intrusion Formation
Interruption of magma ascent is necessary to form igneous intrusions and magma chambers. Horizontal intrusions (sills) can form because of density contrasts, rigidity contrasts, or changes in stress state (Menand 2008). The density control hypothesis postulates that magmatic dykes cannot penetrate far through rocks that have lower density than the penetrating melt and that stably stratified density interfaces promote intrusion (Taisne et al. 2011). In a density-stratified crust, this hypothesis favours a depth-dependence of magma composition, with denser basaltic magmas intruding deeper than intermediate and silicic magmas (Putirka 2017 this issue). Major density interfaces, such as the geophysical Moho and Conrad discontinuities, could be preferred locations for intrusion. Sill formation is also favoured at interfaces where high-rigidity rocks overlie low-rigidity rocks. In this way, intrusions can localise around major geological discontinuities such as the brittle–ductile transitions in the mid crust. Finally, a vertical minimum stress (s3), as is common in compressional tectonic settings, favours sill formation.
Predicting conditions that promote substantial intrusions requires consideration of both the thermal evolution and the “room problem”. Thermal models of growth by incremental intrusion (Annen et al. 2015) find that magma chambers can form when a magma flux threshold is exceeded and when there are high surrounding temperatures. The condition of high ambient temperature suggests that magma bodies should form most readily in the lower crust, leading to models of lower crustal mixing and assimilation (Hildreth and Moorbath 1988) and of deep crustal hot zones where regions of partial melt can be sustained for long times and where extensive magma differentiation may occur (Annen et al. 2006). In the colder upper crust, formation of magma chambers and associated crystal mush requires magma fluxes that greatly exceed estimated time-averaged fluxes, implying that magma chambers are transient and episodic features. This analysis further suggests that the proportion of plutonic rock relative to country rock should increase with depth, an hypothesis that is borne out by analysis of whole crustal sections through arcs and orogenic zones (Paterson et al. 2011).
The geometry and location of magma accumulation depends on geological structures, strain, strain rate, causative stress systems and rheological properties. While dykes are the main means of vertical magma transport, there is limited direct evidence for dyke-like magma reservoirs, even in extensional settings such as mid-ocean ridges (e.g. Sinton and Detrick 1992). At mid-ocean ridges, dykes form in response to tectonic stresses, cool quickly when they reach the shallow brittle crust, and feed thin and shallow melt lenses. The potential for forming large magma systems by dyking may be limited to continental rifts. For many tectonic environments, however, accommodation space remains low.
Horizontal intrusion has a much greater potential for space creation and has been invoked in several recent tomographic studies (e.g. Huang et al. 2015). In the shallow and brittle upper crust, intrusions can lift the Earth’s surface by faulting and folding. In the deep crust, room is created by viscous deformation of host rocks. Much of the literature has emphasized upward bulk magma transport by diapirism. However, two other mechanisms can operate in the ductile mid- to lower-crust. First, horizontal intrusions may act like viscous gravity currents that displace surrounding crust by viscous flow. Second, components of intruding magma can become denser by crystallization and physical segregation of basal mafic-rich cumulates, which cause the intrusion floor to sag (Roman and Jaupart 2016). In this mechanism, downward migration of dense cumulates balances upward migration of more buoyant magma (Paterson et al. 2011). With typical viscosities of 1017–1018 Pa s and reasonable magma flux rates, intrinsically slow ductile deformation of hot, but sub-solidus, rocks is expected for magma transport in the middle and lower crust.
Magma and Mush Processes
Within a vertically extensive magmatic system, the room problem is minimised and a new mechanism of forming magma chambers can be invoked. Melt-bearing mush accumulations are highly susceptible to segregation of melt-rich layers by compaction and shear. This creates a magma–mush system that is inherently unstable. We suggest that within such a system, melt layers will episodically move upwards, amalgamate and mix together; at the same time, the surrounding mush will subside (Fig. 3). Importantly, there is no volume change associated with the vertical re-organisation of incompressible phases (melts and crystals). Two timescales underpin magma accumulation by internal mush processes. First, long timescales are associated with melt segregation and are typically 103–105 years (Solano et al. 2012). Second, short timescales are associated with instability, re-organisation and amalgamation and are governed by the relatively low viscosity of mush (1013–1014 Pa s for 40% melt) (Costa et al. 2009). To address timescales of instability for growing buoyant melt-rich layers within a melt-rich mush we can apply Equation 26 in de Bremond d’Ars et al. (1995). An assumed wavelength of 5 km yields approximate timescales of destabilisation of months to several years. These are the timescales of volcanic eruptions.
Dynamic States Of Magmatic Systems: Dormancy, Unrest, Eruption
Above, we identified two different ways of forming magma chambers: (1) incremental intrusion into sub-solidus rocks, distinguishing between brittle and ductile environments and (2) involved the segregation of melts (and fluids) from mush and the amalgamation of these melts during periods of instability (Fig. 3). Although individual volcanic systems can be dominated by either or both mechanisms, low rates of magma supply, as would be expected during the early stages of volcanism, will favour intrusive processes, while in mature long-lived systems mush processes can dominate. In this section, we focus on mush system dynamics in mature systems.
Herein, we identify three dynamical volcanic regimes that are, or have been, supplied by mature transcrustal magmatic systems: the dormant state, the state of unrest, and the eruptive state (Fig. 3).
- The dormant state is characterised by slow melt (and fluid) segregation from the mush and growth of multiple melt-rich layers. This regime is driven by an influx of mantle magmas, and likely involves development of melt or fluid layers. The volcano itself will appear dormant, with no surface activity beyond fumaroles and hydrothermal activity.
- In a state of unrest there is significant instability over short timescales; melt and/or fluid layers amalgamate and ascend as the residual mush moves downward. Increases in pressure that accompany the decompression of fluids and fluid-saturated melts should generate detectable changes in the near surface, such as surface deformation, enhanced seismicity and magmatic fluid release. However, in the state of unrest, no eruption takes place and the unrest dies down as the system stabilises.
- In the eruptive state the system is so perturbed that magma and fluids break through to the surface. The two regimes of unrest and eruption involve fast processes that enable shallow bodies of eruptible magma to accumulate and sometimes erupt. Qualitatively, episodic volcanism can be explained as the alternation of long repose intervals (centuries, millennia or longer) and short periods of unrest and eruptions (days to decades), reflecting the coupling of slow and fast processes.
An important question for forecasting is whether periods of unrest will develop into eruptions. We envisage no fundamental difference between the processes that cause unrest leading to eruption and processes that cause unrest but no eruption. In fact, unrest is widely acknowledged as much more frequent than eruptions at many volcanoes (Moran et al. 2011). Multiple episodes of unrest without eruption may record progressive development of a shallow magma chamber of increasing volume on a pathway to eventual eruption. In this case, eruptions are triggered by reaching a critical overpressure sufficient to form magma pathways to the surface and are treated as either time- or volume predictable. That these simple models do not always apply provides evidence of other important factors, including processes acting throughout the transcrustal magmatic systems.
A major challenge in volcano monitoring is how to use signals collected from instruments at the Earth’s surface to infer movement of magma and fluids at depth. Historically, seismic signals have provided the first warnings of volcano unrest, signals that have improved with new instruments and denser networks. Volcano-related earthquakes record rock breakage and reflect changes in stress caused by pressure changes within magmatic systems, by movement of fluids within hydrothermal systems, by discharges of fluid from magmatic systems, and by intrusion and magma flow along conduits to the surface. Changes within the magmatic system may activate far-field crustal faults and fractures; conversely, tectonic stresses may trigger changes in the magmatic system.
Deformation can be measured by geodetic measurements on the ground (global positioning system, electronic distance metres, tiltmeters), satellite-based interferometric synthetic aperture radar (InSAR) and strain meters in bore-holes. Patterns of deformation are typically interpreted by applying Mogi models for elastic deformation from changing source pressure. Although usually attributed to magma intrusion into a shallow chamber, pressure changes inferred from deformation can also reflect volatile processes (exsolution or degassing) or changes in shallow hydrothermal systems. As shown below, in unstable and vertically extensive magmatic systems upward magma transport in itself does not cause large volume changes because the upward flux of magma is compensated by the downward transport of mush. Under these circumstances, volatile exsolution and decompression may be the primary cause of volume changes, pressurisation and deformation (e.g. Chang et al. 2010; Christopher et al. 2015).
Gas emissions can be directly sampled at fumaroles, and indirectly sampled by remote sensing from satellites and ground instruments. Routine gas monitoring focuses on SO2 because its low atmospheric abundance allows routine and accurate monitoring. Instruments to investigate other gases, such as halogens, CO2 and water, are typically deployed in campaign research projects and are not currently part of real-time monitoring systems. Increases in SO2 can provide robust evidence of magmatic gas release in some volcanoes (e.g. Christopher et al. 2015). Implications of gas data for eruption potential are, however, commonly ambiguous. A decline in SO2 might imply a decrease in eruption potential or it might mean that gas has become trapped and the volcano is preparing to explode.
Implications For Monitoring And Forecasting Different volcanic states
A dormant volcano is one that shows no detectable signs that can be construed to indicate an impending eruption: there is either no geophysical activity whatsoever or there is a low-level state of background activity (e.g. earthquakes, minor deformation and active fumaroles). Background activity includes long-established hydrothermal systems on some dormant volcanoes can cause pronounced seismic activity, as well as deformation and discharge of fluids, experience may suggest that they do not indicate a pathway to magmatic eruptions.
While there may be no upper crustal unrest in a dormant volcano, the deeper magmatic system can be active, as manifested by slow processes of melt and volatile segregation in mush and by crystallization and gas exsolution in slowly cooling magma. During compaction, melts slowly move upwards, decompress, and, if the melt is volatile-saturated, exsolve fluids. For this reason, volatiles in transcrustal magmatic systems will continually exsolve and percolate towards the surface (Fig. 3 top). Volatile compositions are strongly depth dependent such that low-solubility volatiles (notably CO2) exsolve at relatively deep level whereas high-solubility volatiles (notably water and sulphur species) exsolve at relatively shallow levels. The dramatic density and viscosity contrast between melts and fluids causes them to decouple (Christopher et al. 2015). Thus, large volumes of fluids can be generated, separated from their parent magmas and then accumulate as discrete layers within magmatic systems. This fluid may leak directly to the surface as fumaroles and regions of diffuse degassing or may add magmatic fluid components to near-surface hydrothermal systems. However, magmatic fluids can also be stored for long periods of time within the magmatic system, to be released much later, perhaps in association with eruptions.
The above “slow” processes occur in hot ductile environments at low strain rates. They may be largely aseismic, although transient movements of magmas and fluids may be monitored by tracking long-period earthquakes in the lower and middle crust (Nichols et al. 2011). Even in dormant volcanoes, slow exsolution and upward migration of compressible volatiles can cause positive volume changes and deformation, as well as feeding fumaroles and diffuse surface gas emissions and contaminating hydrothermal systems. An example is the 1978–2004 period of unrest at the Long Valley caldera in California (USA) (Hill 2006), where more than three decades of non-eruptive unrest has been attributed in part to fluid fluxes from a transcrustal magmatic system.
State of Volcanic Unrest
Unrest is common at volcanoes around the world. Although signs of unrest may precede volcanic eruptions, unrest in volcanic regions neither implies nor requires eruption (Moran et al. 2011). Unrest that is not immediately related to eruption can be triggered by changes in hydrothermal systems that are unrelated to magmatic disturbances such as unrest triggered by tectonic activity. As outlined above, non-eruptive unrest may also record input of fluids from fast processes deep within magmatic systems that accompany destabilisation, upward movement of magma and fluids, and general rearrangement of magma, fluid, and mush (Fig. 3 middle). In this scenario, the deep processes may be difficult or impossible to detect if related to ductile processes or if resulting from earthquakes that are either too small to detect or that are dissipated by the mush itself. Exceptions include deep (mid-crustal) long-period earthquakes, which are commonly interpreted as movement of either magma or fluid (Nicholls et al. 2011). In the shallow crust, pressure changes associated with rapid decompression of fluids can be large enough to deform the crust elastically. The result can be earthquakes, surface deformation and enhanced fumarole emissions.
Exsolved volatiles can become strongly decoupled from their parent magmas. Implicit in this concept is the notion that large volumes of “old” magmatic volatiles could separate into volumetrically significant accumulations within dormant systems, only to be released during periods of destabilisation (e.g. Hill 2006; Christopher et al. 2015). Thus, volcanic unrest could arise by destabilisation of magma layers, by volatile accumulations or, indeed, both together. Purely volatile-driven unrest should be considered as, for example, sources of pressure for ground deformation.
We suggest that many eruptions are the end result of major and rapid destabilisation and reorganisation of transcrustal magmatic systems. The transcrustal model stands in contrast to models of incremental magma chamber growth. Nevertheless, both end-member models clearly represent points along a spectrum of possible behaviours. One important consequence of the rapid destabilisation model is that multiple layers of melts, magmas and volatiles could merge to form larger eruptible magma bodies (Fig. 3 bottom). If pressure changes caused by destabilisation are large enough, then the magmas and fluids break to the surface and erupt. Instability can also continue during an eruption if eruption-related pressure fluctuations control the evolution of activity. The question then becomes one of recognizing the signs of magmatic unrest that indicate large-scale destabilisation. Here, we frame this problem by considering, in particular, timescales of likely processes and the critical role of volatiles.
Short timescales are anticipated for internal re-organisation of transcrustal magmatic systems in a ductile dynamic regime. An approximate timescale for the brittle–ductile transition is the ratio of viscosity (m) to the elastic modulus (E). Mush systems have approximate viscosities between 1013 Pa s (high melt fractions) and 1017 Pa s (low melt fractions), so for E ~1010 Pa, the timescales above which ductile behaviour can be expected are tens of minutes to a few months. Similar short timescales can be anticipated for volatile exsolution in response to upward migration. These processes have potentially major implications on how to interpret ground deformation signals.
Commonly, surface deformation signals are attributed to magma intrusion (replenishment) into shallow magma chambers, although usually no explanation is given for that replenishment. The volume changes observed from deformation are then used to estimate volumes of new (intruded) magma. However, the resupply of a shallow magma chamber by destabilisation of a transcrustal magma system (Fig. 3) provides an alternative interpretation: that ground inflation is caused primarily by volatile exsolution and expansion (or, conversely, that ground deflation may record degassing), and that the rearrangement of incompressible melt and crystal components plays little to no role. The implications of volatile-generated, rather than magma-generated, pressures as an explanation for measured surface deformation are hugely different from those of traditional (recharge) models. Consider the volumes of magma implied. If shallow magma chambers are supplied by a destabilisation and amalgamation event, but volume change is created by exsolution of only a small mass fraction of volatiles, then the implied volume of magma addition may be one to two orders of magnitude larger than if it were assumed to reflect magma alone. From this, we suggest that replenishment models that do not fully consider the role of volatiles may both over- and under-estimate the amount of magma responsible for the observed deformation. Over-estimates can result if the exsolved volatiles segregate from the magmatic system, or if they recharge the shallow magma chamber and eventually leak from the system. Alternatively, eruptible magma volumes may be greatly under-estimated if exsolved (and compressible) volatiles are retained within the recharge magma.
Although we have emphasised the role of internal processes in transcrustal magmatic systems, volcanic activity can also be modulated by tectonic conditions (e.g. Wilson and Rowland 2016). A given destabilisation scenario will only lead to eruption if the stress regime is suitable for developing dyke systems to the surface, and this in turn depends both on the strain rate and the stress state of the crust. For this reason, development of truly comprehensive volcano forecasting models also requires the incorporation of processes that account for the interplay between internal igneous processes, strain rates and stress systems that can be developed by both tectonic and magmatic processes.
The emerging concepts of transcrustal magmatic systems provide the basis for understanding controls on the dynamics and magnitudes of volcanic eruptions. The idea of slow segregation of melts and fluids over long periods of time followed by rapid destabilisation provides a chemical and physical framework for anticipating recurrence timescales. Episodic volcanism and unrest reflect fast processes associated with magma and fluid transfer; repose in the dormant state reflects slow processes governed by the high viscosity of the crust and mush and modulated by external factors such as tectonics, melt generation, crustal structure and heat transfer. Igneous processes in the middle and lower crust are, we think, associated with ductile deformation, and this represents a major challenge: there needs to be found ways of detecting, quantifying, and interpreting these deeper processes, e.g. identifying deep long-period earthquakes (Nichols et al. 2011).
As an example of how these new concepts impact on forecasting in real cases, we briefly discuss the 1995 to 2010 eruption of the Soufrière Hills Volcano (Montserrat) drawing principally from Christopher et al. (2015). Since the last phase of dome extrusion in 2010, the volcano has continued to inflate and emit an average of about 370 tons of SO2 per day. Key questions on Montserrat for management and planning are whether the dome extrusion and associated hazardous activity will restart or if, in fact, the eruption has ceased. While the new concepts cannot provide a definitive answer to these questions, they do inform long-term probabilistic assessments. For example, if the inflation and continued high SO2 emissions are interpreted to represent magma recharge, then the observations are more ominous than if the activity reflects the release of volatiles from a (previously) accumulated fluid layer formed within an unstable transcrustal magmatic system. Such admittedly qualitative notions can be fed into expert elicitation methods for prognosis and advice. The point is that changing concepts of how volcanoes work feed directly into interpretation and forecasting. And, developing and testing new conceptual models are critical for improving volcanic hazard assessment.
Annen C, Blundy JD, Sparks RSJ (2006) The genesis of intermediate and silicic magmas in deep crustal hot zones. Journal of Petrology 47: 505-539
Annen C, Blundy JD, Leuthold J, Sparks RSJ (2015) Construction and evolution of igneous bodies: towards an integrated perspective of crustal magmatism. Lithos 230: 206-221
Bachmann O, Huber C (2016) Silicic magma reservoirs in the Earth’s crust. American Mineralogist 101: 2377-2404
Blundy J, Cashman KV, Berlo K (2008) Evolving magma storage conditions beneath Mount St. Helens inferred from chemical variations in melt inclusions from the 1980–1986 and current (2004–2006) eruptions. In: Sherrod DR, Scott WE, Stauffer PH (eds) A Volcano Rekindled: the Renewed Eruption of Mount St. Helens, 2004-2006. US Geological Survey Professional Paper 1750: 755-790
Chang W-L, Smith RB, Farrell J, Puskas CM (2010) An extraordinary episode of Yellowstone caldera uplift, 2004-2010, from GPS and InSAR observations. Geophysical Research Letters 37, doi: 10.1029/2010GL045451
Christopher TE and 6 coauthors (2014) Petrological and geochemical variation during the Soufriére Hills eruption, 1995 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 317-342
Christopher TE and 7 coauthors (2015) Crustal-scale degassing due to magma system destabilisation and magma-gas decoupling at Soufrière Hills Volcano, Montserrat. Geochemistry, Geophysics, Geosystems 16: 2797-2811
Cooper GF, Wilson CJN, Millet M-A, Baker JA, Smith EGC (2012) Systematic tapping of independent magma chambers during the 1 Ma Kidnappers supereruption. Earth and Planetary Science Letters 313: 23-33
Costa A, Caricchi L, Bagdassarov N (2009) A model for the rheology of particle-bearing suspensions and partially molten rocks. Geochemistry, Geophysics, Geosystems, doi: 10.1029/2008GC002138
Davidson JP, Morgan DJ, Charlier BLA, Harlou R, Hora JM (2007) Microsampling and isotopic analysis of igneous rocks: implications for the study of magmatic systems. Annual Review of Earth and Planetary Sciences 35: 273-311
de Bremond D’Ars J, Jaupart C, Sparks RSJ (1995) The distribution of volcanoes in active margins. Journal of Geophysical Research: Solid Earth 100: 20421-20432
Hawkesworth C, George R, Turner S, Zellmer G (2004) Time scales of magmatic processes. Earth and Planetary Science Letters 218: 1-16
Hildreth W, Moorbath S (1988) Crustal contributions to arc magmatism in the Andes of Central Chile. Contributions to Mineralogy and Petrology 98: 455-489
Hill DP (2006) Unrest in Long Valley Caldera, California, 1978–2004. In: Troise C, De Natale G, Kilburn CRJ (eds) Mechanisms of Activity and Unrest at Large Calderas. Geological Society, London, Special Publications 269, pp 1-24
Holzman BK, Groebner NJ, Zimmerman ME, Ginsberg SB, Kohlstedt DL (2003) Stress-driven melt segregation in partially molten rocks. Geochemistry, Geophysics, Geosystems 4: doi:10.1029/2001GC000258
Huang H-H and 5 coauthors (2015) The Yellowstone magmatic system from the mantle plume to the upper crust. Science 348: 773-776
Menand T (2008) The mechanics and dynamics of sills in layered elastic rocks and their implications for the growth of laccoliths and other igneous complexes. Earth and Planetary Science Letters 267: 93-99
Moran SC, Newhall C, Roman DC (2011) Failed magmatic eruptions: late-stage cessation of magma ascent. Bulletin of Volcanology 73: 115-122
Morgan DJ, Costa F (2010) Time constraints from chemical equilibration in magmatic crystals. In: Dosseto A,
Turner S, Van-Orman JA (eds) Timescales of Magmatic Processes: From Core to Atmosphere. Wiley-Blackwell, p 125-159
Nichols ML, Malone SD, Moran SC, Thelen WA, Videle JE (2011) Deep long-period earthquakes beneath Washington and Oregon volcanoes. Journal of Volcanology and Geothermal Research 200: 116-128
Paterson SR, Okaya D, Memeti V, Economos R, Miller RB (2011) Magma addition and flux calculations of incrementally constructed magma chambers in continental margin arcs: Combined field, geochronologic, and thermal modeling studies. Geosphere 7: 1439-1468
Putirka KD (2017) Down the crater: where magmas are stored and why they erupt. Elements 13: 11-16
Roman A, Jaupart C. (2016) The fate of mafic and ultramafic intrusions in the continental crust. Earth and Planetary Science Letters 453: 131-140
Rubin AM (1995) Propagation of magma-filled cracks. Annual Review Earth and Planetary Sciences 23: 287–336
Sinton JM, Detrick RS (1992) Mid-ocean ridge magma chambers. Journal of Geophysical Research: Solid Earth 97: 197-216
Solano JMS, Jackson MD, Sparks RSJ, Blundy JD, Annen C (2012) Segregation in deep crustal hot zones: a mechanism for chemical differentiation, crustal assimilation and the formation of evolved magmas. Journal of Petrology 53: 1999-2026
Taisne B, Tait S, Jaupart C (2011) Conditions for the arrest of a vertical propagating dyke. Bulletin of Volcanology 73:191-204
Tarasewicz J, White RS, Woods AW, Brandsdottir B, Gudmundsson MT (2012) Magma mobilization by downward-propagating decompression of the Eyjafjallajokull volcanic plumbing system. Geophysical Research Letters 39, doi: 10.1029/2012GL053518
Wilson CJN, Rowland JV (2016) The volcanic, magmatic and tectonic setting of the Taupo Volcanic Zone, New Zealand, reviewed from a geothermal perspective. Geothermics 59: 168-187