Volcanoes have a life cycle. They are usually conceived by melting in the mantle, and hence their locations are controlled by plate tectonics and mantle convection (Box 2.1). The silicate melts can then ascend to the surface directly, or accumulate in the crust where their volumes and compositions change as they interact with their surroundings. Magma can have a complex history underground. The eruption of magmas creates volcanoes and affects other surface environments such as the hydrosphere and atmosphere. The interactions between melting, storage, accumulation, eruption, and geologic setting give rise to the great diversity seen in eruptions and volcanic landforms.
Each volcano has its own distinct life cycle, often with multiple episodes of repose, unrest, and eruption. Yet the evolution and eruption of all volcanoes are still governed by the same set of processes intrinsic to the magma and influenced by geologic setting. Thus a central challenge to understanding how magma is generated, is stored, ascends, and erupts is to disentangle the unique features of the birth, life, and death of each volcano from the common processes governing their life cycles. This chapter summarizes current understanding of how volcanoes work and identifies key questions and research priorities in three areas: (1) processes that move and store magma beneath volcanoes; (2) how eruptions begin, evolve, and end; and (3) how a volcano erupts.
The path magma takes to the surface is poorly understood. Magma is buoyant and rises through the crust, sometimes erupting at the surface. At hotspots such as Iceland, Hawaii, and some volcanoes in the western United States, magma can ascend directly from the mantle to the surface. But much of the time, magma stalls and forms reservoirs that later erupt or freeze (Figure 2.1). Magma cools because the crust is cooler than the magma, and magma decompresses as it rises. Cooling leads to crystallization and increased viscosity. Decompression may lead to increased buoyancy due to the formation of bubbles from gas originally dissolved in the melt. A loss of volatiles also increases the melt viscosity. Storage and ascent are influenced by the mechanical properties and behavior of the crust, including its ability to deform, flow, or fracture. These properties evolve over the life cycle of the volcano. The competing drivers that force magma to rise and also to resist movement are partly what makes magma movement and eruption so difficult to forecast (Melnik and Sparks, 2006). Will magma stall because of increased viscosity? Or will bubble expansion accelerate magma to the surface in an explosive eruption? The processes that move and store magma are thus fundamental not only to the transfer of mass from the interior to the
exterior of Earth over its history, but also to the style, intensity, magnitude, and duration of volcanic eruptions (Acocella, 2014).
Most volcanoes are not continuously active but spend much of their lifetime at rest, sometimes for thousands of years before erupting again. Prior to an eruption, the movement of magma and fluids may cause earthquakes beneath the volcano, gas emission into the atmosphere or aquifers, and uplift of the ground surface. Importantly, however, these signs of volcanic unrest do not always presage an eruption. Even at rest, volcanoes are unstable landforms, prone to rapid ero-
sion and collapse, creating hazards even in the absence of eruption. Thus, the life cycle of volcanoes involves alternating periods of repose and unrest punctuated by eruption. We still do not know which signs of unrest signal magma versus gas movement. Which are precursors to eruption? What is normal background activity of volcanoes over their life cycle?
Detecting Magma Under the Ground
A major challenge in studying magma movements is that we are unable to see directly where magma is stored and how it moves under the volcano. From this perspective, it is fortunate that magma does not move quietly or gently. Instead, magma, its movement, and the stresses it generates in the surrounding rock can be detected using the deformation of Earth’s surface, the location of earthquakes (Figure 2.2), the frequency of ground shaking, the direction of displacement on faults, and the way seismic waves propagate through the crust. For example, as magma rises and pressurizes subvolcanic reservoirs, it causes the ground surface to rise (inflate). Ground deformation at the scale of millimeters can be sensed with satellite radar, Global Positioning System (GPS), tiltmeters, and strainmeters. Using these tools, it is possible to constrain a combination of the depth and shape of magma reservoirs. Some erupting volcanoes are observed to “breathe,” as the subsurface inflates prior to eruption, then deflates after the eruption due to magma removal (Figure 2.3). However, some breathing cycles are not coupled to eruptions and may originate in the hydrothermal systems surrounding the magma.
Geophysical imaging provides additional observational constraints on the location, geometry, and state of magma stored in the crust. Recent advances in seismic tomography, including combinations of body and surface waves (e.g., Syracuse et al., 2015), full waveform inversion and imaging of reflected signals (Arnulf et al., 2014), and the use of the ambient seismic wave field, allow four-dimensional (space and time) imaging of crustal magma reservoirs (e.g., Brenguier et al., 2016; Greenfield et al., 2016; Huang et al., 2015; Jaxybulatov et al., 2014). Tomographic images reveal seismic wave speeds which, when combined with experimental data, yield estimates of temperature and the percentage of partial melt present (e.g., Figure 2.4). Attenuation tomography measures the decay in seismic
wave amplitude and can be particularly sensitive to the presence of melt (e.g., De Siena et al., 2014; Lin et al., 2015). Magnetotelluric surveys provide complementary information on the presence of melt and hydrothermal reservoirs (e.g., Desissa et al., 2013; Peacock et al., 2016). Joint inversion of magnetotelluric, seismic, gravity, and other geophysical data has the potential to tighten bounds on subsurface magma systems because the different data types are sensitive to distinct physical properties of the magma and host rocks. Including petrological constraints on composition, temperature, and volatile content (e.g., Comeau et al., 2016) reduces the uncertainty and makes interpretations more physically plausible.
Gases emitted before and during eruption, diffusively and from vents, provide clues about the locations, masses, and histories of magma in the crust (e.g., Aiuppa et al., 2011; Lowenstern et al., 2014). Real-time sampling of the erupted material provides information about syn-eruptive changes in intrinsic magma properties, such as temperature, viscosity, density, and pre-eruptive gas content (e.g., Burgisser et al., 2012). Importantly, these different techniques provide different and complementary images of the subsurface (e.g., Chiodini et al., 2012).
Geologic and geochemical tools have also been developed to study magma underground. Drilling has provided rare samples of magma near the surface, as well as hot rocks and fluids that indicate the temperatures and permeability of the shallowest subvolcanic regions (Elders et al., 2011; Friðleifsson et al., 2013; Marsh et al., 2008; Mortensen et al., 2014; Nakada et al., 2005; Zierenberg et al., 2012). The occurrence of volcanic deposits that are very large and chemically monotonous attests to the existence of large, homogeneous magma reservoirs that supply some giant eruptions (e.g., Bachmann and Bergantz, 2003). The duration of storage, the rate of movement, pressure, and temperature are potentially recorded in erupted crystals (Kahl et al., 2013; Putirka and Tepley, 2008), pockets of melt (now glass) trapped in crystals (Kent, 2008; Sides et al., 2014; Wallace, 2005), and the sizes and shapes of bubbles and crystals (Hammer, 2008; Marsh, 2007; Sable et al., 2006). Interpreting these records is not always straightforward. In particular, the depth of magma storage is difficult to determine and calls for experimental calibration of new crystal and
Despite this large geophysical and geochemical toolkit, resource constraints mean that few volcanic eruptions have been recorded using more than the most basic seismic, deformation, and gas instruments, and both long-term and real-time measurements are often absent or incomplete.
To Ascend or to Stall?
Physical processes and the rheology of the crust govern whether magma ascends from its source to erupt, or stalls and accumulates without erupting. In general, magma will stall if it loses buoyancy, increases in viscosity, or can no longer open and flow through vertical cracks in the surrounding crust. Thus, the intrinsic characteristics of the magma matter: its chemical composition, dissolved and exsolved gas content, temperature, and crystallinity, all of which affect magma density, compressibility, and rheology. Extrinsic parameters also matter, including the density, strength, and stress state of the surrounding crust, and pre-existing weaknesses and structures in the crust. The tools of volcano science are starting to be able to sense magma movement and storage in many regions, and yet it is not known which combination of intrinsic and/or
extrinsic parameters control where magmas stall and accumulate prior to eruption.
The time scales for magma storage and ascent are only now beginning to be quantified. In some settings, magma can be stored for tens to hundreds of thousands of years in magma reservoirs (e.g., Barboni et al., 2016; Kaiser et al., 2017). Some volcanoes erupt magma that has traversed the entire crust (40 km on average) in a few hours to days (Demouchy et al., 2006). Ascent from crustal magma chambers can take only a few hours (e.g., Castro and Dingwell, 2009) or as little as a few minutes (e.g., Humphreys et al., 2008). Constraining these time scales is critical for improving forecasting.
Most magma transport through the crust takes place through cracks known as dikes. Dike propagation involves coupling between fluid flow, solid deformation, and heat transfer (Rubin, 1995). If the melt cannot flow sufficiently rapidly it will cool, become more viscous, and eventually freeze (Rubin, 1993). The direction of dike growth and the focusing of magma toward or away from a central volcano are controlled by the crustal stress state and therefore can be influenced by magma reservoirs (Buck et al., 2006; Karlstrom et al., 2009), surface loading from volcanic edifices (Muller et al., 2001; Pinel et al., 2010), and large-scale stresses and faults. Whether dikes reach the surface depends on magma chamber overpressure, crustal stress, and density structure (Rivalta et al., 2015). As dikes move upward they push the crust aside, often leading to detectable signals in GPS, tilt, strain, and Inteferometric synthetic aperture radar (InSAR) data (Aoki et al., 1999; Segall et al., 2013; Wright et al., 2006). In the shallow brittle crust, this motion is accompanied by rock breakage, in some cases leading to spectacular propagating swarms of earthquakes that can be used to image the passage of magma (Ebinger et al., 2010; Rubin et al., 1998; Sigmundsson et al., 2015).
Recent studies using dense arrays of seismometers have located small earthquakes in vertical clusters, some as deep as the mantle, which could reflect magma transport in dikes (Figure 2.2). Earthquakes below the brittle–ductile transition may be produced by the high strain rate from dike intrusion (White et al., 2011). In volcanic arcs and thick continental crust, magma usually stalls and accumulates within the crust, cools, mixes with other magma, and chemically evolves before erupting. Geodetic, seismic, and petrologic observations typically point to magma storage at depths between 2 and 7 km (Chaussard and Amelung, 2014). Why is there an apparent “sweet spot” for magma storage? Is this where magma reaches neutral buoyancy and is primed for eruption because it saturates in volatiles, reaches a critical viscosity, or encounters a change
in crustal stress or strength (Plank et al., 2013)? Detecting where magmas are stored, how they are distributed in space, their intrinsic properties, and the properties of the surrounding crust will require improved imaging using seismic and electromagnetic tools hand in hand with better laboratory measurement of the geophysical properties of partially molten rocks. Understanding the mechanical properties of the crust and magma are essential to answering this basic question.
How Are Eruptible Bodies Assembled and How Long Do They Persist?
Revealing how eruptible magma accumulates and evolves requires determining the time scales for key processes, including the persistence of magma systems and the time to replenish and pressurize magma bodies. There is considerable debate about the length of time that liquid-dominated magmas exist within the crust compared to the longevity of the magmatic system as a whole. Studies of plutons, magma bodies preserved “frozen” in the geologic record, frequently record hundreds of thousands or millions of years of crystallization (Coleman et al., 2012; Miller et al., 2011). Crystals that erupt from volcanoes also yield radiometric ages that can be as old as thousands to hundreds of thousands of years, demonstrating the extreme longevity of many magma systems (e.g., Cooper, 2015; Kaiser et al., 2017; Peate and Hawkesworth, 2005; Reid, 2003; Schmitt, 2011; Zellmer et al., 2005). Moreover, some volcanoes have long repose periods between large eruptions (tens of thousands to hundreds of thousands of years), indicating that it can take substantial time to develop the conditions needed for eruption (Figure 2.5). For other volcanoes, the time to develop those conditions is shorter (e.g., Allan et al., 2013; Cooper and Kent, 2014; Druitt et al., 2012). For example, while the growth of magma bodies may take hundreds of thousands of years, the time between recharge events and eruption for large caldera-forming eruptions may be less than 100 years (e.g., Druitt et al., 2012). Magma may recharge reservoirs in less than days to months
for smaller systems or when magma viscosity is lower (Albert et al., 2016; Rae et al., 2016). Gas emissions often exceed that which can be derived from only the erupted magma (e.g., Christopher et al., 2015; Shinohara, 2008) and suggest degassing and extraction from nonerupted magma remaining at depth. All of these lines of evidence point to the importance of magma accumulation in reservoirs, where it stages prior to eruption, and where a significant proportion of magmas crystallize without erupting. What controls the fraction of magma that eventually erupts? The fraction of magma that erupts is difficult to constrain, exposures of both erupted and unerupted magma are few and far between, and patterns and controls remain difficult to quantify. An upper bound on the long-term average eruption rate is the melt production rate, but in many settings, only a small fraction of magma erupts (White et al., 2006).
The dynamics of magma bodies may be complex and varied. Large and hot magma reservoirs may convect and mix (Bergantz et al., 2015). Thus, the thermal and geometric states of the reservoir are critical to its dynamics and longevity, as they affect viscosity, crystallization, gas exsolution, and freezing (e.g., Gudmundsson, 2012). The primary parameters that control longevity are the temperature of the magma, the mechanical properties surrounding rocks, and the magma flux into the system (Figure 2.6). The
longevity, magnitude, and melt content of eruptible magma bodies have significant implications for hazard assessment and detection of melt prior to eruption and are therefore important targets for future study using combinations of models, geological mapping, geophysical imaging, rock physics measurements, and detailed studies of crystals.
How Quickly Is Magma Mobilized Prior to Eruption?
The regions where magmas stall may help set the course for eruption by influencing the development of overpressure, the accumulation of gas, and the segregation of melt from crystals (Pioli et al., 2009). Magmatic systems that have been in repose for thousands of years may quickly mobilize to eruption following injection of new (“recharge”) magma with fresh gas (Bachmann and Bergantz, 2006; Huber et al., 2010). Remobilization might occur several times during transit through the crust, and only the final remobilization may lead to eruption (e.g., Reid and Vazquez, 2017). Evolving conditions during eruption may also mobilize magmas through progressive connection of previously isolated melt lenses (Cashman and Giordano, 2014). New microanalytical techniques (Box 2.2) have recently revealed rich chemical records inside volcanic crystals that may record the timing of injection events days to years before eruption (e.g., Rosen, 2016). Physics-based models predict different triggering mechanisms depending on the magma flux into the reservoir and the behavior of the surrounding crust (Figure 2.6).
Magmatic temperatures and evolution can be constrained using crystal-melt chemical thermometers and diffusion chronometry, but magma recharge volume and history are difficult to constrain. Progress in monitoring magma migration through the crust, accumulation in shallow reservoirs, and the approach to the tipping point for eruption require integrating geophysical measurements, the geochemical and petrologic record preserved in erupted materials, and models that account for the evolution of magma bodies and their interaction with their surroundings.
Key Questions and Research Priorities on Processes That Move and Store Magma
Anticipating when an eruption will begin, how it will evolve over time, and when and why it will end are among the greatest challenges in volcano science. Most volcanoes are not continuously active but spend much of their lifetime at rest, sometimes for thousands
of years before erupting again. Prior to an eruption, the movement of magma and fluids may cause earthquakes beneath the volcano, gas emission into the atmosphere or aquifers, or uplift of the ground surface. However, these signs of volcanic unrest do not always presage an eruption. Similarly, some eruptions occur without precursory unrest detectable with our current methods. We still do not know how to interpret the signs of unrest unequivocally. Which are precursors to eruption? What is normal background activity of volcanoes over their life cycle?
Eruptions begin when magma ascends toward the surface, either by propagating in a new dike from the storage region (Section 2.1) or by rising through a preexisting conduit, potentially displacing and interacting with older magma. For volcanic systems in repose, it is commonly assumed that eruptions are preceded by pressure increases within shallow magma reservoirs. The ultimate trigger for eruption can be transfer of additional magma from deeper in the crust (recharge) or changes in the volatile budget (e.g., Girona et al., 2015; Tait et al., 1989). However, the resulting pressure changes may require years, decades, or even centuries to initiate eruption at the surface (Druitt et al., 2012; Morgan et al., 2006). Volcanic unrest that precedes eruptions (run-up) may occur over hours to years (Figure 2.5), although such signals may precede eruptive activity by years to decades (Biggs et al., 2014; Phillipson et al., 2013).
The duration and nature of precursors also depend on the physical setting (tectonic environment and rheology of the crust) and the rheology of the rising magma (e.g., Roman and Cashman, 2006). In general, low-viscosity basaltic magmas ascend rapidly and with brief precursory activity (e.g., Albert et al., 2016; Passarelli and Brodsky, 2012). High-viscosity silicic magmas, in contrast, often have longer run-up periods and may begin with weak gas-driven or phreatic eruptions. This precursory activity acts to construct a magma pathway to the surface (a volcanic conduit). When the conduit is fully developed, buoyancy and the pressure difference between the magma storage region and the surface drive eruptive activity (e.g., Scandone et al., 2007). Some volcanoes, in contrast, maintain connections between magma storage regions and the surface over periods of decades to centuries. These are commonly referred to as open-system volcanoes and may emit gas continuously and erupt with little to no precursory activity. A central challenge is to explain and understand the great variety of styles and durations of all eruptions, and then to incorporate this understanding into eruption forecasting.
Eruptions may initiate either explosively or effusively, and commonly pass from one style to the other during an eruption. Whether exsolved gas escapes from the conduit, favoring lava effusion or causing ascending magmas to stall, or remains physically coupled to the magma, promoting explosive eruption, is strongly influenced by the properties of the melt and the specific nature of the volatile species. Melt viscosity modulates the rates of volatile segregation both prior to and during eruptive activity; for this reason, silicic, high-viscosity magmas are more prone to highly explosive activity than mafic, low-viscosity magmas.
Fragmentation of magma—breakage into small pieces—is a set of critical processes that are required, though not sufficient (Gonnermann and Manga, 2003), for explosive eruption. Rising and decompressing magma will explode, or fragment, if the bubbles contained with the melt cannot expand sufficiently rapidly to accommodate the change in pressure and remain trapped in the melt (Gonnermann, 2015; Proussevitch and Sahagian, 1998; Sparks, 1978; Zhang, 1999), or if strain rates are high enough to drive a viscous magma through the glass transition (Burgisser and Degruyter, 2015; Cashman and Scheu, 2015; Dingwell, 1996; Spieler et al., 2004). Alternatively, rapid decompression and expansion of low-viscosity magma stretches the melt into thin sheets and filaments that are hydrodynamically unstable and tear into fragments (Houghton and Gonnermann, 2008; Namiki and Manga, 2008). Explosive fragmentation can be triggered by sudden decompression during flank failure or collapse of viscous lava flow fronts (Alidibirov and Dingwell, 1996). The former is best exemplified by the May 18, 1980, eruption of Mount St. Helens, when a collapse-triggered lateral blast (Kieffer, 1981; Ongaro et al., 2011) was followed by downward migration of a decompression wave that ultimately intersected the magma storage region and resulted in several hours of sustained eruptive activity (Criswell, 1987; Figure 2.7).
Magma may also fragment nonexplosively, for example by disruption of the crusts of lava flows or in fail–heal cycles during shear deformation of highly viscous magma (e.g., Tuffen et al., 2003). These various fragmentation processes are each fundamentally different, and produce dramatically different particles (Rust and Cashman, 2011).
Conversely, magma will erupt effusively if strain rates remain small enough, if bubbles can expand freely in response to decompression, or if bubbles rise buoyantly or escape through permeable pathways (Gaunt et al., 2014; Jaupart and Allègre, 1991; Okumura et al., 2009; Rust and Cashman, 2004). All of these processes are favored by low-viscosity mafic magmas such as basalt. Nevertheless, there are sustained eruptions of mafic magma (e.g., Vinkler et al., 2012) that challenge this basic understanding.
Many eruptions initiate explosively, often suddenly, even when preceded by weeks to months of precursory unrest. By its nature, the initiation of eruption is difficult to observe in detail, although both seismoacoustic (Johnson and Lees, 2000; Patanè et al., 2013) and radar (Gouhier and Donnadieu, 2010; Scharff et al., 2015) measurements are now being used to characterize the opening seconds to minutes of eruptive activity and the short-term fluctuations in pulsed eruptive activity.
Another common way eruptions begin is steam-driven explosions that occur when magma at high temperature comes into contact with external water (Sheridan and Wohletz, 1983; Zimanowski et al., 2015). Rapid transfer of heat causes water to flash to steam and expand (Kokelaar and Durant, 1983) and magma to quench and fragment (Mastin et al., 2009b; Wohletz et al., 2012). Such phreatomagmatic activity can use thermal energy very efficiently and produce heterogeneous mixtures of juvenile particles, magmatic gas, steam, wall-rock particles, and often, liquid water (Murtagh and White, 2013). Complete mixing is rare and so the products of single explosions may include liquid water droplets and steam as well as both cold and incandescent pyroclasts (Houghton et al., 2015). Open questions remain about (a) the actual mechanism(s) of fragmentation by water interaction as a function of its source (groundwater and surface water) and physical state (ice, liquid, or vapor); and (b) the role of the state of the ascending magma, such as its viscosity and bubble content (e.g., Liu et al., 2015).
When magma ascent is sufficiently slow, eruptions may start effusively. Under these conditions, gas can segregate and outgas from the rising magma at a rate commensurate with that of magma ascent. In fluid magma, this may occur by buoyant rise of large bubbles. In slow-ascending viscous magmas, outgassing occurs through a permeable network of stretched and/ or coalesced bubbles and fractures (e.g., Castro et al., 2012; Lavallée et al., 2013). In the former case, eruptions form fluid lava flows. In the latter case, effusion takes the form of thick lava flows or, when decompression also triggers extensive crystallization, viscous lava plugs and domes.
Improved understanding of eruption initiation requires physical and chemical models informed by petrologic, geophysical, geochemical, and observational constraints. As described in Section 2.1, petrology provides a powerful tool for deciphering conditions of magma residence in upper crustal magma reservoirs (e.g., Turner and Costa, 2007). The same petrologic and geochemical tools can be applied to processes in shallow conduits (Rutherford, 2008). For example, micron-scale analysis of crystal-hosted melt inclusions, matrix glasses, and crystals can be used to track the extent of disequilibrium, and hence the time scales, of processes that are responsible for transitions in eruptive activity (e.g., Costa et al., 2003; Humphreys et al., 2008; Lloyd et al., 2014; Watkins et al., 2012; Zellmer et al., 1999). When combined with textural analysis of bubbles and crystals in pyroclasts, these data can be used, in theory, to reconstruct a complete picture of shallow conduit processes (e.g., Cashman and McConnell, 2005; Liu et al., 2015). In a few cases microanalytical data have been linked directly with real-time data from eruptive observations (e.g., Albert et al., 2016; Blundy et al., 2008; Rae et al., 2016; Saunders et al., 2012). These data, however, have yet to be fully integrated into evaluation of precursory activity. Thus, a major challenge is to integrate analytical and experimental data streams into physics-based models for eruption processes that are testable and that can be used to simulate potential eruption scenarios on short time scales.
As eruptions progress, the style and intensity of the volcanic activity are determined, at least initially,
by patterns of shallow release (exsolution) and retention or escape (outgassing) of gases initially dissolved in the magma (Burgisser and Degruyter, 2015; Castro and Gardner, 2008). These processes drive much of the rich diversity in eruptive behavior (Gonnermann and Manga, 2012; Figure 2.8). Volatile species have very different solubilities, so that low-solubility CO2, for example, will start to exsolve at much greater depths than higher-solubility H2O. Eruption style is also modulated by the depth and geometry of the magma reservoir and volcanic conduit that connects the magma reservoir to the surface. Volcanic products such as lavas and pyroclasts provide our best views of this shallow subsurface magmatic system. A key challenge is to use this erupted material to interpret those subsurface processes that cannot be characterized directly.
The dynamics of eruptive activity can change dramatically with time. For example, initial explosive activity may evolve to short lived (Pinatubo) or long lived (e.g., Kilauea, Santiaguito) effusive eruptions. Alternatively, protracted effusion may be punctuated by larger explosions (e.g., Pallister et al., 2013), and open-system volcanoes may experience rare paroxysmal
explosions following unloading by lava effusion (e.g., Ripepe et al., 2015) or passive degassing (Girona et al., 2015). Effusive eruption and explosive ash venting may also occur simultaneously (Castro et al., 2012). Thus, a major challenge is to understand both sudden and progressive shifts in activity within eruption episodes. Broadly speaking, shifts in eruptive activity may derive from changes in the source (particularly loss of overpressure from magma discharge, and increases from recharge and exsolution), from changes in the conduit geometry (Michieli Vitturi et al., 2008; Wilson et al., 1980), or from rheological transitions within the magma. Loss of overpressure at the source manifests as an exponential decrease in mass eruption rate, as shown by the 1984 effusive eruption of Mauna Loa volcano, Hawaii, the 1988–1990 effusive eruption of Lonquimay, Chile, and the 2004–2005 effusive eruption of Mount St. Helens, Washington (Figure 2.9). Importantly, all three eruptions showed continuous activity. When activity is discontinuous and accompanied by episodic recharge from depth, the mass eruption rate curves may look quite different, as illustrated by the 1995–2010 eruption of Soufriere Hills volcano, Montserrat. The shape and dimensions of the shallow conduit evolve syn-eruptively by erosion and implosion (Eychenne et al., 2013; Kennedy et al., 2005; Sable et al., 2009). In explosive eruptions, conduit geometry modulates both the eruptive flux and whether the erupted plumes are buoyant or collapse.
Effusive activity is often cyclical (Figure 2.10). Cycles of activity may be generated by elastic deformation (Costa et al., 2007; Maeda, 2000) or stick-slip behavior on conduit walls (Costa et al., 2012; Denlinger and Hoblitt, 1999; Iverson et al., 2006; Ozerov et al., 2003). Cyclical behavior may also be generated internally by nonlinear feedbacks between crystallization and gas loss by permeable flow (e.g., Melnik and Sparks, 2002) or other rheological changes (e.g., Michaut et al., 2013). All of these interactions change with magma ascent rate (Figure 2.8), which controls how bubbles and crystals nucleate and grow, how quickly gas segregates from the melt, how magmas heat frictionally along conduit walls, and how magma may pass from fluid to brittle behavior. Feedbacks between processes are common and include the following:
- Changes in crystallinity can cause magma to cross rheological thresholds, localizing deformation, promoting fragmentation, and changing eruption style
- Changes in gas segregation and gas pressure can cause rapid shifts between degassing regimes and changes in melt rheology
- Changes in heating by friction or crystallization can alter the mechanism of magma ascent
- Transitions in deformation behavior can cause magma to break rather than flow
One critical step for improving our understanding of eruption initiation, modulation, and termination is to quantify the key physical processes in the shallow conduit that are not yet well understood or are poorly constrained by data. These include rheological changes caused by changes in the abundances of bubbles and crystals, interactions among bubbles (including coalescence) and crystals, conditions and rates of permeable degassing, thermal effects of flow and phase transitions, mixing and interaction with host rocks, frictional behavior of magma, and modulation of frag-
mentation and transport processes by interacting with water (i.e., ice, liquid, and vapor). Some of these gaps can be filled by experiments on natural magmas (e.g., Kueppers et al., 2006; Lindoo et al., 2016; Mangan and Sisson, 2000; Okumura et al., 2009; Pistone et al., 2012; Takeuchi et al., 2008) and analog materials (e.g., Castruccio et al., 2013; Cimarelli et al., 2011; Mueller et al., 2011; Oppenheimer et al., 2015; Valentine and White, 2012). Understanding other processes requires dynamic, time-varying models of multiphase flow that couple large-scale transport with processes at the scale of particles. The range of flow behaviors that may arise from nonlinearities in these complex systems is illustrated by the huge oscillations in flow rates and flow regime affected by small perturbations in two-phase flow systems (Melnik and Sparks, 2002; Pioli et al., 2012).
New observational research is also needed to understand controls on the evolution of eruptive activity. In the past few decades, observations of volcanic eruptions have improved dramatically thanks to new satellite-based observations and high-precision geophysical instruments. High-speed visual, thermal, and ultraviolet cameras now permit measurement of key parameters (eruption velocity, mass eruption rate, particle size, and gas flux) on time scales greater than 1 Hz, appropriate for quantifying fine-scale variations in explosive activity (Taddeucci et al., 2012). Effusive activity is well characterized by this new technology, as are Strombolian and, to a lesser extent, Vulcanian explosions. More challenging is acquisition of equivalent high-resolution data sets for sustained explosive eruptions (i.e., subplinian, Plinian, and Hawaiian high-fountaining eruptions). Such events occur
less frequently and typically last only hours to days. Plinian explosive eruptions, in particular, produce large, dynamic, and optically opaque plumes. Characterizing them in real time will require rapid-response deployments and direct links to sample collection and deposit-focused studies with fine-scale temporal resolution.
One of the most difficult challenges in volcano science is to determine when an eruption is over, especially when it includes multiple episodes and long pauses (Sheldrake et al., 2016). In the simplest case (e.g., effusive eruptions), an eruption may tap a single, isolated pressurized magma chamber, eruptive activity is continuous, the mass eruption rate decreases exponentially with time (Figure 2.9), and the end of the eruption can be anticipated with some degree of accuracy (e.g., Kauahikaua et al., 1996). Often, however, eruptions tap more than one magma storage region (e.g., Tarasewicz et al., 2012), or magma is resupplied to the system between eruptions (Figure 2.10), or the system becomes “open,” so that influx balances output (Poland et al., 2014). Under these conditions, eruption terminations are currently impossible to anticipate, yet the answer is important for forecasting, especially when unrest persists long after the eruption. New insights may come, however, from emerging conceptual models of magmatic systems. In particular, by considering the broad range of scales in magmatic systems, from the crystal- and bubble-scale to the scale of magma bodies, it is possible to develop more comprehensive models for long-term patterns of eruptive behavior whereby magma reservoirs at all depths interact with each other (e.g., Christopher et al., 2015).
Ultimately, the evolution and end of a volcanic eruption may be dominated by processes acting in the shallow conduit. These processes often occur under conditions that are far from equilibrium and that are currently poorly constrained by observations, experimental data, or models. Research advances in observational data will come from new high-density monitoring networks and targeted drilling opportunities. Advances in laboratory experiments will come from real-time and in situ measurements at the high temperatures and relevant pressures of magmatic systems.
Key Questions and Research Priorities on How Eruptions Begin, Evolve, and End
Volcanic eruptions distribute lava and volcanic particles over Earth’s surface, sometimes to distances of thousands of kilometers. In this sense they are unusual among natural hazard events. Impacts range from highly localized, associated with individual lava flows
and near-vent processes, to global in scale when giant calderas form in super-eruptions. Understanding transport dynamics and dissemination of volcanic products over this extreme range of scales is necessary not only for responding to volcanic crises, but also for interpreting the record of prehistoric eruptions (preserved on land and in marine and lacustrine sediments and ice cores) and assessing their impact on Earth systems.
The fate of materials erupted both explosively and effusively is studied using several techniques, including real-time observations of active eruptions, detailed documentation of the physical and chemical properties of volcanic deposits, and physics-based modeling. An overarching goal in volcano science is to understand the links between observed or modeled dynamic phenomena and the deposits they leave behind in the geologic record: this includes plumes and their far-flung deposits, more proximal and highly destructive pyroclastic density currents, and the lava domes and flows produced by effusive eruptions.
Explosive Eruptions: Jets, Fountains, Plumes, and Drifting Clouds
Explosive subaerial eruptions form jets and plumes, consisting of volcanic particles and a mixture of volcanic and atmospheric gases. Plumes may rise buoyantly in the atmosphere, sometimes to stratospheric heights (8–17 km or higher), or collapse under their own weight to produce fountains of ejecta or hot ground-hugging pyroclastic density currents that create distal and near-source hazards, respectively (Section 1.6). These processes interact with both the natural and built environment in complex ways. For example, the otherwise cold and benign falling ash particles that are sucked into airplane engines are reheated and melted, and can create hazards to aviation as well as respiratory problems (e.g., Horwell et al., 2015) and building collapse. The extent to which the jet mixture incorporates and heats the surrounding air controls whether an eruption column rises buoyantly or collapses (Figure 2.11). Models of plume behavior can explain first-order relationships between vent conditions and plume height (e.g., Sparks, 1986; Wilson et al., 1978; Woods, 1988) and collapse thresholds (e.g., Wilson et al., 1980).
Aspects of plume behavior that are not currently well understood fall into three categories:
- The role of evolving vent conditions, including variations in eruption rate (Clarke et al., 2009; Formenti et al., 2003), overpressured jets and shock waves (Ishihara, 1985; Valentine, 1998), and vent erosion (Solovitz et al., 2014);
- Dynamics of complex plumes including the generation of pyroclastic density currents and secondary plumes from those currents (e.g., Di Muro et al., 2004; Lara, 2009; Figure 2.11) and their contributions to long-range ash transport (e.g., Eychenne et al., 2012); and
- The effect of small-scale processes, such as temporally varying grain size and density (e.g., Dufek et al., 2012), and thermal and mechanical energy exchange between gases and volcanic particles (Neri and Macedonio, 1996; Stroberg et al., 2010; Valentine and Wohletz, 1989).
Studies assessing these processes are in their infancy, yet they are critical for quantifying controls on mass partitioning under different eruption conditions.
When carried high into the atmosphere in a plume, volcanic particles are sorted by size and density, with the coarsest and/or densest particles (and aggregates of smaller particles) falling out near the vent within the first few hours of an eruption. Satellite measurements and in situ sampling of volcanic plumes reveal that gases (e.g., SO2), fine ash, and secondary aerosol particles (e.g., sulfate) may reside in the atmosphere for months to years, and can be distributed around the globe (e.g., Figure 1.1; Carn et al., 2016; Mackinnon et al., 1984; Vernier et al., 2016). Processes that can modify the depositional pattern, but are poorly understood, include ash aggregation (e.g., Brown et al., 2012; Rose and Durant, 2011), ice nucleation (e.g., Van Eaton et al., 2015), hydrometeor formation (Durant et al., 2009), development of gravitational instabilities from particle boundary layers (e.g., Carazzo et al., 2015; Manzella et al., 2015), and orographic effects (e.g., Watt et al., 2015). These processes can remove much of the fine ash prematurely from eruption columns and produce distinctive medial deposits such as a secondary increase in deposit thickness. Aggregation and fallout of aggregates of ash, water, and/or ice from eruption columns are also likely responsible for plume electrification and volcanic lightning in a “dirty thunderstorm” (e.g., Behnke
et al., 2013; McNutt and Williams, 2010; Van Eaton et al., 2016). Enhanced fine ash deposition reduces ash hazards to aviation and prevents distal ash deposition and preservation in archives such as ice cores. The sequestration of gases by particles in volcanic clouds (Durant et al., 2009) can strongly affect dissemination, residence time, and atmospheric loading of volcanic gases that affect climate (SO2) and ozone depletion (e.g., Sigmarsson et al., 2013).
Computational models include some of these small-to medium-scale processes (e.g., Oberhuber et al., 1998; Schwaiger et al., 2012; Suzuki and Koyaguchi, 2013), but models are subject to numerous simplifying assumptions (Scollo et al., 2008a,b) and not all processes are sufficiently quantified or even understood. Furthermore, most models of tephra dispersal treat fine volcanic particles and gases as passive tracers in the atmosphere such that the plume itself has no impact on atmospheric temperature and wind patterns, an assumption that may be violated in moderate to large eruptions.
Near-real-time modeling of dispersal processes would also benefit from syn-eruptive measurements of key eruption source parameters. These include eruption onset and end times, changes in the mass eruption rate over time, the total grain size distribution for a range of eruptive styles (e.g., Cashman and
Rust, 2016), the altitude and vertical distribution of gas and ash in an eruption column (e.g., Kristiansen et al., 2015; Mannen, 2014), and particle characteristics such as size distribution, shape, density, and settling velocity (e.g., Alfano et al., 2011; Beckett et al., 2015; Mastin et al., 2009b). Satellite- and ground-based measurements are crucial to determine some of these parameters, either from direct observations or derived from inverse modeling techniques (Eckhardt et al., 2008; Schneider and Hoblitt, 2013). However, future research is needed to develop methods and establish protocols for these difficult measurements. Meteorological parameters, such as wind speed and direction as a function of height and relative humidity, are also required to improve plume dispersion modeling. Research in this area would benefit from strong links to the atmospheric science community and continuous data streams.
Explosive Eruptions: Pyroclastic Density Currents
Field data show that pyroclastic density currents grade from concentrated granular flows to dilute turbulent flows (Branney and Kokelaar, 2002), and often the two occur simultaneously, with a dilute portion overriding a denser basal portion (Valentine, 1987; Figure 2.12). Field studies, scaled experiments, and numerical simulations have been combined to explain depositional and transport processes for a range of different flow regimes (e.g., Breard et al., 2016; Burgisser et al., 2005; Esposti-Ongaro et al., 2012; Roche et al., 2016; Wilson, 1980). Motivating these studies is a set
of long-standing and still open questions. How, where, and why does a flow separate into dilute and dense regimes, and what is the corresponding density stratification and mass partitioning? How do these different regimes and partitioning translate into diagnostic deposit characteristics? How should friction in the concentrated granular flow and corresponding erosional power be characterized and quantified? Which types of deposits accumulate principally by aggradation and which are emplaced by en masse stopping? Answers to these questions would inform the preparation of hazard and risk assessments, forecasting areas likely to be impacted and anticipating the consequences of pyroclastic density currents.
To answer these questions, a host of processes must be better understood, characterized, and quantified. Critical small-scale processes include sedimentation (e.g., Bursik and Woods, 1996; Charbonnier and Gertisser, 2008; Komorowski et al., 2013), resuspension (Benage et al., 2016), and particle breakup and comminution (Dufek and Manga, 2008). Such particle-scale processes can lead to order-of-magnitude variability in estimates of runout distances (Fauria et al., 2016) or can even reverse the expected direction of flows (Dufek et al., 2007). Larger-scale processes that require additional research include incorporation of air by entrainment (e.g., Andrews, 2014) and the thermal evolution of the currents (e.g., Caricchi et al., 2014), substrate interaction and erosion (e.g., Brand et al., 2014; Calder et al., 2000; Pollock et al., 2016), and interactions with topography (e.g., Andrews and Manga, 2012; Fisher et al., 1993).
Three approaches will facilitate advances: (1) documenting pyroclastic density current depositional processes in the field, (2) measuring depositional processes in the laboratory, and (3) developing numerical simulations that capture all length and time scales of pyroclastic density current processes (Figure 2.12). The hostile interiors of active pyroclastic density currents have been inaccessible to direct observation; new laboratory-, field-, and drone-based instruments would be transformational in probing these dynamic flows.
Effusive Eruptions: Lava Flows
Effusive eruptions create lava flows and domes. Our understanding of the dynamics of simple, single-lobed lava flows has advanced through a combination of detailed field studies, analog experiments, satellite observations, and numerical modeling (e.g., Harris et al., 2016). However, flows are rarely simple, and quantitative controls on whether a flow will consist of a single lobe or multiple breakout lobes are not confidently defined (Maeno et al., 2016; Figure 2.13). This complexity was highlighted by the limited ability to predict the pattern of the June 27, 2014, lava flow from Kilauea that advanced toward the town of Pahoa, Hawaii.
A number of processes that affect lava flow emplacement need to be quantified, including the rheology of crystal- and bubble-bearing lava that evolves during transport and cooling (e.g., Castruccio et al., 2013; Moitra and Gonnermann, 2015; Sehlke et al., 2014); the effect of unsteady effusion rates on the style and distance of flow propagation (Cappello et al., 2016; Favalli et al., 2009; Tarquini and de’Michieli Vitturi, 2014); the mass partitioning between advance of the flow front, breakout lobes, and inflation (e.g., Poland et al., 2014; Tuffen et al., 2013); and the interaction with a sometimes rapidly evolving topography (e.g., Dietterich and Cashman, 2014; Kubanek et al., 2015; Mattox et al., 1993).
Particularly exciting developments in the study of lava flows are new satellite and airborne remote sensing technologies, such as thermal infrared, lidar, and unmanned aerial vehicles, that can provide high-resolution and high-frequency topographic and thermal data for real lava flows (Cashman et al., 2013; James et al., 2007, 2010; Wadge et al., 2014). The ability to quantify rapidly varying effusion rates would complement measurements of flow dynamics enabled by new imaging technologies.
Historical lava flow eruptions do not exceed tens of cubic kilometers (e.g., 1783 Laki, Iceland; Thordarson and Self, 2003). The geologic record, in contrast, shows that prehistoric flood basalt eruptions have discharged thousands of cubic kilometers, with sequences of these large flows (large igneous provinces) comprising millions of cubic kilometers of lava covering hundreds of thousands of square kilometers (Coffin and Eldholm, 1994). Because we have never witnessed such events, we know little about the conditions of eruption, including both instantaneous and long-term effusion rates (e.g., Self et al., 1997), nor are the geometries of storage reservoirs well understood (e.g., Karlstrom and
Richards, 2011). Addressing our observational bias represents an important challenge, not only to improve our understanding of the dynamics of large-volume events, but also to understand their impact on Earth systems (e.g., Black et al., 2014).
Effusive Eruptions: Lava Domes
Silicic and crystal-rich lava domes are the most viscous type of effusive eruption. Emplacement dynamics have been studied extensively in both the laboratory and during several recent and well-observed eruptions (e.g., Mount St. Helens, Soufriere Hills, Merapi, Santiaguito, Chaiten, and Cordon Caulle). Laboratory experiments have demonstrated that the time scale of lava effusion relative to cooling controls dome and flow morphology, and field studies have shown that the theoretical and experimental framework transfers effectively to effusive eruptions (e.g., Buisson and Merle, 2002; Griffiths and Fink, 1997).
Although effusion rates are typically low and lava flows are typically short in length, domes can suddenly collapse or explode to form pyroclastic density currents and lateral blasts (e.g., the 1997 event at Soufriere Hills volcano, Montserrat; see Belousov et al., 2007; Hoblitt et al., 1981; Sparks and Young, 2002), or vertical eruption columns (Carn and Prata, 2010; Druitt et al., 2002; Robertson et al., 1998). A number of factors can influence collapse, including effusion rate (Calder et al., 2002; Carr et al., 2016; Nakada et al., 1999); dome volume, geometry, or strength (Loughlin et al., 2010; Simmons et al., 2005); permeability and pressurization (Fink and Kieffer, 1993; Voight and Elsworth, 2000); and rainfall (e.g., Carn et al., 2004; Elsworth et al., 2004; Matthews et al., 2002; Taron et al., 2007). However, we still cannot predict the dimensions, style, and timing of such events (e.g., Miller et al., 1998; Watts et al., 2002).
Dome-forming eruptions also tend to be long lived (years to decades) but may be episodic with lengthy pauses in eruption (e.g., Soufriere Hills volcano, Montserrat; Wadge et al., 2010; Figure 2.13). The controls on the tempo of eruption and magma supply (Section 2.2) remain poorly understood. New types of measurements promise to provide critical insights. For example, during extrusion hiatuses, measurements of gas emissions can provide constraints on continued magma supply from depth (e.g., Christopher et al., 2010). Sudden transitions from effusive to explosive activity in these long-lived eruptions remain among the most challenging characteristics to explain and forecast.
Secondary Processes: Lahars
The products of eruptions are subject to a range of secondary processes often operating on far longer time scales than the parent eruptions (Major et al., 2000). Principal among these are volcanic mudflows (lahars) and floods produced when large masses of water mix with volcanic sediment and sweep down the slopes of volcanoes, incorporating additional water and sediment (Vallance and Iverson, 2015). The effects of lahars and floods often extend well outside the primary footprint of eruptions. For example, the 1985 eruption of Nevado del Ruiz, Columbia, was relatively small (Volcano Explosivity Index [VEI] 3), but it generated a syn-eruptive lahar that was 10 times larger in volume and traveled up to 100 km, killing more than 23,000 people (Pierson et al., 1990). The VEI 6 Pinatubo eruption in 1991 was followed by a decade of devastating floods and lahars extending in space and time well beyond the pyroclastic density current deposits that spawned them (Rodolfo et al., 1996).
Lahars and floods share a number of common transport and deposition processes with pyroclastic density currents. However, the complex rheology of lahars is unusual in the range and extent of downstream flow transformations produced by the competing effects of dilution (addition of water), bulking (erosion of sediment), deposition, and infiltration of water into the substrate. No single lahar can be uniquely assigned a flow state that is applicable over its entire depth range and lifespan, yet this assumption is frequently adopted for models and hazard assessments. The timing of lahar events is largely unpredictable at present, and models for their flow do not have satisfactory equations to describe the evolution of flow density and bed erosion with time and distance (Vallance and Iverson, 2015).
Magma ascending through the crust often interacts with external water, such as groundwater, lakes, oceans, and ice. At one extreme, phreatic eruptions
occur when groundwater flashes to vapor upon contact with hot rock or magma, but no juvenile magma is erupted. Recent phreatic activity at Te Maari craters, New Zealand (Breard et al., 2014), and Ontake and Aso volcanoes, Japan (Kaneko et al., 2016; Kato et al., 2015), highlight the hazard of these events, which can be highly explosive and often occur without apparent warning. Phreatic events are often interpreted as critical precursors to magmatic eruptions, although they may also occur in isolation. Eruptions driven primarily by the explosive interaction between magma and water are termed phreatomagmatic (Morrissey et al., 2000). Such eruptions are characterized by violent explosions, volcanic plumes, ejection of large ballistic blocks, dilute pyroclastic density currents that spread radially (pyroclastic surges), and lahars (White and Houghton, 2000), and
the resulting landforms include tuff rings, tuff cones, and maars (White and Ross, 2011). Ash generated by phreatomagmatic eruptions tends to be finer grained than ash from purely magmatic explosive eruptions due to highly efficient fragmentation (Walker, 1973). As a result, ash will stay in the atmospheric longer unless counteracted by enhanced ash aggregation and premature deposition in wet eruption plumes (e.g., Brown et al., 2012). The eruptions themselves tend to be unsteady, often pulsating at high frequency, and they can be highly destructive, since thousands of pyroclastic density currents can be generated during a single eruptive episode (Brand and Clarke, 2009). Under the right conditions, magma–water mixing produces repeated explosive bursts caused by rapidly expanding water vapor along with magma quench and fragmentation, a process distinct from, although possibly aided by, purely magmatic fragmentation (Büttner et al., 1999; Zimanowski and Büttner, 2003). Quantitative advances require experimental, numerical, and field studies focused on the coupled mixing and fragmentation processes, fine ash formation, and the resulting style, scale, and duration of eruption.
Submarine eruptions represent another extreme end member of magma–water interaction. Such eruptions represent 75 to 80 percent of all magma erupted on Earth, with basaltic magma erupting at mid-ocean ridges to form the oceanic crust and at intraplate hot spots to form ocean islands and seamounts, and more silicic magmas erupting at submarine volcanic arcs. The hydrothermal systems overlying submarine volcanoes can reach very high temperatures due to the high hydrostatic pressure, and their fluids support unique chemosynthetic ecosystems (e.g., hot vents, cold seeps, mud volcanoes, and sulfidic brine pools) and concentrate valuable metal ores.
Submarine eruptions are predominantly effusive along mid-ocean ridges and produce a range of forms, including pillow basalt flows, broad thin sheets, or domes. Due to their inaccessibility, they are understudied relative to their counterparts on land. However, hydrophone networks and a new cabled observatory on the Juan de Fuca ridge (Barnes et al., 2007), have increased our capacity to detect effusive eruptions at mid-ocean ridges and to study their products using deep sea robotic and manned submersibles (e.g., Chadwick et al., 2016; Rubin et al., 2012; Soule et al., 2007; Wilcock et al., 2016). Given the simplicity of mid-ocean ridge volcanoes (e.g., known magma supply, known crustal thickness, and simple tectonic stress field), understanding how these volcanoes work may be a more tractable problem than understanding their subaerial counterparts. To do so, however, requires a better understanding of eruption sizes and frequencies.
When formed by volatile-rich subduction-zone magmas, submarine eruptions can be highly explosive (e.g., Fiske, 1963; Moore, 1967). Explosive eruptions can initiate with either magmatic and phreatomagmatic fragmentation and range from Strombolian scale to caldera forming (Cas and Giordano, 2014). Recent small-scale explosive eruptions have been observed in the western Pacific using remotely operated vehicles (NW Rota-1, Chadwick et al., 2008; West Mata, Resing et al., 2011). Although they rarely pose a direct threat to human populations, explosive subaqueous eruptions may breach the ocean surface (e.g., 1952–1953 eruption of Myojinsho, Fiske et al., 1998; and the 3 km3 eruption of Havre in 2012, Jutzeler et al., 2014), and produce large pumice rafts that can adversely affect shipping.
Two central questions about submarine explosive eruptions remain. First, how does water depth affect explosivity? Second, how do wind, ocean currents, and particle settling properties affect geochemical fluxes, pyroclast dispersal, deposit characteristics, and their postdepositional reworking? Theoretical and experimental studies on the interaction between seawater and magma, from the particle to the eruption-column scale, can address these questions about explosivity, transport, and deposition. Monitoring of submarine volcanoes, repeat high-resolution bathymetric surveys with autonomous vehicles, sampling submarine deposits with human-occupied and remotely operated vehicles, and ocean drilling would expand our understanding of the history and nature of submarine volcanism.
Key Questions and Research Priorities on What Happens When Volcanoes Erupt
A common theme in many of the questions and priorities in this chapter is the importance of developing models to interpret the new generation of high-resolution observations and to enhance understanding of magmatic and volcanic processes. Community-wide model intercomparison and validation exercises can lead to important advances and also highlight deficiencies that need to be addressed by future research. Equally useful is validating models with controlled laboratory experiments and well-constrained field data sets. Two examples in volcano science include a conduit model comparison study (Sahagian, 2005) and an intercomparison of plume models (Costa et al., 2016). Such exercises are particularly valuable when combined with suites of data from laboratory experiments, observations of the geologic record, and targeted real-world case studies.
The largest-volume explosive eruptions have yet to be characterized quantitatively. It remains uncertain how effectively, if at all, our observations of volcanic plumes and pyroclastic density currents from relatively small eruptions scale up to very large eruptions. For example, the rate and processes of radial spreading of large plumes in the atmosphere, both primary plumes and secondary plumes, may vary with the scale of the eruption (e.g., Baines and Sparks, 2005), and the roles of pulsating activity in the largest volcanic eruptions are uncertain (e.g., Self et al., 1984). Numerical models of explosive eruptions provide the means to assess the consequences of yet undocumented eruptions. The combination of modeling and observations provides the basis to overcome the biased understanding of the full spectrum of magmatic and volcanic behavior on Earth.