Magma reservoir architecture and dynamics

Santorini is one of the most explosive arc calderas in the world, having had over ten plinian eruptions, and at least four caldera collapses, over the last 350,000 years. It lies on continental crust about 25 km thick. This paper reviews what we have learned about the plumbing system of Santorini through a combination of sample characterisation, phase equilibria experimentation, melt inclusion volatile barometry, and multi-mineral diffusion chronometry. Mantle-derived basalt intrudes into the lower to middle crust, where it fractionates to andesitic, then silicic, melts in gabbroic to dioritic mush layers. Immediately prior to a plinian eruption, a large volume of evolved melt is extracted from melt lenses within these mushes and ascends to the top of the subcaldera pluton through high permeability channels. The bubbly melt (plus entrained antecrysts) then intrudes to form a large, sill-shaped reservoir that in turn feeds the plinian eruption less than a few centuries to decades later. These upper crustal reservoirs are transient holding chambers that possibly form where vertical gradients in crustal rheology cause the ascending melt to stall. During ascent from the middle to upper crust, the silicic melts grow phenocrysts by vapour-saturated decompression upon entrained, deeper-derived antecrysts. Melt extraction from depth is probably triggered by destabilisation of mush layers, due either to tectonic forces or to spontaneous breakdown of the solids framework once the melt fraction reaches a critical fraction. The smaller silicic eruptions of interplinian periods appear to be characterised by similar processes and timescales. Interplinian eruptions preceding and following plinian eruptions can tap mid-crustal mush layers different from those feeding the plinian eruptions themselves, as shown by trace element and isotopic signatures of the magmas. We favour a model in which the rheological behaviour of the subcaldera crustal column modulates magma fluxes and storage levels through the crust, and influences eruption styles.

Constraining magma reservoirs is important for understanding eruptive processes and the formation of continental crust. Geophysical imaging is a particularly powerful tool as it provides information on 3D structure, which most other methods cannot address. Geophysics however, is far from a silver bullet and has limitations arising mostly from finite resolution and non-uniqueness of geological interpretations. Recent advances in geophysical inversion and interpretation promise to revolutionize the amount of information and level of detail that they can gain from geophysical data. Michele will discuss the work that his group are doing at Imperial College to advance the state of the art in volcano tomography. The group focus on three aspects: 1) joint inversion of multiple geophysical observables (e.g. seismic and gravity data); 2) full-waveform inversion of seismic data; 3) quantitative interpretation. Seismic velocities are affected by many factors (composition, porosity, temperature, melt content), therefore interpretation of P-wave velocity models can rarely provide robust estimates on melt distribution in the crust. Joint inversion of independent geophysical datasets allows one to recover multiple parameters and start reducing the non-uniqueness of interpretations. Full-waveform inversion (FWI) exploits the information contained in the full record of natural or man-made seismic sources, resulting in 3-5 times better resolution than traditional seismic inversion techniques. FWI also allows the recovery of a wider range of elastic parameters including attenuation, anisotropy and S-wave velocities. Quantitative interpretation uses these multiple parameters to place stronger constraints on melt content and distribution and the composition of the host rock. These techniques are being used to image Santorini Volcano, Greece, using data from the NSF-funded PROTEUS experiment (Plumbing Reservoirs Of The Earth Under Santorini). The experiment was designed for high-density spatial sampling of the seismic wavefield to allow us to apply state-of-the-art 3D inversion methods, and includes 89 ocean bottom seismometers, 65 land stations and 13,300 airgun shots.

The central Andes is a key global location to study the architecture of transcrustal magmatic systems as it currently hosts the world's largest zone of silicic partial melt in the form of the Altiplano-Puna Magma or Mush Body (APMB).  Further, the APMB is dynamic, with a ground deformation pattern 150 km in diameter that has been ongoing for several decades, centered on Uturuncu volcano, Bolivia.  I will discuss results from the recently completed international PLUTONS project, which focused a suite of geophysical, petrological, and geomorphological techniques at Uturuncu with numerical modeling to infer the subsurface distribution, quantity, and movements of partial melts. We find geophysical anomalies extending from the base of the crust to the surface  indicating multiple distinct reservoirs of magma and/or hydrothermal fluids with different physical properties. The characteristics of the geophysical anomalies differ somewhat depending on the technique used - reflecting the different sensitivity of each method to subsurface melt (or fluid) of different compositions, temperature, connectivity, and volatile content, and highlighting the need for integrated, multi-disciplinary studies. We offer a new interpretation of the cause of ground deformation related to reorganization of the magma mush that causes no net ground deformation over hundred to thousands of years that is consistent with the lack of tilted shorelines around Uturuncu.

The orders of magnitude difference in viscosity between crystal-rich regions (mush) and melt-rich regions (chambers) controls the behaviour of vertically-extensive magmatic systems. High-resolution satellite geodesy provides a unique insight into the patterns of magma movement over short (weeks-decades) timescales, reflecting the geometric distribution of the mobile, eruptible component of the system. In this talk, Juliet will discuss the insights provided by 25 years of InSAR observations of deformation patterns at volcanoes around the world. To the first order, co-eruptive deflation and pre-eruptive inflation have been seen but typically the deformation is temporally variable, suggesting pulsed magma supply within viscoelastic rheology. In addition, there are numerous examples of co-eruptive uplift, typically attributed to endogenous growth of the edifice, and inter-eruptive subsidence, attributed to cooling, degassing or hydrothermal processes. The spatial pattern of deformation reveals that many systems are composed of multiple, interacting magma chambers, which can be tens of kilometers from the eruptive vent. The spatial patterns of subsurface magma chambers relates to surface alignments suggesting that pre-existing structures, such as faults, play an important role on the growth of magma reservoirs. This suggests a revised model of the development of shallow magma bodies based on growth and coalescence along fault structures.

Distinguishing time dependent magma chamber pressurization from relaxation of a viscoelastic aureole surrounding the chamber using geodetic measurements has remained challenging.  Elastic models with mass inflow proportional to the pressure difference between the chamber and a deep reservoir predict exponential re-inflation.  For a spherical chamber surrounded by a Maxwell viscoelastic shell with pressure dependent recharge, the surface deformation is the sum of two exponentials, dependent on refilling and Maxwell times.  

GPS displacements following eruptions of Grímsvötn, Iceland in 2004 and 2011 exhibit rapid post-eruptive inflation, followed by inflation with a much longer time constant.  Markov Chain Monte Carlo inversion with the viscoelastic model shows the time series can be fit with viscosity of ~ 2e16 Pa-s, and a relatively incompressible magma. The latter appears to conflict with the ratio of erupted volume to geodetically inferred source volume change of 10, obtained for the best fitting Mogi source, which implies a very compressible magma.  

Re-examination of the co-eruptive GPS and tilt data with a more general ellipsoidal model reveals that the best fitting source are oblate and deeper, consistent with seismic tomography, with much larger volume changes, relative to spherical models, consistent with incompressible melt.  This talk will also show that non-Newtonian magma rheology can, at least qualitatively, explain the data.  

The transfer of heat between a cooling magma body and its host rock induces metamorphic thermal aureoles. In the case of tabular intrusions, simple analytical models predict that the thickness of thermal aureoles scales linearly with the thickness of magma intrusions. However, some aureoles are much thicker or much thinner than predicted by analytical models. With numerical simulations, several factors affecting the size of thermal aureoles can be examined. A contrast in the rate of heat transfer between the magma and the country rock significantly impacts the maximum temperatures reached in the aureole. This contrast in heat transfer can be due to differences in conductivities or to the effect of convection, either in the magma or in the country rock. The emplacement dynamics of the magma also control the size of an aureole. Incremental emplacement, where magma injection is focused closed to one contact produces asymmetric aureoles. In comparison with a similar intrusion that is instantaneously emplaced, fast incremental emplacement induces a thicker aureole whereas slow incremental emplacement induces a thinner aureole. Inversion of aureole sizes can shed light on the associated intrusion dynamics provided that all other parameters affecting heat transfer are taken into account.

Arc magmas typically represent a complex blend of heterogeneous material from diverse sources within the volcanic plumbing system. The complexity and provenance of this material can provide a wealth of clues to understand the crustal architecture of the magmatic system, the mechanisms of magma assembly, and the degree of chemical heterogeneity within individual volcanic systems. Foreign crystals, (xenocrysts) are straightforward to identify as they display a variety of strong disequilibrium textures. Glass-bearing crystal clusters or clots are interpreted to represent larger fragments disaggregated from either plutonic equivalents of the erupted magmas, or intrusive mafic magmas. Here I show that we can identify the petrological signatures of at least two further mechanisms of accumulating material. Firstly, fine-scale textures show suppression of crystal growth in the vicinity of impinging adjacent grains. These textures indicate slow growth, with the crystals representing partially solidified plutonic material from deeper in the system. Secondly, the trace element compositions of mafic minerals such as amphibole can be inverted to reflect the composition of their equilibrium melt source. The resulting variations in elemental ratios record interaction with (or crystallisation from) melts with a wide range of trace element geochemistry despite ‘normal’ major element composition. This shows a record of anomalous trace element compositions reflecting the deeper (and older) roots of the volcanic system. This talk will illustrate this using examples from arc magmas worldwide. 

Crystals have long been used to understand conditions and processes within magma reservoirs, yet making use of the archives contained within crystals requires integrating different kinds of information derived from the same crystals and comparing them to other types of studies of magmatic processes. Challenges include relating information at the crystal- or sub-crystal spatial scale (micrometers to millimeters) to the scale of a magma reservoir (tens of meters to tens of kilometers), and relating processes occurring over years to decades to processes occurring over millennia to hundreds of thousands of years. This talk will discuss successes and discrepancies in comparing different types of crystal records and also comparing crystal records with other methods of investigating magmatic processes. For example: 1) Contrasts between radiometric (absolute) ages of crystals and time scales of diffusion of trace elements between zones within crystals, 2) Contrasts between crystallization temperatures and inferred crystal storage temperatures, and the implications for heating and cooling within a reservoir, 3) Comparisons of thermal conditions within the reservoir from numerical modeling and crystal records, and 4) Comparison of numerical modeling of mush reactivation times with the reactivation and assembly times indicated by intra-crystalline diffusion.

It is increasingly recognized that chemical differentiation in subduction zone settings occurs within crystal-dominated mushy systems that span the entire crust, rather than, discrete melt-dominated magma chambers in the traditional sense. Mushy systems involve the independent movement of buoyant melts and fluids through a deformable crystalline matrix, consistent with observations at active arc volcanoes. Percolating fluids and melts react with their solid matrix, undergoing chemical modification en route, as melts and fluids from deeper parts of the system encounter materials (solids, melts, fluids) in the shallower part of the systems. Thus, volcanoes are the final expression of processes that operate over a considerable time and depth range. Trans-crustal mush systems have implications not only for magma differentiation and crust formation, but also for hydrothermal ore deposition because they provide an effective means to focus the flow of melts and fluids sourced from a large magmatic system through a narrow conduit. The long-term stability of transcrustal mushes and their sensitivity to external or internal forces remains poorly understood. In the context of volcanic hazards a better understanding of mush dynamics is imperative for improved interpretation of monitoring data at restless volcanoes.

In this talk, a series of idealised models will be presented to explore some of the fluid mechanical controls on the formation of mushy zones, including cumulate layers, as a magma cools and crystallises. These models will be used to inform the boundary conditions for the subsequent intrusion of more primitive and relatively hot magma from deeper in the crust. The interaction of this new magma with pre-existing, relatively cool, mushy zones, will be explored, assessing the potential for remobilisation of the mushy zone and any associated mixing between or overturn of the different layers of melt. The implications of these processes on the products of a subsequent eruption will also be explored, including their implications for the composition of mafic inclusions.

Magma degassing play a significant role in the thermo-mechanical evolution of shallow magma reservoirs. In this presentation, we focus on two questions related to outgassing (1) how does the migration of an exsolved fluid phase influence the eruptibility of a heterogeneous magma body? And (2), what are the physical processes and optimal conditions that allow efficient outgassing of crystal-rich magma as they solidify to form plutons? We use a pore-scale numerical approach to study the migration of an exsolved fluid phase in crystal-rich and crystal-poor magmas and show that the fluids migrates efficiently in crystal-rich parts of a magma reservoir and accumulate in crystal-poor regions. Moreover, we parameterized our pore-scale results for fluid migration and inserted them in a thermomechanical model to study outgassing under dynamical conditions where cooling controls the evolution of the magma reservoir. We find that buoyancy-driven outgassing allows for a maximum of 40–50% volatiles to leave the reservoir over the 0.4–0.7 crystal volume fractions, implying that a significant amount of outgassing must occur at high crystal content (>0.7), through veining and/or capillary fracturing.

There is controversy about the duration, dynamics and physical conditions of magmatic systems. Conflicting geological, geochemical and geophysical evidence suggest that magmas and migmatites are either short-lived or can persist in a crystal-rich state for up to a million years. In addition the mechanisms of melt migration from magma mushes is poorly understood as the processes of porosity reduction are not uniformly consistent with the commonly invoked crystal-plastic or viscous compaction. We will first review the expressions of crystal-rich processes as preserved in an intact deep arc crustal section from the Famatinian orogeny in Argentina, and describe the progressive scales of melt extraction and migration that culminate in the formation of canonical, compositionally stratified arc crust. This will motivate us to introduce and exemplify measures of microstructural contact, force and mineral shape fabric and their anisotropy using the respective tensors. We will introduce two dimensionless groups, the viscous and Sommerfeld numbers, which can be used to describe the state of the mush and the controls on mixing and modes of mechanical lock-up. We will close with a proposal of an integrated research program at the crystal-to-outcrop scale combing discrete numerical and Paterson rig experiments, with EBSD, color CL and other measure of microstructures, to aid progress in understanding the distinct character of magmatic processes and products.