Author Query Form

Book title: (SP422) Chemical, Physical and Temporal Evolution of Magmatic SystemsAuthor: C. P. Montagna et al.Chapter: SP422-1162

This is an Online First paper. Until book publication and page numbers are confirmed, pleasecite this chapter as:

C. P. Montagna, P. Papale & A. Longo [year]. Timescales of mingling in shallow magmaticreservoirs. In: Caricchi, L. & Blundy, J. D. (eds) Chemical, Physical and Temporal Evolution ofMagmatic Systems. Geological Society, London, Special Publications, 422. First publishedonline [month] [date], [year], http://doi.org/10.1144/SP422.6

Where square bracketed text should be updated with actual dates.

The following queries have arisen during copy-editing your manuscript. Please provide ananswer in the right-hand column below and on your proofs. Many thanks for your help.

QueryNo.

QueryResponse

Q1 Please check this proof carefully for other errors because onceit is published online the only further change will be the additionof page numbers. In particular, check that figures, tables andequations are correct. Also, please check your proof to ensureyou have acknowledged your funding source (if applicable).

Q2 Please check that the wording describing the supplementarymaterial is correct and as required.

Q3 We have deleted the reference to the supplementary materialhere as per style, ok?

Q4 Figures must be listed in the correct order in the text but hereFigure 4 is listed before Figure 3, delete Figure 4 here?

Q5 Please check that your meaning here has not been changed.

Q6 We have changed ‘allows not to speculate’ to ‘removesspeculation’, is this what was meant? Please clarify.

Q7 Please provide better quality artwork for Figure 2, if available.

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758

Timescales of mingling in shallow magmatic reservoirsQ1

C. P. MONTAGNA*, P. PAPALE & A. LONGO

Istituto Nazionale di Geofisica e Vulcanologia, Sezione di Pisa via della

Faggiola 32, 56126 Pisa, Italy

*Corresponding author (e-mail: chiara.montagna@ingv.it)

Abstract: Arrival of magma from depth into shallow reservoirs has been documented as one of thepossible processes leading to eruption. Magma intruding and rising to the surface interacts with thealready emplaced, degassed magmas residing at shallower depths, leaving chemical signatures inthe erupted products.We performed two-dimensional numerical simulations of the arrival of gas-rich magmas into shallow reservoirs. We solve the fluid dynamics for the two interacting magmas,evaluating the space–time evolution of the physical properties of the mixture.

Convection and mingling develop quickly into the chamber and feeding conduit/dyke, leadingon longer timescales to a density stratification with the lighter, gas-richer magma, mixed withdifferent proportions of the resident magma, rising to the top of the chamber due to buoyancy.Over timescales of hours, the magmas in the reservoir appear to have mingled throughout, andconvective patterns become harder to identify.

Our simulations have been performed changing the geometry of the shallow reservoir andthe gas content of the initial end-member magmas. Horizontally elongated magma chambers, aswell as higher density contrasts between the two magmas, cause faster ascent velocities and alsoincrease the mixing efficiency.

Supplementary material:Q2 Videos showing the evolution of magma composition with time in theshallow chamber for simulations 1–4 are available at www.geolsoc.org.uk/SUP00000.

One of the main processes characterizing the mag-matic feeding systems at active volcanoes is theinteraction between magmas of different chemicalcompositions. A typical scenario is that of a primi-tive, volatile-rich, hot magma rising from depthand intercepting on its way upwards an alreadyestablished, volatile-poor, more evolved batch ofstalling magma (Ewart et al. 1991). Indeed, thetwo magmas typically differ not only composition-ally but also in terms of volatile content, and theirtemperatures can also be appreciably different.Such general scenario can have a variety of declina-tions, depending on the specific setting and physico-chemical characteristics of the magmatic mixturesinvolved. The shallow magma can be highly crystal-line, a mush, that can be rejuvenated by the heatfrom the incoming component (Bachmann & Ber-gantz 2008); or, at the other end, it can still be hotand more fluidal, especially if injection episodes arefrequent (Voight et al. 2010), giving rise to mixingand mingling phenomena (Longo et al. 2012b).

Depending on the timescales of the processesand on the thermal conditions of the magmaticsystem, the two magmas can interact on a purelyphysical basis, in which case the term ‘mingling’is used to describe their mechanical mixing; orthey can have a chemical exchange, by which theiroriginal compositions are modified, in which case

the proper term defining the process is ‘mixing’(e.g. Flinders & Clemens 1996).

In recent years, the importance of magma mixingand mingling processes has been recognized moreand more widely (Perugini & Poli 2012). Mixinghas been identified as one of the major processesgenerating the diversity of igneous rocks (e.g. Ber-gantz 2000), and it has often been invoked as aneruption trigger (Arienzo et al. 2010). Evidence ofmingling and mixing has been found in eruptive pro-ducts in many different settings, such as calderas(Arienzo et al. 2010; Druitt et al. 2012), basalticshield volcanoes (Sigmarsson et al. 2005) and stra-tovolcanoes (Wiesmaier et al. 2013). Previous stud-ies regarding magma mixing phenomena havemostly tackled the petrological and geochemicalissues, with the aim of identifying parent magmasusing compositional patterns (Moune et al. 2012);more recently, experimental studies have been per-formed in order to reproduce mixing patterns andcharacteristics in detail, and determine the behav-iour of the different chemical species involved(Morgavi et al. 2013). Chaotic dynamics havebeen documented both in experimental and naturalsamples (Petrelli et al. 2011), and the morphologyof mixing/mingling patterns has been linked tothe time evolution of the feeding systems (Peruginiet al. 2010).

From: Caricchi, L. & Blundy, J. D. (eds) Chemical, Physical and Temporal Evolution of Magmatic Systems.Geological Society, London, Special Publications, 422, http://doi.org/10.1144/SP422.6# 2015 The Geological Society of London. For permissions: http://www.geolsoc.org.uk/permissions.Publishing disclaimer: www.geolsoc.org.uk/pub_ethics

5960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116

This paper presents results obtained fromnumerical simulations of the arrival of primitive,gas-rich magma into a shallow reservoir, containingmore evolved, degassed magma with a differentcomposition. Patterns of convection and minglingemerge as a distinctive outcome of magma-injectionprocesses, and they are effective on timescales ofhours to days. The properties of the system, particu-larly geometry and volatile content, appreciablyinfluence the efficiency and duration of the mechan-ical mixing process.

The physical model

The magmatic system

The focus of this work is the fluid dynamics ofshallow magma chamber replenishment; specifi-cally, the quantification of the mechanical mixingprocesses that occur between magmatic mixtureswith different oxides compositions and volatilecontent. We refer, as an archetypal case, to the Phle-graean Fields magmatic system, where seismicimaging and attenuation tomographies have ident-ified a huge (probably around 10 km wide) magmareservoir at a depth of around 8 km (Zollo et al.2008; De Siena et al. 2010), while a variety of geo-physical and geochemical evidence suggests thatsmaller (probably less than 1 km3), shallowerbatches of magma have been forming throughoutthe caldera history at virtually any depth less than9 km (Arienzo et al. 2010; Di Renzo et al. 2011).These shallow magma bodies have been identifiedas actively involved in past eruptions, which, atleast in some cases, shortly followed the arrivalof volatile-rich, less differentiated magmas fromthe deep feeding system (Arienzo et al. 2009;Fourmentraux et al. 2012). Chemical compositionsof erupted magmas range from shoshonitic to tra-chytic to phonolitic; geochemical analyses on meltinclusions suggest a variety of processes contrib-uting to this variability, such as recharge fromdepth, intra-chamber mixing and syn-eruptive min-gling (Arienzo et al. 2010; Fourmentraux et al.2012). The same analyses show that deep magmasare typically rich in gas, especially CO2 (Mangiaca-pra et al. 2008).

To study the magmatic dynamics occurring as aconsequence of a recharge event, we simplify themagmatic system retaining its most peculiar fea-tures. We model the injection of CO2-rich shoshoni-tic magma coming from a deep reservoir into ashallower, much smaller chamber, containing moreevolved and partially degassed phonolitic magma;the two chambers are connected by a dyke. Deal-ing with its aperture and propagation is beyond thescope of this paper, thus we assume that the

magma of deep provenance is able to find a wayto shallower depths (e.g. Lister & Kerr 1991).

The set-up

Figure 1 shows the system domain for the numeri-cal simulations. We assume one of the horizontaldimensions of the magmatic system to be muchlarger than the other, so that our domain is two-dimensional (2D). The deep chamber is elliptical,1 km thick and 8 km wide; its top is at a depth of8 km. The geometry of the shallow chamber hasbeen varied, as shown in Figure 1, keeping its sur-face area fixed. In the elliptical cases, the semi-axesmeasure 400 and 800 m, respectively, while the cir-cular chamber has a radius of 283 m.

The initial condition of the system is also shownin Figure 1. The shallow chamber hosts a differen-tiated, volatile-poor phonolitic magma. The volatilecontent varied from 0.3 wt% CO2 and 2.5 wt% H2Oto 0.1 wt% CO2 and 1 wt% H2O. In the feeding dykeand deep reservoir is a less evolved, basaltic shosho-nite, containing 1 wt% CO2 and 2 wt% H2O. Table 1lists the oxide content of the two magmas. Volatilepartitioning between the gaseous and liquid phasesis computed following Papale et al. (2006) as a func-tion of composition and pressure. Pressure at time0 consists of a depth-dependent magmastatic contri-bution superimposed onto the host-rock confiningpressure. The interface between the two magmas,at the inlet of the shallow chamber, is gravitationallyunstable, the lower magma being less dense owingto its higher gas content. To start the dynamics, theinterface itself is not horizontal, but has a sinusoidalprofile. The dynamics is solely driven by buoyancy,without any external forcing. Density contrastsat the interface vary for each simulated scenario asthey depend both on volatile content and their parti-tioning between the liquid and gaseous phases. Vol-atile exsolution, in turn, depends on pressure, andthus on the depth at which the interface is placed,which is different for each geometry of the shallowchamber (Fig. 1). Table 2 summarizes the condi-tions for each of the simulation runs. Tempera-ture differences between interacting magmas areoften negligible (Sparks et al. 1977), particularlyat Phlegraean Fields (Mangiacapra et al. 2008),thus the system is assumed to be isothermal. As aresult, there is no need to speculate on the thermalstatus of the surrounding rock, thus reducingmodel uncertainties.

Mathematical formulation

Interaction between the two magmas develops as aconsequence of the initial gravitational instabilityat the interface. We solve numerically the 2Dspace–time evolution of the system, consisting of

C. P. MONTAGNA ET AL.

117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174

a mixture of two different magmatic components,each of them including a liquid (melt) and a gase-ous (exsolved volatiles) fractions. The equationsof motion for the mixture express conservation ofmass for each component (k ¼ 1, 2) and momentumfor the whole mixture (Longo et al. 2012a):

∂(ryk)

∂t+∇(ruyk) =−∇(rDk∇yk)

∑k

yk = 1 (1)

∂(ru)

∂t+ u∇(ru)

= −∇p +∇ m ∇u + (∇u)T − 2

3∇u

[ ]{ }+ rg.

(2)

In the above, t is time, r is the mixture density, yk

is the mass fraction of component k, u is the fluidvelocity, Dk is the k-th coefficient of mass diffusion,

p is pressure, m is viscosity and g is gravityacceleration.

The magmatic mixture is considered ideal. Itsdensity is evaluated as the weighted sum of thecomponents’ densities; for each component, den-sity is calculated using the non-ideal Lange (1994)equation of state for the liquid phase, real gas prop-erties and ideal mixture laws for multiphasefluids. Mixture viscosity (under the assumption ofNewtonian rheology) is computed through standardrules of mixing (Reid et al. 1977) for one-phasemixtures and with a semi-empirical relationship(Ishii & Zuber 1979) in order to account for theeffect of non-deformable gas bubbles. Liquid vis-cosity is modelled as in Giordano et al. (2008),and it depends on liquid composition and dissolvedwater content. The generalized Fick’s law is used todescribe mass diffusion (Taylor & Krishna 1993).Volatile partitioning between the gaseous andliquid phases is evaluated at every point in thespace–time domain as a function of the mixture

Fig. 1.

Colour

online/

colour

hardcopy

Initial conditions for the numerical simulations of magma fluid dynamics. On the left, the whole domain isshown. On the right, the upper portion of the domain, indicated by the black rectangle, is zoomed in to show the threedifferent geometries explored.

Table 1. Major oxide compositions for the two end-member magmas

SiO2 TiO2 AlO2 Fe2O3 FeO MnO MgO CaO Na2O K2O

Shoshonite 0.48 0.012 0.16 0.021 0.063 0.0014 0.10 0.12 0.028 0.015Phonolite 0.54 0.0060 0.20 0.016 0.032 0.0014 0.017 0.068 0.047 0.079

MAGMA MINGLING IN SHALLOW RESERVOIRS

175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232

composition and pressure, as in Papale et al. (2006).All of the physical properties of the two magmas areevaluated at every point in the space–time domaindepending on the local conditions of pressure, vel-ocity and mass fractions, which are the unknownsin equations (1) and (2). The equations are solvednumerically using GALES, a finite-element C++code specifically designed for volcanic fluid dynam-ics (Longo et al. 2012a).

Convection and mingling

The evolution in space and time of the system iscomplex and presents a number of interesting fea-tures. Figure 2 summarizes the results regardingmagma dynamics, showing the evolution of compo-sition with time in the shallow chamber for the fivedifferent simulation scenarios.

The initial inverse density contrast at the contactinterface between the two magmas gives rise to con-vective mass transfer from the deeper parts of thesystem to shallower depths, and vice versa. Theunstable density contrast is solely due to the differ-ent volatile content of the two mixtures: the shosho-nitic melt has, indeed, a higher density than thephonolitic. The role played by volatiles is crucial,and it is the exsolved gases that ultimately deter-mine the buoyant dynamics (Ruprecht et al. 2008).A Rayleigh–Taylor instability develops (Chandra-sekhar 2013), which acts to bring the system to grav-itational equilibrium by overturning it: a mechanismsimilar to what has been invoked as an eruptiontrigger as a consequence of chamber rejuvenation(Bain et al. 2013). The instability develops startingfrom the perturbed interface, with a first plume oflight material that rises into the chamber. Dependingon the initial density contrast, as well as on the geo-metry of the shallow chamber, the initial plumestarts developing at different times, as recorded inTable 2. The dynamics is strongly enhanced byhigher density contrasts; geometry also plays animportant role when density contrasts are similar(simulations 1–3, 4 and 5), with horizontally

elongated, sill-like chambers favouring convectionwith respect to more dyke-like set-ups (see alsoFigs 2 & 4 Q4).

Plumes of light magma coming from depth keepentering the shallow reservoir as discrete filaments,following irregular trajectories and showing typicalconvective patterns. The lighter material tends torise into the chamber, thereby decreasing more andmore its density as volatiles exsolve in the lower-pressure environments. However, the denser mag-matic mixture initially residing in the chamber sinksinto the feeder dyke, increasing its density by thereverse process of volatile dissolution at higher pres-sures. The plumes thus progressively increase theirbuoyancy, enhancing their expansion and accel-eration. During the rise, vortexes form at the headof the plumes and subsequent plumes interact withthemselves, further favouring mixing. The dyna-mics create complicated patterns that maximizethe interaction between the two different magmaticmixtures (Petrelli et al. 2011). Mingling is evidentfor all simulated conditions both within the cham-ber itself and even more in the feeding dyke, andit is strongly intensified by the chaotic patterns thatform as a consequence of deep magma injection.

Independent of system geometry or density con-trast at the interface, mingling is very efficient in thefeeding dyke, more than inside the upper chamber.Figure 2 shows that, since the very beginning ofthe simulations, the magma entering the chamber isalready a mixture of the two initial end members,and not the pure shoshonitic composition.

As the dynamics proceed, faster for higher-density contrasts and sill-like set-ups, the gas-richmixture tends to accumulate at the top of the cham-ber, thereby originating a stable density stratifica-tion that has, indeed, been demonstrated in variousmagmatic systems (Arienzo et al. 2009). The strati-fication is more prominent in vertically elongated,dyke-like reservoirs (Fig. 2). The density profilealong the vertical direction, evaluated averagingalong horizontal planes (Fig. 3), illustrates that aquasi-stable profile is reached after some hours ofsimulated time.

Table 2. List of simulations and their initial conditions; start time of the convective dynamics

Simulation CO2

shoshonite(wt%)

CO2

phonolite(wt%)

H2Oshoshonite

(wt%)

H2Ophonolite

(wt%)

Geometry of theshallow chamber

Density contrast atinitial interface

(kg m23)

Firstplume

(s)

1 1 0.3 2 2.5 Horizontal 30 102 1 0.3 2 2.5 Vertical 20 3303 1 0.3 2 2.5 Circular 25 2204 1 0.1 2 1 Horizontal 120 105 1 0.1 2 1 Vertical 100 100

C. P. MONTAGNA ET AL.

233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290

As time proceeds, convection slows down due tothe smaller buoyancy of the incoming mixed com-ponent, and the instability proceeds asymptotically

with time: the more the two end members havemingled, the less intense is convection. A time-dependent ‘convection efficiency’, hC, can be

Fig. 2.

Colour

online/

colour

hardcopy

Snapshots

Q7

of the variation in composition with time in the shallower parts of the system for the differentsimulations. Columns correspond to the different simulations (simuations 1–5); rows correspond to the different times,respectively, t ¼ 1, 3 and 5 h. Q3

2150 2200 2250 2300 2350 2400 2450−9

−8

−7

−6

−5

−4

−3

Density (kg m–3)

Dep

th (

km)

t = 0 t = 1000 s t = 7000 s t = 27950 s

Fig. 3. Total mixture density averaged over horizontal planes as a function of depth for simulation 1, at different times.

MAGMA MINGLING IN SHALLOW RESERVOIRS

291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348

defined as:

hC = |mR(t) − mR(0)|mR(0)

. (3)

It represents the relative variation of the massof the initially resident magma (mR) in a certainregion of the domain. Figures 4 and 5 show the timeevolution of convection efficiency in the shallowchamber and in the dyke. They confirm that ahigher-density contrast, as well as horizontallyelongated geometries, favour convection and mix-ing, speeding up the processes. They also highlightthat the dynamics do not start instantaneously attime 0, but at different times depending on the sys-tem conditions (Table 2). The first part of all thecurves is, indeed, flat, indicating no mass exchangebetween the different regions of the domain.

The lower parts of the system are not involved inthe system dynamics by the end of theQ5 simulatedtime (c. 6–8 h), thus we cannot comment on theevolution in those regions.

Figures 4 and 5 clearly show that mingling isvery effective over relatively short timescales, ofthe order of hours. In the more vigorous cases, whenbuoyancy drives faster dynamics, more than 40% ofthe original end-member magmas have mingled inthe feeder dyke within 4 h of the arrival of thegas-rich magma from depth. The asymptotic behav-iour seems to be reached earlier in less efficient

set-ups: after 4–5 h from the onset of the instability,the system seems to reach a quasi-steady state. Inthese cases (simulations 1–3), a much smaller pro-portion of the two magmas has mingled, around10–20%.

In the final stages, the two magmas have becomevery efficiently mixed throughout; therefore, it isharder to identify convective patterns and risingplumes as the overall dynamics reach a quasi-steadystate (Fig. 1).

Discussion

Some important simplifying assumptions haveinevitably been made to obtain the results presentedabove. Nonetheless, some general remarks can beextracted that might provide insights into genericmagmatic processes.

One of the most important issues is that onlymechanical mixing is taken into account. To thisregard, it must be pointed out that at these time-scales, chemical interactions between differentmagmas occur on length scales that are very smallcompared to the size of the system described here(Morgavi et al. 2013). This brings about a consider-able technical issue if we are to describe chemicalmixing in detail: the much smaller numerical gridsize requested would imply a much higher compu-tational effort, which is not sustainable within thescope of this work. The grid size for the simulations

0 2 4 6 80

5

10

15

20

25

Time (hours)

Con

vect

ion

effic

ienc

y (%

)

simulation 1simulation 2simulation 3simulation 4simulation 5

Fig. 4. Variation in convection efficiency with time for the different simulated scenarios in the shallow chamber.Higher density contrasts and sill-like reservoirs allow for higher convection efficiency.

C. P. MONTAGNA ET AL.

349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406

described above ranges from 20 cm to 10 m depend-ing on the dynamical behaviour of the differentdomain areas. This results in order 105 computa-tional elements, which is at the limit of our presentcomputational capabilities. Neglecting chemicalexchange between the two end members is accepta-ble at the length scales that the study is exploring indetail, which is the decameter-size scale of the over-all reservoir convective dynamics. Further studiesare planned to specifically address the chemicalinteraction occurring as mingling proceeds, to becarried out on much smaller domains. The large-scale dynamics are in very good agreement withthe experimental results obtained on similar sys-tems (Morgavi et al. 2013), providing good confi-dence in their validity.

Bi-dimensionality of the system has been as-sumed mainly to keep computational efforts afford-able. Extension to full 3D systems should not bringabout any substantial difference in the convectivepatterns observed, given the short timescale of oursimulations and the low numerical noise. The thirddimension would give rise to different dominantmodes of the Rayleigh–Taylor instability (Ribe1998), thus possibly modifying (either increasingor decreasing, depending on the amplitude of theinitial perturbation) the efficiency of transport, butnot substantially (Kaus & Podladchikov 2001).

The magmatic compositions used as end mem-bers for the simulations are quite similar (Table 1),and they both have quite low viscosities, even when

including the effect of gas bubbles. They rangefrom 500 Pa s in the deeper parts of the system,where most volatiles are dissolved, to 3500 Pa s inthe shallow chamber for simulations 4 and 5 wherethe volatile content is higher and, thus, more bubblesform in the upper region of the domain. As stated inthe earlier subsection on ‘The magmatic system’,these conditions apply to typical Phlegraean Fieldseruptions (Mangiacapra et al. 2008). Other volcanicsettings show mingling and mixing between mag-mas that have very different compositions, ofteninvolving rhyolites that are also characterized bymuch higher viscosities (Morgavi et al. 2013) andresulting in a higher viscosity contrast between thetwo magmas. Viscosity contrast has been demon-strated to play an important role in determiningmixing efficiency; nonetheless, the main featuresof the dynamical processes remain the same asdescribed here (Jellinek et al. 1999). The dynamicsproceed slightly faster and mingling is more effi-cient when the viscosity contrast is not extreme, asin the cases considered in this work. For com-positionally very different magmas, mixing is thusexpected to be slightly slower and less efficient.

The assumption of an isothermal system stemsfrom the specific case of Phlegraean Fields: as inter-acting magmas are so similar in composition, theywere also found from melt-inclusion data to havevery similar temperatures (Mangiacapra et al. 2008).Moreover, thermal exchange is expected to be verylimited in the relatively short (hours) timescales

0 2 4 6 80

10

20

30

40

50

Time (hours)

Con

vect

ion

effic

ienc

y (%

)

simulation 1simulation 2simulation 3simulation 4simulation 5

Fig. 5. Variation in convection efficiency with time for the different simulated scenarios in the dyke region.

MAGMA MINGLING IN SHALLOW RESERVOIRS

407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464

over which our simulations develop. Beyond thespecific case of Phlegraean Fields, temperaturedifferences between magmas involved in convec-tion and mixing have often been found to be smal-ler than 10% (Sparks et al. 1977). The thermalstate of the shallowly emplaced magma does playan important role (see the introduction to this paper)in determining the overall dynamics of the sys-tem. If it is much cooler (hundreds of K) than theintruding component, and thus largely crystallized,the dynamics would develop in completely differ-ent ways (Bain et al. 2013): the dyke could risethrough the crystal mush and spread out on top ofit, causing a gravitational instability as it degasses,and slowly reheating the underlying material. None-theless, such a scenario could still possibly cause,over much longer timescales (years to tens of years),buoyant mixing and overturn (Bachmann & Ber-gantz 2008; Druitt et al. 2012). The isothermalassumption reduces the computational challenges,and removes speculationQ6 on the thermal status ofthe rock surrounding the magmatic system.

As mentioned in the introduction, the issuesrelated to dyke propagation and injection are notdealt with in detail. A propagating dyke is character-ized by a relatively low (order of 106 Pa) magmaoverpressurization at the tip (Lister 1990), whichcould act as a trigger for gravitational instabilityas it intercepts a magmatic body (Eichelberger &Izbekov 2000), giving rise to the buoyant dynamicsdescribed here. Short trial simulations with anoverpressure imposed to the rising magma show nosubstantial difference in terms of the evolutionof the system, as described here. The destabilizinginterface perturbation acting as a trigger has a wave-length of 40 m, double the thickness of the dyke, andan arbitrary amplitude of 6 m. This configurationaims at reproducing the tip of a propagating dyke.Trial simulations performed with a flat interfaceshow that the system becomes destabilized num-erically, but takes much longer. As the focus ofour discussion is on how magma chamber properties(specifically, geometry and volatile content) affectthe dynamical evolution, the trigger mechanismhas been kept as simple and effective as possible.

Field and geophysical observations do not pro-vide unique constraints (Woo & Kilburn 2010) onthe possible thickness of the intruding dyke. It hasbeen set to 20 m in order to avoid issues with freez-ing (Rubin 1993). Apart from issues with closing, amuch smaller dyke (an order of magnitude) wouldhave drastically reduced the efficiency of transportowing to the highly increased viscous resistancefrom the walls, where a no-slip boundary conditionis imposed.

Notably, the results obtained in this work arein remarkable agreement with previous field andexperimental studies on magma mingling and

mixing (Petrelli et al. 2011; Morgavi et al. 2013),both in terms of the chaotic features of the convec-tive dynamics and the timescales for the minglingprocess, specifically at Phlegraean Fields (Peruginiet al. 2010). The strength of the study presentedhere is that, contrary to what can be achieved inthe field and laboratory, the large-scale character-istics of the natural system can be retained, thusextending and consolidating the value of the resultsobtained on smaller samples.

Concluding remarks

The arrival of fresh magma into an already emplacedreservoir and the consequent internal dynamicshave often been invoked as possible eruption trig-gers, especially at Phlegraean Fields (Arienzo et al.2010). The results presented in this work clearlyshow that the timescales for mechanical mixingprocesses to be effective in magmatic reservoirscan be relatively short, of the order of hours. Thisis consistent with what has been observed from theanalyses of erupted products (Fourmentraux et al.2012), as well as from experiments on diffusivefractionation (Perugini et al. 2010), and opens up acompletely new perspective in terms of eruptiontimings. Indications of mixing in erupted productssuggest that recharge events can occur within avery short time frame from eruption, otherwisethe evidence would be wiped out by the efficientmixing process.

Longo et al. (2012b) have shown that convec-tive mingling dynamics in magmatic reservoirs isassociated with ultra-long-period seismic signals,characterized by frequencies in the range 1022–1024 Hz. Given the short timescales over whichthe aforementioned processes can be effective andlead to eruption, it would be beneficial to be ableto routinely detect such signals for eruption fore-casting and mitigation actions. This is especiallytrue for long-dormant volcanoes such as PhlegraeanFields, one of the highest-risk volcanic areas in theworld given the large population living within thecaldera borders (Arienzo et al. 2010), for whichthere is still no widely accepted means of discrimi-nating the precursors of an impending eruption(Druitt et al. 2012).

This work has received funds from the European Union’sSeventh Programme for research, technological develop-ment and demonstration under grant agreement numbers282769 VUELCO and 308665 MED-SUV, and from theItalian DPC-INGV V2 project.

References

Arienzo, I., Civetta, L., Heumann, A., Woerner, G. &Orsi, G. 2009. Isotopic evidence for open system

C. P. MONTAGNA ET AL.

465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522

processes within the Campanian Ignimbrite (CampiFlegrei – Italy) magma chamber. Bulletin of Volcanol-ogy, 71, 285–300.

Arienzo, I., Moretti, R., Civetta, L., Orsi, G. &Papale, P. 2010. The feeding system of Agnano –Monte Spina eruption (Campi Flegrei, Italy): draggingthe past into present activity and future scenarios.Chemical Geology, 270, 135–147.

Bachmann, O. & Bergantz, G. W. 2008. The magmareservoirs that feed supereruptions. Elements, 4,17–21.

Bain, A. A., Jellinek, A. M. & Wiebe, R. A. 2013.Quantitative field constraints on the dynamics ofsilicic magma chamber rejuvenation and over-turn. Contributions to Mineralogy and Petrology,165, 1275–1294, http://doi.org/10.1007/s00410-013-0858-5

Bergantz, G. W. 2000. On the dynamics of magmamixing by reintrusion: implications for pluton assem-bly processes. Journal of Structural Geology, 22,1297–1309.

Chandrasekhar, S. 2013. Hydrodynamic and Hydro-magnetic Stability. Courier Dover Publications,New York.

De Siena, L., Del Pezzo, E. & Bianco, F. 2010.Seismic attenuation imaging of Campi Flegrei: evi-dence of gas reservoirs, hydrothermal basins, andfeeding systems. Journal of Geophysical Research,115, 1–18.

Di Renzo, V., Arienzo, I., Civetta, L., D’Antonio, M.,Tonarini, S., Di Vito, M. A. & Orsi, G. 2011. Themagmatic feeding system of the Campi Flegreicaldera: architecture and temporal evolution. ChemicalGeology, 281, 227–241.

Druitt, T. H., Costa, F., Deloule, E., Dungan, M. &Scaillet, B. 2012. Decadal to monthly timescales ofmagma transfer and reservoir growth at a calderavolcano. Nature, 482, 77–80.

Eichelberger, J. C. & Izbekov, P. E. 2000. Eruptionof andesite triggered by dyke injection: contrastingcases at Karymsky Volcano, Kamchatka and MtKatmai, Alaska. Philosophical Transactions of theRoyal Society, Series A, 358, 1465–1485.

Ewart, J. A., Voight, B. & Bjornsson, A. 1991. Elasticdeformation models of Krafla Volcano, Iceland, for thedecade 1975 through 1985. Bulletin of Volcanology,53, 436–459.

Flinders, J. & Clemens, J. D. 1996. Non-lineardynamics, chaos, complexity and enclaves in granitoidmagmas. Transactions of the Royal Society of Edin-burgh: Earth Sciences, 87, 217–223.

Fourmentraux, C., Metrich, N., Bertagnini, A. &Rosi, M. 2012. Crystal fractionation, magma stepascent, and syn-eruptive mingling: the Averno 2 erup-tion (Phlegraean Fields, Italy). Contributions to Miner-alogy and Petrology, 163, 1121–1137.

Giordano, D., Russell, J. K. & Dingwell, D. B. 2008.Viscosity of magmatic liquids: a model. Earth and Pla-netary Science Letters, 271, 123–134.

Ishii, M. & Zuber, N. 1979. Drag coefficient and relativevelocity in bubbly, droplet or particulate flows. AiChEJournal, 25, 843–855.

Jellinek, A. M., Kerr, R. C. & Griffiths, R. W. 1999.Mixing and compositional stratification produced

by natural convection: 1. Experiments and their appli-cation to Earth’s core and mantle. Journal of Geophy-sical Research, 104, 7183.

Kaus, B. J. P. & Podladchikov, Y. Y. 2001. Forward andreverse modeling of the threedimensional viscous Ray-leighTaylor instability. Geophysical Research Letters,28, 1095–1098.

Lange, R. A. 1994. The effect of H2O, CO2 and F on thedensity and viscosity of silicate melts. Reviews ofMineralogy, 30, 331–369.

Lister, J. R. 1990. Buoyancy-driven fluid fracture: simi-larity solutions for the horizontal and vertical pro-pagation of fluid-filled cracks. Journal of FluidMechanics, 217, 213–239.

Lister, J. R. & Kerr, R. C. 1991. Fluid-mechanicalmodels of crack propagation and their application tomagma transport in dykes. Journal of GeophysicalResearch, 96, 10 049–10 077.

Longo, A., Barsanti, M., Cassioli, A. & Papale,P. 2012a. A finite element Galerkin/least-squaresmethod for computation of multicomponent compres-sibleincompressible flows. Computers & Fluids, 67,57–71.

Longo, A., Papale, P. et al. 2012b. Magma convectionand mixing dynamics as a sourceof Ultra-Long-Periodoscillations. Bulletin of Volcanology, 74, 873–880,http://doi.org/10.1007/s00445-011-0570-0

Mangiacapra, A., Moretti, R., Rutherford, M.,Civetta, L., Orsi, G. & Papale, P. 2008. The deepmagmatic system of the Campi Flegrei caldera(Italy). Geophysical Research Letters, 35, L21304,http://doi.org/10.1029/2008GL035550

Morgavi, D., Perugini, D., De Campos, C. P.,Ertel-Ingrisch, W. & Dingwell, D. B. 2013. Mor-phochemistry of patterns produced by mixing of rhyo-litic and basaltic melts. Journal of Volcanology andGeothermal Research, 253, 87–96.

Moune, S., Sigmarsson, O., Schiano, P., Thordarson,T. & Keiding, J. K. 2012. Melt inclusion constraintson the magma source of Eyjafjallajokull 2010 flankeruption. Journal of Geophysical Research, 117,B00C07, http://doi.org/10.1029/2011 JB008718

Papale, P., Moretti, R. & Barbato, D. 2006. Thecompositional dependence of the saturation surfaceof H2O + CO2 fluids in silicate melts. ChemicalGeology, 229, 78–95.

Perugini, D. & Poli, G. 2012. The mixing of magmas inplutonic and volcanic environments: analogies anddifferences. Lithos, 153, 261–277.

Perugini, D., Poli, G., Petrelli, M., Campos, C. P. &Dingwell, D. B. 2010. Time-scales of recent Phle-grean Fields eruptions inferred from the applicationof a diffusive fractionation model of trace elements.Bulletin of Volcanology, 72, 431–447.

Petrelli, M., Perugini, D. & Poli, G. 2011. Transition tochaos and implications for time-scales of magmahybridization during mixing processes in magmachambers. Lithos, 125, 211–220.

Reid, R. C., Prausnitz, J. & Sherwood, T. 1977. TheProperties of Gases and Liquids, 3rd edn. McGrawHill, New York.

Ribe, N. M. 1998. Spouting and planform selection in theRayleigh–Taylor instability of miscible viscous fluids.Journal of Fluid Mechanics, 377, 27–45.

MAGMA MINGLING IN SHALLOW RESERVOIRS

523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580

Rubin, A. M. 1993. On the thermal viability of dikesleaving magma chambers. Geophysical ResearchLetters, 20, 257–260.

Ruprecht, P., Bergantz, G. W. & Dufek, J. 2008. Mod-eling of gas-driven magmatic overturn: tracking ofphenocryst dispersal and gathering during magmamixing. Geochemistry, Geophysics, Geosystems, 9,Q07017, http://doi.org/10.1029/2008GC002022

Sigmarsson, O., Condomines, M. & Bachelery, P. 2005.Magma residence time beneath the Piton de la FournaiseVolcano, Reunion Island, from U-series disequilibria.Earth and Planetary Science Letters, 234, 223–234.

Sparks, R., Sigurdsson, H. & Wilson, L. 1977. Magmamixing: a mechanism for triggering acid explosiveeruptions. Nature, 267, 315–318.

Taylor, R. & Krishna, R. 1993. Multicomponent MassTransfer. Wiley, New York.

Voight, B., Widiwijayanti, C., Mattioli, G. S., Els-

worth, D., Hidayat, D. & Strutt, M. 2010.Magma-sponge hypothesis and stratovolcanoes: case

for a compressible reservoir and quasi-steady deepinflux at Soufriere Hills Volcano, Montserrat. Geophy-sical Research Letters, 37, L00E05, http://doi.org/10.1029/2009GL041732

Wiesmaier, S., Troll, V. R., Wolff, J. A. & Carracedo,J. C. 2013. Open-system processes in the differen-tiation of mafic magma in the Teide-Pico Viejo suc-cession, Tenerife. Journal of the Geological Society,London, 170, 557–570, http://doi.org/10.1144/jgs2012-016

Woo, J. Y. L. & Kilburn, C. R. J. 2010. Intrusion anddeformation at Campi Flegrei, southern Italy: sills,dikes, and regional extension. Journal of Geophys-ical Research, 115, B12210, http://doi.org/10.1029/2009JB006913

Zollo, A., Maercklin, N., Vassallo, M., Dello

Iacono, D., Virieux, J. & Gasparini, P. 2008. Sei-smic reflections reveal a massive melt layer feedingCampi Flegrei caldera. Geophysical Research Letters,35, L12306, http://doi.org/10.1029/2008GL034242

C. P. MONTAGNA ET AL.