Andes Volcano Undershooting Experiment

Introduction

Goal: To understand melt generation, melt transport and their effects on the crust in continental-type arc volcanism.

Technique: Bottom-to-top (slab-to-surface) seismic images of P and S velocity, and P and S attenuation, using both direct and converted/reflected waves from regional earthquakes, supplemented by measurements of shear wave splitting, geodynamical modeling and geochemical analyses.

Arc volcanism is a poorly understood process. Many different aspects of the process by which arc volcanism occurs - and modifies the continental lithosphere - are controversial (figure 1):

What part of the mantle wedge melts? While adiabatic decompression can cause melting in ridge and hot spot environments, the cool temperatures and downward mantle flow driven by a subducting slab would seem to preclude significant upwelling in the mantle wedge beneath an arc. Geochemical evidence indicates that water-rich fluids from subducting sediments play an important role in initiating melting [Kay, 1980; Stolper and Newman, 1994]. Yet mantle upwelling is suggested by the large regional heat flow anomalies [Gill, 1981] and major-element systematics [Plank and Langmuir, 1988]. The overall pattern of this upwelling is unknown.

Does melt pond at the Moho? Basaltic melt, while less dense than the mantle, is typically denser than the continental crust [Herzberg, 1983]. Hence there may be a tendency for melt to pond near the Moho. This would provide a significant mechanism for heating the lower crust. Such ponds of magma, however, have never been directly observed.

Does melt underplate the continents? Fyfe [1992] argues that ponded magma may travel long distances (hundreds of kilometers?) along the Moho. Heat would be transferred laterally away from the arc to cause lower crustal metamorphism and secondary volcanism in distant places. There is some seismological evidence for past underplating beneath hot spots [Caress et al. 1995] and oceanic arcs [Suyehiro et al., 1996]: sub-Moho material with seismic velocities characteristic of basalt or a mix of peridotite and basalt. Deep electromagnetic sounding [Gough, 1981] also provides some evidence for deep layers of melt: regional lower-crustal high electrical resistivities. However, while some case can be made for the existence of underplating, little can be said about the particulars (i.e. thickness of melt layers, etc.).

Does the lower crust melt? The temperature of a basaltic melt is typically much higher than the melting point of granitic (or granodioritic) crustal rock, particularly if the rock contains significant water. Some melting of crustal rocks is likely to occur. Strong along-arc gradients in volcanic rock chemistry (e.g. Nd and Sr isotopic ratios) occur along some arcs (e.g. the Andes). Hildreth and Moorbath [1988] put forward the controversial Melting Assimilation Storage and Homogenization (MASH) hypothesis, which attributes these gradients to a crustal - as contrasted to mantle - source. They postulate that ponded basaltic magma causes a large (~10%) amount of remelting of the lower crust, which is assimilated into the primary mantle-derived magma.

Are intrusives or extrusives dominant in the upper crust? Calderas and other collapse features on (or near) arc volcanoes have long been understood to be indications of shallow crustal magma chambers. Because of their proximity to the surface, these magma chambers have proved relatively easy to study. Many different techniques (e.g. geodesy, seismic imaging) have been used to estimate their depth, shape and volume (e.g. Lees' [1992] study of Mt. St. Helens and Ohmi and Lees' [1995] study of Mt. Unzen). Dikes injected from the magma chamber into the surrounding rock [Brandsdottir and Einarsson, 1979], cumulate layers formed at their base, and plutons and plugs are all important in adding mass to the surrounding crust. However, the relative importance of extrusive and intrusive processes in building the arc edifice is still debated. The way in which these magma chambers are fed from below is also unknown.

Proposed Experiment

We propose an Andes Volcano Undershooting Experiment. The research that we are proposing couples new seismological and geochemical observations with geodynamical modeling of mantle flow, melting and transport. We focus on the Andes at a latitude range of 32-36 deg S (figure 2, left). The arc of volcanoes, which extends northward from southern South America, terminates abruptly in this region, with Tupungatito (33.4 deg S, 6000 m elevation), because of an abrupt change in slab dip. Hence this region contains a sort of natural "control". Our experiment, which straddles the boundary, will be able to compare structures in both the volcanic and non-volcanic parts of the Andes. Some "unexplained" Pliocene/Holocene basaltic backarc volcanism occurs in the southeastern part of the study area, which also provides some evidence for deep lateral melt transport. The goal of the experiment is to provide new data relevant to the distribution of melting in the mantle wedge, the presence of melt near the Moho (including detecting underplating and MASH scenarios), and the upper crustal plumbing system. The field work is mainly seismological, but contains a modest effort to obtain petrological samples of volcanic rocks in the latitude 33.7-34.8 deg S range (e.g. volcanoes San Jose, Maipo and Tinguiririca), which was least well-sampled in former studies and in which steep gradients in geochemical properties were found.

Seismic experiment to use regional seismicity. The study area is one in which numerous earthquakes occur, both in the subducting stab and in the crust. Furthermore, crustal seismicity is not limited to the western (trenchward) side of the arc - significant seismicity occurs up to 300 km east (landward) of the arc. Thus we can use natural seismicity to both "undershoot" the mantle wedge and crust and to probe them from the sides. We will also need to accurately locate this seismicity in order to be able to effectively use it. To this end we are planning to supplement the Chilean and Argentine national arrays with a 12 station broadband network (figure 2, right), optimized to providing precise earthquake hypocentral locations (particularly for the subcrustal earthquakes). The second key part of the seismic experiment is the multiple deployment of a 50-element linear array along a sequence of both north-south (arc parallel) and east-west (arc crossing) lines. As we will argue further below, this array will not only provide high spatial resolution data for conventional velocity and attenuation tomography, but will also allow the reliable identification of secondary converted and reflected seismic phases. Such phases, especially Moho-reflected PmP, PmS and SmS waves from crustal earthquakes and PS and SP Moho-converted phases from slab earthquakes, will allow us to effectively probe for melt near the Moho. .

A previous Andes Prototype Experiment (January 1997) was successful. In order to assess the viability of a large seismic experiment in this part of the Andes, one of us (Menke), in conjunction with Seismologist Enrique Triep of the National University of San Juan (Argentina), conducted a small, prototype experiment in the field region. We drove 3300 km of roads and found them to provide excellent access to the region, and found many sites appropriate for seismometers. We operated a single broadband seismometer (REFTEK 72A07G / Guralp CMG40T) for one day at each of seven sites, during which time we recorded 150 regional earthquakes and two teleseisms. We will discuss some of these data below. Our overall conclusion was that this region is well-suited for a seismological field experiment.

Petrological sampling targets region of strong geochemical gradients. Hildreth and Moorbath [1988] base their MASH hypothesis on the existence of strong geochemical gradients in the volcanic rocks in this part of the Andes. The 87Sr/86Sr ratio, for instance, changes from 0.7042 at 35 deg S to 0.7053 at 33.8 deg S. However, whether this change represents an along-arc gradient or a sudden jump is unclear, because of a paucity of samples in the critical 33.7-34.8 deg S interval centered about Maipo volcano. Thus we propose a modest sampling/analysis effort to provide high resolution data on geochemical gradients in this critical interval. One of the high-resolution seismic array lines crosses Maipo, thus we will be in a good position to compare seismologically-determined gradients in lower crustal and wedge structure with geochemical measurements of along-axis variation in volcanic rock chemistry.

Geodynamical models for understanding the controls on melting. Geodynamical models that include viscous flow, heat transport, melting and melt transport are relatively easy to implement using modern numerical techniques (figure 3). Such models have been widely applied - with considerable success - to ocean ridges [e.g. Spiegelman and McKenzie, 1987; Sparks et al. 1993] and hotspots [e.g. Menke and Sparks, 1995]. Unfortunately, their application to arcs has so far been limited because of the more complex nature of the problem. Melting beneath arcs depends on three poorly-understood processes: the generation of fluids within the slab and subducted sediments, the migration of that fluid into the mantle wedge, and progressive mantle melting in the presence of volatiles. Different models for the rate and depth of slab dewatering lead to strikingly different predictions of mantle flow pattern and the amount and distribution of melt [e.g. Davies and Stevenson, 1992]. Seismological constraints on the distribution of melt in the wedge will enable us to restrict the "space of possible models", and thus to begin to understand which processes (dewatering, decompression melting, etc.) are the controlling factors in the generation of primary arc magmas.

How the experiment addresses the goal

Melt in the mantle wedge can be detected through velocity and attenuation imaging. Seismic velocity decreases as temperature crosses the solidus (e.g. compressional velocity decreases by -0.14 km/s per % partial melt [Faul et al., 1994]). Large slow velocity regions (up to 4% in shear velocity) have been detected in the mantle beneath hot spots such as Iceland [Wolfe et al. 1997] and used to define the zone of mantle upwelling. We will use earthquake arrival times, together with 3D joint velocity inversion and hypocentral relocation, to define regions within which melt occurs. Seismic attenuation is also very sensitive to melt, with the quality factor of olivine-rich rock dropping from ~100 to ~10 across the solidus [Sato et al. 1989]. We will thus use spectral analysis to define path-averaged quality factors of compressional and shear waves, and then tomographically invert them for 3D variations in quality factor (using raypaths estimated during the velocity tomography). Data from The Prototype Experiment (figure 4) indicate that slab earthquakes observed at stations within the volcanic arc (at Maipo Volcano) and north of it (at Agua Negra Pass) experience vastly different shear wave attenuation.

Mantle flow directions can be detected through measurements of shear wave splitting. Mantle flow can lead to the preferential alignment of olivine crystals, thus giving the material a macroscopic anisotropy that can be detected through measurements of shear wave splitting [e.g. Vinnik et al. 1984; Silver and Chan, 1991; Zhang and Karato, 1995]. Previous studies in the Andes [Russo and Silver, 1994] have identified a broad pattern of along-axis flow in the wedge, possibly associated with large-scale transport of mantle material from beneath South America. They publish data for a single station in our study region (at 33.1 deg S, just north of the northernmost volcano) indicating an oblique flow direction of N128E. A measurement from our Prototype experiment (at 30.3 deg S, figure 5) indicates a more nearly along-axis fast direction of N162E. Hence there is some indication that regional scale variation in flow direction occurs. The high station density and numerous slab earthquakes in our experiment will allow us to map out any small-scale variations in flow direction, and use them in the geodynamical modeling.

Moho phases can be used to detect melt near that boundary. The reflection and transmission coefficients for PmP, PmS, SmS Moho reflections and PS and SP Moho conversions are very sensitive to the seismic velocity contrast across the Moho, and thus to the presence of melt near that boundary [Garmany, 1989]. The high-resolution arrays have proved a very valuable tool for detecting and identifying such secondary phases (figure 6), and measuring their amplitude-. range pattern [Menke et al., 1997]. Waveform modeling will then be used to extract estimates of boundary impedance. These phases also provide a means for mapping out crustal thickness variations, and have been successfully used by Regnier et al. [1994] to determine that the crust is about 50 km thick in the Andean foreland, deepening towards the Andes. The array deployments (figure 2, right) are chosen to be able to detect converted and reflected waves that sample Moho at a wide variety of locations. For instance, in the case of PmP (which is best observed in the 200-300 km range, figure 7), bounce points will be observed west of, beneath and east of the central Andes. The region of backarc volcanism will provide one area of special interest for these studies.

Traveltime and attenuation tomography will be able to detect melt within the crust. We have estimated the typical traveltime anomalies that might plausibly be associated with melt in the lower and upper crust (figure 8). The many different source/receiver geometries give adequate ray coverage, and the typical magnitude of the anomalies of 0.2-0.7s are easily observable. This "undershooting" technique has proven a reliable method of detecting magma in other volcanic settings [e.g. Brandsdottir et al., 1997], especially when supplemented by measurements of shear wave attenuation or shadowing (figure 9).

Tectonic stress data (from focal mechanisms) will test magma migration models. Thermal anomalies and aseismic deformation associated with magma migration can be expected to generate stress concentrations and intense seismicity in the neighboring cooler - and more brittle - material. These stresses, as well as regional tectonic stress patterns may offer ways of testing predictions on the pattern of magma migration, since the magma migration needs to be consistent with these stresses. We will compare the imaging and geodynamic modeling results with the hypocenter distributions and focal mechanisms.

Field Work

Duration. We propose a 90 day experiment, during which time we estimate that we will observe 600 usable regional earthquakes, 10-20 teleseisms, and 100 mine blasts.
Equipment: 65 REFTEK 72A07G's with Guralp CMG40-T broadband geophones, GPS clocks and solar panels, borrowed from the IRIS/PASSCAL pool. Some (up to 25) Mark Product L22D short period geophones could be substituted, should there be a shortage of CMG40-T's.
Seismicity network: 12 element network of broadband seismometers, supplemented by Argentine and Chilean national arrays.
High resolution array: 50 element roving linear array of broadband seismometers (or intermingled broadband and short period seismometers), used in eight 11-day deployments: Three crossings of the Andes, (2 deployments each, 3 km spacing), two in the volcanic arc and one north of it. Two along-axis lines, one on each side of the Andes (1 deployment each with 5 km spacing).
International participation. Scientists from both Argentina and Chile will be active participants in all aspects of the research, including the field work: Dr. Enrique Triep from the Seismological Laboratory of the National University of San Juan (Argentina) and Dr. Emillio Vera of the Geophysical Department of Chile University are the lead scientists, with other personnel and students from these institutions participating as well. Dr. Triep is an experienced mountaineer, and is intimately familiar with the terrain, as well as being well-acquainted with South American seismicity. Dr. Vera has considerable experience in the acquisition and waveform modeling of seismic refraction data. Personnel from the Tambo Mine (Compania Mineral El Indio), a source of large mining explosions, have kindly offered to supply us with shot times and locations.
Field Teams: Four field teams, each with vehicle and 3-4 personnel (staffed variously with LDEO, Argentine, and Chilean personnel) to deploy instruments. Our previous experience is that one team can deploy about 10-12 instruments per day.
Petrological Sampling: Geochemistry graduate student in charge of sampling will piggy-back on seismic field team, but will make side trips when necessary. Special effort will be made to sample small volcanic constructs as well as the three major volcanoes.

Data Analysis

Preliminary characterization of data. The data will be organized into event gathers. Some additional digital data from the Chilean and Argentina networks may also be included. Preliminary event locations and focal mechanisms will be computed using traveltime and first-motion data.
Tomographic imaging: A joint 3-d inversion for compressional and shear velocity and event location will be performed. We have access to several good tomographic inversion codes, including one developed in-house. Special effort will be made to identify statistically-significant regions of high compressional/shear wave velocity ratios (i.e. Vp/Vs > 1.9), as these can be diagnostic of partial melt. P and S wave spectral ratios will also be used to determine path-averaged quality factors, which will then be inverted for images of the three-dimensional attenuation structure. Here we rely on the use of moderate-sized (magnitude<3) events, whose corner frequencies are higher than the 1-20 Hz band in which attenuation is measured (see, for example Menke et al. [1995]).
Identification and modeling of reflections (e.g. PmP, SmS) and conversions (e.g. PS, SP). We have found that the secondary phases are relatively easy to identify on broadband record-sections, using standard signal-enhancement techniques. WKBJ synthetics will be used to model the amplitude and waveform variation of the phases.
Shear wave splitting directions will be measured for each station using slab events (which have sub-vertical ray paths) as sources using standard waveform rotation / cross-correlation techniques.
Geochemical analyses on an estimated 40 samples will include major elements (by DCP), trace elements (by ICPMS), and selected isotopes (Sr, Nd, Pb) (by TIMS).
Geodynamic modeling. One of us (Sparks) specializes in the modeling of the mantle flow and melting, and has written the necessary codes to perform the modeling.

Summary

Benefits. The goal of this research is to achieve a better understanding of continental arc volcanism. Such an understanding is important from several scientific perspectives: 1) Melting in the mantle wedge is also intimately related to volatile recycling by the subduction process, which has long term impact on mantle evolution; 2) Crustal remelting by arc volcanism may be a major factor in the evolution of the continental crust, and may be one of the mechanisms for enriching it in sialic components; 3) The melting may also enhance lower crustal deformation, and thus be a mechanism for thickening the continents during compressional events; 4) Deep lateral magma transport mechanisms (such as underplating and dike intrusion) may explain volcanism that occurs far from the actual volcanic arc. The experiment that we have proposed will gather new data targeted specifically for addressing these questions.

Ancillary Studies. We note that the array will also provide data relevant to several other interesting scientific problems: Measurements of crustal thickness, made in the course of the underplating studies, are also relevant to the question of how the crust has been thickened under the Andes; and Locations and mechanisms of slab earthquakes may also shed light on the question of the precise nature of the sudden change in slab dip (tear or bend?) that occurs north of latitude 32S.

Management Plan

Personnel:

Field Management. Menke and Seeber will lead the field effort. Both have extensive experience in managing seismic field experiments in a variety of foreign settings (e.g. Iceland, New England, Pakistan, California).
Field team will consist of Menke, Seeber, 2 LDEO graduate students, 2 LDEO Field Technicians plus several LDEO, Argentine and Chilean field assistants.
Seismic analyses: Menke will lead the seismic analysis, with Seeber, Tolstoy and a seismology GRA participating.
Geodynamic modeling. Sparks will lead the flow modeling effort.
Geochemistry. Langmuir will lead the geochemical analysis effort, with a geochemistry GRA participating.
Synthesis. All scientists and students (including Argentine and Chilean participants) will be involved in the final interpretation and analysis.

Timing:

Field effort, (3 month) during the Southern Hemisphere summer (e.g. starting December 1997 or 1998)
Analysis effort, the subsequent 9 months.

Dissemination of Results:

Delivery of data to IRIS DMC within 1 year of its collection
Meeting presentation (e.g. AGU)
Publications of results in a major professional journal (e.g. JGR)
Online "research notebook" describing the details of the experimental results. We have done this with a previous experiment, see:
http://www.ldeo.columbia.edu/~menke/b96
Discussion with Argentine and Chilean groups.

References

Brandsdottir, B and P Einarsson, Seismic activity associated with the September 1977 deflation of the Krafla central volcano in NE Iceland, J. Volcano. Geotherm Res. 6, 197-212, 1979.

Brandsdottir, B., W. Menke, P. Einarsson, R. White and R. Staples, Faeroe-Iceland Ridge Experiment 2 - Crustal structure of the Krafla central volcano, J. Geophys. Res. 102, 7867-7886, 1997.

Caress, D. MK McNutt, RS Detrick and JC Mutter, Seismic imaging of a hotspot-related crustal underplating beneath the Marquesas Islands, Nature 373, 600-603, 1995.

Davies, JH and DJ Stevenson, A physical model of the source region of subduction zone volcanics, J. Geophys. Res. 97, 2037-2070, 1992.

Faul, W., D Toomey, and H Waff, Intragranular basalt melt is distributed in thin elongated inclusions, Geophys. Res. Lett. 21, 29-32, 1994.

Fyfe, WS, Magma underplating of the continental crust, J. Volcan. Geotherm. Res. 50, 33-40, 1992.

Garmany, J., Accumulation of melt at the base of young oceanic crust, Nature 340, 628-632, 1989.

Gill, J., Orogenic andesites and plate tectonics, Springer-Verlag, 1981.

Gough, DI, Magnetometer arrays and geodynamics, in RS O'Connell and WS Fyfe (eds), Evolution of the earth, AGU geodyn. Ser. 5, 87-95, 1981.

Herzberg, CT, WS Fyfe and JJ Carr, Density constraints on the formation of the continental Moho and crust, Mineral. Petrol 84, 1-5, 1983.

Hildreth, W. and S. Moorbath, Crustal contribution to arc magmatism in the Andes of central Chile, Contrib. Mineral. Petrol. 98, 455-489, 1988.

Kay, RW, Volcanic arc magmas: implications for melting-mixing model for element recycling in the crust-mantle system; J. Geol. 88, 497-522, 1980.

Lees, J The magma system of Mt. ST. Helens: nonlinear high resolution P wave tomography, J. Volcan. Geotherm. Res. 53, 103-116, 1992.

Menke, W., V. Levin and R. Sethi, Seismic attenuation in the crust at the mid-Atlantic plate boundary in south-west Iceland, Geophys. J. Int. 122, 175-182, 1995.

Menke, W and D Sparks, Crustal acretion model for Iceland predicts 'cold' crust, Geophys. Res. Lett. 22, 1673-1676, 1995.

Menke, W, M West, B Brandsdottir and D Sparks, Along-strike seismic structure of the northern Volcanic Zone, Iceland, submitted to Geophys. J. Int., 1997.

Ohmi, S and J Lees, Three dimensional P and S wave velocity structure below Unzen volcano, J. Volcan. Geotherm. Res. 65, 1-26, 1995.

Plank, T. and CH Langmuir, An evaluation of the global variations in the major element chemistry of arc basalts, Earth Planet. Sci. Lett. 90, 349-370, 1988.

Regnier, M, and 4 others, Crustal thickness variations in the Andean foreland, from converted waves, Bull. Seism. Soc. Am 84, 1097-1111, 1994.

Russo, RM and PG Silver, Trench-parallel flow beneath the Nazca plate from seismic anisotropy, Science 263, 1105-1111, 1994.

Sato, H, S Sacks, T Murase, G Muncill and H Fukuyama, Qp-melting temperature relation in peridotite at high pressure and temperature, J. Geophys. Res. 94, 10647-10661, 1989.

Silver, P and WW Chan, Shear wave splitting an subcontinental mantle deformation, J. Geophys. Res. 96, 16429-16453, 1991.

Sparks, D, EM Parmentier and J Phipps Morgan, Three dimensional mantle convection beneath a segmented spreading center, J. Geophys. Res. 98, 21977-21995, 1993.

Spiegelman, M and D Mckenzie, Simple 2D models for melt extraction at mid-ocean ridges and island arcs, Earth Plan. Sci. Lett. 83, 137-152, 1987.

Stolper, E and Newman, S, The role of water in the petrogenesis of Mariana trough magmas, Earth Planet. Sci. Lett. 121, 291-325, 1994.

Suyehiro, K, N Takahashi, Y Ariie and seven others, Continental crust, crustal underplating and low-Q upper mantle beneath an oceanic island arc, Science 272, 390-392, 1996.

Vinnik, LP, GL Kosarev LI Makeyeva, Anizotrpiya litpsfery po nablyudeniyam voln SKS i SKKS, Docl. Akad. Nauk. SSSR 278, 1335-1339, 1984.

Wolfe, C, I Bjarnason, JC VanDecar and S Solomon, Seismic structure of the Icelandic mantle plume, Nature 385, 245-247, 1997.

Zhang, S and S Karato, Lattice preferred orientation of olivine aggregates deformed in simple shear, Nature 375, 774-777, 1995.