V12A-01 10:20h
The Viability of Three Simultaneous Melting Mechanisms at Volcanic Arcs
Since the advent of plate tectonics, three separate melting mechanisms have each been argued at times to generate the magmas for the observed volcanism at island arcs: decompression, wedge hydration, and slab melting. Despite recognized cases of anhydrous melts and somewhat ubiquitous geochemical sediment melt signatures (e.g., Th and Be) at arcs, the currently dominant view strongly favors wedge hydration with the other two mechanisms occurring only in atypical environments, such as near slab edges or in relation to arc-parallel extension. The notion of wedge hydration as the foremost melting mechanism is supported by two main lines of evidence: 1) Some arc lavas have extremely high water contents, and 2) the problem of producing the other two mechanisms in geodynamic models (flow is overall downwards and slab surface temperatures tend to be fairly cold). However, as geodynamic models become more realistic, decompression and slab melting behaviors emerge. Inclusion of temperature-dependent viscosity triggers decompression melting as the viscous lower portion of the overriding plate is ablated and replaced by hot, upwelling asthenoshpere. Temperature-dependent viscosity also results in slab surface temperatures 100-200 degrees C warmer than isoviscous models. Recent developments to allow the fault zone and overriding plate to evolve to their preferred geometry rather than remain fixed with preset dimensions can further increase slab surface temperatures. More cold, lithospheric material is sequestered in the upper plate and restricted from entrainment by the down going slab, so the slab surface can exceed the sediment solidus with this model formulation. Inclusion of shear heating produced at the fault interface will only increase the amount of sediment melting. These results suggest that volcanic arcs could possibly be simultaneously produced from all three different melting mechanisms without requiring an atypical tectonic environment.
V12A-02 INVITED 10:35h
Mantle anisotropy beneath Costa Rica and Nicaragua and the TUCAN Broadband Seismometer Array
From Nicaragua to Costa Rica, arc volcanics manifest large variations in major and trace element geochemistry, and these trends may be explained by several hypotheses, including decreases in the depth and extent of melting and the amount of slab-derived fluids. To better understand melt generation and transport processes in this subduction zone, we deployed 48 broadband IRIS/PASSCAL seismometers in four intersecting lines across and along the arc in Nicaragua and Costa Rica in July and August of 2004. The array is part of the NSF MARGINS program and will remain in place until March of 2006. Plans for analysis of array data include imaging of velocity and attenuation in the mantle wedge, subducting slab, and overlying plate with body and surface wave tomography, scattered wave migration and guided phases in combination with earthquake relocation. This presentation will focus on anisotropy in the mantle wedge constrained by shear-wave splitting in waveforms from local earthquakes, using data recorded at permanent stations and during the initial months of the TUCAN experiment. In other subduction zones, the orientation of the fast symmetry axis of anisotropy in the mantle wedge beneath the arc ranges from arc-parallel in many to arc-normal in a few. In some cases, the fast direction of anisotropy varies laterally from the arc to the back-arc, such as the arc-parallel to arc-normal trend observed in Tonga. These observations may be explained by three-dimensional solid flow in the mantle wedge, or, as suggested by recent laboratory studies, a strain field that is closer to two-dimensional corner flow but in which the development of olivine lattice-preferred orientation is influenced by higher volatile content or concentrated zones of partial melt. Our eventual goal is to interpret observed anisotropy in conjunction with velocity and attenuation patterns in order to place better constraints on flow and the distribution of melt, volatiles, and temperature within the mantle wedge.
V12A-03 10:50h
The Effect of Solid Mantle Flow Above a Subducting Plate on Melting and Fluid Migration
Arc magmas are generally thought to be derived from hydrous fluids rising buoyantly from the subducting plate causing hot overlying mantle to melt. The purpose of our study is to model the migration of water released by the dehydration of slab minerals, the subsequent melting of the mantle wedge at subduction zones and the overall distribution of water in the mantle wedge. Hydrous fluids with wetting angles smaller than 60\deg (e.g. Mibe et al, 1999) form a connected network along mineral grain edges, hence allowing for Darcy flow with grain size-melt fraction dependent permeability (Wark et al, 2003). The effect of solid flow pressure gradients on melt migration can be neglected for mantle viscosities smaller than 10$^{20}$ Pa-s. The volume of water released by the dehydration of minerals of the slab between depths of 80 and 150 km is derived from the estimates of Schmidt and Poli (1998). The volume of melt generated by the interaction of the hydrous fluids and the mantle is parameterized using the results of the program pMELTS (Ghiorso et al, 1998). Using a temperature-dependent viscosity, we explore for varying subduction rates and overlying crust thicknesses the effect of solid flow near the wedge corner on the temperature distribution of both the mantle wedge and the subducting plate. The temperature distributions thus obtained restrict in turn the region in the mantle wedge where wet melting can occur. The amount and distribution of melt in the mantle wedge is also strongly influenced by depleted mantle material. Indeed, mantle depleted by melting flows through the wedge corner and is entrained along the top of the down-going plate, hence constraining the region where wet melting can occur. Melt production is therefore dependent on mantle flow rate. The strong dependency of fluid migration on permeability requires that we explore a range of parameters, particularly grain size. Grain size in deforming solid mantle is expected to depend on stress, and mantle flow models indicate that a large range of grain sizes should be present. Small grain sizes in high stress regions near the top of the slab can enhance the amount of water carried downward into the mantle where it can be incorporated in transition zone mineral phases. Furthermore, the competition between the increasing solubility of water with increasing pressure and the buoyant upward fluid migration could control the amount of water that rises and causes wet melting region or that is entrained downward to the transition zone.
V12A-04 INVITED 11:05h
Progress towards an integrated computational model of magma genesis and transport in subduction zones
We present progress towards an integrated computational model of magma genesis and transport in subduction zones that incorporates variable viscosity mantle creep, hydrous melting and reactive porous flow of magma. These complexities will permit self-consistent prediction of a variety of geophysical and geochemical observables, including the time-scales of melt transport. There now exists an abundance of disparate observations relevant to subduction zones processes and time-scales: from heat flow and gravity profiles to measurements of radiogenic and cosmogenic nucleides, trace elements and volatiles in arc lavas. Taken together, these can provide powerful constraints on an integrated model. The petrological and computational complexity of assembling such a model is significant, however, and requires a component-wise approach. The hydrous melting parameterization that we have developed and published is the first of three basic components. The second is a functioning code to calculate dislocation/diffusion creep flow and thermal structure in subduction zones. Results from these simulations are distinct from those of isoviscous codes; a much hotter wedge corner and upwelling flow are evident. The third component is the calculation of reactive porous flow of hydrous magma through the mantle. Anhydrous reactive flow codes have demonstrated behaviour such as melt channelization with important implications for geochemical transport. Similar results are expected for hydrous reactive flow, however the inverted mantle temperature gradient present in subduction zones may play a modifying role. We will discuss each of these components and the challenges of assembing and interpreting an integrated model of subduction zones.
http://www.ldeo.columbia.edu/~katz/
V12A-05 11:20h
A New Hygrometer based on the Europium Anomaly in Clinopyroxene Phenocrysts in Arc Volcanic Rocks
Water is arguably the most important chemical component in arc magmas, affecting everything from liquidus temperatures to crystal fractionation trends to melt rheology. Water concentrations in arc magmas provide a first-order constraint on water contents in the mantle wedge, and the mechanisms of wet mantle melting. However, measuring the water content of primary arc magmas has been difficult, or in some cases impossible, due to the near complete degassing of volcanic rocks, and the scarcity of olivine-hosted melt inclusions in many arc volcanoes. We have thus developed a new hygrometer using the composition of clinopyroxene phenocrysts, which are common in arc basalts and andesites. The hygrometer is based on the well-known suppression of plagioclase by water dissolved in the melt, and the effect on the rare earth element (REE) pattern of coexisting phases, such as clinopyroxene. Dry melts saturate in plagioclase early, and the preferential partitioning of Eu2+ in plagioclase causes a negative Eu anomaly to develop in coexisting melts and clinopyroxene. In wet magmas, clinopyroxene crystallizes before plagioclase, and so initially appears with a negligible Eu anomaly. Clinopyroxenes then record water content in the delayed development of their negative Eu anomaly, caused by the delayed appearance of plagioclase along the cotectic with increasing water. We have tested this model using tephras from the 1723 eruption of Irazu, the ET3 unit of Arenal and the 1995 eruption of Cerro Negro volcanoes in Central America, with known water contents of $\sim$ 3, 4 and 5 wt%, respectively, based on olivine-hosted melt inclusions. Clinopyroxene phenocrysts separated from these samples vary in Mg\# from 86-72, and in some cases span the entire liquid line of descent of the volcano. REE patterns were determined by laser ablation ICPMS on 150 micron spots. A marked increase in the magnitude of the negative Eu anomaly occurs in clinopyroxenes with Mg\# $<$ 84 in Irazu, $<$ 82 in Arenal and $<$ 80 in Cerro Negro. These decreasing Mg\#'s correspond to decreasing liquidus temperatures at plagioclase saturation, in proportion to the known water contents of the magmas. Such systematics are broadly reproduced by predictions based on pMELTS, and experimental phase equilibria. This technique thus holds promise as a rapid method of determining the water content of virtually any arc volcanic rock, even those altered or ancient, with a resolution $<$ 0.5 wt% H2O.
V12A-06 11:35h
Estimating Temperatures and Dissolved H$_{2}$O Contents of Arc Andesite and Basalt Magmas
New experimental data from water-saturated melting experiments on Mariana - Izu arc basalts along with recent published experiments allow refinement of the temperature - H$_{2}$O content estimation technique first developed by Sisson and Grove (1993, Contrib Min Pet 133: 167-184). The model (SG93) used H$_{2}$O -saturated experimental data on basaltic liquids saturated with olivine + plagioclase +augite. SG93 estimated pre-eruptive dissolved H$_{2}$O contents and temperatures using 3 equilibria: olivine - liquid, plagioclase - liquid and an empirical expression for multiply saturated liquids. The Mariana -Izu data and other new experiments provide 32 new equilibria for a total of 149 statements of equilibrium. In addition, Moore et al. (1998, Am Min 83: 36-42) provide an expression for estimating the H$_{2}$O contents of the vapor-saturated experiments and this has been used in the model. The changes improve the model's ability to estimate temperature and H$_{2}$O contents of andesite liquids. The revised expression for SG93 for multiply saturated liquids is: T $\deg$C = 1125 - 26.7$\times$H$_{2}$O +0.089$\times$ (P-1) +272$\times$Al\# - 104.2$\times$NaK\# +213$\times$Mg\# (r$^{2}$ = 0.97 and average error is \pm 14 $\deg$C) This revised expression provides a substantial improvement by extending model coverage over a broader range of SiO$_{2}$ contents ( 49 - 63 wt. %). Along with updated olivine - melt and plagioclase - melt equilibria the new calibration can provide a useful tool for estimating pre-eruptive H$_{2}$O contents and temperatures. The hygrometer permits temperature - H$_{2}$O estimates when coexisting mineral and melt compositions are known or can be estimated. The model also allows estimation of dissolved H$_{2}$O contents when melt inclusions data are not available. Calculations using the updated SG93 indicate that basalts and andesites from the Mariana - Izu, Central America and the Aleutian arcs contain up to 5 to 6 wt. % H$_{2}$O, consistent with melt inclusion data from those areas. Determining pre-eruptive H$_{2}$O contents for arc magmas remains a challenge, but these petrologic tools provide improved quantitative estimates that are essential for understanding the flux of volatiles through the mantle wedge above subduction zones.
V12A-07 11:50h
Melt Inclusion Evidence for Both Water-Fluxed and Decompression Melting at Galunggung, Indonesia
A water flux from the slab into the mantle wedge has long been cited as the major cause of melting beneath arc volcanoes, but evidence increasingly favors pressure-release melting as contributing significantly to arc magmatism. Galunggung, Indonesia, is a classic example of an arc-front volcano where decompression melting plays an important part. We present new SIMS measurements of volatile elements (H$_{2}$O, CO$_{2}$, S, Cl, F) in olivine-hosted melt inclusions (n=60) from two high-Mg basalt bombs from the 1982-83 Galunggung eruption. These inclusions are <100 m dia. and each consists of naturally clear, brown glass with a single vapor bubble. A previous melt inclusion study using these bombs revealed H$_{2}$O-poor basaltic melts atypical of arc-front volcanoes (Sisson & Bronto, 1998), but the new data show two populations of magmatic volatile compositions at Galunggung. The dominant group comprises nearly dry (0.08-0.52 wt.% H$_{2}$O) melts similar to those reported by Sisson & Bronto (1998). The SIMS data are also the first to reveal a small group of melt inclusions in the Mg basalt (n=11) that are wet (1.4-2.5 wt.% H$_{2}$O) and more typical of pre-eruptive H$_{2}$O concentrations measured at mafic arc volcanoes. The CO$_{2}$, S, Cl and F contents, however, overlap between the wet and dry melts. CO$_{2}$ (850-0 ppm) and S (2700-70 ppm) co-vary along a trend consistent with simultaneous degassing, but Cl (1000-2000 ppm) and F (200-400 ppm) cluster together, suggesting the inclusions were trapped before Cl or F degassing initiated. Early stages of the 1982-83 Galunggung eruption yielded amphibole-bearing andesites, so the discovery of hydrous melts is not unexpected. The presence of discrete wet and dry melts at the same volcano (or within the same hand sample), however, indicates that decompression and water-fluxing both cause significant melting beneath this arc. Widespread mantle wedge upwelling may sufficiently drive pressure-release melting, but fluid infiltration is also prevalent, enhancing extents of melting and producing hydrous primary magmas. Portions of the upwelling wedge that escape fluid infiltration could yield dry melts that exploit the same shallow magma feeder systems as hydrous magmas.
V12A-08 12:05h
Subduction Dynamics and Mass Transfer: A Synthesis Model
A full understanding of arc volcanism requires a synthesis of several key aspects, including slab and wedge mineralogy, dehydration reactions, water addition to the wedge, and melt generation within the wedge. Separate models exist to explain parts of the problem that independently address, for example, the thermal regime, the phase relations as a function of pressure and temperature, or the solid viscosity and melt properties as a function of major and trace element abundances. However, it is necessary to fully combine these separate models into a self-consistent, adaptable system that allows conclusions to be drawn simultaneously about both dynamics and mass transfer. We present a model that combines the petrological model pHMELTS with a 2-D thermal and variable-viscosity flow model of a subduction zone. We are able to define the thermal state and phase equilibria of the subducting oceanic slab and constrain a fluid flux into the wedge. This allows us to describe the process of hydration of the mantle wedge adjacent to the slab, including hydrous mineral stability, water content in olivine, melt generation, and physical properties of the wedge such as viscosity and density. This is the first step leading to predictions of major element chemistry and evolution of arc volcanics, thus providing a critical connection between model and observation. We have examined two end-member examples: slab thermal ages of 25 and 100 Myr, with a convergence rate of 3 cm/yr and a slab dip of 45\deg. We adopt an iterative approach to find compositional fields consistent with a given thermal model, updating the viscosity field to reflect its dependence on water content (1), and obtaining an updated solution of the velocity and temperature fields. We assume that vertical advection of fluid is instantaneous relative to the time-scale of solid advection (3). Allowing the viscosity to be both compositionally and thermally dependent permits a consistent linkage between the effect of water-addition on the mantle wedge and the wedge velocity field, possibly leading to large-scale changes in the flow field of the hydrated wedge relative to the anhydrous estimate. The importance of combining the influence of water with iterative viscosity calculations within a single model is obvious when considering the implications of the low-viscosity wedge scenario proposed by (4). We observe similar results to previous investigations (2), including the effect of the variable viscosity model on the thinning of the isotherms immediately above the slab, producing a hotter regime immediately adjacent to the slab as compared to the isoviscous case. 1) Hirth, G. & Kohlstedt, D. L. (1996) Earth Planet. Sci. Lett., 144 (1-2), 93-108. 2) Van Keken, P.E. et al. (2002) Geochem, Geophys, Geosyst., 3 (10), 1056, doi:10.1029/2001GC000256. 3) Scott, D.R. & Stevenson, D.J. (1989) J. Geophys. Res., 94, 2973-2988. 4) Billen, M.I. & Gurnis, M. (2001) Earth Planet. Sci. Lett., 193, 227-236.