LARGE-SCALE SILICIC VOLCANISM-THE RESULT OF THERMAL MATURATION OF THE CRUST

Shanaka de Silva, Department of Space Studies, John D. Odegard School of Aerospace Sciences, University of North Dakota, Grand Forks, ND 58202-9008, USA George Zandt, Department of Geosciences, University of Arizona, Tucson, AZ, 85721-0077, USA. Robert Trumbull, GeoForschungZentrum Potsdam, Telegrafenberg, P.B. 4.2, 14473, Potsdam, Germany. Jose Viramonte, Instituto GEONORTE and CONICET, Universidad Nacional de Salta, Buenos Aires 177, Salta 4400, Argentina.


Introduction
Intense episodes of explosive silicic volcanism, known as ignimbrite flare-ups, characterize active continental margins in particular points of space and time. These flare-ups are represented in the geologic record by large silicic volcanic fields (LSVF) like the Sierra Madre Occidental of Mexico (Ferrari et al., 2004), the Mogollon-Datil Mountains of New Mexico, (Elston, 1984;McIntosh et al., 1992) ;San Juan Mountains of Colorado, USA (Lipman et al., 1978), and volcanic fields of the Great Basin of Utah and Nevada (Gans et al., 1989;Best et al., 1989;Sawyer et al., 1994). The main components of these are regionally extensive ignimbrite sheets, typically dacitic "monotonous intermediates" (Hildreth, 1991), and large, complex, multi-cyclic, calderas. One of the youngest and best preserved LSVF's is the Altiplano-Puna Volcanic Complex of the Central Andes (de Silva, 1989a,b). This large silicic volcanic field occupies an area of about 70,000 km 2 located between 21 and 24 o S around the intersection of political boundaries of Chile, Argentina, and Bolivia ( Figure 1). Here, an ignimbrite flare-up produced at least 30,000 km 3 of ignimbrites between 10 and 1Ma. The long-lived, nested, resurgent "calderas" and ignimbrite shields that were the eruptive sites appear to be the surface manifestation of a major intrusive complex, the remnants of which have recently been revealed by geophysical techniques (e.g. Zandt et al., 2003;Brasse et al., 2002;Götze and Krause, 2002;de Silva and Zandt, 2005; Figure 1). The most recent activity from the APVC has been a series of Quaternary (<100 ka) large volume (5 to 26 km 3 ) autonomous flow/domes that bear a strong family resemblance to the ignimbrite producing magmas (de Silva et al., 1994;Watts et al., 1999). These, in addition to the numerous geothermal sources and evidence for active surface deformation, suggest that the APVC remains active with potential for future eruptions.
In this paper we highlight the temporal development of the APVC ignimbrite flare-up. Integration of the geochronological data with petrological and geophysical data allows us to show that the ignimbrites represent the integrated history of intrusion, assimilation and melting of crust, and subsequent differentiation (e.g. Hildreth, 1981;Hildreth and Moorbath, 1988;Johnson, 1991). This history is common to many other LSVF's and suggests a consistency of process and mechanism in their development. We suggest that the fundamental control is the impact of intrusion on the mechanical strength of the crust. The prodigious volumes erupted at LSVFs imply high magma productions rates, and these in turn require high intrusion rates of parental magmas. The advection of heat into the crust must have a profound effect on the mechanical state of the thin crustal "lid" above the accumulating magma bodies, allowing the possibility that eruptions are triggered by mechanical failure of the crust. Figure1. Distribution of major ignimbrite sources of the APVC and surrounding areas keyed to age. This shows the apparent focusing of activity with time to the western part of the APVC. Also shown is the extent of the Altiplano-Puna Magma Body (APMB) constrained on the basis of seismic and gravity data. Note that most of the ignimbrite sources are located in the Lipez region of Bolivia -the least well known part of the APVC -but are coincident with the area of the APMB. Cerro Galan in Argentina is shown for reference. (Modified from de Silva and Zandt, GSA Today, forthcoming) The Altiplano-Puna Volcanic Complex The APVC forms a barren, contiguous highland with an average elevation of ~4000 m and individual volcano summits of nearly 6000 meters. It is located about 100 to 250 km above the Benioff zone of the Nazca plate at the southward transition of the ~3.8 km high Altiplano plateau and the 4.5 km high Puna, a change associated with a rapid southward thinning of the South American lithosphere without an attendant change in slab dip (Whitman et al., 1996;Allmendinger et al., 1997). Crustal seismicity levels are very low in the APVC, as they are throughout the Central Andes, due to the unusually thick and weak crust Zandt et al., 1996;Beck and Zandt, 2002). This is the result from crustal thickening between Late Oligocene to Early Miocene time in a series of eastward-migrating shortening episodes (e.g. Isacks, 1988;Allmendinger et al., 1997;McQuarrie, 2002;Horton et al, 2001;DeCelles and Horton, 2003). The main phase of shortening and thickening of the Andean crust (the Quechua phase) largely preceded the ignimbrite flare-up. The lower lithosphere beneath the Puna may have delaminated as a consequence of the crustal thickening Kay et al., 1999). This is consistent with the geophysical evidence for the thinner lithosphere and would have initiated a catastrophic thermal input into the sub-APVC crust.
The ignimbrites (and domes) of the APVC are typical of the large "monotonous intermediate" genre (c.f. Hildreth, 1991) and are crystal-rich, calc-alkaline, high-K dacites to rhyodacites, with volumetrically very minor rhyolites. The magmas contained a low-pressure assemblage with quartz, plagioclase, amphibole, biotite, and Fe-Ti oxides as the major mineral phases with occasional sanidine, and ubiquitous apatite, and titanite. These volcanics are clearly related to subduction and formation of the volcanic arc as they show trace element ratios like high Ba/Nb, Ba/La etc. that are characteristic of subduction related magmas. The abundance patterns of rare earth elements (REE) are only moderately steep as shown by the ratios Ce/Yb ~ 50, and La/Yb n ranging between 10 and 15, and there is little evidence of depletion of the heavy rare earth elements (HREE). It has long been established that Central Andean ignimbrite magmas were dominated by continental crust (e.g. Klerkx et al., 1977) but several lines of evidence indicate they are most likely 30:70 mixtures of mantle-derived magmas and melts of the crust respectively (de Silva, 1989a;Coira, et al., 1993;Ort et al., 1996). The similarity of the APVC ignimbrite isotope compositions with the range of available basement compositions (e.g. Lucassen et al., 1999; supports a dominantly crustal source. Depth of origin of the magmas is constrained between 15 and 30 km based on isotope and trace element systematics of Sr and the lack of HREE depletion ( Figure 2). Later magma evolution at between 110 and 220 MPa, or between 4 and 8 km, is required by pressure estimates using the Al-in-hornblende geobarometer, saturation pressures for H 2 O in melt inclusions, and equilibration pressures determined for the high-Si rhyolitic glass matrix in pumices (e,g, de Silva et al., 1994;Lindsay et al., 2001;Schmitt et al, 2001). Figure 2. Summary of the Sr-and Nd-isotopic characteristics of the APVC ignimbrites. It is generally accepted that these are best reconciled with a model where at least 30-40% mantle-derived magma is mixed-in with crustal melts. Lower crustal composition is from Taylor and McLennan (1985). The similarity of the APVC ignimbrite isotope compositions with the range of available basement compositions (e.g. Lucassen et al., 1999) is consistent with a dominantly crustal source. The lack of a residual garnet signature suggests that the depth of origin of the magmas has to be less than ~30km. For the Purico system, Schmitt et al. (2001) suggested that magmas originated in the depth range of 30 to 15km; the upper limit being set by assimilation and fractional crystallization calculations that require an assemblage poor in plagioclase; Sr would therefore be incompatible (D Sr <1). Note the apparent regional difference in isotopic compositions between eastern and western APVC centers. Cerro Galan (Francis et al., 1989) shown for reference. Data sources are all mentioned in the text. (From de Silva and Zandt, GSA Today, forthcoming).
The presence of a very active benioff zone beneath the APVC has allowed P-to-S (seismic wave) converted phases (receiver functions) from distant earthquakes and underlying subduction zone events, to be used to identify a very-low-velocity layer, the Altiplano Puna Magma Body (APMB), beneath the APVC (Chmielowski et al., 1999;Zandt et al., 2003;Leidig and Zandt, 2003). Modeling of the converted phases from the top and bottom of the magma body constrain the shear velocity to be ~ 1 km/s and this yields a thickness of 1-2 km and a depth of between 17 and 19 km. High surface heat flow and a broad highly conductive zone in the middle crust detected by magnetotelluric studies further support the interpretation of the APMB (Springer, 1999;Brasse et al., 2002). Schilling and Partzsch (2001) marshaled arguments from several different data sets to estimate the percentage of partial melt to be ~20 vol%, and further argued that surface heat flow constrains the melt temperature to be significantly below 1000 °C. They proposed that the melt was felsic rather than basaltic in composition. A strong negative anomaly over the APVC (Götze and Krause, 2002), provides still more evidence for the APMB and constrains the main portion of the body to be centered in southern Bolivia underneath the Cerro Guacha caldera system, with a narrowing northward extension that parallels the active volcanic arc (Figure 1). Although the resolution of the seismic and gravity data is insufficient to determine whether the APMB is a single contiguous body or several smaller bodies with small differences in depth and partially overlapping edges, we can use the area of the APMB (~45,000 km 2 ), a conservative thickness of 2 km, and the minimum melt percentage of 20% to estimate a minimum volume of melt as 18,000 km 3 . While it likely that this melt is dispersed through a mush this range of values establishes the APMB as one of the largest zones of partial melt on Earth discovered to date (de Silva and Zandt, 2005).
In summary, the petrologic-geochemical and geophysical characteristics of the APVC indicate that the volcanic field is the surface manifestation of a major episode crustal magmatisim. The ignimbrite-producing magmas were formed predominantly from a mixture of invading mafic magmas from the mantle and crustal melts at depths of between 30 and 15 km. The geophysically imaged APMB is probably the time-integrated remnant of the main zone of accumulation and storage of these magmas. From the APMB magmas subsequently rose to depths of between 4 and 8 km where they crystallized further before erupting (pumice crystallinities are typically ca. 50%).

Temporal Development of Ignimbrite Volcanism in the APVC
A compilation of available age data for the APVC shows that ignimbrite-producing volcanism in the APVC initiated ~10Ma (Figure 3) with several large (>100 km 3 ) but regionally restricted units (de Silva, 1989b;Petrinovic, 1994;Coira et al., 1996;Ort, 1993). Figure 3. Age-volume data for known major ignimbrite eruptions in the APVC. Solid rectangles are for eruptions that have been well characterized by age dating, correlation criteria, and volcanological studies. Width of rectangles represents 2σ errors. Dashed rectangle outlines are for known eruptions but where the volumes are poorly known. All volumes are deposit volumes and are regarded as minimum estimates based on known distribution as no estimate of coignimbrite ash is included. Based on measured density data, DRE volumes will be about 60 to 70 % of the volumes of the units shown. Activity continued for 10 Ma to the recent but was clearly punctuated by pulses of activity at 8, 6, and 4 Ma with eruption rates of 750 km 3 /Ma, 2000 km 3 /Ma, and 5000 km 3 /Ma respectively (de Silva and Zandt, 2005). At ~4 Ma, eruptions of La Pacana, Guacha, Chaxas calderas and the source of the Tara ignimbrite (Lindsay et al, 2001) most probably in Bolivia, represent the climactic stage of the APVC (Figure 1). The temporal trend is paralleled by a spatial trendactivity migrated from widely distributed volcanism of relatively small volume prior to about 6 Ma to much larger eruptions focused on the western part of the APVC from about 6 to 4 Ma ( Figure 1). Activity since 4 Ma has been comparatively minor (eruption rate of 200 km 3 /Ma.) with the largest eruptions being those of the Purico and Laguna Colorado centers at ~1 Ma. Three key observations arise from the available radioisotopic and volume data (Figures 1 and 3): (i) Ignimbrite eruptions pulsed with ~2 myr periodicity.
(ii) Eruptions trend toward larger volumes over time, culminating at ~ 4 Ma.
The locus of most recent activity in the APVC is focused on the western margins and to the south of the Pastos Grandes caldera complex (Figure 1). Here, a group of large late Pleistocene domes, exemplified by the Chao dacite in Chile (Guest and Sanchez, 1968;de Silva et al., 1994), and the Chascon-Runtu Jarita complex in Bolivia (Watts et al, 1999), share a strong physical and petrological resemblance to the ignimbrites and are probably "leaks" from the same magma system that fed the large ignimbrites.

Discussion
It has long been suggested that the large ignimbrite producing eruptions are the natural culmination of a long history of intrusion into the crust (e.g. Elston, 1984;Lipman, 1984) and our observations from the APVC confirm this. We find geophysical evidence of a large intrusive complex, the APMB, beneath the volcanic field. The petrological evidence suggests that the APMB is not the pre-eruptive magma chambers but the remnant of a deeper magma accumulation zone, that is probably large mush zone with little eruptible magma (de Silva and Zandt, 2005). The APMB is the integrated result of at least 10 ma of crustal processing by delamination related mantle derived melts. A catastrophic event like delamination is consistent with the sudden onset of the flare-up and the elevated thermal input from the mantle that resulted punctuated the ongoing background subduction related flux of magmas. Crustal processing in the lower to middle crust resulted in melts accumulating at their neutral buoyancy level, the APMB. A persistent flux from beneath (mantle-derived) likely maintained and grew the APMB during the ignimbrite flare-up. As magmas continued to inflate the APMB and to ascend from there to the 4 -8 km level from which they erupted, the thermal profile of the sub-APVC crust would have been considerably altered.
The ascent and emplacement of thousands of cubic kilometers of hot magma to shallow crustal levels results in advection of heat throughout the crustal column and this would have exerted a profound effect on the rheological and mechanical state of the crust. In Figure 4 we show simply that persistent intrusion into mid-and upper-crustal levels results in major changes to the thermal profile of the crustal column. Elevation of the geotherms results in a corollary change in the strength profile of the crust. In particular, the brittle-ductile transition, taken as the 450 o C isotherm, moves rapidly upward, resulting in a very thin elastic lid above the upper level chambers where all the strength of the crust is focused. Time (000's years)

Depth
Emplacement of the thousands of cubic kilometers of magmas at 8 to 4 km is accommodated by both lateral growth (slab-like chambers) and uplift (the APVC has the highest base elevations in the CVZ), which further extends the thin, hot, crustal lid. Analyses like that of Kuznir and Park (1987) show that under normal strain rates the strength of the continental lithosphere strongly depends on heat flow. So, the accumulation of the APMB and its higher-level derivatives would have caused upward compression of the thermal gradient and a highly elevated heat flow. Locally elevated brittle-ductile transitions developed at the tops of the pre-eruptive magma chambers. This process, added to the regional extensional stress associated with plateau collapse, and local uplift, resulted in failure of the crust above the shallow-level derivatives from the APMB. Foundering of the roof into the magma was likely aided by a density inversion resulting from an upper "foamy" magma zone that resulting from volatiles concentrating in the upper parts of the The increasing size of the eruptions suggests that successively larger batches of magma had to accumulate before they were erupted. Growth of larger magma bodies may have been facilitated by increased thermal softening by successive pulses of intrusion into the pre-eruption levels (8 -4 km). This may lead to the wall rocks of the growing magma body to behave in an increasingly viscous way and accommodate the growth of larger bodies later in the evolution of the system (e.g. Newman et al.,2001,;Jellinek and DePaolo, 2003) .
The model developed for the APVC may have applications in other large silicic volcanic provinces. Workers in other regions have recognized the potential role of thermal softening and failure of the crust as a trigger for ignimbrite flare-ups (Gans, et al., 1989;Best and Christiansen, 1991). Flare-ups in space and time share major characteristics that suggest a communality of origin and evolution. They involve the eruption of prodigious volumes of ignimbrite over short periods of time. They are often associated with large negative bouguer anomalies (Lipman, 1984), and involve catastrophic crustal processing in response to transient thermal events that result in high crustal magma production rates. These thermal events have been attributed to catastrophic thermal input into the crust from upwelling asthenosphere in response to delamination (Francis et al,. 1989), slab detachment (Ferrari, 2002), slab-roll back (Humphries, 1995). Best et al. (1991) showed that ignimbrite flare-ups in the western US have a sudden onset, and pulsing history. In detail it is remarkable how similar the volume-time graphs of different volcanic fields are to that of the APVC ( Figure 5).  Sawyer, et al., 1994). These are representative of data from other ignimbrites flare-ups such as the Indian Peaks volcanic field (e.g. Best et al., 1989) that show strongly episodic nature and increasing intensity of volcanism followed by rapid diminution.
The data for the Mogollon-Datil Volcanic Field of New Mexico (McIntosh et al., 1992), the Southwest Nevada Volcanic Field (Sawyer et al., 1994), and the Great Basin (Best et al., 1989) reveals the strong resemblance to the APVC and confirms that flare-ups evolve slowly but build to a catastrophic episode followed by markedly diminished activity and that eruptive activity is episodic, increasing in intensity with time. This resonance suggests that the model of increasing thermal impact of the presented for the APVC maybe applicable to other ignimbrite flare-ups.

Concluding remarks
The Altiplano Puna Volcanic Complex (APVC) of the Central Andes is one of the best preserved large silicic volcanic fields on the Earth. Spatiotemporal, petrological, and geophysical evidence reveals that the ignimbrite flare-up is the surface response to the development of a regional mid to upper-crustal intrusive complex now represented by the Altiplano-Puna Magma Body. This geophysically imaged "mush"zone is approximately 1-2 km thick with an upper surface at ~17 km. The prodiguous magma production and intrusion rates implied by the erupted volumes of the APVC and the inferred extent of the APMB require an equally profound input of heat, which was triggered by delamination of the lower lithosphere some time prior to the initiation of the ignimbrite flare-up ~10 Ma. Delamination resulted in a catastrophic increase in mantle power input (flux of basaltic magma) that punctuated the background subduction related flux. The preflare-up APVC crust had been thermally prepared (elevated geotherms) through ongoing arc magmatism and crustal thickening, thus the increased mantle power input resulted in catastrophic and widespread crustal magmatism. The impact of this transient thermal input was an intense period of generation, transport and storage of thousands of km 3 of magma in the mid crust. Advection of heat through the crustal column had a profound effect on the rheological and mechanical state of the upper crust. Ignimbrite eruptions are preceded by ascent and storage of large volumes of felsic magma at 4 to 8km in the crust. Preliminary thermal modeling shows that magma accumulation causes rapid elevation of the brittle-ductile transition with the consequence that the strength of the crust is concentrated in an ever-thinner elastic layer (lid) above the magma chamber. The massive eruptions represented by ignimbrites are triggered by mechanical failure of the roof of the magma chamber and initially driven by foundering of the roof into the magma chamber. The similarity in character of the APVC ignimbrite flare-up with other well-studied flare-ups from the western US and Mexico suggest a consistency of process and we suggest that the model presented for the APVC may be generally applicable.
London, the Geological Society of America, NASA, the NSF and the DFG. We are grateful to them all.