Geochronological imaging of an episodically constructed subvolcanic batholith: U-Pb in zircon chronochemistry of the Altiplano-Puna Volcanic Complex of the Central Andes

Zircons from 15 crystal-rich monotonous intermediate ignimbrites and 1 crystal-poor rhyolite ignimbrite erupted during the 11–1 Ma Altiplano-Puna Volcanic Complex (APVC) ignimbrite flare-up record multiscale episodicity in the magmatic history of the shallowest levels (5–10 km beneath the surface) of the Altiplano-Puna Magma Body (APMB). This record reveals the construction of a subvolcanic batholith and its magmatic and eruptive tempo. More than 750 U-Pb ages of zircon rims and interiors of polished grains determined by secondary ion mass spectrometry define complex age spectra for each ignimbrite with a dominant peak of autocrysts and subsidiary antecryst peaks. Xenocrysts are rare. Weighted averages obtained by pooling the youngest analytically indistinguishable zircon ages mostly correspond to the dominant crystallization ages for zircons in the magma. These magmatic ages are consistent with eruptive stratigraphy, and fall into four groups defining distinct pulses (from older to younger, pulses 1 through 4) of magmatism that correlate with eruptive pulses, but indicate that magmatic construction in each pulse initiated at least 1 m.y. before eruptions began. Magmatism was initially distributed diffusely on the eastern and western flanks of the APVC, but spread out over much of the APVC as activity waxed before focusing in the central part during the peak of the flare-up. Each pulse consists of spatially distinct but temporally sequenced subpulses of magma that represent the construction of pre-eruptive magma reservoirs. Three nested calderas were the main eruptive loci during the peak of the flare-up from ca. 6 to 2.5 Ma. These show broadly synchronous magmatic development but some discordance in their later eruptive histories. These relations are interpreted to indicate that eruptive tempo is controlled locally from the top down, while magmatic tempo is a more systemic, deeper, bottom-up feature. Synchroneity in magmatic history at distinct upper crustal magmatic foci implicates a shared connection deeper within the APMB. Each ignimbrite records the development of a discrete magma. Zircon age distributions of individual ignimbrites become more complex with time, reflecting the carryover of antecrysts in successively younger magmas and attesting to upper crustal assimilation in the APVC. Although present, xenocrysts are rare, suggesting that inheritance is limited. This is attributed to basement assimilation under zircon-undersaturated conditions deeper in the APMB than the pre-eruptive levels, where antecrysts were incorporated in zircon-saturated conditions. Magmatic ages for individual ignimbrites are older than the 40Ar/39Ar eruption ages. This difference is interpreted as the average minimum Zr-saturated melt-present lifetime for APVC magmas, the magmatic duration or Δ age. The average Δ age of ca. 0.4 Ma indicates that thermochemical conditions for zircon saturation were maintained for several hundreds of thousands of years prior to eruption of APVC magmas. This is consistent with a narrow range of zircon saturation temperatures of 730–815 °C that record upper crustal conditions and Zr/Hf, Th/U, Eu/Eu*, and Ti that reveal protracted magma differentiation under secular cooling rates an order of magnitude slower than typical pluton cooling rates. In concert, these data all suggest that the pre-eruptive magma reservoirs were perched in a thermally and chemically buffered state during their long pre-eruptive lifetimes. Trace element variations suggest subtle differences in crystallinity, melt fraction, and melt composition within different zones of individual magma reservoirs. Significant volumes of plutonic rocks associated with ignimbrites are supported by geophysical data, the limited compositional range over 10 m.y., the thermal inertia of the magmatic systems, and the evidence of resurgent magmatism and uplift at the calderas and eruptive centers, the distribution of which defines a composite, episodically constructed subvolcanic batholith. The multiscale episodicity revealed by the zircon U-Pb ages of the APVC flare-up can be interpreted in the context of continental arc magmatic systems in general. The APVC ignimbrite flare-up as a whole is a secondary pulse of ~10 m.y., with magmatic pulses 1 through 4 reflecting tertiary pulses GEOSPHERE GEOSPHERE; v. 12, no. 4 doi:10.1130/GES01258.1 11 figures; 3 tables; 4 supplemental files CORRESPONDENCE: desilvas@ geo .oregonstate .edu CITATION: Kern, J.M., de Silva, S.L., Schmitt, A.K., Kaiser, J.F., Iriarte, A.R., and Economos, R., 2016, Geochronological imaging of an episodically constructed subvolcanic batholith: U-Pb in zircon chrono chemistry of the Altiplano-Puna Volcanic Complex of the Central Andes: Geosphere, v. 12, no. 4, p. 1054–1077, doi:10.1130/GES01258.1. Received 1 September 2015 Revision received 4 April 2016 Accepted 4 May 2016 Published online 27 May 2016


INTRODUCTION
The eruptive histories of ignimbrite flare-ups are useful as a proxy for pluton formation at depth (Elston, 1984;Lipman, 2007;Best et al., 2013;Lipman and Bachmann, 2015;Christiansen et al., 2016).However, such eruptive histories provide an incomplete picture as magma could be emplaced but not erupted.Identifying the pre-eruptive history of the magmas is critical to efforts to better understand the volcanic-plutonic connection.At the forefront of these efforts are recent studies that take advantage of advancements in high-spatial-resolution U-Pb and U-Th zircon dating in the past 15 years that allow for identification of zircon crystal populations that significantly predate eruption ages (Reid et al., 1997;Brown and Fletcher, 1999;Reid and Coath, 2000;Vazquez and Reid, 2002;Schmitt et al., 2002Schmitt et al., , 2003;;Charlier et al., 2005;Bryan et al., 2007).These studies have successfully used pre-eruption zircon crystallization histories to infer the behavior of magmas prior to eruption.In a silicic magma, zircon begins to crystallize once a magma of appropriate composition cools below the zircon saturation temperature determined by the melt composition (Watson and Harrison, 1983;Boehnke et al., 2013).Because the zircon saturation temperature is often reached prior to eruption, zircons can record an extended magma history between crystallization and eruption.In tandem with eruption ages, the crystallization ages determined by U-Pb or U-Th in zircon can provide critical insight into the magmatic history of the system being studied.
Cooling ages determined with the 40 Ar-39 Ar technique are the most common data interpreted as eruption ages for volcanic systems older than a few thousand years (Spell and Harrison, 1993;Gansecki et al., 1996;Renne et al., 1997;McDougall and Harrison, 1999).U-Pb and U-Th zircon dating of young (younger than 10 Ma) silicic systems compared to 40 Ar/ 39 Ar eruption ages shows that zircon can crystallize as much as hundreds of thousands of years prior to eruption (Bachmann et al., 2007a(Bachmann et al., , 2007b;;Simon et al., 2008;Costa, 2008, Schmitt, 2011;Cooper, 2015).The difference between zircon crystallization ages and eruption ages represents the duration of zircon crystallization and therefore melt presence if zircon crystallization can be shown to be continuous until the time of eruption (Reid et al., 1997).As such U-Pb (and U-Th) zircon ages can be used to delineate the pre-eruption magmatic history as a complement to the eruptive records and in doing so provide a more complete view of magmatic history.
This study explores the zircon crystallization histories recorded in ignimbrites from the Altiplano-Puna Volcanic Complex (APVC) of the Central Andes ( de Silva, 1989a; de Silva et al., 2006a).The APVC is the result of one of the youngest ignimbrite flare-ups on Earth, erupting >15,000 km 3 of magma during multiple supereruptions from 11 to 1 Ma that define an episodic volcanic history (Fig. 1).Here we present an extensive data set of high-spatialreso lution U-Pb zircon ages that define zircon crystallization histories over the ~10 m.y.spatial and temporal span of the APVC ignimbrite flare-up and use this to complement and extend our understanding of the development of continental magmatic systems.The U-Pb data are complemented by reconnaissance zircon trace element data that targeted sample subsets representative of the diversity of composition and eruptive style in the APVC.In particular, we explore the implications of these data for the development of the plutonic system underlying large silicic volcanic fields.

GEOLOGIC BACKGROUND
The APVC in the Central Andes occupies an arid, high-elevation plateau at the political triple junction between Argentina, Chile, and Bolivia (Fig. 1A).The dry climate preserves multiple large caldera centers and their accompanying ignimbrites, providing an unparalleled view of some of the largest explosive silicic eruptions in the world.The ignimbrite plateau of the APVC developed as part of a regional ignimbrite flare-up during the Neogene development of the Central Volcanic Zone of the Andes ( de Silva et al., 2006;Salisbury et al., 2011;Freymuth et al., 2015).The APVC is the most intense and youngest locus of the central Andean flare-up, where over a 10 m.y.period multiple large ignimbrite sheets from several large caldera sources were emplaced over an area of 70,000 km 2 ( de Silva, 1989ade Silva, , 1989b)).Today, the APVC is broadly studied from physical volcanological, geochemical, and geophysical viewpoints to better understand the development of large silicic volcanic fields and the potential for future APVC eruptions.
The APVC is situated on the Altiplano-Puna plateau, which is rivaled only by the Tibetan Plateau in height and extent.The Altiplano-Puna plateau is a major element of the Central Andes, resulting from convergence of the Nazca and South American plates.Significant crustal shortening and magmatic addition over the Cenozoic produced a 70-km-thick crust and an average elevation of 4000 m, with some stratovolcanoes on the plateau reaching 6000 m (Allmendinger et al., 1997).Since the mid-1990s several geophysical studies have identified a thick zone of ultralow seismic velocities, negative gravity signature, and high conductivities with a similar areal extent to the APVC that has been called the Altiplano-Puna Magma Body (APMB; Chmielowski et al., 1999;Zandt et al., 2003;Götze and Krause, 2002).The most recent work identifies a coincidence between the APVC, a negative Bouguer anomaly of 400-500 mGal (Prezzi et al., 2009), and a 500,000 km 3 volume of anomalous seismic (shear) velocities <2.9 km/s between 9 and 31 km below the surface (Ward et al., 2014) (Fig. 1).Estimates of the volume of partial melt in this zone are >20% ( Comeau et al., 2015), and plutonic:volcanic ratios of ~20:1 have been calculated (Ward et al., 2014), supporting the contention that the APMB is the intrusive equiva-Area of -400 mGal Bouguer anomaly (Prezzi et al., 2009)  lent of the voluminous APVC volcanics (de Silva et al., 2006).This upper crustal MASH (melting, assimilation, storage, homogenization; after Hildreth and Moorbath, 1988) zone (Burns et al., 2015), represents the largest known continental crustal magmatic zone (Zandt et al., 2003;Ward et al., 2014).Continued magmatic activity in the APMB is inferred from satellite interferometry studies (Pritchard and Simons, 2002;Fialko and Pearse, 2012) that have identified a zone of surface uplift centered on the composite volcano Uturuncu with an outer ring of subsidence that extends the surface deformation to ~150 km in diameter (Fig. 1).Further correlations include the strong spatial coincidence between the APVC elevated asthenospheric depths of 60-80 km and high-heat-flow values of >100 mW/m 2 (Prezzi et al., 2009).These support the long-standing geodynamic model that major changes in the geometry of the subduction zone conspired to produce voluminous mafic melt that in turn triggered crustal melting and ignited the ignimbrite flare-up in the APVC from ca. 11 to 1 Ma (de Silva, 1989a).The onset of volcanism in the APVC is correlated in time with a large degree of steepening of the subducting Nazca plate from nearly flat-slab subduction prior to 16 Ma to an ~30° dip today (Barazangi and Isacks, 1976;Allmendinger et al., 1997).The onset of steepening (slab roll-back) is thought to have induced decompression melting in the mantle and facilitated delamination of the base of the continental lithosphere (Allmendinger et al., 1997;Kay and Coira, 2009).

GENERAL PETROLOGY AND GEOCHEMICAL BACKGROUND
The magmas erupted during the APVC ignimbrite flare-up are broadly similar in their petrological and geochemical characteristics.The ignimbrites are typical of the large monotonous intermediate genre (e.g., Hildreth, 1981) and evince that ~95% of the >15,000 km 3 of magma erupted is crystal-rich, high-K dacite to rhyodacite in bulk composition.The magmas were dominantly calc-alkaline, although ~1000 km 3 of peraluminous dacites were erupted in the eastern APVC (Ort et al., 1996;Caffe et al., 2002).Rhyolitic bulk compositions are rare.Andesitic compositions are only found as rare individual pumices in some ignimbrites or more commonly as bands and inclusions in pumices in most ignimbrites ( de Silva, 1991;Lindsay et al., 2001aLindsay et al., , 2001b;;Schmitt et al., 2001;Burns et al., 2015;Wright et al., 2011).Although these are volumetrically insignificant, they provide evidence for the role of intermediate magmas in the evolution of the dacitic magmas, representing the thermal and material input from the APMB into the pre-eruptive reservoirs (Burns et al., 2015).The mineralogy of the pumice demonstrates that the dacite magmas crystallized a low-pressure assemblage with quartz, plagioclase, amphibole, biotite, and Fe-Ti oxides as the major mineral phases.Sanidine is found in a few ignimbrites but is absent in others; when present, it is always subordinate to plagioclase.Ubiquitous accessory minerals include apatite, zircon, allanite, and titanite.Monazite is found in some heavy mineral separates.Temperatures of equilibrium range from 700 to 800 °C with f O 2 of Ni-NiO (NNO) + 1 based on Fe-Ti oxide thermometry (Lindsay et al., 2001a;Schmitt et al., 2001;Burns et al., 2015;Grocke et al., 2016).Using these temperatures, TiO 2 activities, aTiO 2 (relative to rutile saturation), of ~0.8 are estimated from the dominant trend for arc-related volcanic rocks in Ghiorso and Gualda (2013).
The ignimbrites display clear arc affinities in their geochemistry with characteristic trace element ratios like high Ba/Nb and Ba/La (de Silva et al., 2006;Kay et al., 2010;Folkes et al., 2011;Freymuth et al., 2015).Isotopic and geochemical studies constrain the magmas as a mixture of mantle-derived melts and crustal melts in crustal to mantle ratios from 1:1 to 7:3 ( de Silva, 1989a;Coira et al., 1993;Ort et al., 1996;Schmitt et al., 2001;Lindsay et al., 2001a;Schilling et al., 2006;Kay et al., 2010;Folkes et al., 2013), depending on the composition of the mantle source and contaminant chosen (see McLeod et al., 2012, for further discussion).Peraluminous compositions in the eastern APVC have the most crustal Sr and Nd isotopes ( 87 Sr/ 86 Sr > 0.712; 143 Nd/ 144 Nd ~ 0.512200) and other characteristics attesting to the influence of a metapelitic basement in the east (Caffe et al., 2012).
The petrological, geochemical, and geophysical data thus indicate a crustal magmatic system that had three distinct but linked levels that become successively more silicic upward through the crust ( de Silva et al., 2006): (1) a lower crustal MASH zone where mafic magmas pond and mix with crustal melts to produce basaltic andesite compositions (e.g., Hildreth and Moorbath, 1988) that then (2) accumulate at ~15-30 km (the APMB).Further differentiation through recharge, assimilation, and fractional crystallization of this basaltic ande site magma produces andesitic and dacitic melts that then separate and (3) accumulate and differentiate further in the upper crust (220-110 MPa;Schmitt, 2001;Muir et al., 2014aMuir et al., , 2014b;;Grocke, 2014) between 10 and 5 km to produce the crystal-rich dacites and their rhyolitic derivative magmas that eventually erupt in the caldera-forming eruptions.Early peraluminous rhyolites may have been generated through contamination of calc-alkaline intermediate compositions with metapelite at depths ~20 km within the APMB (Caffe et al., 2012).

SURFACE RECORD OF VOLCANISM IN THE APVC
The surface volcanic record of the APVC is dominated by large-volume ignimbrite eruptions composed primarily of high-K dacites and rhyodacites with minor rhyolites.Ignimbrite eruptions are sourced at large, multicyclic, nested calderas and smaller ignimbrite shields (see de Silva et al., 2006;de Silva and Gosnold, 2007).Seven eruptions from the La Pacana, Guacha, Pastos Grandes, and Vilama calderas exceed 1000 km 3 of magma, with the 2400 km 3 eruption of the Atana ignimbrite at 3.98 Ma from the La Pacana caldera being the largest yet recognized in the region.The space-time-volume record of APVC ignimbrites suggests that peaks of intense eruptive activity ca.8.4, 5.5, and 4.0 Ma are bracketed by a period of waxing activity from ca. 11 to 8.4 Ma and a waning stage from ca. 2 Ma to recent (Fig. 1; Table 1).These were organized into four clusters of eruptions or eruptive stages in de Silva et al. (2015).We detail these eruptive stages in the following as a prelude to presenting the zircon data.The waxing of the flare-up is designated eruptive stage 1.This initiated ca.11 Ma and culminated with two supereruptions ca.8.4 Ma (this combines pulses 1 and 2 of de Silva and Gosnold, 2007).The earliest eruptions at 11-9 Ma occurred at separate, distant centers along the periphery of the APVC and produced small-volume ignimbrites.These include the Artola ignimbrite (9.40 ± 0.03 Ma; all average age uncertainties reported 2s) erupted from a buried source in the San Bartolo-Rio Grande area, and the San Antonio ignimbrite (10.33 ± 0.64 Ma) that crops out sparsely near Volcan Uturuncu (Fig. 1).Other poorly exposed ignimbrites of this age range include the Lower Rio San Pedro (10.71 ± 0.14 Ma; 40 Ar/ 39 Ar age on biotite; Salisbury et al., 2011) and the Divisoco ignimbrite (10.18 ± 0.15 Ma; 40 Ar/ 39 Ar age on biotite; Salisbury et al., 2011) in the westernmost APVC (de Silva, 1989b) around Volcan San Pedro.Of similar age are the Granada ignimbrite (K-Ar age on biotite 10.1 ± 0.4 Ma; Caffe et al., 2008) and Pairique volcanics, including the 11.28 ± 0.03 Ma Coyaguayma ignimbrite ( 40 Ar/ 39 Ar age on biotite; Caffe et al., 2012).These early eruptions suggest that an extensive magmatic system was already developing beneath the APVC by at least 10 Ma ( de Silva and Gosnold, 2007;Salisbury et al., 2011).Two large-volume eruptions that deposited the 8.33 ± 0.06 Ma Sifon and the 8.41 ± 0.02 Vilama ignimbrites define the climax of this first waxing stage 1 of the APVC flare-up ( de Silva, 1989a;Soler et al., 2007;Salisbury et al., 2011).The ~1000 km 3 (dense-rock equivalent, DRE) Sifon ignimbrite erupted from a buried fissure source; a vent location beneath the Piedras Grandes and Copacoya domes on the western edge of the APVC near El Tatio is the most probable source (de Silva, 1989b).The 1400 km 3 (DRE), crystal-rich, pumice-poor Vilama ignimbrite erupted from the Vilama caldera on the Bolivia-Argentina border (Soler et al., 2007).A single exposure of an unnamed 8.35 ± 0.03 Ma rhyolite ignimbrite was found on the western flanks of Cerro Chaxas ( de Silva, 1989b; K-Ar age on biotite).The ignimbrites of this stage are the lowest in the sequence with basal contacts with Paleogene to lower Miocene and Paleozoic basement throughout the APVC ( de Silva, 1989b;Ort, 1993;Lindsay et al., 2001;Soler et al., 2007;Caffe et al., 2008Caffe et al., , 2012)).A minimum total volume of ~3000 km 3 erupted during this first stage.
Eruptive stage 2 of the APVC flare-up began ca.6.7 Ma and extended until ca.5.2 Ma (pulse 3 of de Silva and Gosnold, 2007).Early eruptions in this stage are found in the east of the APVC, where the 6.79 ± 0.02 Ma Panizos ignimbrite (600 km 3 DRE; Ort, 1993) and the Coranzuli ignimbrite (6.6 ± 0.15 Ma K-Ar age on biotite; Seggiaro, 1994; 650 km 3 deposit volume) erupted from their eponymous sources.Several relatively small (~100 km 3 ) eruptions from the vicinity of the Piedras Grandes-Loma Lucero region of Chile deposited the Toconce Formation between 6.52 ± 0.19 and 6.33 ± 0.12 Ma ( 40 Ar/ 39 Ar ages on biotite; Salisbury et al., 2011).Stage 2 culminated with >3000 km 3 of magma erupting in <250 k.y., marked by the eruptions of the >500 km 3 Pujsa (5.60 ± 0.02 Ma; K-Ar age on biotite; de Silva, 1989b), 5.65 ± 0.01 Ma Guacha (1300 km 3 ) and the 5.45 ± 0.02 Ma Chuhuilla (1200 km 3 ) ignimbrites.The Guacha ignimbrite, which covers >5800 km 2 , is sourced from the Guacha caldera in southern Bolivia near the Argentinian border.The Chuhuilla ignimbrite is the first eruption from the Pastos Grandes caldera in the northern part of the APVC (Fig. 1; Salisbury et al., 2011).The Pujsa ignimbrite is the oldest eruption known from the La Pacana caldera (Lindsay et al., 2001b).A distinct high-Si rhyolite eruption from the vicinity of the Salar de Carcote in Chile at 5.4 ± 0.4 Ma (K-Ar age in biotite; Baker and Francis, 1978) deposited the extensive but poorly studied Carcote ignimbrite.Together these eruptions define a distinct 250-km-long north-northwest-south-southeast trend (335°) from La Pacana to the Salar de Carcote.The final eruption from this stage is the >10 km 3 Alota ignimbrite (5.23 ± 0.01 Ma; 40 Ar/ 39 Ar on sanidine) that forms an ignimbrite shield capped by the Juvina dome on the northern flank of the Pastos Grandes caldera.At least 4000 km 3 of magma erupted during this second stage.
Stage 3 of the flare-up extends from 4.09 to 2.89 Ma (pulse 4 of de Silva and Gosnold, 2007).The period initiated with ~3500 km 3 of magma erupted in ~100 k.y.At 4.09 ± 0.02 Ma the 800 km 3 Puripicar ignimbrite erupted from the area of the Chaxas dome complex, while soon after the massive 2700 km 3 Atana-Toconao eruption at 3.96 ± 0.02 and 4.0 ± 0.01 Ma, respectively, followed to form the La Pacana caldera (Lindsay et al., 2001b;Salisbury et al., 2011).At 3.49 ± 0.01 Ma the 800 km 3 Tara ignimbrite erupted from the Guacha caldera and terminated an intense period of eruptions from sources within 50 km of each other.The final major eruption of stage 3 is the 2.89 ± 0.01 Ma eruption of the Pastos Grandes caldera that deposited 1200 km 3 of magma as the Pastos Grandes ignimbrite.Note that each of the Atana, Tara, and Pastos Grandes  2011) and de Silva and Gosnold (2007), respectively.Volumes are reported as dense rock (magma) equivalent.† Ar-Ar data in Salisbury et al. (2011) were calculated relative Taylor Creek rhyolite sanidine at 28.34 Ma, which is equivalent to Fish Canyon Sanidine at 28.02 Ma (Renne et al., 1998).
eruptions are the second eruptions from the La Pacana, Guacha, and Pastos Grandes calderas, respectively, that all erupted in stage 2.
The waning of the ignimbrite flare-up is marked by a significant reduction in eruptive volumes in the past 2.5 m.y.; we designate this as eruptive stage 4. The most extensive of these young ignimbrites are the largely unstudied Patao ignimbrite (2.52 ± 0.06 Ma; 40 Ar/ 39 Ar on plagioclase; Barquero-Molina, 2003) and the spatially and temporally proximal Talabre-Pampa Chamaca eruption of southern La Pacana at 2.42 ± 0.06 Ma ( 40 Ar/ 39 Ar on plagioclase; Barquero-Molina, 2003).Collectively these cover significant areas in the southernmost, least well studied area of the APVC.Our reconnaissance mapping and the work of Gardeweg and Ramírez (1987) suggest several hundreds of cubic kilome ters of ignimbrite extending beneath the Punta Negra volcanic complex and into the Socaire area.Several small to intermediate volume ignimbrites erupted from sources in the center of the APVC, including the Laguna Colorada (60 km 3 ), Puripica Chico (10 km 3 ), and Purico (100 km 3 ) ignimbrites (Salisbury et al., 2011).The Laguna Colorada ignimbrite forms the radiating flanks of the Laguna Colorada ignimbrite shield, erupted between the Pastos Grandes and Guacha calderas.Puripica Chico is generally associated with the Guacha caldera, although its source is a small lava dome on the western edge of Guacha (Salisbury et al., 2011).The Purico ignimbrite erupted from the Cerro Purico ignimbrite shield complex near the Chile-Bolivia border that also includes a compositionally identical lava dome (dome D; Schmitt et al., 2001).The ~50 km 3 Tatio ignimbrite (0.7 ± 0.01 Ma; 40 Ar/ 39 Ar on biotite; Salisbury et al., 2011), the Filo Delgado ignimbrite (younger than 1 Ma based on stratigraphy; Lindsay et al., 2001b), and the southern locally exposed Tuyajto ignimbrite (0.53 ± 0.17 Ma; 40 Ar/ 39 Ar on biotite; Barquero-Molina, 2003) are the youngest ignimbrites so far identified.A total erupted magma volume of 500 km 3 is a conservative estimate for this stage 4. The youngest eruptions in the APVC are effusive eruptions younger than 200 ka that occur around the western margins of the volcanic province, interspersed with the arc composite cones, some of which locally share temporal and chemical affinity with this waning stage of the APVC (Godoy et al., 2014).The transition to effusive eruptions might suggest that the waning of the flare-up signaled the death of the volcano-plutonic system as a whole.However, a cluster of five lava domes (studied by de Silva et al., 1994) exhibit concordant eruption and zircon age and chemistry characteristics that require a shared thermal history within a magmatic system of superuption scale (Tierney et al., 2015).Thus even though the erupted volumes only total <50 km 3 , a vigorous plutonic system of supervolcanic proportions appears to have underpinned these for several hundred thousand years prior to eruption (Tierney et al., 2015).
Overall, a total of >15,000 km 3 of magma erupted over ~10 m.y. has been estimated ( de Silva et al., 2006;de Silva and Gosnold, 2007;Salisbury et al., 2011).This estimate includes room for the unknown volumes of intracaldera ignimbrite from several large ignimbrites including the Pujsa, Puripicar, and Sifon, the sources of which have not been studied in detail.Moreover, distal ash deposits from APVC eruptions are only recently being recognized (Breitkreuz et al., 2014), attesting to significant unaccounted volumes of co-ignimbrite and maybe co-Plinian ash.These considerations notwithstanding, the 15,000 km 3 likely reflects only a small portion of the entire volume of the magma system that fed the APVC, an assertion that is consistent with the significant plutonic volumes estimated based on the geophysical studies mentioned above (e.g., Ward et al., 2014).The APVC eruptions discussed here tapped the uppermost reaches of a crustal-scale magmatic zone where the pre-eruptive magmas staged, differentiated, and equilibrated.Preliminary work by Schmitt et al. (2002) and Folkes et al. (2011) suggests that each major eruption taps magma that resided and crystallized at pre-eruptive levels for at least a few hundred thousand years before eruption.Here we explore this pre-eruptive magmatic history further to reveal the magmatic development of the APVC through the lens of zircon geochronology.

Zircon as a Geochronometer
Zircon has long been recognized as an excellent geochronometer because of its high partition coefficients for uranium and thorium radionuclides used in radiometric dating and its low affinity for Pb (e.g., Mahood and Hildreth, 1983;Blundy and Wood, 2003).Extremely slow diffusion rates of tetravalent cations enable zircon to retain geochemical and isotopic information through multiple melting events (Cherniak et al., 1997).In addition, zircon is a common accessory phase in silicic magmas, making it an abundant and readily obtained phase for U-Pb geochronology in crustal rocks (Harrison and Watson, 1983).
Zircon crystallization begins when a silicic melt cools below its zircon saturation temperature, which has been constrained experimentally for a wide range of metaluminous to peraluminous melt compositions and zirconium concentrations (e.g., Watson and Harrison, 1983;Boehnke et al., 2013).Zircon with dimensions typically analyzed (~100 µm) can crystallize over time scales of 1000-10,000 yr once saturation conditions are met, and it can be resorbed in a zircon-undersaturated melt over equivalent durations (Watson, 1996), with larger zircon crystals surviving preferentially over smaller zircon crystals.When zircon crystallization conditions are reached anew in a magma undergoing cycles of recharge and cooling, surviving zircon crystals are expected to provide nucleation sites for additional zircon growth.In this way, zircon can record multiple magmatic events, often with hiatuses corresponding to episodes of resorption or crystal residence at subsolidus conditions (e.g., Storm et al., 2011).Reid et al. (1997) pioneered the use of high spatial resolution, in situ U-Th zircon geochronology to understand the time scales of magma storage based on the principles that (1) zircons are frequently found as inclusions within major phenocryst phases, suggesting that they formed early during the crystallization process and (2) zircon crystallizes once a magma cools below its zircon saturation temperature, which is higher than eruption temperatures.Based on these observations, which are backed by many recent high-precision U-series and U-Pb zircon crystallization ages with accompanying high-precision 40 Ar/ 39 Ar eruption ages (see following examples), zircon dating can be an effective way to constrain the pre-eruption history of a magma in the shallow crust (Simon et al., 2008;Schmitt, 2011;Cooper, 2015).

METHODS Sample Collection
Pumice samples from 16 ignimbrites representative of the temporal and spatial span of the APVC were chosen for this study (Table 1).The focus of this study is the eruptions that define the peak ignimbrite flare-up.In this strategy we sampled 9 of the 10 main eruptions of stages 2 and 3.The only large ignimbrite not studied was the ca.5.6 Ma Pujsa ignimbrite (Gardeweg and Ramirez, 1987;de Silva, 1989b) erupted from La Pacana (extractable pumice blocks were not available from this unit).In addition to these nine main units, four eruptions from stage 1, the waxing stage, and three eruptions from stage 4, the waning stage, were studied.Pumice was selected as opposed to bulk ignimbrite as representative of the juvenile magma to avoid any zircons foreign to the magma being entrained during eruption or emplacement (Schmitt et al., 2002).New samples collected for this study in 2007 and 2010 from the Guacha, Puripica Chico, Pastos Grandes, Chuhuilla, and Laguna Colorada ignimbrites in southwest Bolivia were supplemented with samples from around the APVC collected by our group over 25 years of field work and mapping in the APVC.Selected samples were fistsized or larger and had minimal visible alteration.Another part of our strategy was to use, as far as possible, the same samples (the same mineral separates) processed for Ar-Ar analysis (Salisbury et al., 2011), but several samples postdate that study.Another key strategy was to focus on the dominant pumice type.We have found that each ignimbrite contains one dominant pumice type with limited bulk and mineral chemical variation in each ignimbrite ( de Silva and Francis, 1989;Ort et al., 1996;Lindsay et al., 2001a;Schmitt et al., 2001).We chose large single or multiple smaller pumice blocks from the dominant pumice type in each ignimbrite.In a few cases two or more samples from widely separated locations from the same eruption were processed and analyzed separately to check for intraunit and intersample variability.

Mineral Separation
Zircons were separated using standard separation techniques.For dating, a minimum of 20 but more often 50-60 representative zircon crystals were hand-picked for each ignimbrite.Grains were cast in 2.54-cm-diameter epoxy disks and ground to the approximate mid-sections of grains, and polished to a 1 µm finish.

U-Pb Secondary Ionization Mass Spectrometry Analysis
High-spatial-resolution secondary ionization mass spectrometry (SIMS) has proven to be a highly effective tool for U-Pb zircon dating, in particular for complexly zoned grains.Catholodoluminescence (CL) images (Fig. 2; Supplementary Fig. 1 1 ) were taken of each zircon mount using a Leo 1430 VP secondary electron microscope at the University of California, Los Angeles (UCLA).U-Pb zircon ages were obtained using the UCLA CAMECA IMS 1270 ion probe using analytical conditions similar to those applied by Schmitt et al. (2003).A 10-20 nA 16 O -beam was focused on a 25-30 mm spot with a total depth resolution of ~0.5 mm for individual analysis spots.Secondary ions were extracted at 10 kV using an energy band pass of 50 eV.The mass spectrometer was tuned to a mass resolution of ~5000 (measured at 10% peak height) to resolve molecular interference in the mass range analyzed.The relative sensitivity between Pb and U was calibrated to allow U concentrations for unknown analyses to be calculated based on a value of 81.2 ppm for zircon standard 91500 (Gehrels et al., 2008).U-Pb ages were calibrated to zircon age standard AS3 (1099.1 Ma; Paces and Miller, 1993), which was measured throughout the analytical session.All 206 Pb/ 238 U ages are reported with a correction for 230 Th disequilibrium.The correction was based on the ratio of zircon-melt partitioning values for U and Th (Schärer, 1984) using an average whole-rock Th/U = 2 for aphyric APVC rhyolites, which better represent melt compositions compared to crystal-rich dacites (Lindsay et al., 2001).The resulting uncertainty from reasonable variations in melt Th/U is negligible relative to other sources of analytical uncertainty which for young zircon are dominated by the effects of common Pb.   1 for our complete database of images.
Downloaded from https://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/12/4/1054/4178070/1054.pdf by guest Both interior and rims on polished grains were analyzed for a minimum of 20 total analyses per sample, with an ideal goal of 50 analyses per sample.Rims were preferentially measured in order to examine more closely the last stages of zircon crystallization.Interior measurements composed ~15%-20% of analyses per sample.
All data are presented in Supplementary Table 1 2 and summary data are in Table 2.For each unit, U-Pb ages were calculated as weighted mean ages.All errors in the text and Table 2 are quoted as 2s standard error, while the full data set in Supplementary Table 1 is given as 1s standard errors.Data are presented as rank order plots of individual analyses and probability density function curves generated using Isoplot (Ludwig, 2012).Weighted averages were obtained by pooling the youngest analytically indistinguishable zircon ages.Only ages with radiogenic 206 Pb > 80% were used in age calculations.

Zircon Trace Element Analysis by SIMS
A subset of zircons from selected samples analyzed for U-Pb ages was also analyzed for trace elements including Y, rare earth elements (La, Ce, Pr, Nd, Sm, Eu, Gd, Tb, Dy, Ho, Er, Tm, Yb, Lu), P, Ti, Fe, HfO 2 , Th, and U. Three units were chosen; two crystal-rich dacites (3.96 Ma Atana ignimbrite and the 1 Ma Purico ignimbrite, representative for the peak and waning stages of the flare-up, respectively) and the aphyric Toconao rhyolite (4.0 Ma).SIMS spots for trace elements overlapped with or were as close as possible within the same CL zone as the U-Pb analytical pit.Analyses were conducted using the UCLA CAMECA IMS 1270 following analytical procedures in Schmitt and Vazquez (2006).Relative sensitivity factors in relation to Si (assumed to be present in stoichiometric abundances in zircon) were calibrated using NIST SRM 610 glass (Pearce et al., 1997) and SL13 (6.32 ppm Ti; Harrison and Schmitt, 2007); zircon standard 91500 was analyzed as a secondary reference.Significant matrix effects between glass and zircon were only detected for Ti (~-11%), and consequently SL13 was used as the primary standard for Ti, and NIST SRM 610 for all other elements.Additional elements (Mg, Mn, and Fe) were monitored to detect potential beam overlap onto non-zircon phases (e.g., adherent glass) or inclusions, but not quantified due to a lack of certified standards for these elements in zircon.Analyses were discarded where these monitoring elements were significantly elevated relative to the zircon references.

Zircon Trace Elements
Trace elements in zircons from three representative units in the APVC, the Atana, Toconao and Purico ignimbrites, are typical of trace elements from zircon in the continental crust (Fig. 3; Supplementary Table 2    The weighted mean age is the magmatic age of the ignimbrite.The ∆ age is the estimate of the magmatic duration calculated using the difference between the magmatic age and the eruption age given by the 40 Ar/ 39 Ar ages reported in Table 1.See text for further details.The 1σ of the ∆ age was computed using the square root of the sum of the squares of the 1σ uncertainties of the two ages differenced.MSWD-mean square of weighted deviates.See Supplementary used as an index of differentiation; because Zr is more compatible than Hf in zircon and Zr/Hf ratios in zircon are much higher than in the melt, progressive zircon fractionation results in decrease of Zr/Hf in the melt and therefore in later crystallizing zircon crystals.The Th/U, Ti, and Eu/Eu* all decrease with decreasing Zr/Hf.The covariability of all these parameters is compatible with crystal fractionation operating during zircon growth.Because Ti partitioning into zircon is temperature dependent and correlates with indirectly temperature-dependent compositional parameters (e.g., Zr/Hf), the decrease in Ti is a useful proxy for decreasing temperature.Although the exact temperature remains difficult to constrain from Ti-in-zircon data alone because of uncertainties in Ti activities and the experimental calibration, the limited range of Ti-in-zircon implies that zircon crystallized, and the magma differentiated extensively, over a very narrow range in temperature; ~150 °C for aTiO 2 of 0.8 ± 0.1 (Fig. 3).In this framework the decrease in U/Th and Eu/Eu* with Zr/Hf suggests that plagioclase and possibly allanite fractionation accompanied fractionation of zircon in the APVC systems.

U-Pb Zircon Ages
Crystal size varies greatly between ~50 and ~300 mm, and cathodoluminescence images reveal zoned zircon crystals in all ignimbrites (Fig. 2; Supplementary Fig. 1).Two morphological populations of zircon crystals are present in each ignimbrite: equant, subhedral crystals and elongated, prismatic crystals.There is no significant difference in the rim ages between zircon populations (Supplementary Fig. 2).CL distinct interiors are often present in zircon crystals; however, these usually have parallel or subparallel boundaries with apparent rim overgrowths and little evidence for resorption.
Probability density function (PDF) curves from all of the APVC ignimbrites show dominant peaks in zircon age density a few hundreds of thousands of years prior to the eruption age (Fig. 4).These peaks approximate a range of strongly overlapping zircon ages with no significant age gaps and indicate the highest density of zircon ages.Most ignimbrites also show small inflections off the dominant peak or smaller peaks slightly older than but overlapping with ages constituting the dominant peak.These represent subpopulations within the overlapping zircon age spectrum.In some cases, interiors yield ages outside of the APVC age span, but only 24 zircon ages (~3% of the total population) plot outside the 40 Ar/ 39 Ar age span of the APVC (Fig. 5).
In most cases, significant differences between rim and interior analyses are absent, and they appear to be randomly distributed along the continuous age spectra of individual samples (Supplementary Fig. 1).Interior and rim analyses on the same crystal are also often analytically indistinguishable, but interior ages predating the rim ages by several hundred thousand years have been detected in some crystals (Fig. 2; Supplementary Table 1).
The youngest zircon ages in each age population overlap, within uncertainty, the eruption age for each ignimbrite inferred from 40 Ar/ 39 Ar ages of sanidine and in most cases the 40 Ar/ 39 Ar ages of biotite.The amount of overlap between zircon and 40 Ar/ 39 Ar eruption ages varies between ignimbrites; in some cases, the youngest zircon ages overlap the eruption age (i.e., Guacha, Tara), and in other ignimbrites overlap is within error of only the youngest zircon ages (i.e., Pastos Grandes).In all cases it is clear that zircon crystallizes over a significant age range from a few hundred thousand years before eruption until eruption, and the youngest zircons are often concordant with the 40 Ar/ 39 Ar age (see also Bachmann et al., 2010).
There is one anomaly, the oldest ignimbrite in the APVC, the San Antonio ignimbrite.A 40 Ar/ 39 Ar biotite age of 10.33 ± 0.66 Ma was obtained (Salisbury et al., 2011; weighted mean age of 9 of 12 single crystal fusions) for this ignimbrite, which is stratigraphically consistent.However, the weighted mean   (Hora et al., 2010;Salisbury et al., 2011); these works reported that 40 Ar/ 39 Ar ages from sanidine are often younger than those from coeval biotite crystals, and the discordance was interpreted as the result of extraneous 40 Ar being retained in biotite, requiring that apparent ages from biotite be treated with caution.Similar concerns with biotite ages from other ignimbrites were reported by Smith et al. (2008), Lipman andMcIntosh (2008), andBachmann et al. (2010), among others, and 39 Ar recoil, inheritance, or excess 40 Ar are posited as the likely causes.In particular, like our case of the San Antonio ignimbrite, Bachmann et al. (2010) found that biotite 40 Ar/ 39 Ar ages of the units in the Kos-Nisyros volcanic complex are older than the youngest zircon U-Pb ages; they argued that in this context the youngest zircon U-Pb ages are a better approximation of the maximum eruption age than the 40

Xenocrysts
Zircon xenocrysts in the APVC ignimbrites have U-Pb ages older than 12 Ma (maximum age defined by the autocrysts and antecrysts; Fig. 5).Most samples yielded at least a single xenocryst age (24 in total); however, this number may be biased due to spot selection typically targeting interiors of polished crystals clearly visible in CL, whereas we avoided analyzing highly irregular interiors.Only the Panizos and San Antonio ignimbrites from the eastern edge of the APVC yielded multiple zircon xenocrysts, the Panizos xenocryst population comprising almost half of the analyses from that sample.The 206 Pb/ 238 U ages of zircon xenocrysts include 2 early Miocene zircons, 1 Oligocene, and 20 xenocrysts spanning the rest of the Phanerozoic, and a single Precambrian xenocryst (Fig. 5).Many of these ages, however, are discordant as a result of beam overlap onto different age domains.

DISCUSSION
In the following, we interpret an individual zircon U-Pb age to indicate that that zircon-saturated melt was present at that time.The fact that individual zircon rim analyses define a continuous population to within error of the eruption age supports this interpretation over one where the U-Pb zircon age represents the age of magma and zircon cooling and later rejuvenation.If all the zircons in a magma share a crystallization history over time scales equivalent to or faster than the analytical resolution, the weighted mean age of the population should reveal the prevalent zircon crystallization age.However, zircon spectra are notoriously complex and record multiple origins of zircon crystals that need to be recognized and those populations deconvolved (Miller and Wooden, 2004;Miller et al., 2007;Folkes et al., 2011;Storm et al., 2011).Informed by previous work (e.g., Bacon and Lowenstern, 2005;Charlier et al., 2005;Miller et al., 2007;Walker et al., 2010), we use the following framework.Zircon autocrysts crystallize in the magma that accumulates prior to an eruption and are used to constrain the final pre-eruptive crystallization history of a magma.Zircon antecrysts (Hildreth, cited in Charlier et al., 2005) are incorporated from material left from older episodes of magmatic activity at the same center indicating that zircon-saturated magma was present earlier in the history of the magmatic system.An erupted record of this history may or may not be present.Xenocrysts are incorporated from surrounding country rock and have no genetic relationship to any magma pulses associated with autocryst or antecryst production (e.g., Brown and Smith, 2004).Within this framework we interrogate the zircon ages recorded in the APVC ignimbrites.

U-Pb Ages of Autocrysts Reveal the Magmatic Age of APVC Ignimbrites and Model Duration of Zircon Crystallization
We note that only 6 of the 16 zircon populations for individual units yield MSWD values that plot within the 95% confidence interval for the given n value, meaning that most of the ages show dispersion beyond assigned ana-lytical uncertainties (cf.Mahon, 1996).For each ignimbrite, this population of zircon rim, and some interior, ages that predate the eruption forms the dominant peak in each of the zircon spectra (Fig. 4).Because there are zircon autocryst ages that are distinctly older than the eruption age, and the autocryst data show large age dispersion, this implies that zircon was growing well before eruption.The weighted mean age of this population is the most common crystallization age for autocrysts prior to eruption, and using this we estimate a magmatic age (Table 2).
These model magmatic ages are all consistent with stratigraphy and the Ar-Ar ages of the units, giving us confidence that these mean ages are a geologically reasonable estimate of when zircon-saturated melt was present in the system.Accordingly, we calculate a minimum pre-eruptive crystallization history, the D age (Table 2), as the difference between the 40 Ar/ 39 Ar eruption age for each ignimbrite and its zircon magmatic age.From these it is clear that minimum pre-eruption crystallization histories of several hundreds of thousands of years are recorded in these zircons (longer durations being possibly indicated by antecrystic zircon).With one exception, the Toconce ignimbrite, model durations range from ~150 to 500 k.y., with a weighted mean of ~400 k.y.(Table 2).Because these model magmatic ages are only slightly older than eruption compared to the dispersion of ages, this implies significant crystallization just prior to and continuing until eruption.
While we are mindful of the potential problems of biotite versus sanidine 40 Ar/ 39 Ar eruption age for these ignimbrites, we note that if biotites give older eruption ages than sanidine (Hora et al., 2010), this would reduce the difference with zircon magmatic ages.We find no systematic differences in the durations calculated using eruption ages derived from sanidine or biotite; units with preferred eruption ages from biotite or sanidine both give magmatic durations <400 k.y. and >400 k.y.This, and the fact that the dominant zircon peak extends further back in time, suggesting a much longer magmatic prehistory, gives us confidence that average minimum pre-eruption magmatic histories of ~400 k.y. are recorded in the zircons of the APVC ignimbrites.Long pre-eruptive magmatic histories, complex zircon population distributions, and excess scatter recorded as high MSWDs are thought to be the natural consequence of the long thermal lifetimes and complex magmatic architecture of these large mature silicic magmatic systems (e.g., Simon et al., 2008;de Silva and Gregg, 2014;Lipman and Bachmann, 2015).

Magmatic Groups Define Pulses of Magmatism within the APVC Flare-Up
Based on qualitative peak matching, ignimbrites with overlapping age spectra can be grouped into four magmatic age groups (Fig. 3 color groups) with only insignificant overlap in age spectra between the groups.The four magmatic age groups are defined by the age ranges 11-8.3Ma, 8-5.5 Ma, 5.2-2.9Ma, and 1.7-1.0Ma leading into and merging with the four eruptive stages Downloaded from https://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/12/4/1054/4178070/1054.pdf by guest described herein.In each group, each successive eruption contains a zircon age spectrum that overlaps the previous eruption in the group.These qualitative groupings can be assessed statistically using the Kolmogorov-Smirnov (KS) test to quantitatively determine whether zircon populations comprising the overlapping age spectra sample the same distribution (Press et al., 1992).The KS test measures the maximum distance between the cumulative probability function of two populations to determine the probability (P) that the populations come from the same distribution, with the null hypothesis being that they come from the same distribution.The null hypothesis is rejected if P < 0.05; P values > 0.05 do not indicate higher confidence for similarity between populations as P approaches 1.This is a conservative approach because KS tests tend to overemphasize heterogeneity when in fact two populations are identical (Saylor and Sundell, 2016).With this caveat, we use the KS test as a first-order statistical tool to assess age homogeneity versus heterogeneity within APVC magmatic groups.
Magmatic group 1.This group (11-8.3Ma) consists of overlapping U-Pb age distributions of the San Antonio, Artola, Vilama, and Sifon ignimbrites (Fig. 6).This magmatic age group correlates with eruptive stage 1, the waxing stage of the APVC flare-up.We have reassigned the eruption age of the San Antonio ignimbrites to 9.09 ± 0.60 Ma, the age of youngest zircon in the dominant peak.This notwithstanding, group 1 contains the earliest record of continuous zircon crystallization in the APVC, beginning ca.11 Ma with the oldest zircons in the San Antonio ignimbrite and continuing until the eruption of the Sifon ignimbrite ca.8.2 Ma.
The San Antonio U-Pb in zircon age spectrum consists of a dominant peak centered at 9.5 Ma and a smaller peak at 10.8 Ma.Artola yields a single dominant peak.Because the same relative uncertainties in U-Pb dating translate into larger absolute age errors for older ignimbrites compared to younger units, this may obscure some intrapopulation heterogeneities in San Antonio and Artola zircons.Overall, the San Antonio and Artola ignimbrites show nearly simultaneous zircon crystallization peaks for their dominant peaks (Fig. 6).Comparing the distributions of the zircon ages composing the prominent peaks using a KS test yields an acceptable probability of relationship (P = 0.715) between Artola and San Antonio zircon age distributions (Table 3; Fig. 6), suggesting that the magmatic development of these systems was contemporaneous, despite the fact that their sources are >100 km apart (Fig. 3).
The PDF peaks of the U-Pb age spectra for both the Vilama and Sifon ignimbrites overlap almost entirely.Zircon ages from the Vilama ignimbrite  define a single peak while those from the Sifon show some small inflections and irregularities (Fig. 6).Autocryst populations composing the prominent peaks from Sifon and Vilama are statistically indistinguishable (P = 0.213), suggesting contemporaneity of zircon crystallization despite the distance between the two sources (separated by >100 km).Their 40 Ar/ 39 Ar ages also overlap within error.Both these younger ignimbrites contain antecrysts that correlate in age with the Artola and San Antonio peaks (Fig. 6), suggesting that the earlier phase of magmatism represented by the Artola and San Antonio ignimbrites is also recorded in the magmatic history of the Vilama and Sifon ignimbrites.
Magmatic group 2. This group (8-5.5 Ma) is defined by overlapping zircon age spectra from the Panizos, Toconce, Guacha, and Chuhuilla ignimbrites (Fig. 7).There is no significant overlap in zircon ages between the youngest zircon crystals from Sifon and Vilama ignimbrites of magmatic stage 1 and the oldest antecrysts in this second group, those in the Panizos ignimbrite.Thus this second period of continuous zircon crystallization is defined from the onset of Panizos crystallization ca. 8 Ma until the eruption of the Chuhuilla ignimbrite ca.5.45 Ma.This magmatic stage covers eruption ages ranging from 6.5 to 5.45 Ma.
The Panizos age spectrum defines a single peak ca.7.1 Ma.Several younger ignimbrites have zircon ages that correlate with Panizos.An antecrystic peak from Guacha is within the spread of the Panizos PDF and inflections off the prominent peaks in both Chuhuilla and Toconce correlate well with the Panizos peak (Fig. 7).The Guacha PDF curve consists of a dominant peak ca.5.8 Ma with a slight inflection ca.6.3 Ma represented by a kink in the slope of the rank order plot of zircon ages (Fig. 7).This correlates with Chuhuilla zircon ages.The spread of the Chuhuilla PDF is noticeably skewed toward older ages with a large proportion of older crystals relative to the other ignimbrites in this period, with a significant number of crystals predating the eruption by almost 2 m.y.(Fig. 8).Rank order plots show a slight stepped pattern compared to the continuous slopes seen in Guacha and Panizos.The oldest Chuhuilla zircons are well within the range of Panizos ages, while the youngest overlap with the youngest Guacha zircons, suggesting that Chuhuilla zircons record crystallization over nearly the entire span of the second period.
A KS test comparing the Chuhuilla zircon distribution to the combined zircon distributions from Guacha, Toconce, and Panizos suggests age equivalence at P = 0.242.This suggests that zircon crystallization recorded in the Chuhuilla coincides with that in at least three different ignimbrites erupted from separate centers over 2 m.y.The distance between each of these centers spans nearly the entire spatial distribution of the APVC (Fig. 1).Panizos is the easternmost eruption in the APVC, Toconce erupted from a buried source on the western edge, the Guacha caldera is located to the south, and Chuhuilla erupted from the northernmost caldera, Pastos Grandes.This magmatic age group correlates with eruptive stage 2, part of the peak of the APVC flare-up.
Magmatic group 3.The third set of overlapping age spectra defines magmatic group 3 and includes the Puripicar, Toconao, Atana, Tara, and Pastos Grandes ignimbrites, delineating a magmatic age range of 5.2-2.9Ma (Fig. 8).A few isolated zircon ages from Puripicar and Toconao overlap in age with the  zircon histories from Panizos, Chuhuilla, and Guacha spectra in the preceding period, but most of the age spectra are clearly distinct.Zircon crystallization over this period began ca.5.2 Ma and continued until the Pastos Grandes eruption at 2.89 Ma.Eruption ages range from 4.09 to 2.89 Ma.The Puripicar and Toconao zircon spectra overlap considerably and have eruption ages that are indistinguishable within error (Fig. 8).While Toconao has a uniform peak, the Puripicar PDF shows some irregularities in the peak that do not correlate with any other zircon population.A KS test also shows an acceptable probability of fit (P = 0.779) between the Puripicar and Toconao zircon crystallization histories, suggesting that they developed contemporaneously (Fig. 8).Their eruption source vents are ~50 km apart.The Atana ignimbrite zircon age spectrum is the narrowest in the APVC, despite the large number of crystals (n = 61) analyzed (Fig. 8).The shoulder and minor peak at ages older than the dominant peak for the Atana ignimbrite consist of five zircon ages that overlap with the Toconao and Puripicar zircon ages.
The two youngest eruptions in stage 3, the 3.49 Ma Tara and 2.89 Ma Pastos Grandes ignimbrites, display continuous zircon crystallization spectra that overlap within error ca.3.8 Ma (Fig. 8).We interpret this as indicating that both ignimbrites had synchronous zircon crystallization histories until the Tara eruption at 3.49 Ma; zircon crystallization continued at Pastos Grandes until eruption at 2.89 Ma.Zircon ages responsible for the inflection ca.3.6 Ma in the Pastos Grandes peak overlap with the dominant peak of the Tara eruption.The Tara peak is largely uniform, whereas the Pastos Grandes peak has a shoulder ~500 k.y.older than the dominant younger peak.There is very little overlap between the dominant peak from the Pastos Grandes and Tara ignimbrites and the older Puripicar, Toconao, and Atana age spectra (Fig. 9).However, the older tail in the Pastos Grandes PDF and obvious antecrysts, one of which is also present in the Tara ignimbrite, indicates a synchronous magmatic history with the three older systems located almost 200 km south of the Pastos Grandes caldera.This magmatic age group correlates with eruptive stage 3, the peak of the APVC flare-up.
Magmatic group 4.This group is defined by a period of overlapping age spectra from the Laguna Colorada, Puripica Chico, and Purico ignimbrites (Fig. 9), and zircon ages are continuous between ca.2.7 Ma until the eruption of the Purico ignimbrite at 0.98 Ma.The oldest zircon age from the Laguna Colorada ignimbrite overlaps within error the Pastos Grandes eruption age.Laguna Colo rada and Puripica Chico are unusual because their age distributions show a minor, younger population that forms a shoulder offset from the dominant peak by ~300 k.y.Purico and Puripica Chico both contain many minor peaks consisting of multiple zircon ages that overlap the Laguna Colorada age spectrum.The Purico ignimbrite is the only APVC ignimbrite lacking a single dominant pre-eruption peak.Instead, Purico contains several distinct zircon age populations, the oldest of which overlaps significantly with the Laguna Colorada age distribution.Overall, Purico contains a nearly continuous zircon record extending over 1.5 m.y.; a similar age range is present in Dome D lavas (Supplementary Table 1).This magmatic age group correlates with eruptive stage 4, the waning of the APVC flare-up.One small eruption that belongs to eruptive stage 4 but has a magmatic history extending back to magmatic stage 3 is Cerro Bola, the youngest dome from the La Pacana caldera.This small high-Si rhyolite dome has an eruption age of 2.7 ± 0.2 Ma ( 40 Ar-39 Ar biotite; Lindsay et al., 2001b) that would make it concordant with the eruption of the Pastos Grandes ignimbrite eruption ~100 km away.The autocryst age population of Cerro Bola yields a weighted mean age of 2.57 ± 0.12 Ma within error of the 40 Ar-39 Ar biotite age, although the youngest zircons of Bola extend to ca. 2.4 Ma.Thus part of the autocryst history correlates with stage 4, overlapping the early zircon history of Laguna Colorada.However, a second distinct peak at 3.02 ± 0.04 Ma overlaps with the peak of the Pastos Grandes ignimbrite.These correlations suggest synchroneity between centers over a 150 km north-south distance.

Spatiotemporal Development of Magmatism in the APVC and the Development of Magmatic Foci
The spatiotemporal distribution of eruptions in the APVC is interpreted to have evolved from diffuse to focused over the four eruptive stages ( de Silva and Gosnold, 2007;Salisbury et al., 2011).The magmatic record of zircon crystallization histories enhances this view with contemporaneous locations of magmatism (as recorded by zircon age data) without eruptions.By combining these records with petrological and geochemical constraints we can provide some constraints on the crustal depths where zircon crystallized.
The crustal compositions of the zircons (Fig. 3) and available zircon d 18 O data (Folkes et al., 2013) indicate that zircons grew in magmas that were isotopically enriched and well homogenized.Geobarometry (Al-in-amphibole phase equilibria, melt inclusions; Schmitt, 2001;Grocke et al., 2012), geothermometry (Fe-Ti oxides phase equilibria, Al-in-amphibole phase equilibria, and zircon saturation), experimental phase equilibria (Muir et al., 2014a), and geochemical modeling of trace element and isotopic compositions ( de Silva, 1989;Lindsay et al., 2001;Schmitt et al., 2001;Kay et al., 2010;Muir et al., 2014b;Burns et al., 2015;Freymuth et al., 2015) all indicate that APVC eruptions evacuated upper crustal magma reservoirs that stalled, crystallized, and acquired their trace element and isotopic compositions in the shallow crust.In contrast, magmas that are interpreted to be derived from deeper in the APMB (10-30 km) are isotopically less enriched, hotter, andesitic, and zircon undersaturated, obviating the APMB as a source of zircons.These considerations and U-Pb zircon rim ages that suggest zircon crystallization continued right until eruption convince us that the U-Pb zircon ages extend our understanding of the spatiotemporal evolution of the volcano-plutonic system of the APVC to shallow crustal depths of the pre-eruptive magma chamber ~5-10 km beneath the surface, the uppermost reaches of the APMB (Fig. 10).
Each of the time slices in Figure 10 portrays distinct ~2-3 m.y.pulses of magmatism during which spatially distinct but magmatically contemporaneous magma reservoirs developed.The spatial footprint of the shallow magma system varied with time, and some of the constituent magma bodies erupted.

Salar de Ascotan
Other magma bodies had delayed eruptions, but their continued magmatic history is revealed in the zircon age spectra.The time slices clearly support the impression from the eruptive record that magmatism focused and intensified with time from the earliest pulse to the third with diminishing breadth and intensity by the fourth pulse, corresponding to the waxing, climax, and waning of the flare-up.
Within the magmatic record is evidence that three magmatic foci developed during the peak of the flare-up and account for the bulk of the volume of magmatism.The surface manifestations of these are the major resurgent calderas of La Pacana, Guacha, and Pastos Grandes.All three appear to have been initiated during magmatic pulse 2 as the flare-up peaked, and were active through pulse 3.During this time each was the source of two sequential eruptions, i.e., Pujsa-Toconao-Atana, Guacha-Tara, and Chuhuilla-Pastos Grandes.The older eruptions were broadly contemporaneous ca.5.6 ± 0.2 Ma.These were followed by younger eruptions of the Atana (4.09 Ma), Tara (3.49Ma), and Pastos Grandes (2.89 Ma) ignimbrites that reveal significant departure from contemporaneity compared to the older eruptions.However, their zircon spectra evince a more robust correlation of their magmatic histories (Figs. 7 and 8).Thus, contemporaneous magmatism at these centers separated by large distances (50-200 km) is not always correlated with contemporaneous eruptions.This appears to be true throughout the history of the APVC.During the waxing pulse, the San Antonio and Artola ignimbrites and the Vilama and Sifon ignimbrites were erupted from widely separated sources, and show strong correlation in magmatic history; however, the former pair represents relatively small localized eruptions, while the latter are two supereruptions that occurred almost a million years later.This suggests that while spatially, and thus physi cally, separated magma systems may develop in parallel, interruption of the magmatic evolution by eruption is unique to each center.Some may erupt while others may continue to develop and erupt at a later time.In these large mature magmatic systems eruption is thought to reflect the breach of a thermo mechanical threshold (e.g., Gregg et al., 2012).Thus the different histories of otherwise similar contemporaneous systems may reflect differences in local magmatic flux and consequent thermomechanical evolution of the upper crust in which the magma reservoirs developed.

Integrating Magmatic and Eruptive Histories: Episodic Development of the APVC Magmatic System
Correlations between the four pulses of magmatism and eruptive stages are revealed by pulses in zircon crystallization with distinct breaks between magmatic pulses on the regional scale (Fig. 11).All four eruptive stages have overlapping zircon ages, and dominant peaks in the younger pulses are preceded by distinct peaks in zircon crystallization that are interpreted to reflect separate events of magma accumulation in the upper crust.Available U-Pb ages indicate that these precursor events initiated 0.5-1 m.y.prior to erup-  tions, with possibly continuous presence of zircon saturated magma until each eruption.Thus, each of the four magmatic pulses consists of spatially distinct, temporally sequenced (with overlap) subpulses or magmatic events.Therefore, an episodicity at multiple scales is defined and led de Silva et al. (2015) to characterize the pattern as fractal (sensu latu).Looking at flare-ups in continental arcs as a whole, they suggested a hierarchy of pulses with each scale reflecting the time scale of processes occurring at different levels in the arc crust.This tempo of continental arc magmatism is interpreted to reflect modulation of the mantle power input as it is progressively filtered through the continental crust.In this framework, the APVC ignimbrite flare-up as a whole is a secondary pulse with magmatic pulses 1 through 4 reflecting tertiary pulses, and the individual ignimbrite zircon spectra defining quaternary pulses.In summary, the secondary pulse of the 10 Ma ignimbrite flare-up of the APVC consists of four tertiary pulses of intrusion and/or eruption with a periodicity of ~2 m.y.The tertiary pulses consist of three or more distinct quaternary pulses, each of <1 m.y.duration.The tertiary pulses in the APVC indicate processes that not only divide the secondary pulse into shorter pulses (~2 m.y.magmatic events in the APVC that culminated in eruptions), but produce the characteristic space-time-volume pattern of waxing, climax, and waning of the flare-up.The tempo of these tertiary pulses may ultimately reflect the influence of a combination of mantle, crustal, and upper plate tectonic modulation (or other nonmagmatic upper crustal process).The waxing-climax-waning pattern seen in the tertiary pulses in the APVC appears to be a characteristic pattern in ignimbrite flare-ups (de Silva et al., 2006b;Bachmann et al., 2007c;Lipman, 2007) and may reflect the evolution of crustal melt production with time, thereby mimicking the mantle pulse topology (e.g., Elston, 1984;Best and Christiansen, 1991).Alternatively (or in addition), the pattern may reflect a progressive evolution of crustal rheology resulting from the thermal signal from the mantle progressing through the crust by intrusion and advection (e.g., Best and Christiansen, 1991;Gans et al., 1989;de Silva and Gosnold, 2007;de Silva and Gregg, 2014).This progression may result in the formation of a middle to upper crustal MASH zone and magma staging area and an elevation of the brittle-ductile transition to shallow levels in the crust.Over time, these processes combine to allow successively larger magma bodies to be built in the uppermost crust before they erupt (e.g., de Silva and Gosnold, 2007).
The quaternary pulses have time scales of ~1 m.y. and define the magmatic evolution of the upper crustal pre-eruptive reservoirs and the thermomechanical evolution of the host rock-reservoir system.As magma accumulates in the upper crust and evolves to zircon saturation, feedbacks between temperature, host-rock mechanics, and chamber pressurization result in ductile host-rock rheologies that promote storage and growth over eruption and eventual eruption is due to crossing a mechanical threshold (Gregg et al., 2012).The broad correlation of magmatic and eruptive tempo of the APVC supports a strong coupling between the magma dynamics and thermomechanics of periodically constructed long-lived upper crustal magma reservoirs that form the shallowest part (5-10 km) of the APMB.

Zircon Insights into Magmatic Processes and Magma Dynamics
The presence of abundant zircons in the pumices of the APVC ignimbrites and their age spectra indicate that the thermochemical conditions for zircon saturation were maintained for several hundreds of thousands of years for each system.Zircon saturation temperatures based on the most recent experimental studies by Boehnke et al. (2013) range from ~730 °C to 815 °C, broadly concordant with the 750-850 °C estimated from Fe-Ti oxide phase equilibria and Al-in-amphibole ( de Silva, 1991;Lindsay et al., 2001;Schmitt et al., 2001;Grocke, 2014).Trace element data from zircons suggest magma differentiation over a temperature range of ~150-200 °C derived from Ti-in-zircon (Fig. 3).Taking a conservative 400 k.y. as the duration over which zircon saturation was maintained, the temperature estimates suggest extremely low secular cooling rates of 4-5 × 10 -4 °C/yr.Such cooling rates are an order of magnitude slower than typical conductive cooling rates of plutons (~4 × 10 -3 °C/yr) that consider the temperature dependence of thermal diffusivity (a) and heat capacity (C P ) (Whittington et al., 2009;Nabelek et al., 2012;Gelman et al., 2013) or realistic mature thermal gradients in which such systems might develop ( de Silva and Gosnold, 2007;de Silva and Gregg, 2014) suggesting that the pre-eruptive magma reservoirs were thermally buffered.Such conditions are typical of development of the pre-eruptive magma reservoirs in the elevated geothermal gradients (~50 °C/km) typical of the upper crust during the APVC flare-up ( de Silva and Gosnold, 2007).

Carryover of Zircon Antecrysts: Evidence for Upper Crustal Assimilation
A consequence of the thermal longevity of the pre-eruptive magma reservoirs is that assimilation-fractional crystallization was a dominant process during the evolution of the APVC ignimbrite magmas (e.g., see Kay et al., 2010;Folkes et al., 2013, and references therein).During this process restite of assimilated country rock in the form of antecrysts or xenocrysts should be recorded.Zircon is quite robust during magmatic recycling and reworking, and carryover or scavenging of antecrystic zircons is frequently observed in studies assessing the magmatic development of volcanic systems (Watson, 1996;Charlier et al., 2005;Bacon and Lowenstern, 2005;Walker et al., 2010;Folkes et al., 2011).In these examples, younger eruptions incorporate antecrysts from a previous magmatic cycle.Xenocrysts may also be incorporated during crustal assimilation.The relative proportions of antecrysts versus xenocrysts in the APVC ignimbrites may therefore offer some insight into the pre-eruptive magmatic evolution.
A characteristic of the zircon age spectra for the ignimbrites (Figs. 3 and  6-9) is that they become more complex with time.The oldest ignimbrites show largely unimodal zircon peaks with few inflections or smaller peaks (i.e., Artola, Vilama), although analytical uncertainties may mask variability that is visible in younger ignimbrites.Progressively younger ignimbrites show more complex zircon crystallization histories with frequent satellite peaks off the prominent peak (i.e., Guacha, Toconao), isolated older peaks consisting of single zircon ages, as well as satellite peaks (Puripicar, Tara); the youngest ignimbrites yield the most complex and variable PDF curves (i.e., Pastos Grandes, Puripica Chico, Purico).For these units, the precision of individual U-Pb zircon ages can resolve differences in magmatic ages of zircon crystals from individual eruptions, and we note that even in the oldest magmatic pulse, evidence for zircons with U-Pb ages outlying the main peak is found.In all but the two oldest eruptions, these antecrystic ages correlate with older eruptions, the antecrysts in the Artola and San Antonio representing the earliest magmatic history of the APVC for which there is apparently no volcanic equivalent (see also Schmitt et al., 2002).We suggest that the change in complexity of the zircon populations over the history of the APVC likely reflects the development of a multistage, multicyclic plutonic system in the shallow crust (~5-10 km).Each magma is built by incremental intrusion from the APMB into the shallow crust and the final stages of the pre-eruption accumulation history are partly recorded in the zircon autocrysts.Zircon antecrysts record earlier history of the magma system, with or without eruption.Younger ignimbrites contain more antecrysts because the upper crustal magmas feeding these eruptions are emplaced in a more mature magmatic system with a longer history made up of a multicyclic mush; there is simply more of a history to be scavenged by later magmas.It is interesting that the Purico ignimbrite, erupted from the Purico ignimbrite shield located just to west of the Guacha and La Pacana calderas and to the south of the Chaxas complex from which the Puripicar ignimbrite was erupted, has the most complex U-Pb zircon age spectrum that we have found.However, there is no hint of any antecrysts of these three older systems, suggesting a very focused relatively small magmatic system that was established between the giant systems beginning ca. 2 Ma.Similar nonsystematic behavior is mirrored by the somewhat older Cerro Bola dome at La Pacana.The peaks in 206 Pb/ 238 U ages of autocrysts and antecrystic zircon interiors are both significantly younger than the caldera-forming Atana ignimbrite at La Pacana.However, as noted here, the autocryst and antecryst peaks of Cerro Bola overlap with the early history of the Laguna Colorada ignimbrite shield ~100 km away and the peak of activity at the Pastos Grandes caldera ~150 km to the north, respectively, suggesting contemporaneity of magmatism, but not eruption, at these three centers.
Despite the strategy of preferentially measuring near-rim domains over interiors, zircon antecrysts were found to be a ubiquitous but minor component of the zircon populations of APVC ignimbrites.These reveal unequivocal evidence for carryover of antecrysts in all of the investigated APVC ignimbrites.In association with the evidence presented here for the upper crustal evolution of the APVC magmas, we interpret these antecrysts as attesting to upper crustal cannibalization of remnants of progenitor magmas.

Crustal Zircon Inheritance
True xenocrysts, defined here as pre-APVC aged zircon inherited from regional basement rocks, make up a scarce 3% of the analyzed populations (Fig. 5).This overall proportion of xenocrystic zircon is further reduced when considering that 13 of 24 xenocrysts are from a single unit (Panizos).Moreover, obvious CL patterns suggesting the presence of inherited crystal interiors are scarce, and if they were present, they were targeted by interior-rim pairs of analyses.We therefore conclude that scarcity of zircon inheritance (with the notable exception of Panizos) is a true characteristic of the APVC.This contrasts with the whole-rock isotopic (Sr, Nd, O) evidence for a significant crustal contribution (as much as 70%) in the APVC magmas.These observations can be reconciled if crustal zircons became resorbed in the APMB, which is the parental reservoir that supplied magmas to the shallow pre-eruptive magma reservoirs where most zircon crystallized.The preservation of xenocrysts in the Panizos ignimbrite magma may be due to the peraluminous character of the magmas.Caffe et al. (2012) demonstrated that these peraluminous magmas form through contamination of calc-alkaline dacitic magmas, similar to typical APVC ignimbrites by metapelite at APMB depths (≥18 km depth).Caffe et al. (2012) further determined that crystallization of the rhyolite magma started at ~5 kbar and 800 °C, and continued almost isobarically to 720 °C.These conditions are much cooler than the typical APMB conditions, allowing the possibility that along the eastern flanks of the APMB, where peraluminous magmas were being produced, conditions were more conducive to preserving inheritance than the rest of the APMB, where andesitic compositions with temperatures of 900-1000 °C were typical (e.g., Burns et al., 2015).The limited evidence for true inheritance in the APVC magmas is then attributed to basement assimilation occurring primarily in the zircon-undersaturated environment of the APMB at deeper levels than pre-eruption levels.This is consistent with the consensus that the APMB is the major upper crustal MASH zone in which the APVC baseline chemical and isotopic compositions signature is developed.The pre-eruptive magma reservoirs in which the autocryst and antecryst record develops are fed from the deeper parts of the APMB in which few xenocrysts survived except in the east.This also implies that for the main volume of the APVC, zircon crystallization only started after emplacement in the shallow (~5-10 km) magma reservoirs.

Zircon-Based Perspective for APVC Magmatism
As the APVC flare-up developed, the synchroneity of crystallization histories of APVC eruptive centers separated by >100 km suggests a significant level of temporal connectivity, and it might be tempting to look at the time slices in Figure 10 and imagine APVC-wide connectivity of magma at the pre-eruptive levels.This might be supported by the fact that the general chemical, petro logical, and isotopic compositions are concordant in all these magmas, including their zircon trace element compositions (Fig. 3) that suggest a close kinship.The significant spatial separation of the eruptive sites might give one pause, but this could be rationalized as local control of the eruption sites above a broad APVC wide sill at 5-10 km below the surface.However, in detail the U-Pb zircon age spectra and subtle petrochemical details suggest that each erupted magma body developed uniquely.For example, if the Vilama and Sifon ignimbrites were erupted from vents separated by >100 km above a shared magma reservoir, the younger centers that subsequently developed between the respective eruptive sites should record the presence of antecrysts of Sifon and Vilama ages.However, none of the younger ignimbrites ( Guacha, Chuhuilla, Tara Pastos Grandes, and Laguna Colorada) record any antecrysts of appropriate age.Similarly, while the Puripicar and Toconao share very similar U-Pb age characteristics, the Toconao rhyolite is volcanologically and petrochemically linked to the Atana ignimbrite dacite magma, which is chemically distinct from the Puripicar dacite magma ( de Silva and Francis, 1989;Lindsay et al., 2001aLindsay et al., , 2001b)).The Panizos is clearly quite chemically distinct from its pulse 2 ignimbrite cohort.Thus the correlations of U-Pb age spectra point to synchroneity, not physical connectivity; and each eruption is interpreted to have tapped a unique pluton of a developing composite upper crustal batholith.
Although isolated at shallow levels (between 5 and 10 km), it is very probable that separate systems were interconnected at a deeper level, likely deeper within the APMB (Fig. 10).The extent of overlap of zircon age spectra suggests that shallow crustal magma reservoirs separated by tens to hundreds of kilome ters were undergoing simultaneous thermal or material input from deeper levels, particularly once the flare-up reached peak intensity during pulse 2 and 3 when the magmatic foci of the three major calderas developed.With each successive input of material, magma accumulates in the shallow crust and each coexisting magma records a very similar zircon history.Magmas either develop and erupt nearly synchronously, as in the case of the Vilama and Sifon, Toconce and Panizos, the Pujsa, Guacha, and Chuhuilla, and the Puripicar and Toconao-Atana ignimbrites, or they share a common history until one erupts while the other continues to crystallize, as inferred from the zircon crystallization histories of the Tara and Pastos Grandes ignimbrites.A significant volume of nonerupted magma drove resurgent uplift and postcaldera volcanism, and eventually matured into uneruptible mush or solidified into plutons to be recycled in later episodes of magmatism.
As the flare-up progressed, thermal maturation of the APVC upper crust led to an elevated thermal gradient (e.g., de Silva and Gosnold, 2007) in which the upper crustal pre-eruptive chambers were thermally buffered and zircon saturated, leading to comparatively long durations of zircon crystallization.Assimilation in the upper crust results in carryover of antecrysts that form the nucleation sites for autocrystic zircon growth recorded in rim ages.Zircons crystallized in serial and in parallel over protracted time periods in different zones within larger magma reservoirs with slightly different cooling histories (e.g., de Silva and Gregg, 2014), but were well mixed either due to pre-eruptive stirring (e.g., Huber et al., 2009) or eruptive mixing (e.g., Kennedy et al., 2008), accounting for the nonanalytical dispersion of ages (indicated by elevated MSWD values > 1) in the dominant autocryst population.At present it appears that the shallowest parts of the APMB MASH zone, from 5 to 10 km, has matured into a composite batholith with some melt-rich zones that may lead to local diapiric intrusion and rapid surface uplift, as observed at Uturuncu (Pritchard and Simons, 2002;Fialko and Pearse, 2012;del Potro et al., 2013;Muir et al., 2014aMuir et al., , 2014b)).

CONCLUSIONS
U-Pb ages of zircon in magmas erupted during the 11-1 Ma APVC ignimbrite flare-up were determined by SIMS.These data reveal complex age spectra with a dominant peak of autocrysts and subsidiary antecryst peaks.Xenocrysts are rare, except for a peraluminous ignimbrite from the eastern margin of the APVC.Model magmatic ages, calculated as the weighted mean of the youngest population of zircons with overlapping ages, are consistent with eruptive stratigraphy.In combination with pressure-temperature estimates of the magmas, the U-Pb in zircon ages record multiscale episodicity in the magmatic history of the shallowest levels (5-10 km beneath the surface) of the APMB.
The ages fall into four groups defining distinct pulses of magmatism that correlate with eruptive pulses but indicate that magmatic construction often initiated ~1 m.y.before eruptions began.Magmatism was initially (11-8 Ma) distributed diffusely on the eastern and western flanks of the APVC, but spread out over much of the APVC as activity waxed before focusing in the central part during the peak.Each pulse consists of spatially distinct but temporallysequenced subpulses of magma that represent the construction of pre-eruptive magma reservoirs.
Three long-lived calderas are the eruptive outlets for the main magmatic foci during the peak of the flare-up and show broadly synchronous magmatic development but some discordance in their later eruptive histories.In the context of recent models of caldera mechanics, this suggests that eruptive tempo is controlled locally from the top down, while magmatic tempo is a more systemic and deeper bottom-up feature.Synchroneity in magmatic history at distinct upper crustal magmatic foci implicates a shared connection deeper within the APMB.
Model magmatic ages are on average ~400 k.y.older than the eruption ages, revealing minimum pre-eruptive magmatic durations of APVC ignimbrite magmas.These indicate that thermochemical conditions for zircon saturation were maintained for extended (several hundreds of thousand year) periods.Zircon saturation temperatures record upper crustal conditions and trace elements reveal protracted magma differentiation under secular cooling rates an order of magnitude slower than typical pluton cooling rates, suggesting that pre-eruptive magma reservoirs were thermally buffered, consistent with the extended minimum magmatic time scales >400 k.y.we have estimated.
Zircon spectra become more complex with time, reflecting the carryover of antecrysts in successively younger magmas and attesting to upper crustal assimilation in the APVC.Although xenocrysts are present, they are rare, suggesting that inheritance is limited.This is attributed to basement assimilation occurring under zircon-undersaturated conditions deep in the APMB in contrast to the pre-eruptive levels where antecrysts were incorporated under zircon-saturated conditions.
The multiscale episodicity revealed by the zircon U-Pb ages of the APVC flare-up can be interpreted in the context of continental arc magmatic systems in general.The APVC ignimbrite flare-up as a whole is a secondary pulse, with magmatic pulses 1 through 4 reflecting tertiary pulses, and the individual ignimbrite zircon spectra defining quaternary pulses.This hierarchy of pulses is thought to reflect how a magmatic front, driven by the primary mantle-power input, propagates through the crust, producing sequentially smaller spatial and faster temporal scales in the upper crust of the Central Andes from ~30 km to the surface.

Figure 3 .
Figure 3. Secondary ionization mass spectrometry-determined trace elements in zircon from three ignimbrites from the Altiplano-Puna Volcanic Complex.Directly temperature-dependent trace elements (Ti) and trace element ratios that track differentiation in the magma (Th/U, Eu/Eu*) are plotted against Zr/Hf, which measures melt evolution.Ti-in-zircon temperatures were calculated using the calibration by Ferry and Watson (2007) and aTiO2 = 0.8 (see text).Field (dashed outline) demarks the composition of zircons of continental crustal provenance with the outline demarcating ~80% of the data in the compilation by Carley et al. (2014).

Figure 4 .
Figure 4. Eruptive history of the Altiplano-Puna Volcanic Complex represented by a rank ordering of the 16 ignimbrites that are the subject of this study.Ignimbrites are ranked based on 39 Ar/ 40 Ar eruption ages (Table 1) represented by the bold vertical lines with 2σ errors in the dashed lines when discernible.Corresponding probability density function plots of U-Pb in zircon ages for each ignimbrite are superimposed in colors that correspond to Figure 1.Groupings of zircon age spectra with overlapping distributions into four magmatic pulses are indicated by the vertical master index color panels.Details of U-Pb in zircon data for each group and the constituent ignimbrites are presented in Figures 6 through 9 and associated text.

Figure 5 .
Figure 5. Concordia diagram for zircon ages outside of the Altiplano-Puna Volcanic Complex (APVC) time span (>10 m.y.) interpreted as inherited xenocrysts from the basement.Post-analysis imaging suggests that discordant zircons result from beam overlap between xenocrystic and APVC-aged domains.Table inset indicates number of xenocrystic zircons per unit.
Figure 6.Rank order plots (ROP) and zircon 206 Pb/ 238 U ages as probability density function (PDF) curves for magmatic group 1 ignimbrites.Ignimbrites are rank ordered from oldest to youngest from bottom to top, with PDF curves from previous eruptions for reference.Each shaded PDF curve represents the distribution for the ROP displayed.Ages stated are the weighted mean and 2σ error for the dominant peak in the distribution (Table 2).Colors of curves as in Figures 1 and 4. Red lines and text represent the 40 Ar/ 39 Ar eruption ages with 2σ errors for each ignimbrite.MSWD is mean square of weighted deviates.
Figure 7. Rank order plots (ROP) and zircon 206 Pb/ 238 U ages as probability density function (PDF) curves for magmatic group 2 ignimbrites.Ignimbrites are rank ordered from oldest to youngest from bottom to top, with PDF curves from previous eruptions for reference.Each shaded PDF curve represents the distribution for the ROP displayed.Ages stated are the weighted mean and 2σ error for the dominant peak in the distribution (Table2).Colors of curves as in Figures1 and 4. Red lines and text represent the 40 Ar/ 39 Ar eruption ages with 2σ errors for each ignimbrite.MSWD is mean square of weighted deviates.

Figure 8 .
Figure 8. Rank order plots (ROP) and zircon 206 Pb/ 238 U ages as probability density function (PDF) curves for magmatic group 3 ignimbrites.Ignimbrites are rank ordered from oldest to youngest from bottom to top, with PDF curves from previous eruptions for reference.Each shaded PDF curve represents the distribution for the ROP displayed.Ages stated are the weighted mean and 2σ error for the dominant peak in the distribution (Table2).Colors of curves as in Figures1 and 4. Red lines and text represent the 40 Ar/ 39 Ar eruption ages with 2σ errors for each ignimbrite.MSWD is mean square of weighted deviates.

Figure 9 .
Figure 9. Rank order plots (ROP) and zircon 206 Pb/ 238 U ages as probability density function (PDF) curves for magmatic group 4 ignimbrites.Ignimbrites are rank ordered from oldest to youngest from bottom to top, with PDF curves from previous eruptions for reference.Each shaded PDF curve represents the distribution for the ROP displayed.Ages stated are the weighted mean and 2σ error for the dominant peak in the distribution (Table2).Colors of curves as in Figures1 and 4. Red lines and text represent the 40 Ar/ 39 Ar eruption ages with 2σ errors for each ignimbrite.MSWD is mean square of weighted deviates.

Figure
Figure 10.Spatiotemporal development of the pre-eruptive Altiplano-Puna Volcanic Complex (APVC) magma system through time.Each frame represents a single time slice corresponding to each of the magmatic pulses.Magma presence in the subvolcanic system between 5 and 10 km is indicated by the presence of zircon U-Pb ages of appropriate age for each time slice.Colored shapes represent the approximate distribution of the discrete subvolcanic magma reservoirs as estimated by their location and approximate size.Color scheme is consistent with Figures 1, 4, and 6-9.The outline of the Bouguer anomaly and low velocity zone from Figure 1B are shown.Star is Uturuncu volcano.Locations of key salars and lakes are shown.APMB-Altiplano-Puna Magma Body.Note: Interactive PDF (see figure at left) allows the reader to toggle through the magmatic stages of the APVC.

Figure 11 .
Figure 11.Zircon U-Pb age based magmatic pulses compared to eruption ages.(A) The magmatic record is represented by histograms and associated probability density function curve for all magmatic zircon ages determined for Altiplano-Puna Volcanic Complex ignimbrites.(B) Eruption ages and volumes (from Fig. 1;Table 1).

Eruptive record of the Altiplano Puna Volcanic Complex (APVC). (A) Map of the APVC in the Central Andes defined by the approximate outlines of the mapped extent of each of the major ignimbrites. Solid lines and normal font are for those focused on in this study. Dashed lines and italics are for other eruptions in the re- gion that are not focused on in this study. Ages are the latest 40 Ar-39 Ar or K-Ar ages reported in the literature (Table 1). Colors are propogated throughout the manu- script and reflect different stages of the development of the APVC. Some erup- tions ( Divisoco, San Antonio, Toconao, and Alota ignimbrites) are not shown for clarity. Data sources are given in the text. (B) Locations of the known eruption sites for each of the ignimbrites are color keyed to their stage of formation (given in C). Sizes and shapes are based on mapping and image interpretation. Map extent of seismic low-velocity zone (Ward et al., 2014) and negative gravity Bouguer anom- aly (Prezzi et al., 2009) approximates the present extent of the Altiplano-Puna Magma Body. (C) Age and eruptive vol- umes of the ignimbrites focused on in this study. DRE is dense rock equivalent. Four stages of eruptions are identified. Width of the vertical bars is a measure of the age uncertainty. Small numbered arrows on the x axis are for eruptions that are not featured in this study. Numbers refer to the following ignimbrites. 1-Tatio; 2-Pampa Chamaca; 3-Alota; 4-Pujsa; 5-Carcote; 6-Coranzuli; 7-Lower Rio San Pedro; 8-Granada; 9-Divisoco; 10- Coyaguama
. Downloaded from https://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/12/4/1054/4178070/1054.pdf by guest

TABLE 1 .
SUMMARY OF IGNIMBRITES USED IN THIS STUDY Eruption ages and volumes were reported inSalisbury et al. (

table 2 : APVC ignimbrite trace element zircon results
3 ).The Zr/Hf can be 2 Supplemental Table 1.Complete zircon U-Pb data for this study.Please visit http:// dx .doi.org/10 .1130/GES01258 .S2 or the full-text article on www .gsapubs.org to view Supplemental Table 1. 3 Supplemental Table 2. Complete zircon trace element data for this study.Please visit http:// dx .doi.org/10 .1130/GES01258 .S3 or the full-text article on www .gsapubs.org to view the Supplemental Table 2.

Table S3 -Magmatic ages -difference between Weighted mean U-Pb zircon crystallization age in Table 2 and Weighted mean Ar-Ar eruption age inTable 1
ERROR calculated as the square of the sum of the squares of the errors of the two ages differenced.

TABLE 2 .
SUMMARY OF ZIRCON 206 Pb/ 238 U IN ZIRCON AGE DATA FOR THE ALTIPLANO-PUNA VOLCANIC COMPLEX IGNIMBRITES USED IN THIS STUDY Table 3 4 .N-number of zircon analyses used in the magmatic age calculation.
*Determined using U-Pb zircon ages as the difference between 9.09 ± 0.60 Ma (youngest surface age) and the weighted mean U-Pb zircon magmatic age of 9.53 ± 0.32 Ma.Downloaded from https://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/12/4/1054/4178070/1054.pdf by guest Ar/ 39 Ar biotite age.We believe that the discordance in the zircon U-Pb age and the 40 Ar/ 39 Ar biotite age of the San Antonio ignimbrite is real and due to the biotites being compromised.Therefore we suggest that the youngest autocryst age for the San Antonio ignimbrite of 9.09 ± 0.60 Ma (2s; BOL06033 grain 30; Supplementary Table 1) is the best estimate of the maximum eruption age for the San Antonio ignimbrite.